Public Disclosure Authorized ~~ 9950 S~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Public Disclosure Authorized | ~~~~~~~~~~~ Public Disclosure Authorized edited by Graham Pyatt Jeffery I. Round Public Disclosure Authorized A WorldBank Symposium COPY FILE I I I Social Accounting Matrices A World Bank Symposium I I I I Social Accounting Matrices A Basis for Planning Edited by Graham Pyatt and Jeffery I. Round THE WORLD BANK Washington, D.C., U.S.A. Copyright © 1985 by the International Bank for Reconstruction and Development / The World Bank 1818 H Street, N.W., Washington, D.C. 20433, U.S.A. All rights reserved Manufactured in the United States of America First printing September 1985 The World Bank does not accept responsibility for the views expressed herein, which are those of the authors and should not be attributed to the World Bank or to its affiliated organizations. The findings, interpretations, and conclusions are the results of research supported by the Bank; they do not necessarily represent official policy of the Bank. Library of Congress Cataloging-in-Publication Data Main entry under title: Social accounting matrices. Bibliography: p. 1. Developing countries-Economic conditions- Mathematical models-Congresses. 2. Social accounting-Developing countries-Methodology- Congresses. 3. Social accounting-Developing countries-Case studies-Congresses. 4. Economic development-Social aspects-Developing countries- Case studies-Congresses. 5. Regional planning- Developing countries-Mathematical models-Case studies-Congresses. I. Pyatt, Graham, 1936- II. Round, Jeffery I. (Jeffery Ian), 1943- HC59.7.S57 1985 306'.0724 85-12041 ISBN 0-8213-0550-6 (pbk.) Contents Preface ix Participants xi Introduction The Methodology of Social Accounting 2 Country Studies 5 Multipliers and SAM-Based Models 8 Some Concluding Conuments 13 PART I. THE METHODOLOGY OF SOCIAL ACCOUJNTING 1. What Is a SAM? 17 Benjamin B. King A Primitive Example: The Robinson Crusoe Economy 19 Sri Lanka 1970 21 Botswana 1974-75: The Flow of Funds 37 The Uses of a SAM 44 Learning by Doing 49 2. Social Accounting Matrices for Development Planning 52 Graham Pyatt and Jeffery I. Round Background to the Studies 53 The Iran Case Study 55 The Sri Lanka Case Study 59 The Swaziland Case Study 63 Concluding Comments 66 3. The Flow of Funds as a Tool of Analysis in Developing Countries 70 Alan R. Roe The Statistical Format 70 Prevailing Analytical Uses 73 The Role of the Financial System in Economic Development 77 The Implications for Flow-of-Funds Data Systems 82 Conclusions 83 v vi Contents 4. Regional Accounts in a SAM Framework 84 Graham Pyatt and Jeffery I. Round An Aggregate Malaysia SAM in Outline 85 Aggregate Regional Accounts in a SAM Framework 87 The Treatment of Interregional Commodity Trade 89 PART II. COTRY STUDIES 5. A Social Accounting Matrix for Sri Lanka, 1970 99 S. Narapalasingam The Current Policy Emphasis 99 The Data Base 102 Some Conceptual Problems 103 Summary and Conclusions 106 6. A Social Accounting Matrix for Swaziland, 1971-72 108 S. J. Webster The Swaziland Scene in 1974 109 The SAM: Structure, Classifications, and Conceptual Issues 110 Data Problems and Solutions 117 A Social Accounting View of Swaziland's Prospects 120 Issues of Administration and Organization 123 7. A Social Accounting Matrix for Botswana, 1974-75 126 C. C. Greenfield The Matrix and Its Broad Features 126 Detailed Features of the Matrix 131 Concluding Remarks on the Format of the Matrix 133 Flow of Funds 133 Estimation and Balancing 136 Conclusion 139 Appendix A: Building the Botswana SAM 140 Appendix B: Price Effects 141 PART III. MULTIPLIERS AND SAM-BASED MODELS 8. The Disaggregation of the Household Sector in the National Accounts 145 Sir Richard Stone Activities and Sectors 145 Households 146 A Numerical Example 148 Whatever Became of the Multiplier? 156 The Interpretation of the Matrix Multipliers 162 Contents vii Where Do We Go from Here? 178 Conclusions 181 Appendix: Multipliers for Quesnay's Tableau 181 9. Accounting and Fixed-Price Multipliers in a Social Accounting Matrix Framework 186 Graham Pyatt and Jeffery I. Round The Social Accounting Matrix 187 Decomposition of Accounting Multipliers 192 Fixed-Price Multipliers 197 Decomposition of Fixed-Price Multipliers 201 Empirical Results 201 Conclusions 205 10. The Social Accounting Matrix and Consistency-Type Planning Models 207 Erik Thorbecke The Social Accounting Matrix as a Data Framework 208 Comparative Evaluation of First-Generation Models 217 Comparative Evaluation of Second-Generation Models 232 Concluding Remarks and Suggestions 251 11. Social Cost-Benefit Analysis in a Semi-Input-Output Framework: An Application to the Muda Irrigation Project 257 Clive Bel and Shantayanan Devarajan The Model 258 Applying the Model to the Muda Irrigation Project 263 Estimation of National Parameters for Social Cost-Benefit Analysis 266 Social Cost-Benefit Calculations 269 Conclusions 272 References 276 I I Preface The conference in Cambridge, England, that was the origin of papers in this volume has been described as a meeting of a subculture. This character was recognized before the event and little occurred to alter the perspective. Many, but not all, of the participants were well known to at least some of the others, so there is a particular debt to those who were unfamiliar with the fraternity that otherwise defined the group. We would here like to make special mention of C.P. Ezeife, who throughout the proceedings contributed much to the focus of discussion and always stood firmly for a measured practicality at points where one individual or another was bent on pursuing some esoteric irrelevance. Both the spirit and content of his participation were greatly appreciated. The full list of conference participants follows this preface. Unfortunately, some others who we hoped could attend were unable to do so. This was a pity, not least because social accounting as practiced by many of the present authors had been developing rapidly during the 1970s, and it was thought to be time that their ideas and achievements were exposed to a wider and experienced audience. This is one reason the World Bank External Research Program was prepared to finance the conference and is also a reason for publishing this volume based on the proceedings. What started as a small and speculative exercise in Iran has now developed a life of its own. At the same time, we are conscious of legitimate complaints over lack of docu- mentation on what social accounting matrices (SAMs) are, what they can do, and how one (or preferably a team) sets about their construction. The conference and this volume are one response. We hope the papers will not encourage a stultifying conformity in future efforts, since the challenge of defining one's own SAM frame- work, in order to capture the main macroeconomic and planning issues to be faced in a particular country, is perhaps one reason SAMs have attracted so many enterprising and creative spirits. Of course, SAMs have roots, and Sir Richard Stone traces these to eighteenth-century France and Quesnay's Tableau Economique. Others at the conference would be largely content to go back only as far as Stone's personal contributions to national accounting. There are important contributions by others, especially in the interwar years, which should be acknowledged. But Stone is the great architect of SAMs, not only as national accounting frameworks but also in relation to macroeconomic planning models. Our individual personal and intellectual debts to him have been yet further increased by his lively and authoritative contributions to the confer- ence and this volume. A further debt is to Benjamin B. King, who has helped at various stages from inception of the proposal to hold a conference through to completion of the editorial tasks. As one of the few nonmembers of the SAM subculture at the conference, he has been willing to write an introductory chapter from the outsider's point of view under the daunting title, "What Is a SAM?" His willingness to do so has considerably eased our task as editors. Anne McKenna organized the Cambridge conference and the production of this volume through its early drafts. Beyond these essential contributions, her advice often extended into more subjective areas. Nancy Ribeiro also helped in many ways to administer the volume through the latter stages. We are greatly indebted to them both. Our sincere thanks also go to Cathy Bau, who painstakingly undertook some of the editing work, and to Ann Van Aken and Ann Robeson, who assisted in the production of the final text. ix x Preface We would like to thank the conference participants and regret that it is not possible to include all their papers in this volume. To many we also owe a debt of friendship forged in particular research studies that could not have been achieved without the bond of a shared enthusiasm. We hope that these friends gain some satisfaction from seeing the sum total of past efforts- to the extent that this volume is able to capture it. One friend in particular, Harry Fell, merits special mention. We hope he will accept the dedication of this volume to his example. Finally, we must thank the editors of the Review of Income and Wealth, the Economic Journal, and the Pakistan Development Review for permission to reproduce previously published mate- rial in chapters 2, 9, and 11, and disclaim any responsibility of the institutions to which we and other authors of papers in this volume are affiliated for the views expressed. Graham Pyatt Jeffery I. Round Participants The affiliations shown are those at the time of the conference, April 16-21, 1978. Terence Barker, University of Cambridge Clive Bell, The World Bank, Washington, D.C. Alan Brown, University of Oxford Ross Bull, Institute of Social Studies, The Hague Colin Dunn, Coopers and Lybrand Limited, U.K. C. P. Ezeife, Central Planning Office, Nigeria Harry Fell, retired from the Overseas Development Ministry, U.K. K. Gnasegarah, Department of Statistics, Malaysia Colin Greenfield, Overseas Development Ministry, U.K. Carol Hayden, University of Warwick Benjamin B. King, The World Bank, Washington, D.C. Robert Lindley, University of Warwick Frank Lysy, The Johns Hopkins University S. Narapalasingam, Ministry of Finance and Planning, Sri Lanka Graham Pyatt, The World Bank, Washington, D.C. Gerry Rodgers, International Labour Office, Geneva Alan R. Roe, University of Warwick Jeffery I. Round, University of Warwick Sir Richard and Lady Stone, University of Cambridge Erik Thorbecke, Cornell University Michael Ward, Institute of Developm'ent Studies, University of Sussex Stanley Webster, Coopers and Lybrand Limited, U.K. Donald Wilkes, Coopers and Lybrand Limited, U.K. xi I I I Introduction The data base for macroeconomic policy and planning is hardly a fashionable topic among economists. The subject has received relatively little attention since the publication of The Social Framework (Hicks, 1942), and subsequent efforts from the same era to establish concepts and standards for national income accounting. Such accounts provide the information needed for macroeconomic analysis, especially the short-term analysis of demand management, and for the longer-term, post-Keynesian, dynamics of the Harrod-Domar model and the two-gap model of Chenery and Strout (Chenery and Strout, 1986). To go further calls for a richer source of information, and the main extension has been in interindustry economics, with particular emphasis on the Leontief input-output model. This has led to much sophistication in the construction of models, as illustrated by the many applications of activity analysis to devel- opment planning, by Tinbergen's semi-input-output analysis (Tinbergen, 1966), and by the extensive application of the Balassa-Corden formalizations of effective protection (Balassa and associates, 1971, and Corden, 1966). Beyond these are many more extensive models of growth and development in particular economies, of which the Stone and Brown Cambridge growth model is a primary example (Cambridge University, Department of Applied Economics, 1962- 74 and 1975-). In all cases, the starting point is a view of the economy that sees aggregate production broken down into component activities. These activities are interdependent, since each requires inputs of raw materials which have to be purchased from some other activity. They also require imports to the extent that necessary raw materials are not available domes- tically. Since the rest of the world is typically an alternative source of supply for many of the goods that the domestic economy is capable of producing, the domestic activities are also in competition with imports. It follows that domestio production structures and patterns of trade are intimately connected. Detailed accounting of these connections can enrich analysis of aggre- gate demand effects on output levels and the balance of payments. Research in recent years has led to some considerable progress in our understanding of these matters. At the same time, the results have been unsatisfactory in an important respect. Development is about raising the living standards of people. Accordingly, the framework for data and models must recognize the central importance of people, not commodities, if it is to serve best the interests of policy design. Development economists have always recognized that their main concern is the improvement of living standards. It has not always been recognized, however, that this objective might call for the adoption of policies which differ from those that would maximize the rate of output growth. Today, while most would agree that output growth is a necessary condition for sustained improvement in living standards, it is also generally recognized that economic policy must simultaneously concern the distribution of benefits arising from growth, to the point where faster growth overall might be sacrificed for the sake of faster growth in the living standards of particular groups, espeoially poverty groups. This perception of economic development provides an underlying philosophy for many of the papers in this volume. Iltimately it calls for devel- opments in the theory of economic and social change which would embrace both the traditional concerns of growth economios and the agenda of issues that follows from a focus on income distribution, employment, and poverty alleviation. How this might best be done remains a priority for research. Meanwhile, pragmatists may be more concerned with the facts of the matter and will want to know how living standards for different groups actually change in the 1 2 Introduction process of economic development, even though some of the reasons for change are as yet unresolved. Indeed, such an approach can be seen as a useful first step toward an understanding of those reasons. The social accounting approach illustrated in the volume records particular attempts at meeting this practical need within the limitations of (a) currently available statistics, and (b) current perceptions as to the best way of assembling them. Beyond this, a number of the papers discuss uses of data once they have been arranged in a social accounting format. These uses inevitably involve an imposition of behavioral and technical assumptions on the data them- selves. To that extent such applications of the data cannot proceed without theory. But the concern here is not with the theory itself. Rather it is to show how the data base provided by a social accounting matrix (SAM) can be taken as a statement of initial conditions in an economy and how theoretical analysis can proceed from this starting point. Not least, the approach serves to emphasize the fact that the distribution of employment opportunities and living standards in a society is inextricably interwoven with the structure of production and the distribution of resources. The eleven papers in this volume have been grouped into three sets. The first four papers, which compose part I of the volume, illustrate the methodology of social accounting as a discipline within economic statistics. Part II then records some country experience in the construction and use of social accounts and is followed, in part III, by a,final set of four papers illustrating the step from data systems to models in a SAM context. THE METHODOLOGY OF SOCIAL ACCOUNTING The first of the four papers in part I, by Benjamin B. King, addresses the question "What is a SAM?" without assuming anything more of the reader than a rudimentary knowledge of national income accounting and a willingness to entertain the psychological novelty of having such accounts presented in an unusual format. Starting with a very simple formulation, he presents, step by step, many of the technical complexities that feature more prominently in subsequent contributions. This, then, is an introduction to the subject. The next paper, by ourselves, was written some time before the conference, although it featured in the discussion. The paper is included here because it contains relevant background concerning not only the nature and importance of SAMs, but also the early evolution of their practical application in addressing development problems, beginning with a SAMfor Iran (in 1970), and subsequent studies of Sri Lanka (also in 1970) and Swaziland (in 1971-72). The paper readily admits to the process of learning by doing, which was involved in these successive studies, and it begins by setting out the reasons for embarking on them. Prominent among these reasons is a concern to answer the question "Who gets what, and how much, as a result of the economic process of income generation?" This question is set in the context of conventional national income accounts and input-output analysis, with data displayed in the single entry matrix format, which is essential to a SAM and distinguishes it from the more traditional form of double entry bookkeeping. The innovation is to obtain a disaggregation of the household sector within this format so that income distribution is captured as differences between the incomes of various socioeconomic groups, in much the same way that the structure of production can be captured by disaggregation of productive activity into output levels for each of a number of different industries. This paper also includes discussion of some technical problems that arise in attempting to construct a social accounting matrix when both the production sector and the household sector are disaggregated. In the production and household sectors, some disaggre- gations are much more difficult to achieve than others. But it is also the case that some disag- gregations are much more interesting than others. Accordingly, the paper ends with comment Introduction 3 on appropriate criteria for disaggregation in a context of equal concern for "who gets what?" and "what and how much of it is being produced and consumed?" The studies of Iran, Sri Lanka, and Swaziland had the limited ambition of disaggregating the current account for households simultaneously with production. The next obvious step is to disaggregate capital accounts and hence to integrate balance sheets and flow of funds with data on the flows within the real economy. Alternative ways of disaggregating capital accounts are discussed in the third chapter on SAM methodology, by Alan Roe. His contribution draws on specific experience in Botswana to illustrate both conceptual and empirical issues that can arise, depending on the format adopted. Roe then discusses current analytic uses of such flow- of-funds data in planning and in macromonetary models. This discussion accordingly illustrates the links between data systems and models. And insofar as Roe's argument leads him to conclude that the current generation of planning models fails to capture some of the crucial monetary issues of development, he is led automatically to a discussion of the data that are needed on flow of funds, that is, to the classification that would be required to capture the essential role of the financial sector in stimulating savings and allocating their use. Roe argues that market segmentation should be the basis of classifications in the financial accounts. His paper therefore serves to push forward an area for SAM development which is in its early stages, yet seems to depend on considerations that are comparable to those that have arisen earlier in analysis of the real economy. Flow-of-funds accounts are of considerable interest in their own right. To have a set of such accounts fully integrated with the real flows captured in previous SAMs adds to their value by forging the link between the real and monetary aspects of the economy. But it may be worth stressing that there is some virtue in obtaining a SAM without flow of funds as a first step; to then make the extra effort not only facilitates the flow-of-funds compilation, but also potentially improves the data on the real economy, not least insofar as they refer to savings. To set this point in perspective it is worthwhile to consider the basic steps by which national accounts and SAMs are compiled. For national accounts, the best procedure is to work with a balance equation for demand and supply of each commodity. Total supplies comprise imports plus domestic production. Total demand is final demand-exports, investment, and domestic consumption-plus intermediate demand or raw material requirements. The latter can be reconciled with domestic production by using input-output computations. Hence, if the supply side details are known, consistent figures for the demand side can be obtained by treating some component of demand (such as the stock change component of investment) as a residual balancing item or, more generally, by adjusting the data for each commodity so that a balance is struck between demand and supply. That balance should be consistent, to the extent possible, with available data sources and commonsense relationships such as the input-output formu- lation of raw material requirements. This approach is as much as is usually attempted to secure consistency. It produces the major aggregates needed for national accounting purposes, and these aggregates will satisfy the obvious accounting constraints, notably that total value added for all activities (defined as gross outputs minus intermediate demands) will be equal to aggregate final demand. Moreover, the approach implies a particular value added figure for each activity and a particular final demand figure for the goods that each activity produces. These details are consistent in aggregate, and in principle they are consistent for each activity. But are they? By disaggregating the accounts for factors and institutions the SAM approach embodies a further check. In the Sri Lanka study, for example, the factor disaggregation requires that the value added figures from national accounts must be broken down into payments to different types of labor, capital, and, in principle, natural resources. Adding up these details across activities produces a breakdown of total value added into its factoral distribution. Next, the SAMapproach requires that the factor incomes be paid out to institutions accordirng to the factor services they supply: the wages of particular types of labor go to the households 4 Introduction which supply the corresponding labor services; corporate profits go to the private corporate sector or to government if the enterprise is state owned. In this way total value added maps into the disposable income of each institution-that is, each type of household, company, and branch of government-before transfers. Current transfers between these institutions must then be estimated, thus leading to the distribution across various institutions of actual dispos- able income. To carry out these calculations requires data on incomes by source for various institutions. The main requirement is for household survey information on the relevant flows. Often this information is judged unreliable, so that the household data are adjusted to be consistent, for example, with national accounts data on the wage component of value added in each sector. This adjustment may be appropriate, although the evidence is growing that household survey data on income by source, and national accounts data on value added payments by activity, are seemingly unrelated in many, if not most, cases. In taking the national accounts figures as the firmer estimates, the authors of SAM studies may well have misplaced their trust, and data on the incomes of different institutions may have been adjusted too severely so as to permit the control totals provided by the national accounts to be retained. Be this as it may, the national accounts figures are directly challenged at the next step in SAM compilation. Having arrived at disposable income for each institution, the next step is to allocate these incomes (net of the current transfers already determined) either to current expenditures or to savings. But the current expenditure element of this process has already been determined to the extent that national income estimation has already required an esti- mation of consumption by commodity groups. This has to be reconciled, one way or another, with the consumption expenditures of government and the various household types whose disposable income was previously estimated. Household surveys provide a real check here to the extent that their consumption figures are generally thought to be more reliable. Suppose that the national accounts survive this check, that is, that observed discrepancies can be removed by adjustments elsewhere in the matrix which seem sensible and, at the same time, allow the national income figures to be retained at their original value. Then much of the weight of adjustment falls inevitably on the remaining elements of disposable income, namely, savings. In aggregate savings must be equal to investment. But in detail there is no control up to this point of the distribution of savings across institutions. Foreign savings, government savings, and corporate savings may all be known to some extent. If so, the burden of adjustment falls on household savings. The typical experience is that the figures initially obtained are so obviously worthless that there has to be some backtracking over the sequence of calculations which lead to this point so as to provide a reasonably sensible set of savings estimates for different types of households. Of course, what is reasonable and sensible is a subjective matter, and SAM studies have shown much courage and some ingenuity in coming up with answers. They have also typically left an uncomfortable feeling that, at some point, the principle of maintaining previously computed national income estimates should have been abandoned. Forging ahead into the flow of funds imposes yet a further round of consistency checks and a potentially firm basis for evaluating SAM estimates of savings by each institution. Depending on the quality of financial data, this check would be most valuable if there were some reasonably rigid structure to the flow of funds, comparable to that for input-output analysis. But in fact many financial flows are highly volatile, so that data problems, derived from the timing of payments and discordance of the financial year between institutions, weaken the feedback that the financial data could have in pointing up inconsistencies in previous estimates for the real economy. While these data problems are a pity, they should not detract from the value of the data in their own right, nor from the fact that there is some feedback. Just how much depends partly on the classifications adopted, and it is interesting to note in this context that an extremely rigid structure to the flow of funds may be justified, as in the formulation of Ahluwalia and Introduction 5 Chenery (1974), if the classification of institutions within the capital accounts can match that which is adopted for the current accounts also. In his foreword to Pyatt, Roe, and associates (1977) Stone pointed to the ultimately subjective nature of the numerous adjustments which have to be made to different data sets in developing a SAM or, for that matter, the national accounts. He refers to a much earlier paper by Stone, Champernowne, and Meade (1942), which proposed the application to social accounting of the well-known technique for the adjustment of conditional observations by the method of least squares based on a subjectively estimated covariance matrix of the errors. A little later a general formal statement of the problem was provided by Durbin but never published. At that time the state of the art of numerical analysis, together with the capacity limitations of computers, precluded the practical application of formal methods, but the subject was kept alive by Stone, who included a series of small constructed examples in a number of his papers. But times have changed, and more recently Byron (1978) has reformulated the problem in terms of a quadratic loss function and shown that, by using the conjugate gradient algorithm, even very large SAMs can be adjusted without great difficulty. The Byron method has in fact been applied in one of the studies reported in this volume (see chapter 11), although generally the authors have favored what have come to be known as RAS (or biproportional matrix transformation) procedures applied either to sub-blocks of the SAM or to the matrix as a whole, as in the Botswana case. The choice involves various issues which lie outside the scope of this volume. Although these issues are mainly technical, they also include the desire, especially among model builders, to obtain a data set that is purged of the "residual error" categories which some national income statisticians regard as the hallmark of their integrity. The choice of methods is also expanding, and a linear programming approach has much in its favor over constrained least squares, not least with regard to the greater control over results and the "zeroing-in" on a final solution which it facilitates. This approach is strengthened by the work of von Saleski (1977), who shows that the network specialization of linear programming allows much of its flexibility to be retained, yet with considerable savings in computational costs. The development of formal techniques for data reconciliation is clearly a subj ect for the future.' At present we have little experience in producing the national accounts from scratch as part of a more general SAM estimation. However, there are other areas in which progress has been made, and the final paper in this section addresses the conceptual problems that arise in constructing a multiregional SAM (in this case two regions) so as to capture the various flows that take place interregionally. The paper, by ourselves, is drawn from a more extensive study of Malaysia which is reported elsewhere (see Chander and others, 1980). COUNTRY STUDIES For developing countries, perhaps the first studies to have been undertaken with a simul- taneous focus on income distribution and production structure were those for Iran and Sri Lanka, which have been referred to previously. Both arose from the activities of the World Employment Programme, sponsored by the International Labour Office (ILO). The third study, which dealt with Swaziland, was sponsored by the Overseas Dwelopment Ministry, London, which subsequently launched a program of work in collaboration with Warwick University. This program has involved a series of exercises in Botswana and work in both Kenya and Fiji. 1. Van der Ploeg (1982) and others are developing more efficent computational procedures for tackling essentially the same class of adjustment problems. 6 Introduction Increasingly, other countries have undertaken SAM investigations, so that by 1984 there were SAMs for Cyprus, the Arab Republic of Egypt, Indonesia, Malaysia, the Philippines, Saudi Arabia, the Republic of Korea, Thailand, and Turkey. This list notably excludes any Latin American countries, but studies are now under way in Brazil and Mexico. To illustrate all this activity, and because of their pioneering nature, the three country studies which are discussed in this volume are those for Sri Lanka, Swaziland, and Botswana. Together they provide the primary materials for much of the discussion in part I of this volume and, together with the Iran study, may be regarded as the basis for work subsequent to the foun- dations laid by Stone and his associates on the Cambridge Growth Project. The Sri Lanka study has been fully documented in book form (Pyatt, Roe, and associates, 1977). Given this documentation and the material drawn from it in part I of this volume, the discussion of Sri Lanka in part II is a retrospective view by S. Narapalasingam, who was a member of the original study team. As a former deputy secretary, Ministry of Finance and Planning, in Sri Lanka, Narapalasingam reflects on the usefulness of the study for policy issues and ways in which the data base provided by the SAM could be improved from this perspective. He therefore builds on the essential notion that the purpose of social accounting is to inform current policy debates and that the construction of a SAMis not a once-and-for-all effort. Rather, it requires continuous updating to reflect changes in both economic structure and its environ- ment, and because of shifts in policy emphasis. In Sri Lanka the SAM for 1970 has not been maintained, while the focus of policy has shifted somewhat as the development problems of that country have hardened. The initial effort has contributed little that is now useful, beyond the demonstration of what could be done.2 The theme of Narapalasingam's paper is therefore one of regret that previous capital has been allowed to depreciate. At the same time, he points to a number of respects in which a new effort could represent an improvement for policy purposes. He also suggests a number of organizational and administrative changes with respect to primary data requirements that would be necessary to set up and maintain a data base for planning, which could also serve the needs of policy. In chapter 6, S. J. Webster presents a brief but comprehensive view of the Swaziland SAM, in terms of actual achievements and the motivations behind this particular study. He begins with a sketch of the country, and accordingly invites the reader to go through the subsequent discussion of compilations and data problems keeping one eye on the issues and perspectives that the statistical exercise was to capture. Next, a definition of classifications-for households, factors, and production activities-allows the salient features of Swaziland to be manifest. Some novelty is involved here, reflecting the particular socioeconomic structure of the country. It is to be noted that this study, in particular, expresses the view that in systems of classification and other respects social accounting should reflect the circumstances of the country it is designed to serve and not adhere to extraneous norms for the sake of an ultimately illusive comparability. Inevitably, desirable classifications have to be compromised by the limitations of available data. Webster discusses data limitations both from this perspective and with regard to timeli- ness. He argues, on the basis of Swaziland experience in particular, that much can be done in a small country even with relatively limited resources. The proof he offers is mainly in terms of the SAM actually implemented, some work on its updating, and the consequent perspective on development issues and strategies that emerged from the study. Finally, as Narapalasingam has done, Webster addresses some administrative and organi- zational questions. The Swaziland SAM, like the Sri Lanka SAM, has been allowed for the most part to lie dormant over subsequent years. Webster expresses concern over this and rehearses 2. We understand that moves are under way to resurrect the earlier work. Introduction 7 some of the possible causes. He makes explicitly a point that is implicit in much of what Narapalasingam has written, to the effect that a SAM approach to macroeconomic data is potentially a lively and constructive link between statisticians and economic planners. The SAM approach therefore offers relevance and dialogue to a relationship between two professional groups which is all too often observed to be muted. The next country study is of Botswana and was undertaken in 1977 by a team under the leadership of Colin Greenfield. His paper, chapter 7, covers a number of the topics previously discussed by Narapalasingam and Webster but adds new dimensions, which derive partly from the accumulating experience on which the Botswana study was able to draw and partly from the fresh perspective introduced by Greenfield himself and others for whom the Botswana SAM represented an initial involvement in the SAM approach. Beyond these fresh perspectives on issues of classification and data problems, Greenfield introduces three new elements of considerable interest. One of these is a highly technical matter, concerning the choice of price systems for recording commodity flows in a SAM. The discussion of this is largely contained in the second appendix of his paper, but it is in fact quite central. Secondly, Greenfield organizes much of his early discussion of the Botswana SAM around a comparison of how it differs as a conceptual framework from that recommended in the United Nations System of National Accounts (known as the UN SNA). Hence, the question is directly posed as to how the SAM approach illustrated by the studies of Sri Lanka, Swaziland, and especially Botswana differs from the framework currently recommended to developing countries by the United Nations Statistical Commission (United Nations Statistical Office, 1968). Stripped of qualifications, Greenfield's answer is that SAMs are simpler, and hence have a better chance ofbeing "generally understoodby economists and others concerned with planning the economy." He goes on to express a sentiment that most who attended the Cambridge conference would also want to emphasize: "It would be wrong, however, to conclude with an impression that one has nothing but criticism for the SNAformat. Quite the reverse-one has nothing but admiration for the pioneering and meticulous work recorded in the SNA which provided an invaluable guide and source of reference in preparing the Botswana SAM." The conflict, if there is one, can be traced back to the common root previously identified, namely, the Cambridge Growth Project. The UN SNA has emerged from this project through the social accounting matrix developed in Cambridge as the counterpart of a growth model for the United Kingdom. Data systems and models are inseparable, and the SAM studies recorded here simply reflect a different, albeit implicit, model of what the issues are in a developing country, while retaining the same essential principles of macroeconomic accounting. The differences, of both issues and practicalities, argue for a simpler framework and for some greater attention to classifications within the framework. A single household sector will not do if we are concerned about who benefits from development; thus the Botswana SAMfollows the priorities of policy in having nine different types of house- hold, with the differences corresponding to real socioeconomic distinctions: periurban house- holds; urban households (split by density of housing area and with residents of servants' quarters as a separate category); three types of rural households (defined with respect to their ownership of cattle); and finally, migrant workers abroad. The same spirit is reflected in the disaggregations chosen for the corporate sector and production activities. But the SNA prin- ciples and, not least, the matrix approach are rigorously adhered to throughout the Botswana SAM. Iltimately there is no conflict. The issues essentially concern only the route to be followed from the existing data base toward improved macroeconomic accounts. While previous SAMs had all attempted some disaggregation of the current account for insti- tutions, that is, for households, companies, and government, they had made little or no effort to disaggregate the corresponding capital accounts and thereby capture details of the flow of funds. The third striking feature of the Botswana SAM, directly prompted by the priorities of the decisionmaking process, was to include no less than seven separate capital accounts for 8 Introduction institutions, together with thirteen accounts for financial claims. Greenfield discusses this aspect of the Botswana SAM in some detail, which is justified by its importance. Inevitably, the innovation led the team into a new set of issues of classification and data reconciliation. Perhaps the main points to emphasize here are that many of the problems were overcome and that the value of the SAM to policymakers in Botswana was considerably enhanced as a result. 3 This is not surprising, given the well-recognized need for such a development, which was expressed by Narapalasingam in his comments on the limitations of the Sri Lanka study. At various points during discussion at the Cambridge conference, questions were raised and opinions expressed about the merits and limitations of the SAM approach to national economic statistics, not just as a research or planning exercise, but as a thematic, organizing framework that governments might choose to establish on a permanent basis. Harry Fell and Colin Green- field, in particular, made some interesting points in this context, drawing on their specific experience with SAMs in Swaziland, Botswana, and Saudi Arabia and their many years of experience as practicing statisticians in numerous countries. It was instructive, therefore, to hear their views on the SAM approach as a highly effective method of making the best use of "dirty" and "gappy" data. Too bad that the data are not perfect; they never will be, and the only answer is to learn to live with this fact while always pressing for better primary sources. They also point to the various levels of sophistication of a SAM, each appropriate to its purpose, and come close to expressing the view of Pyatt and Thorbecke (1976) that the SAM approach is ultimately based on a philosophy of quantitative analysis that is open and honest in trying to establish the facts and then cautious in moving toward analysis and projections from this basis. More than anything else, a SAM is a generator of agenda for decisions: about what the facts are, about the priorities for improved information, and about the repercussions that a particular course of action might have. In this sense an underlying dynamic for change emerges from the static matrix of numbers. This, in itself, is a reason for not worrying too much about how bad the basic data are since, as Fell and Greenfield point out, either the SAM will not be used or the process of using it will create the forces for improvement. Experience also shows that whether a SAM is used depends in part on whether a country can move beyond an initial effort for some base date to the creation of a continuing capability to generate and maintain future SAMs, so that up-to-date information is available. Equally, however, such a development is unlikely unless the SAM is used. Thus data and analysis are mutually dependent. It is therefore encouraging to learn of instances in which SAMs are proving to be practical aids to policy formulation. Such instances seem more likely to occur when SAMs have been perceived at an early stage as a meeting ground for statisticians and planners to their mutual advantage. MULTIPLIERS AND SAM-BASED MODELS The relationship between SAMs and models is twofold. On the one hand, modeling is a major area of application of SAMs, and the four papers in part III of this volume illustrate the point. On the other hand, models are important as a formalization of particular conceptual frame- works. Without such frameworks, data gathering is largely an empty exercise. The early work on SAMs in the Cambridge Growth Project was guided by the needs of a particular planning model, and later developments in the form of the U1 SNA were similarly conditioned by a somewhat less complete framework of analysis, namely, input-output. In the same spirit, the 3. A brief discussion of various uses to which the SAM has been put in analyzing some policy issues in Botswana can be found in Hayden and Round (1982). Introduction 9 1972 SAM for Iran was conceived as the data base for a planning model that solved simulta- neously for incomes and production levels. Subsequent SAMwork has been directed essentially to the task of facilitating more sophisticated modeling treatment of this and other issues. Analysis, if not formal modeling, is therefore the ultimate driving force. This general position on the relationship between SAMs and modeling underlies the four papers presented here. The first of these, by Sir Richard Stone, provides a ready link from the country studies in part II to the subsequent contributions in part III. Stone is the primary architect not only of the UN SNA but also of the System of Social and Demographic Statistics (SSDS). The treatment of income distribution questions in the latter is conceptually consistent with the treatment of the household sector in the former. The SAMs discussed in this volume can be seen as an extension of the SNA by disaggregation of the household sector, and it is this perspective that Stone has adopted. The first part of his paper carefully enumerates the conceptual differences between the SNA definition of households and the narrower, everyday use of the term. (The former includes nonprofit institutions, foreign visitors, and certain aspect of life insurance, or mutual benefit, societies.) He then illustrates how disaggregated accounts for households proper-he uses seven groupings according to income levels-can be integrated into a SAMfor the United Kingdom for 1968 which otherwise stays close to the established SNA format. Next Stone asks, "What became of the multiplier?" He proceeds, after a short history of the concept, to apply and extend a set of computations which we ourselves initiated in chapter 4 of Pyatt, Roe, and associates (1977) and applied to the Sri Lanka SAM. Stone's application is to the U. K. SA.Mpreviously derived, and the extension is to provide an additive version of a previously multiplicative decomposition of (I - A) - l. But here (I - A) - I does not refer to the familiar Leontief inverse: that is only a part of it. Rather (I - A) in the present context is a multiplier matrix which links all endogenous income levels in a SAM to exogenous injections. Hence, if the endogenous incomes include those of factors, institutions, and activities, then (I - A) - 1 embraces not only what happens within these broad groups of accounts (for example, the interindustry transactions among production activities), but also what goes on between them, not least the full circular flow of income around the familiar macroeconomic loop of demands on activities, leading to demands for factors, hence to the incomes of institutions, and from there back to demands on activities. Stone provides an extensive set of tables for his illustration which, together with the textual commentary, provide a full exposition of a method for manipulating the data in a SAMso as to make explicit many of the structural characteristics that might otherwise be missed. Finally Stone asks, "Where do we go from here?" and points to the obvious step of more sophisticated modeling of economic structure. Beyond such a step, he focuses on various aspects of the household sector which ultimately require further elucidation: the dynamics of household formation and demise; the role that households play as producing units, especially in developing countries, and the imputation of incomes from asset ownership; and, finally, the need for further research on household consumption behavior and the integration of household balance sheets into the SAM framework. This ambitious list is additional evidence of the tendency for each stage of SAM development to suggest an agenda for further progress. "It seems to me that of all the interesting and useful things that could be done to improve the national accounts, the one most worthy of consideration is the disaggregation of the house- hold sector." This is Stone's main conclusion and, in large measure, it is what the Cambridge conference was all about. In a short appendix to his paper, Stone demonstrates that Quesnay's Tableau Economique belongs to the class of SAMs that focus on links between the structure of production and the distribution of income. He applies multiplier decomposition analysis to the Tableau, providing a set of calculations which incidentally were made the weekend before the conference in prep- 10 Introduction aration for it. The final sentence of this appendix is worth noting: "Short of providing an initial stimulus to artisans, the next best method of helping them is to stimulate (nonagricultural) activities. This will do less for landlords, farmers, the state, and the church than a stimulus of the same size applied to any of the other accounts." So much for the economic structure of France in the 1750s. One is reminded by this of some conclusions, similarly based on a simple SAM and multiplier analysis, which emerged from the Iran study referred to in chapter 2: "Undoubtedly our most important conclusion regarding economic policy in Iran relates to the performance of the livestock and agricultural sectors . .. the potential contribution of agriculture to the general development of the economy is very great ... for Iran to ignore these sectors could be disastrous" (Pyatt and others, 1972). It would seem, then, that SAMs and their asso- ciated multipliers have relevance to political economy in recent times as they did in the eigh- teenth century. Chapter 9, again by ourselves, takes our earlier work on multipliers a stage further and adopts Stone's additive version of multiplier decomposition. The data used refer to Sri Lanka. The extension here is to contrast the accounting multipliers that can be obtained directly from a SAMwith the incremental multipliers that can be obtained from a fixed price model. The link is thus made from accounting structures to simple models, and the different results in the two cases are shown to depend on income effects in those instances where income or expenditure elasticities differ from unity. Because the SAM recognizes separate accounts for urban, rural, and estate households, the multiplier analysis serves to point up the extent to which the estate sector in Sri Lanka is isolated from the rest of the economy. More generally, the multiplier decomposition confirms an observation made by Stone, yet in an obviously different economic environment. The decomposition of multiplier effects distinguishes the first- and second-round implications of an injection from the third- and higher-order effects which depend on the complete cycle of circular income flows. The observation is that the distribution of these higher- order effects across households of different types is essentially independent of where in the economy the initial injection takes place. It follows that a changed pattern of injections will change income distribution to the extent that different groups are the immediate beneficiaries. But to the extent that there is a multiplier effect, so that income increases exceed the size of Initial injections, the distribution of benefits is rather rigidly determined. In other words, the distribution of income is insensitive to a degree with respect to the pattern of injections: what matters ultimately is the pattern of skill endowments across household types, the role of govern- ment, and the distribution of wealth. The threshold from accounting to simple models having been crossed, chapter 10, by Erik Thorbecke, takes us deeply into macromodel applications. His long and valuable paper is presented in four sections. The first gives a general exposition of the relationship between SAMs and macromodels. At one level, a SAM is a static picture of numerical structure at some base date. It provides the initial facts, and hence much of the information that a model must be calibrated to reproduce if it is to replicate the base period accurately. In addition, the SAM can be seen as a modular conceptual framework, with different blocks of the SAM representing the different modular components which, together, can describe how the economy will move in response to exogenous change or as otherwise dictated by its own internal dynamic. It is this second aspect to which Thorbecke particularly applies his pedagogic skills, and it sets the scene for the two sections which follow. The next sections describe six macroeconomic models, each set in a SAM framework, and each therefore an illustration of the dictum that every economic model has an accounting framework. Thorbecke splits the six models into two groups of three and refers to them as first- and second-generation models. The distinction is that for the first-generation models, the SAM as a data framework can equally be interpreted as the reduced form relationships of the model structure: the behavioral relationships have an essentially linear relationship to the accounting Introduction 11 structure. This is not so for second-generation models, which are nonlinearly related to the corresponding SAMs. The three first-generation models are (a) the Iran model by Pyatt, in association with Bharier, Lindley, Mabro, and Sabalo; (b) Thorbecke's joint work with Sengupta on a model for Colombia; and (c) a model by Ng for the Philippines. Since Ng was a student of Thorbecke's, and since the Pyatt-Thorbecke collaboration over several years is well known, the selection of material may, perhaps, be questioned. It might be useful therefore to record that the first two papers have a common root in the ILO World Employment Programme, and that it was an awareness of each other's individual and independent responses to the issues for macroeconomic planning, which this program raised, that fostered the Pyatt-Thorbecke collaborative relationship. Beyond that, we are not aware of any earlier work that has attempted a simultaneous determination of income distribution, production structure, and economic development in quite the same way.4 Both these first two models assume fixed prices, while prices are endogenous in the Ng model. By assuming that technology can be adequately explained in terms of generalized Cobb-Douglas models, Ng's model has much in common with the pioneering contribution to development planning of Johansen (1960). In contrast, the second-generation models are known in the jargon as CES-CGEmodels: CES indicating that constant elasticities of substitution (not neces- sarily equal to unity) are assumed to characterize production relations; and CGEbecause these are computable general equilibrium models in which prices adjust so that markets are cleared, essentially according to Walras' law. The three second-generation models are (a) Adelman and Robinson's model of the Republic of Korea; (b) the Lysy and Taylor model of Brazil; and (c) the Ahluwalia-Lysy model of Malaysia. All three models have been developed as part of the World Bank's research program, which also sponsored the Cambridge conference as previously noted. Thorbecke's discussion of these six models will probably be found invaluable by those who wish to absorb trends in macromodeling for development planning and to form a perspective on the state of the art. His synthesis is admirable from this perspective, and his final summing up is an equally admirable, low-key commentary on some of the misgivings that others have expressed less temperately on the usefulness of the second-generation developments. One perspective is this: the first-generation models make no claim to ultimate realism. They can be based explicitly on a SAM,which is openly available for comment and improvement. They then perform some sensible calculations in the spirit that the assumptions made are strong simpli- fications. Hence they arrive at suggestive conclusions which are a starting point for policy debate. The second-generation models, in contrast, are way down the spectrumr toward the black- box end. If one is really concerned about living standards and development, the art of macro- modeling has some way to go before it can be trusted on these issues. For the present, the simpler approaches may be the best, and inspired research is urgently needed. More recent developments in the modeling arena tend to confirm the views and perspectives expressed in Thorbecke's paper. The literature on CGEmodels has become extensive, although much of it has now been consolidated in the book by Dervis, de Melo, and Robinson (1982). Nevertheless, such models remain black boxes to many potential users and, moreover, can be costly to develop within developing countries. As a result, current research is being directed toward alleviating these problems. For instance, the transactions values, or TV, approach can easily accommodate a range of behavioral relationships between economic agents selected by the user (see Drud, Grais, and Pyatt, 1983, and Grais, 1982). But because it is explicitly a SA.M- based system, it has the virtue that it is highly "user-friendly." Finally, a paper by Clive Bell and Shantayanan Devarajan reflects work subsequent to the Cambridge conference at which Bell presented earlier results that he and associates have obtained in tracking the early SAM macroeconomic studies in an extensive study of a large irrigation 4. The nearest precedent we have been able to identify is Desai (1961). Table 1. 1. King's "Scheme of the Income and Expense of the Several Families of England Calculated for the Year 1688" (in pounds sterling) NJumber Total of the Yearly Total of Ranks, degrees, titles Heads per Number of Yearly Income estates or income Expense Increase increase families and qualifications family persons per family income per head per head per head per annum 160 Temporal lords 40 6,400 2,800 448,00 70 60 10 .. 64,000 26 Spirituallords 20 520 1,300 33,800 65 55 10. . 5,200 800 BaronLets 16 12,800 880 704,000 55 51 4 51,200 600 Knights 13 7,800 650 390,000 50 46 4 .. 31,200 3,000 Esquires 10 30,000 450 1,200,000 45 42 3 . .90,000 12,000 Gentlemen 8 96,000 280 2,880,000 35 32 10 2 10 .. 240,000 5,000 Persons in offices 8 40,000 240 1,200,000 30 27 3 . 120,000 5,000 Persons in offices 6 30.000 120 600,000 20 18 2... 60,000 2,000 merchantsand tradersby sea 8 16,000 400 800,000 50 40 10... 160,000 8,000 Merchants and tradersby sea 6 48,000 200 1,600,000 33 28 5 .. 240,000 10,000 Persons in the law 7 70,000 140 1,400,000 20 17 3 . 210,000 2,000 Clergymen 6 12,000 60 120,000 10 9 1 .. 12,000 8,000 Clergymen 5 40,000 45 360,000 9 8 1 ... 40,000 40,000 Freeholders 7 280,000 84 3,3650,000 12 11 1 .. 280,000 140,000 Freeholders 5 700,000 50 7,000,00 10 9 10 .. 10 .. 350,000 150,000 Farmers 5 750,000 44 6,600,000 8 15 8 10 .. 5 .. 187,500 16,000 Persons in sciencesand liberal arts 5 80,000 60 960,000 12 It 10 1 10 40,000 40,000 Shopkeepersand tradesmen 4-1/2 180,000 45 1,800,000 10 9 10 . 10 . 90,000 60,000 Artisans and handicrafts 4 240,000 40 2,400,000 10 9tO. 10 .. 120,000 5,000 Naval officiers 4 20,000 80 400,000 20 18 . 2. , 40,000 4,000 Military officers 4 16,000 60 240,000 15 14 I.1... 16,000 511,586 5-1/4 2,675,520 67 34,495,800 12 18 12 ,. 18 .. 2,447,100 Decrease 50,000 Common seamen 3 150,000 20 1,000,000 7 . 7 10 . 10 , 75,000 364,000 Labouringpeople and outservants 3-1/2 1,275,000 15 5,460,000 4 10 4 12 . 2 . 127,500 400,000 Cottagers and paupers 3-1/4 1,300,000 6 10 2,000,000 2 . 2 5 .. 5 325,000 35,000 Common soldiers 2 70,000 14 490,000 7 . 7 10 lo10. 35,000 849,000 3-1/4 2,795,000 10 10 8,950,000 3 5 3 9 .. 4 .. 562,000 ........ Vagrants ... 30,000 ... 60,000 2 .. 3 .. I. .60,000 849,000 3-1/4 2,825,000 10 10 9,010,000 3 3 3 7 6 . 4 6 622,000 So the General Account is: 511,586 Increasingthe wealth of the Kingdom 5-1/4 2,675,520 67 34,495,800 12 18 12 . . . 18 . 2,447,100 849,000 Decreasingthe wealth of the Kingdom 3-1/4 2,825,000 10 10 9,010,000 3 3 3 7 6. . 4 6 622,000 1,360,586 Met totals 4-1/4 5,500,520 32 43,505,800 7 18 7 11 3 . 6 9 1,825,100 Source: Gregory King (1696). Introduction 13 project in the Muda region of Malaysia. The paper included here develops a model framework that began with a standard multiplier analysis and proceeded to a more elaborate cost-benefit formulation. Similarities between the approach and Tinbergen's semi-input-output framework were discussed in Cambridge. These similarities are explicit in the paper presented here, which goes a long way toward a synthesis of Tinbergen's primal approach to cost-benefit analysis and the dual (shadow-price) approach of Little and Mirrlees (1974). The model is SAM-based in the sense used by Thorbecke, as well as with respect to its data base. Accordingly, it furthers the direct application of SAMs to regional projects as explored by Round in relation to a project in Swaziland, referred to by Webster in chapter 6. SOME CONCLUDING COMMENTS It has unfortunately not been possible to include in this volume all of the papers presented in Cambridge. However, the selection and supplementary material assembled here will give a fair impression of the work discussed and, more generally, of the SAM approach as it has been developing over the years. In many respects it should be left to the reader to draw conclusions on what has been achieved and the fruitfulness of the SAM approach to the issues that have been addressed. However, it may be worth recording here that there was a consensus among the conference participants on a number of issues which others might like to entertain in reaching their own evaluation. Four of the major points are set out below. First, incomplete data of variable quality inevitably face the quantitative analyst. A SAM is an invaluable tool in bringing together whatever data there are and in helping to fashion a quantitative description of the initial position in an economy. Those who have tried it doubt whether those who have not can appreciate fully the extent and nature of SAM's advantages in these terms. The method is so superior that other approaches are even suspect, both because they imply inferior data (accounting constraints reduce the variance on estimates, given reason- able luck and judgment) and because it is most unlikely that a comparable sense of the data limitations can be developed and documented in any other way. Documentation of SAMs has not featured in the papers in this volume, mainly because it is a highly specialized and somewhat tedious business. But, to the best of our knowledge, such detailed documentation has been written up for all the SAM studies discussed in the various papers. It leads to an immediate sense of priorities for future developments in statistics and the opportunity to test how new information changes results. Second, the emphasis in the present set of SAMs is toward disaggregation of the institutional accounts but otherwise to simplify the SNA guidelines to national statistical authorities. The feeling is that little is lost by the simplifications, while much is gained in making the results more intelligible to nonspecialists. To this gain can be added the increased interest which follows from a disaggregation of the household sector, especially when the disaggregation is according to socioeconomic criteria. Those concerned with policy are much more likely to be intrigued by such a SAM than they are by conventional national accounts, and this increased possibility for communication should not be dismissed lightly. More broadly, the concern to put "people" at the center of national accounting is a theme which has already been addressed. But, oddly, since none of the papers makes reference to it, the fact that national accounts had such an orientation at their inception can be noted by reproducing in table I.1 the national income of England in 1688 as set out by Gregory King. The third point to be emphasized in this final summing up is that taxonomies matter. A SAM is not a model; however, accounting constraints are part of a model and typically an important part. Moreover, the classifications adopted within a SAM are really quite crucial. The choice of classifications is potentially much more important than the choice, say, of a consumer demand 14 Introduction system, especially when it is recognized that the adding-up criterion must be satisfied by the latter because the accounting framework dictates this. Market imperfections or behavioral differences can only be captured if classifications and disaggregations are suitably defined. All too often the responsibility of a model builder to define and justify his choice of classifications is honored in the breach. The SAM approach makes this breach less likely, according to the evidence of these papers. Finally, while SAMs are not models, they are clearly a stepping-stone in that direction. Much remains to be done in the development of model frameworks at both national and regional levels; many problems exist in model formulation as well as in SAM construction. The question for policy analysis and planning that is raised here is whether or not the SAM approach to such matters is a useful line to pursue. PART I The Methodology of Social Accounting I I 1 What Is a SAM? Benjamin B. King A social accounting matrix, familiarly known as SAM to the limited fraternity familiar with it, has two principal objectives. The first is concernedwith the organization of information, usually information about the economic and social structure of a country in a particular year, though it could as well be about a region in a country, a city, or any other unit one is interested in; the unit of time, though convenient, is arbitrary. Complaints about the inconsistency and unreli- ability of economic and social data in developing countries have reached the point of being trite. Although there is justification for these complaints, they are not the whole story. There is often information, dispersed or fragmentary, which is not used for lack of a framework to make the maximum use of the available information and to pinpoint with greater accuracy and specificity the salient gaps and inconsistencies. Once the data in a particular country for a particular year have been organized in the form of a SAM, they present a static image which can reveal much about the country's economic structure. Even so, the image is only a "snapshot." In order to analyze how the economy works and to predict the effects of policy interventions, more is needed than just a static image. A model of the economy has to be created which can simulate, for example, the effects of inter- ventions. This is the second objective of a SAM: to provide the statistical basis for the creation of a plausible model. The principle of a SAM is really nothing more than that of double entry bookkeeping in accounting. A SAMis a series of accounts in each of which incomings and outgoings (or income and expenditure in many cases) must balance. What is "incoming" into one account must be "outgoing" from another account. In this respect, a SAMresembles traditional national accounts. In fact, as will be demonstrated later, a SAM embodies the information normally included in national accounts and much more. In a SAM the double entries are achieved by only a single entry in a matrix which resembles an oversized chessboard. Each account consists of one row across the board and one column down it; both are identically numbered. We shall explain how this works shortly. How large the matrix is depends on the limitations of the available data and the motivation one has for constructing it. In principle, there is no limit to the fineness of detail. In practice, both the data and the effort available for constructing the SAM impose limitations. One of the original motivations for the elaboration of SAMs has been the growing interest in issues of poverty and basic needs. If one wishes to show how different activities affect or are affected by different socioeconomic groups in society, the amount of detail must correspond to the differ- entiation one wishes to make. This paper does not attempt to examine the problems arising in the construction of SAMs or the methods of economic analysis which use the assembled data. It is not addressed to specialists but to the broader audience of those who need an introduction to SAMs. In the rest of the paper we shall proceed with simplified examples of SAMs, based on more elaborate ones published elsewhere. We shall start at the simplest level with a purely imaginary economy and proceed by increasing the size and the complexity with examples of SAMs worked out for real economies, Sri Lanka and Botswana. The reason for using two different vehicles, rather than 17 Table 1.1. The SAM for Robinson Crusoe _ _ _ _ _ Expenditures Total 2 34 5 6 R I Income ~~I l _ _ I l _ _ 1 _ I | _ 1,000 _ 1- _ _ _ I -- _ 1,000 _ c 2 Demand | 1,000 l 1 l l 1,000 i p 3 ...... _ _ _- _ - _ - 1 _ 1 _ _ . t 4 Production _____l_|_ 1,000 | ll 1,000 6 Total 1,000 1,000 I 1,000 .1 WhatIs a SAM? 19 carrying the reader nonstop on one, is that they illustrate the fact that SAMs may be constructed in different ways-or, more properly, with different accents-for different purposes. Some points of interest may occur in one and others in another. A PRIMITIVE EXAMPLE: THE ROBINSON CRUSOE ECONOMY As an expository device, the Robinson Crusoe economy has perhaps become rather shopworn. Nevertheless, for want of something better, we shall use it again as a point of entry into the description of SAMs. We shall assume that Robinson Crusoe engages in only one production activity, the picking of coconuts. In some given period, he picks 1,000 coconuts. This represents at the same time the level of production, the level of his income, and the level of his demand for products (sometimes called "wants"). All three are equal, as they must be in such an economy. The structure of this economy is set out in table 1.1, in which two columns and rows have been left blank because they are not being used for the time being. The final column and row around the border show the total of each row or column. Within the border there are three entries, each of them equal to 1,000. These constitute the SAM for Robinson Crusoe. In a SAM the rows represent incomings and the columns outgoings. For example, row 1, which is labeled "income," receives 1,000 from column 4, labeled "production." In other words, Robinson Crusoe's income derives from production and equals 1,000. Now, turning to column 1, we see the corresponding entry 1,000 which represents the outgoings of income in row 2, which is labeled "demand." Column/row 1, in effect, describes Robinson Crusoe's role as an income earner. In column/row 2, we look at him as a consumer. His demand arising from income in row 2 is balanced by what he spends on production in column 2 (row 4). The third leg of this process is in row 4 and column 4, where he turns up a third time as producer: his demand for production is 1,000 and income arising from production is also 1,000. These various identities could be set out in the form of double entry accounts, but, although not so apparent yet, the matrix is more economical since it requires one entry for each item, whereas conventional accounts require two. It wild be noted that there is a circular process. If one put the three entries in the table in coordinate form with the row first and column second, they would appear like this: (1, 4); (4,2); and (2, 1). Thus, the matrix illustrates the circular process of demand leading to produc- tion leading to income, which in turn leads back to demand. Of course, this rather complicated way of setting out the trivial structure of the Robinson Crusoe economy might well be considered much ado about nothing. We present it this way, however, because it is so self-evident and can serve as an introduction to the more complex relationships in an actual economy. In real life, Robinson Crusoe as a member of a society may, indeed, fill all three roles-as income earner, consumer, and producer-but he would do so as a member of different sorts of units or subdivisions, according to his function. In the accounts for a whole society, income may be subdivided into many different categories, among which income to labor and income to capital are only the first tier. That income accrues to a variety of domestic institutions, which are the source of demand: households with different character- istics, firms, and government (central or local). The outgoings or expenditures of these insti- tutions are spread over a variety of products, as indeed Robinson Crusoe's must have been; production thus can be divided into as many sectors or subsectors as is desirable or practical. In the next section, we turn to a description of a SAM which has been worked out for an actual economy. Although progressive complexities are brought in, the fundamentals remain the same. 0 Table 1.2. An Initial Aggregate SAM for Sri Lanka, 1970 (in miflions of rupees) Expenditures Total ____ 1 __2_3_4_ 1 2 1 3 _ _4 1 5 6 R 1 Factors of production I l_l _ |_ 11,473 | 11,473 | | c | 2 Institutions | 11,360 l 885 |497 12,342 I 3 Surplus or deficit l -425 1 425 _ I t | 14 p ~~~~~~~~~~~~~~~~~~~II Production | 11,312 | _ ' 2,113 13,425 I | 15 |___ Rest of the World 113 1,455 | j 1,067 | 2,635 | 6 _ Total _ 11,473 t Total 12,342 | lls473|~~~~~ 0 I 13,425 2,635 | I I_______________ ____ I ____ ____ I ____ I ____I~~~~~~~~~~~~~~~~~~ What Is a SAM? 21 SRI LANKA 1970 Among the first SAMs constructed was that for Sri Lanka. Sri Lanka is a country with a low average income per capita, but an unusually equitable distribution of income and high standards in meeting 'basic needs." Having successfully achieved a high quality of life for such a low income, Sri Lanka's need for more rapid growth of income and reduction of unemployment implied structural change; it also implied better understanding of the existing economic struc- ture. All the data used in this section are taken or adapted from one of the earliest publications documenting the construction of a SAM (Pyatt and Thorbecke, 1976). We start with a highly aggregated and simplified version, shown in table 1.2. In this table all six pairs of rows and columns are used, although one of them (3) has no entry in the column. Apart from the border totals, there are now eleven entries. Two of the titles used in the Robinson Crusoe economy were not the conventional ones used in SAMs,because in such a primitive economy they would have sounded rather fatuous. Column/row 1 is now called "factors of production" and column/ row 2, "institutions." A word of explanation is in order about these two fundamental ingredients of a SAM. Factors of production consist primarily of the labor and capital that are used in the process of production and receive income from it. But the production process draws these factors from where it can, without being overly concerned with the entities to which their owners belong. It is these entities that constitute the "institutions" in column/row 2. Foremost among them are households, which we may wish to study in different categories. Households may supply labor and capital through one or more of their members but act as a unit when it comes to spending the income from it. Other institutions are firms or corporations, both public and private, which provide capital. A third type of institution is government, central or local. It, too, may provide capital, but it has another important role in the production process-at least in the determination of the value of production-namely, the levying of indirect taxes. (It will be noted that these are all domestic institutions.) The two new accounts 3 and 5 (that is, the pairs of row and column) are there for different reasons. The account for the "rest of the world" (5) is necessary because Sri Lanka, though an island, is not an isolated one like Robinson Crusoe's, but has many transactions with the rest of the world. The "surplus or deficit" (3) is a direct consequence of these transactions; the transactions are normally not equal and must be balanced by borrowing or lending, or by the use of reserves. How these two new accounts fit in will be explained in a tour of the matrix. We shall start with the production account (4). In the row are the "incomings": the proceeds of Bales, at producers' prices, of 11,312 to institutions within Sri Lanka in column 2 and of exports of 2,113 to the rest of the world in column 5. These are exactly balanced by the "outgo- ings" or the cost of production in column 4: in row 5 payments of 1,067 for imports of materials going into production; in row 2 payments of 885 made to institutions during the production process, which are in fact indirect taxes on intermediate goods or imported materials; and finally, the value added by factors of production of 11,473 in row 1. (All figures in this section are in millions of Sri Lankan rupees.) If we move to the income account (1), the 11,473 just referred to is now interpreted as income to factors of production (again not differentiated as yet). In column 1 we see the disposition of this income to factors of production, consisting malnly of income of 11,360 accruing to insti- tutions in Sri Lanka and the small balance of 113 to the rest of the world. Moving now to the third main account-the one for institutions-we have in row 2 two incoming items already mentioned: income from factors of production (11,360) in column 1 and indirect taxes (885) in column 4. There is one additional receipt-transfers from the rest of the world (97) in column B. The components of this will be described later. Column 2 shows how the receipts of institutions are spent. In rows 4 and 5 there are goods Table 1.3. A Revised Aggregate SAM for Sri Lanka, 1970 (in millions of rupees) _ _ _pExpenditures Total) 1 1 _ 1 2 _ _ 3 1 423 45 1 6 1 R |le |1 I c 12 Factors Institutions of Production l 11,360 | 2,441 I | 11,473 885 97 I 11,473 14,783 | i i 3 Surplus or Deficit 1 -425 l 425 0 | t s 4 Production _ 11,312 | 4,660 2,113 118,085 I J5 |____ Rest of the World 113 1,455 _ 1,067 2,635 | 6 Total 11,473 14,783 0 18,085 2,635 l _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ __ _ _ _ _ _ _ _ _ _ _ _ _ _ _ __._ _ _ _ _ _ _ _ _ _ _ _ _ _ I __ _ _ _ _ _ _ _ _ _ _ _ _ _ __l_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ __ _ _ _ _ _ _ _ _ _ _ _ _ _ _l__ _ _ _ _ _ _ _ _ _ _ _ _ _ . What Is a SAM? 23 and services purchased: 11,312 being the value of goods produced in Sri Lanka, and 1,455, the value of goods imported in final form as opposed to those used as materials in the course of production. In addition, in row 3, the deficit of -425 is due to the fact that expenditure by institutions in Sri Lanka exceeds their receipts. In the course of this brief tour, we have now referred to every item in the two new accounts, 3 and 5, with one exception. This is the surplus of 425 in row 3 of the account of the rest of the world (5). The latter's total income in the row exceeds its expenditure in the column by this amount; the rest of the world's surplus exactly balances Sri Lanka's deficit. Consequently, Sri Lanka had to borrow or use its reserves to cover its deficit of 425 with the rest of the world. It is worth noting that the three largest entries in the matrix, by a wide margin, are in boxes (2, 1L), (1, 4), and (4, 2); these were the only entries in the Robinson Crusoe SAM. The core of the matrix is still the circular process of demand (2), production (4), and income (1). Most of the remainder are required by the existence of the rest of the world. The circular process is dual in nature. One can move round a circle clockwise or counterclockwise. If we go in the counterclockwise direction, corresponding to the order just given, we are implicitly following the flow of money. In (1, 4), factors of production, such as labor, receive money from production; in (2, 1), institutions, such as households, receive money from factors of production; and in (4,2), production receives money from institutions. We may think of the other direction as the supply of goods or services: factors of production to the production process, production to institutions, and institutions (by a slight stretch of the meaning) to factors of production. We can now give an example of how the SAM could be set up in the form of double entry accounts, of which there would be five. For example, the production account shows: Revenue (row) Expenditure (column) Domestic sales 11,312 Payments to factors of production 11,473 Exports 2,113 Indirect taxes (on production) 885 Imported materials 1,067 Total 13,425 Total 13,425 This is more or less the form of traditional national accounts. The complete accounts are fully articulated in this double entry form in the sense that each item in one account appears on the opposite side of another account. It can be readily appreciated that, as the number of accounts is multiplied, their interconnections can become rather hard to follow. A SAMis more economical in that it has only half as many entries as a series of double entry acccounts, and the interconnections between accounts are obvious. Diagonal Entries There are two modifications to table 1.2 in the next version-table 1.3. Both are diagonal entries: one at (2, 2) and the other at (4, 4). Nothing else in the table has been changed except, of course, the totals for rows/columns 2 and 4, which are each increased by the amount of the corresponding diagonal entries. What meaning is attached to a diagonal entry, which appears both as a revenue and as an expenditure to a particular account? It clearly can only mean that, in one case, institutions make certain payments to themselves or, in the other case, production units do the same. If we were only concerned with institutions or production units as a group, this would not serve much purpose. But that is not the case. In fact, we shall in due course split both accounts into several subaccounts. The diagonal elements represent the total of transactions among these subaccounts plus new diagonal elements within the subaccounts themselves, if they too are aggregations. By including diagonal entries, we have changed the meaning of the totals. In table 1.2, the total of column 4 was the total value of all goods and services (value added in Sri Lanka plus Table 1.4. Further Development of the Aggregate SAM for Sri Lanka, 1970 (in millions of rupees) ___ ___ Expenditures _ Total _____ ______ 1 | 2abc | 2d | 3 | 4 5 | 6 l I I 1 Factors of production I _ [ _ | 11,473 _ | 11,473 I e 2abc Institutions i 11,360 2,052 1,368 l 3 14,783 _ e | 2d Indirect taxes _ 389 | _ j 885 94 1,368 | P 3 Surplus or deficit -425 | l 425 0 | s |4 Production l 11,312 1 | 4,660 2,113 | 18,085 | 5 Rest of the World 113 1,455 1 | 1,067 | 2,635 6 Total 11,473 14,783 1,368 | 18,085 2,635 What Is a SAM? 25 import content) produced in Sri Lanka without duplication. In table 1.3, it represents the sum of the outputs of all production units. There is duplication in this sum to the extent that one production unit sells to another; the diagonal entry is the amount of this duplication. Similarly, the diagonal amount of 2,441 in row/column 2 represents moneys paid from one institution to another. Such transfers are a part of total incomes and total expenditures. Indirect Taxes In table 1.4 we have made an initial split in the institutional account (2). Instead of one line, there are now two: 2abc (foreshadowing further splits into 2a, 2b, and 2c), which we have given, rather lamely, the same title as before; and 2d, which is labeled indirect taxes. Part or all of each of the entries in the old line 2 have been extracted and entered on line 2d. Thus in column 2abc the original figure of 2,441 has been split into two entries, 2,052 and 389; in column 4 the old entry of 885 is now on line 2d; and the old entry in column 5 has similarly been split up. The amounts so extracted from row 2 have been replaced by a single figure equal to the sum of the parts extracted-1,368 in column 2d. Consequently, the total for row (and column) 2abc is the same as before. Why do we need a separate account for indirect taxes? The main reason is that they should be distinguished from direct taxes. The latter are extracted from the income stream and, there- fore, constitute a transfer to the government from other institutions. Indirect taxes, on the other hand, are levied on the expenditure of the final purchaser of goods and services or, earlier in the production chain, on intermediate goods purchased by producers. In column 2 of table 1.3, the total cost of goods purchased at home (11,312) and abroad (1,455) is not what the buyer pays. Indirect taxes (or their opposite, subsidies) must be added in (or subtracted). The same is true for exports in column 5. The symmetry with column 4, where indirect taxes on intermediate goods or materials are clearly separated, should be evident. At the same time indirect taxes, which appear as an expenditure in columns 2, 4, and 5, are also a source of revenue to the government (one of the institutions of Sri Lanka) and so appear in column 2d on the appropriate row (2abc). Of course, indirect taxes are not an "institution" in any reasonable sense. But separating them out has sufficient advantage in understanding the structure to justify a separate subac- count. The logic of the matrix is in no way altered. The total still appears as part of the revenue of the government, while the parts are allocated to the relevant types of expenditure. This is an illustration of the flexibility of a SAM. (And here it can be noted that another way of dealing with indirect taxes, which may be preferable in some contexts, is to insert all of them into the production process, column 4.) Savings and Investment Before proceeding to a subdivision of the various accounts, we will make one more change. In column 2 (or 2abc) expenditure so far has included all kinds of goods and services, whether for consumption or investment. The change introduced in table 1.5, in effect, separates consump- tion from investment goods. Column 2abc now becomes a true current account for institutions and has been so labeled. The initial change is a reallocation between columns 2abc and 3. We extract the investment goods in rows 4 and 5 from column 2abc and enter them in column 3, renamed "combined capital account." For example, in row 4, the entries in columns 2abc and 2d-9,350 (consump- tion goods) and 1,982 (investment goods)-correspond to the total 11,312 in row 4 of table 1.4. The related indirect taxes similarly shift on row 2d. There is also a small shift on row 2abc Table 1.5. A Final Aggregate SAM for Sri Lanka, 1970 (in millions of rupees) a Expenditures |Total| | 1 |2abc 2d | 3 | 4 | 5 | 6 j | | 1 Factors of Production _ 11,473 _ 11,473 | I I I . I I R 2abc Institutions: Current 11,360 2,009 1,368 43 | 3 14,783_| | c 2d Indirect taxes | 119 | | 270 885 94 1,368 | i 3 Combined Capital | 2,214 1 1 425 2,639 t s_ | 5 4 __ Production _ Current _ Rest of the World _ ___ I 113 __ 9,350 1,091 t __ _ 1,962 _ 364 _ I 4,660 1,067 2,113 18,085 | 2,635 I | 6 Total | 11,473 | 14,783 | 1,368 2,639 | 18,085 2,635 1._ _ _ '__ _ _ _ _ _ _ I _ _ _ I _ _ _ I. _ _ _ I _ _ _ I _ _ _ I _ JI_ What Is a SAM? 27 for institutional reasons which are specific to Sri Lanka and do not merit detailed explanation 1 here. The total of all these changes in column 3 is equal to expenditure on investment-namely 2,639. Expenditure in column 2abc is now reduced by this amount. The balance of -425, which formerly appeared on line 3, is altered accordingly by the amount spent on investment: - 425 + 2,639 = 2,214. This is now the difference between income and current expenditure or consumption rather than income and total expenditure. It constitutes the savings of domestic institutions. Total savings derive from domestic sources (2,214) and foreign sources (425). These together finance investments (account number 3). In double entry form, the combined capital account is: Revenue (row) Expenditure (column) Domestic savings 2,214 Domestic investment goods 1,962 Foreign savings 425 Foreign investment goods 364 Indirect taxes 270 Other payments 43 Total 2,639 Total 2,639 A General Framework In table 1.6 we have divided two accounts (1 and 2) into subaccounts. The account for factors of production (1) is divided according to the two main factors: labor (la) and capital (lb). In account number 2, the three main sets of institutions have been distinguished: households (2a), corporations (2b), and government (2c). Indirect taxes (2d) had already been separated in table 1.4. There is little to say about the subaccounts for factors of production, which show income to labor and capital separately. Income to capital is more than half the total, but, as will be seen from column lb, the greater part of it accrues directly to households. This consists largely of income from small enterprises including farms, plus a substantial element of imputed income from housing. The entries in the "box" bounded by rows and columns 2a, 2b, and 2c are in the aggregate equal to the diagonal entry (2, 2) in table 1.5. Payments among institutional sectors all fall into the category of transfer, since they do not constitute income directly received from produc- tion. Payments arise from a variety of causes: ownership of certain assets (such as debt or equity investment in firms), direct taxation or government subsidies to households (not related to goods or services), or even voluntary transfers. Payments to the government include, of course, direct taxes by both households and corporations, but they also include social security contributions, pension fund contributions (actual or imputed), and dividends and the like from corporations (mainly public ones in Sri Lanka). Payments by the government include pensions, interest on the public debt, and a substantial amount of direct transfers to public corporations. The large payments from corporations to households (2a, 2b) are mainly payments to owners of capital (debt or equity), although they include some private corporate pensions. It will be noted in this table that, on line 4, there is no entry for consumption in column 2b (corporations). Following standard national accounting practice, only households and the government consume. Goods and services used by firms as inputs in the course of production are included in the final value of output; they are only consumed when the final output is consumed. Firms may invest, but their investments are included in column 3. Some observations are in order about some of the other accounts. We can see why account 3 1. They have to do with the system of Foreign Exchange Entitlement Certificates (FEECs), a system granting or charging a premium for foreign exchange, here applled to debt payments. Table 1.6. A Partially Disaggregated SAM for Sri Lanka, 1970 (in milions of rupees) _ _ _ _ - _ _ _ _ - _ _ _ _ - _____ Expenditures I la | lb 2a 2b 2c 2d 3 4 5{ 6 l| | la to Labor Income I I { ]|5569 5569 lb Income to Capital _ _ _ _ _ _ | 5904 | |5904 | e e P | 2a | 2b | 2c Household Current Corporations Current Government Current I _ | 5569 | _ 1575 _4216 | 447 | 644 376 | | _248 294 | _ _ 1368 ij _ 43 r | 18 15 | 1854 2234 _10695_| s 1 2d Indirect taxes I _ 119 | 01 885 94 1368_| 3 Combined Capital _ ._ |_ 1337 834 43 425 2639 | 4 | Production l 1 7701 | 1649 1962 4660 2113 18085.| |_ |_5 Rest of the World 113 1091 364 1067 26351 6 6 Total Total 5569 5904 10695 1---- 1854 2234 |5569 |5904 | 0695 1 185412234 1368 I 411067 8 5 26111 1368(| 2 6 3 9 jl 8 0 J 2635| What Is a SAM? 29 is now labeled "combined capital account." It combines the capital accounts of households, corporations, government, and the rest of the world. In principle, there is no reason why each institution should not have a capital account of its own. In practice, this was not possible in this case because of data limitations. Later, we wil show an example of a disaggregated capital account in another SAM. The production account (4) has not yet been subdivided. The blank lines suggest, correctly, that it will follow in the next table. The principal reason for not dividing it at this stage is that we do not wish to make too many changes at once. There is also another reason. The 8JAMin its present form gives all the information necessary to compile the "Consolidated Accounts for the Nation" as outlined under the UN SNA system (United Nations Statistical Office, 1968, p. 29). The accounts in this form are set out in table 1.7. They do not contain any information that is not already included in table 1.6. The latter is therefore a more compact form in which to present these consolidated accounts. It also, perhaps, is more useful insofar as the intercon- nections among accounts are more obvious in table 1.6 than they are in the double entry format. Each of the consolidated accounts can be identified with either a single column or row in table 1.6, or a combination of both. The two accounts which correspond to a single columnlrow can easily be compared. The capital account (c) corresponds to column/row 3; the rest of the world account (d) corresponds to column/row 5. Outgoings in the consolidated accounts corre- spond to columns and the incomings to rows. The other two accounts are less easy to identity, since they correspond to combinations. The reader who is not interested in the details of the identification is advised to skip the next five paragraphs. The "domestic product and expenditure" account (a) corresponds to SAM accounts 2d and 4 combined, with two adjustments. In order to show the correspondence, we have reproduced in table 1.8 rows/columns 2d and 4 of table 1.6 and added the implications of making the adjust- ments, as described below. First, the elements common to both row combination and column combination have been omitted, these are the figures 885 in row 2d and 4,660 in row 4 of column 4 in table 1.6. They can be left out because, when accounts 2d and 4 are combined, as they need be to form the domestic product and expenditure account, all transactions between accounts 2d and 4 become both a receipt and an expenditure of the combined account. They therefore do not affect the balance of receipts and expenditures for the domestic product and expenditure account. Second, we have added an extra row and column, 7, which does not appear as such in table 1.6. The row corresponds to direct imports for final consumption and for investment. (The figures 1,091 and 364 appear in row 5 in table 1.6.) The column has the total 1,455 required to balance the account. Row 8 in table 1.8 is the sum of the combined entries in rows 2d, 4, and 7. This row corresponds to the incomings, or the left side of consolidated account (a), except for the imports, which have in effect been transferred, changing the sign, from the opposite side. Column 8 includes these imports and the outgoings, or the right side of account (a). The "national disposable income and outlay" account (b) corresponds to SAM accounts 2a, 2b, and 2c. Table 1.6 has been reproduced with only these rows and columns in table 1.9, but the diagonal transfer elements between them have been eliminated. Here we represent the totals by two rows and columns, 8a and 8b, instead of one, because, in several cases, the simple addition of the elements in a particular combination does not correspond to an entry in the consolidated account (c). For example, income to capital in column lb appears not as the sum of 4,216 and 1,575 (5,791) but as the differencebetween total capital income (5,904) andcapital income to the rest of the world (113); that the two are the same is obvious from column lb in table 1.6. In column 8b, the three elements of private consumption marked by an asterisk appear combined in consolidated account (b) of table 1.7: indirect taxes (119), domestic produc- tion (7,701), and imports (1,091), making a total of 8,911. Once one has performed these Table 1.7. Consolidated National Accounts for Sri Lanka, 1970 (in millions of rupees) A. Domestic Product and Expenditure Incomings Outgoings Government consumption 1,649 Income to labor 5,569 Private consumption 8,911 Income to capital 5,904 Investment 2,596 Indirect taxes 1,368 Exports 2,207 Less Imports -2,522 12,841 12,841 B. National Disposable Income and Outlay Incomings Outgoings Income to labor 5,569 Government consumption 1,649 Income to capital 5,904 Private consumption 8,911 Indirect taxes 1,368 Savings 2,214 Current transfers from abroad 3 Transfers from capital a/c 43 Less property income transferred abroad -113 12,774 12,774 C. Capital Account Incomings Outgoings Savings 2,214 Capital formation 2,596 Less transfers from capital a/c -43 Foreign borrowing 425 2,596 2,596 D. Rest of the World Account Incomings Outgoings Imports 2,522 Exports 2,207 Property income Transfers 3 transferred abroad 113 ______ Surplus (of ROW) 425 2,635 2,635 Note: ROW signifies rest of world 30 Table 1.8. Reconciliation Table for Sri Lanka, 1970 (in millions of rupees) la lb 2a 2b 2c 2d I 3 4 5 7 81 I- j I I I I I I 5569 I I 55691 la Incomes of Labor _ _ _ lb Income to Capital I I 1 I 5904 1 59041 |2a Household Current 1 I _ _ _ I _ I I _ _ _ |2b Corporations Current 1 |l _ 2c Government Current 1 _ L1 j 1 1368 1 _ _ _ _ 13681 |2d Taxes Indirect 1191 1 I T2701 | 94 | I I I I I I I1I1 3 CombinedCapital = _ 4 Production 7701 1649 1962 2113 _ 5 Rest of the World 1067 1455 2522 7 Final Imports 1091 | 364 8 Subtotal 1 8911 1649 2596 2207 15363| Note: Column 8 is the total of columns 2d, 4. and 7. Row 8 is the total of rows 2d, 4, and 7. Table 1.9. Revised Reconciliation Table for Sri Lanka, 1970 (in millions of rupees) r | la |ilb 2a | 2b 2c| 2d 3 4 5 8a 8b la Incomes to Labor lb Income to Capital 2a 1 ousehold Household Current |5569 I 4216 | I 1I _ _ I 1I _ 1 18 | I- I I 2b Corporations Current | |1575 1 -15 | 2c Government Current _ I I 1 ) 136 8 | 43 | 2d Indirect taxes 1 119 . 119* 3 CombinedCapital I 1337 834 43 I 2214 4 Production I -I ---'1 [7701 1 111649 11 1 f1 1 1 6 |1649 7 70 15 Rest of the World I 1091 | I I i |1091* 18a 15569 159041 I I 113681 43 3 |8b l l | -113 12774) Note: Columns 8a and 8b together equal the total of columns 2a, 2b, and 2c. Rows 8a and 8b together equal the total of rows 2a, 2b, and 2c. *The total of these three equals 8,911 (private consumption). What Is a SAM? 33 arithmetic tricks, the elements in the row again correspond to the consolidated account incom- ings and the column to outgoings. An Input-Output Matrix The final step in this exposition of the Sri Lanka table is to subdivide the production account (4) into six sectors; these are listed in table 1.10 in rows 4a to 4f. The "box" formed by these rows and their equivalent columns constitutes the core of what is commonly known as an input-output matrix. The accounts of each sector follow essentially the same logic as before. For example, if we take the agricultural sector (4a), the column in total is equal to gross output at producers' prices (which exclude sales taxes and the like). This total of 5,903 includes not only value added (in rows la and lb), taxes on intermediates (in row 2d), and imported materials (in row 5), but also inputs from other sectors. The largest volume of inputs is on the diagonal (4a, 4a) and consists of internal purchases and sales within the sector; in fact, it is largely paddy (one subsector) sold to ricemills (another subsector). Just as column 4a shows the sources of gross output to which payments are made, row 4a shows the destinations of gross output from which the production units derive their revenue: in column 2a the amounts going to consumption; in column 3 to investment; in columns 4a to 4f as intermediate goods to other sectors; and in column 5 to exports. Table 1.10, with eighty-five entries in it, is a long way from the pristine simplicity of the Robinson Crusoe economy. The circularity of the process of production-income-expenditure is no longer so evident. But this fundamental attribute of the SAM is still there, and it can be illustrated by reference to the input-output tables. Table 1.11 shows an input-output table with six sectors corresponding to those in table 1.7. Table 1.11 is the same as table 1.10, with two exceptions. The first, which is trivial in substance, is that it has been rearranged. Columns and rows 4a to 4f in table 1.10, which were in the lower right corner, occupy the top left corner in table 1.11; columns and rows 1, 2, and 3 are now below or to the right of them. The substantial difference is that the lower right corner of table 1.11 is blank. The input-output table captures only the relationships between the production accounts and the other accounts (factors of production, institutions, capital accounts, and rest of the world). Interrelationships among these accounts, most of which are in the top left corner of table 1.10, are not there. If we were to specify a new set of final demands different from those in table 1.11, techniques, based on specific assumptions about intersectoral relationships, exist for deriving the implied pattern of production in each subsector. That implied pattern, of course, will include income to factors of production (rows 1 and 2 in table 1.10). The results, however, give no guarantee that there is any relationship between incomes generated and the ensuing demand. The complete SAM, in principle, provides the missing link-or at least the data to establish it. Table 1.10 is as far as we shall go here in subdividing the accounts. In the original source, much greater detail is shown. In one complete matrix, the following appear: * Three labor groups (urban, rural, and estate) * Three capital groups (public, private, and housing) * Three household groups (urban, rural, and estate) * Two kinds of corporations (private and state) * Eleven production sectors. However, still more detail lies behind that matrix. For example, the production account is based on detailed accounts for forty-eight subsectors. Household data are based on information for six income brackets within each group. This completes the first part of this introduction to the exposition of social accounting matrices. For those who may, unexpectedly, have been titillated by the subject, we have added one more section on a SAM for a different country, Botswana. That example illustrates an attempt to Table 1.10. A Complete Disaggregated SAM for Sri Lanka, 1970 (in miUions of rupees) Expenditures Total la I- |lb [ 2a 1 2b 2 ic 2d 13 4a 4b 4c| 4d 4e [ 4f 5 6 | I R |lb la Income to Labor Income to Capital I 5 4 _ 2015 2009 561 245 909 564 1275 5569 _ 909 734 1423 829 = 5904 1 = c 12a Household Current 5569 4216 644 248 _ _ _ 18 10,695 eI I~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1i 12b Corporations Current 1575 j 294 _ I -15_ 1854| |2c Government Current: 447 | 376 431368 27 5 | 2234.| 2d Indirecttaxes _ 119 I 1 270 80 |504_ | 66 1301 76 | 29 94 1368| | 3 CombinedCapital |_ |_ 1337 | 834 43 | I | I |4 425 2639| | 4a Agriculture 2861 4 104 1191 354 3 1 62 40 1288| 5903 1 |4b Industry _ 1824 _ 109 254 815 417 _172 66 74 335 4066| I 14c Construction - 1595 -59 1 | 7 50 92 | 1745_ 114d Trade and Transport _ 1606 154 1 135 344 206 96 42 59 203 28451 1 I 1 I 4e PrivateServices 1410 | | 37 4| 9 38 55 371 287 | 1877| 4f Government Services _ 1649 1 | j| 1649| 5 Rest of the World 113 1091 | |364 | 182 574 65 70 133 43 42635 __6 Total | 5569 | 5904 | 10,695 | 1854 2234 1368 | 2639 1 59031 4066 | 1745 2845 | 1877 1649 2635 | Table 1.11. Input-Output Matrix for Sri Lanka, 1970 (in millions of rupees) Sectors Final Demand Trade & Priv. Govt. Sub- | Agr. Ind. Constr. Transp. Services Services Total Cons. Inv. Gov't. Exports Total |Agriculture 1,191 354 3 62 40 1,650 2,861 104 1,288 5,903 Industry | 254 815 417 172 66 74 1,798 1,824 109 335 4066 |Construction 1 7 | 501 92 150 1,595 1,745 Trade & Transportj 135 | 344 206 96 42 59 882 1,606 154 | | 203 2,18451 I_ 4 _ I _ ___ _ T I 1Priv. Services 37 9 38 55 37 180 1,410 287 1,877 |Govt. Services _ __O 0 1,649 | 1,649 Subtotal 1,617 | 1,518 i 635 313 275 302 | 4,660 7,601 1,962 1,649 2,113 j 18,085 Vaiue addedIII (labor) 2,015 561 j 245 909 564 1,275 5,569 Valueadded (capital) 2,009 909 | 7341 | 829 | 5,904 iIndirect taxes 80 504 66 130 76 | 29 885 |Imports 182 | 574 65 70 1 133 43 1,067 | 1 6 75 2 5 17 I 1 9 |Total j 5,903 | 4,066 1,745 |2,845 | 1,877 | 1,649 18,085 I _ _ _ _ I _ _ I _ _ I _ _ _ _ Table 1.12. An Aggregate SAMfor Botswana, 1974-75 (in millions of pulas) Expenditures Total 1 ~2 3 4 5 6 78 | RI 1 2 Factors of production InstitutionsCurrent Account | 220.71 29.2 80.6 1171.4 4.5 23.0 . 18.3 4.6 0.7 02 2.01 224.3 330.7.| c I3 ProductionAccount _ 1 78.01 82.8 117.9 68.4 | 2.7 349.8 |i 14 ROW Current Account | 3.8 87.2 91.4 | | 37.0 | 9.2 147.7j IPtII I I I I I I t 5 Combined Capital f I I I I I s Account 55.4 69.4 |-83 1165 6 ......... I I.... I 6 1 I I I I I | 7| Errors and Omissions -0.2 0.3 -0.3 j 6.5 | ____ 6.3 8 Total | 224.3 330.7 349.8 228.6 116.5 | 6.3 Note: ROW signifies rest of world. WhatIs a SAM? 37 bring into the framework of the accounts not only the various transactions already described but also the financial counterpart of these real transactions and the flow of capital funds into investment. In some ways this is one of the most intriguing uses of a SAM, because it brings together two related aspects of development, the real and the financial. The parsimonious use of financial data and the neglect of its relationship with the real economy have often been noted (see chapter 3). BOTSWANA 1974-75: THE FLOW OF FUNDS Our purpose at this point is to illustrate how to introduce financial transactions into a SAM. The example we use is taken from a report on a SAMconstructed for Botswana (United Kingdom, Ministry of Overseas Development, 1977). (All figures in this section are in millions of pulas.) The Botswana SAM, shown as table 1.12, is conceptually the same as table 1.5 for Sri Lanka, except for three differences: * There is no separate line for indirect taxes; for simplicity, they have been included with the central government, which is itself included under institutions (row/column 2). * The combined capital account has been moved down to row/column 5 and the two accounts previously on lines 4 and 5 have been promoted to 3 and 4, respectively. * New lines 6 and 7 have been added. One of them has been left blank for the moment. The other consists of errors and omissions. Totals are now shown in rows and columns 8. The reason for not having a separate line for indirect taxes is to avoid unnecessary clutter. Having made the point once, we do not need to repeat it. The change in order is not a change in substance, as there is no magic in any particular order. The best order is the one that follows a reasonable logic and makes the SAM intelligible to the reader; thus there can be differences. The order here conforms rather closely to the original, more detailed version of the SAM. The account for errors and omissions (7) is present for two quite different reasons. The first set of errors is attributable entirely to rounding. Many individual figures in the original table have been added together to form subaggregates; this process inevitably involves rounding errors. There are, however, several large specific errors, which appear as such in the original SAM, mainly in rows/columns 4 and 5. These errors are akin to residual errors often left in the balance of payments or national accounts. An SAMs have such errors at some stage of their construction. How these particular errors arose and why they were left in are explained in detail in chapter 7. Purchase and Sale of Assets The main change in table 1.13 from table 1.12 is that row/column 6 has been labeled "financial account" with one entry each in the column and the row. In addition, there is a new diagonal element in row/column 5. These are basically the only differences. An previous tables, whether on Sri Lanka or Botswana, have dealt with the consequences of current activity during the year: the production-income-expenditure cycle. Savings are savings out of current income, and the investment they finance encompasses only new investment. However, this does not exhaust the totality of transactions. Institutions may buy or sell existing physical assets, particularly land and buildings. They also lend or borrow, thereby creating financial assets or liabilities. The new entries are intended to acknowledge these facts. The diagonal element of 30.3 in row/column 5 expresses the fact that institutions bought existing assets of that amount and also that they sold them. Obviously, the two must balance. When looked at as an entry in the column, the transaction appears as a purchase; when looked at as an entry in the row, it appears as a sale. XTable1. 13. Introduction of a Financial Account into the Aggregate SAM for Botswana, 1974-75 (in millions of pulas) Expenditures Total 1 2 3 4 5 6 7 8 1 Factors of production 29.2 171.4 23.0 0.7 224.3 2 Institutions Current Account 220.7 80.6 4.5 18.3 4.6 2.0 330.7 R e c 3 Production Account 78.0 82.8 117.9 68.4 2.7 349.8, e i p 4 ROW Current Account 3.8 87.2 91.4 37.0 9.2 228.6 5 Combined Capital Account 55.4 69.4 30.3 127.3 -8.9 273.5 6 Financial Account 126.6 0.6 127.2 7 Errors and Omissions -0.2 0.3 -0.3 6.6 -0.1 6.3 8 Total 224.3 330.7 349.8 228.6 273.5 127.2 6.3 Note: ROWsignifies rest of world. What Is a SAM? 39 The new account (6) reflects all financial activities on capital account, such as borrowing and lending. Most, but not all, of these are carried out by banks and financial enterprises. Current activities and new physical investments of these enterprises are already included as part of the accounts of institutions. The two new entries, at the intersection of row 5 and column 6 (127.3) and the intersection of row 6 and column 5 (126.6), are identical except for the errors and omissions. In principle they must be. They express the fact that institutions incur financial liabilities to the financial sector (for example, by borrowing) and also acquire financial assets from that sector (such as currency or bank deposits). These two must balance, because a liability automatically creates a corresponding asset. They must balance in the aggregate, but, as we shall see, they need not balance for any individual subset of institutions. We can now set out the capital account implied by the entries in row/column 5 in the familiar double entry form as follows: Incomings (row) Outgoings (column) Domestic savings 55.4 New investments (rows 2-4) 110.0 Foreign savings 69.4 Purchase of existing assets 30.3 Sale of existing assets 30.3 Acquisition of financial assets 126.6 Financial liabilities incurred 127.3 Errors and omissions 6.6 Errors and omissions -8.9 Total 273.5 Total 273.5 Decomposition of the Capital Account In table 1.14 we have divided row/column 5 into three parts. The first element is labeled 5abc and includes the group of institutions identified in Sri Lanka, namely, households, enterprises (or corporations), and government; the fact that it has an "abc" at the end implies that it will be broken down still further at a later stage. The second element, labeled 5d, consists of a new category of financial enterprises, such as banks, which loom quite small in the production processes we have considered so far, but loom much larger in the financial transactions that we are considering now. They do, in fact, represent the capital and money markets through which most of the financial transactions take place. The final element in row/column 5e is the capital account of the rest of the world (ROW). The sales of existing physical assets are now identified (5abc, 5e) mainly as sales by insti- tutions in Botswana to the rest of the world and, to a smaller extent, vice versa. The latter transactions are explained by the interest of foreign corporations in mining enterprises in Botswana. We see that individual accounts do not have to balance in their financial transactions. Those who save may put their savings into real or financial assets. Investors in real assets may borrow in order to finance them. This is the purpose of a capital market. For example, we can set out the capital account of domestic institutions other than financial enterprises to illustrate this point: Incomings (row) Outgoings (column) Savings 53.9 New investment (rows 2-4) 108.6 Sale of physical assets 23.8 Purchase of existing assets 6.8 Financial liabilities incurred 90.7 Acquisition of flnancial assets 53.7 Errors and omissions 0.5 Errors and omissions -0.2 Total 168.9 Total 168.9 A similar account could be made for financial enterprises (5d). It would include rather modest amounts for savings and new investment (1.5 and 1.4, respectively). The principal elements in this account would be the entries in row/column 6, that is, the sale or purchase of financial assets. In practice, these are bound to balance, except for the minor difference between savings Table 1.14. A SAM with Disaggregated Capital Accounts for Botswana, 1974-75 (in millions of pulas) Expenditures Current Accounts Capital Accounts Total 1 2 3 4 5abc Sd 5e 6 7 8 I Factors 29.2 171.4 23.0 0.7 224.3 o 2 Institutions 220.7 80.6 4.5 18.3 4.6 2.0 330.7 uc: e_ 3 Production 78.0 82.8 117.9 67.5 0.9 2.7 349.8 R X e 4 ROW 3.8 87.2 91.4 36.5 0.5 9.2 228.6 c e i 5abc Institutions 53.9 0.3 23.5 90.7 0.5 168.9 t s Financial 5d enterprises 1.5 31.8 -9.6 23.7 0 o 5e ROW 69.4 6.5 4.8 0.2 80.9 X4 Financial . 6 Transactions 53.7 15.4 57.5 0.6 127.2 Errors and 7 Omissions -0.2 0.3 -0.3 -0.2 6.9 -0.1 -0.1 6.3 8 Total 224.3 330.7 349.8 228.6 168.9 23.7 80.9 127.2 6.3 Note: ROW signifies rest of world. What Is a SAM? 41 and new investment, but, in the table, balancing is achieved through the errors and omissions row/column. This arises, for example, because occasionally in the year-end accounts of a borrower and a lender, a liability and its corresponding asset may be valued differently (see chapter 7). Decomposition of the Financial Transactions The account for financial transactions (6) in table 1.14 has been broken down further in table 1.15 into four different categories: domestic currency, 2 bank deposits, and the like; domestic borrowing or lending; and foreign borrowing or lending. These are identified by rows/columns 6a through 6d. As might be expected, institutions acquire additional resources (row 5abc) by incurring liabilities through domestic or foreign borrowing (columns 6c and 6d). To the extent that they do not spend these resources on physical assets (old or new), they retain them for the most part in the form of financial assets: either currency or deposits (rows 6a and 6b in column 5abc). The account of financial enterprises (5d) in table 1.14 is of particular interest here. In presenting it in the usual form of outgoings and incomings, we make a minor modification by changing titles to "change in assets" and "change in liabilities," respectively. We could do the same for other capital accounts, but the change in this case brings out more clearly the nature of outgoings (acquisition of assets) and of incomings (incurring of liabilities). The account thus reads: Change in liabilities (row) Change in assets (column) Physical assets 1.4 Savings 1.5 Currency 0.4 Deposits received 22.9 Deposits made 0.5 Domestic borrowing 3.2 Domestic lending 14.5 Foreign borrowing 5.7 Errors and omissions 6.9 Errors and omissions -9.6 Total 23.7 Total 23.7 Two items appear on both sides of the account: deposits and domestic lending or borrowing. There are at least two reasons for this. First, financial enterprises cover more than commercial banks. Some of them make deposits in commercial banks. These deposits appear as a liability to the banks, but an asset to the depositing enterprises. Second, financial enterprises may both borrow on the market and lend to their customers. They therefore increase their assets by lending and their liabilities by borrowing. The account is, in fact, analogous to the change in an enterprise's balance sheet from one year to the next, except that it applies to a set of enterprises. Savings here correspond to the increase in equity investment attributable to retained earnings. Decomposition of Institutions In table 1.16, the final one for Botswana, institutions have been broken down into the same three constituent parts as for Sri Lanka: households, enterprises (or corporations), and central government. This has been done both for the current account (2) and the capital account (5abc). Except for the greater amount of institutional detail, there is no change in principle from table 1.15. However, these new data now show how each set of institutions contributes to the flow of capital funds through the system. This information shows, for each of the three sets of 2. Actually, at the time there was no independent Botswana currency; the currency then circulattng was the South African rand. Table 1.15. Extension of the SAMfor Botswana to Include Separate Accounts for Financial Assets, 1974-75 (in millions of pulas) Expenditures Current Accounts Capital Accounts Financial Accounts Errors Total 1 2 3 4 5abc 5d 5e 6a 6bX 6c 6d 7 8 to 1 Factors of Production 29.2 171.4 23.0 0.7 224.3 0 2 Institutions 220.7 80.6 4.5 18.3 4.6 2.0 330.7 c- 3 Production 78.0 82.8 117.9 67.5 0.9 2.7 349.8 R 4 Rest of World 3.8 87.2 91.4 36.5 0.5 9.2 228.6 o c Sabc Institutions 53.9 0.3 23.5 38.8 51.9 .5 168.9 0 C u e 5d Banks & financial ie enterprises 1.5 22.9 3.2 5.7 -9.6 23.7 co 5e Rest of World 69.4 6.5 5.4 -0.6 .2 80.9 t ' s a 6a Domestic Currency 5.0 0.4 5.4 S0 u 6b Bank deposits 22.2 0.5 .2 22.9 .-4 6c Bank advances & other c domestic borrowing 27.2 14.5 .3 42.0 6d Foreign borrowing -0.7 57.5 .1 56.9 7 Errors -.2 .3 -.3 -.2 6.9 -.1 -.1 6.3 8 Total 224.3 330.7 349.8 228.6 168.9 23.7 80.9 5.4 22.9 42.0 56.9 6.3 Table 1.16. A Final SAM for Botswana, 1974-75 (in millions of pulas) Expenditures __= Current Accounts Capital Accounts Financial Accounts Errors Total I 2a 2b 2c 3 4 5a 5b 5c 5d 5e 6a 6b 6c 6d 7 8 Factors of Production 7.3 21.9 171.4 23.0 0.7 224.3 la 2a Inst. Households 181.5 11.2 9.3 .7 202.7 a 37.4 5.2 8.1 6.5 .4 5.4 .4 63.4 Y 2b Enterprises 1.8 15.7 24.6 4.1 12.9 .2 3.9 0.5 .9 64.6 c 2c Cent. Govt. 67.2 .6 10.2 82.8 117.9 14.1 32.5 20.9 0.9 2.7 349.8 a 3 Production R u 3.8 0.5 9.2 228.6 e 4 Rest of World 3.8 79.9 .3 7.0 91.4 1.4 31.3 23.1 0.1 1.7 .2 25.1 c 5a Households 5b Enterprises 11.7 35.3 41.2 1.7 89.9 i 19.1 0.1 0.1 23.5 1.8 10.7 -1.4 53.9 p Sc Cent. Govt. t - 1.5 22.9 3.2 5.7 -9.6 23.7 s Sd Banks & fin. ent. 69.4 6.5 5.4 -0.6 .2 80.9 5c Rest of World a 6a Domestic Currency 4.5 0.5 0.4 5.4 6b Banks 6.3 3.0 12.9 0.5 .2 22.9 I* other deposits _ & other 0.4 6.2 20.6 14.5 .3 42. Bank advances - i a 6c domestic borrowing borrowing -1.7 5.7 -4.7 57.5 .1 56.9 _ 6d Foreign Errors -. 2 .4 -. 1 -. 3 -. 2 .1 -. 1 6.9 -. 1 -. 1 6.3 7 224.3 202.7 63.4 64.6 349.8 228.6 25.1 89.9 53.9 23.7 80.9 5.4 22.9 42.0 56.9 6.3 8 Total 44 The Methodology of Social Accounting Table 1.17. Changes in Assets and Liabilities of Households, Enterprises, and Government in Botswana, 1974-75 (in millions of pulas) Households Enterprises Government Total Change in Assets (column) Physical assets (new) 15.7 67.7 25.2 108.6 Physical assets (existing) a/ - 6.7 -23.7 -17.0 Financial assets 9.5 15.4 28.8 53.7 Errors and Omissions -0.2 0.1 -0.1 -0.2 Total 25.0 89.9 30.2 145.1 Change in Liabilities (row) Savings 23.1 11.7 19.1 53.9 Financial liabilities 1.7 76.5 12.5 90.7 Errors and Omissions 0.2 1.7 -1.4 0.5 Total 25.0 89.9 30.2 145.1 a. Sales of existing assets have been brought over from the "liability" side as a negative item. This is a net figure. institutions and for the total, the changes in assets and liabilities. In an alternative format, it can be presented as shown in table 1.17. Households, as is often the case, saved more than they invested in physical assets and, conse- quently, put the difference into financial assets. Enterprises were the principal investors in physical assets and, since their savings (or equity participation) were small relatively, most of their investments had to be financed by borrowing. In this particular year (1974-75), the role of the government was unusual; it financed most of its new investment by the sale of existing assets. THE USES OF A SAM The effort required to put together a SAM is not trivial. Data must be ferreted out, wherever they may be available. Conflicting sources must somehow be reconciled. Rows and columns do not conveniently come to the same total in the first instance.3 What does one get out of it all except a rather complicated and impressively tidy collection of numbers? Because social accounting matrices have not been in existence for long and there are not many of them, to say what they are useful for is partly an exercise in conjecture. Nonetheless, there seems to be sufficient foundation to make a few plausible suggestions. In the first place, a SAM is clearly a step forward in the upgrading of statistics. Recent comparisons of micro- economic information obtained from household surveys with national accounts have shown that the discrepancies between these two sources of information can be very large. How do we 3. Techniques exist for making adjustments to achieve balance at a minimum cost in terms of variation from the original; and new or improved techniques are being worked on. What Is a SAM? 45 choose between them? Or, should we choose between them? While construction of a SAM is certainly not going to reveal the ultimate truth, at least it forces attention on inconsistencies in a way that brings one closer to the root of their cause. Judgment, to be sure, has to be used in imposing ultimate consistency, but it can be done in such a way as to keep adjustments within plausible limits and so avoid a purely Procrustean process of fitting one set of data to the dictates of another. The concept of a SAM goes further than the improvement of statistics for their own sake. It could be said to be the common ground of economic planners or development economists, on the one hand, and statisticians, on the other. A SAM is cast in a form that, given the fineness of detail with which it is constructed, makes the most of existing information. Economic models of an economy, which maybe designed for particular purposes, nevertheless imply the existence of an underlying SAM. Parts of this implied SAM may be aggregated and parts highly disag- gregated, but it is, nevertheless, a SAM. The existence of an actual SAM, against which to test the behavioral assumptions of a model and the SAM they imply, is, on the face of it, a useful way of testing the model's validity. Much has been and could be said about the relationship between models and SAMs. Examples from the growing literature are de Melo, 1979, and Dervis and Robinson, 1978. Here we shall only touch on some of the simpler applications of a SAM to the understanding of the way in which an economy works. The uses of a SAM fall into two categories: those in which the whole corpus of information in the SAMis used and those in which only a part is used. Of course, in the latter case, it is not necessary to have the complete SAM. But the construction of the complete articulated SAM means that one has at one's disposal a multipurpose tool and does not have to construct separate subsets of accounts for each purpose. An illustration of the use of part of a SAM has been documented in the case of the Sri Lanka SAM that we have described in aggregate terms. The purpose of the exercise in question was to establish the order of magnitude of the total fiscal incentives for exporting in various sectors (see Pyatt, Roe, and associates, 1977, ch. 6). At the time (in 1970) substantial direct fiscal incentives were given to encourage nontraditional exports. At the same time, many industrial subsectors in Sri Lanka, among which, it might have been expected, would be found some potential exporters, received fairly high nominal protection. The input-output matrix within the Sri Lanka SAM was used to convert nominal protection rates into effective protection rates to these industries; in most cases, these were substantially higher than the nominal protection rates. The incentives in the tariff system that implicitly encouraged production for the domestic market could then be compared with the export incentives. In many cases, they greatly outweighed the export incentives and, in other cases, reduced them to fairly small proportions. These findings, which were unexpected, could have been reached without a SAM; its existence, however, made the task easier. One of the principal ways in which the whole corpus of information in a SAM can be brought to bear is through multiplier analysis, which shows how changes in one or more elements of a SAM generate changes elsewhere in the matrix. Here we will only consider a simple example to illustrate the approach. The starting point is to assume a simple economy and a highly aggregated SAM which has accounts only for the private sector, government production, the rest of the world, and a combined capital account. To keep things even simpler, it is assumed that only the private sector buys goods from the rest of the world. Corresponding to this simple economy, we assume we have a SAM for some base period, and we ask the question, "What would happen if demands on production activities were increased by increasing government expenditure (by an amount i2 ), investment (by an amount i3), and exports (by an amount i5)?" Without loss of generality we can put the sum of the i's equal to one. The first part of the answer to this question is that whatever the processes of consequential changes might be, the end result wlfl be a new SAM for our simple economy. Moreover, those 4> Table 1.18. Multiplier Effects in the Form of an Incremental SAM Expenditures Total 1 ~~~2 3 4 5 6 1 Private Sector _ MM R 2 Government MP2 _4P2 e 3 Capital Account MP3 MP2 -i2 MP5 i5 i3 p t 4 Production MP4 i2 i3 i5 M 5 Rest of World MP5 Mp5 6 Total M MP2 i3 . Note: M(I- p4) =p 1 = Xi pe = marginal propensity to tax p3 = marginal propensity to save p4 = marginal propensity to consume (domestic) p3 = marginal propensity to import M= multiplier ie = impulse from increased government expenditure i 3 = impulse from increased investment i, = impulse from increased exports WhatIs a SAM? 47 elements of the initial SAM that are zero by definition will remain zero. Because our model assumes that only the private sector buys goods from abroad, the purchases of such goods by the government, for example, will remain zero. Given the accounting rules and model assumptions, the difference between the new SAM and the original one will imply an incremental SAM in which many cells have zero entries. This incremental SAM is shown in table 1.18. At this stage we know the items i2, i3, and iS,because these are the changes that we have exogenously postulated. We also know that the blank entries in the table are zeros, because these follow from our model and accounting conventions. The question then is, "What can be said about the nonzero entries apart from i 2 , i3, and is?" Not much can be said about these nonzero entries without making further assumptions about what will happen, for example, to prices, monetary policy, and how people choose to spend any extra income. We will assume, for simplicity, that private sector income goes up by an amount M, and then explore what the incremental SAM in table 1.18 can say about the relationships between M and the i's. Because in this simple model the private sector gets all its income from production activities and because these activities pay all value added to the private sector, row 1 and column 4 of the incremental SAM are very simple and contain zeros apart from the entry in row 1, column 4, which is M. The increase in private income (row 1) must match the increase in private expenditures in column 1. The latter must now be spread over the different components of private expenditures. This spread is assumed to take place in the proportions P2 , P3 , p4 , and PS,which can be referred to as the marginal expenditure propensities of the private sector. Because all the extra income M must be spent or saved, the accounting balance for row/column 1 implies that p2 + p3 + P4 + p5 = 1. We might also be prepared to assume that these propensities are constant. But if we do, then this is clearly a behavioral assumption, not an accounting rule. Because MP2 is the only increase in income for the government, row/column 2 of table 1.18 must have sums Mp2 . From column 2, this implies that the entry in row 3, column 2-the increase in government savings-must be Mp. - ip. For now, we skip over the details of accounts 3 and 4 and move to the rest of the world account (5). Here, the only increase in receipts is Mp5 because only the private sector imports in this model. This then implies that the entry in row 3, column 5, must be Mp 5 - is. This entry measures the extent to which foreign savings, or a reduction in domestic reserves of foreign exchange, finances the increased investment, i3. Returning now to account 4, the fact that row and column sums must be equal implies that M = Mp4 + i2 + i3 + i5 = Mp4 + 1, or M = 1l(1 - p 4 )- In other words, the value-added M must be equal to the aggregate increases in government expenditure, investment, and exports, inflated by the factor 1/(1 - p 4 ). This factor is the familiar expenditure multiplier; it is the reciprocal of the complement of the marginal propensity to consume domestic goods. Hence, while the SAM does not tell us what value to give to M or to p4 , it does show that once one value is fixed, the other is also fixed; and in that sense it defines the relationship between the initial increments in expenditure (the i's) and the increase in total value added (the M). At this stage we have discussed the balancing of four of the five accounts of the incremental SAM. That is all that is necessary, because it is always true that within a SAM the last account 48 The Methodology of Social Accounting will balance if all the others balance. To illustrate this point, the rule requires that, for our account 3, i3 = Mp 3 + (MP 2 - i2) + (MP 6 - , or i2 + i3 + i5 = (P 2 + P3 + PS) M- Since the sum of the i's and the sum of the p's are each equal to one, this can be written as: 1 = (1 - p 4 ) M. Hence the condition for account 3 to balance is the same as that for account 4, that is to say: M = 1/(1 - p 4 ). This result simply repeats that obtained previously. If all but one account are balanced, then all accounts are balanced, and the story of the incremental SAM shown as table 1.18 is thus completed. The application of multiplier analysis with a complete SAM is little different in principle, though it is far more complex. It takes into account all the interactions within each step of the process of linkages among incomes, expenditures, and production. The linkages could include, for example, the effects on other industries of expansion within a particular industry. There is, however, no longer a single multiplier, but an entire matrix of multipliers, which potentially shows the effect of expansion in one cell of the original SAM on any other cell. How these effects are to be interpreted must always be approached with care because the effect of one variable on another ultimately depends on economic behavior and not just on accounting constraints. However, the approach has some value in distinguishing accounts or subaccounts that are likely to be affected from those that are likely to be bypassed. This distinction may well have importance in considering the effect of exogenous changes on the distribution of income. The analysis may also serve to identify the important elements that result in changes in government accounts or in the balance of payments. Several different applications of this type of analysis have been made using the Botswana SAM. In such applications the SAM relationships can trace the complex interactions inherent in the circular process. If initial changes in prices or wages are involved, the analysis can show-at least in orders of magnitude-how the initial changes affect the prices in different industrial sectors and the consumption patterns of different household groups; if interindustry relations are complex and if, as is more than likely, household consumption patterns are very different, the resulting pattern may be difficult to predict. Such analysis, however, is based in the first instance on the assumption that patterns of production and consumption are unaffected by price changes. Adaptation to take into account assumed responses can, however, be intro- duced. This adaptation, in fact, is essential in modeling an economy. In a small open economy or a region, interindustry relations tend to be weak and leakages in the multiplier process large. SAMs have been constructed in each of these contexts: for example, for Swaziland and for the Muda Valley in Malaysia. (See chapter 6 for a discussion of the Swaziland SAM and Bell, Hazell, and Slade, 1982, for the Muda Valley SAM.) The construction of accounts for a region, as opposed to a country, is likely to reveal features of the regional economy that were little appreciated before. This is obviously so because regions do not possess "national" accounts and other data normally associated with an economy as a whole. The construction of the SAM for the Muda Valley is a good illustration. In this region a large irrigation project had more or less doubled the output of rice, the main crop. Several "downstream" effects of the resultant increase in farmers' incomes are of interest. Perhaps the main one is the very large outflow of capital from the region to the rest of Malaysia. This fact What Is a SAM? 49 and other data in the SAM are consistent with the theory that the principal downstream effect was to increase the incomes of nonfarm households, such as traders, who were in effect "import- ers" from the rest of Malaysia. Leakages from the regional economy were thus substantial. It is perhaps significant that there was still a substantial number of poor landless laborers. Although the SAM may have been constructed too soon after the completion of the irrigation system to allow for opportunities for reinvestment in the region, it is, nevertheless, a clear reminder that downstream effects can simply not be taken for granted. Clearly, no SAM can ever be constructed to answer questions except in the broadest sense. Specialists in any particular subject may have a much better idea of specific consequences, based on their accumulation of intimate knowledge, than a SAM alone could provide. But no one since Thomas Jefferson and his contemporaries can be a specialist in everything. A SAM can be used to bring out what is likely to be important in any given context and, therefore, to order the consultation of specialist knowledge to the occasion. A highly disaggregated SAM, such as that reported in Pyatt, Roe, and associates (1977) for Sri Lanka, would, of course, be physically difficult to reproduce on one sheet of paper. Even if it were possible to do so, the result would be comprehensible only to a very limited group. The great advantage of a SAM is that one can select for any occasion those parts of it to be aggregated and those parts where detail is to be preserved. LEARNING BY DOING There are more detailed accounts of the SAMs used in this paper, which the reader can consult. But, while further reading may give the reader a fuller taste of what a SAM is all about, there is probably no substitute for learning by doing. A do-it-yourself SAM does not have to be on the scale of the SAMs described in this paper. National accounts, balance of payments, and financial data (such as the central bank's balance sheets and consolidated statistics for commercial banks) are often readily accessible and are sufficient to start the construction of a rudimentary SAM or even a series of SAMs for different years. The data, at first, may appear inconsistent or inadequate even to this task, or other questions may crop up. But once one starts asking questions about the data, one will begin to appreciate some of the reasons it is useful to adopt a SAM framework for the numbers. Data of this kind can often be found in the appendices of World Bank reports. The example shown in table 1.19 is taken from a report on the Yemen Arab Republic. The SAM itself is only a first cut using eight tables in the appendix for basic information and two others to make very crude estimates of the allocation of indirect taxes (line 2b) and imports (line 4) to the two production sectors (columns 3a and 3b) (World Bank, 1979; tables 2.1, 2.3, 2.4, 3.1, 3.8, 5.1, 6.1, and 6.3 were used for the basic data, and 3.4 and 3.5 were used to make the crude estimates). Anyone with access to the original data could easily improve on the SAM. The point here is to show how even a limited exercise can throw up questions of substance and consistency for fArther investigation. There are forty entries in the core of the SAM and fifteen totals (each appearing twice); of these, twenty-six, including the rough estimates, could be directly entered in the core of the SAM, and ten could be entered in the totals. The rest followed easily by simple addition or subtraction, making, it is true, some arbitrary assumptions about the location of small residuals; there was one independent check on the outcome. A special feature of the table is the row/column lb devoted to remittances. A large part of the Yemeni labor force was working in Saudi Arabia and the Gulf states at that time. In the year in question, gross remittances (there was also some reverse flow) were equal to nearly 50 percent of factor income. The proportions were changing extremely rapidly; the corresponding A Table 1.19. A Simple SAM for the Yemen Arab Republic, 1975-76 (in millions of rials) Expenditures Total Current Accounts Capital Accounto Financial Accounts 1l lb 2a 2b 3a 3b 4 5 6a 6b 7a 7b 8 9 10 11 la Factor Income 4220 508 100 4828 lb Reaittancee 2363 2363 a 2a Private Sector 4789 2057 6846 0 o 2b Public Sector 39 72 338 115 564 1 3- Production (consumption) 4900 681 293 5874 R 3b Production (capital) 814 262 94 1170 c 4 Rest of World 306 1316 547 2169 e 5 Private Sector 1874 -11 121 178 2162 p t* a 6a Public Sector(budget) -117 15 609 507 x 6b Public Sector (other) 9 -15 103 97 7a Rest of World (BOP) -587 1288 701 7b Rest of World (other) -45 181 136 8 Central Bank 886 258 108 1252 9 ercial Banks 462 -13 3 452 10 Foreign Borrowing 712 712 Total 4828 2363 6846 564 5874 1170 2169 2162 507 97 701 136 1252 452 712 Note: BOP signifies balance of payments. What Is a SAM? 51 figure in the previous year was less than 25 percent and in the following year over 75 percent. The response of the private sector to this rapid increase in resources was to save more than 25 percent (column 2a). Of these savings, less than half was invested in physical assets; the rest, plus a substantial amount of borrowing, was retained in the form of currency and deposits with commercial banks (column 5). At the same time, the government was borrowing abroad, more than enough to finance its investment. (It, too, accumulated funds in the central bank, row 8, column 6a.) Two other features of the SAM are the consequence of the form in which the data were available. The public sector capital account has been split into budgetaxy and "other" trans- actions (rows/columns 6a and 6b). Government expenditure (investment plus consumption) in the national accounts exceeded expenditure in the budget. Similarly, official loans or grants in the balance of payments exceeded borrowing recorded in the government budget. There were evidently "government transactions" outside the budget, a large part of which must have been due to investment in public enterprises. The division is intended to draw attention to this point; there is some gratification to be had from the fact that the adjustments required to balance the line and column were trivial. Similarly, there are two subdivisions of the rest of the world capital account. One is for the official balance of payments. Again, however, there were additional transactions, as is evident from the increase in foreign assets (row 7b) over the officially recorded reserves in the central bank (row 7a). This increase has been balanced by a corresponding inflow of capital, here allocated mainly to the private sector. One residual in the table is the intersection of row 8 and column 9. This represents an increase in deposits of the commercial banks with the central bank. Obviously, there should be little difficulty in checking this figure, but our purpose here has simply been to present what can be done with a particular set of information rather than going beyond it. The time invested was not large, about half a day of uninterrupted time. (The compiler, moreover, had no previous familiarity with the country concerned.) Yet this was sufficient for the preparation of SAMs for six successive years. It seems a small price to pay for an articulated set of accounts which reveals, at least in order of magnitude, the salient features of the economy. It is arguable that six successive matrices of this kind give a better appreciation of change- in this case kaleidoscopic change-than do individual tables of the traditional kind. If this final example strikes a mundane note on which to finish, that may not be inappropriate. The SAM approach is a flexible tool which can be deployed with varying degrees of sophistication and for a variety of purposes, once an initial investment has been made to learn how. Although economists have long since understood that their analyses can each be set within a framework of accounts, this aspect has not usually been developed. The general point is that an economist who understands SAMs will probably be better equipped to tackle a variety of problems than one who does not. 2 SocialAccounting Matricesfor Development Planning Graham Pyatt and Jeffery I. Round It is well known that accounts for transactions within an economy can be presented in matrix as well as double entry format. Such a matrix is known as a social accounting matrix (SAM) and must be square. I Within it each row records the details of receipts by each particular account while the columns (which follow the same ordering as the rows) record the corresponding expenditures. Thus the entry in row i, column j, represents receipts by account i from account j or, alternatively, expenditures by account j that are paid to account i. Within such a general schema, SAMs can take a wide variety of forms, depending on how the constituent accounts are defined. A particular and most important variant is provided by the United Nations System of National Accounts (SNA), which has set down guidelines for deriving national income statis- tics as part of a more comprehensive social accounting matrix approach (United Nations Statis- tical Office, 1968). It is noteworthy, however, that only a small part of the text of the SNA is directed toward the specific needs of the developing countries, and even then the discussion is downgraded to "suggestions" rather than "guidelines" for implementation. A full implemen- tation of the SNA has frequently been questioned as a statistical priority, as has the need for a SAM approach to macroeconomic information systems. Our view is that the underlying philos- ophy of the SNA and the SAM approach is thoroughly appropriate to statistical systems for developing countries, but that some flexibility and a less mechanistic approach are needed for its actual implementation. In particular, we consider detailed disaggregation of factor and household accounts-implying, for example, separate accounts for different types of labor and for different types of household-as a priority. This position is developed in the course of this paper. Meanwhile, there are not many examples in which the SAM approach has been applied to developing countries, and our main purpose here is to outline and compare the results of three studies with which we have been associated. These have led to SAMs for Iran in 1970, for Sri Lanka, also in 1970, and for Swaziland in 1971-72, all of which attempt disaggregation of households or factors in one form or another. (References to sources and methods used in these studies are provided as part of the discussion of each.) Tables 2.1, 2.2, and 2.4 give a preliminary impression of the results that these studies have yielded. Further detail of each of the studies is provided in the third, fourth, and fifth sections of this paper, where some of the practical difficulties encountered in our work are described. Before coming to these studies, we discuss in the next section some of the reasons for under- taking this work. This is necessary for a number of reasons. One is the contention that the need for data systems derives from concern for quantitative advice on policy and that the characteristics of such systems feed back onto the nature of advice that can be offered. Such considerations explain why our studies depart from SNA recommendations in some respects. Specifically, the motivation of our work has been the need for an information system to advise on the issues of employment opportunities and income distribution, which have challenged the conventional emphasis in macroeconomics on growth alone. This need has been clearly iden- Note: This chapter has been published with minor modifications in the Review of Income and Wealth, series 23, no. 4, December 1977. The authors are indebted to Dudley Seers and Stanley Webster for comments on an earlier draft. 1. Nonsquare formats can be defined, but these are always derived, conceptually at least, from a square matrix, which is the basic format. 52 SAMs for Development Planning 53 tifiedby the International Labour Office,World Employment Programme (ILO WEP), and implies the view that economic growth is inadequate as a policy objective unless its content, in terms of the living standards of different groups within society, is spelled out (International Labour Office, 1976). Acceptance of this position implies that conventional data systems which derive from a preoccupation with aggregate growth or average living standards must also be judged inadequate. Accordingly, we greatly regret the separation of the UN SNA from the System of Social and Demographic Statistics (United Nations Statistical Office, 1975) and have made a start in our work toward the integration of the two. Thus in a narrow sense the SNAis inadequate for our purpose. 2 This point, however, is subsidiary to the fact that developments or modifi- cations of the system, such as we have explored, are greatly helped by the underlying philosophy, that is, by the SAM approach. If the SNA is interpreted as having championed this approach, rather than in its specific detail, then we would see it as having a great deal to offer developing countries, which they may ill afford to be without. Meanwhile our three case studies illustrate the feasibility of making progress in this direction. BACKGROUND TO THE STUDIES The historical origins of the SNA, going back 300 years, are set out briefly by Stone in his foreword to Pyatt, Roe, and associates (1977). Our discussion can start from a more recent event, namely, the inception in 1960 of work on the Cambridge Growth Project, which was 3 initiated by Stone in association with Brown. This work produced the first SAM, as we now know it, as the information system counterpart of early versions of the Cambridge growth model (see Cambridge University, Department of Applied Economics, 1962-74 and 1975). At this time the structure of the welfare state in the United Kingdom was well established, so that questions of employment opportunities and care of the needy were not pressing. The issues that caused most concern were those of economic growth, or rather a comparative lack of it. The focus of the work was therefore on industrial structure. To carry it through called for various developments on standard input-output analysis so that contributions were forthcom- ing, such as the R ABmethod of updating technology matrices and the use of "make matrices" to supplement commodity-by-industry specifications of technology. 4 The latter especially is now firmly established in the SNA recommendations. It is important to emphasize this link between policy, data, and models because it is essential and permeates our own work. In the SNA the link between data and models is fully explicit, and both aspects build on the earlier Cambridge work. Unfortunately, it is in the nature of affairs that the policy applications of the SNA have to be taken largely as read. Much of the complication of the revised SNA seems hardly worthwhile if the purpose is simply to get better estimates of national income. At least some criticisms of it might be muted if it is realized that the purpose is to describe an economy in detail with a view to changes, or to make sure it remains on course. In this view the heart of the SNA is the model that the data serve to calibrate, in much the same way that the economics of Keynes is the rationale of conventional national 5 accounts. 2. It is noteworthy, however, that the draft report "Complementary Systems of Statistics of the Distribution of Income, Consumption, and Accumulation," adopted by the TINStatistical Commission in 1972, proposes ways of integrating household income distribution data and the SNA. 3. We are advised that there are antecedents from work in Norway and the Netherlands dating back to the 1930s and 1940s. 4. See Cambridge University, Department of Applied Economics (1962-74), vols. 1 and 3, for early references on these subjects. The RAS method was subsequently developed by Bacharach (1970). The project has pioneered a number of other contributions, Those cited, however, are the ones that have become most firmly established in statistical-as opposed to modeling-work. 5. The economics of Karl Marx leads to the net material product concept, as opposed to national income. 54 The Methodology of Social Accounting The links between policy, models, and data in our own work explain its special characteristics. The ILO WEP sent a comprehensive employment strategy mission to Colombia in 1970, under the leadership of Seers. The report of this mission (International Labour Office, 1970) raised the question of whether its recommendation and those of other such missions could be set in a comprehensive consistency framework .6 The next WEP mission-this time to Sri Lanka and again with Seers as its leader-provided an interesting opportunity to pursue the issues for two reasons. One was the fact that Seers has been a pioneer in this field for many years and his national accounts for Zambia, for example, are a fascinating and unconventional attempt to address the data requirements of a developing country and to reconcile them with what is possible (see Frank, 1967, for a discussion of Seers's approach). From his subsequent writings (Seers, 1975), it may be fair to classify Seers as a critic, if not an opponent, of the SNA. His unconventional system for Zambia, however, can in fact be rearranged to be a more or less conventional SAM, while his criticisms of the SNA can all be embraced by it in its SAM incar- nation. The second factor to make the choice of Sri Lanka propitious for the issues under discussion was that a considerable amount of time and energy had been spent in deriving a credible series of national accounts, as part of an earlier planning project by the UN Development Programme (UNDP). These were complemented by an input-output table estimated by Narapalasingam, who subsequently used these data to build a planning model of Sri Lanka along the lines of the Cambridge growth model (see Narapalasingam, 1970). A particular feature of this work involved experimentation with the effects of income redistribution. As such it was a pioneering effort. The case study of Sri Lanka discussed in the fourth section of this paper was undertaken as part of WEP research to resolve some of the issues that this earlier mission had raised. In the interim, however, there was another WEP country mission, to Iran. The respective economic circumstances of Sri Lanka and Iran imply that the issues of growth, employment, poverty, and income distribution arise in quite different settings. In Iran a crucial question concerned the extent to which policies for growth might need to be modified in order to do more for the poor, especially in rural areas. The modeling of income distribution questions was therefore important, and a data system that embraced them was needed accordingly. Nara- palasingam had been able to avoid this need because his model of Sri Lanka looked only at how a change in income distribution influenced consumer demand, and hence the structure of production. He did not consider how production structure influenced factor payments and hence income distribution. In this sense his model was incomplete. In the case of Iran, both directions of causality were thought to be crucial. The model and data system were therefore designed to capture them both, otherwise building on earlier work in Sri Lanka. 7 The need to introduce income distribution Lnto models and social accounts has meant going beyond the realm of the SNA into the province of the TN System of Social and Demographic Statistics (SSDS). This has raised a number of questions, some of which are touched on in what follows. Meanwhile we have already mentioned our regret that this development of economic and social statistics should be separate from the SNA. Our work indicates that it is relatively straightforward to integrate aspects of both systems at the conceptual level, which is perhaps not surprising since Stone is the prime architect of the SSDS as well as the SNA. 5 This facility is illustrated with respect to income distribution by the case studies discussed in the next 6. One early response to this question has been a paperby Thorbecke and Sengupta (1972). Subsequently, Thorbecke has set out his views as part of the evaluation of the first five WEP comprehensive employment strategy missions. See International Labour Office (1973b). 7. Apart from the Narapalasingam study some modeling work was undertaken as part of the mission In Sri Lanka. Some of this is reported in International Labour Office (1971), vol. 2, technical appendix 4. 8. Stone's early thoughts on what eventually became the SSDS were presented as the Radcliffe lectures at Warwick University in 1973. SAMs for Development Planning 55 section. In other areas, such as housing, nutrition, education, and wealth, we have no empirical results as yet. But much thought has been given to the issues, and a preliminary report is available (Pyatt and Thorbecke, 1976; see also United Nations Statistical Office, 1972). Essen- tially our view is that the integration must go ahead if the data system is to serve the current policy debates on these questions. Sen (1973) has made reference to the problems for economists of abandoning the welfare principles of Pareto and incorporating income distribution questions into their thinking. But there is no intellectual problem in integrating the income distribution component of the SSDSinto the SNA.Our experience is that the SAMframework is an invaluable aid through the empirical problems of doing so. It is also our experience that the end product is widely perceived to be relevant in a way that the standard SNA is not. Indeed our Swaziland case study derives directly from interest among individuals of the Overseas Development Minis- try, London, in the replicability of the Sri Lanka study. THE IRAN CASE STUDY The case study of Iran resulted in the smallest and most confused of the SAMs presented here--some learning by doing has been involved through successive studies. The basic frame- work of accounts and estimates for 1970 are given in table 2.1, which is extracted from the original source (Pyatt and others, 1972). The table resembles a conventional input-output transactions matrix; the first twelve rows and columns relate to the incomings and outgoings of a set of production activities. The remaining rows and columns record receipts and expen- ditures for other accounts in the system (a la SNA) and show relationships between domestic and foreign institutions, as well as relationships between these institutions and production activities. In the usual way, the first twelve columns show the outgoings of twelve domestic activities. These consist of raw material purchases, payments to institutions of value added, imports of raw materials, and indirect taxes on inputs. The revenue of production activities derives from the intermediate sale of commodities, plus institutions' current expenditure, exports, and sales of capital goods. Four institutions are distinguished: three types of households and government. Value added is shown as a direct accrual to these institutions, so that company profit is included as income distributed to households and government. This feature is important for comparison with other case studies. Apart from value added payments arising out of production activities, incomes are also created by households in purchasing domestic services and by government through wage and salary payments for public administration. Gross national product at market prices-771.2 billion rials, as shown in cell (18,26)-comprises these elements of value added, together with net income from abroad, and indirect taxes; this is shown in column 26, distributed over the four institutions, as total incomings to their current accounts. Outgoing values from these accounts appear in columns 14 through 17. Expenditures of the institutions on domestic commodity outputs and on imports, and on payments of indirect and direct taxes are shown. 9 Separate company accounts, however, are not shown, and company profits are recorded as an ultimate receipt by households and governments. It follows that households and governments are deemed to make investment expenditures on behalf of the companies they own. A feature of this framework is the distinction drawn between three categories of households. Rural households consist of all people in the rural areas, which in turn are defined in the population census as places with less than 5,000 inhabitants. The urban population is divided into two groups: households in the top 30 percent of the urban expenditure distribution are 9. For the government, direct taxes are a negative olutgoing (a receipt of taxes from households) since they are a transfer between institutions. R .. ipt. IxPital ~ e., ~ o. Households a 02k- ou - ~ ~ ~ ~ . . .... M. K. f - -~~~~~ ot~. . . . . .. t. C-Pt-dC .. . . . . . . . . .+ . . . . .2d ~ ~ ~~ C...~ . . .. .. . .. .. . -hip f -Ilig ~ 1 t. 12 ~ ~ R-1~~~~~~~~~~~~~~~~~~~~~ce o 0' 0' 0 0'0 0 0' 0 0' 0 0' - 0 0' ' 0. - LiU, toI * ~ . . . .. . . . . . u.. l08-t .o 0 . . .. . .0'.0. . . . . 0' . 1tOfU I -~~ 0' 0' 00000'0' 0 0' 0 0 0 0 0 S orouu~~~21tu.u 25 .otro Planning SAMsfor Development 57 classified as Urban II, and those remaining, as Urban I. In other respects the SAM is fairly aggregative; for instance, only twelve production sectors are distinguished. However, these sectors are chosen on a wider set of criteria than simply homogeneity or similarity of products. The level of technology (modern or traditional manufacturing) and ownership (resident or nonresident oil sectors) are also taken into account. 10 The payment of factor income directly into institution accounts raised several difficulties. Although data on aggregate value added by activity were available from published sources, the allocation of these sums directly to the three household types and to government had to be subjectively estimated within constraints set by the classifications themselves (for example, rural wages had to accrue predominantly to rural households) and by controls on total incomes from all sources which were known or could be estimated. On the expenditure side of the institution accounts, separate commodity expenditures are shown for three types of household and for government. The allocation of private consumption between Rural, Urban I, and Urban II households was made on the basis of a family budget survey for 1965 carried out by the central bank. An important point to note is that after allowing for indirect and direct taxes and for imports, the difference between household expenditures and gross household incomes yields household savings. Government savings are similarly derived. The capital accounts are highly aggregative. The three household accounts are consolidated, so that three savings figures are shown as incomings to a single capital account. In consequence, the flow of funds, shown in the three penultimate rows and columns of the SAM (23 to 25), has a very simple structure. The balance of payments deficit (36.0 billion rials in 1970) is financed by capital transfers to households ( 16.7 billion rials) and to government ( 19.3 billion rials). Domestic savings are supplemented by these capital transfers from abroad to finance domestic investment. Financing of the private and public components of domestic investment is facilitated by a capital transfer from households to government of 41.6 billion rials. The resulting SAM for Iran is a very simple design. It is worth noting, however, that it was produced in a matter of days, rather than weeks, by twvopeople on the basis of (a) considerable relevant knowledge of a generalized kind; (b) reliance on published sources almost entirely; and (c) work in association with others." The feasibility of the exercise depended crucially on the rigors imposed by the accounting constraints in a SAM, since only in this way could the better data be fully exploited to support the weaker and more doubtful figures. Of course, better data would have been used if available. In this sense table 2.1 presents the best that could be obtained for Iran in 1970 without new primary information. Its quality owes much to the discipline of working in the context of a SAM, and it was judged adequate to support a modeling exercise for Iran which considered the two-way links between production structure and income distribution referred to earlier. Aspects of the model used in ^onjunction with table 2.1 and some of its analytic properties are discussed in the original source and elsewhere (see Blitzer, Clark, and Taylor, 1975, and chapter 10). These need not concern us here beyond noting that the "single entry" accounting that characterizes a SAM requires that the treatment of items as a receipt be consistent with their treatment as an expenditure. In a model context this means that the effects of income distribution on production must be consistent with the effects of production on income distribution if the results of the model are to be expressed as a new SAM in the format of table 2.1. 10. The nonresident oil sector is excluded from the production activities distinguished in the table. Its contribution to national product is treated as an indirect tax receipt on exports, that is, in row 20, column 19. 11. The prime calibrators of the Iran matrix were Julian Bharier and Robert Mabro. Other members of the team were Robert Lindley, Graham Pyatt, and Yves 8abolo. The team report (Pyatt and others, 1972) discusses sources and methods in detail. That such an approach was worth trying emerged from a preliminary reconnaissance of the issues with Abdul Meguid, who had previously constructed a more conventional input-output model not available to us. ?-I. Table 2.2. A Social Accounting Matrix for Sri Lanka, 1970 (in millions of rupees) Expenditures 2 3 4 5 6 7 Factors of Production Institutions Production Activities Labor 1 Other Current A,cou,,ts Households c 0 ' v r Zeo n c ie Z _ _ _ _ D t w 0 Oc a - 07a a i - n 3 8 ot 2 j Z Z 3~* a *a.o.0 P ci .a.. . . 'i ,ffl 0 U ciA-' Qa 0 50 ci ee 8 iS c 00 s X >r ' a5t a -a~~~ - a aoo a3 >5 -c 5 0 45 55 a'0 0a ' > s,c .0 54 - 0 0 S.C a u GI~~~~~~~~~~~~t ad 0 -4 on iicc e Urban 5 5 9 25 75 46 182 81 414 276 555 1673 a oi Rural 43 158 67 706 247 68 259 159 487 276 715 3185 5 .a Estate 526 133 11 5 4 2 5 8 12 5 711 0. o S~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~3 Housing 633 63 5 o Other Private 13 24 442 282 1259 184 604 742 1424 123 -113 49S4 Public v_ -11 1.2 1.09 '8 -1 73 174 Si0 Urban .1673 137 662 434 91 6 3003 2 Si o ~ Rural 3185 330 3026 203 151 6 6901 o Estate 711 31 30 7 6 6 791 Si0 0 0r 3 ei c Psivate Corporations 135 1266 571443 o' 3 v . State Corporations 174 237 411 4 Gover-icent 368 194 4 272 104 313 33 4 14 10 19 288 216 66 130 76 29 94 2234 S co:nbined Capital Account 519 807 11 527 301 43 425 2639 Tea 14 55 7 -55 2 2 839 664 o Rubber 25 8 341 374 a Coconxuts 54 208 27 29 239 8 6 4 2 577 i Rice 158 760 102 105 1082 2 15 18 22'2 t h:ber Agriculture 357 980 139 11 1 2 95 63 34 3 39 16 106 1846 6 Food and Drink 253 541 82 37 9 11 24 188 29 8 94 1276 Other I-dustry 258 621 69 72 97 24 9 35 69 49 554 417 172 37 66 241 2,55 j Coistruction 1595 1 7 50 92 17.'5 r Trade assd Transport 410 1074 122 154 50 10 8 44 23 95 249 206 96 42 59 203 2S'5 o Private Services 405 920 85 11 3 7 15 1 4 9 38 55 37 287 1877 _ Government Services 1649 1649 7 Rest of the World 207 741 143 364 75 12 10 32 53 204 370 65 70 133 43 2522 Total 1673 3185 711 633 4984 174 3003 6901 791 1443 411 2234 2639 864 374 577 2242 1846 1276 2790 1745 2845 1877 1649 2522 Note: This SAM is an early version of the results published in Pyatt, Ro0e,and associates (1977). SAMs for Development Planning 59 THE SRI LANKA CASE ST'lUDY The experience of the Iran study indicated that data requirements, and not modeling, were the initial obstacle to progress with planning techniques that embrace employment and distri- bution questions. Accordingly, our subsequent efforts have focused in this direction. The SAM for Sri Lanka shown in table 2.2 indicates some of the innovations achieved in this study. First, it is immediately apparent that there is a new set of accounts, not included in the Iran matrix, relating to factors of production. Second, the accounts have been rearranged so that, for example, the factor accounts lead in the rows and columns. A third difference, not readily apparent from the table, relates to the compilation of accounts. Each of these will be discussed in turn below. But first it needs to be emphasized that the study in Sri Lanka was a much Larger exercise than that in Iran. Hence table 2.2 is only a summary of results for the latter, while table 2.1 is more exhaustive of the output of the Iran study. (The Iran study also involved an estimated SAM for 1972 and detailed analysis of labor statistics, as well as the modeling work previously referred to.) This difference in order of magnitude is discussed later. Meanwhile, further comments on each of the three points referred to is in order. The factor accounts in table 2.2 are in addition to the production and institution accounts shown. hitherto. Their main purpose is easily stated: it is to receive factor payments, from both domestic production activities and from the rest of the world. These in turn are mapped into the household and other institutions accounts, thereby recording the factor income component of the gross income receipts of institutions. Nonfactor incomes, such as current transfers between institutions and transfers from the rest of the world, augment factor incomes to yield gross incomes of institutions. The distinction between factor and institutions accounts serves two important purposes. In the first place, a clear distinction can be made between factor income and nonfactor income that arises from the redistributive process within the economy. These redistributive forces are likely to be a centerpiece of policy and planning strategy, and therefore need to be captured in this way. In the second place, the classification of factors can be entirely divorced from insti- tutional classifications. The latter can be determined by a range of socioeconomic considerations; for households these may include location and socioethnic factors as well as income level; for other institutions "ownership" or "purpose" might be appropriate. The Sri Lankan institutions shown in table 2.2 are in fact an aggregation of more detailed accounts. Thus in the full study each of the three household classes is further subdivided into six income groups, and the government accounts are disaggregated into ten categories for income receipts and nineteen heads (or accounts) for expenditure. Similarly, in the full study the factors are classified accord- ing to the kinds of economic agents employed by production activities and thereby receiving factor returns. In table 2.2 only six factors are shown, but this is a more aggregated version of the classification that was in fact utilized. For example, three kinds of labor are distinguished in tab:Le2.2 (urban, rural, and estate labor), although a disaggregation of labor income by nine occupational groups was also achieved. Three nonlabor factor accounts are distinguished: the factor "housing" simply receives imputed rents on owner-occupied housing, while all other returns are divided between private and public ownership of capital. Each of the six factor accounts has a row sum which accumulates all domestically generated factor incomes, together with net factor income from abroad which is shown to accrue to the factor "other private capital." The total of all factoral row sums conveniently shows gross national income (GNI) at factor cost (Rs. 11,360 million in 1970), while the arrangement of the table puts the individual factor accounts first, so that the decomposition of GNI into its factoral distribution is also given prominence. 60 of SocialAccounting The Methodology The arrangement of accounts in the table is a conscious attempt to capture the circular flow of income-from income generated by activities to factors; from factors to the institutions that provide factor services; and from the expenditure of income by institutions to demand on activities, and hence income generation. It is also an attempt to give prominence within the SAM to the things that matter most. For us, these are employment and income distribution questions set in a framework of the level and structure of activity. Thus our accounts start with factor incomes and move to the incomes of households and other institutions in the economy. These, not the structure of production, are our primary concerns. A third innovation in our Sri Lanka study concerns the practical procedures for compilation of the SAM. In many ways the data base for Sri Lanka differed from the situation encountered for Iran, although the basic notion that the SAMimposes a discipline for data consistency was sustained. Some details of the methods used to construct the Sri Lanka SAMmay be of particular interest, since these involve several novel features for social accounting in developing countries. (The most recent national accounts for Malaysia make use of the basic approach described herein; see Malaysia, Department of Statistics, 1975.) In close correspondence with SNA guidelines, our Sri Lanka SAM is built up from data on the supply and disposition of commodities, traced through the rows and columns of the produc- tion accounts. Using a 1965 input-output matrix to depict the approximate structure of produc- tion, the commodity balances for Sri Lanka in 1970 were achieved in a systematic, yet nontrivial way. The complicating factors were not only incomplete or uncertain data, but also multiple estimates of some elements. We were faced, for example, with two (or more) estimates of most value added components-one from Ceylon, Department of Census and Statistics (1973), the other from Ceylon, Central Bank (1971). Furthermore, available figures on gross outputs were recognized to be subject to a substantial degree of error. The final use components of commodity requirements (household and government expenditures, fixed capital formation and exports) were more reliable, except for the vector of changes in commodity stocks, and this was used as a residual in the commodity balance calculation. The derivation of a set of commodity balances was achieved in four stages. The first set of commodities considered were those which make no sales whatsoever on intermediate account: in Sri Lanka these sectors were tea, bread, other bakery products, and tobacco manufacturing. These were investigated first because final sales must also be equal to gross sales and gross output for these sectors. On the input side, with gross outputs thus ascertained, the input technology (determined principally by the 1965 input-output matrix) gives some value added estimates as well as estimates of intermediate input requirements. Alongside these four sectors were those known to have few sectors to which intermediate sales were made: rubber and fishing, for example. The second group of sectors to be investigated were those belonging to what we term "process loops." The coconut group is one example. There is a linkage between the three sectors: coconut; dessicated coconut and copra; and coconut fiber and yarn. For these sectors intermediate sales and purchases exist but are largely confined within the process loop. These intermediate sales can easily be estimated from the final demand estimates, given some knowledge of the nature of the interaction between various sectors within a loop. The gross output of the sector of coconut fiber and yarn, for example, can be set equal to its final demand entry since it has almost no intermediate sales. Working through the backward linkages in the loop, we found it possible to iterate to feasible (though, admittedly, not unique) solutions for the gross output, value added, and intermediate transactions of each of the sectors involved. 12 The third stage of the commodity balance procedure was to investigate the remaining sectors. 12. Sectors belonging to a process loop form a natural aggregate sector, of course, but it is convenient on occasions to distinguish between them, as when, for example, the outputs serve different export markets as well as further stages In the product process. Planning SAMsfor Development 61 Clearly some intermediate sales had been determined from the first two stages, which limited the problem. This knowledge allowed sectors to be ordered so that those with relatively few undetermined intermediate transactions would be considered first. The fourth and final stage was to review the feasibility of the initial estimates of value added, gross output, and intermediate transactions. Inevitably this stage revealed some anomalies, and it proved necessary to repeat the first three stages, in an iterative manner, eventually converging to an overall feasible matrix that was not unacceptable on the basis of the facts.13 The balanced production accounts determined a consistent set of value added estimates for the forty-eight production activities in the study. The way in which they were derived took full cognizance of the various prior estimates of value added. Changes were made to the firmest of these estimates if there occurred even firmer estimates of commodity supplies and dispositions not consistent with the value added data. It is perhaps worth noting that in this regard our experience is at variance with the contention that estimates and "guesstimates" should not be placed side by side (Barkay, 1975). The SAM consistency framework forces a confrontation between various data sources, which can never reasonably be expected to be of equal quality. Reconciliation of data of varying qualities is therefore unavoidable. Moreover, unless data are literally useless they can add something to SAM calibration. There is no sensible alternative to setting all sources (with prior judgment regarding their relative reliability) alongside one another and executing an "optimum" balance. One consequence of doing so is that the SAlM approach teaches a great deal about statistical priorities for new information. After the production accounts, the next step was to obtain a balance of all the institution and factor accounts. From several standpoints the government accounts and the accounts governing transactions with the rest of the world were the firmest. They therefore formed a basis for this part of the matrix. We were particularly fortunate in being able to utilize a socioeconomic survey, or SES (Ceylon, Department of Census and Statistics, 197 1), which enabled us to obtain disaggregations of household expenditures according to urban, rural, and estate subdivisions. Not surprisingly, SES estimates of household (and factor) incomes implied negative savings in all household groups, confirming our expectations of underrecording of incomes in household surveys. In this situation, assumptions about the economywide capital/output ratio, the rela- tionship between business investment and retained profits, and the implicit constraints of the SAMwere all utilized to obtain a feasible solution to the remaining cell entries, including revised estimates of household incomes and savings. The complete table has eighty-seven rows and ninety-six columns, as shown in table 2.3, which gives listings of the detailed accounts in the full study. Justification for these and the full results are set out in Pyatt, Roe, and associates (1977), which attempts to record the man- year of work that went into the study.14 It can be noted, however, that this man-year was in fact collapsed into a period of less than three months, with the team involved averaging some six people through this period. 15 There are two further comments to be made on the Sri Lanka study. First, it will be recalled that the purpose of the exercise was not to build a model as such, but rather to push forward 13. In his foreword to Pyatt, Roe, and associates (1977), Stone has suggested that more formal techniques of data reconciliation may have advantages. He refers in particular to statistical techniques that iterativelybalance the accounts subject to initial estimates and sets of constraints, both of which maybe subject to uncertainty. This issue is the subject of continuing research in the Development Research Center, World Bank. Some initial results are given in Byron (1978). 14. The 87 x 96 matrix was estimated in full detail subject to one caveat. This caveat arises because estimates of current transfers between institutions could not be obtained at the level of detail of table 2.3 but only at the more aggregate level of table 2.2. Otherwise the 87 x 96 matrix was estimated in full. Thus at the full level of detail only the eighteen household current accounts are incomplete, but fufl accounts for three aggregate household types, and therefore for a 72 x 81 matrix, were obtained. 16. The team comprised S. Narapalasingam and Neil Karunaratne, respectively from the Ministry of Planning and Employment and the Industrial Development Board, Sri Lanka; Alan Brown and Robert Mabro from Oxford University; and Robert Lindley and Alan Roe, in addition to ourselves, from the University of Warwick. Table 2.3. Summary of the Extent of Disaggregation of the Full Sri Lanka SAM Aggregate Accounts Number and Nature of Component Number and Nature of Component as in Table 2 Accounts Shown in the Rows Accounts ShowTn in the Columns (1) Factors of Production 6 (three accounts for 6 different employment statuses and three accounts for the factor of production, capital)* (2) Firms Current 3 (Private, Public Financial 3 Institution, Other Public Companies) (3) Households Current 18 (Urban, Rural and Estate 18 and within each of these, six income classes) (4) Government Current 10 (seven categories of tax, 19 (eight accounts for expendi- one account for current transfers, ture on goods and services, one account for Local Government nine accounts for transfer and a summary account) payments, one account for Local Government and one summary account) (5) Consolidated Capital Account 1 1 (6) Production Activities 48 48 (7) Rest of World - Current 1 1 TOTAL 87 96 Note: Alternative factor accounts were also produced showing nine categories of skill as well as capital. Planning SAMsfor Development 63 the possibilities for modeling by resolving some of the problems of data system design and availability. However, Pyatt, Roe, and associates (1977) include a number of empirical exercises using the data and showing its immediate relevance for policy issues. These exercises include a description of the economy with reference to income distribution; the output, income, and employment multipliers in the economy; an analysis of export incentives and of effective protec- tion; and a study of the structure of household expenditure in Sri Lanka focusing on its sensi- tivity to the income distribution. These, then, are some of the products that can be obtained short of a full-scale model once data has been set up consistently in a SAM framework. The second and final comment on Sri Lanka is that increasingly, as countries come to adopt the SNA (and hence the commodity balance approach), the preceding discussion of the way we were able to implement such an approach may be of interest. The methods that proved successful in Sri Lanka were nonetheless challenged by the subsequent study in Swaziland. Comment on these methods is reserved, therefore, until after the description of our work in Swaziland. THE SWAZILAND CASE STUDY Interest in the replicability of the Sri Lanka case study led to the formation of a group of similar size to spend six weeks in Swaziland and about the same time subsequently in an attempt to set the major economic statistics in a SAM context.'- As in the Sri Lanka case, there was no initial intention to undertake modeling work immediately, and the focus was an endeavor to contribute directly to policy discussion on the basis of an understanding of the economy, for which the SAM exercise was to be the catalyst. In several respects Swaziland offered the opposite of Sri Lanka in available data. Not least, the SAM framework which we intended to estimate, and which was broadly comparable with the Sri Lanka matrix in its dimensions, was not fully determined by available data. Although the details of achieving SAM estimates differed markedly between Sri Lanka and Swaziland, it is interesting to note that a common approach was sustained, and the discipline underlying the SAM was revealed to be of unquestionable value in deriving estimates. Table 2.4 sets out aggregate accounts for Swaziland for the year 1971-72. These are aggre- gative in the sense that more detailed estimates were obtained corresponding to disaggregations of some of the accounts shown. Thus, although the nine factor accounts and the seventeen institution accounts of the study are shown in full detail, table 2.4 consolidates into one account each of the forty-four commodity and twenty-five production activity accounts that were distin- guished. The distinction of activity accounts from accounts for the commodities that they produce is one of the main differences between the Swaziland and the Sri Lanka matrices. This distinction is, of course, very much in line with the SNAguidelines, and we found that it afforded a conceptual flexibility in the definition of activities and commodities which was also advantageous in the estimation of the matrix elements. Before considering the commodity/activity distinction and the determination of the commodity balances for Swaziland, we must mention several of the classifications embodied within the factor and institution accounts. The nine factor accounts, distinguished in the first nine rows and columns of the matrix in table 2.4, are novel in several respects. It should first be understood that the organizational aspects of the supply of agricultural factor services within Swaziland are complex: part of the land is held by the Swazi nation, and the remainder is still owned by individuals-often non- Swazi--and is generally farmed according to modern agricultural technology. Within the Swazi 16. The group was financedby the Economic and Social Committee on Research (ESCOR)of the Overseas Development Ministry (ODM), London. It comprised Harry Fell and Stanley Webster, on secondment from the ODM;Graham Jones and Malcolm Waamsley from the Swaziland Department of Statistics, Ministry of Finance and Planning, Mbabane; and the same four Warwick colleagues who undertook the Sri Lanka study, plus Paul Stoneman, also from Warwick. Table 2.4. A Social Accounting Matrix for Swaziland, 1971- 72 (in mniIAonsof emalangeni) 1 ~ ~ ~~ 3 ~~~~~~ 45 67 ~~~2 I Fact-r f Pc.Odccton Iatit.tul... Curen Arront 4 .TotAl- T.o.F. Urban Custoon - 3-1 Nation Na-LOD.A. 9.26 9.26 Soni Noti-n 0.D.6. 0.93 0.93 I IA1,Odaa T--a Foo 1.1 1.15s Orbart,s L- od Not. R. -o..0.56 0.56 Zia..~Co n.WanIo 0.1 .22 1s 4.81 32805801 Saf owl,- 8 N8.1.3 1.83 0th .83.91 - 3.01 ZOthar Capit.1 - 800 9.67 96 L 00.30 20~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ i Nt &.D.A. 9.26 2.19 0.39 0.01 ).06 0.03 0.02 0.36 0.04 12.30 1.28 N.D.A. 0.93 0.25 0.04 2 - . 8.5 lon 1.12 0.22 4.03 0.12 0.17 0.15s 0.02 6.53 1-. boo i 4.24 0.02 0.46 .,06 0.02 4800 818Ina. U' 0.03 0.08 20.04 1.20 2.27 0.95 0.25 24.82 ioN- Inca 7.86 0.40 0.70 .06 0.05 0.02 0.85 3 N-.ftBt.0.05 0.13 0.18 9 -hoi NatLor1 Cooci 0.11 0.29 0.23 0.63 0 aPablic 2.01 4 ~ ~ ~~010.hO ~ 0.07 -0.16 7.66 8.04. 1.01 0.29 0. 2.00 DIoac 0.18 0.10a I ~~~~ InAirart ~~~~~~~~~~~~~~~~~2.094 .a.-....---- 0.06 0.03 0.00 0.52 0.07 2.60 6.20 a! ..a ConcItd 0.10 6.20 2.04 0.52 S Othor cIrot- To-s .0.) 0.127 1.21 1.38 flirarn Tan.. ~~~~~~~~~~~~~~~~~~~~-.---. 1.89 -.--- 0. 94 2. 90 5.2 Oth-a GosoocA -o 0.27 0.-.--. 0.21 - ,0.76 0.22 0.3 .7 Consltdatd 00000 8052 1238 5082 1.79 -0.62 -3.93 -1.46 .9.09 2.41 6 Co..bi-d Cdo..tl U-ot ~- 7.0 --.... ~. 0.63 ~.05 12.26 1.05 2.41 -18.29 -2.69 -5.06 7 C-mdlict. . 44.23 2.0941.30 3.65 17.77 2. 62 6. 6.9231 N Fe-adacton ntili~tla 168.03 0408.03 9 C-bita R-a of W-rld Acoao 0.96 1.02 6.36 0.53 0.49 0.00 56.02 65.36 Total. 8.20 0.93 1.15 0.58 38.01 1.03 3.91 9.67 20.35 12.50 1.20 6.53 4.00 24.82 8.85 0,906 .1 70 25 .10 6.20 9.52 1.30 5.02 0.79 2.41… … … … … … 203.15 140.03 165.36 SAMsfor Development Planning 65 nation land, rural development area (RDA) schemes are currently being introduced and repre- sent a significant break from traditional methods.' 7 To avoid an arbitrary division of the returns to land and labor of the self-employed in the agricultural sector, we defined a composite factor for each of these three types of land: Swazi nation (traditional), Swazi nation (RDA), and individual tenure farms. Labor receiving employee compensation is shown as a separate cate- gory, as is self-employment income from nonagricultural activities. It is also worth noting that in the Swazi context one individual may be supplying his services in the form of two factors in the course of a normal year. For example, he may be working on a rural homestead, thus accruing factor income of the first kind, and may also be receiving employee compensation for casual work in urban areas. More typical is the instance in which members of a rural homestead (household) will be supplying a variety of factor services in both rural and urban districts. In such situations the importance of distinguishing factors from institutions and carefully choos- ing the classifications for each of them is obvious. Finally, returns to other capital (that is, capital other than land) are distinguished as Swazi and non-Swazi controlled, a distinction which is of considerable importance in the policy context. The first two institution accounts relate to two forms of traditional Swazi households: those outside and those within the RDAs. Each household receives the major part of its gross income from the factor income derived from its traditional agricultural activity, although typically this is supplemented by employee compensation and self-employment incomes. The main sources of the supplementation are employment on Swazi nation land, in small traders' establishments, and in rural education and health services. Note that teachers, for example, who teach in schools and live on Swazi nation land are included in this part of the rural sector. Households on individual tenure farms and in urban areas are further subdivided into high- and low-income groups according to whether their aggregate income is greater or less than E600 a year. (E is the standard reference for the currency unit, the emalangeni.) The remaining institutions are nonhousehold institutions. We attributed separate categories to nonprofit bodies (which receive transfer income from government expenditures on health and education) and the Swazi National Council (which essentially receives income in the form of rent and mineral royalties). Three corporate categories were distinguished, which allowed separate accounts for large and small corporations. The merit of the distinction within a SAM between factors and institutions is more clearly seen in Swaziland than in our other studies. At the same time, flexibility is important since the distinction is most informative (and useful for modeling applications) when the classifi- cations are chosen to reflect the particular structure and organization of each country. Accord- ingly, we are against stereotypes for the detailed classifications. The classifications appropriate for the Swazi economy are only approximately replicated in other developing countries. In particular, the Swazi nation is a unique organizational form, and while it is important to reflect this aspect in the national accounting system for Swaziland, there may be no analogue for other countries. The data base was constructed from a variety of detailed sources. The national accounts statistics were a crucial input, but, unlike the Sri Lanka situation, there were relatively few instances in which multiple estimates were available for the major elements. For the most part, however, value added payments by production activities (in aggregate) were readily obtained. An important exception was the difficulty of making a distinction between RDA and non-RDA factors and households. Simply, the level of activity of RDAs was globally estimated to be 10 percent of other traditional agriculture, and this percentage was applied throughout the accounts in order to distinguish RDA from non-RDA classes. This was the most arbitrary of the assump- tions made and would have been avoided had it not been considered important in demonstrating 17. For 1971-72 this distinction within Swazi nation land is not of great significance, but for monitoring progress within RDAs it is ultimately a distinction of considerable policy interest. 66 The Methodology of Social Accounting how a SAM could be designed to monitor the future progress of RDAs when independent data become available. The derivation of the commodity balances differed substantially between Swaziland and Sri Lanka in several respects. Nonetheless, these balances still proved to be a useful starting point for the framework and probably provided some of the firmest estimates in the accounts as a whole. Without a previously derived input-output table, it was necessary to construct matrices showing the intermediate requirements of commodities by activities (absorption matrix) and the domestic supply of commodities by activities (make matrix) from the available evidence. This included detailed statistics of the commodity inputs and outputs of manufacturing sectors, together with much detail on commodity imports.' 8 At the present stage of development, Swazi- land has a very simple commodity output mix, so the buildup of an absorption matrix took the form of commencing with the structure based on imports, and then allocating the domestic supplies along its rows. The whole operation was tentative because total intermediate inputs, obtained by netting value added from gross outputs, provided a set of constraints. Since imports of consumer goods and capital goods were distinguishable, the only major problem was iden- tifying the stock elements of each purchase. As in the Sri Lanka case study, the vector of increases in stocks tended to be derived as a residual of unallocable items. The detailed procedure for allocating domestic supplies by both sector and use followed a theme similar to that of the Sri Lanka study. That is, many commodities could be readily identified as being produced essentially for final use (usually as exports) or as part of particular process loops. This considerably aided what otherwise might have been a formidable task. Our experience also showed that a high degree of commodity detail helped us identify the using sectors more easily. Swaziland has not carried out a household income-expenditure survey comparable to that of Sri Lanka. Consequently it proved impossible to derive disaggregations of household expenditure on commodities, except for an overall urban/rural distinction and separate treatment of high- income individual tenure farmers. Even this disaggregation was only possible by "borrowing" a set of expenditure coefficients from a rural household expenditure survey undertaken in nearby Lesotho. A further consequence of this lacuna is that no detail could be obtained on the savings propensities of the various household groups, although corporate savings and consol- idated government savings were defined more explicitly. Thus the Swaziland study serves not only to endorse the advantages of an approach that starts with commodity balances, but also to underline the importance of multipurpose household surveys in seeking to obtain an integrated set of accounts. As it is, the table was not completed. Nonetheless, it is interesting to note that the table was available before the national accounts for the same year. A final point on Swaziland is that the data base has been used subsequently for some model work. This use arose from the need to consider some specific investments, which were nonmar- ginal to the economy. The evaluation took the form of a project appraisal based on a macro- economic model. This was new ground which, experience suggests, was well prepared by the SAM approach. CONCLUDING COMMENTS Our work has raised a number of issues that are barely touched in the preceding discussion but that we would like to suggest as avenues for future work. We begin by summarizing the main points of that discussion. 18. The need for import data arose mainly because of the need to estimate Swaziland's revenue entitlement from the Southern African Customs Union. Planning SAMsfor Development 67 First, the SAM approach has proved in our experience to be a practical working tool of considerable merit in making the best use of available data and in providing a quantitative basis for analysis. It inevitably involves using data of variable qualities and calls for skills in data reconciliation which have not required the same emphasis in the past. It would undoubtedly be of value (and also of comfort) to have available formal techniques for pooling data. Meanwhile the informal methods must suffice, and the exercise of reconciliation gives a valuable focus to discussion of statistical priorities. Second, the SNArecommendation that SAMs should be approached through commodity balances has served us well. Furthermore, the refinement of having separate commodity and activity accounts is valuable both for implementation of a SAMand as an aspect of subsequent modeling. Next, we have concluded not only that it is possible to disaggregate the household sector, and hence to build income distribution into the macroeconomic picture, but also that it is desirable. At one level this is simply a matter of classifications-in this case, of institutions. But in taking the step from national accounts to a SAM some extra effort is obviously needed. At the same time policymakers are concerned about income distribution, and considerable effort is therefore going into data collection in this field. In our approach there is no conflict between the two competing claims: the extra costs of bringing income distribution into the major macroeconomic statistical picture are relatively small, and there seems to be a much wider interest in the product when households are disaggregated, rather than treated as a single sector. Our Swaziland study makes the fairly obvious point that it is not easy to include a disaggre- gated household sector unless a multipurpose household survey-covering income received as well as expenditures-is available. Even then, the problems of data reconciliation are consid- erable in our experience, and this is confirmed by the work of Altimir (1975). Do the problems exposed imply margins of error hitherto unsuspected in survey research or in national accounts? Regardless of the answer, the interests of better data are well served by the discipline of trying to reconcile household surveys and national accounts. It is not entirely adequate to resolve this question in favor of national accounts data on the grounds that the savings behavior measured by household surveys often implies that the rich dissave. In none of the case studies that we have conducted is there even approximate empirical support for the logical certainty that savings equals investment. This implied inaccuracy of data is not a trivial matter. As Ahluwalia and Chenery have emphasized (1974), savings behavior plays a crucial role in both growth and redistribution. There is no escaping the fact that this sensitive area has hardly begun to be charted by statisticians in developing countries. Mean- while, since policy never waits, a SAM approach at least forces guesses to be consistent with whatever macroeconomic data are more precisely known. While we realize that reference to guesstimates is unpopular in government statistical circles, the need to accept them as a part of macroeconomic statistics is unavoidable. This point goes beyond the early arguments in favor of using data from all sources to calibrate a SAMframework. As applications of our Swaziland study demonstrate, economic planning in developing countries is largely about structural change. Our view of a SAMis as concerned with the picture of future economies that might exist as it is with the initial position in which any particular economy might be. Accordingly, if statistical effort is to focus on reducing the standard errors on forecasts relating to policy alternatives, it is not at all clear that scarce resources should be devoted to more accurate estimation of the historical position. To us, not least of the virtues of the SAM approach is its ability to make the best use of those primary sources that might happen to exist. If these need to be filled out for the time being by guesstimates, there is nothing new in doing so that is attributable to the SNA except, perhaps, the relevance of the statisticians' work to the policy model. None of this is intended to detract from the importance of good basic data. The fact is that the SAM framework is not just a statistical tool: it is also a framework for economic analysis. The essential point, therefore, is that SAMs are not the preserve of the statistician but a potential 68 of SocialAccounting The Methodology bond in common with the economist. This potential, then, is the fullest implication of our earlier suggestion that the heart of the SNA is an economic model. Appreciation of SNA origins in the Cambridge growth model makes this point rather obvious. One further point which should be exposed in the present context is the importance of classifications. It can be rehearsed in relation to production activities, although it extends throughout the SAM framework. The literature of development has always seen duality in production techniques as an essential element of economics. More recently the question of vintage of technology has been found to be a powerful element of economic theorizing. We have already referred to the link between development planning and structural change. Is it not plausible, then, that this technological dimension of production units is just as important as the goods they produce? Indeed the SNAalready recognizes that separate commodity and activity accounts are needed. Once this separation is accepted it is simply inefficient not to ask what the most informative classification of production might be. An answer in terms of principal products does not seem to be self-evident and requires some justification. Making the best use of data by choosing appropriate classifications is also important and provides another avenue for enhancing the value of what limited resources are capable of producing. In conclusion, it should be emphasized that we do not consider any of the three SAMs discussed here to represent a best data framework for the countries in question or even the best use of the available data. And there are important omissions from the discussion, such as the treatment of imputed transactions, and the virtues of trying to obtain complete data for a SAMgiven that this will involve time and effort in estimation of some details that may be essentially irrelevant. With respect to imputed transactions, the narrow answer is that we have simply followed national conventions throughout, since our concern has been to fill out existing national accounts rather than to produce new figures. But the broader answer, and the answer to questions concerning the best SAM design for a particular country, is that such questions cannot be answered without reference to a model: only a model of the economy can define the correct basis for imputation, facilitate the distinction between what is important detail and what is not, and suggest the best data system to serve the needs of policy and planning. Thus in our view questions concerning the design and implementation of a data system cannot be divorced from the model such systems are intended to serve. We would prefer such models to be explicit, but that may not be essential, and a data system may need to serve more than one model. Accordingly, there may be disagreement over what is relevant detail. But, meanwhile, we do not see model construction as the primary task, even though the model (or models) is ultimately preeminent. In our view progress is to be made by iterative-or better, simultaneous-attention to a priori or model considerations, on the one hand, and empirical measurement and calibration, on the other. Enough has been written in the literature of development economics on the importance of institutional structure and dual- ities to justify the view that an examination of data systems in the light of such considerations may be timely. And if this point is not conceded, then it must surely be agreed that recent concern for distributional issues is sufficient justification for an attempt to measure some aspects of this dimension of an economy consistent with other continuing concerns, such as the balance of payments or the rate of investment. Although we lack a full articulated model of how all these different dimensions come into play in determining the actual path of development, we know enough to be sure that consistency is not, of itself, enough and that an integrated picture of interdependence in the different dimensions is required. Hence we have attempted, on the empirical side, to integrate detailed accounts for factors and households into an otherwise conventional SAM framework. Within this framework we have views about preferred classifications, which have been touched on in the text. But our empirical work has been circumscribed on two counts. First, it has been necessary to work largely with secondary sources that tabulate data on the basis of classifi- SAMs for Development Planning 69 cations in current use. These may or may not be ideal, which points to the second limitation, namely, the lack of a model to resolve such outstanding issues. Differences in the existing data base, as well as the effects of learning by doing, explain the differences in the three country studies reported here. We have not yet reached the point of wanting to prescribe what data ought to be collected in the future and how it ought to be arranged. Thus potential conflicts between country data systems and international standards for comparability are not resolved in our perspective. Simply, this paper attempts to set out details of the directions in which our research has been leading and to demonstrate the empirical feasibility of going beyond existing national accounts in three specific cases toward something more interesting and useful for policy purposes. 3 The Flow of Funds as a Tool of Analysis in Developing Countries Alan R. Roe The preparation of relatively detailed flow-of-funds accounts for developing countries is becom- ing common. Long series of such accounts are now available for India, Korea, and other coun- tries, and partial exercises are available for several more. This comprehensive format of financial information is being used both in descriptive work on financial questions and in the calculation of financial implications of development policies. The work, however, is often compartmentalized and kept substantially separate from statistical and analytical work on the real economy. We begin with a few remarks about the methodology of constructing flow-of-funds accounts and about the narrowly statistical advantages of their integration with SAM data proper. Our main purpose is to explore the possible analytical uses of flow-of-funds data for developing countries and to form a view about how well the existing data serve these functions. A central conclusion is that substantial improvements are needed in both sectorization and disaggre- gation of claims, in order to obtain quantitative insights about the role of finance in the devel- opment process. THE STATISTICAL FORMAT A statistical format for the flow of funds, which is most general and corresponds closely with that used in the SNA,is set out in table 3.1. The format is that used in Stone (1966). Note that subscripts refer to the order of matrices and vectors shown as entries in the table. This structure has two elements that link it back to the real economy in a purely statistical sense, namely, the savings vector (including the balance of payments deficit) and the real investment vector. This linkage involves certain difficulties. For example, the sectoral savings of table 3.1 ought to include the capital gains/losses of each sector even though these would normally be excluded from the national income accounts. Similarly, the real investment of table 3.1 ought to include the second-hand purchases/sales of each sector even though these would normally be excluded from the national accounts. If these problems can be overcome, the structure of table 3.1 provides an additional set of accounting identities that can contribute a good deal to the identification and elimination of inconsistencies between different data sources. For example, from the explicit identities of table 3.1, we can derive the following: (3.1) si = a! - Ljkik, that is, sectoral saving is identically equal to the total of assets acquired by the sector less the liabilities issued, and (3.2) zi = li - that is, sectoral real investment is identically equal to the total sources of each sector less the financial assets acquired. Since, in all SAM exercises reported so far, the estimation of sectoral saving as the residual of income and expenditure accounts is recognized as an extremely inaccurate procedure, the 70 The Flow of Funds as a Tool of Analysis 71 Table 3.1. The Structure of the Flow-of-Funds System Financial Cla ims Total Savings Institutions (liabilities) Liabilities Reall| ll Investment zj l l Institutions sj l I Ljk kik Financial l l l l Claims Ajk akj (assets) l ll Total aj =Zj + Ajkik lk Note: aj =l ik=ak sj = the vector of savings of the j sectors Z, = the vector of real investment of the j sectors L:k = the matrix of new issues of liabilities (of the k types) by the j sectors Ajk = the matrix of financial assets (of the k types) purchased by the j sectors aj = total asset purchases (real and financial) of the j sectors (that is, total uses) lj = total financial issues plus savings (that is, total sources) lk = total issues of claims of the k types ak = total purchases of claims of the k types. supplementary information provided by equation (3.1) and the associated financial data are clearly of potential value. In the Overseas Development Ministry and University of Warwick social accounting exercise in Botswana, for example, the flow-of-funds data, while themselves full of difficulties, proved of value in identifying certain initiVial errors in the estimation of items in the income and expenditure accounts of the SAM and in nroviding a subsequent basis for an upward adjustment of household income (see chapter 7 '. This advantage to appending a flow- of-funds element to the SAM structure illustrates a gen,.;:i. proposition: the more identity constraints that can be brought to bear upon a particular ement of the structure, -.-. the smaller will become the confidence interval surrounding the estimation of that element. The obvious qualification that needs to be attached to this proposition concerns the accuracy with which the additional constraint can itself be specifie4i. More specifically, we have to ask whether the flow of funds for a "typical" developing country can be estimated with sufficient accuracy to service the functions that our preliminary estimates discharged in the Botswana exercise. The answer is probably country specific, but a few remarks about the situation as it was encountered in Botswana may nevertheless be of interest. Botswana is a small and institutionally simple economy, and it was therefore a relatively straightforward task to extract ful balance sheet information for all of the public corporations, 72 of SocialAccounting The Methodology banks, other financial institutions, andmajor mining corporations. First differencing of adjacent balance sheets (in the absence of any significant asset revaluations) yielded information about the financial flows of our 1974-75 data period. The second stage of the exercise involved identification of the other party to all financial transactions taken directly from balance sheets. This was possible by virtue of (a) direct information (for example, the balance sheets of the banks specify the sector to which loans are made and from whom deposits are received) and (b) some knowledge of individual sector-to-sector relationships (for example, the public corpo- rations only receive loans from central government). In a residual, but small, number of cases, the second party to each directly observed transaction had to be deduced by a combination of assumptions and guesswork. The third stage of the estimation involved completion of sectoral accounts for households, including unincorporated enterprises; nonmining and nonpublic corporations; and the central government. The procedure simply entailed identifying those transactions that were obviously important but had not been captured by the previous two stages and then constructing the best possible estimate of their magnitude. For the central government this procedure was relatively straightforward. For households and nonmining and nonpublic corporations it was extremely speculative. The overall impression was that, since the first two stages of this procedure couldbe completed with a reasonably high degree of accuracy, and since the really difficult residual items were relatively few in number, our final estimates of sectoral saving as deduced through equation (3.1) were probably more accurate for some sectors than estimates derived from income and expenditure accounts. The method we have outlined seems, in broad terms, to be the one used in flow-of-funds estimation in other country contexts (see Venkatachalam and Sarma, 1977), with households, unincorporated enterprises, and all but the very large corporate enterprises generating all the serious estimation difficulties. It can easily be replicated in countries for which no flow-of- funds data currently exist and can be done at relatively modest cost. However, it is a methodology clearly capable of providing statistical feedback to the rest of the SAM only in the simpler economies. In more complex cases, the residual estimation problems associated with the flow- of-funds system may not permit it to define narrow, and therefore useful, constraints on the estimation of the rest of the SAM. The final point on the purely statistical side concerns the layout of the flow-of-funds infor- mation. There are two possible alternatives to what might be termed the sector-claim format of table 3.1. The first is the sources-uses table in which both sector and claim detail is retained but in columnar form (a sources and uses column for each sector), rather than in the A and L matrices of table 3.1. The sources-uses approach essentially rearranges the A and L matrices of table 3.1 in such a way that any sale of an asset, or acquisition of a liability, by a sector is shown in the sources column for that sector, while any purchase of an asset or repayment of a liability is shown in its uses column. In effect, both the set of sources columns and the set of uses columns are matrices (having the dimensions sectors times type of claim), but the normal practice is to arrange the sources and uses columns of each sector next to each other so that these two matrices are interwoven into one. The advantage of this procedure is that the full picture of any sector's capital transactions is easily seen. The main disadvantage is that some of the interesting aggregates that are obtain- able from the format shown in table 3.1 (for example, a, and lj) are lost. However, since both formats cross-classify the available information by sector and type of claim, one can normally be constructed from the other with minimal difficulty. The third approach, in contrast, completely suppresses the financial claims dimension of the presentation and shows a matrix of lender-borrower flows; the lending sectors are shown as columns of the matrix and the borrowing sector as rows. (A good example of this form of The Flowof Fundsas a Toolof Analysis 73 presentation is given in Bhatt, 1969.) Thus, if a sector issues, say, three types of financial claim--currency, securities, and savings bonds in the case of government-then any sector acquiring any one or more of these would be shown as lending to government the amount of the acquisition. No detail regarding the type of claim through which this lending was channeled would be provided. In the Botswana case, we were able to demonstrate that the sector/claim format could be translated into a lender/borrower format with little difficulty since most claims in issue were associated with just one sector. Thus, for example, the knowledge that the house- hold sector was accumulating rand currency in 1974-75 indicated that the sector was, in effect, lending to the rest of the world. The snags with the lender/borrower format are, first, that it cannot be used to infer anything about the channels and instruments through which these funds are flowing and therefore about the manner in which the flows might be influenced by changes in the conditions attached to certain instruments. Second, this format provides no link whatsoever with sectoral portfolio structure and with the sectoral behavior patterns that influence flows of funds. Any analysis based on it has to be largely mechanistic. Finally, it is impossible to move from it to either of the two alternatives. PREVAILING ANALYTICAL USES Although pioneering work on flow-of-funds data began nearly four decades ago, far more effort is still devoted to the data than to the analysis. However, analytical uses are becoming relatively common in the developing countries, and in this section we focus on three of the most important. First, comprehensive flow-of-funds data are increasingly used as the framework for historical descriptions of the evolution of an economy's financial system and for comparative analysis of financial developments in different countries. (An IBRD financial sector mission to the Ivory Coast, for example, has used the flow of funds as its descriptive framework.) It is, of course, interesting to know something about differences in the rate of creation of financial assets either through time or across countries, about the changing relative importance of different instru- ments in the total of financial assets, and about the manner in which a given structure of real investments was financed. In organizing these and other pieces of financial information, histor- ical flow-of-funds data serve an extremely useful purpose. What is invariably lacking in such studies, however, is a consistent theory concerning the manner in which the financial events have interacted with the production, consumption, and income creation in the real economy. In particular, such studies rarely provide insights into how production, levels of income, and other real economic variables would have been affected if the financial flows had evolved differently.' Lacking the theory to make connections for historical periods, 2 descriptive flow-of-funds work also provides relatively little policy guidance for those who would attempt to intervene in the evolution of the financial systems of the developing countries. The second common application is in the broad area of financial planning, particularly in 1. There are, of course, notable exceptions to this as a general statement about the analysis of the contribution of the financial sector to economic development. See, for example, Patrick (1966). 2. This statement should not be interpreted to mean that no attempts have been made to establish these connections. Goldsmith, in particular, has done prodigious amounts of work in this area but without arriving at any really general conclusions. For example, his book, Financial Structure and Development, ends with the statement "... The question [does finance make a difference?] certainly cannot be settled before the theory of finance is developed much further in the direction of analyzing the process of financial development and its relation to economic growth in operational testable terms and before we possess a substantial number of intensive case studies for different representative countries and periods that use the framework of such a financial theory" (Goldsmith, 1969). 74 The Methodology of Social Accounting calculations of the sectoral financing implications of medium-term development plans. This work has traditionally been carried out on a rather ad hoc basis with particular attention directed at the manner of financing government deficits. There is, however, an increasing interest in a more comprehensive and a more formal approach to the problem. A good example of this approach is work conducted by Venkatachalam, Bhat, and others at the Reserve Bank of India (see, for example, Bhat, 1972). Their procedure develops the original idea of Stone (1966), in which an exogenous vector of sectoral investment is applied to coef- ficient matrices constructed from the A and L matrices of table 3.1. Assuming constancy of all elements of the coefficient matrices and enforcing all the identities implicit in table 3.1, Stone's procedure generates the vector of sectoral saving that is consistent with the exogenous invest- ment and the portfolio composition indicated by the coefficient matrices. The interesting elab- oration injected by the researchers of the Reserve Bank of India is estimation of marginal portfolio coefficients based on time trends and their use, rather than the average coefficients, in the coefficient versions of the A and L matrices. Applying the procedure to the investment targets of the Fourth Indian Plan, they compute consistent sectoral savings figures, as well as the implied increase in the amounts outstanding of each major financial claim. By comparing their derived savings vector with the target savings vector from the plan, they are able to identify the trends that are necessary to ensure the attainment of these targets. A similar procedure, based on a lender-borrower format of the flow-of-funds system, is described in the papers by Bhatt (1969) and Divatia (1969). Divatia's paper is more formal and takes as constant the coefficients constructed by relating the elements of the lender/borrower matrix to their row totals (or column totals). With either sectoral saving (or investment) taken as exogenous, he generates the consistent sectoral pattern of investment (or saving). Bhatt's approach is similar in structure, but he incorporates additional information to avoid the assumption that the coefficients of the lender/borrower matrix will remain constant at their historically estimated values. These papers are valuable in identifying possible inconsistencies between the sectoral savings and investment targets of the Indian plan, which they analyze. Indeed, the work was used to modify the initial estimates of the investment targets of that plan. A similar exercise, but for another continent, is described in a paper by Bhatia and Engstrom (1972) on the Second Nigerian Plan. These papers are valuable in extending the consistency analysis of development plans to their financial dimensions. There is still a need for more explicit procedures whereby inconsistencies, once uncovered, can be eliminated. There is little doubt, however, that the comprehensive flow- of-funds approach to the problem represents a considerable improvement over the partial equilibrium approaches emphasizing government. These first two analytical uses of flow-of-funds information potentially involve the full gamut of financial claim and sectoral detail. A third example is rather different in that it draws on certain key elements of the flow-of-funds picture. Werefer to the burgeoning literature concerned with the empirical formulation for developing countries of monetary models of the balance of payments. In this literature, the central sectoral flows are those of the commercial and central banks shown in simplified form in table 3.2. Consolidation of these two flow accounts, the netting out of central bank borrowing from government (that is, government deposits) against its loans to government, and the netting out of the currency held by the banks against amounts issued, yields the following identity: (3.3) A Currency held outside banks + A Commercial bank deposits = A Loans and advances to government + A Loans and advances to private sector + A Foreign reserves which is equivalent to: (3.4) A Money supply = /v Domestic credit + A Foreign reserves. The Flow of Funds as a Tool of Analysis 75 Table 3.2. Flow of Funds of Commercial and Central Banks A. Commercial banks Changes in Liabilities Changes in Assets Private Deposits Currency Government Deposits Balances with Central Banks Loans and Advances to Government and Private Sector Total = X Total = X B. Central bank Changes in Liabilities Changes in Assets Currency in Circulation Loans and Advances to Government Government Deposits Loans and Advances to Private Balances of Commercial Banks Sector Foreign Reserves Total = X Total = X2 Since the majority of developing countries are highly open to trade, the change in the money supply is taken to be strongly endogenous in the sense that an expansion in domestic credit, to the extent that it stimulates aggregate expenditure, will generate an increase in imports and a fall in reserves. Thus the two terms on the right side of (3.4) will partly offset each other and so reduce the final money supply expansion associated with any given initial change. The recent monetary models for the developing countries have mostly been concerned with speci- fying, elaborating, and testing this basic insight in order to form a view as to the nature of the interaction between monetary and credit factors, on the one hand, and income levels, prices, and the balance of payments, on the other. The original model along these lines for developing countries was developed by Polak (1957) a*adinvolved the endogeneity of nominal income, imports, and the money supply. The income velocity of circulation of money was rigidly constant in the classical manner; exports, capital inflows, and domestic credit were all exogenous; and real income was held constant through a full employment assumption. Subsequent work has introduced greater endogeneity into the model by relaxing many of the rigid classical assumptions and thus providing fuller insights into the division of monetary effects between real and inflationary consequences. A brief summary of the additions to the basic Polak structure introduced by five papers in this area is given in table 3.3. A couple of examples will suffice to indicate the sort of insights that these models can provide. Thble 3.3. Summary of Characteristics of Five Macromonetary Models Short-term Some Links Supply De- Government: Price Price Capital Rate of Rate of from Nominal Inflation termination Private Sensitivity Sensitivity Flows Interest Interest to Real Rate Explicitly Sectori- of Imports of Exports Endogenized Involved Endogenized Magnitudes Endogenous Specified zation Aghevli (1977) (a) (b) INDONESIA v v x v x x v x V Khan (1974) VENEZUELA X x _ _ x x Aghevli and Khan (1977) Whole of balance of payments X X X I X v INDONESIA is exogenous Otani and (c) Park (1976) v X X v X V v X KOREA Otani (1975) (e) (d) PHILIPPINES V X L K K x x x Notes: J = a characteristic that is included. X = a characteristic that is not included. a. Applies to non-oil exports only. b. Sectorization only in the sense that government revenues and expenditures appear explicitly. c. Nonprimary sector only; the primary sector is exogenous. d. A trend value of supply confronts monetarily determined demand to generate price and inflation effects. e. Real income is endogenous. The Flow of Funds as a Tool of Analysis 77 In a joint paper by Aghevli and Khan (1977) on Indonesia, the authors argue that inflation itself feeds the growth of the money supply (as well as being fed by it) by causing government expenditure to rise faster than revenues, thereby generating the need for increasing amounts of deficit financing. To focus attention on this central point they simplify the basic Polak structLure by regarding the balance of payments as exogenous, so that the only endogenous influence on the money supply is the government deficit. Given the exogeneity of real income and the absence of any interest rate term in the demand for money function, an initial rise in the money supply can only be matched by a change in demand to the extent that the domestic price level is increased (thereby reducing the excess of real money balances). However, the rise in price stimulates a larger increase in nominal government expenditure than taxes and increases both the government deficit and the monetary base. That generates a second round of monetary expansion, the effects of which are exacerbated by the adaptive adjustment of inflationary expectations and the resulting fall in the equilibrium holdings of real money balances. Similarly, in the Otani and Park (1976) model for Korea, the initial response to a monetary expansion is a rise in the domestic price level to clear the money market. However, in their model this response stimulates imports, which also appear as an explicit input in the production function of the nonagricultural sector. Thus the real output of that sector rises. This rise causes an increase in the nominal demand for money which moderates the earlier price increase, as does the reduction in the money supply associated with the deteriorating balance of payments. Thus output, monetary variables, prices, and the balance of payments are inextricably woven together. The models just described are not flow-of-funds models in the conventional sense; they do not involve a full specification of the intersectoral flow of funds and do not enforce all the financial accounting identities implicit in table 3.1. Thus they commit some of the sins of implicit specification discussed at length by Brainard and Tobin ( 1968). The models can also be criticized because the strong link between aggregate demand and imports (that is, the openness assump- tion), the central element in most of them, is of dubious validity when the import demand of so many "open" developing economies is in fact heavily dictated by direct central controls.3 Finally, their presumption that there is a real possibility for developing countries to have "pure" monetary policy-that is, a change in the money supply that does not require changes in government expenditure or taxation-either flies in the face of reality or misleadingly applies the label "monetary" to something that, with equal validity, might be regarded as "fiscal." THE ROLE OF THE FINANCIAL SYSTEM IN ECONOMIC DEVELOPMENT Clearly, all the applications we have just discussed are extremely useful in their own different ways, and it would be quite wrong to argue otherwise. Of themselves they justify the continuing effort to produce flow-of-funds data for a wider range of countries and to use the classifications and the methodologies already established. However, if we take either the view that the financial sector is a potentially important and independent source of dynamic in a country's economic development or the view that inappropriate financial policies can, of themselves, damage a dynamic established by innovation in some other sector, then a somewhat more critical stand- point might emerge. In the rest of the paper we explore the implications of these viewpoints and conclude that if we wish to quantitatively analyze the innovative (or repressive) influences of financial policies, then we need to think harder about what sort of flow-of-funds data will be most appropriate. We begin with the view that the financial system of a developing country at any time is a 3. I am grateful to Wliam Dellalfar for this point. 78 of SocialAccounting The Methodology part of its inherited technology in the same way that the production systems of its factories are part of that technology. The financial technology services two essential functions, namely, the facilitation of the movement of funds from surplus to deficit sectors and the provision of incentives to stimulate future as opposed to present consumption and so help move society closer to some "optimum" mix of these two quantities. In other words, its functions are to stimulate and allocate savings. Now it is clear that the financial technology of a typical devel- oping country services these two functions extremely imperfectly. This imperfection is not so much because the resources available to existing financial institutions do not work to good effect; rather it reflects certain barriers to the expansion of these institutions to anything like an optimal scale of operation. Furthermore, these barriers may be inherent in the nature of a developing economy or erected by some form of government intervention. If this canbe accepted as a startingpoint,we can easilymove onto two importantpropositions. The first is that beneficial developments in the real economy (for example, a technological breakthrough in agricultural production) may be completely frustrated or have their benefits moderated by imperfections of the financial markets. The second is that innovations in the financial markets, to the extent that they reduce the inefficiencies of these markets, can generate independent benefits in the real economy. In either case, the financial system is important to understanding the behavior of the real economy. But the nature of that importance can only be understood by first acquiring a clear comprehension of the causes and consequences of the imperfections of the financial markets and the manner in which these might be overcome. (This point is clearly ignored by most of the literature that attempts to incorporate a monetary asset into standard growth models.) Most sources of capital market imperfections in developing countries can be summarized under two main headings: interest rate restrictions (including restrictions on the terms of lending and borrowing other than the interest rate) and transaction costs. The first speaks for itself, but the second needs elucidation. Financial transactions, by definition, involve the intercession of time (for example, time between borrowing and repayment dates), as well as the contractual interlinking of persons unknown to each other to varying degrees. Thus both uncertainty and imperfect information arise to drive a wedge between the demand price of real capital assets and the price at which funds can be borrowed to finance their acquisition. This is in addition to the wedge between these two prices arising from the operating costs, including profits, of the financial institutions. Since imperfect information only arises as a problem in an economy having transaction costs (otherwise borrowers and lenders could spend infinite time in gathering the information to make correct borrowing and lending decisions), we can theoretically represent the effect of any given imperfection of information by the amount of transaction costs needed to overcome it. Similarly, uncertainty can be represented as a transaction cost (for example, the uncertainty of the lender might be represented as his evaluation of the probability of default of the borrower, multiplied by the amount of the proposed loan). Finally, recognizing that there are transaction costs in the broad sense incurred by both borrowers and lenders, we can construct a catalog of elements that form total transaction costs, and thereby contribute to capital market imper- fections. The elements are as follows: Borrower (or demand) side of the market 1. The administrative costs and opportunity costs of the borrower (for example, traveling time to geographically distant banks). 2. The borrower's risk as evaluated by the borrower, that is, the difference between the expected marginal efficiency of an investment project and the rate at which the borrower is prepared to borrow funds to finance it. The Flow of Funds as a Tool of Analysis 79 Figure 3.1. The Effects of Transaction Costs Interest rates and yields D 2, \ /Total supply Rnvestment, borrowing, and lending Lender (or supply) side of the market 3. The administrative costs of the lending institution (for example, wages and profits). 4. The interest premium that the lender requires as compensation for his evaluation (correct or otherwise) of the riskiness of the loan or the absence of conventional security. The effects of these sources of inefficiency can be seen very simply in figure 3. 1.4 The demand curve D1D1 represents the marginal efficiency of capital schedule, while the first downward shift of this curve, D2 D2 , represents its adjustment for borrower transaction costs, and the final shift, to D3D3 , represents its adjustment for borrower and lender transaction costs. In the absence of transaction costs, the economy will attain an equilibrium at point A, with total lending equal to OS 1 and the borrowing and lending interest rate equated at R1. With allowance for transaction costs, the equilibrium moves to point B, total lending falls to OS 3, and the lending and borrowing rates of the financial institutions ("banks") diverge by the distance Bt,DB.Above all, the transaction costs impose a loss of surplus on the economy equal to the shaded area in the figure, in addition to certain d;ynamic losses associated with a lower level of real investment. Equally it is clear that a reduction in transaction costs associated with greater efficiency in capital market transactions in any of the four senses identified in the classifi.cation given above will provide the economy with real gains in income and output. 4. In this exposition, we are ignoring the possible effects of monopoly power in the hands of the financial institutions. Such a monopoly would have effects similar to those of the four sources of inefficiency that we do consider. 80 of SocialAccounting The Methodology This much is clear and probably fairly obvious. The story becomes more interesting when we allow for some sectorization based upon the differential transaction costs of borrowing and lending of different sectors. For purposes of exposition imagine a "traditional" sector that has relatively high transaction costs and a "modern" sector that has relatively low transaction costs. Through an extension of figure 3.1, it is a relatively simple matter to demonstrate the following propositions (see Roe, 1978): 1. In the absence of transaction costs the marginal efficiency of capital is equated in the two sectors. 2. Transaction costs impose a larger fall of lending, and of surplus, on the traditional sector than on the modern. 3. If the banks are allowed to discriminate by price, transaction costs will result in their charging a higher interest rate to the traditional sector than to the modern. 4. If price discrimination is not allowed, the same effect will probably be achieved by imposing more stringent nonprice conditions on lending to the traditional sector. In short, as well as imposing an aggregate economywide reduction in welfare, transaction costs will cause a disproportionate reduction in lending to the traditional sector and conse- quently a reduction in that sector's investment, its income, and the output that it produces. Thus, just as economywide analysis would clearly wish to specify the effects of taxes and the effects of removal of taxes having uneven incidence, so ought such an analysis to specify the effects of differential sectoral transaction costs in financial intermediation. There is no theo- retical reason to expect these effects to be trivial or sectorally neutral. The second source of capital market inefficiency, namely, the administering of the interest rate, is also easily analyzed using figure 3.1. Assume that both lending and borrowing rates are moving the quasi equilibrium of the economy to B' when there are transaction fixed at a level RPa, costs and to A' when there are not. In fact, this "equilibrium" will be characterized by excess demand of ED' when there are transaction costs, and ED when there are not. Since the excess of ED over ED' depends only on transaction costs, it is clear that in this situation the salutary effects of an improvement in the financial technology will lead to an increase in excess demand only and not to any improvement in social welfare. In this same situation, any upward shift in the schedule of the marginal efficiency of capital (because of a green revolution, for example) is likely to have its beneficial effects on the economy moderated by the inelasticity of the supply of funds with respect to the new lending opportunities. Thus, whether a technological improve- ment is in the financial sphere or elsewhere, one condition for its full benefits to be realized would seem to be the absence of any initial excess demand in financial markets. Two final issues need to be considered before we attempt to draw out the data implications of this line of reasoning. They both concern the nature of the supply schedule of figure 3.1. If we conceive of this schedule as representing the supply of funds to the formal (or organized) financial markets, then our analysis rather begs the question of the role of the informal (or unorganized) financial markets of developing countries. We take the view that these markets have two fundamental distinguishing characteristics. First, their transaction costs of borrowing and lending are far lower than those of formal sector institutions serving the same clientele. This difference is partly because they lack expensive infrastructure, but largely because they operate on a narrow local basis and therefore have far lower costs of information. The second characteristic is that they are successful, either by law or by practice, in avoiding administrative restraints on lending and borrowing rates. In short, the informal institutions are distinguished by their comparative advantage in relation to both sources of capital market inefficiency referred to earlier. What this means in practical terms can be seen by referring back to figure 3.1. In the case in which the interest rate is administered at R,, the informal sector, to the extent that it can The Flow of Fundsas a Toolof Analysis 81 tap into the part of the supply of funds denied to the formal sector by the interest rate restriction, will move the equilibrium of the economy away from point C and toward point B. It will thereby enable the economy to recapture some of the surplus lost because the interest rate is controlled. Furthermore, since the credit rationing associated with an administered interest rate will almost certainly impinge disproportionately on disadvantaged sectors (the traditional sector in our analysis), the intercession of the informal sector may moderate that distributional effect. However, although it provides these obvious aggregative and distributional benefits, the infor- mal sector may have such a large gap between its lending and borrowing rates that it appears exploitive. Similarly, if we consider an initial equilibrium at B, then the lower transaction costs of the informal sector will enable it to tap into the supply of funds to the right of B and move the equilibrium nearer to A. Again this confers an aggregate social benefit and provides certain distributional benefits in that the informal sector lending will undo part of the disadvantage of our traditional sector arising from differential transaction costs. The social functions of the unorganized financial markets are thus clear: they restore part of the social welfare lost by virtue of the two capital market imperfections referred to earlier, and they redistribute income in favor of the traditional sector. If they could discharge these second-best functions perfectly, then we would have no cause to concern ourselves with the imperfections of the formal markets. Unfortunately, the informal markets are themselves defec- tive, most seriously because they are heavily localized in their spheres of operation. This severely restricts their access to the economywide supply of funds, as well as their ability to allocate the funds they do raise in the most efficient manner. The second issue concerning the supply schedule of figure 3.1 relates to the distinction between aggregate savings and financialized savings (that is, savings mobilized through the financial system). The supply schedule of figure 3.1 would be controversial if it was interpreted as referring to aggregate saving as the supply concept, because the empirical evidence on the interest elasticity of saving is inconclusive (see Saito, 1977). We prefer, therefore, to interpret it as referring only to financialized saving, which implies that the upward slope of that schedule relies upon a diversion of saving away from nonfinancial assets. This diversion is far easier to demonstrate empirically than interest elasticity of saving (see Roe, 1978). The alternatives to putting savings into financial assets differ for different groups in an economy, but some of these alternatives will be directly productive of real income. Thus the reduction of social welfare associated with the shaded area of figure 3.1 is potentially overstated to the extent that the downward move along the supply schedule from A to B may permit increased direct investment of savings in tangible and productive form, and thus offset the reduced real investment associated with reduced financialized saving. At the risk of some over- simplification, the most important alternatives to the financialization of saving for different groups of a "typical" developing country are the following. 1. For rural households and small businesses: unproductive items such as precious metals, jewelry, and inventories of their own output in excess of consumption needs; productive items such as land, agricultural and handicraft tools, and livestock. 2. For urban households: real estate and various consumer durables. 3. For government: public sector investment projects and overseas investments. 4. For modern sector companies: corporate real capital and overseas investments. Even a casual inspection of this list reveals that some items will certainly not substitute for financialized saving in the sense that the income and output to which they will give rise will equal those arising from the real investments that financialized saving makes possible. Other items on the list are ambiguous in this respect. However, it follows from the earlier analysis that improvements in relation to either or both forms of capital market inefficiency will move 82 The Methodology of Social Accounting the economy upward along the supply schedule of figure 3.1. Whether this movement will produce the potential benefits referred to earlier depends on both the source from which the financialized savings are diverted and the output and income forgone by virtue of the reduced direct investment associated with that diversion. THE IMPLICATIONS FOR FLOW-OF-FUNDS DATA SYSTEMS The previous section has attempted to identify some aspects of the role that the financial system can play in the process of economic development. We also hope to have identified some of the critical changes in the flow of funds which accompany this development. In this section we attempt to identify some of the implications for data systems which would arise from attempts to do empirical work based on our essentially theoretical argument. We do not begin this task with an expectation that many of our suggestions will be implemented, at least not in the near future. We proceed rather with the view that it is legitimate for a coherent theory to define a data requirement which, if met, could significantly improve knowledge of how partic- ular financial changes might improve economic well-being. If we accept the proposition that finance is part of an economy's technology, we require classifications that reflect this proposition. In particular, we need a classification of borrowing institutions which reflects the institutions' differential advantages or disadvantages that arise from the transaction cost source of inefficiency. The analogy between these costs and a tax has already been made: we need to be able to identify any group that would receive a disproportionate advantage from the lowering of such a tax. The traditional/modern distinction in our example was not necessarily intended to imply an analogy with the traditional and modern sector of conventional usage. However, it is clear that in many developing countries the small-scale, traditionally organized businesses do suffer a transaction cost disadvantage, and it would seem appropriate, as an initial step, to try to separate this group for flow-of-funds purposes. To develop this classification, we need fuller documentation of the nature and magnitude of the transaction costs faced by each group in the classification. This is a difficult task since only some of these are tangible financial costs. As indicated in the classification presented earlier, certain of these costs involve subjective evaluation on the part of the borrower, the lender, or both and are not readily quantifiable. In practice, flow-of-funds estimation falls far short of the sectoral detail that the proposal of the previous paragraph implies. Indeed, it has often proved difficult to produce separate accounts for the household and the corporate part of the private sector, let alone any finer disaggregation of these two elements. Where separate household accounts are available they normally depend disproportionately on a residual approach to estimation and incorporate substantial elements of unincorporated business activity. So long as no finer sectorization can be achieved from the data, analyses based on those data seem certain to miss much of the fundamental importance of financial processes. A second point relates to the portfolio choices that flow-of-funds data reflect. In most appli- cations the ex post measures are wholly concerned with formal sector financial assets and have little to say about (a) informal sector assets and liabilities, and (b) tangible assets that appear in portfolios. Given our theoretical judgment as to the fundamental importance of the informal financial sector in a situation characterized by capital market imperfections, it is critically important that we have some information about how this sector intercedes in the borrowing and lending transactions of different groups. We need to know much more about the portfolios and the potential portfolio choices of those who lend to informal sector financial institutions and about the categories of transactors to which these funds are ultimately made available. Without such information we have no real hope of quantifying the consequences of the imper- The Flowof Fundsas a Tool of Analysis 83 fections in formal capital markets or the consequences of eliminating such imperfections. In particular, we have a quite inadequate basis for analyzing the effects of ending the administra- tion of interest rates at subequilibrium levels. Similarly, flow-of-funds information needs to give far more attention to the tangible assets that appear in the portfolios of the various sectors. If, as the argument of the previous section implied, the move along the supply schedule of figure 3.1 is merely a diversion of funds from tangible assets to financial assets, then it is important to know something about the nature of tangible assets in each sector and the scope for substitution away from them. In particular, we need some information about the quantities and types of tangible assets that are productive of output, relative to those that are not, in each sector. The standard flow-of-funds practice of incorporating only one catch-all category of tangible asset permits no analysis of this problem. Clearly nothing can be done about these particular recommendations unless the raw data are available in the appropriate form. Since this will normally not be the case, the first step toward implementing the recommendations is likely to be the elaboration of a survey program directed at households and small-scale businesses, to provide a systematic documentation of the financial aspects of their operations. This chapter has tried to pinpoint the questions that ought to be asked and the analytical uses to which the answers might be put. CONCLUSIONS Several of the arguments in this chapter are worth highlighting. First, there is a strong statistical case for fully integrating the flow-of-funds system within a SAM.While this approach is not likely to improve the overall accuracy of every SAM, it will do so in many cases. Second, our partial survey of analytical uses of flow-of-funds data suggests several established uses but notes two fundamental weaknesses as well: either the analyses fail to provide any explicit mechanism whereby the financial system can affect real economic events, or they fail to reflect any of the important distinguishing characteristics of the financial systems in the developing economies, or both. We have attempted to identify these distinguishing characteristics and, in that light, to exam- ine the possible beneficial effects of an improved financial system on the real economy. The view we take is that these effects are potentially large, both in aggregate social welfare and in distributional benefits. This is the third conclusion we wish to underline. The fourth and final conclusion is that data required to quantify the potential effects of financial improvement are not provided by the standard flow-of-funds tables available for devel- oping countries. This is largely due to the statistical difficulties of measuring these flows and implementing the sectorizations needed to properly analyze the effects of an evolving financial system. A clear conception of the importance of the financial system to economic growth is needed as a starting point for addressing these statistical problems. Some of the ideas in this chapter may be a small contribution toward the development of such a conception. 4 RegionalAccounts in a SAM Framework Graham Pyatt and Jeffery I. Round There have been many attempts to identify, as well as to solve, the special problems of social accounting at the regional level. To the extent that only a single region is involved, whether it is part of a nation (a subnational region) or a group of nations (a supranational region), the conceptual problems are not really very different from those arising at the national level. This is not to understate the practical problems, which are often severe and derive largely from the fact that regions do not enjoy the statistical advantages that are by-products of the existence of well-defined national boundaries, such as customs controls or currency areas. The point is rather that a multiregion system poses a different range of problems, and these extend beyond those normally encountered for a single region system. This paper will consider some extensions of the SAM framework as it applies to a two-region system. The particular study used for illustration concerns East and West Malaysia, but the conclusions are applicable to other regional systems. Not least, the approach to a two-region system can readily be generalized to the many- region case. By way of background, the east-west split of the Malaysian economy is one that arises quite naturally. The Federation contains twelve states, of which two-Sabah and Sarawak-make up East Malaysia. Historically, they are the most recent members of the federation, joining the other states in 1963; geographically, they are separated from the peninsular (West Malaysian) states by some 500 miles of sea. The regional distinction is easy to make statistically, since the sources for West Malaysia are independent and, incidentally, are also substantially better than those for East Malaysia, both in coverage and in quality of data. For example, national accounts have been estimated for West Malaysia for some years, while those for East Malaysia have, until relatively recently, been nonexistent. Our underlying concern has been to integrate the social accounts for Malaysia as a whole, while retaining the regional detail. In the first instance, this involves a disaggregation of the external transactions for each region into transactions between East and West Malaysia (interregional transactions) and transactions by each region with other parts of the world. Such disaggregation of external transactions is of some general interest. Many countries are already developing their balance of payments statistics toward showing separately the links with different international trading or financing partners, since there are obvious policy implications to be drawn from differences in the pattern of external transactions with different currency areas, countries in different stages of development, or different economic or political blocs. It should also be noted that the situation as it exists between the two regions of Malaysia is complicated by the position of Singapore as an entrepot for a variety of commercial activities. A good deal of merchandise trade between East and West Malaysia is still routed through Singapore, while its position as a financial and commercial center means that many nonmer- chandise transactions and monetary transfers are cleared through financial intermediaries in Singapore. In consequence, there are very real difficulties in distinguishing actual east-west flows as opposed to those that involve the intermediation of Singapore. In general, the existence of entrep6ts in regional systems is not unusual; in our particular case, we are perhaps fortunate that the entrepot is a single location and that the trade flows through it have previously been identified for us by the Malaysian Department of Statistics, Kuala Lumpur. 84 Regional Accounts 85 Before embarking on the intricacies of the regional accounting scheme we shall first briefly describe the structure of the Malaysian SAM.As already noted the estimation of separate accounts for West and East Malaysia actually preceded the derivation of a SAMfor the whole of Malaysia. Nevertheless, the aggregate All Malaysia SAM will provide a useful basis for describing the basic structure of the accounts upon which subsequent discussion will revolve. AN AGGREGATE MALAYSIA SAM IN OUTLINE The aggregate structure of the framework underlying the full Malaysia SAM is shown in table 4.1.1 Table 4.1 shows the SAM for Malaysia in schematic form: its entries simply describe the nature of the transactions that take place between the various accounts. With the same frame- work is shown our estimates of the aggregate transactions for Malaysia in 1970. The eleven sets of accounts contain seven broad groups of accounts as follows: wants (account 1), factors (account 2), domestic institutions current accounts (accounts 3-5), a consolidated capital account for domestic institutions (account 6), rest of the world (accounts 7 and 8), production (accounts 9 and 10), and indirect taxes (account 11). The wants accounts are designed to focus on the fact that development policy objectives ultimately reduce to the welfare of individuals (subject to the need to supply public goods to provide for the national interest), and that these are met by the provision of "wants" or needs. Ideally such wants might include items that go beyond those normally included within an estimate of private consumption expenditure in the national accounts. However, such impu- tations were not attempted, so the first row of table 4.1 has only one entry, showing receipts from the supply of wants defined as categories of private consumption expenditures to house- holds. The corresponding numerical entry in the table shows the aggregate level of consumer expenditure to be 7,528. (Amounts are all in millions of Malaysian dollars.) The second row and column of table 4.1 refer to a set of factor accounts which receive factor incomes (domestically generated by production activities and those received from abroad) along the rows and pay these out down the columns to domestic institutions (households, companies, and government) and abroad. The incomes of institutions can be seen to comprise factor and nonfactor (that is, transfer) receipts. As with the wants accounts, in the most disaggregated form of the SAMmany factor and household accounts are distinguished. In contrast, the current accounts for companies and government and the combined domestic institutions capital account are maintained in consolidated form throughout. The current and capital accounts for the rest of the world are shown as accounts 7 and 8. In accordance with normal practice, the balance of payments current account deficit, from the point of view of Malaysia, is shown as a transfer of 39 from current to capital account (that is, a rest of the world "saving" of 39). Overall balance in the rest of the world account for 1970 is achieved via net lending abroad of - 39 (that is, a reduction of assets held abroad by Malaysia of 39), which arises out of net disinvestment abroad of 124, plus net capital transfers from abroad of -85. Our treatment of the production accounts follows SNA practice in recognizing separate sets of accounts for commodities and activities. Although only aggregative flows relating to commod- ities and activities are shown in table 4.1, our most detailed SAM involves substantial disag- gregations of these accounts for fifty-nine commodities and thirty activities. One feature of the SAM accounts is especially worth noting given our concern in this paper for a proper recording of regional and international commodity trade and regional transactions in general. It relates 1. A more detailed summary description based on this table is presented in Chander and others (1980). Full detail is to be found in Pyatt and Round (1978). Table 4.1. A SAM for Malaysia, 1970 (in millions of Malaysian dollars) EXPENDITURES .. 12 3 | L 5 :onolidtl6 7 | 8 9 | 10 11 1 Institutions' Current Capital Rest of World Production Wants Factors Accounts &ceount for ___Indirect )omestic TOTAL Households Coopanies Covernent Institutcin Current Capital Commoditie Activities Taxes 11 Wants Want acqui- ~~~~~~~~~sitions by Total q aqii ant r' Wants households 7528 tioms 7i28 Factor in- alued adde, Total 2 _ Factors comesrecd payments to factor 2 FaIctors ~ from abroad 232 factors 10601 incomes 10833 Wae -Inter-house Govern=ant CurrentToa rprtdhodta ages Distributed t.far' c 3 C Households bnold tran- proditsntracpa r transfer t rasfers household v Hoshlbusiness in- ferr rft households from gibroadinoe come S696 476 109 196 16 __ _ 9493 Operating Governoent Current Total 4 4n C paessurpluses Companies fcopn transffts C maie1 transfers company comCpa1nty3 So f companie to co8 anie 2 from abroad incomes 1551 134 15 1700 Diet Direct Current Net Total S GCovernment axe; tae transfers indirect government 992 ~~~~~from abroad taxes Income 15 1802 3197 Household Company Governsent Capital Total Consolidated Capital transfers capital 6 Account for Domestic savings savings Savir.Fs from abroad receipts Institutions 1101 369 855 -85 2240 F5 ____ __ _Toa . .___ __.- u- Factor Household Company Government Tot aelur 7 r-urrent 40 ~~~incomes transfers transfers transfers Imports -aroadpymt X A-' 7 Current aid abroad abroad abroad abroad 4798 abroad __________ 586 230 27 84785641 > _ _ N~~ ~ ~~~~~~~~~ Balance of ~ ~~~~~~~~et _r __ otal capita L o aia ivsmt 'mnoPayments 8 n., Capital _ _ current eabroad def- abroad - e~~~~~~~~~~~~~~~~~~~~~ ~~~~~-124 icit 39 -85 of watsiO Coverreent roa Iiedntermediat Total fOantcurrent tiltal fot ivi t commodit 9 Coodities into coood- ton &stoco nt 1 ties expenditur raa 5324 demands demends 7528 TOTAL t 2364 564L fro_ 9311 26512 9 Domestic GrRn o0 0 Aciiiscosmmodity Activities ~~~~~~~~~~~~~~~~~~~~~~supplies output. 20403 20403 _____ I Jin~~~~~mooi-ty in- Non commom Total net I t~~~~~~-ract tazes ity Indi- indirset Ii indirect taxes leat of sub- rect razes taxes idies Ttal want Total Total house. lotal Total Total in- Tottarlcur- Total capi- Total Gross Toaot 12 ~~~TOTALnoe acquiai- fco pamns hldxpandi trs opn s.!iu Coverrlent expenditure vatat220 atrCeiPt rmabroad ital from abroad commodity receipt supples Impure inirc 7528 10833 43 1700 3197 24 5641 -85 26512 20403 taexs Source: Pyatt and Round (1978)- Regional Accounts 87 to the valuation of the commodity balances. In the Malaysia SAM commodity transactions are recorded at market prices. In the SNAthey are valued at basic prices. Such prices are essentially factor costs (that is, market prices net of indirect taxes on sales) less trade and transport margins on sales of output. However, for certain conceptual and practical reasons discussed elsewhere (see Pyatt and Round, 1978, and Chander and others, 1980), rather than follow this route a market price valuation is retained throughout. Its relevance to the issues addressed in this paper is obvious: the margins for trade and transportation include the costs of moving goods between producer and user locations. This means, for example, that the want acquisitions in row 1 of table 4.1 represent consumer expenditures of 7,528 at market prices. Much of the difference between this convention and the use of basic prices is not apparent at this aggregate level of the SAM, but it is seen more readily at a disaggregated level where accounts for trade and transportation services are separated from those of other commodities. The next section of this paper shows how the SAMframework can be applied to the regional accounts for Malaysia. The system will continue to be viewed in a fairly aggregate way. In a final section we shall focus more explicitly on the particular conceptual difficulties concerning the treatment of interregional commodity trade. AGGREGATE REGIONAL ACCOUNTS IN A SAM FRAMEWORK The development of a regional account system may be approached from two standpoints. One approach is to disaggregate a SAM for the economy, taken as a whole, into its constituent regional components. The alternative approach is to combine SAMs for two (or more) regions into an integrated system. As already indicated, our starting point has been separate SAMs for East and West Malaysia, which we have then articulated into a system for Malaysia as a whole. At the aggregate level of accounts, this combination of separate regional SAMs requires only one major conceptual step, namely, that the external accounts for each region have to be disaggregated so as to show interregional and true international transactions (and transfers) separately. As will be seen more clearly at the disaggregated level, this step involves problems beyond those of obtaining a consistent set of data from alternative initial estimates for each of the regions. There are several alternative formats for the presentation of regional accounts, corresponding to more or less exacting demands for primary data. Table 4.2 shows one arrangement, which is among the least demanding in its data requirements. The accounts within table 4.2 can be viewed as comprising three principal blocks. One block relates to East Malaysia; the second to West Malaysia; and the third to the rest of the world. Each of the blocks for East and West Malaysia contains nine accounts, which are the nine domestic accounts in the basic eleven-account SAM. The numerical ordering of the eleven- account SAM is retained, so that the external accounts for each region (accounts 7 and 8) are excluded. Transactions between each region and the rest of the world are shown in accounts 7* and 8* which are labeled as such in recognition of the fact that they are true international transactions and exclude those that take place between regions. One consequence of representing the accounts in the format shown in table 4.2 is that trans- actions between the domestic accounts for East Malaysia, on the one hand, and West Malaysia, on the other (that is, the interregional flows), appear as diagonal entries in the off-diagonal partitions of the table. This simply illustrates the fact that interregional transfers are simul- taneously an outgoing from an account in one region and an incoming to the same account in the other region. Not all of these accounts have interregional transfers associated with them, however, and for those that do the actual transactions may be zero. Definitional and actual zeros are distin- Table 4.2. Aggregate Regional Accounts for East and West Malaysia, 1970 (in millions of Malaysiani dollars) laPooditt.- last IbIayol. Malaysia Uco~~~~~~~~~~~~~~~~W.t Rest of Utld uint. Pattot. itti. s Caro lod Cm-o Attie1- ledieset wat Plt.t lasttitlti-s' cr,wert ovabiad cmwa- Aetill- [Nditot aC CuItt Total Uo..- co- Cot. Itol dtles tie. Taaot Mao- tt. to-.r..- Capital ditl.. tie. lass.. hold. pooto. sot h.ld ole aen 1 2 ~ ~ ~~3 6 3 6_ 9 10 it 1 2 3 4 1 6 9 10 12 * 6 1 ~~~~~~1179.0 1179.0 pant.,. 2 ~~~~~~ ~~~~~~~ 0 12.8 ~ ~~~~~~~1563.0 1575.6 - a1aaa 3 1376.0 74.2 10.9 23.0 o 48. - 0 C.ot8 149. . 0 3.2 160.0 3 6...roasat ~~36.1 32.5 233.0 226.6 13 2.2 SSO.6 3. C..aidd .pto ai,. 160.4 36.6 C-oodtt6 9 I11". 363.0 369.0 1421.9 91.1 1211.1 4901.6 Lotfttttos to 3108.7 306 loiLToct Iso11 129.2 123.8 253.0 u. 1 6369.2 6369.2 Watt... 2 0 ~ ~~~~~~~~~~~~~~~~~~~~~ ~~~~~~ 219.6 9257. 8.00soheId 3 0 7319.7 401.7 9.0 171.2 16.0 30. * C. 000900 41401.9 125.9 11.9 15339.7 Caain 235.0 331.9 939.1 1349.0 12.9 3107.9 coaaalwtod CapItal 6 0 903.8 262.3 818.6 -63.0 1399.9 Cosoditla 9 - ___218.2 6349.2 1742.0 2014.9 7691.3. 4,111.8 2227.4 Aclvtt… 17296.1 .17296j>\ IdI.t ir.. it ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~1182 366.71 1349.0 - j cot~~~.at ~7. 103O00 . 1046.7 5335.9 220.1 23.6 3750.9 9640.6 jjCapital 8- 63.0 T -208.5 36.8 -33.0 Irmal 17W 33. 461 16. 9.4 460 4023 30. 230 66.2 92575 306.6 ~1339.7 3107.9 13899.9 22327.6 17296.1, 1369..0 364.3 -83.0 RegionalAccounts 89 guished in table 4.2 by blanks and zero entries, respectively. Thus, for instance, the interregional transfers for three accounts are definitionally zero. The wants accounts (account 1) simply define a mapping between household requirements and the commodities that they implicitly demand; this mapping has no geographical dimension. The appropriate vehicles for trade in goods and services are the commodity accounts. By a similar reasoning neither activities nor indirect taxes (accounts 10 and 11) can have any interregional transfers: the sole function of activities is to produce commodities; similarly, the indirect tax accounts serve only to collect such taxes and eventually pay them to government. For some other accounts there appear to be zero interregional transfers, such as those between factors or households. While it is perhaps unlikely that flows actually were zero in all these instances, they accord with estimates provided to us by the Department of Statistics, and can be interpreted as being negligible. An important feature of the regional SAM depicted in table 4.2 is the distinction that we have made between functional flows and geographical (interregional) flows. All transactions between accounts are shown as taking place within each region. This is not to deny the possibility of some intraaccount transfers occurring within a region (such as household-to-household trans- fers in West Malaysia of 401.7), which are still functional flows in our sense. Interregional transfers simply augment the receipts of an account in one region and simultaneously deplete the same account in the other region. Clearly, since gross flows are being represented, it is possible for an east-west transaction to take place at the same time as a west-east flow. All of this implies, of course, that when interregional transactions are aggregated with corresponding flows to the rest of the world, the totals obtained give the aggregate external transactions for each region separately. An interesting permutation of the scheme shown in table 4.2 is to order the accounts initially by functional type rather than by geographical region. This has the effect of showing each cell of the eleven-account structure as a 2 x 2 submatrix, with the two rows and columns relating to East and West Malaysia, respectively. With this approach, all such submatrices would neces- sarily be diagonal for transactions between accounts i and j for i, j = 1, ... , 11; i # j. Only in the diagonal block of submatrices (account i to account i) can off-diagonal elements appear with this arrangement, and these would represent the interregional transactions. The approach has the advantage of showing even more explicitly than table 4.2 that the interregional trans- actions are transfer payments between accounts of the same type. And as transfer payments, these interregional transactions would be netted out if the regional accounts for East and West Malaysia were consolidated back into a SAM for Malaysia as a whole. THE TREATMENT OF INTERREGIONAL COMMODITY TRADE The aggregate structure of regional accounts as in table 4.2 clearly shows that regional interdependence involves transactions between a number of accounts. There are few conceptual difficulties involved in identifying these flows, even though the practical difficulties may be substantial. One exception to this general rule, however, is the treatment of interregional commodity trade. In general terms, imports of commodities are portrayed in the SAMframework as augmenting "domestic" commodity supplies. At a disaggregated level commodities may be distinguished by qualitative differences or even by price differences if the incidence of commodity taxes or distri- bution margins lead to a variation in user price. It should be recalled that commodity purchases are valued at market prices in our Malaysia SAM framework. This means that commodity transactions are valued at the prices that prevail at the location of the user, rather than the location of the producer. It follows that, on the supply side of the commodity accounts, the value of total commodity supplies comprises the following elements. First, there are domestic commod- Table 4.3. Definitions for Components of Imports and Exports Exports of Exports of lisports of Inmports of goods and ser- goods and ser- goods and ser- goods anidser- Imports fromii Imports from vices origina- vices origina- vices origina- vices origina- Rest of World Rest of World thig In East ting in West ting in West ting in East to East to West Malaysia to Mlalaysia to Malaysia to Malaysia to Mlalaysia Malaysia Ite eml West Mlalaysia East Malaysia East Malaysla West Malaysia (c.I.f.) (c.i.f.) (f.o.b.) (f.o.b.) (c.i.f.) (c.i.[.) X([ -F 1W) x(W + E) x(W -+ E) x(E -1.W) r(W -F E) r(E -, W) + +- + +I I + e(E -F W) w(W - E) e(W - E) e(E W) Imports of Imports of goods and gpods and Defin ition + - + + services services orlgliating orlginating e(W -t E) w(E -W W) W(W -F E) w(E W) In Rest of In Rest of World to East World to + tI- Malaysia West Malaysia (c.i.f.) (c.i.f.) r(W + E) r(E + W) The Flow of Funds as a Tool of Analysis 91 ity outputs supplied by production activities, which are valued at basic, or ex-factory, prices. Second, and also part of the "activity to commodity" matrix (the so-called make or mix matrix), are the margins supplied by the trade and transport activities that are required to deliver commodities from the point of production to the point of consumption. In our SAM system, we view these margins as components, supplied by trade and transport activities, of commodity outputs, rather than as the supply of a separate trade and transport commodity. This is because trade and transport margins are an intrinsic part of the value of a commodity when considered from the user's perspective. The third component is the commodity tax element, which is specific to each commodity account and which must be included in the market price valuation. The final component is that of imports c.i.f. (cost, insurance, and freight), the implication being that c.i.f. prices are equivalent to ex-factory costs from the user's point of view, so that when appropriate components of domestic distribution margins and commodity taxes are added, imports are revalued from prices at the point of entry (c.i.f.) to those actually paid by the user, namely, market prices (see Pyatt and Round, 1978, ch. 5). The conceptual problems of interregional commodity trade concern the treatment of insur- ance and freight margins on commodity flows, which are simultaneously recorded as a part of one region's exports (at f.o.b., or free on board, prices) and of the other region's imports (at c.if. prices). To accommodate these within a SAMframework, the insurance and freight services that are required to move commodities from the market in one region to that in the other need to be distinguished according to who supplies them; that is, the distinction must be made as to whether these services are supplied by East or West Malaysia or by the rest of the world. The notation E -- W will be used for the flow of goods from East Malaysia to West Malaysia; and W - E for flows in the opposite direction. The notation x(E -* W) is then used to denote the value of goods shipped from east to west, with the valuation being in terms of the market prices at the point of dispatch, East Malaysia in this case. By the time these same goods arrive in West Malaysia, their value will be increased to the extent of the freight and insurance charges involved in shipment. Such charges for the flow of goods from east to west will be denoted by e(E -> W), w(E --> W), and r(E -- W), depending on whether the required shipment services are provided by East Malaysia, West Malaysia, or the rest of the world. Similar notation x(W - E), e(W-+ E), w(W -- E), and r(W -+ E) can be defined to cover the parallel movement of goods from West Malaysia to East Malaysia. With this notation, various magnitudes can be defined so as to recogrnize that, in our SAM compilations, exports of goods are valued f.o.b. while imports are valued c.if. 2 Table 4.3 sets out the definitions needed. The first four columns of table 4.3 decompose the imports and exports for East and West Malaysia that originate in, or are destined for, the other region. One problem that immediately arises is the question of the appropriate treatment of the elements e(W - E) and w(E -* W). As we have seen, these refer to freight and insurance services on imports when their services are supplied by the importing region. In the balance of payments statistics, in order to maintain imports strictly at c.i.f. values, the convention is to include these elements both in the imports and the exports of a region. It is as though these services are exported and simultaneously reimported. This convention is reflected in table 4.3 where the element e(W - E) appears in columns 1 and 3, and likewise, element w(E -- W) appears in columns 2 and 4. In other respects, freight and insurance charges appear in the exports of one region and the imports of the other region just as one would expect. For example, x(E - W) is shown in column 1 (as an export) 2. PLrists will notice that we have not captured all possibilities, having excluded the East or West Malaysian supply of shipping services on goods imported from or exported to the rest of the world: e(R - W), w(E - R), and so forth. These elements can easily be handled by considering the accounts for a closed, three-region system: east, west, and the rest of the world. We limit ourselves to an open, two-region system basically because our focus is on Malaysia rather than the world as a whole. We are grateful to Ann Harrison for raising this point. Table 4.4. Final Interregional and External Commodity Flow Matrix Commodity Accounts Rest of World Total East Malaysia West Malaysia of Domestic Exportsof goods a commodity and services Total I transactionsIn x(E * W) originatingin commodity East Malaysia + East Malaysiato requirementsIn + e(E -. 1 W) Rest of World East Malaysia oS1 e(W E) (f.o.b.) o 1 -A 10 Domestic Exportsof goods o a commodity and services Total x(W - E) transactionsin originating in commodity o S + West Malaysia West Malaysiato requirementsin v(W - E) + Rest of World West Malaysia w(E + W) (f.o.b.) Importsof goodsand Importsof goods and services originating servicesoriginating in Rest of World to in Rest of World to Rest of World East Malaysia West Malaysia (c.i.f.) (c.i.f.) + + r(W-E) r(E- W) Total commodity Total commodity Total suppllesin suppliesin East Malaysia West Malaysia RegionalAccounts 93 and column 4 (as an import). Even freight and insurance services provided by the exporting region, such as e(E - W), are consistently classified in this way, although (and this is a second probleim to be dealt with) from the point of view of the exporting region they are an export of a service, while from the view of the importing region they are embodied as part of the c.if. values of individual goods. Table 4.4 displays these same component items within and between the commodity markets of East and West Malaysia in a matrix format. It emphasizes interregional and external flows of the commodity accounts, showing commodity requirements along the rows and commodity supplies down the columns. The basic SAM rule of full articulation of the accounts is clearly captured in table 4.4. It means that the exports (f.o.b.) from East to West Malaysia [x(E -* W) + e(E -* W)] are the true interregional flow from East to West; while from the importing region's point of view the insurance and freight charges not included, that is, w(E - W) and r(E -- W), must be handled separately. Of these additional freight and insurance items, table 4.4 shows quite clearly that r(E -* W) is an amount that is imported from the rest of the world. The item w(E -> W), which represents West Malaysia's freight services on its own imports, is both a receipt and an expenditure of the external accounts to West Malaysia when East Malaysia and the rest of the world are treated in aggregate. But when these are distinguished, as in table 4.4, the SAM rule that each transaction should be recorded only once comes into force. Accordingly, w(E - W) becomes a diagonal element in this format. West Malaysian commodity accounts receive income for providing services in shipping goods from East Malaysia to West Malaysia, and this income is also a cost of the West Malaysian commodity accounts because it is part of the amount that must be paid for the supply of goods and services available in West Malaysia. By way of a check on this treatment of interregional trade, it is useful to consider the effect of aggregation across regions and the resulting consolidation of regional accounts. If the East and West Malaysian accounts were consolidated, then the interregional transfer of commodities would be netted out, since they appear on both the credit side of one region's commodity accounts and the debit side of the commodity accounts for the other region. This is not the case, however, with the items e(W -- E) and w(E - W), which would not net out through the consolidated process, since they are not an interregional transfer. Table 4.5 shows estimates of the interregional and external commodity flows for Malaysia, and it illustrates numerically the schemes we have just discussed. It broadly corresponds to table 4.4 in its entries; the exceptions are the elements corresponding to intraregional commod- ity flows. In table 4.5 these only include the elements e(W -- E) and w(E -- W), which amount to 0.1 and 2.0, respectively, all other "domestic" commodity transactions being netted out. This means that row and column sums for each in table 4.5 have a direct interpretation. The row sums represent total exports f.o.b., so that exports from East Malaysia (f.o.b.) are 1,310.0. The column sums equal total imports c.i.f. into each region, which for East Malaysia amount to 1,265.0. The commodity trade flows between each region and the rest of the world could be decomposed in an analogous way. Such decompositions, however, are only important in seeking a consistent treatment of trade flows in a system of two or more interacting regions, that is, when imports of one region are being matched against the exports of another. In general, the rest of the world is not treated in the same way as the domestic economy, and the decompositions discussed above are not usually warranted. Moreover, it should also be noted that the balance of trade for either region with the rest of the world is unaffected by the treatment of trade flows we have proposed. A region's earnings in providing freight and insurance services on its own imports have been deducted from both total exports and total imports as normally recorded, so the difference between them is unchanged. Having noted a certain amount of rearrangement of the decomposition of trade flows, and in 94 The Methodology of Social Accounting Table 4.5. Interregional and External Commodity Flows for Malaysia, 1970 (in millions of Malaysian dollars) Commodity Accounts East Malaysia West Malaysia Rest of World Total _ .~~~~~0 a 98.1 1211.8 1310.0 U DU^ 8 218.2 2 0 (a) 4111.8 4332.0 Rest of 1046.7 3750.9 4797.6 World Total 1265.0 3851.0 5323.6 a. These items refer only to a region's earnings in providing freight and insurance services in bringing imports into the region from the other region. particular the extraction of freight and insurance services supplied by an importing region from the apparent interregional trade flows, it remains to discuss in more detail the way in which disaggregated interregional commodity flows can be formalized within a SAMframework. Commodity trade between two regions is represented in the SAM framework as a transaction between the commodity accounts of one region and those of the other region. Therefore, the general representation is by two matrices X(E -* W) and X(W - E), where, for example, the former is a matrix whose rows represent the commodity accounts of East Malaysia and whose columns are the commodity accounts of West Malaysia. Now consider the structure of X(E -* W) and suppose for simplicity that all freight and insurance services are supplied by the exporting region; we thus assume w(E -* W) and r(E - W) are zero. The element x(E - W) is the total amount of goods exported from east to west. At a disaggregated level, the individual commodity exports at f.ob. prices, which sum to x(E - W), would comprise diagonal elements of X(E - W), otherwise the diagonal would be zero. From the point of view of the importing region (West Malaysia in this case), these goods should include elements of freight and insurance margins. This can be shown in the matrix X(E - W) in the following way. From the point of view of East Malaysia, freight and insurance services are a commodity export. The situation is therefore Regional Accounts 95 analogous to the treatment of distribution margins in the SAM framework that was referred to earlier. These services have to be allocated across the East Malaysia rows of the commodities "freight and insurance services" to the whole range of West Malaysia commodity accounts columns according to the respective c.i.f. markup. The matrix X(E -* W) has the following properties. The row sums of X(E - W) would be a vector of exports from East Malaysia, f.o.b., written x(E W). (4.1) [X(E - W)]i = x(E- > W) where x(E W) contains as one (or more) commodities the total exports of freight and - insurance services to West Malaysia. The column sums of X(E - W) are elements of a vector of the same commodity trade, but shown as imports into West Malaysia, c.if., written m(E -* W). (4.2) [X(E -- W) ]'i = m(E - W) where the freight and insurance margins are now embodied within the c.i.f.value of the commod- ity imports. The difficulties that are often encountered in matching trade statistics, and which especially manifest themselves in multiregional analysis, can therefore be accommodated in the SAM framework in a way that is entirely consistent with the distinction of basic prices, factor prices, and market prices. The proposed convention amounts to using the trade matrices X(E -- W) and X(W -> E) to transform commodities defined at the border prices of the exporting region into commodities defined at the border prices of the importing region. Moreover, the conceptual structure of these trade matrices is relatively easy to put into practice with recourse to available data. I I PART II Country Studies I 5 A Social Accounting Matrix for Sri Lanka, 1970 S. Narapalasingam The most disaggregated SAM currently available for Sri Lanka relates to the year 1970. Some details of it have been discussed in chapter 2, and a full description is available in Pyatt, Roe, and associates (1977). In this chapter, I shall comment on conceptual problems that are somewhat peculiar to Sri Lanka's national accounts. My focus will be on the paramount need to improve existing social accounting methods and strengthen the data base, so as to make the SAM an effective tool in development planning. I shall also, in the wake of the social and economic changes in Sri Lanka since 1970, draw attention to those aspects of the SAM that need to be expanded further to provide empirical answers to questions of current interest to policymakers in Sri Lanka. In doing so, I shall touch upon the question of how to determine a frame for the SAM, because the usefulness of the SAM in development planning hinges on the choice of the frame.' This leads me into a discussion of the statistical needs of the SAM. Greater disaggregation of the SAM obviously places increased demands on the statistical system, and this aspect cannot be ignored. Because there are usually constraints on the quality of available data, greater disaggregation implies greater effort in reconciling different accounts. Obviously, there is a point beyond which disaggregation proves to be meaningless, unless the statistical system has developed to an extent that justifies greater reliance on the basic estimates themselves. Statistical techniques that iteratively balance the different accounts are meaningful only in the context of a strong data base. These are some of the fundamental problems that I wish to address. To motivate the discussion, I need to provide a brief introduction to the current policy emphasis in Sri Lanka, and I have set out my perception of the general context in which the need for a SAM arises as a schematic flow diagram, shown as figure 5.1. THE CURRENT POLICY EMPHASIS Over the years, Sri Lanka has achieved major advances in income distribution, health, educa- tion, mortality, and fertility comparable to those achieved by more developed countries. It is estimated that Sri Lanka has about one and a half times the life expectancy, about three times the literacy, a quarter the infant mortality, and half the fertility that would be expected at its per capita level of income. There have also been similar gains in income distribution: the Gini coefficient has declined from 0.45 in 1963 to 0.35 in 1973, according to the best available estimates. In the Overseas Development Council's index of the physical quality of life, Sri Lanka occupies a place among the developed countries, securing as high as 83 points out of a possible 100. This compares with an index of 75 for Mexico, 68 for Brazil, 80 for Korea, 41 for India, 42 for Algeria, and 16 for Ethiopia. Nonetheless, Sri Lanka's per capita income of around $150 a year is one of the lowest in the world. The most severe problem currently facing Sri Lanka is unemployment. The unemploy- ment rate (about 20 percent) is high by any standard and has been accentuated in recent years by a mismatch between the high aspirations of educated new entrants to the labor force and the types of jobs available. An explanation for the increase in unemployment is found not only 1. This point is dealt with in greater detail by Pyatt and Thorbecke (1976). 99 Figure 5.1. The Determinants of a SAMSchema for a Developing Country Existing structures Social and economic Objectives and special features policies requiring of the SAM of the economy emphasis in planning o h A SNA concepts Social accounting . Social accounting rules (noting special features of the economy) Programs for the ~~~~~~~~~~~~The SMschema b development of for the country statistical services X .. o h onr The SA Sri Lanka, 1970 101 in inadequate resource mobilization for economic development, but also in misallocation of scarce resources. The allocation of limited resources in favor of consumption, as against invest- ment, has not only increased unemployment, but has also made it impossible to sustain the various social welfare programs that were initiated after independence and that enabled Sri Lanka to achieve the impressive gains already mentioned. 2 The need to change the pattern of allocation of resources away from consumption and toward investment-in order to provide solutions to unemployment and economic growth problems- has been recognized by government. Changes to the food subsidy scheme enabled the govern- ment to divert substantial resources from consumption to investment in the 1978 budget. Until the beginning of 1978, almost the entire population was entitled to the food subsidy. Thereafter coverage was restricted, and the current beneficiaries of the food subsidy are those families (comprising about half the population of Sri Lanka) with incomes less than Rs. 3,600 a year (about US$20 a month). As a result of the nationalization of tea, rubber, and coconut estates of more than 50 acres under the Land Reform Programme and the establishment of new corporations in the public sector since 1970, 30 percent of the value added in total gross domestic production in 1975 came from the public sector, which also provided nearly one-quarter of total employment. More than two-thirds of the tea industry profits and one-third of the rubber industry profits now accrue to the public sector. The principal capital asset owned by the public sector, after the implementation of the Land Reform Programme, is in agriculture, including almost a million acres of land. The physical assets of the public sector in agriculture and industry have the potential to yield much greater returns than in the recent past, when efficiently and fully utilized. However, performance of the public sector in the recent past has been very disappointing. Many corporations have consistently failed to make any profits, so that nearly all capital expenditures by the public corporations have had to be financed by transfers from the govern- ment. Because the government's net current accounts position has been in deficit in recent years, not only all of the public sector's investment, but also part of its current expenses, have had to be financedby domestic andforeignborrowings. Even if the public corporations incurring operational losses are omitted, the rate of return on capital employed in public corporations was below 10 percent-much less than private capital would have yielded. This is because public corporations have been used as instruments of pricing and employment policies, in disregard of some basic commercial principles. The point to be emphasized is that at least some of the decisions that have previously aggravated Sri Lanka's financial problems and adversely affected the long-term growth prospects could have been avoided had the implications been formally studied. The political will might not have permitted alternative decisions, but this is not the immediate concern of planners. What is at issue here is how all the implications concerning income distribution, employment, and pricing policies could have been studied, so that sound advice could have been given to the authorities. Ronnie del Mel, minister of finance and planning, remarked in his budget speech of November 15, 1977: "Since 1956 we have been experimenting with piecemeal, patchwork, ad hoc solutions to our economic problems. The age of cosmetic solutions to economic problems is over, and this Government is determined to offer real and long-term solutions for the problems of our nation." Sri Lanka has paid dearly by not developing further the planning foundations laid in the late 1960s with the assistance of the United Nations Development Programme (UNDP), which subsequently contributed to the compilation of the 1970 SAM. The framework suggested for economic statistics in that SAM, though still relevant in the present stage of development of the economy, requires further disaggregation to take account of institutional and structural 2. "A static redistribution of income through consumption transfers, for example, would not alter the structure of production and resources in the initial period and, therefore, could not lead to any sustainable changes in the generation of incomes over time." See Pyatt and Thorbecke (197e). 102 CountryStudies changes since 1970. Moreover, the estimates provided in the 1970 SAM are now outdated; work on the compilation of a fresh SAM needs to be initiated. In this regard, the new emphasis given to employment generation and the need to reallocate resources from consumption to investment and to improve the financial viability of the public sector as a whole (including public corporations) have to be noted. Thus more disaggregated accounts are required in the new SAM. The traditional small-scale sector should be separated from the modern sector, since the former tends to be more labor intensive. Public sector accounts need to be disaggregated further, identifying separately (a) public corporations, (b) public enter- prises, and (c) government departments. Similarly, separate capital accounts are needed for households, private businesses, public corporations, government, public enterprises, and finan- cial institutions, so as to identify the sources of funds for financing investment by public corporations and by government. All these developments call for a pioneering effort to set up detailed flow-of-funds accounts for Sri Lanka. 3 In the 1970 SAM, there was only one consolidated capital account (including the rest of the world), and public corporations were grouped under the respective production activities, along with private firms. Today, public corporations assume greater importance in Sri Lanka because of the dominance of the public sector in all major industrial activities. The distinction between public and private corporations is important for another reason. The effi- ciency of capital employed in the two sectors differs considerably. If performance in public corporations is to be monitored closely, with a view toward improving capital and production efficiencies, disaggregation will be essential. The 1970 SAMidentified nine distinct occupational groups for compiling the manpower matrix. The usefulness of this matrix would be enhanced if technical skills were further disaggregated to distinguish engineers, welders, fitters, carpen- ters, bricklayers, and similar jobs that have become critical in recent years because of large- scale migration to the Middle East. THE DATA BASE Given the present state of statistics in Sri Lanka, the framework suggested for social account- ing is undoubtedly ambitious. It is also a great pity that, over the years, there has been a deterioration in the flow of information needed to update the estimates contained in the SAM. The quality of statistics on national income and components of national expenditure has not improved either. Moreover, the methodology used to compile the available estimates has remained the same, despite changes in the intersectoral dependencies, cost structures, and final demand patterns. In the estimation of gross domestic product (GDP), less than one-quarter of value added is estimated directly, while the rest either is assumed to grow at a trend rate or is estimated on the basis of circumstantial evidence. Value added in construction is calculated as a multiple of the value of certain materials inputs (local and foreign), while investment is derived largely as the value of output of construction plus a multiple of the c.i.f. value of imported machinery and equipment (that is, the value including cost, insurance, and freight). Even estimates of output of major crops and industries are suspect, because reliable estimates of acreage planted are not available for tea, rubber, and coconut crops; and the coverage of industries varies from year to year depending on the response to questionnaires that are sent out. In several instances, domestic consumptiorn in real terms is derived from estimates of constant per capita consump- tion obtained several years earlier, despite significant increases in the prices of consumer 3. It is inadequate to treat savings and investment as a mere overall accounting identity. The level of investment made by a given sector of the economy depends not only on its own surplus generated within, but also on the volume of investible funds made available by other sectors either directly or through credit institutions. Again, see Pyatt and Thorbecke (1976) and chapter 3 of this volume. Sri Lanka, 1970 103 commodities over the years. Estimates of the changes in inventories are simply not available for most commodities. Despite the overwhelming concern with unemployment and the cost of living, statistics on employment and consumer prices are unfortunately dubious. Anmong the several papers that have pinpointed the deficiencies of the statistical system in Sri Lanka, the most comprehensive is a report preparedbyPetterJacob Bjerve, statistical adviser to an ILO World Employment Programme mission to Sri Lanka in 1971 (International Labour Office, 1971). Bjerve examined the availability, adequacy, and timeliness of the statistical data and outlined a program for the development of statistical services in Sri Lanka. The recom- mendations contained in his report continue to be relevant, notwithstanding what Dudley Seers, the chief of the mission, wrote at that time: "To govern in Ceylon, given the present state of its statistics and the present structure crisis, is like driving a racing car without headlights along a winding road at night" (p. 153). The data used in the 1970 SAMwere not obtained from any one source: the more disaggregated the SAM,the wider the range of information sources is likely to be. It follows that, where there is some lack of comparability between concepts, definitions, sampling methods, and valuations used in the different sources, reconciliation becomes difficult. This indeed was the experience in the compilation of Sri Lanka's SAM. There was also the added problem of adjusting data to account for price changes, because the surveys from which estimates were obtained were carried out in different years. The control totals for the 1970 SAMwere the national income statistics compiled by the central bank. In a few instances, the control figures were rejected in preference to survey estimates. This is not surprising, since national income statistics compiled without reference to a SAM conceptual framework are unlikely to prove internally consistent when individual components of incomes and expenditures are compared. If different surveys are conceived and planned within a SAM framework, the prospects for achieving overall consistency in concepts and definitions are enhanced. What Sri Lanka and possibly many other developing cou.mtries lack is a commitment to the implementation of statistical programs covering a medium term of four to five years. In Sri Lanka's case, such a commitment on the part of the authorities is of paramount importance in expanding and strengthening the data base, so that a SAM could be used as an effective tool in development planning. There is a yawning gap in Sri Lanka between data supply and requirements for the compilation of a SAM. To fill the gap, I strongly recommend the preparation of plans for a medium-term statistical program, which will ensure that adequate funds are allocated annually for statistical work. (In fact, the lack of funds has been one of the major factors contributing to the deteri- oration in the supply of data useful for the compilation of a SAM.) Such plans would also ensure that surveys are designed to deploy available manpower resources optimally; that all data requirements are studied in advance, within a SAM framework, to avoid duplication; that the financial resources allocated for data collection are used efficiently; and, above all, that the SA-M is kept "alive" by feeding in new information. Ideally, statistical services must improve at a faster rate than the development of planning techniques. But if the former are allowed to stagnate, no useful purpose would be served by increasing the degree of sophistication of the latter. It has been argued that improvements in planning techniques wiln stimulate the expansion of the data base and the quality of data supplied. From Sri Lanka's experience this certainly has not been the case. SOME CONCEPTUAL PROBLEMS The new System of National Accounts (SNA) recommends the disaggregation of imports into two types: competitive and complementary. The latter are goods that cannot be obtained 104 Country Studies locally. In the 1970 SAM, all imports were treated as primary inputs and hence as comple- mentary imports and were valued c.if. The need to distinguish between protective and other import duties, as recommended in the SNA, therefore did not arise. Nevertheless, if the (cur- rent) input-output matrix contained in the SAMis to be used to forecast the outputs of different industries corresponding to a given level of final demand, a distinction must be made between competitive and complementary imports. If the objective is simply to obtain insight into interdependencies that existed in the past such a distinction may not be that important. However, for many developing countries, the scope for increasing incomes and employment (at least for several years) lies more in the area of import substitution than in export promotion. The import classification can therefore be crucial in economic forecasting and development planning. Thus import classifications raise a basic issue regarding the use of a SAM in devel- opment planning. Sri Lanka had a dual exchange rate regime from 1968 to 1977. The rates were unified from November 15, 1977, and then allowed to float on the basis of underlying market forces. Before this turning point, except for a few selected items (such as rice, flour, drugs, books, fertilizer, and some condiments), most goods were imported at the higher rate. Similarly, exports, except for tea, rubber, and coconut products (copra, coconut oil, and dessicated coconuts), were also valued at the higher rate. In the 1970 SAM, the premiums paid on imports were treated as indirect taxes (that is, as import duties) and those received by exporters of nontraditional products were regarded as export subsidies. In the rest of the world (current) account of the 1970 SAM, imports and exports were both recorded at the lower rate of exchange. These valuations of the external transactions raise two problems, given that the Sri Lanka rupee (at the official, lower rate) was overvalued in relation to other currencies in 1970. First, domestic savings are overstated to the extent that the conventional rest-of-the-world current account deficit failed to reflect the magnitude of the true deficit. The true deficit in 1970 would have been higher on the basis of the weighted average exchange rate. Domestic savings were derived by subtracting the balance of payments deficit (obtained by valuing all transactions at the lower rate) from estimated gross domestic capital formation and were therefore overstated. Moreover, the estimate of gross capital formation used in this calculation was valued at market prices. It therefore included the premium exchange payment made for imports of capital goods and accentuated the overestimation of savings. Furthermore, in the absence of independent estimates of savings in different parts of the economy, there was no way of checking the overall savings estimate. Second, the principle of homogeneous pricing in a SAMis sacrificed if the exchange rate used is at variance with the valuation of a competitive commodity produced locally. This problem is not peculiar to Sri Lanka alone but would arise in all countries that have multiple exchange rates. The SNA does not explicitly deal with problems arising from the use of multiple exchange rates. The problem, insofar as residual estimation of domestic savings is concerned, stems mainly from the use of different exchange rates for capital goods imports as against other current transactions in the external account. The dual exchange rate problem affects not only the estimation of domestic savings, but also the current account position of the government. The inclusion of the entire premium on foreign exchange in current receipts, including those generated by public debt amortization and the government's purchases of capital goods, leads either to an overstatement of the surplus or to an understatement of the deficit. Price distortions have been pervasive throughout the Sri Lankan economy, because of the previously prevailing dual exchange rates, government-administered prices, and price controls. Under these conditions, the task of ensuring homogeneity in prices in the valuations of the several entries in the SAMis formidable. Two different producer prices were available for rice- Sri Lanka, 1970 105 depending on whether the farmer sold his produce to the government or to the private trader (middleman). The effective price used in the SAM is the weighted average of the two prices. Likewise, the same commodity sold in different markets (at most there were three: the "official" market, the open market, and the black market) fetches different prices. This price differential poses a problem in regard to accounting prices. If the basic accounting principle of valuing private consumption at market prices rather than at artificial prices is accepted, then, when the government sells goods and services below open market prices, the question arises of whether there has been a transfer from the government to households of resources that would otherwise have been available for alternative use. In Sri Lanka's case, several examples can be found of goods that have been sold below market prices. The SAM has not captured these as transfers, because it did not use market prices for valuing goods that entered into private consumption. This is in fact a weakness in Sri Lankan national accounts, and the 1970 SAM has simply followed the practice. The loss incurred by the government in selling food, for example, either free of charge or at subsidized prices has been treated as a negative indirect tax, in line with the practice followed by the Sri Lankan authorities responsible for compiling the national accounts. It is more correct to treat this amount as a current transfer from government to households, since the food subsidy is not strictly a producer subsidy. It can be noted here that in the IMF format for presenting government accounts, the food subsidy is in fact treated as a current transfer item. If market prices are used instead of subsidized or cost prices to value consumption of commod- ities sold below market prices, the estimate of the implied transfer from government to house- holds will increase. This can be illustrated by an example. Let us take rice distributed free of charge to consumers and for simplicity assume that consumers purchase the balance of their requirements of rice in the open market. According to the present way of estimating the food subsidy on rice, the value of the subsidy will be exactly equal to the total cost incurred by the food commissioner in purchasing local and foreign rice and distributing it to consumers. Let us say this amounts to Rs x. Under the suggested valuation procedure, the income transfer from government to households will be Rs (x + y), where Rs y is the value of the margin that the food commissioner would have received had he sold the rice at the average open market price of rice. In other words, Rs (x + y) represents the market value of rice given away free of charge by the government. The upshot of this imputation is that the food commissioner's department would be construed as a government trading enterprise, and the margin (Rs y) realized on the distribution of rice would be charged to revenue of the government accounts. On the debit side of the income and outlay account of the government, an identical entry has to be made to increase the value of transfers to households to Rs (x + y). Thus the estimate of government saving remains unaffected by explicitly providing for the relevant entries in the SAM. These revisions would also entail adjustments to factor accounts. The food commissioner's margins (Rs y) should be reflected in the value added for wholesale and retail trade. This approach to valuation of private consumption would ensure that the estimates are comparable from year to year. If, for example, one assumed that the free rice was eliminated altogether and no other change occurred, the estimate of private consumption expenditure would remain unaffected by this change. The difference will show up only in the amount of current transfers receired by households from the government. Besides, as already mentioned, the SAM would reveal more accurately the extent of the transfer involved in selling goods below market prices. All these imputations, though not necessary for balancing the different accounts, are useful to policymakers and planners and may be shown explicitly in the SAM. Where open market prices are nonexistent, cost prices can be used to estimate the value of current transfers from the government to the household sector. In Sri Lanka's case, educational, health, and even passenger transport services are provided by the government either free of charge or at highly subsidized rates. These subsidies help to reduce the inequalities in income 106 CountryStudies - distribution. 4 However, the 1970 SAM for Sri Lanka does not show these subsidies explicitly as current transfers; the total value of the services at cost prices is shown as a current expen- diture incurred by the government (institution accounts) in purchasing government services (production activities). This line of argument reflects the view that "the provision of free services of health and education should be allowed for in the calculation of living standards" (Pyatt and Thorbecke, 1976). It is claimed that in the 1970 SAM pensions are treated as a current transfer to households rather than as expenditures on goods and services. However, this is not borne out by the entries made in the relevant rows and columns of the SAM. The reason for this apparent anomaly in the treatment of pensions is that in Sri Lanka there is no separate fund to which contributions are made and out of which pensions are paid, because the pension plan is noncontributory (except for the Widow and Orphans Pension Scheme, or W & OP,where contributions by employ- ees are compulsory). The government annually meets almost all the liabilities on account of pensions from its revenue (except some small amounts coming in as W & OP contributions from the employees only). As recommended in the SNA (paragraph 7.17), in Sri Lanka's case, contributions to social security must be imputed. If the assumption is made that the wages paid to employees currently in service are hypothetically higher by the imputed value of their contributions to social security and that the payments to pensioners are made from this matching contribution, both GDP and the value of government consumption (current expenditure on goods and services excluding transfers) are increased by an amount equal to the pension payments. In order to show pensions as distinct from wages and salaries in a SAM, current transfers from households to government should be increased to reflect the imputed value of the contributions made by employees to social security. Since a corresponding entry of the same magnitude must also be made on the expenditure side of government accounts (that is, a current transfer from government to house- holds), the imputations do not alter the current account position of the government. SUMMARY ANrD CONCLUSIONS The 1970 SAM for Sri Lanka, with the amendments suggested in this chapter, provides a comprehensive and detailed framework for the systematic and integrated recording of flows and stocks. (See also the final SAM scheme suggested by Pyatt and Thorbecke, 1976.) The amendments suggested will certainly help to obtain a better understanding of the hidden and hitherto unrecorded flows, which are important to planners and policymakers. The framework itself will help to improve the quality of Sri Lanka's national income and expenditure statistics, provided a commitment is made to collect economic statistics with reference to this framework. Such a commitment will require a high degree of coordination and planning in the preparation of statistical programs that can be implemented over the medium term. Finally, the suggested disaggregation of certain accounts in the SAMfalls in line with priorities that have emerged since 1970 in the wake of institutional, social, and economic changes that have taken place in Sri Lanka. The disaggregation will require vast improvements in the supply of statistics, particularly in priority areas. The framework underlying a SAM depends ultimately on the availability of statistics for the year chosen; hence, emphasis should be given to the 4. See the appendix by Jayawadena in Chenery and others (1974). Although in Sri Lanka's case, it was assumed, for lack of detailed data, that the entire population irrespective of their income classes benefited from these measures, the Gini coefficient showed a decline from that obtained by ignoring the distribution of educational, health and public passenger transport services in the calculations. It is reasonable to assume that the chiefbeneficiaries of these services provided by the government are those in the lower income groups. SriLanka, 1970 107 collection of economic statistics useful for compilation. Attention should also be drawn to the problem of reconciling different accounts in the SAM,because of the wide range of sources that must necessarily be used to compile the matrix. The gravity of these problems varies from country to country, depending on the quality and supply of statistics. In the Sri Lanka SAM, the control figures were taken from official statistics on the components of national income and expenditure, which in some cases were deficient in regard to concepts and methodology used in their estimation. The reconciliation problem would be mitigated by a well-articulated series of sample surveys conducted at regular intervals by a powerful central authority. That authority would be responsible for coordinating economic statistics on all aspects of the coun- try's social accounts. The other specific suggestions made to enhance the usefulness of a SAM in development planning and economic analysis in Sri Lanka briefly are: * Use of proper accounting rules and prices to show explicitly the hidden transfers from the government to other sectors * Adoption of the SNA's concepts and definitions in Sri Lanka's national income and expenditure estimates * Disaggregation of current and capital accounts to separate public corporations and enterprises from the private sector - Differentiation of traditional and modern sectors of the economy * Disaggregation of the capital account to reveal the variations in capital cost structures among different industries and between public and private sectors e Expansion of the manpower matrix to identify the supply of specific skills that are known to be critical in the development process e Compilation of a flow-of-funds matrix * Aggregation of household income classes into two or three categories, while retaining the urban-rural classification e Disaggregation of imports into competitive and complementary categories to obtain current input-output coefficients for use in forecasting. 6 A SocialAccounting Matrixfor Swaziland, 1971-72 S. J. Webster A team of economists and statisticians drawn from the University of Warwick and the Ministry of Overseas Development, London, visited Swaziland in the summer of 1974.1 Our objective was to examine Swaziland's economic situation and prospects through the medium of a social accounting matrix which we would construct. The matrix was intended to quantitatively describe the structure of the Swazi economy as it was then and to provide a consistent data framework within which the implications of different future development strategies and major economic events could be explored, especially with respect to employment and income distribution. By trying to construct a SAM for a country in the early stages of developing its administrative capacity we were consciously challenging the frequently heard assertion that data limitations preclude useful quantitative analysis of development problems at the macroeconomic level, even where policymakers are willing and able to make use of it. We were also hoping to demonstrate that the social accounting matrix presents a better framework for such analysis than the standard national accounts tables compiled by most developing countries. This chapter provides a brief description of the Swaziland scene as we found it, sets out the matrix we designed, and discusses some of the conceptual issues we had to resolve. It also recounts the major difficulties we faced in using available information to fill in our framework and explains the solutions we adopted. The finished matrix is presented together with some of the analysis it stimulated and a discussion of its usefulness. The final section discusses the lessons of the Swaziland experiment for introducing social accounting as a working tool. Although the work was conceived as a research project, we had hoped to leave behind a usable tool that could be maintained and improved by others. That this did not happen maybe seen as a challenge to our initial assertion that the approach is practical for developing countries; nonetheless, those of us who visited Swaziland believe that the problems can be overcome. In closing, I refer to subsequent work, and to work elsewhere, which is turning out to have been more successful in this respect, perhaps with the benefit of experience gained in Swaziland. It will be clear from the foregoing that the Swaziland experiment should not be viewed in isolation. The nucleus of the team had already used social accounting as the framework for studying development problems in Iran and in Sri Lanka. Various members have since been heavily involved with similar work in Saudi Arabia, Malaysia, and Botswana, and still further work is planned. On each occasion different aspects have come to the fore. The Swaziland experiment should therefore be seen simply as one contribution to the growing body of expe- rience on some, but by no means all, of the issues involved. In particular, the experiment focused on the data framework, and no attempt at carrying the work a stage further into macroeconomic modeling was made. 2 1. The team consisted of Graham Pyatt, Robert Lindley, Alan Roe, Jeffery Round, and Paul Stoneman, all from the Unliversity of Warwick, and Harry Fell and myself, both, at that time, at the Ministry of Overseas Development, London. 2. This remark relates to the mission itself. On a subsequent mission (discussed later) Jeffery Round constructed a simple model. 108 Swaziland, 1971-72 109 THE SWAZILAND SCENE IN 1974 Swaziland is a compact, landlocked country of about 6,700 square miles. Its eastern border lies at the nearest point some forty miles from the Mozambique port of Maputo. (In 1974 Mozambique had not attained independence and Maputo was still called Lourenco Marques.) The administrative capital of Mbabane lies about fourteen miles from the western border with the South African province of Transvaal, which also encloses the country to the north and south. Despite its small size, the country has four geographically distinct regions and is rela- tively well endowed with minerals, perennial rivers, and land suitable for agriculture. The climate varies from the temperate highveld, with a mean annual rainfall of around fifty inches, to the semiarid, subtropical lowveld, where the rainfall is sometimes insufficient to support the traditional subsistence crop of maize. A surge in primary economic activity took place between 1960 and 1964, and by 1974 there had been some additional development of light manufacturing activity, mainly around the commercial center of Manzini, and a very marked increase in the development of the Ezulwini valley (the Mbabane-Manzini corridor) as a tourist center, catering mainly to South Africans from the Johannesburg area. Mbabane itself had seen a considerable expansion in retail and wholesale distribution facilities. This brief description of the range of economic activity perhaps hides the extreme dualism of the economy. For a very large proportion of the population of around 400,000 the way of life in 1974 was a traditional peasant agriculture based on maize and cattle herding, although a few cash crops, notably cotton and tobacco, had made inroads in some areas. This dualism was reflected in a twin system of land tenure and government. At the beginning of this century, European settlers were confirmed by the British administration in freehold title to large areas, including some of the best agricultural land and what became the modern towns. In 1974, six years after political independence, there was and still is a sizable community of European farmers; and it is on the freeheld land that most of the commercial agriculture takes place, with the larger farms now being operated by companies which in some cases also control processing plants, in particular a pulp mill, sawmills, two sugar refineries, and a fruit canning factory. The rest of the land is held under the traditional Swazi system by the nation in trust for the people, and arable land is allocated to individuals by chiefs who also have the power to reallocate it. The Swazis, a single tribe forming about 97 percent of the 1974 population, settled the area in the seventeenth and eighteenth centuries. Their settlements consist of individual homesteads rather than villages, with each homestead traditionally supplying most of its own material wants, although for manyyears individuals have sought employment outside, including temporary emigration to the South African mines, without surrendering their links with tradi- tional household responsibilities such as ploughing and harvesting. Maize and sorghum are the most important arable crops, although the national cattle herd is large, much of it being in Swazi ownership and grazing on common land. As is often the case in Africa, the ownership of cattle has a significance beyond the mere provision of food. After political independence in 1968 the country had a complete Westminster-style govern- ment with the familiar ministries and a parliament in which, however, only the royalist Imbo- kodvo movement gained any seats in the preindependence elections. This government ran side by side with the traditional system of chiefs, the senior of whom formed the Swazi National Council. The two systems met in the person of King Sobhuza II who was, by a long way, the longest-reigning monarch in the world. In 1972 Parliament was disbanded, without much disturbance, but the civil service and the ministries continued to function as before in most respects, with ministers reporting to the King-in-Council. Inevitably, Swaziland has close economic links with South Africa. Since 1910 these had been embodied in a formal customs union, and in addition South African currency was used in 110 CountryStudies Swaziland. In 1969 a new customs union agreement was signed between South Africa and the three newly independent states of Botswana, Lesotho, and Swaziland. This related revenue shares to imports and vastly increased the revenue of the smaller countries to the extent that British budgetary assistance ceased. The provisions of the customs union, of course, have a marked effect on Swaziland's development, and part of the increased revenue was meant to compensate the country for the disadvantages of being in union with a more developed neighbor which was building up industry behind a tariff wall. South Africa is thus by far the most important source of imports for Swaziland, but it is much less important as a market for Swaziland's exports. Sugar, iron ore, asbestos, wood pulp, beef, citrus fruit, and coal were all exported beyond the region. In 1974 negotiations with South Africa to formalize the monetary arrangements were concluded. The Swaziland Monetary Authority came into being, and the lilangeni (plural: emalangeni) replaced the South African rand in general circulation, although the two currencies were closely linked and Swaziland did not aspire to full monetary inde- pendence. The postindependence development strategy of the Swaziland government was expressed in a plan that contained little macroeconomic analysis and was largely a statement of public sector projects and recent economic history. The mining developments and the main agriculture-based industries had existed before independence, and although there was some renegotiation with their foreign owners which gave the Swazis (through the Swazi National Council) a greater share in profits, the general attitude of the government was to create a climate that would encourage further industrial development by foreign companies. Some success was achieved, but the additional employment created was far too small to absorb the rapidly growing labor force. Similarly, although tourism grew spectacularly, the industry had little impact on the welfare of the ordinary Swazi. The Ministry of Agriculture, however, continued its preinde- pendence policy of trying to raise the standards of the Swazi farmer in both crop growing and cattle rearing. Rural Development Areas were set up in which these efforts would be concen- trated, and infrastructure was provided through the development budget. In 1974, the RDA program was just beginning to be implemented. Apart from the question of general development strategy, the Swaziland government faced a number of major issues. The sugar industry was booming at the time under the impact of record world prices and there were plans for expansion, but the iron ore mine which provided the next most important export after sugar was coming to the end of its profitable life. There were largely unexploited coal reserves in the lowveld, and some studies had been made of the feasi- bility of building a thermal power station to export electricity to South Africa as a means of using the coal. In addition, there was political uncertainty as to the future of Lourenco Marques, through which iron ore and most other exports passed on a railway for which Swaziland was still paying out of the returns from the iron ore mine. Finally, the growth of government expenditure on the civil service was beginning to outstrip the revenue from the customs agree- ment, and there was little prospect of continuing to absorb a high proportion of qualified school leavers into government employment as had been the case in the five years following inde- pe,idence, when the civil service was being built up and localized. THE SAM: STRUCTURE, CLASSIFICATIONS, AND CONCEPTUAL ISSUES The social accounting matrix we designed is presented in table 6.1 (This matrix shows differ- ent disaggregations of the accounts from the version of the Swaziland SAM shown as table 2.4 in chapter 2.) This section discusses first its structure and then the classifications through which we hoped to present a picture of the Swaziland economy at work. 1971-72 Swaziland, 111 Structure of the Matrix The first block of rows shows the factoral distribution of income generated in production activities, government employment, and domestic service. In the first block of columns this income is transferred to the institutions that own the factors-households of various types, business enterprises, and government. Something of the structure of factor ownership is thereby revealed. The second block of rows continues the process of collecting the receipts of institutions, which consist of transfers among themselves, including dividends, interest and taxes, and transfers from abroad. The government also receives indirect taxes from production activities. Thus all the sources of income for each type of institution are displayed. However, there is no direct mapping from production activities to household types, so that the activities on which particular households depend for their livelihood are not directly available from the matrix except by inference through the factor accounts. Apart from farming households, where the main activ- ities are obvious in any case, this omission may be said to reflect the view that labor is more mobile among activities than among skill levels, so it is the type of labor rather than the activity in which it is employed at present on which we concentrate. The second block of columns views the institutions as spenders and savers of the incomes received in the rows. Consumer expenditure and government expenditure are shown by commodity and the balance, after providing for the interinstitutional transfers already discussed and trans- fers abroad, appear as savings at the intersection with the combined capital account. The contribution of each type of institution to domestic savings is therefore directly available. The combined capital account of the nation forms the third block, in this case a single row and column. No attempt was made to disaggregate the capital transactions of different insti- tutions except in the case of savings, as already explained. Although considerable efforts had been made to document current transactions, capital transactions had received little attention. Nevertheless, the combined capital account column is able to show real investment by commod- ity. Domestic investment turned out to be less than domestic savings, and thus the capital account column is completed by an entry in the rest of the world account to reflect Swaziland's current balance of payments surplus. We knew in advance that data on the corresponding capital outflow was too scarce to justify an attempt to present it in the matrix, although we did attempt to complete the balance of payments in a separate table. Had our investigations revealed a deficit on current balance of payments, then the entry would of course have appeared along with savings in the combined capital account row as another source of finance for domes- tic investment. Our representation of income generation, distribution, expenditure, saving, and investment is thus complete, and we have the pattern of final demands for commodities which the process creates. These are shown in the fourth block of rows, together with the requirements of produc- tion activities for commodities as intermediate consumption. The fourth block of columns shows the breakdown of the supply of each commodity among domestic production activities and as between domestic production and imports. The pattern of demand for imports by commodity is, therefore, directly related to the structure of production and the distribution of income. The matrix as presented above displays what we saw as the most important mappings of transactions between decisionmaking units for analysis which was to focus on questions of income distribution and employment generation. Clearly, however, not all the tabulations which are potentially of interest to the administrator are available directly from the matrix. For example, even if we had disaggregated the capital account, we should have been able to show only thLepurchases of capital goods by institutions and not the use made of them in different production activities. Therefore, the matrix should not be regarded as replacing the need for Table 6.1. Swaziland SAM, 1971-72 (in tens of thousands of rand) Factors Institutions (Currsnt Accounts) Households Corporations Government 1 2 3 4 S 6 7 8 9 10 11 12 13 14 15 16 17 1 19 20N 21 Swazi naUon land ( labor) I Freehold farmers (9 land) 2 Other land C natural resources 3 Self-employment, n.e8 4 Rmployment 5 40 481 Housing 6 Other domestIc capital 7 Other foreIn Capital 8 Swazi nation land 9 1019 43 244 1 6 5 S ;3 JUpperincome 10 116 30 132 2437 284 110 27 nercome I 8 1220 118 12 7 2 PYrlvte nonprofit 12 13 5 6 e LaOge private, for profit 13 -21 1231 9 Other private, for profit 14 7 -16 766 804 29 181 Public 10 201 Swazi NaUonal Council 18 11 29 23 useoms unlon 17 12 180 92 3 6 8 09 Drewt axes I8 1166 24 299 94 Other revenue raising 19 27 6 44 11 76 22 Consolidated revenue 20 852 054 315 -146 -393 -981 Oombined Capltal Aodunt 21 339 418 40 180 1226 105 63 231 -2098 Maine 22 314 99 182 7 Other fresh food 23 249 178 210 30 Other agrlc produce 24 3 23 mInerals 25 1 62 Petrol Volls 28 57 22 28 3 m lectHw Uwater 27 27 8 9 3 Buildings Y works 28 704 Semi finiabed manufactures 29 5 31 Proceed food, drink 30 136 213 149 37 31 Machiney Vvehicles 31 170 37 990 Other fInished manufactures 32 81 376 233 85 S8 Dwelling services 33 39 333 70 36 Other peonal servIces 34 62 338 86 130 294 Transport & communicatlons 35 22 63 37 135 Dlstributve & bus. services 36 98 332 212 100 Traditonal activIties 37 Other farming & forestry 3H Ming 39 H3 5 Processing agrc. products 40 Other manufacturing 41 Consatruieon 42 Puoblc utbUiti 43 Hotels & restaursats 44 Other trade v transport 45 Other services aetivitie 46 Rest of the World 47 636 182 96 49 61 Total 481 1019 115 56 183 3901 391 967 2033 1339 3133 1360 18 1254 1793 201 63 852 554 315 231 0 0 0 O Commrodities [Production Aotivities Swazi nationland (& labor) I 1019 1019 Freehold farmers (& land) 2 ~~~~~~~~~~~~~115 115 Oilier land #1natural reource 3 45 7 2 1 1 56 Self-employment, ns.e. 4 1 2 5 4 133 35 183 Employment 2 15 759 316 494 94 257 55 90 466 734 3901 Housing 6391 391 Other domesil cepital 7 597 3 19 33 Si 150 54 967 Oilie foreig capItal 8 344 516 781 47 68 £27 3 2035 twaz nation land 9 40 1359 upperincome 10 3135 lowerincome 11 P65als nonpishIt 12 18 -e 6 larg private,for profit 18 27 1254 88 Oilierprivale, Ilir profit 14 22 1793 8 PobOc 15 201 SWas NailonalCouncil 16 63 a* Oustomesunion 17 48 33 52 48 24 1 12 20 40 214 852 Dire tlanse 16 584 -0 Othierrevenue raising 19 -4 10 27 13 11 2 1 7 44 17 1 315 Cossollidated revenue 90 231 CapitalAcountt Comined 21 -509 0 maize 22 21 623 Oilierfresh food 25 42 67 851 1647 Other agric. produce 24 6 165 1 1793 3 13 370 2377 Minerals 25 1 1 17 8 1217 1307 Petrol &olls 26 30 13 67 4 14 4 4 24 11 254. Eiectrlilta' & waler 27 25 34 99 7 1 6 6 7 4 233 Buildings& works 28 10 91 845 Re..i-finiehedmanufacture 29 38 28 17 15 99 i 1307 1941 Procesed food,drInk 307 94 47 91951 2665 Machinery & vehicles 31 68 99 40 11 12 2 224 102 30 1754 Oilierfinished manufwtunre 32 11 444 119 150 93 125 5 38 32 125 169 2157 Dweuinng sevices 33 5 Oiler personal service 349 15 Transport & conmmnications 25 105 2 191 30 1 38e 7 362 140 Dlstrlibtive & bsis.service 3 3 171 185 132 63 110 15 £2 131 34 69 1884 Traditional wavtiole 37 420 300 169 51 110 1090 OilierfarmInig& foredor 35 23 999 1966 3018 MIning 39 1289 14 3 43 1349 Processing gegre.products 40 32 25 9 5 32 1358 1900 10 27 1 9 49 30 Oilier mauwiig41 253 13399 52 Construction 42 697 16 2 715 Publicutlihties 43 194 5 190 Hotels & restaurant. 44 368 6 Other trade & tranport 45 1 10 922 971 1904 Oilherservices acuvtites 45 482 991 308 1751 Rwstofthe World 47 118 95 46 18 2584 90 183 382 1768 1897 4785 0 6536 TOWa 48 523 1547 2377 1307 284 233 845 1541 2585 1784 2157 483 1339 1407 i841 1099 3018 1349 3902 £26 715r 190 36 1904 1762 B Note: HDW signifies rest ofworld, Len.s. signifies specified. not elsewhere 114 CountryStudies standard tabulations. Rather it is a means of focusing attention on the main interrelationships in the economy and thus encouraging analytical use of economic statistics which we find requires a much greater effort of imagination when they are presented separately. Naturally, greater or lesser disaggregation of the accounts shown would be suitable for different purposes, and we turn now to the most detailed classifications that we produced in our attempt to provide a suitable framework for analysis of the situation in Swaziland. Classifications in the Matrix The full set of classifications used is presented in table 6.2. Rather than going through each account in detail, this section aims to discuss the main issues faced in arriving at the final classifications. To some extent it interacts with the data problems discussed in the next section. Institutions. Whether they are households, business enterprises, or government departments, institutions represent people collected together into the real decisionmaking units of the economic system. They are the owners of factors of production, the source of entrepreneurship, and the consumers of commodities produced. For this reason it seems more appropriate to give them first consideration than to follow the traditional pattern of beginning with the production account, and hence focusing on industries or factors. Since the final aim is to improve the living standards of the population, especially the poorest groups, it is to the classification of households that we turn first of all. By far the most numerous household type was the traditional Swazi homestead on Swazi nation land. These were clearly to be our first group and there was a possible subdivision by location since RDAs had been designated and it would be interesting in the future to compare the progress of households living in RDAs with that of other traditional households. We did not attempt any further subdivisions, although given adequate data there would have been enough regional disparities produced by the geography of Swaziland and the historical pattern of communications to justify a regional breakdown. Apart from this we came to the view that the way of life of traditional households was sufficiently standard to deal with them as a single group, despite the difference in income and wealth which the traditional system itself generates. This may reflect a further view that policies designed to act directly on the traditional system would stand little chance of success. The wholly different way of life of households living in urban areas reflects the dual nature of the economy. Here, subdivisions were clearly necessary and partly because of the general scarcity of data we opted in the end for a very simple classification into high income and low income. We thus ignored divisions that might have been based on race, location within urban areas, type of dwelling, or the occupational or educational level of the household head. By doing this we showed directly the proportion of total income going to the poor and, of course, we could compare the average income of poor urban households with that of households living in the traditional manner in rural areas. We were not, however, able to point to any useful subgroups in the population which had features in common other than their level of income, except such as would be revealed through the factors from which their income was derived. This is largely a reflection of the fact that Swaziland's population has little of the complicated ethnic and tribal structure that characterizes many African countries. Furthermore, urban areas throughout the country are fairly similar despite the fact that certain "towns" are owned and operated by companies for their employees. Since we could find no readily identifiable social grouping that had special access to higher occupations or educational facilities and none of significant size which was excluded from them (except by living in rural areas too far away), the remaining possible classifications for urban households were merely correlates of income, and only worthy of consideration if they had any particular advantage with regard to data availability or if we had been asked to consider them for policy reasons. As it is, we did not even find it necessary Swaziland, 1971-72 115 Table 6.2. Classification of the Accounts in the Swaziland SAM Main Account Detailed Accounts Notes Factor Accounts Swazi nation land-traditional Accounts for Swazi nation land and free- Swazi nation land-Rural Develop- hold farmers include land and labor. ment Areas Freehold farms are also referred to as Freehold farmers individual tenure farms. Other land and natural resources Self-employment (not elsewhere specified) Employee compensation (including housing subsidy) Other housing Other capital-Swazi controlled Other capital-non-Swazi controlled Households (Current Accounts) Traditional Swazi households Full detail at this level of disaggregation RPuralDevelopment Area households is available only on the income side of Freehold farm households, high- the accounts. Household expenditures are income estimated at a more aggregate level. Freehold farm households, low-income Urban households, high-income Urban households, low-income Companies (Current Accounts) Private nonprofit organizations The account for public corporations Public corporations comprises five public corporations. The Large private companies account for large private companies Small private companies comprises five companies. Government (Current Accounts) Swazi National Council Customs union account Direct taxes Other revenue raising Consolidated revenue account Health expenditure account Education expenditure account Other expenditure account Capital Account Consolidated capital account Commodities Account Commodity accounts Aggregated into 15 accounts in table 6.1 Production Activities Account Activity accounts Aggregated into 10 accounts in table 6.1 Rest of the World Account Combined current and capital account for the rest of the world to separate urban European households (whether expatriate or citizens of Swaziland) from their African counterparts in the upper income group. They resembled each other quite closely in lifestyle, and with localization proceeding in both public and private sectors, interest in European households from the macroeconomic standpoint was dwindling along with their numbers. We were, however, uneasy with the sharpness of the distinction we had made between Swazi households living in the urban areas and those living in the rural areas. Such is the strength of the household tie in African societies, that we knew them to be quite capable of spanning the rurallurban distinction in their capacity of decisionmakers. FPurthermore, we had iittle hope that the data would allow us to capture most of the interhousehold transactions which arise by separating family members living in towns from the rural homestead. Because the urban 116 CountryStudies areas of Swaziland have only a short history as true centers for most of the present population, and thus few separate, entirely urban-based families exist, it is difficult to evaluate the impact of employment generation in the modern urban sector on the country as a whole. Nevertheless, although there are few urban families without strong rural links, there are enough rural families who as yet have few if any urban links to justify our analysis. Having covered the traditional rural areas and the urban areas, we were left with households living on freehold farmland. Apart from the farmers (mainly European or colored and, in Swaziland terms, in the upper income bracket), these included the households of their regular employees and other households that provided a pool of casual labor but otherwise lived in the traditional fashion, often on farms where the owner was absent and carried on no farming activities. Again, a simple subdivision by income was sufficient to separate the groups. We felt that the classification presented above gave a reasonable picture of the household types in Swaziland in 1974 with which Swaziland's administrators would be concerned. Certainly, even this broad classification stretched the available data to their limits and beyond. We believed, however, that enough reasonable assumptions could be made and tested later by fresh data collection to put the matrix together on this basis. After households, the second institutional group we distinguished was the private corporate sector. There is a very marked gap in the size distribution of private corporations operating in Swaziland. Certain enterprises in mining, agriculture and forestry and agricultural processing activities are very large, relative to the economy as a whole. These giants, all of which are essentially foreign controlled, clearly required a separate category from the general run of private business enterprises, because they would be affected by and react to economic events and changes in public policy in quite a different way. Which criterion of size we used did not matter, and the task of separating out the giants was quite straightforward. Other private corporate enterprises were left as a single group. Most appeared to have a high degree of foreign ownership in any case, and further subdivision by size or type of technology did not seem likely to yield any groupings that would be interesting for policy purposes, given the data and time available. Our third institutional category was the public sector. The Swazi National Council, as an institution that made certain crucial decisions as to the use of Swaziland's natural resources and that had recently become a shareholder in enterprises exploiting them, clearly deserved separate treatment, although we knew in advance that data would be hard to come by. Beyond this distinction, and a separate class for public corporations such as the Swaziland Electricity Board, the main issue was how to classify the revenue-raising activities of the "modern" govern- ment. Revenue from the customs union was of such importance and was generated so differently from other revenue that we decided on a separate customs union account. On the expenditure side, we chose to show health and education separately from other government expenditures, partly for their intrinsic interest and partly because private institutions as well as government were engaged in providing health and education services. Factors of production. We started from the usual view of factors as consisting of land, labor of various types, and reproducible capital assets. Very soon, however, we realized that it was fruitless to try to separate the returns to the labor of the traditional Swazi household from the returns to its land, especially given the use of common grazing and the system of allocating arable land to individuals by the chiefs. We therefore identified a single composite factor of production consisting of the land, labor, and capital operated by a traditional household. In some respects this decision begs important questions, because it creates a one-to-one relation- ship between the factor and institution accounts and precludes, for example, analysis of the difference in returns to those traditional farmers who have received some measure of training or exposure to extension services. However, the decision reflected the fact that the household in traditional Swazi society operates as a producer-in fact, an entrepreneur-as well as a 1971-72 Swaziland, 117 consumer. Furthermore, the household can and does sell some of its labor to production activities which it does not control, and thus the matrix is able to show what proportion of income is derived from wages and what proportion from self-organized efforts. This seemed a much more interesting division than estimating the "pure" returns to land, labor, and capital. We chose to take a similar course with freehold farmers, although since there was a market in freehold land, we could in principle have imputed a rent for land farmed by its owners. In our view this calculation would not have justified the effort of making it. Land other than that farmed by its owners or its occupants under the traditional system was classified with other natural resources as a single factor of production. Apart from the labor factors included above and in other self-employment, we hoped to classify labor by level of skill. This we found to be available by production activity from employment surveys, but data on which to map labor incomes by level of skill into households by type was almost nonexistent. Hence the discussion of employment by level of skill took place in a separate analysis outside the main matrix. The way we used the data on labor incomes by skill in the household accounts is described in the next section on data problems. We had some difficulty in deciding how to treat the "income" of employees in the form of subsidized accommodation. Since some important employers spanned several production activ- ities and provided a whole range of facilities amounting to a complete township we decided to show the benefits as a transfer from the corporate institution involved to the relevant household group, rather than as factor income. Capital as a factor of production was divided into other capital, Swazi-controlled, and non- Swazi controlled. This last was clearly a desirable distinction although, equally clearly, estab- lishing the true pattern of ownership and control of enterprises takes much careful investi- gation. Nevertheless, the Central Statistical Office (CSO)had classified firms according to whether the majority of shares were held by residents of Swaziland, and we followed them in this. Commodities and production activities. Our final classification had forty-four separate commodities and twenty-six produietion activities. The classification followed the lines of stan- dard international practice, except we tried to expand the parts that seemed important and contract those that were less so. It is the number of classes, rather than the scheme of clas- sification, which requires comment. For an economy the size of Swaziland and in its stage of development in 1974, a smaller number of classes in each case would have resulted in a more manageable matrix, and hence in a matrix more likely to be used. Aggregation is, however, always possible, and we had good reasons for starting with a disaggregated list. Although the CSO carried out several Industrial and Agricultural Production Censuses, and also surveyed other sectors in the process of compiling national accounts, the reports contained very little commodity detail, and the industrial classification was inhibited by the need to preserve confi- dentiality in publications. There had been no attempt at producing commodity balances or input- output tables. In contrast, import statistics were collected in a fair degree of commodity detail by surveys of the importers, there being no customs control within the region, and it seemed to us that this accident of the customs union arrangements could be exploited to provide a much more detailed picture and a far better check on the macroeconomic aggregates that were being estimated than had so far been available. What we did is described in the section on data problems, but the point here is that it required us to work at a very detailed level. We decided that our results should remain in detail for the benefit of those we hoped would follow. DATA PROBLEMS AND SOLUTIONS The data provided to us by the OSOcovered a great deal of the necessary ground. It fell short, however, in certain specific areas and in a number of general respects. Of the latter, timeliness 118 CountryStudies was probably the most important, and the second half of this section is devoted to what we did about that. First, we turn to the basic data with which we were presented. The national accounts which were prepared each year contained estimates of gross domestic product by origin and expenditure on GDP, but did not attempt a capital finance account, and the external transactions account was incomplete even as regards the current account. The accounts were, however, based on surveys of various aspects of the economy to which there had been reasonable response. The most important of these for our purposes are: * Surveys of importers * Survey of agriculture on Swazi nation land * Census of individual tenure farms * Census of industrial production * Survey of employment and wages * Rural household survey * Survey of income and rent in Msundusa (a poor area of Mbabane). In addition, we had the government accounts and a great deal of incidental information from the Annual Statistical Bulletin. The list above is not atypical for a small statistics office and represents the fruits of several years of hard work by successive statisticians. Only one item requires special comment. Import statistics were not collected as an administrative by-product of the Customs Department for the simple reason that most imports came from South Africa, and in terms of the customs union agreement there were no customs formalities between the two countries. However, the value of Swaziland's imports is the most important single deter- minant of its revenue under that agreement, and consequently much ingenuity had been expended in designing surveys to measure it. In particular, each modern sector business received a quarterly questionnaire asking it to report the value of direct purchases from South Africa and elsewhere. To assist respondents in the classification of commodities, a guide was issued which corresponded roughly to the three-digit level of the Standard International Trade Classification (SITC). Separate arrangements were made for small Swazi traders, personal shoppers, and postal imports. Since a very high proportion of expenditure went to imports, the fact that we were able to classify them by the production activity of the importer proved critically important when we came to build up input-output tables. The commodity detail also gave us a reasonable guide to end use. The main defect of the list of sources presented above is that an urban consumption survey was lacking. In the national accounts private consumption was estimated as a residual, and it was only through careful analysis of the import data and comparisons with consumption patterns revealed by surveys undertaken in South Africa and Lesotho that we were able to build up a vector of expenditures by households on commodities that we could enter in the matrix. We could not, in fact, sustain the division of households into the five categories that had proved feasible when examining incomes, and we were reduced to three categories-traditional, high- income nontraditional, and low-income nontraditional-for the expenditure entries. Some may question the usefulness of a matrix that contains guesswork of this sort. We would justify the exercise in two ways. First, it makes explicit assumptions which have to be made informally in considering policy options, and the results of such consideration at a broad level are often insensitive to the accuracy of the detailed data that underlie them. Second, we did not (and do not) view the construction of the first SAM for a country as anything more than the start of a process by which the coverage and consistency of its statistics are gradually improved. If the first attempt reveals that certain areas are very weak, then these are the areas on which subsequent data collection should be concentrated. The other defect of the sources from our point of view was that each stood independently as a survey of a particular area, and the mappings from one to another, which lie at the heart of Swaziland, 1971-72 119 the concept of a social accounting matrix, did not exist. Thus, for example, while the employment and wages survey had a perfectly good classification by level of skill, information on skills had not been collected in the household surveys. Similarly, no input-output work of any kind had been attempted. The analysis of imports by importer led us to the solution of that problem, since it gave far more detail on the use of commodities by industry than was available from the Census of Industrial Production. Because domestic interindustry transactions were rela- tively few, and we could identify the main ones separately, this information proved all that was necessary. The problem of skill classifications we solved by using the information on earnings by skill from the employment and wages survey to separate household incomes according to the categories we had defined, although in doing so we lost the ability to show more than one category of labor as a factor of production. This limitation is certainly a serious defect for analysis using the matrix and would have to be remedied by further data collection. We turn now to the problem of timeliness. When we arrived in Swaziland in July 1974, the National Accounts for 1971-72 were still in draft, and there was no hope of selecting a later year as our base. So much had happened both internally and externally between 1972 and 1974 that we felt it essential to try and fill the gap as best we could and so provide a foundation on which the updating of our matrix could be built. It was out of the question to generate in a short time a model that was capable of updating the whole matrix with any degree of reliability, and so we chose instead to tackle the problem piecemeal, using the 1971-72 matrix as a framework. The understanding we had gained of the economy through completing the matrix gave us more confidence in this process than we would have had otherwise for the following reasons. First, we knew that a great proportion of economic transactions had as one party the govern- ment, one of the few giant private enterprises, or one of a similarly limited number of institutions handling the export or processing of agricultural products. Thus we were able to draw up a list of key institutions and arrange a program of visits to be carried out by members of the team, working in pairs. These visits served other purposes as well, in that they enabled us to discuss our interpretation of the basic statistics already available and future prospects for the enter- prises concerned. They gave us solid information on virtually all the exports of the country, and since this accounted for a high proportion of activity other than subsistence farming, they also gave us a base on which to build an estimate of the change in production. Second, with only two commercial banks operating in Swaziland at the time, it was possible to gain a great deal of useful information as to the general growth in commercial activity by talking to their managers, who proved most helpful without breaching the confidence of their individual clients. Third, import statistics were reasonablyup-to-datebecause of their importance in determining customs revenue. Having carried out a very thorough analysis of 1971-72 imports in the process of putting together the production and commodity accounts, we were able to make much use oi later figures to estimate change in expenditure. Furthermore, since such a high proportion of imports came from South Africa, the South African price indices, coupled with information on transport costs, could be used as a supplement to Swaziland's own price indices to provide estimates of real change or to convert volume indicators of change into values. Fourth, having already covered the marketed output of traditional farmers through our visits to processing and marketing institutions, we found that the size of the maize crop and the net increase in the cattle herd were by far the most important remaining items for general welfare. Although the Department of Agriculture relied on the CS0 for surveys to establish the former, their extension workers could make informed comments, and we could supplement these with information from the commercial milling company, which saw the success or failure of the traditional crop reflected in the amount of maize coming to them as a surplus or being bought from them for consumption in the rural areas. In the process of this investigation we were 120 CountryStudies able to show that the CSO's surveys, as then constituted, gave very poor estimates of change, a fact that would have been noticed long before had the data actually been used. Finally, there were several statistical series available, such as electricity sales and installa- tions by class of consumer and registrations of new motor vehicles, which were very up to date. Normally, they were used by the CSO only in the Annual Abstract of Statistics, but we found that, with the other information we had pieced together, they helped us bring our macroeco- nomic picture more or less up to date. The conclusion we reached after going through this process was that, given sufficient incen- tive and a good framework within which to view the separate pieces of information, it would be possible for even the small statistical office to provide a much more up-to-date service than had been the case. Furthermore, as experience in using different indicators developed and was compared with more formal survey information, the results could be expected to increase in reliability. A SOCIAL ACCOUNTING VIEW OF SWAZILAND'S PROSPECTS A condensed version of the SAM we produced for Swaziland in 1971-72 has previously been presented in table 6.1. In order for it to fit within the confines of this book, production activities have been reduced from twenty-six to ten, and commodities from forty-four to fifteen, but the broad outlines of the economy mostly remain clear. Detailed analysis naturally requires the full matrix, but this section will concentrate on the broader issues and uses to which the matrix can be put. First, the matrix shown in table 6.1 provides a very convenient source of data for writing a quantitative description of the economy. For example in the top right corner, the factoral distribution of income generated by different production activities is displayed and the activities can be compared. The two major activities of mining and processing agricultural products are each seen to depend heavily on foreign-controlled capital, in each case to the extent that about 62 percent of the year's value added goes as a return to such capital. Proper interpretation of income from employment naturally requires knowledge of the numbers involved and these are presented by skill level together with incomes similarly distributed in tables 6.3 and 6.4. By the same token full understanding of the household distribution of income shown in the second block of rows requires knowledge of the population in each household group. Swaziland's popu- lation census was eight years old at the time we did our work and of course did not contain any usable data on income. We did, however, have projections of population living on Swazi nation land, on individual tenure farms, and in urban areas. These are presented in table 6.5. With the aid of this additional information it can be seen that there are strong inequalities in the income distribution and that households living on Swazi nation land were participating only to a minor extent in the employment market. The block of columns in table 6.1 devoted to households shows how income was disposed of, in particular who bore the taxes and how the pattern of consumption varied between household groups. From the next block of columns it can be seen to what extent the government was successful in collecting tax revenue from the corporations, how much was retained by them to finance further investment, and in the final entries how much of the income flowed abroad. Government activities are depicted in the next block and capital formation by commodity in the central column. Where the rows for commodities intersect with the columns for production activities there is embedded a matrix of the technical requirements of each domestic production activity followed by a vector showing exports by commodity. The converse of these entries to complete the commodity balance is the make matrix and vector of imports shown in the last block of rows. We would argue that presenting the information in a matrix format like this, rather than in 1971-72 Swaziland, 121 Table 6.3. Payment of Employee Compensation in Swaziland, by Occupation and Industry, 1971-72 (in thousands of rand) Administrative Semi- Production Activities and Technical Clerical Skilled skilled Unskilled Total 1 Traditional Swazi households - - - - 144 144 2 RDA activities - - - - 14 14 3 Individual tenure farms 1,389 - 348 544 4,601 6,882 4 Forestry 360 48 82 265 894 1,649 5 Maize milling 80 58 - 3 39 170 6 Sugar milling 777 152 688 73 893 2,583 7 Cotton ginning 31 8 6 4 22 71 8 Timber pulping 438 49 659 412 267 1,925 9 Timber sawing 110 29 118 94 251 602 10 Meat packing 23 5 43 12 78 161 11 Fruit canning 119 40 46 60 163 428 12 Iron ore mining 211 70 135 320 209 945 13 Asbestos mining 521 78 373 679 632 2,283 14 Coal mining 49 9 9 15 133 215 15 Other mining 16 3 3 7 54 83 16 Other manufacturing 252 73 107 158 363 953 17 Utilities (electricity & water) 230 51 68 106 129 584 18 Construction 204 57 838 712 760 2,571 19 Transport 562 153 481 532 615 2,343 20 Posts and telecommunications 136 129 38 9 20 332 21 Hotels, restaurants, and bars 135 115 133 213 319 915 22 Wholesale trade and modern retail 796 243 112 175 954 2,280 23 Small Swazi traders - - - - 43 43 24 Education and health 4,184 123 70 189 302 4,869 25 Other service activities 1,311 540 274 142 250 2,515 Total 11,932 2,130 4,627 4,724 12,149 35,562 separate tables, has the great advantage that the inevitable inconsistencies that economic data contain have necessarily had to be removed by deliberate judgment and that what remains is not only a complete macroeconomic picture but also a consistent one. Useful as an overall picture of the economy may be, it is by no means the sole object of the exercise. A fuli planning model was beyond the scope of our Swaziland experiment, but some discussion of the problems of development strategy can take place simply by reference to the matrix itself, while in other cases the matrix suggests a line of approach for farther analysis using data more specific to the task in hand. In fact, the latter aspect probably predominated in the report we presented to the Swaziland government, in which major issues such as the evolving employment situation, the customs union, financial matters, and the distribution of income were all analyzed in considerably more detail than that in which they could be treated in the matrix. Turning to the general questions of development strategy, which the matrix throws into relief, we can note that the domestic market is much too small for import substitution to be a viable major aim. Furthermore, if the object is to raise the living standards of the rural poor, more development of industry along existing lines is unlikely to have much impact. For one thing, the number of jobs that could possibly be created is far smaller than the number that would be required to make more than a tiny dent in the situation; and this, coupled with membership of the customs union, which has overriding advantages in providing an assured 122 CountryStudies Table 6.4. Paid Employment in Swaziland, 1971-72 (in thousands of workers) Administrative Semi- Production Activities and Technical Clerical Skilled skilled Unskilled Total Individual tenure farms 480 - 213 1,105 16,383 18,181 Forestry 80 39 42 459 3,152 3,772 Maize milling 12 33 - 4 160 209 Sugar milling 113 64 99 72 1,152 1,500 Cotton ginning 5 5 1 4 60 75 Timber pulping 73 78 172 312 561 1,196 Timber sawing 40 37 55 176 1,040 1,348 Meat packing and canning 4 5 13 7 183 212 Fruit canning 14 16 9 40 260 339 Iron ore mining 27 34 22 180 256 519 Asbestos mining 78 62 54 545 1,050 1,789 Coal mining 11 9 2 26 323 371 Other mining 4 4 3 9 157 177 Other manufacturing 72 83 47 276 730 1,208 Utilities (electricity and water) 44 48 35 115 289 531 Construction 41 49 229 632 2,133 3,084 Transport 131 114 114 383 844 1,586 Posts and telecommunications 56 117 23 12 79 347 Hotels, restaurants, bars 50 53 35 273 899 1,311 Wholesale trade and modern retail 217 190 53 213 1,860 2,533 Small Swazi traders - - - - 360 360 Education and health 2,821 133 64 260 1,242 4,520 Other service activities 200 232 82 184 700 1,398 Total 4,573 1,465 1,368 5,287 33,873 46,566 source of government revenue at virtually no cost in scarce administrative resources (but which places most industries located in Swaziland at a disadvantage compared with similar industries located nearer the large population centers of South Africa), probably means that the general policy of seeking to attract industry by providing sites and tax advantages could not achieve the desired results. We can also note, although not so clearly from the compressed matrix, that the tourist industry may be a useful minor source of government revenue, but its contribution to the general economic well-being of the country is relatively small. Although it creates a few jobs, they are mostly at rather a low level, and surprisingly it places few demands on domestic industry. Most of its requirements are imported, including fresh meat and vegetables which Swaziland could easily supply. It seemed to us that further development as a tourist economy was probably not in the country's long-term interest, particularly given the costs that providing for tourists imposes on a small country in social terms. The mining industry-iron ore and asbestos, but with the opportunity of developing much Table 6.5. Population and Income in Swaziland, by Household Type, 1971-72 Income (millions Average Annual Income Household Type Population of rand) per Capita (rand) Traditional 295,597 13.57 45.90 Individual tenure farms 81,467 10.89 133.67 Urban 72,295 32.21 445.54 1971-72 Swaziland, 123 bigger coal mines than existed when we worked-has clearly been an important source of government revenue either directly or through the customs revenues entailed in its relatively heavy imports of plant and equipment. Also, the need to build a railway to get the iron ore to Lourenco Marques (Maputo) provided Swaziland with an important route to the coast for other export products and an alternative to the expensive route through South Africa for bringing in imports bought on the world market. Equally clearly, however, Swaziland is unlikely to become a mining economy in the sense that the mines could not provide incomes for a sizable proportion of its people, which might then generate more domestic activity in the course of being spent. The high propensity to import and.the lack of viable import substitution oppor- tunities seem to ensure this outcome. Furthermore, the direct domestic linkages of the mining industry are naturally very small. There was, of course, no opportunity to look at one of the government's other strategies in quantitative terms, in that the RDA program was in its infancy when we worked. Its effects could be seen through the matrix if the proper data collection were organized and the matrix reestimated for later years. However, it links quite neatly with what seems to be, on first inspection, a favorable strategy for the government to pursue. This would be to expand the processing of agricultural products either by taking existing processing a stage further or by starting to process things that were being exported in a raw state. There seem to be several advantages in this approach. First, it provides a market and a focus for attempts to introduce more cash crops into the rural areas. Second, as a fertile, well-watered country, Swaziland competes well with others if the products are exported in a form or to markets where transport costs do not provide a barrier. Third, while processing's main domestic link is clearly to agri- culture, it provides at least as much opportunity for other domestic industries to grow up by providing the processing industries with inputs as by any other strategy we could envisage. We were encouraged to note, in fact, that Swaziland already had a small factory making card- board cartons, which were used to export the citrus crop. Fourth, it goes some way to satisfying the aspirations of the government and the people to create an economy in which there are opportunities for wage employment and urban living, as well as peasant farming in the rural areas. There are, of course, difficulties, but at least such a strategy seems feasible when examined in the broad terms allowed by the matrix, while several other strategies do not. A third use can be made of the information in examining the impact of major economic events or development projects. In fact, Jeffery Round, the team member primarily responsible for the production accounts of the system, returned to Swaziland some eight months after the initial work was completed to evaluate a proposal to build a large thermal power station and thus convert the coal resources into a commodity that could profitably be exported to South Africa through their electricity grid network. Similar studies could be made of the impact on various aspects of economic life of events such as the closure of the iron ore mine or a dramatic fall in the price of sugar. The advantage we see in having a SAM for this work, rather than merely using a set of input-output tables together with some supporting economic statistics, is that, apart from consistency, the matrix functions as a checklist to ensure that important aspects are not overlooked. In particular, it helps ensure that aspects such as employment generation and income distribution, which are central to the government's stated policy objectives, are kept to the fore. ISSUES OF ADMINISTRATIONAND ORGANIZATION The process of constructing the SAMand discussing it with Swaziland's administrators brought us into close contact with the administrative as well as the technical problems of providing quantitative advice in a developing country. In fact, part of the initial interest which the civil servants of the team had in the experiment was our belief that the approach might offer a 124 CountryStudies means of bringing statisticians and economists into closer contact and thus improving their joint contribution to solving administrative problems. It had been our experience across many countries that little use was made in government of the work done in statistics offices, apart from a formal chapter in the development plan and transmission of data to international organizations for inclusion in their publications. Stat- isticians often felt cut off from the process of giving policy advice, while some economists complained of lack of hard data on which to base their judgments, not only in planning but also in reacting to major economic events. I do not propose to offer a full discussion of the reasons for this state of affairs, either in Swaziland in particular or for all countries in general. However, it seems clear that in many cases the statistics collected were either not those wanted by economists or they were classified in a way unsuitable for economic analysis. In addition, statistical reports were often three or more years out of date by the time they left the statistics offices and even then there could be serious doubts about their reliability. Some economists argued that, even if reliable, up-to-date macroeconomic statistics were available, they would be of little use in project appraisal which formed the bulk of their work and for which highly specific data on particular organizations is usually required. Although many of the criticisms of the work of statistics offices are valid, we did not share the view that the economist concerned with project appraisal could afford to ignore the macro- economic picture. It may be reasonable to ignore the indirect effects of particular small projects on the grounds that the indirect effects of all small projects on an economy are broadly similar. However, this is scarcely true of a relatively large project such as the thermal power station and the development of the coal field in Swaziland. Furthermore, the assumptions that underlie the appraisal of even small projects seem to require a view of general development strategy if, for example, sensible shadow prices are to be used. This in turn requires a clear picture of the resources available and their present uses; and, as aid donors have become increasingly concerned with trying to improve the lot of the poorest people in the poorest countries, it has become more important when seeking aid to know who gets what, out of what activities. It also seems likely to me that in evaluating the uses of general economic statistics too much stress is placed on formal planning applications. Many of those who are cynical about planning are thereby enabled to question the worth of investing in comprehensive data frameworks. However, it is a feature of developing countries that single, unplanned events can have a dramatic impact on their welfare-world price movements, new trading arrangements, epidemics affecting crops and cattle, the actions of a powerful neighbor-and, even where domestic policy options are shown to be insufficient to deal with the problem, the country is in a much better position to seek help if the probable total effects can be anticipated early in quantitative terms. It is true that informed guesses can sometimes be made by experienced economists, but even these will benefit from being made within a consistent framework. There are enough examples of important mistakes to support the view that a reliable macroeconomic data framework represents a worthwhile undertaking if it can be achieved. We found nothing in our Swaziland experiment to alter the views expressed above. We showed that the resources needed to make a start are not out of the reach of most countries with technical assistance agreements, given the information that is already available in national accounts, and I have tried to demonstrate in the previous section that the SAM is a worthwhile development of such accounts. I would argue that it is worth organizing a sufficiently large input to put the first SAM together quickly, or the effort is likely to be overtaken by events and never brought into profitable use. For the same reason, I would counsel against waiting until the statistics office has produced a complete range of surveys before beginning. The SAM, like the national accounts to a lesser extent, provides a useful indication of where the most serious data deficiencies lie. Furthermore, survey data are never perfect and often one set will benefit from being seen in the context of other available information. Swaziland, 1971-72 125 However, it became clear from our experience in Swaziland that the mere provision of a completed SAM for a base year with some ideas on updating and recommendations as to the future pattern of statistical work by a team of outsiders was not in itself sufficient to stimulate the development in data collection and the application of economic analysis which we thought desirable. Since the social accounting approach is intended as a practical aid to improve capacity, careful thought will have to be given to integrate social accounting into the work of planning and statistics offices against the background of extreme pressure on skilled resources. Since the Swaziland experiment was completed, some members of the team have returned to southern Africa to help the government of Botswana construct a SAM. With British government assistance, a special post of "research statistician" has been established in the CSO and its occupant works closely with a macroeconomic unit in the Department of Economic Affairs in Gaborone. It remains to be seen whether this alternative approach will succeed, although the early indications are very encouraging. In the end, much will depend on whether aid donors give adequate continuous support to countries that are trying to improve their administrative capacity this way. In turn, whether countries are willing to make such an effort will depend to a great extent on whether they themselves can see any return for the expenditure of scarce resources. No one, however, should underestimate the distance still to be traveled in spite of the substantial step forward by the experience in Swaziland and SAM studies like it. 7 A Social Accounting Matrix for Botswana, 1974-75 C. C. Greenfield Botswana is a landlocked country about the size of France but with a population of only 700,000. More than two-thirds of the land surface is covered with Kgalagadi sand, of which a large part is the Kgalagadi desert. Together with Lesotho, South Africa, and Swaziland, Botswana is a member of the Southern African Customs Union (SACUA) and, before 1976, used the rand as its currency. The Bank of Botswana was established in July 1975 as a full central bank and a year later introduced the country's own currency, known as the pula, in place of the rand. The economy is open and relatively simple; cattle farming is the major traditional economic activity. The discovery and mining of copper and nickel and diamond deposits, coupled with high inter- national prices for beef, have brought rapid economic growth in recent years. These developments have meant that economic planning in Botswana can no longer concen- trate on sector and project issues, but must increasingly consider macroeconomic issues as well. This broader approach, together with the government's goal that all citizens should benefit from development, accounts for the desire to produce a SAM for Botswana. The particular features of the matrix that was produced are in part a result of the particular circumstances of the country. THE MATRIX AND ITS BROADFEATURES The Botswana SAM for 1974-75 was constructed in 1977 by a small team working in close association with officials of the government of Botswana.' The date of the SAM was chosen to include the latest year for which national income estimates were available at the time this study was undertaken. Appendix A, at the end of this chapter, sets out some logistic details of the exercise, while table 7.1 presents a version of the final matrix. The major features of the matrix will be discussed here by comparing it with table 2.1 of the United Nations System of National Accounts (SNA) (see United Nations Statistical Office, 1968). The latter is the format that has been most widely publicized and recommended internationally. The comparison is facilitated by table 7.2, which summarizes the principal features of the Botswana SAM and the SNA format. It is apparent from table 7.2 that the Botswana SAM represents a considerable condensation, and a simplification, of SNA table 2.1. In certain cases this result arose from a lack of data, but mainly it arose by choice. Lack of data specifically precluded incorporation of opening and closing assets and revaluations. Although in principle one would wish to include these, the possibility of obtaining sufficiently reliable data, particularly for tangible assets, would seem to represent an unattainable goal in most, if not all, countries. The remaining differences arose by choice, the most striking of these being the current account in the Botswana SAM, which can be compared with the production, consumption, and rest of the world current transactions accounts of the SNAtable. In the latter, production is subdivided 1. The production of the Botswana SAM was a team effort and thanks are due to all who participated. Particular thianks go to Harry Fell for his contribution to the design of the matrix. 126 Botswana 1974-75 127 into commodities and activities, so that a "make" matrix, showing industrial production by commodity, and an "absorption" matrix, showing the use of commodities by industries, are both separately estimated, with imports shown classified by commodity. In order to produce an input-output table from these matrices, the conventional method would be to assume either a commodity-based technology or an industry-based technology, and more exceptionally a mixture of the two, and thereafter to combine the make and absorption matrices into a single input- output table through mechanical algebraic manipulation. In table 7.1, however, there is only an input-output table, located at the intersection of the production activities rows and columns. This input-output table was estimated directly without first producing make and absorption matrices. It was possible to do this in Botswana because of the relatively simple nature of the economy, with only a few large business enterprises for which one can obtain individual estimates of expenditure associated with noncharacteristic production. For the remaining small enterprises, noncharacteristic production was trivial, particularly given the high degree of aggregation used (only seventeen groups of production activities were identified, condensed into fifteen in table 7.1). Although this approach might not be possible in more complex and developed economies, there is merit in employing it whenever possible. The size of the SAM is appreciably reduced in consequence, and ease of interpretation is thereby improved. Other differences in the current account are equally substantial. In the SNAtable 2.1, consump- tion is subdivided into expenditure and income and outlay. Expenditure is subdivided by purpose- for households, government, and private nonprofit bodies-while income and outlay are divided into the subheadings of value added, institutional sector of origin, form of income, and insti- tutional sector of receipt. It would take too long to describe the logic of all the flows arising from these classifications, but let us trace value added through them as an example. Value added is shown as being derived from commodities (protective duties), commodity taxes, industries, and so forth. It is then paid to the institutional sector of origin, which in turn pays its incomings to the form of income account. Form of income then pays to accounts for the institutional sector of receipt and the rest of the world. And finally, institutional sector of receipt pays for its incomings to the consumption expenditure, form of income, and capital finance (savings) accounts. This is hardly a picture that one would visualize as describing the flow of value added through an economy, largely because of the inclusion of the form of income accounts, which in part duplicate and in part disaggregate the value added accounts. In the case of the Botswana SAM, all of the above accounts have been replaced by accounts for factors and institutions. The logic of the flows is that factors receive income from hiring their services to production activities, government, and the other accounts. This income is then channeled to the institutions owning the factors, namely, households, enterprises, government, and the rest of the world. Transfers between domestic institutions then occur, such as payments of interest, dividends, and direct taxes, and are shown in the institution-by-institution subma- trix. The resulting income of institutions is then spent on domestic production, imports, and other payments to the rest of the world, and the balance is saved, as shown at the intersection of the institution capital account rows with the institution current account columns. While some detail has been lost in this presentation as against that of SNA table 2.1, it is suggested that the gain in simplicity, with a description of flows that is intuitively clear, more than compensates for the minor loss of detail. The differences in presentation of the capital account, apart from necessary omissions in the Botswana SAM already mentioned, are more a matter of detail than principle. Thus gross capital formation has been presented in the Botswana SAM at the intersection of the institutions columns with the rows for production activities and the rest of the world. Capital transfers are shown in the institution-by-institution capital account submatrix. The treatment of financial claims is essentially the same as that in table 2.1 of the SNA, and the basic identity-that Table 7.1. A SAM for Botswana, 1974-75 (in million of pulas) Current Account Factors Institutions 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 Adinin. & prof. employees sFpat 02 Locxal, 03 } Skilled manual & clerical employees ELal 04 2 Unskilled employees 05 Self-employed 06 Operating surplus 07 Depreciation 08 Periurban households 09 0.1 0.5 9.7 0.2 0.2 0.2 Urban hhs., high-density housing 10 0.2 8.2 13.5 0.9 0.2 0.2 0.1 0.3 @ Urban hhs., servants' quarters 11 0.3 3.6 0.1 0Urban hhs., med-density housing 12 0.5 0.9 4.2 3.5 4.1 i Urban hhs., low-density housing 13 0.3 11.7 1.3 5.4 0.4 Rural hhs., R <10 cattle 14 2.5 0.1 5.7 0.4 3.1 16.3 0.3 2.4 0.8 1.5 0.3 1.0 0.2 1.5 Rural hhs., 10-80 cattle 15 2.0 3.2 0.7 29.5 2.2 2.5 0.7 1.2 0.3 1.0 0.2 1.0 0 D Rural hhs., >80 cattle 16 0.6 0.9 6.6 9.0 1.0 0.2 Migrant workers abroad 17 18.6 44 2 . Major mining enterprises 18 2.1 7.8 e Other private enterprises 19 11.8 5.9 0.1 0.3 0.1 0.3 0.8 0 Parastatal enterprises 20 7.7 1.4 Private nonprofit & local govt. 21 0.3 0.4 0.1 04 0.6 0.5 1.6 0.6 0.6 Central Transfers 22a 0.1 1.7 0.1 0.2 0.6 2.0 0.1 0.1 0.1 12.9 Goverlent Sales 22b 0.3 0.9 0.7 0.4 0.4 Indirect taxes 22c 0.1 0.1 0.1 0.2 Freehold farms 23 0.3 Traditional farcos-livestock 24 4.0 9.0 2.1 E Irwlitional farms--rops 25 0.1 2.3 3.4 0.5 Traditional farms-other activ. 26 0.2 0.2 3.5 1.6 0.5 M lining (diamnond, copper, other) 27 Botswana lleat Commission 28 Other manufacturing 29 0.4 1.2 02 0.4 0.1 3.1 2.3 0.8 0 Water 6 electricity 30 0.1 0.2 0.1 0.3 0.1 0.5 6 Construction 31 6 Wholesale & retall trade 32 1.5 3.3 0.6 0.8 1.0 34 2.0 1.0 t Hotels, bars, restaurants 33 0.2 0.2 0.1 0.3 Rai transport 34 0.1 0.1 0.1 0.1 0.1 Other transport & communications 35 0.1 0.1 0.1 0.2 0.1 0.2 0.2 Services, n.e.s. (incl. dwellings) 36 0.6 0.9 1.3 2.9 2.2 0.9 Personal & household services 37 0.1 0.4 0.1 0.5 0.3 0.7 1.2 Pri ce Effects 38 0.9 1.9 0.4 0.6 0.9 1.8 1.8 1.6 39 0.1 0.1 0.2 0.2 0.2 0.4 0.2 ROW Goods & noifator services 40 5.7 11.1 2.1 3.7 4.5 10.7 10.9 9.0 1.6 0.7 IFactor services & transfers 41 0.7 3.1 0.6 5.4 -9.3 ROW Capital transactions 42 Households 43 0.2 0.4 0.1 0.8 1.3 2.2 8.0 5.5 4.6 8 Major mining enterprises 44 5.9 i Other private nonfinan. & local govt. 45 S Nonfinan. parastatal enterprises 46 - Banks 47 Other enterprises, irncl. BDC 48 Central government 49 Domestic currency 50 Bank deposits 51 Other domestic deposits 52 Treasury bills 63 Bank advances 54 g c1 9 Other domestic lending & Short term 85 3 borrowing Long term 86 o Common customs area account 57 Short term 58 Other foreign lending LT from govts. 59 6 borrowing LT from orgs. 80 LT from private 61 Offlctia reserves 82 U_nallocated 83 Errors & omissions 84 Total 65 8.2 13.3 42.8 12.5 39.5 53.6 33A 23.2 10.9 23.6 4.0 13.4 19.8 36.3 44.0 27.1 23.0 11.0 Note: ROW slglltes rest of world; n.e.s. signifles not elsewhere specied; BDC is Botswana Development Corporation. 128 Current Account Capital Accounit Price Production Activities Effects ROW ROW Institutions 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 48 46 47 48 49 1.7 3.1 0.1 0.1 01 0.1 0.1 0.3 0.1 05 1.0 3.8 0.1 1.2 0.4 0.4 0.9 0.9 0.5 0.1 1.8 1.8 04 2.2 11.0 0.7 1.6 0.1 1.4 0.3 0.8 0.8 1.0 1.8 0.2 0.4 1.7 01 186 0.2 1.9 0.4 2.7 0.1 1.1 0.3 3.1 0.9 0.1 0.8 085 0.4 0.2 1.7 1.4 1.3. 1.8 0.1 3.6 1.4 3.4 0.4 9.4 4.3 0.8 0.8 0.4 08 3.6 4.4 1.3 26.9 5.2 6.1 2.9 0.8 7.9 0.3 0.1 20 0.1 3.7 9.8 0.4 3.4 0.2 34 4.6 4.7 0.1 -1.9 086 4.1 0.4 0.7 0.7 1.2 0.1 7.9 0.5 0.7 1.0 2.0 1.0 0.2 0.5 1.4 44 04 0.1 0.6 86 0.1 0.9 8.9 0.3 1.2 36 0.8 0.2 02 8.3 0.1 0.2 0.1 0.7~ 4.3 7.4 12.9 0.1 0.2 0.1 0.1 0.1 0.1 0.1 0.5 0.8 0.1 0.2 0.2 16.4 10.5 0.4 03 01 0.9 0.3 18.7 0.3 0.1 13.0 0.7 1.0 -1.9 1.2 0.3 0.8 0.9, 31.0 3.0 0.2 0.2 1.1 32 1 2.7 0.4 585 0.1 0.2 0.3 9 9 [0.5 0.9 8.8 0,8 0.1 0.5 0.2 0.2 0.1 0.2 0.2 0.1 0.1 0.5 6.2 06 1.8 8.8 8.9 7.4 0.8 0.1 19.7 0.1 0.3 0.5 0.3 1.3 1.1 2.6 0.8 4.8 0.3 0.4 0.9 0.3 0.5 13 0.7 0.2 0.8 1.7 0.2 0.4 0.2 0.7 0.3 13 0.1 0.2 0.2 0.3 0.1 0.9 0.8 0.1 0.1 1.1 085 0.1 0 2 14.9 0.8 0.2 7.0 0.1 0.2 0.1 0.6 0.1 0.5 1.4 0.3 1.1 0.1 0.5 0.1 0.3 0.1 1.1 0.4 0.1 1.8 1.9 0.1 0.1 0.5 1.8 0.3 0.4 0.6 0.2 0.8 0.6 1.1 0.7 0.1 0.7 0.3 2.0 0.2 0.2 0.8 0.1 0.2 0.1 0.4 01 02 0.1 01 0.2 1.1 2.5 0.3 0.8 0.2 0.3 0.4 0.1 0.2 0.1 0.1 2.2 0.1 0.2 0.2 0.2 0.3 12 0.5 0.9 56 2.7 11 0.2 16.2 1.4 10.1 3.0 19.9 7.1 1.3 19.7 4.7 1.7 2.3 -80.9 1.4 15.2 14.1 2.0 0.1 0.4 3.8 6.1 0.2 1.4 -11 5 8.5 0.1 1.8 2.7 1.5 1.0 0.3 0.2 19.1 23.5 0.1 0.1 4.8 0.1 0.4 0.4 3.8 1.9 1.0 0.1 0.5 12,9 2.8 0.8 10.9 0.4 2.1 0.9 2.8 -0.1 0.8 0.2 3.1 20.8 -8.2 11.7 -1.7 8.2 0.8 1.8 8.0 4.6 35.2 1 ~~~~4.4 2.4 30.8 10.3 11.6 64.8l 12.9 44.4 6.2 8.1 87.4 36.8 28.8 11.1 81.2 33.8 3.8 20.8 12.6 17.6 10.8 18.4 6.2 1477 1 28.1 41.4 .32.0 18. 17.2 8.8 88.9 (Table continues on the following page.) 129 Table 7.1. (Continued) Capital Account o 0 o 0 e Financial Claims w 0 Total 50 51 52 53 54 55 56 57 58 59 60 61 62 653 64 65 Local 01 6.2 Admin. Y prof. employees ExLal 02 13.3 Local 03 42.6 c Skilled manual & clerical employees E 02.5 S I Expat. 04 12.5 N Unskilled employees 05 39.5 Self-employed 06 53.6 Operating surplus 07 3354 Depreciation 08 23.2 Periurban households 09 10.9 Urban hh1s., high-density housing 10 2536 Urban hhis., servants' quarters U 11 4.0 U Urban hhs med-density housing 12 13.4 Urban hhs., low-density housing U 13 19.8 6 Rural hhs., <10 cattle 14 36.5 Rural hhs., 10-80 cattle 15 44.6 3 Rural hhs., >80 cattle 16 27.1 3 Migrant workers abroad 17 23.0 Major mining enterprises 18 11.0 , Other private enterprises 19 30.5 0 Parastatal enterprises 20 10.3 Private nonprofit & local govt. 21 11.6 Transfers 22a O*ventrnmt Sales 22b 64.6 G Indirect taxes 22c l a Freehold farms 23 l59 Traditional farms-livestock 24 44.4 ; Traditional farms-crops 25 6.2 Traditional farms-other activ 26 6.1 e MIning (dliamond, copper, other) 27 57.4 Botswana Meat Commission 28 36.5 < Other manufacturing 29 25.5 C Water & electricity 30 11.1 3 Construction 31 51.2 i Wholesale & retail trade 32 33.6 H Hotels, bars, restaurants 33 3.6 Rail transport 34 20.5 Other transport & communications 35 12.6 Services, n.e.s. (incl. dwellings) 36 17 6 Personal & household services 37 10.6 58 16.4 Price Effects 39 62 ROW Goods & nonfactor services 40 91 147 7 Factor servioes & transfers 41 ROW Capital transactions 42 5.4 -6.2 5.6 Households 43 0.2 10 0.5 25.1 5 Major minng enterprlses 44 -1.1 11.5 51 21.8 , 41.4 3 Other private nonfinn. & local govt. 45 8.4 2.0 0.9 5.1 11.2 32.0 2 Nonfinan. parastatal enterprises 46 2.2 1.5 9.9 1.4 16.5 3anks B 47 20.0 3.5 2.2 -8.6 -1.0 17.2 Other enterprises, incl. BDC 48 2.9 0.1 1.4 1.7 -01 6.5 Central government 49 0.6 1.2 6.0 4.6 0.1 17 539 Domestic currency 80 5.4 Bank deposits 81 20.0 3 Other domestic deposits 52 2.9 Treasury bills 53 0.6 Bank advances 54 i 1.| Other domestic lending & Short tero S 589 3 borrowing Long term 56 24.51 E Common customs area account 57 1 -6.2 S(hort term 581 17.3 Other foreign lending LT from govts. 691 6.0 & borrowing LT from orgs. e0 4.6 "LT from private 61 35.2 Official reserveE r 2 Unellocated 63 = Errors a omissions 84 l Total e81 5.4 20.0 2.9 0.6 11.0 8.9 24.5-6.2 17.3 6.0 4.6 35.2 81 7.7 Note: ROW signifies rest of world; rie.s. signifles not elsewhere specified, BDC is Botswana Development Corporation. 130 Botswana 1974-75 131 Table 7.2. Summary Structure of the Botswana SAMand Table 2.1 of the United Nations System of National Accounts Botswana SAM SNA Accounts Accounts Current Account Opening Assets Factors Financial claims Institutions Net tangibleassets ProductionActivities Production Rest of the World Commodities Activities Capital Account Consumption Rest of the World Consumergoods purposes Institutions (expenditures) FinancialClaims Income and outlay Accumulation Increase in stocks Fixed capital formation Financial claims Capital finance The Rest of the World Current transactions Capital transactions Evaluations Financial claims Net tangibleassets Closing Assets Financial claims Net tangibleassets Note: SNA accounts are as summarized in table 1.6 of the System of National Accounts (United Nations Statistical Office, 1968). savings plus change in financial liabilities is equal to capital formation plus change in financial assets-is retained. We will return to the subject of financial claims later, however, because although the Botswana SAM does not diverge from SNA recommendations in this respect it is a feature that differentiates the Botswana SAM from other SAMs presented in this volume. DETAILED FEATURES OF THE MATRIX The detailed classifications employed in the matrix naturally differ from those used in SNA table 2.1. The latter table is given in the SNA simply as an illustration and is not concerned primarily with questions of income distribution. The need to consider issues of income distri- bution was, however, one major reason for preparing the Botswana SAM, and this led to the classification of factor income by broad skill group and by local or expatriate labor, and the breakdown of households into rural and urban groups, with further subgrouping among these. For questions of income distribution, it is not sufficient to have a SAM expressed in money 132 CountryStudies values, as presented here. Additional information is required for at least some of the cells on the physical values to which the money values relate. For example, although one can estimate from table 7.1 the total income of high-density and low-density urban households, this estimate is only half of the picture as regards income distribution-one also needs to know the number of households or persons to whom these incomes apply. Consequently, supporting estimates of relevant quantities were also produced in Botswana. The actual classifications employed in a SAM are crucial to its potential usefulness. No one set of classifications can be ideal for all purposes and, unless there are abundant resources to produce a set of SAMs with different classifications, a compromise has to be made to select that set most likely to be required for major policy purposes. The classifications used in the Botswana SAM were arrived at after lengthy discussion with officials in that country. There are only two points to note, namely, the small number of group- ings used for production activities and the fact that a different classification of institutions was used in the current and capital accounts. The classification of production activities was proposed by officials in Botswana. It does not accord with the normal requirement that the members of any given industrial subgroup should have as similar a production technology as possible. The other manufacturing group is partic- ularly heterogeneous. This classification must add some unreliability to any prediction employ- ing the input-output table, but it is unlikely that this will be significant because, even with the fifteen groups given in the SAM (as opposed to the seventeen for which estimates were made), the degree of interindustry dependence is slight, a reflection of the present stage of economic development in Botswana. As regards the different classification of institutions between the current and capital accounts, households have been aggregated as a single group in the capital account simply because there was no information on financial claims for the household groups used in the current account. A different classification of enterprises was employed, however, because it was found that those groupings considered the most useful in the current account were not the most useful for understanding capital account transactions. In particular, it was felt essential to identify banks and other financial institutions in the capital account. Price effects represent an important and unusual feature of the Botswana SAM, although the novelty is partly a matter of terminology since they correspond to "commodity taxes, net" in SNA table 2.1. Some reasons for our preferred terminology are set out in appendix B to this chapter, together with some discussion of associated technical issues. Here it may suffice to note that table 7.1 expresses commodity balances in what the SNA refers to as basic prices and that "commodity taxes, net" or "price effects" have two separate accounts in the table, so that some implications of Botswana's membership of SACUAcan be shown separately. Indirect taxes levied in Botswana by central and local government have been charged in rows 21 and 22(c) to those production activities and households paying them. Import duties are levied by SACUA, rather than the government of Botswana, and have already been charged and hence included in the price of imports, when they reach the border of Botswana. In order to show the c.if. value of imports (that is, cost plus insurance and freight), as it is conventionally understood, it is therefore necessary to remove the estimated duty charges from the cost component. The true value of this content in any given time period is not known, and reliable estimation would require appreciably more information than was readily available. What is known, however, is Botswana's entitlement to revenue from SACUAfor her imports during the period, the calcu- lation of which is fairly complex. The actual estimate of this amount, P16.4 million, has therefore been taken as the duty content of imports, deducted from the value of recorded imports, and entered in row 38. The corresponding revenue payable to the Botswana government has been entered in the row 22(c)/column 38. Insofar as actual receipts from SACUA differed from Botswana's accrued entitlement, a compensating entry is made in the capital account. The Botswana 1974-75 133 remaining component of price effects, row and column 39, relates to these in the more conven- tional setting, that is, estimated as arising from locally raised indirect taxes, plus the additional price raising effects of SACUAduties through interindustry transactions. A final point to note is that intraindustry transactions have been netted out to zero in the input-output submatrix. For the most part these were trivial, the only notable exception being the agricultural sector. The data are available, however, and it is straightforward to convert the matrix and its inverse to one that includes intraindustry intermediate consumption: in retrospect, it would seem preferable to have retained these diagonal elements of the input- output submatrix. CONCLUDING REM-ARKS ON THE FORMAT OF THE MATRIX Many changes in format have been made in the Botswana SAM as compared with table 2.1 of the SNA.The end result is a matrix containing far fewer cells that are blank by definition, and one that is much easier to understand and therefore use. It was considered essential to produce a matrix that could be generally understood by economists and others concerned with planning the economy. Some information is, of course, lost with such condensation, and it is a matter of judgment as to how far one should go. Our own inclination, where more detail is required, would be to provide it in submatrices outside of the main matrix. It would be wrong, however, to conclude with an impression that one has nothing but criticism for the SNA format. Quite the reverse-one has nothing but admiration for the pioneering and meticulous work recorded in the SNA,which provided an invaluable guide and source of reference in preparing the Botswana SAM. FLOW OF FUN-DS Two members of the team went to Botswana in June 1977, to obtain prior agreement on the classifications to be used, the overall structure of the SAM, and the analyses required, so that when the other five members arrived work could commence immediately on estimating the matrix. This approach proved invaluable, as can be seen in the case of flow of funds. Although initially we had not intended to include a flow-of-funds section, we soon became aware of the considerable interest in monetary and other financial matters among local officials. Local offi- cials insisted that flow of funds be incorporated, even though neither the central bank nor the pula had existed during 1974-75; otherwise they would not consider the SAMworth producing. It was not their intention that the first SAM should be an end in itself, but rather that it should be the first in a regular series. In retrospect, we are grateful to them. Their main argument was that national accounts estimates tend to become available too late for many purposes and, moreover, to contain certain data that are, to put it mildly, of dubious reliability. Through estimating financial flows it was possible to obtain some check on the real flows estimated in the national accounts. Many of the financial flows required were generally available with greater frequency, timeliness, and accuracy than the national accounts. Besides providing a consistency check on the national accounts, the flow-of-funds accounts also provide a way of looking at developments in the economy, which is of value in its own right. A system of financial programming was being introduced at this time with assistance from the International Monetary Fund (IMF). This was a short-term forecasting exercise, covering a period of up to one year ahead, based on financial variables. A regular data sheet was being prepared, called the "framework," as a basis for the system, and it was necessary that the treatment of flow of funds in the SAM should be compatible with it. Table 7.3. Reconciliation of Real and Financial Flows Private Sector (:ovu rnie n t B;inking- fiore i-gn Sector Sea tor Sector. Households l nLer- G B F prises hIf pE Savings/Investment Il) (1 - - G) (S - I) i iB) Id (M - X) = 1'inancial Flows: A L A L A L A L A L Money Market A (Including quasi.-..oney) A Foreign Assets Market A (lForeign Reserves) A A Rinks loans mark}et (Credit fromn banks, A AA A domes tic) Government bonds market (domestic) A A A A A S-tock market (domnestic & foreign) A A A A A A A A Capital flows market A A A A A A A A A a) Priva-te I I b) Official A A A A A A Note: A signifies assets, L signifies liabilities, A signifies change during time period. An increase in liabilities or a reduction in assets is positive. Entries in the foreign sector are recorded from the point of view of the foreign sector. In general, cells left blank will not have entries. Botswana1974-75 135 The basic approach can be summarized as follows. Taking the conventional identities: (7.1) Y = Cp + Ip + G + X -M (7.2) Y = CP + T + Sp where Y is national disposable income, C, is private consumption expenditure, Ip is private investment expenditure, G is total government expenditure, T is total taxes collected by govern- ment, X is total current account balance of payments receipts, M is total payments on current account balance of payments and Sp is private savings.2 Then (7.3) CP + T + Sp = CP + Ip + G + X- M or (7-4) (Sp - Ip) + (T -G) + (M -X) = . Since government current expenditure plus government investment is equal to G and govern- ment saving is equal to T minus government current expenditure, it follows that (7.5) (T - G) = (SG - IG)- Given (7.4), it is possible to check the balances on real flows in the economy derived from conventional national accounts estimates against the corresponding financial flows. Table 7.3 gives a summarized illustration of this approach and could be broken down into as much detail as is desired or feasible. In the table, the private sector has been split into households, non- financial enterprises, and banks. The sum of the savings/investment row, that is, the balance of real flows derived from national accounts, will be zero in accordance with equation (7.4), although any individual entry may be positive (or negative), indicating an excess (or shortfall) of saving over investment in a partic- ular sector. The sum of any row in the financial markets will also be zero because any change in assets held by a given sector will be matched by a corresponding change in liabilities of some other sector or sectors. Equally, the sum of change in assets and liabilities in the column for any sector will be identically equal and opposite to the savings/investment balance, showing how that balance is disposed of in the case of a surplus or financed in the case of a deficit. In entering data in table 7.3, failure to achieve these identities indicates errors or omissions in the data which, given knowledge of sources, could also be a useful guide to where the errors are probably located. Even without errors, the table is of value because of the importance of financing the real balances in an economy. It is worth noting that the foreign sector is an integral component of the system, its financial flows representing capital account movements in the balance of payments. This model for reconciliation of real and financial flows fits into the format of table 2.1 of the SNA for dealing with financial claims. The format adopted in table 7.1 is essentially the same as that of the SNA,and translation between the arrangement of table 7.3 and table 7.1 is more or less immediate. The only change required is in table 7.1, where the sum of savings and the change in financial liabilities is totaled: this should equal, rather than cancel out, the sum of investment and change in financial assets. Consequently, the signs for change in financial assets are the reverse of those in table 7.3 (an increase being positive in table 7.1), as are the signs for investment and current balance of payments receipts. Our experience with including flow-of-funds data in the SAM will be covered in the next 2. More precisely, G is expenditure of general government (including "government purposes"); private consumption expenditure and investment includes government enterprises; and T is the net income received by general government from taxes, property income, and transfers. 136 CountryStudies section, but before proceeding there are two further points to mention. First, thirteen categories of financial claim are identified in appendix A (although there is no entry for official reserves, because at the time the SAM was developed Botswana was using the rand and did not have a central bank). These, however, are a summary of the number of categories actually used in the detailed estimation process. In all, eleven categories of domestic claim and twelve categories of foreign claim were employed. These were the categories that would be required in subsequent analytical work, and, while the distinction between domestic and foreign claims was not essen- tial, there was a major advantage in confining the majority of balance of payments transactions to a clearly identified section of the flow-of-funds matrices. In fact, for each entry in row 42, columns 50 to 62, the breakdown of the domestic counterpart change in assets is found in the corresponding rows 50 to 62, against columns 43 to 49. The same applies to entries in column 42, by rows 50 to 62, as against rows 43 to 49 by columns 50 to 62. The second point, although minor, is raised to avoid possible confusion. The current account balance of payments deficit has not been shown as a single number, as is normally done, but has been allocated between the change in external assets and liabilities (row 42/columns 40, 41, and rows 40, 41/column 42) and includes net errors and omissions. The effect is that the totals of the rest of the world capital transactions, assets, and liabilities, are identically equal to zero. ESTIMATION AND BALANCING The problems of estimating the entries in the SAM largely hinged upon availability of data. A surfeit of data could be just as troublesome as a shortage of data, if there were alternative sources providing conflicting estimates of the same items. Apart from certain items in the flow- of-funds section, however, the problems in Botswana arose mainly from lack of data. Never- theless, many different sources were available and were used. The most serious data inadequacies were in respect to households, particularly urban house- holds. The latest available information on household expenditure was for 1968-70, and it had to be used for lack of anything better. Fortunately, tabulations of the data were available which largely corresponded with the classification of households employed. The expenditure survey had also obtained information on savings and, although these were clearly understated, they were used as a guide to the distribution of savings between different types of households. For urban areas, there were no estimates in any official surveys undertaken of the channeling of factor income by broad skill group into households, by type of household. The only survey that had attempted to do this was one undertaken in Gaborone by the University of Botswana and Swaziland. Because of a technical fault in the coding of this survey, however, the attempt had failed, and only the marginal totals of income by type of household and broad skill group were available. A matrix was built up from these, using judgment and an RAS balancing technique. The resulting distributions were then taken to be true for all urban areas. In consequence, no consistency check was possible for household income and expenditure; the estimates auto- matically balanced because of the methodology employed. For rural households, however, the position was far better than one would normally expect. A major survey of income in rural areas had been undertaken in 1974-75, the main results of which were published in Rural Income Distribution Survey (Botswana, Ministry of Finance and Development Planning, 1976a). This detailed and careful study, undertaken with assistance from the World Bank, attempted to estimate all aspects of rural income, in cash and in kind. In many ways, it could be regarded as a model for developing countries interested in undertaking such studies. Tabulations of the survey were prepared for the desired grouping of rural house- holds, so that all the required information on the channeling of rural factor incomes into Botswana 1974-75 137 different types of households was available. In addition, the survey obtained information on transfers of income between rural and urban areas. Although the estimated value of consump- tion of the respondents' own production was covered, cash expenditure was not; for this infor- mation, recourse had to be made to the 1968-70 expenditure survey. For the production activities segment of the matrix, the main sources of information were National Accounts of Botswana 1974/75 (Botswana, Ministry of Finance and Development Planning, 1976b); supporting surveys, particularly the 1974/75 Census of Production and Distri- bution; and work files. Botswana has a very good set of national accounts estimates, covering production, income and outlay, and capital finance accounts, and the report is well presented. The breakdown of compensation of employees by broad skill group is not, of course, available in the national accounts; although it is covered in the annual employment inquiry, the results of the particular relevant inquiry were clearly unreliable for certain sectors. Consequently, for these sectors, projections were made of the 1972 Manpower Survey (Botswana, Ministry of Finance and Development Planning, 1973). Equally, the national accounts do not include an input-output table, and one had not been previously prepared for Botswana. Even though the input-output table produced is relatively small, its estimation was a major task, requiring a detailed knowledge of industry in Botswana and the use of a multiplicity of sources. Fortunately, the member of the team responsible for this section of the matrix had the requisite knowledge, having been responsible for producing both the 1973-74 and 1974-75 national accounts estimates. The main source of information for enterprises and central government in the current account was the national accounts, supplemented in certain cases with information from the accounts of individual enterprises. Savings, taken as a residual from the current account, had to be reallocated in the capital account because of the different classification of enterprises employed there. Overall, there were no major problems. Transactions with the rest of the world, however, did present problems because only recently have there been attempts to produce a comprehensive picture of this aspect of the economy. The collection of foreign trade statistics had not been essential data to Botswana as a conse- quence of its membership in a customs union. In the early years of collecting such statistics, including 1974-75, imports were understated quite substantially. The first balance of payments estimates were produced in 1976, in respect to 1973-74 with tentative estimates for the calen- dar year 1975. These formed the starting point for the 1974-75 estimates in the SAM. Not all of the entries were based on firm data sources because so little work of a continuing nature had been conducted. When the estimates were first completed, there was a very substantial balancing item for net errors and omissions, the only satisfactory explanation for which seemed to be that imports had been considerably underestimated, particularly in respect to expenditure on them by households. Customs Department officials were themselves convinced imports had been understated, and they had independently estimated the probable magnitude. Consequently, the figures for imports were increased in line with this estimate, which left a more reasonable balancing item for errors and omissions. There were several stages in estimating the flow-of-funds submatrices. The first stage sepa- rated out those institutional sectors for which reasonably good balance sheet data were available for both 1974 and 1975. In certain cases, the financial year-end for such sectors was not June as required by the matrix; this presented a significant problem only in the case of the major mining enterprises. The second stage involved taking entries derived from first-stage sources and identifying with whom the transaction had taken place, which was often quite easy. For example, commercial bank data on deposits and loans were readily classified by the corresponding lending and borrowing sectors; Post Office Savings Bank deposits are held almost entirely by households. Thus it was possible to build up a good deal of information on the flow-of-funds entries for 138 CountryStudies sectors for which balance sheet data were not available. Consistency checks arose when balance sheet inforimation covered both sides of a transaction. If one set of information related to June and the other did not, it was assumed that the discrepancy was due to timing, and the June figures were accepted. In the case of data on transactions between the commercial banks and parastatals, the banks' data were preferred because they related to the desired time period. An inevitable consequence, unfortunately, was that balance sheets of rejected sectors no longer balanced. In certain cases discrepancies were found between balance sheets covering the same time period, although timing could still be a reason if certain items were "in transit." By the end of the second stage all domestic financial claims were balanced except for currency and trade debtors/creditors accounts. Figures entered for the latter were notional and cannot be improved until considerably more balance sheet information is obtained regularly from private enterprises. The problem of estimating change in currency holdings arose because the rand was still in use during 1974-75; it was possible to estimate the item only by making certain assumptions, the reliability of which could not be confirmed. The final stage in estimating the flow-of-funds section was the incorporation of items from the rest of world capital account. When all of the above work had been completed, a first draft of the matrix was produced. It showed that the accounts for factors and current accounts of institutions balanced. This result was achieved because in certain cases, we had data with accounting accuracy while in the remaining cases, referred to above, the methodology of preparing the estimates ensured that balance was achieved, so there was no consistency check on them. Production activities did not balance, and in a few cases there were fairly substantial errors. Financial claims balanced, but there were substantial errors in the capital accounts for institutions. In other words, the estimates of savings and investment were not compatible with the independently estimated change in financial claims by institution. If one accepted the financial claims estimates, then household savings were substantially underestimated and the savings of all other institutions were more than correspondingly overestimated. The balancing procedure then followed was, first, checking entries for errors and, then, in the case of production activities, making adjustments by judgment, some of which also served to reduce errors in the capital account. Such adjustments were made until all remaining errors in the production accounts were within 5 percent of the mean total of corresponding rows and columns, with one exception, for a sector that was in any case small in absolute terms. In the case of the households' capital account, the financial claims data and the capital formation estimates were treated as being correct, which meant that the household savings figure was adjusted accordingly. This estimate of household savings appeared to be considerably more reasonable than the original estimate, which had largely depended on the Household Expenditure Survey results. The methods employed to raise household saving (mainly through reallocation of factor income between "operating surplus" and "self-employed") also served to reduce savings of enterprises, as was required, so that balance was also achieved for the account for other private nonfinancial institutions and local government. The account for major mining enterprises had already been balanced. For the remaining institutions, we were left with errors that were classified into two types: "unallocated" and "errors and omissions." The unallocated errors are those for which we were able to identify the cause. These errors arise from incon- sistencies between the accounts of different institutions, for example, because of the timing of transactions or the valuation of assets and liabilities. The errors and omissions are pure errors for which we could not find any satisfactory explanation, all of which, however, were relatively small. It is perhaps surprising to find such errors in the case of the central government account (row 49), but the reason again is due to differences in timing, since government accounts are not available for the July/June accounting year. The final stage in balancing was to use the mechanical, simple R AStechnique on a computer. The objective was to obtain balance in the production activities accounts, and the method of Botswana 1974-75 139 application ensured that estimates which one was unwiUling to have changed were left untouched, particularly unallocated errors and errors and omissions. Although it seemed reasonable to make mechanical adjustments to data where the adjustments so made were certainly within the margin of likely error of the data, we could see no justification for forcing errors out of existence in data that basically have accounting-type accuracy. In the end, questions that naturally arise are: Should the SAM have been produced, given the deficiencies noted in data availability? Given these deficiencies, should the SAM be used? And was the incorporation of the flow of funds in fact worthwhile? The short answer to all of these questions is yes. Until a SAM has been produced for the first time, and the attendant decisions on classifications, for example, have been made, it will almost certainly remain the case that availability of data is inadequate. Producing the SAM pinpoints deficiencies and inconsistencies and gives a clear guide to future statistical requirements. It also demonstrates to economic planners and other users the potential value of the SAM as a planning tool, hence gaining their support for its maintenance, updating, and improvements. In Botswana, measures to rectify the deficiencies in the household sector (noted earlier) are already inhand, and anupdated SAMwas produced in 1978 for theyears 1976-77. The intention is to computerize the SAM, so that as new and better information becomes available updating should be relatively easy. It should be noted, however, the SAM does not call for surveys that would not otherwise be required; virtually everything is required in any case, for other purposes, and the SAM is simply a further reason for obtaining the data. As to whether the SAM is sufficiently reliable for use, given the amount of judgment that had to be employed in estimating it, the ready answer is that plans and decisions that can be assisted by the SAM have to be made in any case, and judgments not constrained to consistency by the formal structure of the SAM will still be made, possibly subconsciously. It seems preferable to recognize this and to make judgments in the light of all available information and knowledge of likely causes, but explicitly and consciously within a SAM framework. In the case of Botswana, inclusion of the section on flow of funds resulted in more informative accounts for the rest of the world, and hence permitted consistency checks on the real flows that would not otherwise have existed. These checks led to changes in estimates which improved reliability. At the same time, however, it must be recognized that there were elements in the flow-of-funds section which were just as difficult to estimate and just as unreliable as certain components in the real flows section, particularly trade debtors and creditors. Furthermore, the flow-of-funds estimates were particularly sensitive to timing and valuation questions. Massive flows could occur virtually overnight, and unless one had all major balance sheets for the same time period, reconciliation was very difficult, if not impossible. These difficulties may reduce much of the potential value of the flow-of-funds accounts. CONCLUSION As has already been stressed, we regard the 1974-75 SAMas simply the first in a continuing series. Many improvements will doubtless be made to it in time, but one that can be recommended now is that a SAM at purchaser prices be produced. At present, the SAMhas been estimated at both producer and true basic prices, each of which has its own uses in different types of analysis. It is also true, however, that there are possible uses for which purchasers' valuation would be the most appropriate. The possible uses of the SAM have not been of concern here. Nevertheless, one would not wish to conclude without some reference to this subject. A number of analyses on a wide range of topics, such as the impact of wage increases and drought on the economy, are being under- taken or have already been completed at the University of Warwick for the government of 140 Country Studies Botswana. Botswana officials have reported that, almost by the day, they are finding the SAM of value in relation to a whole range of issues. APPENDIX A: BUILDING THE BOTSWANA SAM Two members of the 1977 SAM team went in advance to discuss classifications and desired analyses among other things with representatives of the Botswana government. These discus- sions were held with a SAM Reference Group which had been formed by the government of Botswana. This group comprised representatives from the Department of Economic Affairs and the Department of Financial Affairs in the Ministry of Finance and Development Planning and also representatives from the Central Statistical Office and the Bank of Botswana. As a result of the discussions, the format and classifications desired for the SAMwere finalized and, in addition, the following occurred: 1. A seminar was presented to the SAM Reference Group to which all economists and planners in government were invited. Members of the Reference Group were given various reference material, including sections from Pyatt and Thorbecke (1976), and some had the complete volume. 2. A member of the CSOprepared a list of data available from all government departments, the Bank of Botswana, and the University of Botswana and Swaziland, which included the following information: * Department or organization producing the data * Name of person to contact there * Survey involved or other source of the data a Periodicity of its collection * Where published, if published * Coverage and quality of the data. The purpose was to minimize the time spent by team members in following up data requirements and availability for their segments of work. The CSOfound it very useful themselves to have such a list available. 3. A complete set of the latest reports listed under item 2 was located in one room for easy reference by team members. Consequently, when the rest of the team arrived, each member was able to commence work immediately on the particular segment of the matrix that had been allocated to him or her. In order to minimize errors that would arise with individuals working simultaneously on different parts of the matrix, control totals were taken from the national accounts for such items as compensation of employees, operating surplus, capital formation, and so on, and team members were asked to ensure that if it proved necessary to change a control total, they would notify other members and discuss it with them. On one or two occasions this procedure was not followed but, overall, it was certainly worthwhile and reduced balancing errors appreciably. The members finished their tasks at about the same time, and then all but two left Botswana. This was a mistake. It would have saved much time subsequently if they had stayed in Botswana for about a week afterward, so that major discrepancies could be followed up by members whose sections were affected. Their time would not have been wasted, even if there were no major discrepancies, because final detailed notes and worksheets could have been completed. As it was, the balancing was done in London, and final detailed notes and worksheets were completed by members in their respective permanent institutions. Since members were located in the north of England, Warwick, Sussex, and Norway, it was possible to bring them all together Botswana 1974-75 141 again only on one occasion for two days in October 1977 to work on the balancing, which was not finally completed until early November 1977. In producing the final balanced SAM, various intermediate SAMs were produced, as follows: * SAM I was the unbalanced SAM produced by taking entries direct from worksheets. * SAMs II and III contained adjustments to data made on the basis of judgment, local knowledge, and correction of errors. * SAM IV was that used for the RAS input, containing the totals to which one wanted to work. * SAM V was the final, balanced SAM, which also contained some minor adjustments to the RAS results made for balancing purposes. All the above SAMs, together with the detailed notes and worksheets and the final report (United Kingdom, Ministry of Overseas Development, 1977), were returned to the CSOin Botswana at the end of November 1977. It was clearly essential that they have all of this in order to follow what had been done, answer any queries that might arise, and update the matrix in the future. Moreover, the questions of errors in data and adjustments made in balancing warrant more research. If that work is to be done, then a clear record must be left of such errors and adjust- ments. In presenting the final SAM,a second seminar was given for economists and statisticians throughout government to demonstrate the use of the matrix, taking as an example the possible impact of foot and mouth disease on the economy. In all, the exercise took about twelve man-months, including work undertaken outside of Botswana. If, however, only one person had worked on it full-time, rather than a team, I think there is no possibility that he would have been able to complete it in one year. The pace of work within a team during a short period is far greater than could be achieved by one individual. Moreover, the interaction among team members is useful in resolving many problems that might stump someone working alone. APPENDIX B: PRICE EFFECTS Rows and columns 38 and 39 of the Botswana SAM are designated "price effects," a term not employed in other SAMs. The equivalent term in SNA table 2.1 is "commodity taxes, net." Our reason for preferring a different expression requires explanation. The transactions in SNA table 2.1 are valued at approximate, as opposed to true, basic prices. The SNA defines true basic values as "producer's values of the gross output of commodities, industries, etc., less the commodity taxes, net, in respect of the gross output and the direct and indirect intermediate inputs; or the sum of the value of the primary inputs, indirect taxes, net, except commodity taxes, net, and the true basic values of the intermediate inputs in respect of the gross output." The use of true basic values is stressed in the SNA as being essential in order to obtain uniformity of valuation for meaningful manipulation of an input-output table. The standard national accounting convention is followed in estimating basic values, namely, that industries pass on indirect taxes charged to them in full to their customers, through appropriate adjust- ment of their prices. Insofar as commodity taxes are charged on interindustry transactions, then there will be a multiplier effect, which serves to raise the general price level above the level that would result in the absence of taxes on interindustry transactions. The value of commodity taxes charged is passed on to final demand, identically, but the total value passed on will exceed this by the value passed on in interindustry transactions. These calculations consequently require matrix inversion, because of the interindustry effects. If this inversion is 142 Country Studies not done, then the SNA recommends use of approximate basic prices. These approximate basic prices are derived by simply prorating commodity taxes over the values of all transactions, so that the value passed on in total is identically equal to the value of the taxes charged. Conse- quently, the value charged to final demand is less than the total value charged by the amount allocated to interindustry transactions, which is contrary to the national accounting convention used, and quite arbitrary, depending on the relative magnitude of interindustry transactions as against final demand transaction. The SNA defines commodity taxes as "preferably, indirect taxes less subsidies, each of which are proportional to the quantity, or the value, of commodities produced or sold. At least, indirect taxes less subsidies, each of which differ from one disposition to another of a commodity or from one commodity to another of the same class." Thus, commodity taxes are a particular subgroup of indirect taxes and normally would form the bulk of these. They will normally be paid, in the literal accounting sense, by industries (including wholesale and retail trade) to government in the same way industries pay wages and salaries to their employees. Some might be paid by final consumers directly to government, for example, on imports they bring into the country themselves, but this would normally be a very small part of total payments. Commodity taxes shown in SNA table 2.1 are not the taxes paid to government, however, which have already been included in indirect taxes, net (row 32). They are the estimated effects of these taxes through industries passing on the sums charged to them in the prices they charge their customers, using the approximate basic value method of estimation. This is potentially confusing, and the United Nations Statistical Office (1973) seems to fall into the trap. In para- graph 1.12, when describing table 1.2 of the same volume, it states that "industries purchase commodities in matrix 1 and pay the relevant taxes in vector 7," where vector 7 is the inter- section of the industry columns with the commodity taxes row. To better understand the objec- tion to this line of reasoning, suppose it was desired to estimate the effect of wages and salaries on prices charged by producers. An exactly equivalent treatment would be to relabel the commod- ity tax row in table 1.2, "wages and salaries." The actual payment of wages and salaries by employers would appear in the factor income accounts, wages and salaries being the major component of compensation of employees. The entry in the commodity tax row, now called wages and salaries, would be equal in value to the payments that had actually been made and recorded in the factor income accounts and would simply be the estimated value of these payments passed on by sellers to buyers in their selling price (where approximate rather than true values were estimated). The entry in vector 7 of table 1.2 is not the taxes paid by industry, it is the value of the additional price-raising effect of the taxes arising from interindustry transactions. Thus we have used the expression "price effects" in preference to "commodity taxes" in the appendix, because we feel that this is a better description of the items so described and is less likely to be misunderstood. Following the treatment in table 2.1 of the SNA is not helped in any case by what appears to be a printing error. Total indirect taxes, net, raised (in row 32), are not matched by an equal payment of these to government, that is, there should presumably be an entry of 29 at the intersection of row 53, column 32. True basic values have been estimated in table 7.1, rather than approximate basic values as in SNA table 2.1. It has been shown by Greenfield and Fell (1979) that the use of approximate basic values, as defined in the SNA, is inadequate for purposes of achieving uniformity of valua- tion or for estimating price effects in general. They present a better approximate formula which does have the property, for example, that all indirect taxes are passed on to final demand, but conclude in any case that it is preferable to use a true formula. No attempt has been made in table 7.1 to separate commodity from noncommodity indirect taxes and to treat them separately, largely because the bulk of indirect taxes in Botswana are commodity taxes. PART III Multipliersand SAM-BasedModels I I 8 The Disaggregation of the Household Sector in the NationalAccounts Sir Richard Stone When the Statistical Office of the United Nations published a major revision and extension of its system of national accounts (UTNSO, 1968), it was recognized that, although the revised system had greater coverage than the original (described in UNSO, 1953), a number of topics had been left over for treatment in the future. One of these was the distribution, and in particular the distribution among households, of income, consumption, and wealth. The subject is a difficult one, and it was felt that a good deal of discussion would be needed before a generally acceptable scheme could be devised. However, the matter was not overlooked by the Statistical Office,which at the time of the publication of the revised SNAwas engaged in preparing an integrated system of distribution statistics intended to fit into both the SNA and the Material Product System, or MPS (as described in UNSO, 1972); a report on this scheme appeared in UNSO,1977b. In the past decade inequality has become a major issue in many parts of the world, and a great deal of new statistical work has been done, mainly on the distribution of income. Apart from data collection, many analytical studies have been made. An interesting one relating to Britain, which is described in Nicholson (1964) and has since been repeated annually by the U.K. Central Statistical Office in Economic Trends, shows for different types of households in various income groups the changes in original income brought about by taxation and social service benefits. For this kind of analysis it is never necessary to move outside the household sector, and so there are no links between the original incomes of households and the activities in which they are gained or between the expenditures of households and the activities which gain from them. To trace these circular flows it is necessary to embed the distribution statistics in a social accounting matrix. This step has been taken by Pyatt, Roe, and associates ( 1977). In their study of social account- ing for development planning with special reference to Sri Lanka undertaken for the Inter- national Labour Office, they have set out a social accounting matrix in which consumers as well as producers are disaggregated. In what follows I shall describe the main problems encountered in disaggregating the house- hold sector within a social accounting framework, and I shall illustrate my remarks by means of a numerical example relating to Britain in 1968. I shall go on to use this matrix to approx- imate the interconnections between the current accounts of an economic system and to show their effect in increasing the responses of different parts of the system to supposedly exogenous stimuli. Finally, I shall consider some of the more recent analytical tools that ought eventually to be incorporated in an approach of this kind. ACTIVITIES AND SECTORS The SNA, like all systems of national accounts, distinguishes between the given country and the rest of the world and, within the given country, between production, consumption, and accumulation. Production is classified by activities, that is to say, by grouping together estab- lishments producing a range of similar products and responsible for resolving the day-to-day 145 146 and SAM-Based Multipliers Models problems of production. These groups of establishments are usually termed industries, though a place is also given to the producers of government and private services. The institutional classification of consumption and accumulation is different, the units being grouped into sectors responsible for resolving financial problems including the spending and saving of income. The main sectors distinguished in the SNA are nonfinancial enterprises (corporate and quasi-corporate), financial institutions, general government, households, and private nonprofit institutions. In practice the last two categories are often combined into what is termed the personal sector. HOUSEHOLDS Clearly, if the social accounts are to contribute to a discussion of inequality and of the factors that affect it, the household sector must be isolated and divided into categories. Households are appropriate for this purpose because they are the units in which decisions on spending and saving of income are generally made. Isolating the Household Sector In attempting to isolate the household sector the following problems are likely to be encoun- tered. Removing private nonprofit institutions. The construction of an income and outlay account for private nonprofit institutions (PNPIs) is laborious but cannot be avoided because of the importance of PNPIs in the aggregate. An account of British experience in this respect is given in Economic Trends (no. 259). In Britain the sector is composed mainly of universities and colleges, friendly societies, trade unions, housing associations, and charities. The last group is numerous: in England and Wales nearly 77,000 charities were registered with the Charity Commissioners in 1970. A good deal of work, therefore, is needed to yield even a rough set of accounts for PNPIs as a whole; further estimation must be done to fit these figures into a social accounting matrix as set out below. Removing visitor expenditure. In tables of consumer expenditure it is usual to include the expenditure of visitors from abroad along with the expenditure of local residents and, deducting a single figure unclassified by constituent goods and services, to obtain an estimate of the domestic expenditure of residents. For present purposes it is necessary to decompose this figure and deduct its components from the categories of goods and services, since in many countries tourist expenditures are large and we cannot assume that their composition is similar to that of the residents. This, again, is an area in which reliable information is difficult to obtain. In this context visitors include foreign diplomats and armed forces, as well as visiting businessmen and tourists. Adjustments for life insurance. The treatment of life insurance in the national accounts may result in a partial amalgamation of life insurance accounts with those of the personal sector; this amalgamation is intended to represent the beneficial interest of that sector in life funds and to ensure that personal saving through life insurance is properly reflected in the personal income and outlay account. In combining information from the national accounts with data from other sources, which is inevitable if disaggregated household accounts are to be constructed, it is necessary to check the consequences of this partial amalgamation for the entries in the national accounts, since they are not likely to accord without adjustment with those to be found in, say, a family expenditure survey. It may be mentioned in passing that there are many other respects in which the usual of the HouseholdSector Disaggregation 147 treatment adopted in the national accounts is not followed in other sources. For instance, employers' contributions to national insurance are usually routed through households in the national accounts but not in household surveys. Unincorporated businesses. In the national accounts all production accounts are brought together independently of the firm's legal form of organization, but the appropriation account entries of unincorporated businesses will usually be lost in the corresponding entries in the income and outlay account of persons. TIhis is highly inconvenient since some unincorporated businesses are very large; to simplify matters, the concept of quasi-corporate enterprises was introduced into the SNA. In the case of small businesses, little harm is done by the conventional treatment provided that income is defined after depreciation and stock appreciation, so that entries out of place in income and outlay accounts are not included. This, of course, has impli- cations for the corresponding production accounts. Distributions over Households By such means as I have described we may expect to obtain a more or less clean income and outlay account for the household sector as a whole, and the same could be said of the capital transactions account. The remaining tasks are to distribute the entries in these accounts over types of household and to deal as far as possible with the transfers that make their appearance as a consequence of this disaggregation. The income and outlay account should be the easier to distribute because by now there is a good deal of information in many countries from family expenditure and other household surveys on current account items but very little on the entries in the capital account. The distribution can be carried out by taking each entry in the account for the household sector in the national accounts, comparing this entry with the nearest corresponding item in the household survey, and distributing the former by reference to the latter. In general, it is not necessary, or even desirable, to assume that the estimate in the national accounts is accurate and that the estimate in the survey must bear the full weight of any adjustment. Household surveys, where they exist, are usually an important source in constructing the national accounts, but they are not always used systematically; if someone with full access to and knowledge of the data to be compared were to carry out the exercise described below, he might well succeed in improving the national accounts, as well as meeting the objective of disaggregation. Many of the problems arising in this endeavor can be summarized as follows. Definitions and classifications. A detailed comparison of the definitions and classifications in the two sources will provide a means of selecting from the survey the best indicator for each item in the national accounts or, alternatively, of grouping the data from both sources so as to improve their comparability. The classification of households. Many classifications have been proposed: by size of income, by demographic composition, by age or occupation of the head, by region, and so on. Distributions by these criteria serve different purposes. Totals and averages. A survey relies on a sample, and since the national accounts do not usually provide demographic information, this information will have to be introduced. An esti- mate of the total population or the total number of households would enable us to express the national accounts estimates as so much per head or so much per household. At this point we should have to look out for the effect of individuals living in institutions, since at least part of their income and expenditure may get into the personal income and outlay account in the national accounts, whereas they are not likely to be recorded in a household survey. It must be remembered that the survey sample may not be representative of the entire 148 Multipliers and SAM-Based Models population, and it may even be deliberately distorted in order to increase the coverage of certain groups. It would be helpful if more data were available nationally for the total number of households in different categories, as they already are for some classifications in years in which a population census was taken. Matching and the use of indicators. If the total or the average for an item does not match in the two sources or if an item is simply not recorded in the survey, it is necessary either to aggregate items to improve the match or to select from the survey an item or combination of items which can be used as an indicator of the distribution of the item in the national accounts. For instance, if employers' contributions to social security are not recorded in the survey but bear a known relationship to employees' contributions, then the distribution of the latter will provide a good indicator of the distribution of the former. To-whom-from-whom problems. Although we may have no difficulty in distributing an item, such as wages and salaries, over different categories of households, it will usually be impossible to distinguish at the same time the activity in which the payment originates or the sector in which a particular producing unit is located. The same is true of rent, dividends, and interest, which typically will have to be treated as a block without further subdivision. Such problems can be resolved by means of dummy accounts. For instance, we might set up an account for income from employment which received this type of income from each activity and paid it out to various types of household. Thus we would obtain distributions by activity and by household but would have no cross classification. Alternative classifications. Household expenditure should be classified in terms of the goods and services that consumers buy, although eventually this classification will have to be converted into the product classification used in the production accounts of the system. This transfor- mation can be carried out by means of a classification converter, which will be described later. Discrepancies and adjustments. The social accounts are constructed by combining infor- mation from a variety of sources, and discrepancies are certain to appear. If these are not very numerous they may be left as they are and the accounts set up so that they are all shown explicitly, as in the example which follows. Usually there will be many such discrepancies, and these could be adjusted away by the method suggested in my introduction to Pyatt, Roe, and associates (1977). In practical cases a great deal of computing is involved, but this difficulty has been resolved by Byron (1978). A NUMERICAL EXAMPLE Table 8.1 sets out a social accounting matrix relating to Britain in 1968. Its main purpose is to illustrate the disaggregation of the household sector, but it also serves two subsidiary purposes: the capital transactions accounts are set up much as in the SNA so that, with minor modifications, opening and closing balance sheets and the associated flow-of-fands and reval- uation accounts could be added to them; and all residual errors and unidentified items are brought together and shown explicitly in account 62. The order of the accounts follows that adopted in Pyatt, Roe, and associates (1977). This order is convenient for present purposes but is not a matter of principle, since the accounts could be rearranged in a more familiar order by a conformable permutation of the rows and columns. Thus if the fifty-two accounts of table 8.1 form the matrix T and the familiar arrangement is denoted by T*, then T* = PTP-1, Disaggregation of the HouseholdSector 149 where P is the appropriate permutation matrix. Similarly, if A* and A are coefficient matrices derived respectively from T* and T then A* PAP-` and (I -A*)-l = P(I - A)-' P-i. I have chosen 1968 because for that year an input-output table is available (UKCSO,1973) that is keyed into the national accounts as set out in the 1972 Blue Book (UKCSO,1952-). The report of the Family Expenditure Survey for 1968 is set out in the 1969 issue of U.K. Department of Employment (1961-). This material makes it relatively easy to construct a numerical exam- ple for 1968; with more work and subject to difficulties in obtaining input-output information and in keeping the estimates at a common level of up-to-dateness, it would be possible to construct table 8.1 for other years. In the present case the input-output structure is confined to the six branches given in the small example in UKCSO(1973) and is on an industry basis without reference to commodities. Again this is not essential but seems a justifiable simplifi- cation in what is primarily an example. of the Accounts and Entries The NWature As usual, each row and column pair of table 8.1 represents an account with incomings in the row and outgoings in the column. The rows and columns of the whole matrix of order 52 have been divided into eleven categories, thus giving rise to 121 submatrices. The nature of these categories is as follows. Value added: accounts 1-3. The first three accounts relate respectively to income from employment, gross profits and other trading income, and taxes on expenditure (net); and the sum of their entries is equal to the net domestic product at market prices, S39,204 million. Payments to the factors of production are made by the six industries, at the intersections of rows 1 and 2 and columns 34 to 39, and provisions for depreciation are shown negatively at the intersections of row 2 and columns 45 to 47, the capital accounts of the sectors. Taxes on expenditure (net) are received by row 3 from a wide range of accounts, namely 19 to 39, 41 to 43, and 48. The receipts of account 1 are paid out as wages and salaries to account 4 and as employers' contributions to account 5. The receipts of account 2 are divided into income from self-employ- ment and the rest; the former is distributed over households at rows 7 to 13, while the latter is paid into account 6, which collects together rent, dividends, and interest (property income) from all sources and distributes it to all destinations. The receipts of account 3 are paid to the central government at row 17 and to local authorities at row 18. Forms of income: accounts 4-6. The wages and salaries and employers' contributions received by accounts 4 and 5, respectively, are distributed by these accounts to households at rows 7 to 13. The receipts into account 6 come not only from domestic production, at column 2, and from the rest of the world, at column 49, but also from most of the income and outlay accounts of the sectors, at columns 7 to 18. Consumer debt interest is included in these entries. Rent, dividends, and interest are paid out to the sectors, at rows 7 to 18, and to the rest of the world at row 49. The entry at row 52 is the residual error. It can be seen that the net receipt of rent, dividends, and interest by the domestic sectors and the rest of the world is exactly equal to the amount generated in the domestic productive system. Income and outlay of sectors: accounts 7-18. In addition to the types of income already described, rows 7 to 18 show, at the intersections with columns 7 to 18, additional income Ct Table 8.1. A Social Accounting Matrix for the United Kingdom, 1968 (in mllions of pounds sterling) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 1 Income from employment 2 Gross profits & other trading income 3 Taxes on expenditure (net) 1041809 519 6 155 221 195 784 1 45 110 4 Wages & salaries (inci. Forces' pay) 23056 6 P.mployers'contributions 2284 6 Rent,dividends, and interest 5135 8 61 164 212 154 82 175 3534 636 1240 791 7 Households: income £ 0-10 20 42 21 189 38 823 1 8 > £10-20 238 1393 184 771 50 1086 13 9 " " £20-30 418 5529 549 785 31 629 42 10 " " £30-40 482 5982 545 707 19 374 38 11 " £40-50 308 4079 372 572 9 181 17 12 " " £60-60 278 2521 231 268 4 87 11 13 " " £6-+ 973 3510 382 1020 6 112 11 14 Private nonprofit insUtutions 241 13 42 91 103 77 45 116 34 254 15 Companies 6814 16 Pubhc corporations 529 17 Central government 4548 881 19 386137015461179 7621483 915 17 18 Local authorities 1464 560 1898 19 Food 295 B25 125 1291 750 427 547 20 Drink andtobacco 80 361 859 771 501 330 421 10 21 Housing 203 488 800 692 434 230 336 108 22 Fuel and light 114 234 336 273 162 86 111 24 23 Clothing andfootwear 59 223 527 545 349 203 319 45 24 Durable & other household goods 64 221 473 471 322 194 258 45 25 Other miscellaneous goods 59 192 427 405 260 156 236 45 26 Motor vehicles (incl. running costs) 20 137 450 614 379 252 374 45 27 Travel and communications 35 129 254 244 157 98 156 35 28 Entertainments, hotels & restaurants 45 151 352 361 250 159 285 120 29 Other services 71 189 339 334 243 148 405 314 30 Mtlitary defence 2370 31 National health service 1376 32 Other central government purposes 996 33 Local authorities purposes 2976 34 Agriculture etc., mining quarying 796 2 157 45 84 35 Metals and metal products 34 371 112 393 38 Othermanufacturing 2355 645 127 75 1149 713 700 131 252 37 Construction 595 1s 38 Gas, electriciqr, and water 73 948 39 Services 1636 8061925 148 776 628 606 669109011741732 40 Sales by final buyers 2 5 11 13 10 5 14 227 81 172 41 Vehicles, ships, U aircraft 42 Plant e machinery 43 Buildings & works 44 Increase in stocks 45 Personal sector 37 57 -72 192 263 189 693 51 46 Companies 1378 47 Public sector -124 1637 22 48 ProduciUon 8 27 60 66 49 29 76 772 73 16 6 190 115 122 67 17 87 14 49 Consumption 908 2 7 14 14 9 5 9 17 753 179 B0 Accumulation 51 Identified financial assets 52 Unidentified items 29 Total 25340 7882 6012 23056 2284 14074 1134 3735 7983 8147 5538 3400 6014 1016 6614 529 13242 3922 5663 3333 3291 1340 2270 2048 1780 2271 1108 1723 2043 (Table continues on the following page.) (A LA Table 8.1. (Continued) 30 31 32 33 34 36 356 57 38 39 40 41 42 45 44 45 46 47 48 49 50 51 52 Total 1 Income from employment 954 4607 3977 1795 514 13493 25340 2 Gross profits & other tradingcncome 889 1324 1958 733 768 5798 -777 -1411 -1430 7852 3 Taxes on expenditure (net) 33 57 66 185 -183 168 332 238 85 940 60 84 24 26 6012 4 lWages salaries (incl. Forces'pay) 23086 6 Rmployers' Gontributions 2284 6 Rent, dividends, and interest _ . 1882 14074 7 Households: fincomeY 0-10 1134 8 £10-20 3735 9 £20-30 7983 10 £30-40 8147 11 £40-60 5538 12 " " O-60 3400 13 £6OW+ 6014 14 Private nonprofit insttutions 1016 16 Companies 6614 16 Public corporations 529 17 Central government 136 13242 18 Local authorities 3922 19 Food 5663 20 Drink and tobacco 3333 21 Homing 3291 22 lhel and light 1340 23 Clothing and footwear 2270 24 Durable 6 other household goods 2048 28 Other misceUaneous goods 1780 26 Motor vehicles (nel. running costs) 2271 27 Travel and communications 1108 28 Entertainments, hotels & restaurants 1723 29 Other services 2043 30 Military defence 2370 31 National health service 1376 32 Other central government purposes 996 33 Local authorities purposes 2976 34 Agriculture etc., ming & quarrying 3 31 4 45 51 1193 92 386 20 28 14 130 3081 35 Metals and metal products 848 47 124 57 170 669 709 131 610 780 2046 -76 3363 10388 36 Other manufacturing 49 215 86 277 727 1077 915 140 1401 46 60 160 2474 13774 37 ConstructIon 104 17 47 87 78 94 45 35 70 233713 11 25 4959 38 Gas, electricity, and water 23 16 16 76 67 287 278 16 328 169 13 2300 39 Services 10591020 8352438 247 1284 2243 323 197 116 244 410 2 2231 23839 40 SWlesbyfinalbuyers -73 -58-267-222 3 188 47 15 6 55 -276 -76 118 0 41 Vehicles, ships, & aircraft 127 560 161 848 42 Plant & machinery 226 1550 1158 2934 43 Buildings & works 911 851 2473 4235 44 Increase m stocks 29 140 45 214 45 Personal sector 194 1594 48 Companies 460 1838 47 Public sector 427 40 2002 48 Production 324 31 85 33 129 1308 3032 123 38 1124 168 408 103 8700 49 Consumption 372 2289 50 Accumulation 271 271 51 Identified financial assets 1537 -944 -996 403 0 52 Unidentified items -886 1052 -63 -132 0 Total 23701376 996 2976 3081 10388 137744959 2300 23839 0 8482934 4235 214 1594 1838 2002 8700 2289 271 0 0 154 and SAM-Based Multipliers Models received from other domestic sectors and, at column 49, British direct taxes paid by nonresi- dents. In the columns we find (in addition to payments of rent, dividends, and interest, taxes on income, and other transfers) outlays for consumers' goods and services at rows 19 to 29, for government purposes at rows 30 to 33, and saving at rows 45 to 47. The remaining entries are transfers abroad, at row 49, and consumers' expenditure abroad, at rows 40 and 48. The entries in row 40 are somewhat mysterious, but in UKCSO(1973, p. 121) £60 million of sales by final buyers are shown as a component of consumers' expenditure abroad. Consumers' goods and services: accounts 19-29. The rows of these accounts collect together the expenditure by the seven categories of households and by PNPIs on one of the eleven categories of consumers' goods and services. Visitors' expenditures could have been treated in the same way in column 48, but in fact they are converted into demands on industries and included among export sales. Columns 19 to 29 show the cost components of these goods and services in terms of taxes on expenditure (net) at row 3, direct demands on industries at rows 34 to 39, purchases from final buyers at row 40, and imports at row 48. Government purposes: accounts 30-33. These rows collect the expenditure on goods and services for each of the four government purposes distinguished, and the columns show the corresponding cost components,just as in the previous block of accounts for private consumers. Industries: accounts 34-39. These rows and columns represent the production accounts of the six industry groups into which, following table 7 of UKCS0 (1973), the productive system is divided. This simple treatment seems justified for illustrative purposes; there would be no difficulty in introducing much more disaggregation or the familiar commodity/industry treat- ment, but this would needlessly complicate the present example. It will be noticed that these accounts, like all the accounts of table 8.1, are presented in consolidated form. (The diagonal elements that have been omitted are, in order, 490, 5749, 4668, 945, 66, and 1934.) Sales by final buyers: account 40. This account contains in its row positive and negative entries which sum to zero, and its column is left blank. The nonzero entries relate mainly to the sale and disposal of goods and services from final demand accounts, such as used cars sold by businesses to consumers, plant and machinery bought by scrap merchants or exported, and payments to public authorities for various services provided. Investment goods: accounts 41-44. These accounts correspond for investment goods to accounts 19 to 29 for consumption goods and services, that is to say, in the rows they receive finance from the capital accounts of the sectors and in the columns they pay it out to meet the costs of the different categories of goods. Capital transactions of sectors: accounts 45-47. For each sector or group of sectors these accounts receive capital transfers and the sector's saving, and with them finance gross capital formation less depreciation (in row 2), capital transfers to other sectors, and identified acqui- sitions (net) of financial assets. The accounts being consolidated, a figure of 204, representing intragovernmental capital transfers, is omitted from the intersection of row and column 47. The personal and the public sectors are shown in aggregated form mainly because information is not available to fill in row 51 for the divisions of these sectors. The entries in row 52 are not real outlays and would all be zero if complete information were available. The rest of the world: accounts 48-50. Account 48 receives in the row the proceeds from British imports, including British tourist expenditure abroad, and in the column pays for British exports. The negative entry in row 3 represents the remission of taxes on exports, an item of 471 representing reexports is omitted from the intersection of row and column 48, and the Disaggregation of the Household Sector 155 entry of 372 in row 49 represents the excess of British imports and expenditure abroad over British exports and visitors' expenditure in Britain, that is, the rest of the world's favorable balance of trade with Britain. Account 49 receives in addition property income, taxes on income, and other current transfers from Britain and pays out property income, taxes on income, and other current transfers to Britain. The balancing item, 271, represents net lending to Britain by the rest of the world and is paid into account 50, where it is matched by the rest of the world's net acquisition of financial assets, including unidentified items. Financial assets and errcrs: accounts 51 and 52. These accounts have positive and negative entries which sum to zero in the rows and no entries in the columns. If we wished to attach balance sheets to accounts 45 to 47 we should have to show these flows gross and also disag- gregate them by financial asset. The gross acquisitions of financial assets would appear in a set of rows like row 51, and the gross acquisitions of financial liabilities would appear in the corresponding columns, wh.ich in the present example are reduced to column 51 which is blank. It would then be possible to add on revaluation accounts and balance sheets in the manner illustrated in the UN SNA table 2.1. Finally, row 52 brings together the errors and unidentified items left unresolved in the British national accounts. The entry at column 6 is the residual error, the entries at columns 45 to 47 are the unidentified items Ln the financial accounts, and the entry at column 47 is also equal to the balancing item in the balance of payments accounts. This is a convenient way in which to introduce the errors into the social accounting matrix but, as explained in Stone and Stone (1977, p. 66), it is impossible to say just where the errors lie since to a greater or lesser extent all the entries in the national accounts are subject to error. A Word on the Construction of Table 8.1 I do not propose to describe the construction of table 8.1 in detail, but a short account of the procedure adopted may be helpful. The broad picture. Given the general framework, many of the entries of the table can be obtained from the 1972 Blue Book (UKCSO,1952-). My first step was to fill in as many entries as possible from this source. The input-output structuLre. Most of the entries in rows and columns 1-3, 34-40, and 48 come from the 1968 input-output tables given in UKCSO(1973), which are keyed into the 1972 Blue Book. I adopted an interindustry formulation based on table 7 because it seemed the easier method to follow. In fact a, good deal of rather tedious calculation was involved since table 7 does not give all the disaggregation required, while the large tables often provide commodity rather than industry detail. Furthermore, the procedure used by the Central Statistical Office to adjust the small table is not the same as that used to adjust the large tables. The classification converters. The converters used to transform consumer goods and services, government purposes, and investment goods into industrial cost structures are based on tables 0, P, and B of UKCSO(1973). The treatment of table 0 will serve as an example. The first step was to group the entries in. columns 1 to 24 into the eleven categories of consumers' goods and services. The second step was to allocate the entries in columns 27 to 30 row by row over the eleven categories, deducting at the same time expenditure by foreign visitors in column 25; this was done on a pro rata basis with the sole exception of net taxes on mineral oil refining (row 16, column 29), which fall mainly on gasoline used by motor vehicles and only to a minor extent on heavy oils used for domestic heating. The third step was to combine the items column by cohimn to give the elements distinguished in the cost structures. The final step was to adjust the demands on domestic production from a commodity to an industry basis; this was done by 156 Multipliers and SAM-Based Models means of the RAS method which, while not correct in principle, was probably as good as anything else I might have done. The isolation of the household sector. The PNPIs were taken out of the personal sector on the basis of the information provided in Economic Trends (no. 259) and in the 1977 Blue Book, and visitors' expenditures were removed by the method described earlier for the input-output structure. I made no attempt to carry out further adjustments for life insurance or unincor- porated businesses. The allocation of household income and outlay. This allocation was based on information taken from table 3 of the 1969 issue of the Family Expenditure Survey. The twelve household groups (classified by their weekly income) distinguished in the survey were aggregated into seven. For each item of income or outlay the relevant total derived from the 1977 Blue Book was then distributed over these seven groups by means of indicators derived from the survey. As far as possible I constructed a single indicator for each item by adding up what seemed the most appropriate series in the survey: expenditure on food, for instance, was allocated according to numbers 12-42 in the survey; number 43, "meals away from home," was included with expenditure on entertainments, hotels, and restaurants; and so on. In some cases, however, it seemed preferable to subdivide one of my categories of goods and services and use separate indicators to allocate the components. By distributing the Blue Book totals according to the sample total of the indicator in each income group I have implicitly treated the sample as representative. Household saving. I distributed all other items in the household income and outlay accounts, so saving was left as a residual in each income group. All that can be said is that the series, with almost half of household saving in the top income group, is not altogether implausible. The negative total for the income group £20-30 a week may seem surprising, but there appear to be a large number of young married couples and couples with young children in the group; this would be likely to boost expenditure, particularly in view of the fact that durable goods are treated as part of current expenditure. WHATEVER BECAME OF THE MIULTIPLIER? It has been more than fifty years since Kahn published his famous paper on the multiplier (1931), which has opened up so many paths. One path has led to the development of the matrix multiplier proposed by Goodwin (1949), which has the form of a Leontief inverse but can be applied to households and government as well as to industries. A recent example of multiplier analysis in a social accounting framework is provided in Copeland and Henry (1976), which contains many calculations relating to Ireland in 1964 and 1968. The latest development on these lines, which in this section I shall apply to table 8.1, is given in Pyatt, Roe, and associates (1977) and in Pyatt and Round (1979), reproduced in chapter 9 of this volume. A particularly interesting feature of this work is that it shows how a matrix multiplier can be decomposed into multiplicative components each of which relates to a partic- ular kind of connection in the system as a whole. Suppose, as in the following example, that there are three endogenous subsystems; then an injection applied to an account in one subsys- tem may have the following effects. First, it may move around within the subsystem, giving rise to repercussions of the kind measured by a Leontief inverse. Second, it may move around the whole system and return to the subsystem from which it started. And, finally, it may move around and end up in one of the other subsystems. Pyatt and Round refer to the results of these Disaggregation of the Household Sector 157 movements as direct effects, indirect effects, and cross effects. I prefer to call them, respectively, intragroup, intergroup, and extragroup effects in order to avoid any confuLsionwith the direct and indirect requirements of traditional input-output analysis, especially as both types of requirement are included in what Pyatt and Round call direct effects. The example that follows starts off from a coefficient matrix based on table 8.1, which is set out in table 8.2. The whole system is divided into four subsystems, three endogenous and one exogenous, namely: (a) value addedlincome originating, accounts 1-3; (b) forms of income and sectors, accounts 4-18; (c) goods and services, government purposes, and branches of produc- tion, accounts 19-39; and (d) everything else, accounts 40-52. If we write table 8.2 as a partitioned matrix, AO,say, of order 52, then O O A1 3 A1 4 A21 A2 2 0 1A 24 AO = 0 A32 A3 3 ' A34 L 0 A4 2 A 43 A.J where the first three rows and columns relate to the accounts that are treated as endogenous and the fourth row and column relate to the accounts that are treated as exogenous. In the final terminology this means that leakages from the endogenous part of the system are caught in the capital accounts and in the accounts for the rest of the world. If we denote the top left-hand partition of A0 by A, then, say, [o 0 A131 A= A21 A2 2 0 O A32 A33J O O O O O A13 = A22 0 + I2 ° ° O O A33 O A32 ° = B + C. If we denote the vector of totals of the thirty-nine endogenous accounts by y and the vector of the sums of the incoming elements into these accounts from the thirteen exogenous accounts by x, then, say, (8.1) y = A:= + x = By + Cy + x - (I- B)-1 Cy + (I - B)-1 x =[I- (I -B)-I C]-' (I -B)-l x - [I + (I - B)-1 C + (I - B)-1 C (I - B)-1 C] *I - [(I - B)-' C (I - B)-1 0 (I B)-' C]}- (I - B)-1 x - M3 M 2 MI x M.x. Table 8.2. Coefficient Matrix, 10000 A, Derived from Table 8.1 for the United Kingdom, 1968 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 28 1 Income from employment 2 Gross profits & other trading income 3 Taxes on expenditure (net) 184 5428 1577 45 683 1079 1096 4 Wages & salaries (incl. Forces' pay) 9099 6 Employers' contributions 901 6 Rent, dividends, & interest 6540 71 163 205 260 278 241 291 5343 12023 936 2017 7 Households: income £ 0-10 26 18 92 120 374 622 3 8 " 104-20 303 604 806 549 492 520 33 9 " " £20-30 532 2398 2404 559 305 475 107 10 " " 40 614 2595 2386. 603 187 282 97 11 O" £4-50 392 1769 1629 407 89 137 43 12 " £50-60 354 1093 1011 191 39 66 28 13 " £60+ 1239 1522 1673 726 59 85 28 14 Private nonproit institutIons 171 115 112 114 126 139 132 193 51 192 15 Companies 4706 16 Public corporations 376 17 Central governmnent 7565 627 168 1033 1716 1898 2129 2241 2466 1383 321 18 Local authorities 2435 398 1433 19 Food 2601 2209 1914 1585 1354 1256 910 20 Drink & tobacco 705 987 1076 946 905 971 700 98 21 Housing 1790 1307 1002 849 784 676 559 1063 22 Fuel & light 1005 627 421 335 293 253 185 236 23 Clothing e footwear 520 597 660 669 630 597 530 443 24 Durable &other household goods 564 592 593 578 581 571 429 443 26 Other miscellaneous goods 520 514 535 497 469 459 392 443 26 Motor vehicles (incl. running costs) 176 367 564 754 884 741 622 443 27 Travel & communicatios 309 345 318 299 283 288 259 344 28 Entertairnments, hotels & restaurants 397 404 441 443 451 468 474 1181 29 Other services 626 506 425 410 439 435 673 3091 30 Military defence 1790 31 National health service 1039 32 Other central government purposes 752 33 Iocal authorities purposes 7588 34 Agriculture, etc., mining & quarrying 1406 6 1172 253 35 Metals & metal products 103 1812 629 36 Other manufacturing 4159 1935 386 560 5062 3481 3933 37 Construction 1808 38 Gas, electricitiy, & water 222 7075 39 8erviGes 2889 2418 5849 1104 3419 3066 3404 40 Sales by final buyers 18 13 14 15 18 15 23 41 Vehicles, ships, & aircraft 42 Plant & machinery 43 Buildings & works 44 Increase in stocks 45 Personal sector 326 153 -90 236 457 556 1152 802 46 Companies 2083 47 Publicsector -2344 1236 56 48 Production 71 72 75 81 88 88 126 1363 219 49 45 837 562 685 49 Consumption 646 18 19 18 17 16 15 15 167 1138 135 50 Accumulation 51 IdentIfied financial assets 52 Unidentified items 21 Total 10000 10000 10000 10000 10000 10000 10000 10000 10000 10000 10000 10000 10000 10000 10000 10000 10000 10000 10000 10000 10000 10000 10000 10000 10000 (Table continues on the following page.) Table 8.2. (Continued) 26 27 28 29 30 31 32 33 34 36 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 1 Income from employment 3096 4436 2887 3843 2235 5660 2 Gross profits & other trading income 2886 1275 1422 1426 3339 2432 -4875 -7677 -7143 3 Taxes on expenditure (net) 3462 9 261 538 139 414 663 622 -594 162 241 483 370 394 708 286 57 -30 4 Wages U salaries (1ncl. Forces' pay) 5 Employers' contributions 6f Pent, dividends, & Interest 8222 7 Households: income £ 0-10 8 ' £IO-20 9 £20-30 10 £30-40 11 £40-SO 12 " £50-f0 13 £80+ 14 Private nonproftt institutions 15 Companiss 16 Public corporations 17 Central government 594 18 Local authorities 19 Food 20 Drink y tobacco 21 Housing 22 Fuel light 23 Clothing U footwear 24 Durable 6 other household goods 26 Other miscellanous goods 26 Motor vehicles (ind. running costs) 27 Travel U communicaUons 28 Entertainments, hotels U restaurants 29 Other services 30 Military defence I 31 National health service 32 Other central government purposes 33 Locai authorities Purposes 34 Aricultre, etc., mining Oqxrying 488 13 225 40 161 49 866 179 1678 8 66 654 149 35 Metals U metal products 1731 3578 342 1245 192 552 486 1380 570 256 9198 6973 -3551 3866 36 Other manufactrirng 577 1463 207 1562 863 931 2360 1037 1781 609 583 157 142 7477 2844 37 ConstrmcUon 73 439 124 472 292 253 90 33 152 29 78 8767 514 29 38 Gas, electricity, & water 97 116 161 255 217 276 202 31 138 542 18 39 Services 2946 9838 6814 8478 4468 7413 8384 8192 802 1236 1628 629 857 1368 832 968 93 2564 40 buyers SBlesby final 1000 470 842 308 -422 -2661 746 10 181 34 29 26 23 -3255 -259 136 41 Vehicles, ships. O aircraft 797 3047 804 42 Plant U machinery 1418 8433 5784 43 Buildings O works 5715 4630 12353 44 Increase in stocks 182 762 225 45 Personal sector 969 46 Companies 2298 47 Public sector 2679 218 48 ProducUon 295 153 505 69 1367 228 853 111 419 1259 2201 239 165 471 1981 1391 4813 49 Consumption 428 50 Ahcumulatlon 1184 81 Identifed financial assets 9642 -5136 -4975 14871 52 Unidentified items -5558 5724 -315 -4871 Total 10000 10000 10000 100001 10000 10000 10000 10000 10000 10000 10000 10000 10000 100001 0( 10000 10000 10000 100001 10000 10000 10000 10000 10000 0 0 Note: Components do not always add to totals because of rounding. 162 Multipliers and SAM-Based Models The M-notation is used by Pyatt and Round, who also denote (I - B) - 1C by A*. If we write out the expressions in rows 5, 6, and 7 of (8.1) in terms of the submatrices of A, we obtain the following. First, I O O (8.2) o Mt = (I-A 2 2 )-' O . O O (I -A33)- Thus the multiplier effects included in Ml arise from the repercussions of the initial injection within the group of accounts (or subsystems) which it originally entered, and so may be said to measure the intragroup effects. Second, D O O M2 = O E O O O F where D = [I - A13 (I - A 33 )-lA 32 (I -A22)- A2 (8.3) E = [I - (I - A 22)-lA 2 1 A, 3 (I -A33) 1A321 F = [I - (I - A 33)-'A 32 (I - A22Y'A 2 1A, 3]'. Thus the multiplier effects included in M2 arise from the repercussions of the initial injection when it has completed a tour through all three groups and returned to the one that it had originally entered and so may be said to measure the intergroup effects. [ Finally, I -A 33 )-'A~(I A1 3 (I -A 3 3 ) 'A 32 A13 3= (I-A 2 2 ) - A2 1 I (I-A ,'A 2 1 AI 3 (I -AA33) - lA3(I-A22) A2 1 'A (I -A 3 3 ) - 32 I Thus the multiplier effects included in M3 arise from the repercussions of the initial injection when it has completed a tour outside its original group without returning to it, and so may be said to measure the extragroup effects. It is a simple matter to express M in terms of additive components, and for some purposes it is convenient to do this. Thus (8.4) M = I + (Ml-I) + (M 2 -I)Ml + (M3 -I)M 2 Ml. In this expression we start with a matrix of injections, the unit matrix I. In the second term we add on the effects coming from Ml. To this we add on in the third term the effects coming from M2. And, finally, in the fourth term we add on the effects coming from M3 . The matrices Ml, M2, M2 Ml, M3 , M3 M 2M, =M, (M 2 - I)M, and (M3 - I)M 2 M, are set out in tables 8.3 to 8.9. THE INTERPRETATION OF THE MATRIX MULTIPLIERS On the assumption that when an account receives money it spends it in the proportions shown in the A-matrix (table 8.2), the numbers in the columns of M (table 8.7) show the ultimate consequences for each account of 1,000 units received exogenously by the account at the head of the column. For instance, the figure of 2,416 in the top left-hand corner of M indicates Disaggregation of the Household Sector 163 that the exogenous receipt of 1,000 in income from employment will ultimately lead to the generation of an additional 1,416 of such income. How does this come about? Clearly it cannot be an intragroup effect since from (8.2) the first diagonal submatrix of MI is the unit matrix. Equally it cannot be an extragroup effect since we are considering the effect of a change in something on itself. It muist, therefore, be an intergroup effect arising from the movement of the injection through the system and back through its starting point until it has all leaked away. The course it follovvs is shown in (8.3): it moves from group 1 to group 2 to group 3, then back to group 1, and so on. Thus in terms of (8.4), the element in the first row and first columrn of M, Ml, say, can be expressed as Ml, = 1,000 + 0 + 1,416 + 0 = 2,416. In a similar way the injection of 1,000 of income from employment will also generate as an intergroup effect, 671 of profits and other trading income. Thus M2=0 + 0+671 +0 =671. If we now look at the opiposite corner of the matrices we can see how to combine intragroup and intergroup effects and. thus extend the familiar results of input-output analysis. Let us take as an example the element in 63 , the demand for the output of other manufacturing generated by a demand for metals and metal products. An extra 1,000 units of demand for these products gives m,,.,=0+ 123 + 617 +0 = 740. Of the 123 coming from M1, we can see from A that 104 represent direct requirements, leaving 19 for indirect requirements. Of the remaining 617, only 447 appear at the intersection of row 36 and column 35 of M2 , but it must be remembered that M1 , M2 , and M3 combine multiplicatively so that the 447 is only on.e element in the sum. So far, extragroup effects have not arisen because we have been looking at the ultimate effect on accounts belonging to the same subsystem as the account initially stimulated. Let us now ask wrhat would be the effect on household food expenditure of providing different types of households with an addilional 1,000 units of income. The answers for the poorest and the richest households are given in M,, respectively, and these elements can be decom- 9 7 and nil 9 ,13, posed as follows: (8.5) M1 9 .7 =0 + 0+0+619 = 619 and (8.6) m9 3 = 0 + 0 +0+417 = 417. From M3, the poorest households would give rise to an additional expenditure of 260 and the richest to an additional expenditure of 91: these are the proportions that appear in the A- matrix, and so they represent the direct expenditure of the households initially given the extra income. But households do many things besides buying food and so, as we have seen in (8.5) and (8.6), the ultimate effect on food expenditure is 619 if the injection is applied to the poorest Table 8.3. Matrix Multiplier for Intragroup Effects, 1000 M1 , United Kingdom, 1968 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 I Income from employment 1000 2 Gross profits & other trading income 1000 3 Taxes on expenditure (net) 1000 4 Wages & salaries ( incl. Forces' pay) 1000 5 Employers' contributions 1000 6 Rent, dividends & interest 76 75 1507 14 45 65 77 84 81 93 8 832 1818 197 306 7 llouseholds: income E 0-10 17 24 38 1002 8 13 15 16 17 19 39 30 48 70 a 8 " " L10-20 84 103 109 3 1013 20 23 25 26 29 51 72 134 100 26 9 L20-30 255 256 100 2 9 1013 15 17 17 19 32 63 123 64 31 10 " " L30-40 270 249 86 2 6 9 1010 12 12 13 20 52 104 42 27 11 "" 40-50 183 169 66 1 4 6 6 1007 7 8 10 39 80 24 18 12 " " .50-60 112 104 31 * 2 3 3 31003 4 4 18 38 11 9 13 " " L60 + 160 175 112 1 4 7 8 8 8 1009 7 63 136 24 26 14 Private non-profit institutions 20 20 42 12 15 17 19 21 21 28 1003 32 52 29 9 15 Companies 36 35 709 7 21 31 36 40 38 44 4 1392 855 93 144 16 Public corporations 3 .3 57 1 2 2 3 3 3 3 * 31 1068 7 12 17 Central government 217 214 288 20 115 190 211 236 247 273 21 302 381 1070 65 18 Local authorities 34 34 101 3 18 30 33 37 39 43 3 76 127 161 1022 19 Food 20 Drink and tobacco 21 llousing 22 Fuel and light 23 Clothing and footwear 24 Durable and other household goods Z5 Other miscellaneous goods 26 Motor vehicles (incl. running costs) 27 Travel and cousiiunications 28 Entertainme.nts, hotels & restaurants 29 Other services 30 Hilitary defence 31 National health service 32 Other central governmient purposes 33 Local authorities purposes 34 Agriculture erc., mining & quarrying 35 Metals and metal products 36 Other manufacturing 31 Construction 38 Gas, electricity and water 39 Services 1209 .±1 2 1I3 124 I27I28 I2526 29 30 31 32 33i34 35436 _37_38__ _ 1 Income from employment 2 Gross profits & other trading Income 3 Taxes on expenditure (net) 4 Wages & salaries (icl. Forces' pay) 5 Employers' contributions 6 Rent, dividends & interest 7 Households: income L 0-10 p 8 of If LIO-20 10 '" L30-40 11 " 40-50 12 " L50-60 13 " L60 + 14 Private non-profit institutions 15 Companies 16 Public corporations 17 Central government 18 Local authorities woo0 19 Food 1000 20 Drink and tobacco 1000 21 Oheusing 1000 22 Fuel and light 1000 23 Clothing and footwear 1000 24 Durable & other household goods 1000 25 other miscellaneous goods 1000 26 Motor vehicles (inc. running costs) 1000 27 Travel and communications 1000 28 Entertainments,hotels & restaurants 1000 29 Other services 1000 30 Military defence 1000 31 National health service 1000 32 Other central government purposes 1000 33 Local authorities purposes 187 21 21 255 52 40 68 12 9 71 8 19 49 28 39 1029 22 96 40 181 10 34 Agriculture etc., mining & quarrying 47 20 61 73 43 215 101 188 31 34 28 386 72 167 59 79 1014 63 155 80 31 35 Mietalsand metal products 491 218 122 185 551 406 448 101 66 211 58 106 227 175 168 266 123 1043 212 124 67 36 other manufacturing 8 2 185 19.5 6 6 4 4 5 11 50 18 53 34 29 11 7 1004 21 4 37 Construction 21 9 36 722 20 21 19 12 16 17 14 30 30 37 43 32 33 27 14 1011 16 38 Gas, electricity and water 392 282 629 233 443 405 4331 337 10011 727 863 519 7951 896 863-1140 1511 188 1211 133 1017 39 Services Note: than 0.5. an entryofless denotes An asterisk Table 8.4. Matrix Multiplier for Intergroup Effects, 1000 M2 , United Kingdom, 1968 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 I Income from employmiient 2416 1133 1439 2 Gross profits & other trading income 671 1535 672 3 Taxes on expenditure (net) 469 362 1408 4 Wages & salaries (incl. Forces' pay) 1000 1311 1190 1114 1031 964 940 809 1L55 572 1311 5 tEployera' contributions 1000 130 118 110 102 96 93 80 114 57 130 6 Rent, dividends & interest 1006 917 832 778 719 672 656 561 783 376 880 7 Households: incoume . 0-10 1073 67 64 59 55 54 46 63 29 69 3 o LI10-20 237 1216 203 186 176 172 147 205 99 229 9 " It20-30 " 480 437 1409 316 354 3V,5 296 419 206 474 10 " " 130-40 489 444 415 1384 359 351 301 427 210 483 11 " " 140-50 331 301 282 261 1244 238 204 290 142 328 12 " " .50-60 205 186 174 161 151 1147 126 180 88 203 13 E 60 + 383 347 324 300 280 273 1235 333 163 377 14 Private non-profit institutions 65 59 55 51 48 47 40 1056 27 62 15 Comlpanies 432 391 366 338 316 309 264 368 1000 177 414 16 Public corporations 35 31 29 27 25 25 21 29 1000 14 33 17 Central government 870 806 766 713 667 656 557 747 1347 820 18 Local authorities 271 253 243 226 212 209 177 231 104 1249 19 Food 20 Drink and tobacco 21 Hlousing 22 Fuel and light 23 Clothing and footwear 24 Durable & other househiold goods 25 Other miscellaneouis goods - 26 Motor vehicles (incl. running costs) 27 Travel and conmlunications 28 Entertaininiits, hutels & restaurants 29 Other servicas 30 Military defence 31 National hiealth service 32 other central government purposes 33 Local authorities purposes 34 Agriculture etc., muining & quarrying 35 Metals and metal products 36 Other manufacturing 37 Construction 38 Gas, electricity and watar 39 Services 19 j20 21 22 23124125 26127128 29130 3113213313435 36 37- 38 39]. I Income from employment 2 Gross profits & other trading income. 3 Taxes on expenditure (net) 4 Wages & salariet (incl. Forces' pay) 5 Employers' contributions 6 Rent, dividends & interest - - - - - - - - - - - - - - - - - J 8 10 H~~~~~~~~~~~~~7 louseholds: of income 1. 0-10 "s Is" 1.1-20 130-40 11 " 140-50 12 " 150-60 13 " 160 + 14 Private non-profit institutions 15 Comipanies 16 Public corporaLions 17 Central government 18 Local authorities 1007 215 63 2 27 43 43 137 .~10 21 6 16 26 25 244 281 211 268 256 397 19 Food 4 1123 36 1 15 24 25 78 * 6 12 3 9 15 14 144 166 124 158 151 234 20 Drink and tobacco 4 127 1037 1 16 25 26 81 * 6 13 3 10 16 15 142 163 123 156 150 231 21 Hiousing 2 53 15 1000 7 10 11 33 * 3 5 1 4 6 6 58 66 50 64 61 94 22 Fuel and light 3 84 24 1 1011 17 17 54 * 4 8 2 6 10 10 99 113 85 108 103 159 23 Clothing and footwear 3 76 22 1 10 1015 15 49 * 4 8 2 6 9 9 89 102 76 97 93 144 24 Durable & other household goods 2 67 19 1 8 13 1013 42 * 3 7 2 5 8 8 77 88 66 84 81 125 25 other miscellaneous goods 3 82 24 1 10 16 17 1052 * 4 8 2 6 10 9 99 113 85 108 103 160 26 Motor vehicles (incl. running costs) 1 42 12 * 5 8 8 26 1000 2 4 1 3 5 5 48 55 41 53 50 78 27 Travel and communications 2 65 19 1 8 13 13 41 * 1003 6 2 5 8 7 75 86 64 82 79 121 28 Entertainments, hotels & restaurantS 3 79 23 1 10 16 16 50 * 4 1008 2 6 10 9 90 101 77 97 95 144 29 Othierservices 5 161 47 1 20 32 32 102 8 16 1004 12 20 18 92 113 88 113 112 164 30 Military defence 3 93 27 1 12 19 19 59 * 4 9 2 1007 11 11 54 66 51 66 65 95 31 National hiealthservice 2 68 20 1 8 13 14 43 * 3 7 2 5 1008 8 39 48 37 48 47 69 32 Other central government purposes 9 257 75 2 32 51 52 163 * 12 25 7 20 31 1029 109 140 111 144 144 206 33 Local authorities purposes 3 97 28 1 12 19 20 62 * 5 10 2 7 12 If 1098 113 85 109 105 161 34 Agriculture etc., mining & quarrying 6 167 49 1 21 33 34 106 * 8 17 4 13 20 19 136 1160 122 156 151 229 35 Metals & metal products 13 387 113 3 49 77 78 246 1 19 38 10 30 47 44 386 447 1337 430 413 634 36 Other manufacturing 2 52 15 * 6 10 10 33 * 2 5 1 4 6 6 45 52 40 1051 49 74 37 Construction 3 78 23 1 10 15 16 50 k 4 8 2 6 10 9 74 85 65 83 1080 121 38 Gas, electricity and water 321 933 2711 81 1171 1851 188 593 2 45 93 24 71 114 107 785 922 1701 1897i 870123151 39, Services Note: An asterisk denotes an entry of less than 0.5. Table 8.5. Matrix Multiplier for Intragroup and Intergroup Effects, 1000 M2 M1 , United Kingdom, 1968 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 1 Income from employment 2416 1133 1439 2 Gross profits & oLher trading income 671 1535 672 3 Taxes on expenditure (net) 469 362 1408 4 Wages & salaries (incl. Forces' pay) 2289 1289 909 1354 1348 1360 1306 1271 1257 1166 1361 664 1133 1236 1539 5 Employers' contributions 128 1128 90 134 134 135 129 126 125 116 135 66 112 122 152 6 Rent, dividends and interest 968 967 2130 960 983 1009 981 963 950 895 934 1285 2594 1028 1342 7 Hiouseholds: income L 0-10 90 97 88 1077 84 90 89 88 88 84 113 66 110 136 90 8 " to IO-20 317 337 271 247 1257 266 259 255 254 239 293 190 337 318 296 9 to L20-30 727 727 432 498 502 1512 493 482 477 445 527 304 536 512 589 10 " L 30-40 750 729 423 506 508 516 1496 485 479 446 524 298 525 498 595 11 " " 1.40-50 508 494 295 343 344 349 336 1328 324 302 351 205 365 333 403 12 " " t50-60 314 306 173 213 213 215 207 202 1200 186 216 122 214 203 248 13 , " 1.60 + 534 549 375 397 397 402 387 377 373 1347 399 255 463 379 468 14 Private non-profit institutions 84 84 87 79 81 84 84 84 83 85 1068 64 107 88 83 15 Companies 456 455 1002 452 463 475 462 453 447 421 439 1605 1221 484 632 16 Public corporations 36 36 80 36 37 *38 37 36 36 34 35 48 1098 39 51 17 Central government 1092 1088 889 918 1023 1113 1099 1099 1104 1057 907 736 1128 1852 1036 18 Local auithorities 309 309 288 282 303 320 313 309 309 289 277 211 359 401 1318 19 Food 20 Drink aLnd tobacco 21 lHousing 22 Fuel and light 23 Clothing and footwear 24 DuLrable & other household goods 25 Other muiscellaneous goods 26 Motor vehicles (incl. running costs) 27 Travel and conomunications 28 Entertainments, Isoccis & restaurants 29 Other services 30 Military defence 31 National health service 32 Otier central governmient purposes 33 Local authorities purposes 34 Agricultsure etc. , mining & quarrying 35 Metals and metal products 36 Other manufacturing 37 Construction 38 Gas, electricity and water 39 Services 19j20 21 224 23 24 25 2 27 8 29 30 31 32 33 34 35 36 37 38 39 ( I I fI( ~~~~~~~~~~~~~1 from employment Incomae 2 Gross profits & other trading income 3 Taxes on expenditure (net) 4 Wages & salaries (incl. F~orces'pay) 5 Employers' contributions 6 Rent, dividends and interest 7 Households: income 1. 0-10 8 to I" 110-20 9 " 120-30 11 " 140-50 12 150-60 L 13 " 160-i- 14 Private non-profit institutions 15 Companies 16 Public corporations 17 Central government 18 Local authorities 1332 387 419 406 350 366 361 352 427 376 392 368 424 496 448 401 387 345 419 410 434 19 Food 196 1224 246 239 206 215 212 205 252 221 231 217 250 292 264 237 228 203 247 241 256 20 Drink.and tobacco 193 227 1244 237 204 213 210 206 248 219 228 214 247 289 261 234 225 201 244 239 252 21 Housing 79 93 100 1096 83 87 86 84 101 89 93 87 100 118 106 95 91 82 99 97 103 22 Fuel and light 134 153 168 163 1140 147 145 140 172 151 157 148 170 199 180 162 156 138 168 165 174 23 Clothing and footwear 120 139 151 147 127 1132 131 126 155 136 142 133 154 179 162 146 140 125 152 149 157 24 Durable & other household goods 105 121 132 128 110 115 1114 110 135 118 123 116 134 156 141 127 122 109 132 129 137 25 Other miscellaneous goods 134 151 167 163 140 146 144 1139 172 151 157 143 170 199 180 162 156 139 168 165 175 26 Motor vehicles (cine. running costs) 65 75 82 80 69 72 71 69 1084 74 77 72 83 97 88 79 76 68 82 61 85 27 Travel and communications 102 117 128 125 107 112 110 107 131 1115 120 112 130 151 137 123 118 105 128 126 133 28 Entertainmients,hotels & restaurants 121 142 152 149 127 133 132 128 155 137 1143 134 154 180 163 147 140 125 152 151 158 29 Other services 138 231 194 170 153 165 163 190 177 158 169 1153 180 213 193 157 157 142 175 174 179 30 Military defence 80 134 113 99 89 96 95 III 102 92 98 89 1105 124 112 91 91 83 101 101 104 31 National health service 58 97 82 72 65 69 69 80 74 67 71 64 76 1090 81 66 66 60 73 73 75 32 Other central government purposes 125 346 260 218 200 218 216 274 222 201 218 193 231 275 1249 190 195 179 221 221 225 33 Local authorities purposes 322 187 194 420 195 190 216 161 182 223 168 168 221 230 221 1191 178 235 209 348 185 34 Agriculture etc.,*mining & quarrying 239 286 315 309 250 435 318 418 277 253 258 599 320 458 322 305 1236 262 397 319 282 35 Metals and metal products 1023 880 804 836 1116 999 1033 690 749 814 690 695 908 972 889 903 740 1594 883 783 761 36 Other manufacturing 71 86 266 96 72 77 76 76 84 76 85 119 98 147 120 102 83 72 1082 99 85 37 Construction 123 140 168 847 128 135 132 127 147 132 135 143 161 191 182 154 151 133 143 1137 149 38 Gas, electricity and Hater 1498 1784 2082 1590 1631 1660 1676 1640 2418 1983 2186 1742 2217 2567 2375 1446 1426 1331 1515 1508 2456 39 Services '.0 Table 8.6. Matrix Multiplier for Extragroup Effects, 1000 M3,United Kingdom, 1968 1 2 3 4 5 -6 7 8 10 11 12 13 14 15 16 17 18 1 Income from employment 1000 421 375 345 318 297 287 251 386 205 454 2 Gross profits & other trading income 1000 217 188 169 154 143 138 119 176 87 206 3 Taxes on expenditure (net) 1000 122 132 138 132 125 127 103 96 25 77 4 Wages & salaries (illel.Forces' pay) 910 1000 5 Employers'contributions 90 1000 6 Rent, dividends and interest 76 1013 224 1000 7 Houselholds: income L 0-10 18 33 55 1000 8 I " L10-20 85 110 82 1000 9 " " L20-30 -255 124 56 1000 10 ' " 30-40 268 121 39 1000 11 " " L40-50 182 85 22 1000 12 " L50-60 112 57 11 1000 13 " " L60 + 161 200 24 1000 14 Private non-profit institutions 20 35 24 1000 15 Conmpanies 36 477 105 1000 16 Public corporations 3 38 8 1000 17 Central government 217 267 825 1000 18 Ltocal authorities 34 79 371 1000 19 Food 168 113 56 260 221 191 158 135 126 91 20 Drink and tobacco 101 65 27 71 97 108 95 90 97 70 10 21 llousing 96 68 36 179 131 100 85 78 68 56 106 22 Fuel and lightt 38 28 16 101 63 42 34 29 25 18 24 23 Clothing and footwear 68 45 18 52 60 66 67 63 60 53 44 24 DuIrable& other liotseholdgoods 61 41 18 56 59 59 58 58 57 43 44 25 Other miscellaneousgoods 53 36 16 52 51 53 50 47, 46 39 44 26 Motor vehicles (inel. running costs) 70 45 15 18 37 56 75 68 74 62 44 27 Travel and coitmnunications 33 23 10 31 35 32 3C 28 29 26 34 28 Entertainments, hotels & restaurants 51 37 15 40 40 44 44 45 47 47 118 29 Other services 57 48 22 63 51 42 41 44 44 67 309 30 Military defence 39 48 148 179 31 National health service 23 28 86 104 32 Other central governmentpurposes 16 20 62 75 33 L.ocalauthoritiespurposes 26 60 281 _ _ _ _ 759 34 Agricultureetc., mining & quarrying 64 46 40 92 75 65 56 50 47 36 27 11 29 35 Metala and metal products 78 66 107 59 56 56 55 52 51 42 46 89 45 36 Other manufacturing 250 182 165 274 254 241 219 201 194 156 131 56 127 37 Construction 26 22 30 39 29 23 20 18 16 14 25 15 26 38 Gas, electricityanid water 46 36 38 91 61 45 38 34 30 23 31 11 33 39 Services 460 387 566 446 403 374 348 329 320 292 534 243 655 19 20 21 2 .2 4 5 2 7LI2 k29 30 3 32 33 34 35 36 37 38 39_________________ I1111 1 1 1 1 1 1 443 289 384 223 566 3~~~~~~~~~~~~~10 289 127 142 143 334 243 2 1 I Incomefrom employment Gross profits S other trading income 18 543 158 41 68 108 110 345 11 26 541 14 41 66 62 -59 16 24 46 37 39 3 'Taxeson expenditure (net) 282 404 263 350 203 515 4 Wages & salaries (mncl. Forces' pay) 28 40 26 35 20 5i 5 Employers' contributions 4 122 35 1 15 24 25 77 6 12 3 9 15 14 302 166 171 184 363 298 6 Rent, dividends and Interest 1 30 9 4 6 6 19 1 3 1 2 4 3 12 13 11 14 17 20 7 Househiolds:income L 0-10 2 45 13 * 6 9 9 28 * 2 4 1 3 5 5 53 53 42 52 59 78 8 of if lIO-20 1 31 9 * 4 6 6 19 * 1 3 1 2 4 4 112 130 93 119 101 177 9 " 20-30 1 21 6 * 3 4 4 13 I* 2 I i 2 3 2116135D 70IL 961210 1Q3 10L34 * 12 4 * 2 2 2 8 * I 1 1 1 80 92 65 83 70 125 11 " 40-50 * 6 2 ~I 1 1 4 I 1 1 50 57 41 52 44 77 12 " 50-60 * 13 4 * 2 3 3 8 I 1I 1 2 2 106 97 76 92 104 141 13 " 160 + * 13 4 * 2 3 3 8 I 1 * 1 2 1 15 14 11 14 17 21 14 Private non-profit institutions 2 57 17 * 7 11 12 36 * 3 6 1 4 7 7-142 78 81 87171 140 15 Comipanies * 5 1 I 1 1 3 - * I I 11 6 6 7 14 11 16 Public corporations 15 448 130 4 56 89 90 285 1 22 44 11 34 55 51 95 144 121 160 168 220 17 Central government 7 201 58 2 25 40 41 128 * 10 20 5 15 25 23 11 31 30 42 48 53 18 Local authiorities 1000 19 Food 1000 20 Drink and tobacco 1000 21 liousing 1000 22 Fuel and lighit 1000 23 Clothiingand footwear 1000 24 Durable & other houiseholdgoods 1000 25 Othaermiscellaneous goods 1000 26 Motor vehicles (cine. running costs) 1000 27 Travel and commujnications 1000 28 Entertainmnents,hotels & restaurants 1000 29 other services 1000 30 Military defence 1000 31 National health service 1000 32 Other central government purposes 1000 33 Local authorities pnirposes 1000 34 Agriculture etc., mining & quarrying 1000 35 Metals and metal products 1000 36 Other mranufacturing 1000 37 Construction 1000 38 Gas, electricity and water 1000 39 Services Note: An asterisk denotes an entry of less than 0.5. Table 8.7. Matrix Multiplier for All Effects, 1000 M3 M2M1 1000 M 1000 (I - A)-', United Kingdom, 1968 1 F2 3 4 56 7 f 9 T 12 13 14 5 67 18 I Incomne fromn employmnent 2416 1133 1439 1416 1417 1000 1488 1482 1495 1435 1397 1382 1282 1496 730 1245 1358 1691 2 Gross profits & othier trading income 671 1535 672 671 671 471 721 709 711 680 661 652 603 702 343 587 b33 791 3 Taxes on expenditure- (net) 469 362 1408 469 469 312 463 480 496 478 464 464 419 453 223 388 387 476 4 Wages & salaries (ind. Forces' pay) 2199 1030 1309 2289 1289 909 1354 1348 1360 1306 1271 1257 1166 1361 664 1133 1236 1539 5 Employers' contributions 218 102 130 128 1128 90 134 134 135 129 126 125 116 135 66 112 122 152 6 Rent, dividends and interest 968 1721 1105 968 967 2130 960 983 11009 981 19631 950 895 934 11285 2594 1028 11342 7tHousehiolds:incoLue£ 0-10 90 90 125 90 97 88 1077 84 90 89 88 88 84 113 66 11 13 90 8 " " LIO-20 318 295 312 317 337 271 247 1257 266 259 255 254 239 293 191 337 318 296 9 " 20-30 727 501 531 727 727 432 498 502 1512 493 482 477 445 527 304 536 512 589 10 " 30-40 148 504 522 750 729 423 506 508 516 1496 485 479 446 524 298 525 498 595 Ii " 40-SO 507 344 350 508 494 295 343 344 349 336 1328 324 302 351 20S 365 333 403 12 "1 " 50-60 313 218 214 314 306 173 213 213 215 207 202 1200 186 216 122 214 203 248 13 "1 " 60 + 535. 499 401 534 549 375 397 397 402 387 377 373 1347 399 255 463 379 468 14 Private non-profit institutioas 84 86 86 84 84 87 79 81 84 84 84 83 85 1068 64 107 88 83 15 Comipanies 456 810 520 456 455 1002 452 463 475 462 453 447 421 439 1605 1221 484 632 16 Public corporations 36 65 42 36 36 80 36 37 38 37 36 36 34 35 48 1098 39 51 17 Cenitral government 1091 954 1654 1091 1088 889 918 1023 1.1131099 1099 1104 1057 907 736 1128 1852 1036 18 Local authorities 3091 294 624 309 309 288 282 303 320 313 309 309 289 277 211 359 401 1318 19 Food 508 384 396 508 509 328 619 585 564 518 488 475 417 389 231 407 388 424 20 Drink and tobacco 301 224 227 301 300 190 282 310 325 304 296 300 260 234 133 235 220 249 21 sing Hiou 294 226 234 294 295 195 390 345 320 297 286 274 249 334 138 242 230 248 22 Fuel and light 119 92 97 119 120 80 186 150 132 120 114 110 98 118 57 99 96 101 23 Clothing anidfootwoear 204 154 155 205 204 130 196 206 215 210 -204 199 183 197 91 161 150 170 24 Duirable & octherhouzehioldgoods 184 139 141. 184 184 118 187 191 194 188 185 183 161 182 83 146 137 153 25 Othier miscellaneous goods 160 121 123 160 160 103 165 166 171 163 158 156 142 165 72 128 119 133 26 Motor vehiciles (cine. running costs) 206 153 151 206 204 128 162 182 204 218 208 212 191 194 90 158 146 169 27 Travel and commnunications 99 76 77 99 100 65 102 106 105 101 98 97 90 109 45 80 75 83 28 Entertainmnents, hotels & restaurants 154 119 119 154 154 102 151 153 159 155 154 154 149 234 72 126 116 130 29 Other services 181 147 146 1181 1182 127 197 187 182 177 177 175 193 448 90 158 143 156 30 M-ilitarydefence 195 171 296 195 195 159 164 183 199 197 197 198 189~ 162 132 202 332 165 31 National healtlh service 113 99 172 113 113 92 95 106 116 1'14 114 115 110 94 76 117 192 108 32 central governmnent outher purposes 82 72 124 82 82 67 69 77 84 83 83 83 80 68 55 85 139 78 33 1.oeal autburitiespurposes 2351223 473 235 234 2191214 230 243 23812341 235 219 211 160 1273 304 1000 34 Agricultureetc., mininig & quarrying 203 158 179 203 204 137 240 227 221 207 19S 194 114 185 98 170 171 202 35 Mietalsand metal products 284 229 308 284 264 203 2Z6 285 296 2S9 233 262 2u0 270 153 255 312 295 36 Other manufacturing 803 622 714 803 803 539 859 854 859 817 789 777 703 749 386 670 682 811 37 Construction 93 74 95 93 93 66 109 102 99 94 91 89 82 99 48 82 91 107 38 Gas, electricityand wjater 153 121 144 153 153 106 203 178 165 154 148 144 131 152 76 132 137 165 39 Services 1636 1320 1718 1636 1636 1170 1689 1696 1718 1655 1616 1601 1499 1821 856 1458 1598 2094 19 20 21 22 1 25 26 7 128 29 30 131 132 33 34 37 38 39 - 1403 1513 1750 1634 1473 1539 1513 1421 1829 1598 1666 1582 1801 2107 1899 1657 1690 1460 1801 1621 1858 1 Income from employment 684 708 811 963 696 712 712 650 843 750 768 712 841 972 887 929 723 706 807 1013 857 2 Gross profits & other trading income 347 940 592 411 427 481 477 705 441 409 458 389 476 576 523 327 386 351 448 421 447 3 Taxes on expenditure (net) 1277 1377 1592 1487 1340 1401 1376 1293 1664 1454 1516 1439 1639 1917 1728 1508 1537 1329 1639 1475 1691 4 Wages & salaries (incl. Forces' pay) 126 136 158 147 133 139 136 128 165 144 150 143 162 190 171 149 152 132 162 146 167 5 Employers' contributions 877 1043 1086 1191 912 946 943 924 1092 972 1007 928 1095 1273 1159 1139 947 905 1054 1243 1109 6 Rent, dividends and interest 66 102 90 83 72 77 76 85 84 75 80 73 85 101 91 78 75 68 83 85 85 7 Households: income 1 0-10 223 284 287 279 237 249 246 251 285 252 264 245 285 334 302 270 255 231 279 284 289 8 " " E10-20 463 -0. -n 520 ,81 jg,io ,60 n, 487 50,91 i,…66 502 Al4S31cl0 57c1 54 Q 547 I 515 591 69Q,2 6-22 c-55? r- 3414b 1 '.:t.Sr.4 6061 9 of of 20.30 203 473 528 591 571 496 518 511 487 610 535 558 526 603 705 637 570 556 491 598 574 620 10 of is 130-40 321 356 400 388 337 351 346 329 414 364 379 357 410 479 432 388 377 333 406 390 421 11 " " L40-50 199 219 248 242 209| 218 215 203 257 225 235 221 254 297 268 241 234 207 252 243 261 12 ' " 50-60 372 409 459 466 387 402 398 376 474 418 433 407 470 548 496 461 426 385 463 474 482 13 " " 160 + 61 78 78 77 65 68 67 68 77 69 72 66 77 91 82 74 69 63 76 79 78 14 Private non-profit institutions 413 491 511 561 429 445 444 435 514 458 474 437 515 599 546 536 446 426 496 585 522 15 Conmpanies 33 39 41 45 34 36 35 35 41 37 38 35 41 48 44 43 .36 34 40 47 42 16 Public corporations 773 1293 1085 951 858 921 912 1064 986 884 945 854 1008 1192 1050 877 878 795 976 970 1001 17 Central government 230 456 343 284i 263i 287 284i 361i 292 265| 287 254| 304| 362 328| 251| 258 2351 291 2911 2971 18 Local authorities 1332 387 419 406 350 366 361 352 427 376 392 368 424 496 448 401 387 345 419 410 434 19 Food 196 1224 246 239 206 215 212 205 252 221 231 217 250 292 264 237 228 203 247 241 256 20 Drink and tobacco 193 227 1244 237 204 213 211 206 248 219 228 214 247 289 261 234 225 201 244 239 252 21 |lotising 79 93 100 1096 83 87 86 84 101 89 93 87 100 118 106 95 91 82 99 97 103 22 Fuel and light 134 153 168 163 1140 147 145 140 172 151 157 148 170 199 180 162 156 138 163 165 174 23 Clothing and footwear 120 139 151 147 127 1132 131 126 155 136 142 133 154 179 162 146 140 125 152 149 157 24 Durable & other household goods 105 121 132 128 110 115 1114 110 135 118 123 116 134 156 141 127 122 109 132 129 137 25 Other miscellaneous goods 134 151 167 163 140 146 144 1139 172 151 157 148 170 199 180 162 156 139 168 165 175 26 Motor velhicles(incl. running costs) 65 75 82 80 69 72 71 69 1084 74 77 72 83 97 88 79 76 68 82 81 85 27 Travel and communications 102 117 128 125 107 112 110 107 131 1115 120 112 130 151 137 123 118 105 128 126 133 28 Entertainments, thotels& restaurants 121 142 152 149 127 133 132 128 155 137 1143 134 154 180 163 147 140 125 152 151 158 29 Other services 138 232 194 170 153 165 163 190 177 158 169 1153 1t0 213 193 157 157 142 175 174 179 30 |iitary defence 80 134 113 99 89 9o 95 111 102 92 98 89|1105 124 112 91 91 83 101 10i 104 31 Nacional health service 58 97 82 72 65 69 69 80 74 67 71 64 76 1090 81 66 66 60 73 73 75 32 Other central government purposes 175 346 260 216 200 218 216 274 222 201 218 193 231 275 1249 190 195 179 221 221 225 33 Local authorities purposes 322 187 194 420 195 190 216 161 182 223 168 168 221 230 221 1191 178 235 209 348 185 34 Agriculture etc., mining & qurrying 239 286 315 309 250 435 318 418 277 253 258 599 320 458 322 305 1236 262 397 319 282 35 MIetals and metal products 1023 880 804 836 1116 999 1033 690 749 814 690 695 908 972 889 903 740 1594 883 783 760 36 OLher manufacturing 71 86 266 96 72 77 76 76 84 76 85 119 98 147 120 102 83 72 1082 99 85 37 Construction 123 140 168 847 128 135 132 127 147 132 135 143 161 191 182 154 151 133 143 1137 149 38 Gas, electricity and water 1498 1784 2082 1590 1631 1660 1674 1640 2418 1983 2186 1742 2217 2567 2375 1446 1426 1331 1515 1508 2456 39 Services Table 8.8. Additional Effects Associated with M2 : 1000 (M 2 -I) M1 , United Kingdom, 1968 I 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 1 Incone fronm employment 1416 1133 1439 2 Gross profits & other trading inconme 671 535 672 3 Taxes on expenditure (net) 469 362 408 4 Wages & salaries (inel. Forces' pay) 1289 1289 909 1354 1348 1360 1306 1271 1257 1166 1361 664 1133 1236 1539 5 Emiployers' contributions 128 128 90 134 134 135 129 126 125 116 135 66 112 122 152 6 Rent, dividends and intereSa 892 892 623 946 938 944 905 879 869 802 926 453 776 831 1036 7 louseholds: E 0-10 incoIIIe 73 73 50 75 76 77 74 72 71 65 74 36 62 66 82 8 " " LIO-20 233 233 163 244 244 246 2:a 230 223 210 242 118 203 217 270 9 I " E20-30 472 472 332 496 494 498 478 465 460 426 495 242 413 447 557 10 " E30-40 479 479 337 504 502 506 486 473 468 433 504 246 420 456 563 11 " " E40-50 325 325 229 342 340 343 330 321 317 294 342 167 285 309 385 12 " LSO-60 201 201 142 212 211 213 204 199 196 182 212 103 177 192 239 13 L60 + 374 374 263 395 392 395 379 369 364 337 393 192 328 355 442 14 Private non-profit institution1s 64 64 44 6/ 66 67 64 63 62 57 66 32 55 59 73 15 Companies 420. 420 293 445 442 444 426 414 409 377 436 213 365 391 483 16 PtLiblic corporations 34 34 23 36 35 36 34 33 33 30 35 17 29 31 39 17 Central government 874 874 600 897 907 923 888 863 857 785 886 434 747 783 971 18 Local authorities 275 275 187 279 284 291 280 272 271 246 274 134 232 239 296 19 Food 20 Drink and tobacco 21 llousing 22 Fuel and liglht 23 Clothing and footwear 24 DUrable & otlier household goods 25 Other imiscellaneous goods - 26 Motor vehiciles (ici.l ruinning costs) 27 Travel and coIalmunications 28 EnterrainmlSenItS, IloLelS & restaurants 29 Other services 30 Military defence 31 National heIaltI service 32 OLher central governmient Purposes 33 Local authorities puirposes 34 Agrictlture etc, , nlining & quarrying 35 Metals and nletal products 36 Oher manufacturing 37 Construction 38 Gas, electricity and Water 39 Services _ _ _. . _ _ _ __ __ _~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 19 120 21 22 .23 24 25 1 27 [*6 28 29 0 31 32 35 36 39. _ _ _ __8 I Income from employment 2 Gross profits & other trading income 3 Tax:eson expenditure (net) 4 Wages & salaries (innl. Forces' pay) 5 Employers' contributions 6 Rent, dividends and interest 7 Households: income L 0-10 8 " " E 10-20 9 to i L20-30 10 is If 30-40 1I of " 40-50 12 ' l50-60 13 , "e60 + 14 Private non-profit institutions 15 Companies 16 Public corporations 17 Central government 18 L.ocal authorities 332 387 419 406 350 366 361 352 427 376 392 368 424 496 448 401 387 345 419 410 434 19 Food 196 224 246 239 206 215 212 205 252 221 231 217 250 292 264 237 228 203 247 241 256 20 Drink and tobacco 193 227 244 237 204 213 210 206 248 219 228 214 247 289 261 234 225 201 244 239 252 21 |holising 79 93 100 96 83 87 86 84 101 89 93 87 100 118 106 95 91 82 99 97 103 22 Fuel and light 134 153 168 163 140 147 145 140 172 151 157 148 170 199 180 162 156 138 168 165 174 23 Clothing and footwear 120 139 151 147 127 132 131 126 155 136 142 133 154 179 162 146 140 125 152 149 157 24 Durable & other household goods 105 121 132 128 110 115 114 110 135 118 123 116 134 156 141 127 122 109 132 129 137 25 Other miscellaneous goods 134 151 167 163 140 146 144 139 172 151 157 148 170 199 180 162 156 139 168 165 175 26 l2otorvehicles (incl. running costs) 65 75 82 80 69 72 71 69 84 74 77 72 83 97 88 79 76 68 82 81 85 27 Travel and coiiununications 102 1!7 128 125 107 112 110 107 131 115 120 112 130 151 137 123 118 105 128 126 133 28 Entertainmuents,hotels & restaurants 121 142 152 149 127 133 132 128 155 137 143 134 154 180 163i 147 140 125| 152 151| 158, 29 Other services I _I I I I I I § - 1 1 138 231 194 170 153 165 163 190 177 158 169 153 180 213 193 157 157 142 175 17,4 179 30 Military defence 80 134 113 99 89 96 95 111 102 92 98 89 105 124 112 91 91 83 101 101 104 31 National health service 58 97 82 72 65 69 69 80 74 67 71 64 76 90 81 66 66 60 73 73 75 32 OLher central government purposes 175 346 260 216 200 218 216 274 222 201 218 193 231 275 249 190 195 179 221 221 225 33 Local autlhoritiespurposes 135 166 172 165 143 150 148 148 173 153 160 149 172 202 183 161 156 140 170 167 176 34 Agriculture etc., mining & quarrying 192 266 254 236 207 219 217 230 246 219 230 213 248 291 263 226 222 199 243 239 250 35 Metals and metal products 532 662 682 651 565 593 586 589 683 603 631 589 681 798 722 637 617 551 671 659 693 36 Other manufacturing 62 84 82 77 67 71 70 73 80 71 75 69 80 94 85 74 72 65 79 78 81 37 Construction 102 130 132 125 109 114 113 115 131 116 121 113 131 153 139 122 |11| 106 129 127 133 38 Gas, electricity and water 1106 1502 1453 1357 1188 1255 1241 1304 1417 1256 1322 1223 1422 1671 1512 1306 1275 1143 1394 1376 1439 39 Services Table 8.9. Additional Effects Associated with M3: 1000 (M3 - I) M2 Mj, United Kingdom, 1968 1 ._ '-2---3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 1 Income from employm.ent 1417 1417 1000 1488 1482 1495 1435 1397 1382 1282 1496 730 1245 1358 1691 2 Cross profits & other trading income 671 671 471 721 709 711 680 661 652 603 702 343 587 633 791 3 Taxes on expenditure (net) 469 449 312 463 480 496 478 464 464 419 453 223 388 387 476 4 Wages & salaries (incl.Forces' pay) 2199 1030 1309 5 Employers'contributions 218 102 130 6 Rent, dividends and interest 968 1721 1105 7 income N 0-10 liouseholds: 90 90 125 8 " " L1O-20 318 295 312 9 " L.20-30 727 50i 531 10 " " k30-40 748 504 522 11 :" 40-50 507 344 350 12 " L' 50-60 313 218 214 13 " " L60 + 535 498 401 14 Private non-profit institutions 84 86 86 15 Companies 456 810 520 16 Public corporations 36 65 42 17 Central governawnt 1091 954 1654 18 Local authorities 309 294 624 . _ _ 19 Food 508 384 396 508 509 328 619 585 564 518 488 475 417 389 231 407 388 424 20 Drink and tobacco 301 224 227 301 300 190 282 310 325 304 296 300 260 234 133 235 220 249 21 Housing 294 226 234 294 295 195 390 345 320 297 286 274 249 334 138 242 230 248 22 Fuel and light 119 92 97 119 120 80 186 150 132 120 114 110 98 118 57 99 96 101 23 Clothing and footwear 204 154 155 205 204 130 196 206 215 210 204 199 183 197 91 161 150 170 24 Durable & otlher householdgoods 184 139 141 184 184 118 187 191 194 188 185 183 161 182 83 146 137 153 25 Odliermiscellaneousgoods 160 121 123 160 160 103 165 166 171 163 158 156 142 165 72 128 119 133 26 Motor vehicles (tncl. running costs) 206 153 151 206 204 128 162 182 20'. 218 208 212 191 194 90 158 146 169 27 Travel atidcomiuinications 99 76 77 99 100 65 102 106 105 101 98 97 90 109 45 80 75 83 28 Entertaintments, hotel & restaurants 154 119 119 154 154 102 151 153 159 155 154 154 149 234 72 126 116 130 29 Othterservices 181 147 146 181 182 127 197 187 182 177 177 175 193 448 90 158 143 156 30 Military defcnce 195 171 296 195 195 159 164 183 1.99 197 197 198 189 162 132 202 332 185 31 National health service 113 99 172 113 113 92 95 106 116 114 114 115 110 94 76 117 192 108 32 Other central governmentpurposes 82 72 124 82 82 67 69 77 84 83 83 83 80 68 55 85 139 78 33 Local authorities purposes 235 223 473 235 234 219 214 23C 243 238 234 235 219 211 160 273 304 1000 34 Agriculttireetc., wini..g& quarrying 203 158 179 203 204 137 240 227 221 207 198 194 174 185 98 170 171 202 35 Metals and metal products 284 229 308 284 284 203 276 285 296 239 283 282 260 270 153 255 312 295 36 Other manufacturing 803 622 714 803 803 539 859 854 859 817 789 777 703 749 386 670 682 811 37 Construction 93 74 95 93 93 66 109 102 99 94 91 89 82 99 48 82 91 107 38 Gas, electricity and water 153 121 144 153 153 106 203 178 165 154 148 144 131 152 76 132 137 165 39 Services 1636 1320 1718 1636 1636 1170 1689 1696 1718 1655 1616 1601 1499 1821 856 1458 1598 2094 19 20 21 22 23 24 25 2 3 1 32 33 34 35 36 137 38 39 1403 1513 1750 1634 1473 1539 1513 1421 1829 1598 1666 1582 1801 2107 1899 1657 1690 1460 1801 1621 1858 1 Income from employment 684 708 811 963 690 i12 712 650 843 750 768 712 841 972 887 929 723 706 807 lOi3 857 2 Gross profits & other trading income 347 940 592 412 427 481 477 705 441 409 458 389 476 576 523 327 386 351 448 421 447 3 Taxes on expenditure (net) 1277 1377 1592 1487 1340 1401 1376 1293 1664 1454 1516 1439 1639 1917 1728 1508 1537 1329 1639 1475 1690 4 Wages & salaries (incl. Forces' pay) 126 136 158 147 133 139 136 128 165 144 150 143 162 190 171 149 152 132 162 146 167 5 Employers' contributions 877 1043 1086 1191 912 946 943 924 1092 97211007 928 1095 1273 1159 1139 947 905 1054 1243 1109 6 Rent, dividends and interest 66 102 90 83 72 77 76 85 84 75 80 73 85 !01 91 78 75 68 83 85 8C 7 __I- i^.co. I 223 284 287 279 237 249 246 251 285 252 264 245 285 334 302 270 255 231 279 284 289 8 " E10-20 463 528 581 560 487 509 502 483 597 524 547 515 591 692 625 557 543 481 586 564 606 9 " " L20-30 473 528 591 571 496 518 511 487 610 535 558 526 603 705 637 570 556 491 598 574 620 10 " " L30-40 321 356 400 388 337 351 346 329 414 363 379 357 410 479 432 388 377 333 406 390 421 1 " " E40-50 199 219 248 242 209 218 215 203 257 225 235 221 254 297 268 241 234 207 252 243 261 12 " " L50-60 372 409 459 466 387 402 398 376 474 418 433 -407 470 548 496 461 426 385 463 474 482 13 | " L60 + 61 78 78 77 65 68 67 68 77 69 72 66 77 91 82 74 69 63 76 79 78 14 Private non-profit institutions 413 491 511 561 429 445 444 435 5i4 458 474 437 515 599 546 536 446 426 496 585 522 15 Conmpanies 33 39 41 45 34 36 35 35 41 37 38 35 41 48 44 43 36 34 40 47 42 16 Public corporations 773 1293 1085 951 858 921 912 1064 986 884 945 854 1008 1192 1080 877 878 795 976. 970 1001 17 Central government 230 456 343 284 263 287 284 361 292 265 287 254 304 362 328 251 258 235 291 291 297 18 Local authorities 19 Food 20 Drink and tobacco 21 Hiousing 22 Fuel and light 23 Clothing and footwear 24 Durable and oLier houschold goods 25 Other miscellaneous goods 26 Motor vehicles (incl. running costs) 27 Travel and comimunications 28 Hntertainmients, hiotels& restaurants 29 Ottherservices 30 Military defence 31 National health service 32 Other central government purposes 33 Local authorities purposes 34 Agriculture etc.,mining & quarrying 35 Metals anidmectalproducts 36 Ot:hermanufacturing 37 Construction 38 Gas, electricity and water 39 Services 178 and SAM-Based Multipliers Models households and 417 if it is applied to the richest ones. Apart from the immediate responses of the households concerned, the remaining extragroup effects are not very different: 619 - 260 = 359 for the poorest households and 417 - 91 = 326 for the richest ones. This pattern persists in other categories of expenditure. Although the rich spend initially on all goods and services only 573 out of their 1,000 compared with 923 in the case of the poor, the remasining effects are only about 9 percent less: 1,560 compared with 1,714. While I believe that the tables described are useful in gaining some insight into economic interdependence, their limitations will be apparent. In the first place, all the relationships are linear, indeed homogeneous. In the second place, they are static and take no account of the time needed for effects to work themselves out. In the third place, I have chosen, as is often done, to treat as exogenous parts of the system which are not altogether exogenous: in some measure, current spending is likely to induce both new investment and exports. Conversely, in the fourth place, I have treated some accounts as endogenous which for some purposes should be treated as exogenous: for instance, in Britain in the past decade there have been immense variations in central government saving. Finally, to conclude a list which is certainly incomplete, I have not distinguished between quantity changes and price changes: when I say that food expenditure will rise by so much, I do not discuss whether this means more food at the same prices or the same amount of food at higher prices. There is nothing new in all this. There is a limit to what we can hope to get out of even sophisticated multiplier analysis, and if we want to deal with the issues just raised we must build more complicated models. WHERE DO WE GO FROM HERE? There are plenty of examples of such models but in this section it is not my intention to discuss the problems to which they give rise. Rather I propose to discuss briefly some analytical extensions which ought to be considered in modeling the household sector. Models of Income Generation The multiplier analysis of the preceding sections shows the circular flow of payments from consumption to production and back again, and in that way throws some light on the generation of incomes. But my example, at least in the form presented, relies heavily on fixed coefficients: the proportions in which each group of households spends its income are fixed and so are the cost proportions in each of the industry groups; each type of income is collected together and distributed to each of the sectors, and within the household sector to each of the types of household, in fixed proportions; a change in exogenous demand will change output levels and therefore the distribution of the components of value added, but each component will be allocated to household groups in fixed proportions. Thus, whatever happens, the distribution of incomes is not likely to change much. The position would be different if, for instance, income from employment arising in each industry could be divided into earnings groups and if the income in each group could be separately distributed over households. With this information, a shift in demand toward products requiring highly paid labor would tend to be reflected in a relative increase in the incomes of the richer families. Nothing, however, will help us to connect the income distribution of one period with that of the next. For this purpose we need information on the change in individual incomes from one year to the next. Thus, of the households in a given income category at the end of one year, some will remain in that category throughout the next and others will experience a rise or a fall of income within the year and will be found in a different category at the end of it. These may be called surviving households; yet others will disappear as a consequence of emigration or of dissolution for one reason or another. At the same time, new households will make their of the HouseholdSector Disaggregation 179 Table 8.10. A Demographic Matrix Connecting the Opening and Closing Stocks of Year 0 with the Flows during Year 0 State at new Entrants year e in Our country: Closing State at year 0 opening states stocks new year e+1 Leavers in year 0 a d' Our country: closing states b S An Opening stocks n' appearance as a conseqaence of immigration or of formation from individuals in the country concerned. This kind of information can be set out in a version of the standard demographic matrix described in Towards a System of Social and Demographic Statistics (UNSO, 1975), as shown in table 8.10. Let us begin with the row vector, n', the elements of which represent the number of households in each income group at the beginning of year 0. Those which emigrate or dissolve appear in the row vector d' and those which survive appear in the columns of the matrix S. Looking at the central row of the matrix, the elements of the closing stock vector, An, are made up of the entrants in year 0 by immigration or formation, the elements of b, and the survivors from the opening stock recorded in the rows of S. The scalar, a, represents those households that both appeared and. disappeared in the year and so, while present for a time, were not recorded in either the opening or the closing stock. As applied to individuals this scheme is quite straightforward since there is only one thing that can move from stato to state, namely, the individual himself. But in the case of households, there are typically several things that can move, namely, each of the household members. We must decide what constitutes a household and by what it is replaced when it dissolves. This subject is discussed in tUNSO (1975, pp. 64-66). If from the demographic matrix we calculate a coefficient matrix C, say, as C = S A-' that is to say, by dividing the elements in the columns of S by the corresponding element of n, then we can see that (8.7) n = Si + b = Cn + b where i denotes the unit. vector, so that Si denotes the row sums of S. This and similar models are discussed at some length in UNSO (1975, pp. 42-47). In suitable circumstances the model set out in (8.7) can be interpreted as a Markov process and the matrix multiplier (I - C) as the fundamental matrix of an absorbing Markov chain. 180 Multipliers and SAM-Based Models A difficulty which must somehow be dealt with in this kind of model arises from the fact that incomes tend to increase in money terms, so that in a few years the entries in S will move out of the matrix into an area for which no transition proportions are available. This may force us to concentrate on relative movements around a stationary mean income which can then be related to the actual mean income in successive years. The treatment of income generation as a Markov process has been developed in Champer- nowne (1953; 1969, vol. 2, ch. 18; 1973), and an early application is given in Vandome (1958). A study of the income mobility of male employees, in Shorrocks (1976), shows that this mobility is not governed by a first-order Markov process but can be represented by a second-order process. Though much work remains to be done, these models seem to offer a convenient starting point for the study of income mobility. Households as a Branch of Production In the SNA the activities of general government are provided with production accounts, although these accounts are of a somewhat formal kind because the value of output is equated to the cost of the inputs so that, whatever happens, no gain or loss can arise. Households, in contrast, are not provided with production accounts, although provisions are made to ensure that the services of owner-occupied dwellings and of domestic servants are included within the production boundary; they are provided only with income and outlay and capital transactions accounts, as in this chapter. Production accounts for households might be introduced for specific activities with the aim of eventually covering all aspects of household management. Suppose we began with an account for the services of dwellings to owner-occupants. At least in countries with a free market in rented housing it should be possible to impute a revenue for the services of owner-occupied houses, against which the relevant expenses could be set off to give a gain or loss on owner- occupation. Where rent control existed, this method would not be applicable to the categories of property affected, and we should have to try by the use of programming methods to generate the kind of information provided by markets. Alternatively, we could fall back on the formal approach, usually adopted in respect of general government services, in which the provider of the services (in this case the owner-occupant) is assumed either to break even or to earn some conventional rate of return on his investment. This kind of accounting may serve to tidy up the limited amount of data we possess, but if presented without due qualifications it can be misleading, since no amount of mismanagement can ever be reflected in a loss. The next step might be to set up accounts for the services of consumer durables such as cars, major household appliances, and so on. Some progress could certainly be made, and the main difficulty would be to obtain an estimate of the value of the services of these goods which was not built up from costs. But even if the exercise were not a complete success it would provide a measure, in terms of depreciation, of the consumption of durables and also a measure of the stock. Beyond this point we get into still deeper waters if we try to set up an account for household activities involving the unpaid services of housewives and other members of the household. We may be able to get a clear idea of how people spend their time from time budgets, but there are problems of valuing the unpaid inputs as well as the outputs. For the time being this last step seems more a matter for research and data collection than something whose introduction into the national accounts should be given any priority. Consumer Behavior Much work has been done in recent years on consumers' spending and saving behavior. The familiar linear expenditure system is highly convenient, but its main limitation, as shown in of the Household Disaggregation Sector 181 Deaton (1974), is that it erLtails an implausible connection between own-price and total expen- diture elasticities. Much the same can be said of all demand systems based on additive prefer- ences. It is important, therefore, to find an otherwise acceptable demand system not based on this assumption. The latest contribution I know of is in Deaton and Muellbauer (1980). Household Balance Sheets In the early days of social accounting there was a tendency to concentrate on the measurement of flows as opposed to stocks, to construct accounts as opposed to balance sheets. This emphasis is now changing. National and sector balance sheets and revaluation accounts which cover the real and financial assets and the liabilities of institutional sectors form an integral part of the SNA, and provisional international guidelines on the subject have recently been published in UNSO (1977a). In Britain, detailed estimates have been made by Revell and others (1967), Revell and Roe in UKCSO(1953-: no.211) and Roe in Cambridge University, Department of Applied Economics (1962-74; 1975-). This work has been brought up to date, with revisions, by the Central Statistical Office, and provi.sional estimates for the personal sector, in which households and PNPIs are shown separately, have appeared in UKCSO(1953-: no. 291). These estimates cover end-1975 and end-1976 in considerable detail and, for the household sector only, the period 1966-74. The estimates, which include consumer durables along with other physical assets, should be extremely useful in studying the spending, saving, and financial behavior of the household sector. CONCLUSIONS It seems to me that of all the interesting and useful things that could be done to improve the national accounts, the one most worthy of consideration is the disaggregation of the household sector. I believe that my statistical exercise, embodied in table 8.1, which is nothing more than a demonstration of possibilities, could be carried out effectively in many countries; and the work of Pyatt and his associates suggests that developing countries can be numbered among these. As things are, we already have a disaggregation of the productive system in input-output tables and, for a more restricted number of countries, a disaggregation of the financial system in flow-of-funds tables. The missing piece is the disaggregation of income and outlay. In the present climate of opinion it seems likely that efficiency suffers if explicit regard is not paid to equity. The disaggregation of the household accounts would constitute a recognition of this relationship between equity and efficiency and would open the way to studies that cannot at present be carried out. APPENDIX: MULTIPLIERS FOR QUESNAY'S TABLEAU Quesnay (1758) did not produce any multipliers, indeed he did not produce a recognizable transactions table. But later writers, particularly Phillips and Barna, have done better. In a paper entitled "The Tableau Economique as a Simple Leontief System" (1955), Phillips translated the Tableau into a social account matrix of order 3. Continuing on these lines, Barna (1975), in "Quesnay's Tableau in Miodern Guise," produced a matrix of order 9. In this latest version, two accounts, for capital transactions and the rest of the world, can be treated as exogenous. The seven endogenous accounts are divided into two groups: production accounts for two activ- ities-agriculture and everything else; and income and outlay accounts for five sectors-landlords, farmers, artisans, the state, and the church. Table 8A1 is based on Barna's table 1 with an alteration in the order of the accounts and a change in entries ending in 3 to end in 2.5. 00 Table 8.AI. A Social Accounting Matrix for France, about 1750 1 2 3 4 5 6 7 8 9 1 Agriculture 525.0 525.0 300.0 525.0 262.5 150.0 75.0 525.0 262.5 2 All other activities 300.0 525.0 262.5 150.0 75.0 3 Landlords 1050.0 4 Farmers 1050.0 5 Artisans 525.0 6 State 300.0 7 Church 150.0 8 Capital transactions 525.0 9 Rest of the world 262.5 Total 3150.0 1312.5 1050.0 1050.0 525.0 300.0 150.0 525.0 262.5 Table 8.A2. Coefficient Matrix, 10000 A', Derived from Table 8.A1 for France, about 1750 2 3 4 5 6 7 8 9 1 Agriculture 1667 4000 2857 5000 5000 5000 5000 10000 10000 2 All other activities 2857 5000 5000 5000 5000 3 Landlords 3333 4 Farmers 3333 5 Artisans 4000 6 State 2857 7 Church 1429 8 Capital transactions 1667 9 Rest of the world 2000 Table 8.A3. Matrix Multiplier for Intragroup Effects, 1000 Ml, France, about 1750 1 2 3 4 5 6 7 1 Agriculture 1200 480 2 All other activities 1000 3 Landlords 1000 4 Farmers 1000 5 Artisans 1000 6 State 286 1000 7 Church 143 1000 Table 8.A4. Matrix Multiplier for Intergroup Effects, 10000 M2 , France, about 1750 __ _ _ _1 2 3 4 5 6 7 1 Agriculture 3332 1399 2 All other activities 1388 1833 3 Landlords 1666 1166 1166 1166 1166 4 Farmers 666 2166 1166 1166 1166 5 Artisans 476 833 1833 833 833 6 State 190 333 333 1333 333 7 Church 95 167 167 167 1167 Table 8.A5. Matrix Multiplier for Intragroup and Intergroup Effects, 10000 M2Ml, France, about 1750 ________ ~~ ~~1 2 3 4 5 6 7 1 Agriculture 3999 2999 2 All other activities 1666 2499 3 Landlords 2166 1166 1166 1166 1166 4 Farmers 1166 2166 1166 1166 1166 5 Artisans 833 833 1833 833 833 6 State 619 333 333 1333 333 7 Church 310 167 167 167 1167 Table 8.A6. Matrix Multiplier for Extragroup Effects, 1000 M 3, France, about 1750 ____ 11 2 3 4 5 6 7 1 Agriculture 1000 480 840 840 840 840 2 All other activities 1000 286 500 500 500 500 3 Landlords 333 | 1000 4 Farmers 333 1000| 5 Artisans 400 1000 6 State 95 | 1000 7 |Church 48 - - - 1000 183 184 Models and SAM-Based Multipliers Table 8.A7. Matrix Multiplier for All Effects, 1000 M3M 2Ml = 1000 M = 1000 (I - A)-', France, about 1750 __ _ _ _ _ _ _ _ _ 1 2 3 4 5 6 7 1 Agriculture 4000 3000 3500 3500 3500 3500 3500 2. All other activities 1666 2500 2083 2083 2083 2083 2083 3 Landlords 1333 1000 2166 1166 1166 1166 1166 4 Farmers 1333 1000 1166 2166 1166 1166 1166 5 Artisans 667 1000 833 833 1833 833 833 6 State 381 286 619 333 333 1333 333 7 Church 190 143 310 167 167 167 1167 Table 8.A8. Additional Effects Associated with M2 : 1000 (M2 -I) Ml, France, about 1750 I 1 2 3 4 5 6 7 1 Agriculture 2799 2519 2 All other activities 1666 1499 3 Landlords 1166 1166 1166 1166 1166 4 Farmers 1166 1166 1166 1166 1166 5 Artisans 833 833 833 833 833 6 State 333 333 333 333 333 7 Church 167 167 167 167 167 Table 8.A9. Additional Effects Associated with M3: 1000 (M3 - I) M2M1 , France, about 1750 __ __ 11 2 3 4 5 6 7 1 Agriculture 3500 3500 3500 3500 3500 2 All other activities 2083 2083 2083 2083 2083 3 Landlords 1333 1000 4 Farmers 1333 1000 5 Artisans 667 1000 6 State 381 286 7 Church 190 143 The Algebra of the Multipliers In terms of the notation of my paper the present case can be set out as follows. The matrix of the endogenous entries, A, takes the form A- [All A 12 1 LA21 A22 [ 0 A0.] + [o A..] =B + C Disaggregation of the Household Sector 185 and the equation y = Ay + x = [I + (I -- B)-'C][I - (I - B)-'C(I - B)-'C]-'(I -B)-x = M3M 2 M,x. In this system, first, 1 [(I - A,)- (I -A 22 ) Second, M2 = [D 0] where D = [I - (I -All)-' A12 (I -A22)-' A21] and E = I - (I -A22) -1 A2 1(I All) A12. And, third, 3 I (I All) 'A12 33 | (I -A22)-:LA21 I The coefficient matrix, AO,for the whole system is set out in table 8.A2, and Ml, M2 , M2 MI, Ml, M3M2 Ml = M = (I - -A)-I, (M 2 - I)M,, and (M, - I)M 2 Ml are set out in tables 8.A3 through 8.A9. Conclusions From table 8.A7 we can see what ought to be done if we wish to help one of the accounts. The best thing, as we might expect, is to provide the initial stimulus to the account. Failing this it is better, in the case Dfactivities, to stimulate one of the sectors, and it does not matter which one, rather than the other activity. If we want to help landlords or farmers, it is best to stimulate agriculture and worst to stimulate other activities; and the opposite is true if we want to help artisans. If we want to help the state or the church, it is best to stimulate landlords or, failing them, agriculture or, failing that, any one of the other sectors. Stimulating other activities will do least. Short of providing an initial stimulus to artisans, the next best method of helping them is to stimulate other activities. This will do less for landlords, farmers, the state, and the church than a stimulus of the same size applied to any of the other accounts. 9 Accounting and Fixed-PriceMultipliersin a Social Accounting Matrix Framework Graham Pyatt and Jeffery I. Round This chapter is concerned with the relationships between output, factor demands, and income and the decomposition of these relationships into separate effects, as suggested by the repre- sentation of the flows between them within the structure of a social accounting matrix. Since output, factors, and the nongovernment institutions sector (households and companies) are all disaggregated in the system to be examined, it follows that the analysis is concerned not only with output levels and the level of factor and household incomes, but also with the structure of production, the distribution of factor incomes, and the distribution of disposable income both among households and between them and the corporate sector. This is the first sense in which this chapter is concerned with decomposition, and it makes the point that the distribution of income and the structure of production are inextricably interwoven. The closed-loop character of the present formulation implies that the incomes of production activities, factors, and institutions are all derived from injections into the economy that result from a multiplier process. This multiplier is a matrix, M, which can be expressed as the product of three multiplier matrices, Ml, M2 , and M.. The first of these captures the effects of transfers within the economy, such as the distribution of profits from companies to households, and the transfers of goods between activities which is the essence of input-output. The other matrices, M. and M,, capture the consequences of the circular flow of income within the economy. Matrix M3 shows the fuli circular effects of an income injection going round the system and back to its point of origin in a series of repeated and dampening cycles. In contrast, M2 captures the cross-effects of the multiplier process whereby an injection into one part of the system has repercussions on other parts. These cross-effects correspond to open-loop effects and hence to the class of models, such as that of Maton, Paukert, and Skolka (1978), which trace the effects of some exogenous changes in income distribution on output and employment, with no allow- ance for the effects in the reverse direction of changes in output and employment on the distribution of income. The decomposition of M into component parts is the second sense in which we are concerned with decomposition. The first perspective on decomposition is illustrated in the next section by a simplified SAM for Sri Lanka in 1970, which shows balanced accounts for factors, production activities, house- holds, and companies set in the broader framework of a full national accounting system. The following section explores the structure of these accounting balances in terms of a multiplier matrix and its decomposition into transfer, open-loop, and closed-loop effects.' An additive version of the decomposition developed by Stone in the previous chapter is also presented. Note: This paper has been published with minor modifications in the Economic Journal (Pyatt and Round, 1979). We are particularly grateful to Charles Blitzer and Sherman Robinson for comments on an early draft and to Sir Richard Stone for his general support of the line of work reported in this paper. Particular contributions from him are acknowl- edged in the text. We also wish to thank Kenshi Ohashi for computational assistance. 1. This aspect has been treated previously by us in Pyatt, Roe, and associates (1977), ch. 4. However, there is an error in the exposition with respect to the treatment of indirect taxes, which is removed in the present paper. The multiplier decomposition has also been applied in Bell, Hazell, and Slade (1982) and by Stone in the previous chapter. 186 and Fixed-Price Accounting Multipliers 187 The multipliers discussed in this chapter are referred to as accounting multipliers. Their data base is the SAMobserved for 1970, and their role is simply to represent the accounting balances of the SAMin a novel way w:hich gives some insights into economic structure. With the account- ing multipliers as a starting point, it is then possible to consider the potentially more interesting case of multipliers due to income effects in a fixed price model. The argument presented later in this chapter shows that these fixed price multipliers are strictly analogous to the accounting multipliers. The only difference arises from extensive use of marginal expenditure propensities in the fixed price case, while the accounting multipliers are built up from the average expen- diture propensities which can be calculated directly from the SAM. Thus the fixed price multi- pliers can be interpreted as having a data base which is the initial SAMnow complemented by estimates of income elasticities when the latter differ from unity. Our pedagogic procedure of presenting accounting multipliers first, and then the fixed price multipliers, makes it possible to bring out the implications of income elasticity effects, such as Engel's law, within a fixed price system. Indeed, following the decomposition of the fixed price multiplier matrix, we are able to show that the difference between this matrix and the accounting multiplier matrix can itself be represented as a mnultiplicative matrix effect that is dependent on income elasticities which differ from unity. The empirical results presented later illustrate the various components of fixed price multi- pliers and alternative ways of deriving them. The results show that the estates sector in Sri Lanka is relatively isolated. within the economy because its linkages with other sectors are slight. The results also sho-w the extent to which input-output calculations underestimate the linkages between producing sectors in comparison with the case where the full circular flow of incomes is taken into account. More generally, the anatomy and interdependence of income and production structures in the economy are captured by the various multiplier matrices discussed. The inclusion of different; types of households in the present formulation distinguishes the approach from standard closed-loop Leontief systems and allows the distribution of income to be brought into the picture. Because it includes factors as well as households, the present formulation extends the structure of accounting balances as set out by Quesnay (1758) and the previous closed-loop multiplier formulations which have been developed within his account- ing framework. 2 THE SOCIALJACCOUNTING MATRIX The SAM in table 9.1 provides the numerical base for subsequent illustrations. In reading this table it is important tc keep in mind the convention that entries are to be read as receipts for the row account in which they are located, and expenditures or outlays for their column accoun.t. The SAM is square because each account has both receipts and expenditures; and the row and column sums for a given account must be equal because all income must be accounted for by an outlay of one type or another. Eight groups of accounts are shown, some of which are further disaggregated. The partitioning of the eight groups into endogenous and exogenous accounts is discussed following an explanation of the flows depicted in table 9.1. Factors of production receive income from domestic production (shown as the intersection of accounts in row block ]. with column block 4) which in turn is distributed to households 2. The distinguishing feature ol Quesnay's Tableau Economique from the present perspective is that value added in different production activities is paid directly to households of various types as opposed to being routed to them through a set of factor accounts. This simplified approach is also adopted in Desai (1961) and in a model of Iran (Pyatt and others, 1972) which is of the fixed-price multiplier type. In an appendix to the previous chapter Stone has applied the analysis of accounting multipliers in this paper to Quesnay's Tableau. Table 9.1. A Social Accounting Matrix for Sri Lanka, 1970 x0 (in millions of rupees) rJTod.mva"4 Accounts Exogamous Acc"urt &'&.ogo at Production j1ous.1hoj4 Corpoarst Production ActiviislI _ Rest _ _ _~ ~ Labour capital Accounts _Accounts __~E Ce-odity _ _ GrouJps _7_ _ S__ r 1 oF1l Wfl 01 I- a A~~~~- 16 154 31 - 7 _ 122S 1673 u 1 c 13) 1171 54 243 123 1460 3184 x Estate 6is~~~~~~~~~~~~~~~~~~1 34 10 2 4 46 III l r _ btzt' 143 1136 141 586 924 2201 5771 _ idb 3 77 22 72 __1 _, us "TI _ 2 Ooursheul 1673 ttrbis 199 434 92 3004 2 XccountS hT t 3184 3356 204 153 6903 tEstate 111 63 1 7S1 1402 T _ C0_C p"tZ _3_1S402 AccouneS - CSte ITI7 294 4S1 Tas band Rubb 14 56 6 8 2 2 -30 lieu 1238 oth iFUerdculturz 404 1199 16 12 166 1294 66 7 45 20 78 108 3569 4 sh oodear F 275 1065 138 t1 34 188 40 26 90 152 2019 341 904 112 121 113 43 490 248 2310 66 is I64 28?7 Himing n end2 aenst 6 S 236 SS 92 1595 19 201. Services 9E,P815 1933 208 74 56 120 251 234 230 131 154 490 5996 S Gover=naC Curnt AcCOmnt 332 119 236 352 i5 1388 44 2346 : lnt±tutlng CapItal 520 B8O 11 505 329 41 382 259, *7 ndiredct Tazm (nC) 34 72 3 37 35 16 416 80 205 29 270 131 31369 * S Qf ,ransocto 207 741 143 87 92 199 368 77 202 43 364 -279 2244 _ eTorLa Orho; 113 16 36 365 Tor.1 1613 314 711 51731 387 3004 6903 791 11402 481 1238 3568 2019 2887 2014 5996 2346 2596 1388 2244 165 Source: Adapted from Pyatt, Roe, and associates (1977). and Fixed-Price Accounting Multipliers 189 and companies (rows 2, 3 intersecting with column 1) and as net factor income payments abroad (row 8, intersecting with column 1). Factor incomes received by households include wages, unincorporated business profits, and rent on dwellings (row 2, column 1), but households also receive distributed profits from the corporate sector (2,3) and transfers from government (2,8). Similarly, corporate enterprises receive factor incomes in the form of gross profits (3,1), as well as current transfers from government (3,5). Government income is derived from direct tax payments and other transfers by households (5,2) and corporate enterprises (5,3) and from the rest of the world (5,8), as well as intragovernment transfers (5,5), together with net indirect tax payments (5,7), shown as a receipt from a special indirect tax account. The expen- ditures on domestically produced commodities are shown in the row of account 4. They include outlays by households (4,2), government (4,5), investment (4,6), and the rest of the world (4,8), as well as intermediate transactions between production activities (4,4). Indirect taxes on all of these expenditures, and purchases of imported goods, are shown as separate outlays by the various spending units. They are received in row 7 by the account for (net) indirect taxes and in row 8 by the rest of the world revenue account. Finally, outlays on domestic investment (column 6) are matched by domestic and foreign savings (row 6) where the latter (5,8) is the final balancing item in the rest of the world accounts. An important feature to note is that factors, institutions, and activities are all disaggregated in table 9.1, so that the SAM captures the distribution of factor incomes as well as their level. It also shows the distribution of income among household types. To move from a SAM to a model structure requires that each account should be designated as endogenous or exogenous. The accounts in table 9.1 have been ordered so that the endogenous accounts occupy the leading rows and columns of the SAM. This is shown schematically in table 9.2. The notation to be used in subsequent discussion is given with this table, as are a number of accounting relationships, equations (9.1) to (9.11), which follow directly from the SAM structure. Equation (9.1) states that transactions between endogenous accounts, denoted by matrix N, can be expressed as the product of a square matrix, An, of average propensities to consume and a vector of endogenous incomes, yn. Similarly (9.2) equates leakages, L, with the product of a non-square matrix, At, of average propensities to leak and the endogenous incomes, yn. It is important to note that since N, L, and yn are observed in a SAM such as table 9.1, the matrices An and A,, can be obtained directly. Equations (9.3) and (9.4) express the accounting relationship by which endogenous incomes are determined. Equations (9.5) and (9.6) have the same role with respect to incomes of the exogenous accounts, y,. Equation (9.7) sums expen- ditures (columns) of the endogenous accounts. It implies that, for these accounts, row and column sums will be equa], provided equation (9.8) holds, that is, provided column sums of A,, plus those of Ae, add to unity in all cases. Equation (9.9) expresses column sums for exogenous accounts. The requirement that these be equal to row sums (equation [9.6]) yields equation (9.10). Finally, an implication of (9.10) is obtained in (9.11), which states that, in aggregate, injections into the system must equal leakages. From equation (9.4) and the definition of t it follows that (9.12) yn = (I -A.)-lx = M.x and (9.13) t = Ae (I -An) -x = Ae Ma x provided that (I - An) - exists. This inverse is the accounting multiplier matrix M, which relates endogenous incornes, yn, to injections, x. Numerical values for M, and AeM, are given in table 9.3. The existence and decomposition of M, are discussed in the next section. It can be Table 9.2. Notation and Accounting Balances for Equations (1) through (11) Expendi tures Total Endogenous Accounts Exogenous Accounts a w yn n + x (9.3) o N Ay (9.1) X fA a nynA + x (9.4) tJ Z a 0 *C.)o 0 y - Q+ Ri (9 .5 ) bo 0 L - A y (9.2) R 0 <: t n * =A y + Ri (9.6) in yI - (1 A + i'A )Y (9.7) yx' i'X + i'R (9.9) U n it II Total "yu x9i (9.11) g i'A + i'A (9.8) .. A y - X' (R-R')i (9.10) a n n Note: An = Nyn- matrix of average endogenous expenditure propensities Ai = L&.' matrix of average propensities to leak Ni = n = vector of row sums of N = Anyn Xi = x = vector of row sums of X Li = = vector of row sums of T. =A.y,, Xs' = V'A, = vector of column sums of A,, i.e., the vector of aggregate average propensities to leak N = rnatrix of SAM transactions between endogenous accounts X = matrix of injections from exogenous into endogenous accounts L = matrix of leakages from endogenous into exogenous accounts R = matrix of SAM transactions between exogenous accounts. Table 9.3. Estimates of Matrices Ma, AfMa, X, and R Derived from Table 9.1 Expenditure. tndogenCue Accouits Exogenous Account 3 2 3 1. 5 Productlon llousehold Corporate _ rduction Activitle 6 lest R of World Fecttrs of Current Cutrent 0 - l.hour Capital Accounta Accounte Cooasodity Croups 2 | 600 | e Sa X 0 | L;0 w | * j *> 1 12 - W - 14[ u 0 ~ 1 Vol . a | i | I- 0~~~~~~I b ttrb_ 1 1.19 0.22 0.23 0.18 0;19 0.22 0.23 0.O 0.22 °.24 0.22 0.23 0.23 0.39 | ° 0: .1 1 0.38 1.45 0.49 0.37 0.38 0.45 0.49 0.19 0.53 0.74 0.60 0.40 0.45 0.62 0.02 0.02 0.02 0,02 0.01 0.52 0,03 0.03 0.02 0.02 0.03 X _t tce 1 0.02 0.02 1.03 a ft_yes 0.61 0.12 0.78 1.59 0.61 0.72 0.78 0.30 0.60 1.14 0.97 0.13 1.13 0.98 0,02 0.02 0.02 1.00 0.02 0.02 0.02 0.01 0.02 0.02 0.03 0.05 0.04 0.03 C U PwubLic 0.02 2.23 2.44 2.55 2.18 1.00 1.23 1.44 1.54 0.60 2.09 2.17 1.84 1.42 1.87 2.06 Sub Total U4b_ 1.33 0.37 0.39 0.52 1.32 0.37 0.39 0.47 0.39 0.48 0.43 0.38 0.48 0.bO 96 92 6 s 2 ACu geae Rural 0.77 1.90 0.91 1.36 0.71 1.90 0.97 0.52 1.03 1.45 1.20 0.85 L.15 1.23 159 153 6 1.03 0.02 0.53 0.04 0.074 0.03 0.03 0.04 12 6 6 lAa 0.03 0.03 1.03 0.04 0.03 0.03 U c.rpoaeca PTVC 0.35 0.18 0.19 0.39 0.15 0.38 0.39 3.01 0.20 0.28 0.21. 0.18 0.28 0.24 294 294 5 3 Accounts IState 0.02 0.02 0.02 0.02 1.00 0.02 0.02 0.02 0.01 1.00 0.02 0.02 0.03 0.05 0.04 0.03 2.51 2.62 2.33 1.00 2.29 2.51 2.62 2.09 1.00 2.16 2.27 1.94 1.49 1.95 2.15 563 545 l8 Sub TotaL 2 4 3 2.29 0. * tea and 0,01 0.02 0.02 0.01 0.01 0.02 0.02 0,01 1.02 0.02 0.01 0.01 0.01 0.01 1152 2 30 1180 e _ i other 0,53 0.68 0.76 0.54 0,53 0.67 0.76 0.27 0.61 1.62 1.17 0.44 0.49 0.54 206 20 78 108 0.38 0.41 0.30 0.24 0,37 0.41 0.14 0.33 0.32 1.29 0.27 0.21 0.30 268 26 90 152 s Y is Food 0.28 0.18 0.49 0.40 0.36 1.44 0.48 0.39 305 66 75 164 _ 0Other 0,37 0,43 0.46 0.35 0.31 0.43 0.46 0.03 0.03 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 1.14 0.02 1106 92 1595 19 MUt 0.17 0.63 0.69 0.75 0.71 0.33 0.70 0.66 0.62 0.53 0.70 1.66 2035 1371 154 490 Services 0.69 0.75 2.43 L.b4 1.90 2.26 2.43 0.93 3.16 3.02 3.46 2.71 3.09 2.92 5652 1517 1962 2113 Sub Total 1.90 2.26 S 3 er-ao Curanac Accmu 0.19 0.11 0.10 0.15 0.32 0.19 0.11 0.10 0.24 0.32 0.10 0.13 0.12 0.10 0.13 0.14 1507 75 1338 44 0.68 0.39 0.37 0.28 0.54 0.68 0.28 0.37 0.32 0.26 0.34 0.36 423 41 382 6 ActItut 0.39 0.31 0.26 0.40 o.32 o.13 o.13 0o.06 0.16 0.13 0.15 0.25 0.11 0.15 430 29 270 131 4 7 Indirect Tare (n*t) 0.12 0.13 0.13 0.13 0.37 0.46 0.15 0.44 0.34 0.39 0.36 0.33 0.33 128 43 364 -279 S R;et . 0.29 0.37 0.41 0.30 0.29 0.01 0.02 0.02 0.04 0.01 0.02 0.02 0.0'2 0.02 0.03 0.02 0.02 0.03 0.02 36 36 o fiworl: 0th. 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 2524 224 634 13368 31 141 Sub Totat 1.00 1.00 1.00 192 and SAM-Based Multipliers Models noted here that the linkage between injections and leakages as given by equation (9.13) satisfies the requirement (9.14) i'AtM a = X'aMa = i or, in words, that each injection is ultimately accounted for by one or more leakages. 3 In deriving the matrix Ma it has been assumed that the accounts for factors, households, companies, and production activities are endogenous. The corresponding exogenous accounts are therefore those for government current expenditure, investment, 4 indirect taxes, and inter- national transactions. Injections, x, therefore include current transfers to households and companies both from government and the rest of the world, plus the demands placed on produc- tion activities through government consumption, investment, and exports. Direct and indirect taxes, savings, imports, and income transfers abroad constitute the leakages. 5 DECOMPOSITION OF ACCOUNTING MULTIPLIERS From equation (9.4) it follows that for any matrix An of the same size as An and such that (I - An) exists, we can write (9.15) ,Y = Any. + x (An - An)y + AnYn + x = (I - AAn-(An - A_)Yn + (I - x = A*Yn + (I - AD-' x. Multiplying throughout by A* and substituting for A*yn on the right hand side of equation (9.1 5) gives Yn = A*2y. + (I + A*)(I - A>-l x. Similarly, multiplying both sides of (9.15) by A*2and substituting for A*2ynin this last expres- sion, we get (9.16) y, = A*3yn + (I + A* + A*2 ) (I - An) I x = (I - A* 3 ) l (I + A* + A*2 ) (I - A°)-IX provided that (I - A * 3) -1 exists. Comparison of (9.16) with (9.12) shows that the above algebra has decomposed the account- ing multiplier matrix Ma into the product of three separate matrices. This decomposition is quite general. It can become informative by referring to the structure of A,, and choosing A°4 accordingly. Specifically, we can write - O AO3 O O O (9.17) An = A21 A22 ° and An = ° A22 0 0 A3 A33_ 0 A3 3 3. This follows from the equation: i' = i'(Ak + A,) = V'A, + XA = X',(I - A.)D = X.M. 4. To obtain Tinbergen's semi-input-output model as a special case of our analysis, it would be necessary to make investment in the nontraded goods sector(s) endogenous. 5. It can be noted that if the model were to specify the import and indirect tax content of government expenditure, investment, and exports, some elements of R would be determined as a function of x. The equation (9.11) would then be sufficient to determine the balance of trade, government savings, and the current account deficit on the balance of payments. Accounting and Fixed-Price Multipliers 193 so that A* defined by equation (9.15) can be written o 0 A*13 (9.18) A= °j O A3*2 0 where A*3 = A13 A2I = (I -A 2 2 ) 1 A2 1 A32 = (I -A 3 3 ) 1 A32, and where the partitioning of An (and of A-. and A*) corresponds to the separate accounts in the SAMfor factors, the endogenous institutions (households and companies), and production activities. At this point in the argsument it is worth noting that the three-part decomposition of Ma in (9.16) does not require the three-way partitioning of matrix An as in (9.17): An can be parti- tioned into as many (or few) sets of accounts as one wishes. Similarly, there is nothing special from a mathematical perspective in choosing to end the sequence of substitutions which leads to equation (9.16) after three steps. Further substitutions are possible, and the general result is Yn = A(I - *)- (I + A* + A *2 + *+ A*(k - 1) ) (I - A°)-' x. Our choice of three partitions for A,, and the decision to end the chain of successive substitutions after three steps (k = 3) derives from the structure of the SAMin table 9.1. And this structure derives in turn from the conceptual framework of economics. Thus the particular application of the mathematics that i.s illuminating in our context is to have three partitions of An (corre- sponding to factors, endogenous institutions, and production activities) and to choose k = 3 not simply because there are three partitions, but because with this particular trio of partitions, three steps in the sequerLce of substitutions corresponds to one complete cycle in the circular flow of income within the economy. Further reference to the SA.M(table 9.1) shows that with the chosen partitioning of An, its zero submatrices are indeed empty blocks within the accounts. The nonzero submatrix A1 3 reflects payments from activities to factors; A21 corresponds to the mapping of income from factors to households and companies; and the nonzero elements of A32 record the average propensities with which different types of households consume the goods produced by the various production activities. Submatrix A., captures current transfers between endogenous institutions and, in our example, is restricted to the distribution of dividends and interests to households. Submatrix IL33 shows the transactions between activities, that is, interindustry flows. With these conventions we now define (9.19) M., = (I -- An) - 1; M.2 = (I + A* + A* 2 ); M^5 = (I - A*3) - with the implication from (9.12) and (9.16) that (9.20) M. = M.Z M.2 M. 1 Equations (9.17), (9.18), and (9.19) imply, first, that M,, is a block diagonal matrix with 194 Multipliers and SAM-Based Models successive diagonal elements given by I, (I - A 2 2) - 1 and (I - A3 3 ) - . They also imply that [ o AA*3A3 2 0 s r I A*13A3 2 A1 3 (9.21) A*2 = O*3 sothat Ma2 A2 1 A A2AI 3 _A2A2l 0 O A2A'L A32 I and that Maz is also block diagonal with successive diagonal elements given by (I - A*3 AL 2 A2 1) , (I - A2A1 3 A3 2 ), and (I - A*3) The structure of Ma2and Ma 3 derives from that of A*. From (9.18) it can be observed that the pattern of zero and nonzero cells of A* corresponds to a circular permutation matrix of size 3 x 3. Accordingly, if yn is partitioned compatibly with A,, then the structure of equation (9.15) implies that the partitions of yn are related to each other as points on a closed loop. In figure 9.1 these points are shown schematically as the corners of a triangle. Matrix A*j represents the mapping from one partition of yn to another. Starting from any corner of the triangle, three steps in this mapping brings one back to the starting point. Hence the structure of A* implies that our formulation contains a closed-loop system, which is the algebraic statement of the circular flow of income: for example, from activities to factors to institutions and then back to activities in the form of consumption demand. This structure explains why M., is block diagonal and justifies referring to this matrix as the closed-loop or circular multiplier matrix. Matrix Mal is also block diagonal as previously noted. It captures the effects of one group of accounts on itself through direct transfers and is independent of the closed-loop nature of the system. Since there are no direct transfers between factors, the first diagonal block of Ma. is simply an identity matrix. The second diagonal block captures the multiplier effect resulting from direct transfers between institutions, (I - A2 2 ) -. The third diagonal block similarly refers to the multiplier effect of interindustry transfers, (I - A3 3 )- l, which is the Leontief inverse. Matrix Mal can be referred to as the transfers multiplier. If Ma, and Mz3 are block diagonal, all effects between partitions of Yn must be captured by M3,2. This matrix is therefore referred to as the cross-effects matrix or alternatively as the open- loop multiplier matrix. This terminology can be justified by considering the implications of one partition of yn for the others. Take as an example the effect of household and company incomes on both factor incomes and production. This is an open-loop system and equivalent to breaking the closed loop by setting A21 = 0, that is, the effects of factor incomes on the incomes of institutions is ignored. From (9.18) it is apparent that Ajl is now zero, so that all terms in Ma3 and Ma2which involve A2 1 will be zero. This implies that Ma3 will now be an identity matrix. From (9.21), certain cells of Ma, will also be zero. But the columns of Ma, which refer to households and companies will be unaltered. These columns show the impact of incomes in the second partition of yn (endogenous institutions) on factor incomes (the first partition) and activity incomes (the third partition) in an open-loop system. So far the discussion has assumed that the matrices Ma,, Ma2,and Ma 3 exist and that it is legitimate to describe them as multiplier matrices in the sense that each has elements that are not less than the corresponding elements of an identity matrix. To justify this it can be noted that the matrix An is semipositive. 6 It follows that Ma will be a multiplier matrix if it exists. Mathematical conditions for the existence of Ma can obviously be postulated (see Lancaster, 1968, pp. 94-95). If An is a semipositive indecomposable matrix, then Ma will exist if no column 6. This is always possible in a SAM since a negative element in the ith row, jth column can be set equal to zero and balance restored by adding a positive element of the required size in the jth row, ith column. Accounting and Fixed-Price Multipliers 195 Figure 9.1. The Closed-Loop Structure of Accounts X, Factor incomes received from Factor /2 V incomes / 1 =(I - A22 )' x 2 = (I - 3)YAxX3 where x2 =nonfactor wherex 3 = income received export from demands ,abroad \ r d: = (I -- A2*1 A2 2 ) AA3 A-1 Y2 Incomes of A32 Icmso endogenous = (I - A33)-'A3 2 domestic institutions producl n Note: The closed-loop structure presented here is defined by equations (15), (17), and (18). xe = x2 Yn p Y2 where the partitioning is thg same as in equation ( 17) . Table 9.4. The Matrix Product Ms2Mal Origin of Injection 1 2 3 __ Factors of Production Household Corporate Production Activities Current Current Labour Capital Accounts Accounts Commodity Groups- 5 0 _~~ .1 ~~ . ~ ~ . ~ ~ .. ~ ~~ .I . I Cd to P. 0. :a 94 A' t- A ° .4 o .4 °4 l "A oc Urban 1.00 0.09 0.09 0.10 0.04 0.04 0.05 0.07 0.11 0.08 0.22 o o Rural 1.00 0.17 0.20 0.22 0.08 0.14 0.35 0.28 0.16 0.13 0.27 N m __ Estate I.4oo 0 .0 0 -. 0.0 * 0.50 0.01 0.01 0.01 0 o Private 1.00 0.27 0.32 0.35 0. 0. 0.18 05 045 0.34 0.62 0.41 a - o Z Public 1.00 0.01 0.01 0.01 4 04 0.07 0.3 0.02 0.02 Sub Total 1.00 1.00 1.00 1,00 1.00 0.55 0.63 0.69 O.Z6 0,86 0,94 0.82 0.64 0.86 0.92 1 Urban 1.00 0.22 1.00 0.31 0.08 0.17 0.16 0.18 0.21 o.31 Household 2 Cprrept Rural 1.00 0062 1.00 0.15 0.25 0.68 0.56 0.37 0.51 0.52 cAccPunts o _ _ Estate 1,00 0,01 1.00 0.01 0.50 0.02 0.02 0.01 0.01 0.01 Corporate Private 0.25 1.00 0.04 0.13 0.11 0.08 0.15 0.10 3 Current c Accounts StatePu 1.00 1* * 0.01 0.03 0.02 0.01 Sub Total 1.00 1.00 1.00 1.09 1.00 1.00 1.00 1.00 1.47 1.00 0.90 1,01 0.86 0.67 0.90 0.95 Rubeand 1 0.01 0,01 0.01 1. 0 0000 0.01 1.00 * * 1 0.3 v Et Ariculture 0e22 0.31 0.3624 1022 0,31 0.36 0.11 0.02 1.05 0.69 0.08 0.02 0.02 Sub a co t P athe 0,11 0.17 0.19 0. 0.17 0.19 0.06 0.01 0.01 1.02 0.08 0.01 0.01 v aufcue 0,16 0.19 0,20 0,16 0.16 0,19 0,20 0.08 0.12 0,04 0.06 1.22 0.18 0.05 Mining and Aounts Construction 1.00 **0.01 1.0001 * * 00 0 1.13 0.01 4 02 ~Services 0.33 0.33 0.31 0.28 0.33 0.33 0.31 0.15 0.07 0.02 0.08 0.12 0.15 1.05 Not: Note s idcantutiaount 0seik Sub Total ther L0.183tha *es 0 7 1.01 001 1.08 . _ 0.82 **1001.* 0.83 1.01 1.08 0.41 _. 1.28 ** 1.12 1.85 1.51 .3 1.50 00 1.13 Accounting and Fixed-Price Multipliers 197 sum exceeds unity and at least one column sum is strictly less than unity. Expression (9.8) supports the former conditions, and we have only to guarantee a leakage from some accounts for MBto exist, providing of course that A, is indecomposable. It is of interest to note that since (A, - AS) can be viewed as a circular permutation matrix then A, is certainly "block" inde- composable of order three. But this is not a sufficient condition for A, to be indecomposable in the general sense. The existence of Ma is enough to ensure the existence of Mal. This can be shown by first noting thatA, is a semipositive, completely decomposable matrix. If the condition on the column sums hold for the existence of Ma then they will hold for the existence of Mal, since A° is contained within A,. Furthermore, Mal will be a multiplier matrix. It also follows from (9.16) and (9.17) that A* will be semipositive if Ma, exists. Hence from (9.21) MB2 will exist and will be a multiplier matrix. Finally, from (9.20), since Ma, Ma2,and Mal all exist, then Ma, must also exist because it is bounded by finite matrices on both sides. Moreover, A*3 is semipositive, so that Ma3is also a multipliar matrix. A final remark on the existence of these multiplier matrices is to note that they essentially depend upon the designation of at least one exogenous account with at least some injection into, and hence some leakage from, the endogenous accounts that remain. This ensures at least one element of Ma is positive. To provide a useful way of presenting the results of our decomposition, Stone has proposed (see chapter 8) an additivis form of equation (9.20), namely, (9.22) M = I + (Ma, - I) + (M&2 - I)Mal + (Ma3 - I)Ma2Mal so that elements of Ma are accounted for by (a) the initial injection; (b) the net contribution of transfer multiplier effects; (c) the net contribution of open-loop or cross-multiplier effects; and (d) the net contributicn of circular or closed-loop multiplier effects.7 To illustrate this form of the decomposition requires results for the product matrix Ma2M,l (table 9.4) in addition to the details of M'. 8 FIXED-PRICE MULTIPLIERS The accounting multipliers described in the preceding section are interesting for the infor- mation they contain on the structure of an economy as revealed by a SAM. However, because they are accounting multipliers, they cannot be interpreted directly as measures of the effects of changes in injections into the economy on the levels of endogenous incomes. For such a purpose we need to know how different economic agents behave in response to changes. In this and subsequent sections, we shall be concerned with the behavior which generates the expenditure patterns of eindogenous accounts under the assumption that prices remain fixed when income is altered. Since prices may in fact change, the multipliers obtained under this assumption are referred to as fixed-price multipliers. Under the assumption that prices are fixed, it follows from the accounting balance equation (9.3) that (9.23) dy1, = dn + dx 7. The arrangement (9.22) is applied by Stone in the previous chapter to a decomposition M. = M,,M,,M. 1 so that, in comparison with (9.20), the crder of M., and M,, is reversed. This alternative ordering was used by us in Pyatt, Roe, and associates (1977), ch. 4. It is easily checked that both orderings are legitimate. However, the ordering adopted in (9.20) is perhaps to be preferred since it corresponds to the progression from transfer effects to open loops to closed-loop models. 8. It can be noted that, since M.1 is block diagonal, it follows from the structure of M,, defined in (9.21) that setting off-diagonal blocks of M,,M.1 eqiu. to zero reduces this product to the matrix M,1 . 198 Multipliers and SAM-Based Models (9.24) = On dyn + dx (9.25) = (I - Cn)J dx = M 0dx and similarly that (9.26) dt = CGdyn (9.27) = C,(I - Cn>' dx = CeMcdx. The result (9.23) is obtained by taking the total differential of (9.3). Equation (9.24) then follows from the fact that, if prices are fixed, the vector n of incomes received by endogenous accounts as a result of expenditures by these same accounts can be a function of Yn but otherwise is constant. Hence (9.24) follows from (9.23) if the (i,j)" element of matrix Cn is the partial derivative of the ith element of n with respect to the jth element of yn. In this sense, COis a matrix of marginal propensities to consume. If (I - Cn) -' exists, then equation (9.25) shows how elements of yn change as a result of changes in injections. Similarly, the matrix Ce in equation (9.26) is a matrix of marginal propensities to leak, and equation (9.27) shows how leakages change as a result of injections. Equations (9.25) and (9.27) are analogous to equations (9.12) and (9.13). Consequently, under the condition that the matrix Cn is nonnegative, M 0 is a multiplier matrix, to be referred to as the fixed-price multiplier matrix. Matrices Cn and Cf will have column sums that add to unity, and M. will exist under conditions analogous to those for the existence of Ma. Hence, given estimates of the matrices Cn and Ce.both the fixed-price multiplier, M,, and the matrix of marginal leakages, CeMc,can be calculated. These matrices are illustrated in table 9.5 using data for Sri Lanka, which are discussed below. To go further, we need to consider data sources for C, and C,. This can be done with reference to table 9.1, which shows that the outlays of factor incomes primarily generate incomes for the endogenous domestic institutions. The table shows that all urban labor income accrues to urban households. Thus the first column of Ce is zero, and all elements of the first column of C. are also zero except the element in the row for urban households, which is one. Thus the sum of the first column of Ce,plus the sum of the first column of On,is unity, as it must be. The second, third, and fifth columns of Cn and CQare similarly obtained. For the fourth column, there are five different recipients of the income that accrues to private capital. The proportions in which they receive this income will depend on who owns private capital. And if the structure of ownership can be taken as given, then there is no reason to assume other than that incre- ments of income will be distributed in the same proportions as the shares observed in the SAM. On these grounds, columns of C, and CO, which refer to factor outlays, are estimated by assuming that marginal and average propensities are the same. For marginal and average propensities to be equal requires income elasticities of particular expenditures to be unity. This is clearly not true for household expenditures, and table 9.6 sets out the marginal propensities that have been assumed. It is to be noted that the income elas- ticities of demand for imports are unusually low. This is partly because consumer imports in Sri Lanka include imports of the staple foods, rice and wheat, and partly because the observed cross-section elasticity has been lowered in recognition of the restrictions on imports which were in effect at that time. For companies, marginal allocations of income have been assumed to be equal to the average allocations implied by table 9.1. This is in default of any better basis for deciding how corporate taxation, savings, and distribution policy might be responsive to changes in corporate income. It has also been assumed that the allocation of total costs for production activities is the same at the margin as on average. The best way to justify this is as follows. First, the assumption of fixed prices would be reasonable if interindustry technology follows Leontief assumptions, Accounting and Fixed-Price Multipliers 199 Table 9.E6. Estimates of the N.atrices M, and CFM, Derived from Tables 9.1 and 9.6 Otigin .)f InJeicio Feators of Production 4ouehold Corporate Prduction Activicis Current Current Labour Cpcouct* Accounts Co dlty Groups Z el 1ST;. . b a .0 a _ C A = , ,|X 1 4 tU| § - Urban 1.19 0.25 0.2i 0.2.0 0.19 0.25 0.28 0.10 0.26 0.26 0.24 0.24 0.25 0.41 u o Rural 0.34. 1.47 O.S8 0.37 0.34 0.47 0.58 0.18 0.5D 0.74 0.60 0.40 0.45 0.62 1 6 : Eate 0.01 0.02 1.03 0.02 0.01 0.02 0.03 0.01 0.52 0.03 0.03 0.02 0.02 0.03 a _ rivate 0.55 0.76 0.94 1.60 0.55 0.76 0.94 0.29 0.88 1.16 0.98 0.73 1.14 0.98 *~ u EublIc 0.02 0.03 0.03 0.02 1.00 0.02 0.03 0.03 0.01 0.03 0.03 0.03 0.05 0.04 0.04 Sub Total Z.12 2 2.53 2.36 2.21 1.00 2.12 1.53 1.86 0.58 2.26 2.21 1.08 L.44 L.90 2.0B Crbam 1.31 0.41 0.49 0.54 1.31 0.41 0.49 0.47 0.45 0.51 0.45 0.40 0.49 0.62 HousehoLd 2 Curran Rural 0.68 1.94 1.16 1.37 0.68 1.94 1.16 0.50 1.13 1.46 1.21 0.86 1.16 L.23 U _Estaze 0.02 0.03 1.04 0.04 0.02 0.03 1.04 0.02 0.53 0.04 0.04 0.02 0.03 0.04 Corporate Private 0.14 0.19 0.23 0.39 0.14 0.19 0.23 1.07 0.22 0Q28 0.24 0.18 0.28 0.24 a 3 Current e Accounts Scate 0.02 0.03 0.03 0.02 1.00 0.02 0.03 0.03 0.01 1.00 0.03 0.03 0.03 0.05 0.04 0.04 a- C Sub TotaL 2 + 3 2.17 2.60 2.95 2.36 1.00 2.17 2.60 2.95 2.07 1.00 2.35 2.32 1.97 1.30 2.00 2.17 _ __ Ru Rabber 0.01 0.01 0.02 0.01 0.01 0.01 0.02 * 1.O1 0.01 0.01 0.01 0.01 0.01 _ Ocher 0.38 0.59 0.84 0.46 0.38 0.59 0.84 0.21 0.62 1.53 1.09 0.38 O°.4 0.45 ca 4 'rocessir,sr < 0.22 0.38 0.52 0.29 0.22 0.38 0.52 °.13 0.38 0.31 1.28 0.27 0.26 0.28 3 a Other re Manufactu 0.42 0.62 0.64 0.48 0.42 0.62 0.64 0.22 0.63 0.54 0.48 1.53 0.59 0.50 Co..La nd. 0.01 0.01 0.02 0.01 0.01 0.01 0.02 0.01 0.01 0.01 0.01 0.01 1.14 0.02 Services 0.72 O.S6 0.97 0.70 0.72 0.86 0.97 0.35 0.83 0.75 0.69 0.56 0.76 1.73 Sub Total 1.75 Z.47 3.00 1.95 1.75 2.47 3.00 0.92 3.48 3.16 3.56 2.77 3.17 3.19 5 Cov.rment Current Accoust 0.26 0.15 0.16 0.19 0.32 0.26 0.15 0.16 0.27 0.32 0.15 0.17 0.15 0.13 0.17 0.18 o 6 Ace..ount d 0.3 0.44 0.38 0.46 0.68 0.43 0.44 0.38 0.53 0.68 0.35 0.42 0.37 0.30 0.39 0.41 7 tddirect .axes (net) 0.13 0.17 0.19 0.14 0.13 0.18 0.19 0.07 0.20 0.16 0.18 0.27 0.20 0.18 e 8 a 0 1tRet |Caeodiey rassactions 0.16 0.22 0.25 0.17 0.16 0.22 0.25 0.08 0.28 0.22 0.28 0.28 0.22 0.21 Wor Othar 0.01 0.02 0.02 0.04 0.01 0.02 0.02 0.02 0.02 0.03 0.02 0.)2 3.03 0.02 Sub Total 1.0 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1300 so that there are no scale effects, and prices are fixed for given indirect tax rates if import prices are fixed and factor costs per unit of output are constant. These assumptions would make elements of Ae and C. the same in the corresponding columns for production activities, with a similar equivalence of those elements of A, and Cnthat refer to interindustry transactions. With respect to factor payments, profits will have a constant share if value added price, that is, value added per unit of output, is set as a constant markup over labor costs per unit of output. And labor costs per unit of output will be constant if labor is paid at fixed piece rates. Alternatively, it can be assumed that wage rates are fixed and the average product of labor is constant. This alternative assumption is necessary if labor incomes are assumed to be proportional to employ- ment levels. It implies thar, the economy is working below capacity in all sectors. With these assumptions it is not unreasonable to assume that prices are fixed and that columns of O, which relate to activities can be estimated by columns of A,. In aggregate, the above arguments imply that An is equal to Cn (and similarly for Ae and Ce) except for the data in table 9.6. These arguments also illustrate the fact that to estimate C,, and Table 9.6. Average and Marginal Expenditure Propensities of Households in Sri Lanka, 1970 Urban Rural Estate Average Marginal Average Marginal Average Marginal cn Tea and Rubber 0.005 0.002 0.008 0.006 0.008 0.006 4J IX 0 > Other Agriculture 0.134 0.080 0.174 0.134 0.214 0.241 4A 0 0 Food Processing 0.092 0.059 0.154 0.149 0.174 0.246 V. o o 0 ., Other Manufactures 0.114 0.122 0.131 0.204 0.142 0.156 0 Mining and CMnitucind 0.001 0.001 0.001 0.001 0.001 XI 0 0.001 W h O~ Construction o ci _ Services 0,.291 0.315 0.280 0.311 0.263 0.302 SuJb 'rotal 0.637 0.579 0.748 0.805 0.802 0.952 5 GovernotentCurrent Account 0.111 0.164 0.017 0.022 - 0.010 6 Institutions Capital Account 0.173 0.209 0.117 0.135 0.014 0.015 o 7 Indirect Taxes (net) 0.011 0.020 0.010 0.014 0.004 0.005 00 0 Rest CosmoodityTransactions 0.069 0.028 0.107 0.024 0.181 0.018 8 of World Othler - - - - - Sub Total 0.364 0.421 0.251 0.195 0.199 0.048 Source: Pyatt, Roe, and associates (1977), ch. S. Accounting and Fixed-Price Multipliers 201 hence Me,it is only necessary to estimate a SAM and those income elasticities which are different from unity. DECOMPOSITION OF FIXED-PRICE MULTIPLIERS A further implication of the discussion in the previous section is that the patterns of zero and nonzero entries in partitions of O, and A, are the same. Hence the fixed-price multiplier matrix can be decomposed into a transfer effects multiplier, Me1; an open-loop multiplier matrix, Me2 ; and a closed-loop multiplier matrix, Mc3. Furthermore, these effects can be expressed multiplicatively as (9.28) Mc = Mo3MC2Mel or in Stone's additive form (9.29) Me = I + (M. 1 - I) + (Me2 - I)Mcl + (Mel - I)Mc2 M.l- Results for Mel and Mc 2 are shown in table 9.7. With prices fixed, the differences between corresponding elements of Ma and M, must be due to income effects. This can be formalized by writing from (9.24) (9.24) dy. = C. dYn + dx (9.30) = (C. - A.)dy. + Andy. + dx = (I - A)-l[(Cn - An)dyn + dx] = Ma(Cn - An)dyn + MadX (9.31) = [I - Ma(Cn - An)] Madx (9.32) = My Ma dx where (9.33) My = [I - Ma(Cn -An) and MiyMa = Mc- Thus the income effects can be captured in a matrix My which transforms the accounting multiplier matrix Ma into a fixed-price multiplier matrix Me. However, My itself is not a multi- plier matrix because, as can be seen from (9.32), elements of Mycan be negative since elements of C, can be less than the corresponding elements of A,, that is, income elasticities can be less than one. In our example, the matrix My is particularly simple. Since only households have income elasticities which differ from 1, it is only in the columns for households that Mydiffers from an identity matrix. EMPIRICAL RESULTS A number of general points as well as particular features of the Sri Lanka economy in 1970 can be illustrated from the empirical results. Of the general features it can be noted from table 9.5 that the columns for factors contain little information that is not included in the detail for institutions. With respect to labor this is partly because there is a one-to-one relationship between types of labor and types of household, and partly because the basic SAM shown as table 2MC Table 9.7. The Matrix Product MC 1 Origin of Injection 1 ~~~~~~~~2 3 4 Factors of Production hlouselhold Corporate Production Activities Curreat Current Labour Capital Accountsb Accounts Commodity Grouips 4 S~~~~~~~~~~~~ a' I 31.u 4 0 qo b ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~t o oa Purbli 1.00 .09 .01 .01 *04 o4 .I .072.1I .0 .24 0 a Prvate 100.24 .34 ~.1 .4 .18 .5 45 .34 .6244 0 Stib ESbTotateTotal140 1.00 1.0 1.00 11.00 0 .00 1 A00 A .4 .49 6At .67 .81 00 .252 6 .86 A .9 .51 8 .82 6 65 .5 .85 .9 .91 ___ ._ _ __ __ _ _ b _ _3 Urbau1 1.00 1.00 .31 .08 .17 .16 .11 .21 .31 i louselbold .08 u 2 Curreut Rural 1.00 .2 .10 .25 .68 .56 .37 .52 .52 o Acc-ounts Estate 1.00 01 n.0o * .50 .02 .02 .01 .01 .01 Corporae Private .25 1.00 .04 .13 .11 .08 .62 .10 3 g. C..rres.t SbFotao.0 10d.0 100 10 4 6 8 2 .86 .95 1.82 0 .65 0 .85 .2 .91 .1 to _ ccount a State _ _1.00 I.O. l 1.00 1.00 1.00 1.00 1.46 1.00 .88 1.00 .86 .68 .91 .95 j _ _ atid Tc~~~'fa * 01 .01 .01 * .01 .01 * I00 * * * * u ~~~R1bber * Cuprt. u Iae.2 lr4utr .14 .27 .44 .20 .14 .27 .44 .0.04 .09 .02 .03 1.05 .11 .69 .08 .08 .15 .02 1.05 .02 o ra .07 17 27 13 .07 17 .27 .05 08 .1 1 .8 21 .1 42 s.ru Rurl 1.00 1.00 .h5 ..2 .64 .06 .32 .52 .5 aOu0ie .17 .28 .23 .21 .17 .28 .23 .09 .2 .4 .6 12 1 0 u Q HA1X; ]rel k Intitu- Production 2 *" Factors Intiu Pouton TFactors in civte tions233 A3one tcti v vc k3 k J =12 -- sw Or Cther -| 1 2 3 4 | x Oh 1 2 3 5 * 3, C I=cstituticns Factors ~ 1 .~~~~~~~~~~~~~~~~~~ at X- 2. 3 0 Factors Institutions 2 1 kI,k =OU- * Y = :ylL-tl5 .3l51-2h 11 =m i u a n ttroduction Production Activities T 3.2 is o kctivity j A Accmmts 2 x ____ ___ u 1 J~~Atiite x'ote: Matrices are denoted by capital letters and vectors and scalars by lower case letters. Table 10.6A is the SAM transaction matrix where the notation is analogous to that in table 10.3, while table 10.6B gives the corresponding SAM coefficient matrix, i.e., Hk T 2.2:'; Cjk = T,, 2 9', etc. It should be noted that Su in the notation table 10.6matri coeficihent that in table 10.5. Table 10.6 introduces a notation which of inemditteanhfaciiyjeorctvt differs from is followed consistently in the evaluation of all subsequent deand Xk = exogenous (injection) o intiuio foohr I accounts" models. Hkk coefficient matrix of transfers among institutions, = where kt = 3 household categories (i.e., rural, urban poor, and urban rich households) Vj coefficient matrix of value added allocation from activity j to institution kt; j =12 sectors = = coefficient matrix of consumption Cil expenditures by institution kt on activity = coefficient matrix of initermediate Lj demand of activity jfor activityj Xk =exogenous demand (injection) of institution k for "other accounts" Xi = exogenous demand (injection) of activity jfor "other accounts" l, and lj = exogenous leakages out of institution k and activity j, respectively hk = y 2 total income of institution k Pi = y3 total demand (= output) of activity j. Consistency-TypePlanning Models 223 equals marginal propensity to consume of household k on activityj. This type of specification Cjk of consumption allows one to postulate different income elasticities of demand for different income groups. Linear approximations of Engel curves can be obtained. The system as represented in equation (10.9) is moved by the exogenous components which are collapsed into the two vectors x, and xk. These components consist of government expen- ditures, investment, exports, and the government's indirect tax receipts from exports (mainly revenue from the nonresident oil sector). These variables jointly determine the household income distribution and the output mix. An appropriate interpretation of the model is that it generates consistent, disaggregated output and income multipliers for any given change in the exogenous variables. In addition, by linking employment to output levels through Verdoorn-type coefficients, it can also yield estimates of employment. It is therefore essentially a Keynesian model. There is no guarantee that supply is capable of meeting the total demands generated by the system. Consequently, an attempt was made in the Iran model to check the projected gross output levels resulting from running the model according to the assumptions made in the plan against some independent estimates of likely capacity availabilities by sector. In fact, the results of policy simulation indicated that supply constraints were an important potential issue in the Iran context, and that, for example, a failure to plan for agricultural expansion consistent with demand would tend to lower rural incomes in the model and hence aggravate rural-urban income differentials. The major characteristics of the Pyatt model are presented in table 10.4. There is no doubt that this model represented an innovative approach for deriving endogenously incomes and income distribution in an :interdependent fashion. One of its major advantages is that it was embedded from the outset in a SAMdata system which provided the base-year information and permitted the calibration of all linear coefficients appearing in the model. The Thorbecke-Sengupta (].972) Model of Colombia The scheme of the consistency framework developed by Thorbecke and Sengupta (1972) applied to Colombia is set out in figure 10.3. The general spirit of the model closely parallels that of the Iran model discussed previously: certain key exogenous variables such as exports and public expenditure determine other elements of final demand and hence, through input- output, the sectoral values added and employment. However, unlike the Iran study, both these elemenits are important in generating the personal income distribution which, in turn, yields the consumption component of final demand. The Thorbecke-Sengupta, study is divided into two major parts. The first part attempts to describe quantitatively the, macroeconomic and sectoral structure of the Colombian economy over the period 1950-67 in terms of output, employment, and income distribution. The second part projects these variables to 1980 within a consistent framework and under different assump- tions regarding export grciwth and technological change. Part I entailed a number of steps. The first step consisted of building a macroeconomic model of Colombia over the period 1950-67 which determined the paths of the endogenous variables consisting of gross domestic product, total consumption, investment, and imports as functions of exogenous variables such as exports, changes in terms of trade, and public expenditures. The second step consisted of obtaining input-output and employment information on a comparable basis in a ten- to twelve-sector breakdown. From this it was possible to derive the sectoral income distribution and tc design a methodology that provided a mapping between the sectoral and personal income distribution prevailing in the mid-1960s. This methodology will be discussed in detail later. Meanwhile, it is apparent from figure 10.3 that it plays a major part in the model, providing a feedback loop which in practice tends to stabilize the model considerably. OneLine Short 224 Multipliers and SAM-Based Models Figure 10.3. Analytical Schema of the Thorbecke-Sengupta Consistency Framework for Colombia Macroeconomic model Exogenous variables, such as exports, terms of trade EX T Endogenous variables such as GDP, imports, investment Sectoral final Input-output demand matrix , ~~~~EX, + Ci - M fi + Ii + AS, = Fi b :g ej ~ Income elasticities Sectoral f demand output Yd Y E V, Sectoral Personal income Sectoral income Sectoral 4 value distribution distribution employment | added Growth rates of labor productivity Source: Thorbecke and Sengupta (1972). Part II of the study projected the major macroeconomic variables to 1980 within the context of the macroeconomic model mentioned above on two alternative assumptions regarding exports and public expenditure variables. Next, the various components of sectoral final demand were projected in a way consistent with the macroeconomic projections. For example, final consump- tion demand for the various sectors was computed as a function of GDP growth, given likely values of the sectoral income elasticities of consumption demand. Likewise, the sum of the sectoral final demand components-consumption, changes in stocks, investment, exports, and imports-were consistent with (that is, added up to) the projected values of the variables appearing in the macroeconomic model. Third, the sectoral gross output and value added vectors were projected to 1980, given projected final demand and the consolidated input-output table of 1966. Furthermore, on the basis of magnitudes of the growth rates of labor productivity by sector likely to prevail over the proj ec- tion period, the sectoral employment and income distributions were derived. PlanningModels Consistency-Type 225 At thlis stage the methodology for mapping the personal income distribution from the projected sectoral-factorial distribution was used. Since Cobb-Douglas type sectoral production functions were postulated with constant returns to scale, the ratios of sectoral value added to gross sectoral output and the shares of labor income and nonlabor income out of sectoral value added remained constant over time. From this information, both the sectoral and the personal income distri- butions couldbe endogenously derived, and it couldbe determinedwhether the projected changes in the composition of output and employment would affect the personal income distribution. To the extent that changes in the latter were projected to prevail in 1980, revised projections of the final demand compcnents (specifically, consumption of agricultural and manufacturing goods) were undertaken to ensure consistency with the new income distribution. In addition, a fairly rudimentary test was conducted to check whether the alternative output combinations resulting from the projections to 1980 could be produced, given the total investment finds generated by the macroeconomic model. It was found that the investment availability would not constrain the attainrent of the projected sectoral output and value added combinations reached under the two growth alternatives. The whole set of projections described above reflected the likely consequences of maintaining the productive structure of the Colombian economy, since the input-output matrix prevailing in the base year (1966) was used to generate the projections. Their value for policy purposes is that they may provide the policymaker with a quantitative view of the consequences of essentially neutral technological policies.8 The final section of the model was devoted to a simple analysis of the effects of technological changes in agriculture on employment and income distribution. Even though this model is not explicitly based on a SAM framework, it can easily be fitted into one. This is attempted in table 10.7, which uses the same breakdown of accounts as in tables 10.3 and 10.6. Thus, the major differences between the Pyatt model and that of Thorbecke and Sengupta are clearly revealed by a comparison of tables 10.6 and 10.7. In contrast to the Iran model, the value added generated by twelve production activities here accrues to two factors of production (labor income and nonlabor income)-a transformation that is represented by the matrix Vj: in table 10.7B. 9 The next link allocates factorial value added to fourteen household income categories which are defined essentially along sectoral lines except for one entrepre- neurial and managerial class which receives mainly capital income. This mapping is repre- sented by Fk, in table 10.1'B.Thus, in contrast with the Iran model, value added first accrues to the factors of production before it is distributed to household groups. The determination of this last matrix is of particular interest. The totality of labor income outside of agriculture accrues to different income classes, according to the sector in which the workers are employed. (In other words, for each nonagricultural sector, there is a corresponding income class.) In turn, the great bulk of nonlatbor income outside of agriculture is assumed to be received by one entrepreneurial and marLagerial class. Finally, within agriculture, rather than distinguishing between labor and nonlabor value added-a distinction that would be meaningless, at least for the smallholders-total value added is apportioned to four agricultural income groups according to an existing knowledge of the agricultural income distribution. The above mapping generates 8. It is, of course, true that the underlying structure of an economy may change during the projection and planning periods. Thus, a model which describes accurately the performance of the macroeconomic variables over some (histor- ical) sample periods may still not predict well. It is even more likely that the intersectoral (input-output) structure will change over, say, a ten-year planning horizon. It may be very difficult to approximate quantitatively the new structural relations. However, even when it is not possible to make reasonable "guesstimates" of likely changes, it can be ve-ry revealing to simulate the effects of the maintenance of the prevailing structure (at both the macroeconomic and the intersectoral levels) and then to simulate various types of presumed technological changes. 9. It should be noted that table 10.7B gives the transaction matrix of the SAM,in contrast with table 10.6B which gives the coefficient matrix. Thus, for example, V. in table 10.7B denotes the allocation of value added from production activities to factors in absolute amounts, while Vkjin table 10.6 represents the coefficient matrix; that is, Ta.r = VJ in table 10.6B, while T1., = V. in table 10.7B. Table 10.7. A Schematic Representation of the Implicit SAM Structure of the Thorbecke-Sengupta Model of Colombia A. Transaction Matrix 1 B. Transaction Matrix 2 Expenditures Expenditures Endogenous Accounts FXog. Endgenous Accounts |Exag. Instltu- Pr*ut . , -0 Factors I-,uuctlo 0 - Factors In;titu- Production ° ton.s Activities -5 0 Ftos tionls Activities :1 9~~~~~~~ 0 % ~~~~~~~~~~~~~~~0 0 E- " < s =2 k = 14 j=12 1 2 3 4 5 1 2 3 5 Factors 1 T1. 3 X1 y1 Factors 1 V x S vS - ~~~~~~~0 !. r e Institutions 3___________ Production - 0 2 T .D _. 2 2 2 V~~~~~~~~~~~~~~~~~~~~~~~~ - o 0eut 2 Institutions ProucTo 2 F ks ____2 __T2.1 x k h k Activities 3 T 3.2 T3 3 x3 y3 Activities jk Su Su-ot Other ~~Accounts 1 11 2 _ 1' 3 ___~~~~~~~yx I u _X ot Other jAccounts lk 1' sky r y 1 Total 5 Y3 Yx Total 5 vs pi Xy| Note: Both panel A and panel B of table 10.7 are equivalent transaction matrices using two different notations. Matrices are defined by capital letters and vectors and scalars by lower case letters. %j = value added allocation from activity j to factor s; j = 12 sectors and s = 2 factors (labor, capital) Fk = factorial income s accruing to household category k; k = 14 household categories Cp, = expenditures by income class k on activity j (see text for discussion of this transformation) Li, = intermediate demand of activity j for activity j V. = value added income accruing to factor s. Consistency-Type Planning Models 227 fourteen income classes which are assumed, by definition, to be completely homogeneous in the sense that intraclass variance is zero.10 These income classes can be ranked on the basis of average class income, from the lowest to the highest (see table 10.8), The results of this clustering of income classes according to average income reveal the big gap that exists between traditional and modern sectors. The first two agricultural classes, embracing 80 percent of agricultural populations, personal services, craft manufacturing, and perhaps commerce, with average annual incomes ranging from 780 pesos to 2,190 pesos clearly represent traditional activities. At the other end of the spectrum, high average incomes in utilities, finance, government services, modern manufacturing, and the entrepreneurial class reflect modern activities. The last two columns of table 10.8 indicate the cumulative percentage of income and corre- sponding population, respectively, of the derived personal income distribution. A three-param- eter distribution was fitted to these data, which assumed that all incomes exceed a minimum threshold and that the distribution of the excess is lognormal." This distribution yielded for the base year (1966) a Gini coefficient of inequality of personal incomes of 0.55 which corre- sponded almost exactly to two independent estimates of the degree of personal income inequality in Colombia in the same period. The next endogenous transformation, namely, the determination of consumption demand for the various activities which is reflected by the matrix (Cjk)in tables 10.7, needs to be explained briefly. Rather than estimating demand functions by the k income classes for the i commodities, as is implied by this matrix, the demand for the twelve activities was directly related to the personal income distribution. In other words, a consumption relationship for each of the twelve commodities (or activities) was postulated as a function of the mean and the variance of the prevailing lognormal income distribution. 12 The final matrix in table 10.7B is the input-output matrix (I). In most projection runs, the input-output matrix for the base year was used, which, of course, introduced a bias. In a few runs, a very modest attempt was made at incorporating into the It matrix some expected technological changes in agriculture. The Ng (1974) Model of the Philippines Since the starting point for the Ng (1974) model was the Thorbecke-Sengupta Colombia study, it follows that the two have much in common. However, the Philippines model is more ambitious in a number of respects. First, the model disaggregates labor and endeavors to solve for employment by industry and skill. Second, relative prices are solved endogenously within the model. In both these and other respects, this model anticipates the specification of the nonlinear second-generation models which are reviewed in the next section. The Philippines model consists of four major subsystems: (a) the macroeconomic model, (b) the demand subsystem, (c) the supply, price, and employment determination subsystem, and (d) the income distribution subsystem. These four subsystems are interrelated as shown in figure 10.4. The macroeconomic model simulates or projects the values of the endogenous 10. This last assumption is, of course, unrealistic. In fact, every income class has its own distribution around its mean income, which is influenced first by the distribution of income for the employed within a sector, and second by the distribution within a household. Clearly, the procedure used here is, at best, an approximation of the true personal income distribution which can only be derived exactly when the distributional parameters of all income classes are fully specified. The difficult question of the sensitivity of the overall personal income distribution to the intraclass variance is discussed in the original source. 11. The value of the threshold was preassigned at 680 pesos per capita per year, corresponding to whatwas considered to be a subsistence level for the minimum feasible nutritional standard. 12. This yielded twelve consumption functions rather than the 168 functions (= 12 x 14), had consumption functions been estimated for each income class for each activity. Table 10.8. Distribution of Income by Sector and Derived Personal Income Distribution in Colombia, 1966 Average Total Cumulated Income Income Income Cumulated Cumulated Cumulated Income (thousands (millions Population (millions Population Income Population Class of pesos) of pesos) (thousands) of pesos) (thousands) (percent) (percent) x y ~~~~~~~~N AgricultureI 0.78 3,471 4,450 3,471 4,450 5.8 23.9 PersonalServices 1.05 2,732 2,613 6,203 7,063 10.3 38.0 Craft Manufacturing 1.27 1,696 1,337 7,899 8,400 13.1 45.2 AgricultureII 1.30 3,471 2,670 11,370 11,070 18.8 59.9 Conmerce 2.19 3,345 1,528 14,715 12,598 24.4 67.7 AgricultureIII 2.88 3,827 1,330 18,542 13,928 30.7 74.9 Construction 3.05 2,458 805 21,000 14,733 34.8 79.2 Mining 3.17 854 270 21,854 15,003 36.2 80.6 Transportationand Communication 4.31 2,959 687 24,813 15,690 41.1 84.3 ModernManufacturing 4.46 4,505 1,011 29,318 16,701 48.6 89.8 GovernmentServices 5.35 3,872 724 33,190 17,425 55.0 93.7 Finance 8.58 1,762 205 34,952 17,630 57.9 94.8 Utilities 9.86 465 47 35,417 17,677 58.7 95.0 and Entreprcneurial 26.7 24.943 926 60,360 18,603 100.0 100.0 Managerial TOTAL 3.24 60 ,360 18 1603 Note: For methodology, see text. Pesos are 1966 pesos. Source: Thorbecke and Sengupta (1972). Figure 10.4. Major Subsystems 6f the Ng Model of the Philippines Macroeconomic Demand Supply, price, Income subsystem subsystem and employment distribution subsystem subsystem Exognu variables 1' Macroeconomic model Iu matrix tt , ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~r /r X ~~~~~~~ ~ ~ ~_ ~~~~Sectoral Income . f ownership / . ~~~~~~~~~Sectoral gross output _ distribution f c apital Endogenous demand andland variables at . allocation current prices / > ~~~~~~~~~~~~~~~~~~~~~~Sectoral\ policy ( 0 production variabsfunctions Endogenous /in variables at ~~~~~~~~~~~Employme nt .4 and population . constant prices . 0determirrtion < . | ~~~~~~~~~Prices 230 Multipliers and SAM-Based Models variables as a function of two types of exogenous variables, the noncontrollable ones (including the initial values of all variables), and policy means under the control of the policymaker. The next step consists of deriving within the demand subsystem the various sectoral components of final demand consistent with the macroeconomic simulation. A particularly important part of this subsystem is the derivation of sectoral consumption. In the supply and price determi- nation subsystem the sectoral final demand vector is linked to the input-output framework by premultiplying it by the Leontief inverse to obtain the corresponding vector of gross sectoral production. Since the sectoral production functions are of the generalized Cobb-Douglas type, where output is expressed as a function of the intermediate and primary inputs, the labor income shares accruing to the various skill groups as well as the nonlabor (that is, capital and rent) value added are derived as fixed proportions. These income shares, which reflect a functional-skill distribution, are subsequently trans- mitted to the income distribution and employment subsystem. In determining employment, the supply of various occupational groups is related to educational attainments which are given exogenously, as is the pattern of ownership of capital and land. Labor supply by skill group is equated to the corresponding labor requirements after making some assumptions about the underemployment rate prevailing in each labor skill-sectoral class.13 Thus, the pattern of employment is generated within a relatively detailed cross classification by sector and by occu- pational group. This information is then used to obtain the equilibrium input prices (that is, labor income and wage rates of the occupational groups) and the net rate of return on capital and output prices by sectors given the sectoral production functions. Finally, in the income distribution subsystem, the household income (and consumption) distribution is derived indirectly from the endogenously determined earnings distribution of labor income (in terms of seventy-eight occupation-sector classes) and assumptions regarding the distribution of nonlabor income, which is itself influenced by the pattern of ownership of capital and land. In turn, the resulting household income distribution is used as an additional explanatory variable in estimating sectoral consumption in the demand subsystem. This process is continued until the complete system converges after a few iterations. 14 Figure 10.4 indicates that the starting point of the model is the demand subsystem, where the aggregate components of final demand are either determined by the macroeconomic model, such as aggregate consumption, or determined exogenously, such as exports. Furthermore, it is assumed that total consumption is (a) invariate to a change in income distribution and (b) distributed lognormally among households. In turn, the sectoral consumption expenditures of households are specified to depend on total consumption and the household distribution of the latter. Thus, by expressing the sectoral consumption functions as depending in a nonlinear way on the mean and variance of the lognormal household distribution of consumption, the demand subsystem can distinguish between commodities facing relatively high or low income elasticities (for example, luxuries or necessities). Thus, as in the Thorbecke-Sengupta model, no Cjkmatrix exists as such. Rather than specifying demand functions for each of the labor skill-sectoral household classes for each of the commodities, aggregate demand functions for each activity are expressed as functions of the mean and variance of the lognormal household income distribution. 15 Once all the components of sectoral final demand are obtained, the aggregate final demand 13. Since the overall model is broken down into thirteen activities (twelve productive sectors and government consumption) and six skill groups, there are, potentially, seventy-eight such labor skill-sectoral classes. In fact, some of these classes are empty. 14. It should be noted before completing this review of the model that it is only after factor and sectoral output prices have been derived that real GDP (and real values for all other variables) can be determined. 15. Since the structure of the Ng model incorporates the same transformations as the Thorbecke-Sengupta model- although in a more elaborate way and with a greater degree of closure-the reader is referred to table 10.7. PlanningModels Consistency-Type 231 vector is linked to the input-output framework to derive the corresponding vector of sectoral gross outputs. At this stage on,e enters the supply subsystem, which consists of twelve productive activities and government. It is assumed that the cost structure of each of these activities is a fixed proportion of the column totals. Since this is postulated to hold in current prices, the value shares of total input costs are constant. Thus, value added is a constant proportion of gross output in money terms, and the share of each factor of production-six types of labor, capital and rents-is a fixed proportion of value added in each industry. (Hence, the coefficient matrix V.J in table 10.7 is fixed once and for all.) This then corresponds to the assumption that output in each sector is determined by a Cobb-Douglas production function. Similarly, with respect to intermediate inputs (Lj) expenditure on each input is assumed to be a constant proportion of money costs. This implies unitary substitution elasticity between any pair of primary and intermediate inputs. Taking all these assumptions together implies that gross sectoral outputs are a generalized Cobb-Douglas function of all inputs for each produc- tion activity. This is an upper limit to the amount of substitutability likely to exist and is almost certairLly excessive. However, it does make modeling of the first twelve activity columns very simple. After determining gross output for each sector, the next step consists of generating the corresponding income distribution. The subsystem that yields income distribution through the operation of labor markets is the most interesting part of the Philippines model. The mecha- nisms incorporated in this subsystem generate information on employment, wages, and income distribution. The last, in turn, provides the measure of income inequality necessary to determine sectoral consumption. It is assumed that labor supply for the six occupational groups is determined by educational policies with two exceptions, namely, for occupational group 3 (farmers, farmworkers, fish- ermen, and so forth) and group 5 (manual workers) for which supply was taken as exogenously given. Given the supply-which in fact is a stock-of individuals in the six occupational groups, it is postulated that the economic system is forced to absorb this stock. This means that the total supply (or stock) of individuals in each occupational group must be absorbed across sectors requiring or demanding this group, and by the government subject to an exogenously estimated fraction of underutilized labor for each skill group. It is clear that in reality any assumed discrepancy between the labor force (or labor availability) on the supply side and the labor requirements on the demand side is adjusted through a combination of changes in (a) overt unemployment, (b) underemployment, and (c) the labor income rate per full man-year. However, it appears that in many developing countries the main adjustment mechanism takes the form of changes in the underem,ployment level and in the labor income rate rather than in that of overt unemployment. This is clearly the case of the Philippines where the unemployment rates for different occupational groups have remained fairly constant in the past. In the context of the model formulation, the assumption that numbers employed in each occupation follow directly from labor supply means that the adjustment of demand and supply is brought about by wage levels-since the underemployment rates are somewhat arbitrarily taken as constant. On the demand side, the Cobb-Douglas assumptions imply that the expen- diture of a production activity on each type of labor was a fixed proportion of the value of its gross output. Given the latter, the number of jobs offered for a particular type of labor can be derived by dividing the wage bill for that type of labor by the wage rate paid to it by that activity. More generally, given the v alue of gross output for each activity, the numbers employed in each occupational category by each activity can be determined if wage rates by skill and activity are known. Hence, total emplcyment for each skill will depend on the average wage for that skill and on relative differences in the wage paid to that particular skill by the different production activities. Of course, an alternative assumption to the effect that each occupational group is paid the 232 Multipliers Models and SAM-Based same wherever it is employed could have been made. However, the classification that has been used here, namely, (a) professionals, (b) sales and clerical workers, (c) farmers, fishermen, and farm workers, (d) semiskilled and skilled workers, (e) manual workers, and (f) service workers, is too broad to really reflect completely homogeneous skill groups. The specification of truly homogeneous skill groups from a labor market standpoint would have necessitated a much larger number of classes. But data limitations precluded this. Instead, it was assumed that for each skill group the labor income (wage) rates by sector are distorted by a constant factor to reflect the complex mix of skills and labor returns and wages within each of the six occupational categories selected in the model. Incidentally, this same assumption is made in the three second-generation models reviewed in the next section. The implicit assumption is that the subskills within a given occupational group coincide reasonably well with the sector employing that occupational group. In other words, an occupational-sector group is considered to approximate a specific homogeneous skill or subskill relatively well. At this point in the model, given assumptions about education policy and hence exogenous labor supply by occupational groups and given the value of gross outputs in each sector, then the wage rate and numbers employed by sector and occupation are determined. This information provides the basis for determining the distribution of income along the lines pioneered in the Colombia study as discussed in the previous section. However, in the present case there is much more detail available because the labor force in each sector is disaggregated by skill groups, that is, labor income is distributed over seventy-eight (6 x 13) groups corresponding to the six occupations and twelve production activities plus government, in addition to an entrepre- 6 neurial class.' It is possible to derive a rather detailed distribution of earnings for the employed on the basis of the distribution of labor income and nonlabor income among these occupational-sectoral classes and one entrepreneurial class. Under the assumption that each of these classes repre- sents a homogeneous socioeconomic group (that the variance of the individual incomes within each class is zero) the corresponding lognormal income distribution can be derived. A final point to note about this model is that while the assumptions made to determine wage rates may be questioned, the fact that wage rates are endogenously derived implies that there is a basis for generating prices throughout the system. The complete determination of prices requires an exogenous estimate of either the rental price of capital or of quantities of capital employed in each sector which can then be used to deflate the known profits figures to obtain estimates of the rental prices. These questions are treated in more detail in regard to the second- generation models. COMPARATIVE EVALUATION OF SECOND-GENERATION MODELS The three second-generation consistency-type models that are reviewed in this section are (a) the Adelman-Robinson (1978) computable general equilibrium model of the Republic of Korea, (b) the Lysy-Taylor (1977) computable general equilibrium (CGE) model of Brazil,' 7 and the Ahluwalia-Lysy (1979) model of Malaysia. Only the third of these models is based explicitly on a SAMframework, while the first two can be considered as using an implicit SAM. The second-generation models are much more ambitious than the first-generation models discussed in the previous section in at least three ways: (a) many of the behavioral and technical 16. In fact, because some of these groups are empty, the total number of sectoral-skill groups is less than seventy- eight. 17. Lance Taylor and a number of associates buJlt three different models which were applied to the Brazilian situation (see Taylor and others, 1980). The most relevant one from the standpoint of the study is the Lysy-Taylor CGEmodel which is presented in ohs. 6-9 of their book. Consistency-Type PlanningModels 233 relations appearing in these models are nonlinear; (b) the degree of closure of these models in depicting endogenously the interdependence among the various parts of the system is much greater; and (c) the level of disaggregation in terms of the number of classes or categories into which the accounts are subdivided is much more elaborate. In order to bring out the major similarities and differences among these models, the compar- ative evaluation is undertaken on the basis of the same format (links and characteristics) as in the previous section. Table 10.9 provides this comparative evaluation in synoptic form. It will be seen that these models incorporate some relationships or specification features that had already been adopted by the previous set of models. To this extent the detailed treatment of these features is not repeated. The Adelman-Robinson (1978) Model of Korea The Adelman-Robinson model is specified neither as a full neoclassical, general equilibrium model nor as a pure disequilibrium or partial equilibrium model. The economy is supposed to adjust from one set of conditions and institutional constraints to another in a kind of "lurching equilibrium" over time. Even though a number of neoclassical behavioral rules are postulated for firms, consumers, and other economic agents, there are certain elements of the model that are clearly not neoclassical, particularly with regard to the treatment of factor markets. The model is essentially short-run with a time horizon of a decade in which wages, prices, and income distribution are endogenously determined. The model is decomposed into three stages or submodels. Stage I describes the interaction between firms and financiaL markets in determining expenditures on investment goods. The second stage depicts the operation of factor and product markets subject to the investment constraints determined in stage I and various institutional rigidities and specifications affecting these markets. Finally, stage III introduces expectations and, in general, contains the dynamic part of the whole system. Stage II contains the heart of the interactive, general equilibrium part of the model. On the supply side, twenty-nine production activities are identified which are further disaggregated into four firm (or farm) sizes. On the demand side, households are disaggregated into fifteen different types according to socioeconomic criteria. The specification of the production side is based on alternative sectorEl production functions-either Cobb-Douglas or constant elasticity of substitution (CES)-and the assumption that firms maximize profits subject to a number of constraints which are discussed subsequently. In turn, on the demand side, consumption functions for each househcld group and for each type of commodity are specified. One of the distinguishing features of this model is that it incorporates monetary phenomena with the rate of inflation determined endogenously. To better analyze this mcdel and compare it with others, table 10.10 expresses the model in an implicit SAM framework. However, in contrast with tables 10.6 and 10.7, which express the Pyatt and Thorbecke-Sengupta models, respectively, the mappings (transformations) that appear in table 10.10 either as matrices or as vectors do not necessarily represent linear transfor- mations (that is, the elements of these transformations are not necessarily derived through fixed coefficients). In addition, to reflect the greater degree of endogeneity that exists in the Adelman-Robinson model, separate accounts are distinguished for companies, government, capi- tal, and the rest of the world, respectively. Since the model is inter(iependent, its description can start with any link in the system. A natural starting point is the analysis of the specification of the production structure and its mapping into a factorial income distribution.1 8 Before reviewing the form of the production 18. The sequence in which the major links are discussed in this section is the same as the one used with respect to the first-generation models. Thus, to help the comparative evaluation of all of these models, the format of table 10.9 is the same as that of table 10.4. Table 10.9. Second-Generation Consistency and Social Accounting Matrix-Type Models: Comparative Evaluation Links Lysy-Taylor Model Adelman-Ropinson Model Ahluwalia-Lysy Model and Characteristics Brazil (1977) Republic of Korea (1978) Malaysia (1977) Production Activities Value added originating in 25 Production activities are broken down into 29 Production activities are broken down into 15 I production activities is allocated to sectors following the input-output (1-0) sectors after consolidating a 60 sector 1-0 table. Factorial Income (a) 5 labor skill groups defined by classification. Fifteen different factorial groups Twelve different factorial groups are Distribution educational status; (b) family farm are distinguished: (a) 6 labor skill categories distinguished: 5 labor skill groups defined workers and sharecroppers, defined according to occupational and according to educational levels and also racially proprietors, and employers. Capital educational characteristics; (b) 2 groups of self- broken down. Self-employed are included in the income is allocated to proprietors, employed; (e) 5 agricultural groups--4 defined labor skill groups and receive their share of employers, and highly skilled according to farm size plus 1 group for laborers capital (nonlabor) income. Wage differentials for workers. The above distribution (presumably landless); (d) 1 group of the same labor skill can exist across sectors. yields approximately 130 different capitalists; and (e) government workers. For groups of income recipients. wage earners (by skill categories, including hired labor in agriculture) the model generates wages for about 500 groups distinguished according to skill, sector, and firm size. Factorial Income The above groups of income The household income distribution is mapped From the above total income accruing to 12 Distribution recipients are aggregated into 4 almost directly from the above 15 categories. factorial classes is generated. Through a I donsumer classes (1) rural income Thus, prototype households are defined household composition matrix allowing for Household Income recipients; (2) urban uneducated according to the occupation of the head of the households defined according to the skill level of Distribution workers, informal workers, and household for these 15 categories. Each the head to include other skills on a fractional self-employed; (3) urban household class can have other workers or basis, the income distribution in terms of 12 manufacturing workers with fractional workers, The distribution within each household types is derived. It is assumed that primary education, all workers household category is assumed to be lognormal. each type is totally homogenous with respect to with middle-level schooling, and Finally, the overall size distribution of income is income and consumption. recipients of capital incomes from derived from the 15 lognormal household agrieulture; and (4) highly skilled distributions. workers, proprietors in manufacturing, and employers from nonrural production sectors. Household Income Each of the 4 income classes is Each of the 15 household classes is assumed to For each household class savings and taxes were Distribution assumed to have expenditure have expenditure functions for the 29 sectoral deducted and transfers added to obtain total ! functions for the 25 sectoral goods goods and services. The form of these functions consumption. Seven consumption categories Household Expenditures and services and for noncompeting corresponds to Houthakker's addilog function, were specified corresponding essentially to on Wants imports. The form of these Disposable income by household class is wants (e.g., food, clothing, house and furniture, | consumption functions determined after deducting savings and allowing services). In effect, a multilevel structure of corresponds to Houthakker's direct for transfers and subsidies. An important consumer demands is specified. The expenditures Household Expenditures addilog system. It should be noted innovation is that because the monetary sector system used is a combination of the linear on Commodities and that the 4 income classes' is included in a nonneutral way, the demand for expenditure system and multilevel CES. At the Final Demand disposable income is computed cash balances by each class has to be subtracted highest level each consumer allocates his budget assuming different tax and from disposable income after savings to obtain among broad categories of consumption (wants) transfer rates. The other the remaining amount to be spent on goods and such as food. components of final demand are services. Other components of final denaind are Ae determined as follows, sectoral derived as follows: (a) investment is derived in a investment is assumed to be first-stage submodel based on expectations at the suballocated to sectoral products such as determined by differing sectoral level with regard to variables discussed agriculture and food processing. Finally at the expectations or "animal spirits"; below under Savings-Investment Behavior, and lowest level the consumer decides how to government demand and exports real financial factors; (b) imports are broken allocate his spending on the sectoral product- are assumed exogenously given. down into noncompetitive and competitive and say, food processing-between domestically derived through fixed coefficients; (c) exports produced and imported output. Thus, the whole are essentially exogenously determined (see expenditures system consists of 12 household below under Foreign Sector)u (d) government classes choosing among 7 wants to be satisfied purchases are aiso exogenousr by 15 sectoral products (either domestic or imported). Different levels of substitution elasticities are specified. Other components of final demand are essentially exogenously determined (i.e., investment in nominal terms and exports depending on world prices). (Table continues on the following page) U Table 10.9. (Continued) Links Lysy-Taylor Model Adelman-Robinson Model Ahluwalia-Lysy Model and Characteristics Brazil (1977) Republic of Korea (1978) Malaysia (1977) Household Expenditures A neoclassical specification of Different types of sectoral production functions Sectoral production functions are built up along on Commodities and production and cost functions is are postulated: (a) either Cobb-Douglas or two- multilevel CES lines. A given aggregate sectoral Final Demand adopted. Separate subproduction level CES neoclassical specification for output is produced through two (artificial) I functions are specified for manufacturing and agricultural sectors allowing aggregate components representing intermediate Production Activities aggregate labor (5 skill groups + substitution among labor skills at the first level, inputs and "value added." These two aggregates 2 agricultural groups), aggregate and between aggregate labor and capital at the combine in a CES way to produce sectoral capital, entrepreneurial inputs, second level; (b) essentially fixed capital-output output. At a lower level intermediate inputs and intermediate inputs. or labor-output ratios for other sectors such as combine again in a CES way to yield the Substitution is allowed among the services, construction, and utilities. aggregate and the different labor skills and different types of labor within that capital goods combine to determine aggregate aggregate and between aggregate labor and capital. Finally, at the lowest level any labor and capital. With regard to intermediate input is made up of domestic and capital, fixed proportions are imported components which substitute for one maintained among different capital another, and, likewise, any given capital good is types within each sector and made up of a domestic and an imported part. capital cannot be shifted out of a sector once it is installed. Finally, aggregate labor and capital inputs combine to produce value added according to a two-level CBS function. Price Formation Endogenous price determination. Endogenous price determination. Outside Endogenous price determination. Multilevel CES Commodity prices are determined agriculture, wage rates are determined through production functions are converted into as functions of factor prices. Prices the operation of labor markets for each skill corresponding price functions. (In equilibrium are solved iteratively. First it is level. Labor demand equations-derived from the costs are dual to quantities.) Given a value for assumed that levels of value added production functions-and labor supply all wages and for all sectoral outputs one can are temporarily given in each functions are specified and it is assumed that the solve for all other (e.g., commodity) prices. sector, which yields factorial markets are always cleared. Sectoral wage Prices of goods for intermediate and final uses income and consumption levels for differentials are postulated for given skills. The and for value added and output are determined the income groups. Along with agricultural labor market is treated separately in through the multilevel cost functions derived exogenously specified items of final that only agriculture is permitted to hire from the production functions. demand (e.g., investment and agricultural labor. Profits are assumed to be exports), consumption levels by maximized subject to technical constraints and a sector then give estimates of given capital stock. Product prices are sectoral output which in turn determined so as to clear markets. determine sectoral value added. Iteration on output levels is made to yield consistent prices. Nature of Production Sectoral production functions of Various forms are used: Cobb-Douglas and two- Multilevel sectoral production functions (see Function or the CES type are specified. The CES level CES in manufacturing, CES in agriculture, above). The treatment of imports among Relationshup specification is of the two-level and fixed coefficients in services, construction, intermediate inputs and capital is interesting. form (see description above). At and some other sectors. Fixed coefficients are The first branching of the multilevel CES the first level an aggregate labor assumed with respect to intermediate inputs and function is between value added and component is expressed as a CES substitution allowed among labor skills and intermediate inputs. Value added is composed of function of different labor inputs between aggregate labor and capital. In each aggregate labor and capital with the latter defimed according to skiUl and firm total investment expenditure is fixed in further broken down between domestic and educational levels. The aggregate nominal terms which can be aggregated by imported. Intermediate inputs, in turn, are capital component assumes capital goods and sectors through fixed combined among sectors and between essentially a fixed mix of capital coefficients. domestically produced and imported. goods at the semWral level. A. 'lie second level the aggregate labor component can be substituted for the aggregate capital component, again according to a CES function. Substitution Among The sectoral elasticities of Substitution among labor skills and between As seen above, substitution is possible among or Primary Inputs; substitution between different aggregate labor and capital in manufacturing between (a) intermediate inputs where a very Employment labor skills are taken as very high and agriculture. For other sectors, see above. low substitution elasticity is assumed (0.1); (b) Determination (between 4 and 7) while between Employment is determined at the skill level domestic and imported inputs where the sectoral labor and capital they are taken in through the determination of wage rates which range is from 0.1 to 1.5; (c) aggregate the range of 0.4 to 2.2. clear the segmented labor markets. Self- intermediate and value added where (T = 0.1; (d) Employment levels of the various employment in nonagriculture is assumed fixed aggregate capital and labor where the sectoral skiU groups by sector are except for the trade sector serving as an range is 0.4 to 1.5; (e) labor types where (J is determined along neoclassical employment sink. Migration is modeled as a assumed very high, 4-7. rwo alternative lines by equating exogenously function of urban-rural wage differentials with approaches are used to solve the whole model; projected labor supplies with labor migrants coming from the landless and first 2 (i) assume money wages are fixed and solve for demands. Hence, the model forces farm sizes and going into 3 urban labor labor demand, and (ii) equate labor demand to fuU employment of all the categories. exogenous labor supply. members of the economically active population. Nature of Consumption As previously indicated each of the Houthakker addilog form for 15 household As was indicated above, the demand structure is Funmtions 4 income classes is postulated to groups and 29 commodities (see also Household multilevel of the CES type formally analogous to have sectoral expenditures Expenditures). the form used on the production side. The functions of the addilog type. system can be viewed as resulting from consumers maximizing a special type of multilevel utility function, with strong assumptions postulated on pairwise substitutability of goods within and across levels. The introduction of the "want" concept is interesting. (Table continues on the following page) Table 10.9. (Continued) Links Lysy-Taylor Model Adelman-Robinson Model Ahluwalia-Lysy Model and Characteristics Brazil (1977) Republic of Korea (1978) Malaysia (1977) Foreign Sector Nonoompeting and competing Quantitative targets for exports are-fixed as a Imports are determined as follows: (a) final imports as well as exports are share of output for large sectors and in absolute consumption imports are derived from the given exogenously, as is the rate of terms for small exporting sectors. Two kinds of multilevel demand subsystem; (b) intermediate exchange. imports are distinguished: competitive and imports are derived from the multilevel noncompetitive. The first type is assumed to be a production system, as are capital imports. fixed proportion of domestic production while Exports are essentially assumed to respond to the latter are computed with import coefficient relative changes in prices of domestic and world matrices. The model distinguishes between two goods. types of traded goods: goods whose prices are determined in the domestic market or in the world market. Major Policy Means Simulation of changes in tax rates Rural policy means: land reform, changes in Static experiments were undertaken of the and Poliey and the exchange rate, changes in production and productivity, employment following types: (a) variations in the level of Experimentation labor supplies by education type, promotion, human resources (subsidization of exogenous components of aggregate (final) and modifications in profit consumption of poor, population control), anti demand, i.e., investment, exports and distributions and wage structures. overall strategy using all the above policies, government (demand); (b) government Some major conclusions: (a) more Urban policy means: human resources (same as intervention in key relative prices: devaluation, difficult substitution among labor above), employment promotion, technology, tax on labor, tax on capital; and (c) changes in types equalizes earnings and a nationalization, and overall urban development resource endowments of the economy (e.g., narrow wage structure; (b) strategy. Major conclusions of policy increasing the supply of higher skilled labor reduced investment demand leads experimentation are (a) size distribution of classes). These experiments are run under two to fall in price level, improving income is extremely stable: (b) comprehensive alternative model specification and solution distribution since the economy is development strategy combining structural algorithms: (a) Keynesian, assuming under less pressure to produce changes and other policy means is essential, and unemployment of labor and constant given savings; (c) changes in pattern of (c) agricultural terms of trade and rural-urban money wages and solving for the levels of investment can lead to significant migration are important instruments. employment; and (b) neoclassical, assuming improvements in distribution; (d) exogenously determined labor supplies and growth as well as equity solving endogenously for equilibrium wage rates. considerations lead to the conclusion that education programs should be directed to the illiterate. In general model not very responsive to policy shifts. The model is essentially A serious attempt is made at capturing some Model is essentially static or timel6ss at this Static-Dynamic comparative-static. dynamic elements. The overall model is stage. Only counterfactual simulations can be decomposed into three stages. Stage I is used to undertaken for given base year conditions. determine nominal investment on the basis of firms' expectations and accumulated earnings and credit availability. Stage II contains the basic part of the model and deals with the determination of wages, employment, prices and profits, and the demand for products and income distribution. Stage III updates variables and - formulates expectations which enter stage I. It is really the dynamic part which specifies the intertemporal linkages embracing such variables as population, technological change, and migration. Cambridge (England) specification Modeling of financial sector and influence of Introduction of consumer wants to be satisfied Special Features determination with money. Specification of different firm and farm by commodities. Multilevel CES demand and of investment investment depending on sizes. Incorporation of dynamic elements, production systems. Explicit linking of model to sectoral expectations and "animal spirits." SAM data framework. Linear savings functions are Savings functions are postulated for 15 Savings functions are built for 12 household Savings-Investment specified for the 4 income groups. household groups. Nominal investment is categories. Nominal investment is determined Behavior Investment is determined in an determined in nominal terms in stage 1 at firm exogenously and fixed capital matrix linking expectational way-depending on and sector level on the basis of expectations and capital by sectors of origin to sectors of "animal spirits" at the sectoral financial considerations. destination is assumed as well. level. In this specification it is investment which moves the system and income distribution has to adjust so as to be capable of generating the required savings. CES signifies constant elasticity of substitution. w' 1. 240 Multipliers and SAM-Based Models Table 10.10. A Schematic Representation of the Implicit SAMStructure of the Adelman-Robinson Model Expeaditur es Endog Insts. u 4) $ 0I cV o 1 w Od ZsC" I > ri E U 4 q &:) lo 0 X ~~~~~~~~~ ~~~ IJC 00 =4 ; ~~~~ 54,. 0 0a) as s 15 k=15 b = 4 J =29 0 A m= ) and value added (vj) which combine to produce sectoral output (xJ). These two aggregate inputs are in fact artificial CES indices which may not be observable in the real world. The aggregate intermediates index, in turn, is produced as a function of all intermediate inputs entering into the production of activity j. Finally, to complete the discussion of the left- hand branch in figure 10.5, each intermediate input used in the production of activity j consists of a combination of the corresponding domestic and imported inputs. Similarly, aggregate value added, on the right-hand side of figure 10.5, is a CES function of aggregate labor (f,) and aggregate capital. At the lowest level of this CES cascade, aggregate labor is produced as a function of different labor skills, and aggregate capital consists of different capital goods. Different degrees of substitutability among inputs are assumed.2 6 The value added, generated by the fourteen activities, accrues to twelve different factorial 26. Ihus, a very low elasticity of substitution is assumed among intermediate inputs and between the intermediates and value added aggregates (u = 0.1). A much higher substitution elasticity is assumed to prevail between aggregate capital and labor where, depending on the sector, the selected range is from 0.4 to 1.8 and among labor skills where the range is from 4 to 7. 248 Multipliers and SAM-Based Models Table 10.12. A Schematic Representation of the Explicit SAM Structure of the Ahluwalia-Lysy Model Expenditures 5. I-n i-q- U) 1 1 o :W;uc ) g: t X 00 r 0 o0s.0 4r $3 4s) 12 k=-Id12 b= 2d w- CQ 1 0 - 4O . 0,+ )-i~ +o dC) 0 n OH, s k2 b=2 4 1 w7 c1 j =l. 14~, o .C 1 2 3 4 5 6 7 8 9 10 1112 Factors 1 vsj xs vs c holds ks kb k xk k o Companies 3F v Wants 4 wk w 02 ___ _ _ _ _ 1 Cozmmwdities C L k z x c 5 wcj] 9C C C C C Production 6 ._ Activities 6 Djc P Government 7 dk - 1b Capital 8 _k Sb ; _b s Indirect 9 _ ic _____ kCommodities')10 e5va- ____ 1116 i1' mc _ic I X ny 2 counts 11 S k _1' r |y Total 2 v hk vb w c P z Y Note: The major difference between the format of the present table and that of tables 10.10 and 10.11 is that two additional accounts have been added representing wants and commodities. The numbering of the accounts is different from tables 10.10 and 10.11 because of the inclusion of these two new acco,unts. For definitions of symbols, see table 10.10. The new transformations and corresponding symbols which have been added are: W,k = expenditure of household k on want w; k = 12; w = 7 Ccw = satisfaction of want w by commodity c; c 14 D,, = demand for commodity c satisfied by domestic production activity j; j = 15 (the sectoral breakdown of c and J is identical) i = indirect taxes paid by commodity c to government g = total indirect taxes accruing as revenues to government. PlanningModels Consistency-Type 249 Figure 10.5. MuLtilevel Production Structure of the Ahluwalia-Lysy Model of Malaysia Sectoral output i, v,~~~~~~X Aggregate Aggregate intermediates value added 1lj tNs 1, ~~~~~~~~~~~~~~~~~k, Intermediate inputs into j Aggregate Aggregate X gX ~~~~~~~~~labor capital i M . . . idNj iy I' '' iNmi IIJ . .* lnj kii. k,, Domestic and imported inputs Labor skills Capital goods of i into ;j groups, consisting of five labor skill groups defined according to educational levels and one class of proprietors. Each of these classes is, in turn, broken down on racial lines between Malay and non-Malay. This mapping from production activities to factors is represented in table 10.12 as V 61 which yields the factorial income distribution (v8 ). In the next link the household income distribution is derived from the factorial distribution through the allocation of labor income, FkS,and property income, Pkb.It is assumed that there are twelve different household income groups corresponding to the twelve factorial categories. As in the Adelman-Robinson specification, a household composition matrix is used to allow household classes, defined according to the skil level of the head, to include other skills on a fractional basis. It is assumed that each household within a type is totally homogeneous with respect to income and consumption. The income distribution among households (hk) is deter- mined after government transfers (g.) have been added to factorial income. It is in the next link that this particular model introduces a conceptual innovation by mapping household consumption with respect to wants rather than with respect to production activities or commodities directly. This can be seen best by looking at the column of the household account (2) in table 10.12. For each household class savings (sk) and taxes (d{) are deducted from total household expenditures (h{) and the remainder is spent on wants (the matrix WWk). 27 Seven categories of wants are specified, such as food, clothing, housing and furniture, and services. 27. Linear functions are postualated for savings and taxes for each household category. 250 Multipliers and SAM-Based Models In fact, the structure of consumer demand was specified in a multilevel fashion somewhat analogous to its specification on the production side. The expenditure system which is postulated is a combination of the linear expenditure system and multilevel CES.At the highest level, each of the twelve types of consumers allocates his budget among broad categories of consumption (wants) such as food. At the next level a mapping takes place from wants to commodities capable of satisfying these wants (CO,). Thus, the seven wants are suballocated to fourteen commodity groups, which are broken down according to the same sectoral classification scheme as production activities. For example, the want for food, that is, the food budget, is suballocated to sectoral products such as agriculture and food processing. This last transformation from wants to commodities does not distinguish, at this stage, between imported and domestic commodities. Finally, at the lowest level of the multilevel demand subsystem the consumer decides how to allocate his total commodity demand between domestic production and imports. This process is represented by reading down the columns of account 5 of table 10.12 where DJC stands for the demand of commodity c to be satisfied by domestic output activity j and mC' stands for the part of total commodity demand to be imported (the competing imports). In conclusion, the whole expenditure system consists of twelve household classes choosing among seven wants to be satisfied by fourteen sectoral products, either domestic or imported. This multilevel CES demand system postulates different levels of substitution elasticities. The treatment of foreign trade is relatively straightforward. Final consumption imports are presumably competing imports (mc') and are derived from the multilevel demand subsystem as noted previously. Intermediate imports (In° ), which are typically noncompeting, and capital inputs are derived as indicated previously through the multilevel production system. Exports (zc), in contrast, are essentially assumed to respond to relative changes in prices of domestic to world goods and, as such, can be taken as exogenously given. In summary the flow diagram of the model presented in figure 10.2 clearly reveals a pen- tagonal structure. The distinction between production activities and commodities can be justified in this specific case only by the fact that commodity demand can be satisfied by either domestic production activities or competing imports. Otherwise the use of an identical classification scheme for activities and commodities would render this distinction meaningless. 2 8 As in the two previously reviewed models, the process of determining prices endogenously is a crucial feature of the present model. The multilevel sectoral CES production functions are converted into corresponding price functions, since in equilibrium costs are duals of quantities. Thus, the multilevel production cascade given in figure 10.5 can be translated into a corre- sponding set of cost functions. An elaborate and rather complex algorithm is designed to solve for a consistent set of prices. Of particular interest is the treatment of labor markets and wage determination. Two alternative approaches are postulated, a Keynesian approach and a neoclassical one. Under the first approach money wages are assumed to be fixed, and the model is solved for labor demand by skill groups; under the alternative neoclassical specification, labor demand is equated to exogenous labor supply. As in the Lysy-Taylor model, nominal investment is determined exogenously (presumably influenced by "animal spirits"). A fixed-capital matrix linking capital by sectors of origin to sectors of destination is postulated as well. A number of static laboratory experiments were undertaken on the model. The most important exogenous policy changes that were introduced are: (a) variations in the level of demand for investment, exports, and government expenditures; (b) government intervention in affecting key relative prices such as devaluation, taxes on labor, and taxes on capital; and (c) changes 28. Of course, the conceptual distinction between activities and commodities can be meaningful when the classifi- cation schemes are different and the number of commodities is different from that of activities. PlanningModels Consistency-Type 251 in resource endowments of the economy and, in particular, increasing the supply of higher skilled labor classes. These experiments were run under two alternative model specifications and solution algorithms. The first specification, as previously indicated, assumes unemployment of labor and constant money iwages. Given these money wages the model can be solved for the level of employment by skill categories, which is determined by the demand for these skills at the prevailing wage rates. Under the alternative neoclassical specification, labor supply by skills is exogenously determined and the model is used to solve endogenously for the corresponding equilibrinum wage rate. CONCLUDING REMARKS AND SUGGESTIONS On the basis of the preceding comparative evaluation of development planning models based implicitly or explicitly on a SAM framework, a number of suggestions can be made to modify and improve the organizational framework of the SAM and the specification and design of corresponding planning models. These suggestions fall into five categories: (a) the treatment of structural changes and reforms particularly as they relate to changes in asset distribution (such as land reform) and in the resulting factorial and household income distributions; (b) the treatment of the agricultural and informal sectors, particularly with respect to the appro- priate level of disaggregation and the classification criteria used to break down production activities; (c) the incorporation of some regional dimensions; (d) the treatment of basic needs and the measurement and identification of poverty by socioeconomic household classes; (e) the limitations of using neoclassical or Keynesian equilibrium assumptions in closing these models through market clearing mechanisms at the product and factor levels, and the scope for an alternative disequilibrium approach. It will be seen in the subsequent analysis that these five issues are, at least to some extent, interrelated. Changes in Asset Distribution A universal finding shared by the six models reviewed in this chapter, as well as by other such models, is that the size distribution of income remains stable even under significant policy changes. The relative degree of poverty as a whole is much less affected than the composition of the poor. There appear to be two major reasons for this outcome. First, the great majority of the policy experiments and simulation.s undertaken in these models rely on rather conventional packages of policy instruments instead of structural changes or reforms, to use Tinbergen's terminology. Second, the household income distribution is derived rather mechanically from the factorial (functional) distribution. Policy instruments consist of marginal changes in the quantitative magnitude of means under the control of the govern=ent within a structural and institutional setting that assumes as specifically given the asset dlistribution among households. In contrast, structural changes and reforms entail fundamenta]. changes in the underlying institutional setting, including changes in the asset (wealth) distribution. Three different types of wealth might fruitfully be distin- guished: (a) human capital, in the form of educational and skill levels possessed by individuals; (b) the land tenure system, including property rights or tenure arrangements with respect to land; and (c) the ownership of and access to other forms of capital. It is only with respect to humarL capital (changes in the composition of skills among households) and to a very limited extent land tenure (the simiulation of a moderate land reform in the Adelman-Robinson model) that the second-generation models incorporated changes in the asset distribution of their policy simulations. The major policy experiments undertaken in these models are listed in table 10.9. It was seen in the preceding section that both the Adelman-Robinson and Ahluwalia-Lysy 252 and SAM-Based Multipliers Models models used a household composition matrix to map the factorial income distribution into a household income distribution and to reflect exogenous changes in the educational and skill levels possessed by different household categories. To some extent, the household composition matrix is a device for incorporating the wealth distribution by household category explicitly as a major element predetermining the household income distribution. This matrix determines the composition of each household category in terms of the various factorial classes, where the classes could include, at least conceptually, different labor skil groups and different groups of proprietors and farmers defined according to farm size and land ownership. As such, the classes can take into account implicitly, if not explicitly, the differential skill levels possessed by house- holds as well as the average land and capital available to individuals in each of the factorial and household groups. Thus, this formulation is an interesting first attempt at incorporating the wealth or asset distribution into a planning model. However, to the extent that the wealth distribution is a crucial determinant of the household and size income distributions, it appears essential to incorporate explicitly the household wealth distribution as one account within the SAM. This could be done, for instance, by adding an account for domestic factor endowments as indicated previously in connection with table 10.2. The corresponding matrix (T6 9 in table 10.2) would extend the coverage of the household composition matrix to land and capital ownership in addition to skills. The inclusion of the initial distribution (in terms of educational skills, land, and capital) among household groups in a SAM data system and the conceptual treatment of the effects of changes in the asset distribution on the structure of production and technology and thereby on the factorial and household income distribution would greatly enhance the operational usefulness of the SAM approach. Thus, for example, a land redistribution may have an important impact on the cropping pattern and the technology that is adopted. A land reform process that would entail the breakdown of large holdings previously cultivated by tenants and landless workers into small private farms might well trigger a technological shift away from mecha- nization and tractorization to the use of labor-intensive techniques in the production of the same crops. It might also lead, to some extent, to a shift in the composition of output away from relatively capital-intensive crops to labor-intensive products. It is clear that such a process of land redistribution would have a major impact on the factorial income distribution, as rent would accrue to the new small landowners who were formerly tenants and hired workers. In turn, the size distribution of income may be affected in a signif- icant way through, for instance, the reduction or disappearance of landless workers and tenants. In summary, the incorporation of an initial domestic factor endowments account and the design of the link between the asset distribution and the resulting factorial and household income distributions are crucial elements to make the SAM more useful both as a data system and as a conceptual framework. Treatment of the Agricultural Sector and Informal Activities A second important issue is the inadequate treatment of the agricultural sector and, in a more general sense, of traditional and informal activities inside and outside of agriculture. None of the six models breaks down agricultural production activities into more than two sectors-the most disaggregated treatment is that of Adelman and Robinson, who divided the two agricultural sectors into four farm sizes. It can be argued that for a number of reasons this consolidation of all agricultural production into only two activities is, at best, inappropriate and, at worst, an important source of bias through the process of averaging out very different agricultural income classes. The point is that the standard of living of agricultural households is largely determined by their land tenure Consistency-TypePlanning Models 253 status (whether they are small or large landowners, tenants, or landless workers hiring them- selves ouit), their cropping pattern, and the technologies adopted. The cropping pattern, in turn, is related to regional agroecological factors. It seems essential that models explaining the determination of income distribution should attempt to capture the diversity among all three of these elements. This differential production pattern and its implications with respect to income distribution can be approximated by an appropriate classification of agricultural production activities. It is suggested that these activities be broken down according to the following criteria: (a) the nature of the crops and agricultural products; (b) the type of technology used (for example, traditional and almost totally labor-intensive methods, intermediate methods relying on chemical and biological inputs such as fertilizer and high-yielding varieties of seed, and capital-intensive methods relying on mechanical energy such as tractors); (c) the form of organization and the tenure system to reflect, for instance, the impact of different farm holdings by size on the incomes accruing to the cuLtivators; and (d) regional variables such as soil type. These criteria tend to be significantly interrelated in the real world. Thus, it is typical for large farmers to produce export crops under relatively capital-intensive technologies in specific regions. In contrast, small farm owners often produce a combination of food and cash crops using a traditional labor-intensive technology, or alternatively, an intermediate technology in different regions. The fact that these criteria are often systematically interrelated makes it easier to design an operationally useful classification scheme of agricultural production activ- ities. Furthermore, there exists a wealth of information in many developing countries on the determinants of agriculturaL production from such varied sources as farm management studies, farm surveys, and integrated rural surveys. A concrete problem facing builders of SAMs and SAM-type models is the treatment of agri- culture in input-output tabcles. It is either too consolidated in a small number of subsectors or the classification scheme is based on the commodity structure alone rather than on other relevant criteria such as type of technology, form of organization, and tenure system. In such cases, an attempt should bee made to combine various sources of information in order to improve the breakdown of agricultuiral activities. Just as it is desirable to classify agricultural production activities according to technology and form of organization in addition to the nature of the product, such a distinction should be attempted outside of agriculture as well. More specifically, goods and services produced under informal, small-scale, labcr-intensive conditions should be distinguished from those produced under formal, large-scale, capital-intensive conditions. The types of production functions under- lying modern manufacturing and formal services are totally different from those underlying workshop-type output and informal services. Likewise they yield very different flows of value added and resulting factorial income streams. Again, a shortcoming of the models is that they did not introduce this distinction, relying instead on input-output information. In addition to distinguishing between formal and informal activities in the urban areas, a serious attempt should be made to model the rural nonagricultural activities separately from, say, the informal urban activities. Underlying this recommendation is the fact that a large number of rural households derive their incomes from a combination of employment in agri- cultural and nonagricultaral activities. Given the highly seasonal variability in agricultural employment, off-farm aclivities provide a means to absorb what would otherwise have been seasonal underemployment on the part of the agricultural labor force into additional effective employment in a seasonally complementary way. The effects of increasing effective employment of the rural labor force through the design of appropriate rural nonagricultural activities, such as construction and public work projects, can contribute significantly to raising the incomes of certain household groups such as landless workers. In conclusion, the type of breakdown suggested here with respect to agricultural and informal 254 and SAM-Based Multipliers Models activities outside of agriculture appears to be essential in capturing and reflecting the underlying dualism that prevails in most developing countries. A Regional Dimension Another feature that has not yet been systematically incorporated in the planning models is a regional breakdown. In many instances differences in the output mix and the pattern of production in agriculture can be explained by interregional differences with respect to such variables as agroecological factors, soil fertility, the extent of irrigation and rainfall, and the amount of social overhead capital in support of agricultural production. Thus, to the extent that differences in living standards among different groups of households are significantly influenced by regional elements, it is important to include this regional dimension in a SAM. Another important advantage of the explicit inclusion of the regional dimension into a SAM conceptual framework is that a large number of policy means tend to be location-specific. These may include investment projects; current government expenditures on services such as credit, extension, health, and education; and price policies with respect to commodities and inputs at least to the extent that the production of specific commodities is regionally concentrated. Whereas in principle there are compelling reasons for adding a regional dimension to the SAM, in practice there are serious difficulties on the data side. Clearly, in the absence of a set of interregional social accounts, which are not available on a systematic basis in any country (developed or developing), production activities and interregional trade cannot be expressed in the form of an integrated set of interregional input-output tables. Given the data limitations there are two alternative procedures which suggest themselves and which appear to be feasible within the context of many developing countries. The first procedure would be to divide the economy into a very small number of regions and divide production activities into those that are typically national and those that are more regional in character. National activities would be those characterized by economies of scale and reasonably low transportation cost and would in many instances be produced in the modern sector or, in other words, the industrialized region which often coincides with the larger urban centers embracing the capital city. Thus, in this type of specitication the regional dimension would be introduced only with regard to the production accounts which, in turn, would be broken down into national production activities and a small number of regional production activities. A second more modest and perhaps more realistic alternative procedure for capturing some of the regional effects of production consists of breaking down the factor and household accounts along regional lines without using a regional classification of production activities. Hence, the mapping from production activities to factors of production (V., in table 10.11, for example) would allocate the value added generated by the various national production activities to factors where at least some of the factors were defined on a regional basis. A specific example might be the allocation of value added generated by agricultural production activities to regional groups of smallholders. Likewise, the mapping from the factorial to the household income distribution, particularly if it was based upon a household composition matrix, would yield a regional income distribution for at least some household groups. This type of information can be quite valuable in those instances where there exist marked interregional income differences which can be accounted for by differences in the pattern of interregional production-particularly in agriculture. In Kenya, for example, the findings of a large-scale integrated rural survey suggest that very significant interregional income differ- ences prevail among smallholders. Both these interregional and intraregional differences in the standard of living among different classes of smallholders can to a large extent be explained by the pattern of production. Two studies by the author have demonstrated the possibility of an approximate mapping from the structure of production to the factorial and household income distribution (see Crawford and Thorbecke, 1978). PlanningModels Consistency-Type 255 Basic Needs The standard of living of different individuals and households is to a large extent related to their income level or, better, their total consumption levels. Yet it is clear that a scalar concept such as the value of total household consumption expenditures, even when all government services are imputed, reflects only imperfectly the degree to which the basic needs of a household are met. There are at least two reasons why this should be so. First, total consumption may be allocated very unevenly among the goods and services which fulfill the various private needs such as nutrition, shelter, and clothing, and public needs such as education and health. Consequently, it is conceivable that some basic needs might be over- fulfilled, whereas others might be only inadequately met. This imbalance is all the more likely with respect to the public needs, because the imputed consumption of these services depends largely on current and capital expenditures by the government and as such is outside the direct control of any given household. A second reason why total household consumption is an inadequate measure of a household's physical well-being is that well-being can only be expressed meaningfully in terms of physical or quantitative units and not in value terms. Indeed, the adequacy of a diet from a nutritional standpoint needs to be expressed in terms of such indicators as number of calories or grams of protein per day per adult equivalent, while the adequacy of shelter might be expressed as a function of the number of square meters of shelter per adult equivalent and the quality of building material. Likewise, the adequacy of education is reflected by such criteria as the student- teacher ratio, the level of education of teachers, and the average distance from residence to school; health services may be measured by the ratio of doctors to inhabitants, the distance to the nearest health dispensary or hospital, the frequency of diseases, and life expectancy. Since prices differ significantly from one region to another and, in some instances, among different household groups within the same region, any serious attempt at identifying pockets of poverty should include for each of the distinct household categories (which are likely to represent distinct socioeconomic groLups),physical or quantitative measures of the degree of satisfaction for each of the basic needs. In this fashion it should be possible to obtain a profile of need satisfaction for average house- holds, in each of the relatively homogeneous socioeconomic categories. This type of information can be incorporated into a SAM through the wants accounts, as long as the classification of wants embraces the major needs. (See tables 10.2 and 10.12 for examples of SAMs that contain wants as a separate account.) However, it should be recalled that the wants account in a SAM data system contains entries expressed in value terms and not in physical terms. Hence, in the light of the previous discussion, these value terms for the subset of wants consisting of basic needs would have to be coinverted into physical units by using the appropriate prices faced by each category of households. This means, for example, that the matrix T, 3 in table 10.2 (the intersection of the row of wants and the column of households), or alternatively, the matrix W.k in table 10.12 would need to be expressed in physical terms, at least for that part of the matrix representing basic needs. In order to judge the well-being of each household category with regard to such needs, the prevailing conditions as given, for instance, in an initial SAM would have to be gauged against some threshold levels considered minimal with regard to some normative definition of poverty. The incorporation of a way to measure the degree of satisfaction of needs by household categories in a SAM should not imply an endorsement of a so-called basic needs strategy in contrast, for instance, to a strategy of growth with redistribution or asset redistribution with growth. Rather, the reason for incorporating basic needs in a SAM is, first, for diagnostic purposes so as to identify poverty groups. Second, the effects of different development strategies on need satisfaction and other policy objectives can be simulated within a SAM-type conceptual model if the various tranEformations discussed above are explicitly incorporated in the model. 256 and SAM-Based Multipliers Models Equilibrium and Disequilibrium Specifications The six models that have been reviewed relied on either Keynesian or neoclassical specifi- cations upon which, in some cases, institutional constraints were superimposed. Crucial to both of these types of specifications is the concept of equilibrium and maximizing behavior on the part of consumers and producers. Equilibrium in the markets for products and factors requires that markets are totally cleared, that is, that all excess demands are zero. It could be argued that, in fact, equilibrium in that sense is never reached and excess demand, or supply, always prevails. More exactly, the adjustment mechanism that moves the system from one time period to another could be expressed as a function of the prevailing disequilibrium situation. Thus, for example, excess supply of a given commodity in period t would be reflected by increases in stocks and a tendency for prices to fall to a new level in the next period. This fall in price could be related to the rate of increase in stocks. Any change in this rate or reversal would be translated into corresponding price movements. In this sense the system would move from one disequilibrium situation to another over time. A disequilibrium specification appears to reflect best the operation of labor markets. It has already been argued-in the discussion of the Ng model-that any prevailing (or assumed) discrepancy between the labor force (or labor availability), on the supply side, and labor require- ments, on the demand side, tends to be adjusted in developing countries through a combination of changes in: (a) overt unemployment, (b) underemployment, and (c) the wage rate (the hourly or daily rate). For a number of skills, such as farm workers and unskilled workers in nonagricultural activities, the labor markets operate in such a way that (a) overt unemployment tends to be very low and stable and (b) the wage rate remains very stable over a long development stage characterized by labor surplus. Consequently, the burden of the adjustment process falls on the extent of underemployment, that is, the effective employment rate per member of the economically active population. This process has been verified empirically in a number of developing countries. Thus, it has been demonstrated that in agriculture the number of man-days of work per year per member of the labor force tended to rise very significantly over an extended period before the labor shortage point was reached, and to be reflected in an increase in the wage rate.2 9 Neither the Keynesian nor the neoclassical specifications which were adopted to reflect the workings of labor markets in the models reviewed describe the above adjustment process accu- rately. It is, of course, unfair to criticize models for using the only specifications that are founded on existing deductive theories. It may not be unfair, however, to suggest that certain processes may be empirically tested and explained through inductive methods rather than derived deduc- tively from an existing body of theory that does not fit the underlying relationships. It seems that SAM-type models could and should experiment with new specifications which can be empirically tested and, it is hoped, confirmed. The underlying dynamic processes might be tracked more realistically by a recursive (period to period) adjustment mechanism that would not impose a market-clearing condition but would explain empirically changes in prices and levels of inputs and outputs as functions of the tensions which prevail in each period. 30 Clearly, such a disequilibrium approach must be based on a detailed and comprehensive data system to generate and test the empirical relationships among variables. The SAM can help provide this data set upon which more empirical specifi- cations might be built. 29. In Taiwan, for example, the number of man-days of work per year per worker rose from slightly above 100 in 1946 to about 200 in the late sixties at which time the agricultural wage rate started moving up. 30. The Adelman-Robinson "lurching equilibrium" concept described previously is, perhaps, a step In this direction. Nevertheless the underlying specification of product and factor markets is stil essentially neoclassical in their model. Social Cost-BenefitAnalysis in a Semi-Input-OutputFramework: An Application to the Muda IrrigationProject Clive Bell and Shantayanan Devarajan Investment projects affect the incomes of households, firms, and government, not only directly through the value added produced by the projects themselves, but also by inducing additional output through interindustry linkages and the expenditures out of the extra incomes accruing to their beneficiaries. The latter, which are sometimes called the "multiplier" or "downstream" effects of a project, have been discussed in some of the recent literature on social cost-benefit analysis (see Scott, 1976), which has been concerned with the derivation of shadow prices which capture all such effects in full. If these shadow prices are correctly calculated, so it is asserted, then valuing a project's direct inputs and outputs at these prices yields the right measure of its social profitability. This approach is in the spirit of, and consistent with, that of the various manuals on social cost-benefit analysis (UINIDO, 1972; Little and Mirrlees, 1974; Squire and van der Tak, 1975). This paper also deals vrith the problem of allowing fully for multiplier effects in project evaluation, but approaches it on a different tack. Here, the total (direct and indirect) impact of a project on outputs, incomes, savings, and expenditure, which are the elements entering into the calculation of social profits, is derived using a linear model of the economy which is both an extension and an elaboration of Tinbergen's (1966) semi-input-output framework. In other words, we follow the "primnal" route of quantity and income flows rather than the "dual" route of shadow prices. This use of the notion of primal and dual is more than suggestive, for there is an intimate connection between the model developed in this paper and that which underlies the manuals. Although this connection will not be analyzed In detail here, the concepts and formulations that the two approaches share will be discussed as they appear.1 To iLLustrate our methcd, we apply it to an evaluation of an existing large-scale irrigation project in northwest Malaysia, the Muda River scheme. The main reason for the choice of this particular project is that a semi-input-output model of the project's impact on the Muda region already exists (Bell and Hazell, 1980) and there is an excellent data base in the form of a SAM for the Muda region (BelL,Hazell, and Slade, 1982, ch. 5). The general model is set out in the next section. Following sections deal with the specific application to the analysis of the Muda irrigation project, with the estimation of certain national parameters that are indispensable for the calculation of soci.al profitability. Note: An earlier version of this paper appeared in Pakistan Development Review (Bell and Devarajan, 1979) and, with further revision, as chapter 8 of Bell, Hazell, and Slade (1982). 1. Wanhill (1974) and Kuyvenhoven (1976) have tried to synthesize the Little-Mirrlees and semi-input-output approaches without alluding to the primal-dual distinction. In another paper (Bell and Devarajan, 1980), we show why their approach may have dleparted from the spirit of Little and Mirrlees. The only other empirical application of a similar nature known to us is by 9eton (1973), who attempts to calculate shadow prices for Chile, allowing for output- and expenditure-linkage effects. Unfortunately, as noted in Bell and Devarajan, Seton's otherwise commendable paper is marred by a theoretical flaw 257 258 Multipliers and SAM-Based Models THE MODEL The model is a variant of Tinbergen's (1966) semi-input-output framework. At this point, the model is best thought of as one of a national economy, although its application to a regional context is discussed in the next section. The M production activities in the economy are divided into two groups: those producing tradables and those producing nontradables. 2 The former have parametrically fixed output levels, that is, they operate at full capacity, so that variations in demand are met entirely by changes in net exports. Changes in demand for nontradables, however, are mnetsolely by adjustments in domestic output. The latter sectors either have excess capacity, in which case they are producing at constant marginal costs, or they add just enough new plant and equipment to satisfy demand, in which case production takes place at constant average costs. It is also important to note that this classification of sectors is not necessarily a statement about whether the commodities they produce are, in fact, traded. Rather, the distinc- tion is whether changes in domestic demand are met by changes in net exports or domestic output. To keep the national income accounts consistent, it is necessary to introduce a third type of sector-for distribution and transportation services-since export activities, both exog- enous and endogenous alike, purchase those services domestically. Material Balances The subscripts T, D, and N denote the set of tradable, distribution/transport, and nontradable sectors, respectively. The rows and columns of the fixed coefficient technology matrix, A, are ordered so that it can be partitioned as follows: ATT ATD ATN1 A = T DD ADN WANT AND ANN_ The submatrix AT,, for example, represents the inputs of tradables required by the nontradable sectors when the latter are operating at unit activity levels. The material balance equations for the economy are: (11.1) XT = AM-XT +ATDXD +ATXN + CT + GTT + JT + ET (11.2) XD = ADTYX + ADDXD + ADNXN + CD + GD + JD + ItE (11.3) XN = ANTXT + ANDXD + ANNXN +CN + GN + JN + EN, where X = [XT, XD, XN3are gross outputs. C = [CT, CD, ON]are consumption demands. G = [GT, GD, GN] are government demands. J = [JT, JD, JN] are private investment demands. E = [ET, ED, EN] are net exports demands. 2. Tinbergen calls these two types of sectors "international" and "national." To highlight the relationship of our approach to that of Little and Mirrlees, we have adopted their terminology. Semi-Input-Output The Muda Project Framework: 259 Note that ED = 4LET,EN], where p. is the matrix of trade and transport margins on net commodity exports accruing to the domestic economy.3 Now equations (11.1) through (11.3) are expressed in physical units. By choosing units of physical measure such that all domestic prices are unity, the same equations would result if everything were measurecl in value terms, provided the user price of each commodity (net of transport and distribution. margins, which may vary across users) is the same for all users. Incomes The vector of incomes Y = [Y1 , e YL] earned by the L household types in the economy is , determined by direct earnings in production and distributed profits. Let v. and wj be the direct earnings and dividend payments, respectively, accruing to the kth household type for each unit of output produced by industry j. Then, (11.4) Y = (V + Q)X, Q = IllwJ where V = N~vjIl, Consulmption The gross income of household class k is used to pay lump sum taxes, Tk, and income taxes, TkYk; a constant share, Sk, of posttax income is saved. A fixed proportion of its marginal income is also spent on the consamption of each good. These proportions are given by the matrix B, where Pik is household k's marginal propensity to consume good i. Denoting the intercept terms of these consumption furnitions by -yk,and noting that domestic prices are normalized to unity, we have (11.5) C = ru + B(I - s)[(I - r)Y- where F = (I - s) and (I -T) 11yikll; are diagonal matrices whose kkth elements are 1 - sk and 1 - T k, respectively; T = [T1, .. TL];and u is a column vector of ones. Government Government derives its revenue from direct taxes, tariff collections, and ownership of state corporations. In addition to income taxes, the government collects corporate taxes, which are denoted by Tt4'X, where T. is the corporate tax rate and 4' is the vector of profit margins on gross outputs. Letting 0,, be the vector of profit margins on state-owned corporations, the government's receipts from these enterprises are o'X. Finally, let P1 be the ratio of the world price to the domestic producer price of good i, where the vector P is exogenously given. Then government revenue from tariff collections is [P' - (u' + -')]E, where u' is a row vector of ones and jI is the vector of total trade and transport margins on net exports which accrue to domestic producers. 4 3. As these margins are earned on gross rather than net flows, the above formulation is strictly correct only when there are no competitive imports of characteristic commodities which are exported. In the present application, that is a fair approximation to the observed trade patterns. The transport and distribution margins on noncompetitive intermediate and consumer goods are included in ADT, Al)) A,),, and CD,respectively. Noncompetitive imports may be placed in category T, the corresponding element of XTbeing zero. 4. When the sum of the trade and distribution margins (jI) on net exports of a commodity accrue to domestic producers, the extariff fob. (firee on board) price of the exports is (1 + jIg). 260 and SAM-Based Multipliers Models We now assume that the government spends exactly what is received in revenues. 5 The bundle (G) of government purchases is given by: a vector (C.) of obligatory purchases for current consumption (education, health, and so forth), a vector (K) of exogenous outlays on investment goods (project-related expenditure), and a residual whose composition is fixed. The latter is given by the vector 4), where 4i is the government's outlay on goods from sector i out of each domestic currency unit of the revenue remaining after outlays on current consumption and the project in question. We have, therefore, (11.6) G = K + C, + 4)u' (G - K - CG) Equating the government's purchases to its revenue yields (11.7) u'G = t'Y + u'T + (TF4' + 8')X + (P' - (u' + W'))E, where t' (T1, .T., TL). It is now necessary to relate government revenues (and purchases) as defined here to the concept of "uncommitted social income, measured in terms of convertible foreign exchange," which is the numeraire advocatedby Little and Mirrlees (1974, pp. 145-51). Inour model, the impact of a project on government revenue in a particular period is simply the change in u'G associated with the project, that is, the difference between u'G in equilibrium with the project and u'G in equilibrium without the project. As K is the set of government purchases required to put the project into effect, the change in u'(G - K - CG)arising from the project will measure the change in social income available for commitment to other uses only if there is no change in public consumption, 00. However, this change in "uncommitted social income" will indeed take the form of a change in the holdings of foreign exchange reserves only under special conditions. Because the government can exchange one tradable good against another in any quantity it desires at fixed world prices, a necessary condition for its marginal income to be uncommitted, in the sense that the composition of its marginal purchases does not matter, is that there are no allocations for the purchase of nontradables, that is, that 4),and 4), are both zero. Hence, by setting all values of 4),in equation (11.6) equal to zero, except those for noncom- petitive imports (which will be unity), and by keeping CGinvariant with respect to K, the project's effects on uncommitted social income, measured in terms of noncompetitive imports, is simply the associated change in u'(G - K - CG).Measured in terms of convertible foreign exchange, the change in uncommitted social income is, therefore, the change in u'(G - K - CG)times the ratio of the world to the domestic price of noncompetitive imports. 6 Savings and Investment A crucial feature of the model developed here is that private investment is assumed to be financed solely by private (household and corporate) savings. Although this is a plausible assumption in a national setting over the long run, it introduces certain difficulties in the context of a regional economy, which will be discussed in a later section. Taking corporate savings to be equal to undistributed net-of-tax corporate profits, we have: (11.8) u'J = u'S + [(1 - ^;F)C' -u'Q]X, where the vector of household savings is given by (11.9) S = s[(I - T)Y 5. Thle alternative is to assume that the government's total outlays, including those on the project, are given exog- enously. In this case, any difference between revenues and outlays must show up as an equal change in foreign borrowing if private agents are on their budget lines. For a fun discussion, see Bell and Devarajan (1983). 6. This is a consequence of the fact that the economy is In balance of payments equilibrium, as is shown later in this section. Semi-Input-Output Framework: The Muda Project 261 Investment demand in thiLsmodel has two components. First, "balancing" investment may be needed to release capacity bottlenecks in the nontradable sectors as the economy moves from one steady state to another: this is denoted by the vector H. Second, there is "voluntaxy" invest- ment, which is taken to be the residual investable funds allocated in fixed proportions (k) across the various sectors. Thus, (11.10) J = H + *u'(J - H). It should be noted that in order to value private investment in the social cost-benefit calcu- lation, we must know the distribution of investment demands among households (since, presum- ably, investment by poor hauseholds wiLL be socially more valuable than that by rich ones). As our data do not permit this Levelof disaggregation, however, we must represent private invest- ment by a single vector, implicitly assuming that the composition of investment demand is the same across all private institutions. The particular set of assumptions we have made also guarantees balance of trade equilibrium for the economy, that is, P'E = 0. This is readily seen by adding equations (11.1) through (11.10) and setting prices equal to costs at domestic prices. Labor Supply When a project is undertaken, one of its direct effects is to increase the demand for labor, both during the constructiDn phase and when the project is producing output. The laborers thus employed would otherwise have been idle or engaged in some other activity, depending on whether there was full employment to start with.7 As we are dealing with a general equilibrium system, the total effects of a project will, of course, usually differ from its direct effects. Yet the central question remains, Where is the labor coming from? Since the output of nontradables is driven by the level of domestic demand, the ultimate source of labor to the economy, in the sense that all adjustments in the total demand for labor elsewhere in the economy fall upon it, is a subset of sectors producing tradables and the pool(s) of unemployed. 8 If labor is perfectly mobile and homogeneous and its supply is perfectly inelastic, we have e'x + LD = L, where t' = (t1, .. ., (,) is th.e vector of labor-output ratios; Lp is the pool of unemployed workers; and L is the total labor force. This equation implies nothing about which classes of households reallocate their labor in response to changes in output, and it will be consistent with equation (11.4) only if there is but one class of household. When there is more than one such class, there is, in effect, more than on,s sort of labor, or "characteristic factor." The appropriate generali- zation to the case of L household classes is immediate: (11.11) AX + L, = L, where A = 11iieIis the matrix of labor-output ratios, and L, and L denote vectors whose elements are the pools of unemployed workers and total workforces, respectively. Solving the Model The linear system (11.1) through (11.11) has 4M + 2L linearly independent equations in SM + 2L variables. Hence, to solve the system, M of the variables must be specified exogenously. There are two cases of interest. 7. If, initially, there were some unemployment of (mobile) labor and the project were big enough, the project would draw off labor from other sectors as well as exhaust the pool of unemployed. 8. There may be more than one such pool if labor is not perfectly mobile and homogeneous. 262 Multipliers and SAM-Based Models First, if there is unemployment of all kinds of labor (Lp > 0), (11.11) can be ignored. In this case, we fix the levels of outputs of tradable goods. X., and the levels of (net) exports by sectors in the subset N, so that the system is solved for ET,ED,XD,XN, and Y. The system may be written compactly as follows: (11.12) RI, R1 2 | R[3E + R1 41 ( ) + Q) L-~~~~~(V I YG L 0 where R,1 = I - A - ¢(TF' + 8 ) - 4{(1 - Tf)4 - u'f] Rl, = -[B(I - s)(I - r) + (t' + - u's(I -r)] R1 3 = ,,tP' - (u' + Jl')] + I RI, = Eu + B [(I - s) + ( - 4)u']T + (I - ku')H + (I - tsu')(C. + K). Second, for each instance in which there is full employment of one kind of labor, the output of a traded good (or a bundle of traded goods) becomes endogenous, as well as the exports of that good (or goods). This restores the degree of freedom lost whenever an element of Lp is zero, and so (11.11) becomes an effective part of the system, which may be written as: Rll Rl, ° X R1,E + RI4 (11.13) -(V + Q) I O Yp = 0 After suitable transpositions of columns and a simple rearrangement of variables, both variants of the above model can be represented in the following schematic form: (11.14) Z = A*Q + A**u, where Q and Z are the vectors of exogenous and endogenous variables, respectively, and the system's parameters enter into the elements of the matrices A* and A**.The precise form of the scheme will, of course, depend on whether there is full employment of each and every kind of labor. In all of the above cases, the levels of the endogenous variables represent their values when all responses to the exogenous variables are complete. We assume the system adjusts instan- taneously when perturbed by a change in the exogenous variables. Social Cost-Benefit Analysis To determine the impact of a project, we must specify the values taken by the exogenous variables, Q, both in the presence of the project and in its absence, in each case solving for the endogenous variables, Z. If, as is usual, the project affects the technology employed in its own sector, or even elsewhere in the economy, one or both of the matrices A* and A**will be different in the two cases. Denoting the case in which the project is not undertaken by the superscript 0, the difference between the ordered pairs of vectors (Q, Z) and (QO, Z") represents the project's impact on the economy. How can the project's impact on outputs, incomes, savings, and so forth be used to determine the contribution of the project to social welfare in a given year? Following well-established tradition, the arguments of the instantaneous social welfare function are: private consumption; private savings over and above that needed to finance the balancing investments in the nontrad- ables sectors; and uncommitted public income. Here, it will be assumed that the weights for private savings and consumption relative to uncommitted social income (the numeraire) are estimated so as to remain constant over the relevant ranges of these variables, ranges which Framework: Semi-Input-Output The Muda Project 263 may be quite large. Hence, the social profit, AU,associated with the project for a unit time period is given by (11.16) ~AU = u'(G - - K) + '(C - C0) + ('(S - S°) + >F[(l - TF);' - u'lQ](X - XO) -tH, where the vectors Xand t denote, respectively, the social valuations of a unit increase in the levels of consumption and savings for each of the household classes; tF is the social valuation of a unit of retained corporaete profits; tH is the social valuation of the savings which finance the balancing investment, H, whose composition by source may differ from that for savings as a whole; and C = (I - s) [(I - T)Y - T] is the vector of household consumption levels. It is important to note that (11.15) departs from Little and Mirrlees in that the weights A, t, kF, and tH are applied to private consumption and savings measured at market, as opposed to accounting, prices. In the:Lrdiscussion of this choice, Little and Mirrlees (1974, pp. 236-37) seem to be concerned principally with the fact that different consumers may pay different prices for the same good. Faced with this difficulty, they suggest, as a shortcut, that consumers' expenditures be valued at accounting prices. Although we do not find this a very appealing suggestion for making "baskets" of goods purchased at different market prices comparable with one another, it is the only strictly correct way of comparing consumers' expenditures with uncommnitted social income unless the consumer's consumption conversion factors happen to be the same. As Little and Mirrlees recognize, however, an assessment of equity among consumers requires comparisons of their expenditures (or incomes) at market prices, even though this may mean that incomes of two people which differ greatly at market prices may differ little at accounting prices-and hence be of roughly the same social value if this shortcut method is adopted. We have formulated exactly this issue of equity among consumers at the possible cost of introducing some errors into the valuation of their expenditure relative to the numeraire. Indeed, since the method advanced here is designed to finesse the task of estimating accounting prices for nontradables and shadow wage rates, we have no choice in the matter. As for the possibility that different people may pay different prices for the same good, the region under study is blessed with a good transportation system. Moreover, the provincial capital is both modest in scale and readily accessible to all rural households, so that urban and rural cost of living indices probably differ little. To complete the evaluation of the project, the stream of social profits-as given by equation (11.15)-generated by the project throughout its lifetime is discounted at the accounting rate of interest. APPLYING THE MODEL TO THE M A IRRIGATION PROJECT The model set out in the' preceding section describes the flows within a national economy and those to and from the rest of the world. In applying the model to evaluate the Muda scheme, however, we employ data for its surrounding region, which encompasses the state of Perlis and the northern half of Kedah, in northwest Malaysia. Although the basic structure of the model is preserved for a regional economy, some additional considerations are necessary if the results are to be interpreted correctly. First, there are primary commodities, like paddy and smoked rubber sheet, which do not normally enter international trade until they are processed, but instead are shipped in raw form from the producing region to other regions for processing. With the advent of irrigation in 1970, the surge in the region's paddy output resulted in substantial exports of paddy to rice 264 and SAM-Based Multipliers Models mills in southern Kedah and Province Wellesley. As for rubber, less than 16 percent of the output of the region's smallholders and estate sectors is processed within the region itself. Strictly speaking, therefore, these two commodities should be treated as tradables with respect to the region. Yet, the processing of paddy and rubber generates value added within the Malay- sian economy, and the multiplier effects of this value added would go unaccounted for if the model were applied literally. To correct for this, processing activities outside the region are treated in the model as if they took place inside the region, which entails the assumption that technologies, the distribution of value added among households, and the levels of household incomes do not vary across regions. As this comparison involves northern Kedah, on the one hand, and southern Kedah and Province Weflesley, on the other, all of them contiguous, rather poor regions, the accompanying assumption seems defensible. Second, by no means is the whole of private savings invested in the region itself. The outflow of private capital from the region, much of it through the banking system, was especially large in 1972 and for some years thereafter. Very little is known about the uses to which these savings have been put, and for want of anything better it has been assumed that they have been used to finance the purchase of a bundle of investment goods whose composition is the same as that of the region's own investment activities. As all private savings are invested in the model presented above, the procedure adopted here is tantamount to incorporating within the region those activities making deliveries to meet investment demands from outside the region which are also financed by the region's savings. Third, there is a stream of seasonal workers into the Muda region during the paddy harvesting period. Some of these agricultural workers come from southern Kedah and Kelantan, but others come from Thailand. The former belong to households whose incomes are, in all probability, at best equal to those received by landless paddy farm households in the Muda region. Hence, at one extreme, the migrants' seasonal earnings could be simply lumped in with those of Muda's landless households, on the assumption that the output forgone in their home regions is zero. Alternatively, it can be assumed that their seasonal earnings in Muda are exactly equal to the output forgone elsewhere in the economy, both valued at market prices. The assumption under- lying the results presented below rests somewhat uneasily in between, inasmuch as all payments to seasonal workers, Malay and Thai alike, are treated as imports used in paddy production. This is undeniably rough and ready, but the sums involved are small and hardly worthy of more elaborate treatment. However, theire is one respect in which payments to Malay and Thai seasonal workers must be distinguished: presumably the Malaysian government attaches no social value to extra income enjoyed by Thai workers, whereas it ought to smile on additional income accruing to poor Malay agricultural workers. For this reason, payments to Malay seasonal workers must feature in an appropriate way in the calculation of the project's stream of social profits. The next step in the calculation of the project's social profitability is to set up a pair of vectors of exogenous variables for each year of the project's expected lifetime, one denoting the estimated actual values of such variables and the other their estimated values in the hypothetical event that the project had not been undertaken. Construction work on the project began in a small way in 1965, reached a peak in 1967 and 1968, and then tailed off steadily to completion in 1974. Irrigation commenced on a small scale in 1970, about three-quarters of the project command area was served in 1972, and the project attained full maturity in 1974. However, 1974 was also an abnormal year for the ratio of the domestic to the world price of rice, and that ruling in 1975 seems more representative of the conditions likely to hold over the project's lifetime. Hence, assuming that a steady state ruled after 1975, this chronicle of events appears to entail the estimation of no fewer than twenty-two separate vectors of exogenous variables- a daunting task indeed. Fortunately, it turns out that during that part of the project construction period when no Semi-InputbOutput The MudaProject Framework: 265 output is produced, it suffices to estimate the two sets of exogenous variables for just one year. For recall from equation (11.14) that so long as A* and A** are unchanged, that is, there are no changes in technology and. expenditure propensities, the change in the levels of the endog- enous variables is given by: (11.16a) Z - ZO = A*(Q - Q°)- Now if during this phase of the project, the members of the sequence of exogenous vectors Q - QOare scalar multiples of one another-in this case, Q - QO is the project's demand for investment deliveries from the construction sector-the members of the sequence Z - ZOare likewise. Hence, once the sequence of Q - QO has been estimated, it is enough to estimate one member of the sequence of Z - ZOin order to arrive at the rest. The year actually chosen for this purpose is 1967, in keeping with the discussion in Bell and Hazell (1980), from which the estimates of most of the parameters of the system are drawn. Once the project begins to produce output, however, A* will change continuously until the project reaches maturity. Hence, in principle, it is necessary to estimate (11.16b) Z - Z° = A*Q - A*°Q° for all years after 1970, the matrix A** being unaltered by the project. In the present case, the pair of vectors (Q, QO)had already been estimated for the years 1967, 1972, and 1974 in connecticn with the model in Bell and Hazell (1980). To keep the amount ofadditional work within reasonable bounds, it was decided to estimate (A*, A*°) and (Q,QO) for 1970 alone and to use linear interpolation in order to arrive at the values of relevant variables for the intervening "transitional" years, 1971 and 1973. In brief, the paddy production tech- nologies entering into A* and A*owere taken to be output-weighted averages of the correspond- ing technologies for 1967 and 1972. Apart from the obvious difference of deliveries from the construction sector in the presence of the project, Q departs from QO by virtue of the additional paddy output from irrigation and extra throughput in the region's rice mills. There is a small fall in rubber output and processing as labor is drawn off into paddy harvesting activities. Furthermore, exogenous taaes change somewhat as farmers benefit from irrigation subsidies and greater activity in the :region swells tax receipts from vehicle and business registration fees. In treating investment, govermnent outlays, and labor demand endogenously, the model employed here features three sets of parameters, namely, (I, and A, which are not found in Bell and 4', Hazell (1980). As was alreaiy explained, residual government purchases are allocated entirely to noncompetitive imports. The vector 4'denotes the composition of demand for investment goods by the private sector after all necessary purchases of balancing investments have been made. A critical assumption underlying the estimation of IVis that the share of noncompetitive imports in total outlays on such investments stayed constant at its 1972 level. That being so, 'Pvaried little up to 1970, t'hereby keeping A* invariant to all intents and purposes. It showed only slight year-to-year varlation thereafter, so it seems reasonable to assume that there were no fluctuations after 1974, when the project attained its steady state configuration. Turning, at last, to labor supply and demand, we begin by noting that virtually all the people engaged in farming are ethlic Malays, whereas most of the workers employed in the secondary and tertiary sectors are ChiLneseand many of the firms in those sectors are family businesses. Thus, the mobility of labor between the farm and nonfarm sectors is somewhat imperfect, so that the labor market is seigmented-at least, to a degree. In keeping with the evidesnce discussed in Bell and Hazell (1980), the endowments of labor of Muda's paddy farming households are treated as If they were specific to the paddy farming sector. However, it turns out that the project brings about an increase in the demand for labor by this sector which is so large that family labor supplies cannot meet it, and so labor is drawn 266 and SAM-Based Multipliers Models in, first, from the rubber smallholdings on the periphery of the scheme and then from southern Kedah, Kelantan, and Thailand. The losses of output that result elsewhere in the economy have been discussed above; they capture the effects of the project on the allocation of labor in the farming sectors. To complete the story, it is necessary to specify the source of labor to the sectors producing nontraded goods, the output of which will, in general, be affected by the project. If the labor force available to these sectors is not fully employed, even after the project has reached maturity, there will be no effects on output elsewhere in the economy, and the appropriate model is given by equation (11.12). Conversely, if there is always full employment, labor will be drawn off from one or more sectors producing traded goods, and the appropriate model is furnished by equation (11.13). In the present case, the real state of affairs lay somewhere between these two extremes. From what we know of the region's economy a decade ago, it is hard to avoid the conclusion that there was some underemployment of family labor engaged in the sectors producing nontraded goods. At the same time, the boom in employment in the free trade zones in Penang, which began in the early seventies and drew in large numbers of young women from Kedah and Province Wellesley, was attended by a rise in real wage rates. Thus, in the recent past, it is probable that the Muda region and Penang's free trade zones were competing for labor. The evidence is far too shaky to warrant simulations in which there is a switch, at some point, from equations (11.12) to (11.13). Rather, we shall estimate the effects of the project using first the one and then the other. This should suffice to bracket what actually happened. It is also apparent that we are treating the industries in Penang's free trade zone as the source of additional labor to the Muda region's economy, at least where the production of nontraded goods is concerned. At first sight, this may appear somewhat odd to hardened prac- titioners of social cost-benefit analysis, who naturally think of subsistence agriculture as the source of labor to the rest of the economy. However, there is nothing in the celebrated manuals that commands subsistence agriculture to fulfill this role; on the contrary, by examining care- fully how the labor market actually works, we have adhered closely to the spirit of their recom- mendations. For simplicity, we assume that only one sector yields up its labor in response to changes in the demand for workers by other activities. That sector is electronics, which is far and away the biggest employer of labor in Penang's free trade zone. Given the tax and commercial policies toward these zones and the nature of the manufacturing technology in electronics, this sector affects the national economy solely through its demand for domestic labor; for imports of intermediate inputs, final output, and profits all go untaxed, foreigners provide all finance, and the requirements for nontraded goods, such as utilities, are negligible. In effect, therefore, the sole activity of this sector from the point of view of the Malaysian economy is the hiring of labor. In the case of full employment with competitive labor markets, any change in the wage bill for this sector must be accompanied by an equal and offsetting change in the incomes accruing to nonfarm households from the sectors producing nontraded goods.9 ESTIMATION OF NATIONAL PARAMETERS FOR SOCIAL COST-BENEFIT ANALYSIS Four sets of national parameters are used in our calculus: (a) the accounting ratios of traded goods, that is, the ratio of their world price to their domestic price; (b) the social valuation, in terms of uncommitted social income, of a unit of private consumption enjoyed by each household 9. Note that the supply of labor to each sector is assumed to be perfectly elastic, and recall that there are constant returns to scale with no foint production-and no fixed factors outside agriculture. Semi-Input-Output The Muda Project Framework: 267 class (the vector A); (c) the social valuation of private savings in terms of the numeraire (E); and (d) the accounting rate of interest. Veitch (1976) estimates ratios for our ten tradable sectors for 1974. Except for the two rice milling sectors, we assume these ratios were the same throughout the period under study. Since the ratio of the world price to the domestic price of rice changed considerably over the period 1967-75, and rice is the sole direct output of the project, the accounting ratio for rice was estimated separately for each year in this period (Goldman, 1975; and Malaysia, 1974): 1967 0.86r7 1970 0.895 1973 0.716 1968 0.914 1971 0.776 1974 1.110 1969 0.905 1972 0.732 1975 0.901 The sectors producing nontraded goods in this economy that also export significant amounts (albeit in exogenously fixed quantities) are "other agriculture" and "manufacturing n.e.c. (not elsewhere classified)." Their accounting ratios were estimated by using the ratios for those categories in Veitch that correspond most closely to their characteristic commodities. Finally, the accounting ratio for noncompetitive imports was derived by taking a weighted average of the different imports, the weights being obtained from the original SAM tableau. The complete set of accounting ratios by sector for 1974 (from Veitch, 1976) is: Commercial rice mills 1.11 Smallholder rubber 1.22 Small rice mills 1.11 Rubber processing 1.22 Food processing 0.91 Other agriculture 0.91 Fish processing 1.19 Sawnilling 1.32 Paddy production 1.00 Manufacturing n.e.c. 0.80 Fishing 1.19 Noncompetitive imports 0.90 Estates rubber 1.22 (No paddy is in fact exported, and so the value of unity is used solely for accounting require- ments.) To estimate X,the set of social weights for household consumption, two parameters are needed: c, the so-called critical per capita consumption level, at which a unit of additional consumption (valued at market prices) is just as valuable socially as a unit of uncommitted government income; and -q,the elasticity of the marginal utility of consumption. For c, it is reasonable to take the official "poverty line," on the grounds that a prime objective of government policy is to bring all individuals to that level of living as soon as possible. At the time of writing, the value of c in 1972 prices is $375 a year (Visaria, 1979, p. 29), which is about one-third of per capita GDP.1 0 It is difficult to be so definite about r. Following a well-established tradition in this literature, two values 'will be used: unity, which arises from a logarithmic utility function and is mildly egalitarian; and two, which corresponds to Atkinson's (1970) equally distributed equivalent income being the harmonic mean of all incomes and is quite strongly egalitarian. AB the project brought about large changes in the per capita consumption levels of the five classes of households (Bell and Hazell, 1980), the concavity of the utility function implies that potentially serious errors will arise if k is estimated at the steady state values of household consumption. Instead, we proceed as follows. In view of equation (11.15), two additional param- eters of the utility function may be fixed. First, the marginal social utility of a unit of private consumption at c should be unity. Second, the absolute social utility attached to a private consumption level of c may be normalized to c. Hence, we have, for 7 = 1 and TI = 2, respectively, (11.17a) V(c) = c[l + loge(c/6)] (11.7b) V(c) = c[2 - c-c]. 10. The unit of currency is Malaysian dollars ($1 = US$0.40 in 1974). 268 Multipliers and SAM-Based Models Table 11.1. Social Weights for Household Consumption at Market Prices Household Class 1967 1970 1972 1974 X(1) X(2) X(1) X(2) X(l) X(2) A(l) X(2) I Landless paddy 3.25 10.53 2.62 6.90 2.29 5.39 2.14 4.70 worker 2 Labour-abundant 2.37 5.63 2.11 4.45 1.83 3.41 1.61 2.63 paddy farm 3 Land-abundant 1.54 2.36 1.38 1.92 1.19 1.43 1.03 1.08 paddy farm 4 Nonproject farm 1.75 3.09 1.59 2.55 1.39 1.94 1.25 1.42 5 Nonfarm 0.78 0.61 0.73 0.53 0.63 0.41 0.53 0.29 denotes the value of Xevaluated at -i = 1, 2. Note: X(X) The value of A arising from a movement from co to c is, therefore, (11.18) A = [M(c) - V(c 0 )]I[c - c°]. The per capita consumption levels of the various classes of household for 1967 and 1974, both with and without the project, are taken from Bell and Hazell (1980), while those for 1972 with the project have been estimated directly in the course of constructing the SAM for that year. Those for 1970 and 1972 (without the project) were derived by linear interpolation. The resulting values of A are set out in table 11.1. There is no call for the (spurious) accuracy of similar calculations for the intervening years. Up to 1970, the values of A for 1967 will do; those for 1972 will be applied to 1971; and those for 1974 are assumed to hold for 1973 and all years thereafter. In placing a social value on a household's savings, Little and Mirrlees (1974, p. 243) make the suggestion that, as a shortcut, the weight attached to a unit of savings by household class k, gk, be taken as halfway between unity (the weight for uncommitted government income) and the weight for a unit of consumption by that household. Of course, if the latter exceeds unity, the rule of thumb implies that the households concerned are deemed to be saving "too much," given their poverty. Using the Little-Mirrlees rule of thumb, the sets of weights for household savings corresponding to those for consumption follow at once from table 11.1. As for corporate savings, which are ultimately distributed to very rich households or foreign nationals, these will generally have a very low social valuation. In the light of the social weight attached to the savings of nonfarm households, perhaps a weight of 0.2 for corporate savings does not seem objectionable. Finally, there is the matter of the social valuation of the savings needed to finance the set of balancing investments, H. Most of these investments were undertaken in the nonfarm sector of the economy between 1970 and 1974, by incorporated and unincorporated enterprises alike, while others took the form of house construction or improvement by farm households. Accordingly, in rather rough and ready fashion, the value of tx is taken to be 0.7. The argument advanced above rests on the notion that all households have access to a perfect capital market and so earn the same real rate of return on their savings. In practice, poorer households in the Muda region are likely to place their financial savings in bank accounts yielding zero or negative real rates of return, to the advantage of affluent borrowers who have access to highly profitable investment opportunities. If, in the extreme, all household savings were placed in the banking system and then lent to the government, to the richest class of households, or to private corporations, then the social weight attached to all savings, irre- spective of source, would lie somewhere in the range of those for these three institutional Semi-Input-Output The Muda Project Framework: 269 categories, depending on their shares in the total volume of lending. Very little is known about these matters, and it is profi&less to attempt a refined calculation. Moreover, since the purpose of setting up this extreme case is to establish a parameter value for sensitivity analysis, a social weight of 0.7 for all private savings should provide a plausible lower bound on the valuation of this use of income flowing from the project. Turning to the accounting rate of interest (ARI), Anand (1977) arrives at an estimate for Malaysia of 10 percent. An alternative, crude approach is as follows. A nominal rate of 10 percent is fairly close to the rate at which Malaysia was able to borrow on the world market, since the country's balance Dfpayments position was quite strong in the period 1967-74 and foreign reserves have remained high thereafter. However, the model set out here is formulated in constant (domestic) priclas. Adjusting for an average annual inflation rate of 5 percent and assuming exchange rate stability against an appropriate basket of currencies, we estimate the ARI to be around 5 percent. It may be felt, however, that public sector projects should have a higher rate of return than the marginal cost of foreign borrowing, so that this figure of 5 percent is probably a lower bound. To reflect this consideration, we perform sensitivity exper- iments using figures of 5, 10, and 20 percent for the ARI. SCICIALCOST-BENEFIT CALCUIATIONS As shown in equation (1:L.15), the stream of social profits generated by the project is made up of various elements: the project's effects on household consumption and savings, on retained corporate profits, and on uncommitted government income. It is worthwhile setting out the individual trajectories of these elements before combining them to yield the time path of social profits, for the positive basis of the subsequent normative aggregation is then clear. The trajectories for the twvoextreme cases of labor supply facing the nonfarm sectors of the economy, namely, unemployment and full employment, are laid out in tables 11.2 and 11.3, respectively. Qualitatively, there is nothing very startling about them, given the nature of the project. In the construction phase, the government made large outlays on construction, which accrued, in the first instance, as incomes to nonfarm households and corporations, both domes- tic and foreign. (In the ful employment case, of course, output and incomes fell in the "source" sector, so that these househlolds gained little on balance.) Agricultural households participated only marginally in construLction and the extra activity in other domestic sectors induced in its wake. Indeed, the incomes of farm households on the periphery of the scheme fell slightly, since they earned much of their livelihood from rearing livestock, the numbers of which declined in the face of mechanization induced by the irrigation project. After the project came on stream in 1970, the incomes of agricultural households rose sharply. In the case of unemployed labor in the nonfarm sectors, the additional spending by paddy farming households boosted nonfarm incomes and corporate profits as construction work on the project wound down. Uncommitted government income recovered, too, as revenue from income and indirect taxes increased by more than enough to offset the losses arising out of the protection afforded to the paddy sector in all years. (In 1974, an almost mature project and a very high world price of rice combined to yield an exceptionally high payoff for the government.) In the case of fully employed nonfarm labor, the increase in incomes accruing to nonfarm households is naturally rather modest, with attendant repercussions on farm households, which do earn small incomes from the production of nontraded goods. The fall in the output of traded goods as labor is drawn off from the electronics sector results in a fall in uncommitted government income, and a substantial one it has been since the project has attained maturity. The outlays on balancing investments are the higher of the two (ral,her rough) estimates discussed in Bell and Hazell (1980), just to be on the safe side. The flows for 1975 are assumed to be the steady state values of relevant variables Table 11.2. Changes in Incomes Due to the Project: Unemployed Nonfarm Labor (in thousands of 1972 U.S. dollars) Institution 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 Rousehold Claeis 1ousehold Clas 16 44 56 40 648 1,120 1,591 1,858 2,124 2,124 1 Si - S; 0 0 0 0 0 2 3 4 5 5 5 C2 - C2 9 36 102 129 94 2,875 5,345 7,814 10,013 12,211 12,211 2 * 1 5 13 16 12 359 667 975 1,249 1,523 1,523 2 S2 C3 - C; 6 _ 23 64 81 59 8,091 15,383 22,674 28,989 35,304 35,304 S3 - 1 5 14 18 13 1,717 3,264 4,811 6,151 7,490 7,490 C4 - CZ -6 -22 -62 -78 -57 -96 405 906 2,112 3,317 3,317 S4 S -1 -5 -13 -16 -12 -20 85 190 442 694 694 C5 - C; 1,527 5,834 16,468 20,832 15,121 13,904 20,343 26,782 32,271 37,760 37,760 S5 - S; 289 1,105 3,123 3,946 2,864 2,633 3,853 5,073 6,113 7,152 7,152 Change in retained corporate profits 617 2,358 6,664 8,421 6,112 3,911 4,460 5,008 5,354 5,700 5,700 Change in uncom- mitted govern- ment income -6,539 -16,920 -42,483 -51,401 -40,540 -14,874 -29,713 -16,259 -2,861 16,227 11,350 'Balancing' investments -720 -2,160 -5,040 -5,760 -5,040 -5,760 -10,080 -12,240 -14,400 -10,800 0 Total -4,812 -9,725 -21,088 -23,756 -21,334 13,390 15,135 47,329 77,296 118,707 124,630 Table 11.3. Changes in Incomes Due to the Project: Fully Employed Nonfarm Labor (in thousands of 1972 U.S. dollars) Institution 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 Household Class 1 -i C; 2 6 17 21 16 630 1,094 1,558 1,815 2,071 2,071 S. s 0 0 0 0 0 2 3 4 5 5 C2 - -3 -11 -31 -39 -28 2,787 5,220 7,652 9,800 11,947 11,947 2 S2 0 ° I -12 -4 -5 -4 348 651 954 1,222 1,490 1,490 C3 - C. -20 -76 -216 -273 -198 7,904 15,118 22,332 28,541 34,749 34,749 S3 - -4 -16 -46 -58 -42 1,677 3,208 4,738 6,056 7,373 7,373 C4 - C; -28 -106 -300 -379 -275 -255 181 617 1,732 2,847 2,847 S4 - S; -6 -22 -63 -80 -58 -53 38 129 363 596 596 C5 - C5 396 1,513 4,275 5,402 3,921 5,563 8,743 11,923 12,806 13,688 13,688 S5 - 75 287 810 1,024 743 1,054 1,656 2,258 2,426 2,593 2,593 Change In retained corporate profits 561 2,143 6,054 7,650 5,553 3,489 3,878 4,266 4,383 4,499 4,499 Change In uncom- mitted govern- ment Income -6,787 -17,867 -45,161 -54,785 -42,996 -20,942 -37,080 -24,924 -11,801 7,011 2,060 'Balancing' investments -720 -2,160 -5,040 -5,760 -5,040 -5,760 -10,080 -12,240 -14,400 -10,800 0 Total -6,534 -16,130 -39,705 -47,282 -38,408 -3,556 -7,370 19,267 42,948 78,069 83,918 272 Multipliers and SAM-Based Models for the rest of the project's life. In keeping with the procedures used in the final appraisal by the World Bank, the project is assumed to have a salvage value of $100 million in 2000. The streams of social profits associated with the flows in tables 11.2 and 11.3 have been calculated under five sets of assumptions about the social valuation of household consumption, savings, and corporate profits. These correspond to three values of the elasticity of the marginal utility of consumption and two methods of placing a social valuation on household savings and retained corporate profits, as already discussed in the previous section. In particular, to satisfy those who would have no truck with distributional weights to value private consumption and savings in terms of government income, the streams of social profits have been calculated with all such weights set to unity. In this case, the streams are then simply the algebraic sums of the elements making up the individual columns of tables 11.2 and 11.3. These alternative streams are set out in tables 11.4 and 11.5 In all cases, the project's net present social value is handsomely positive at an accounting rate of interest of 10 percent, and it can get by even at 20 percent. Not surprisingly, the project turns out to be more profitable if nonfarm labor is not fully employed. Thus, the great improvements in the material living standards of people in the Muda region which have been brought about by the project-at once evident to the casual observer-are reflected in, and consistent with, the project's high social profitability. Interestingly enough, the value of -1makes very little difference to the project's net present social value, even at the extreme ends of its range. This has come about by chance in that the poverty line, c, where unit increases in private consumption and uncommitted government income are equally valuable, happens to be roughly equal to the mean social value of additional household consumption as actually distributed over the five classes of households. If, for exam- ple, c had been set at the level at which households are just drawn into the income tax net, which is much higher than the official poverty line, the project would show much higher social profits for q = 2 than -q = 0. As it is, in the case of unemployed nonfarm labor, the results for these cases are almost identical (given the ARI), but the net present value is somewhat higher in both than that for - = 1. For T = 2, the higher social values placed on government outlays early in the project's life happen to be counterbalanced by the large net social benefits associated with the high income accruing to the project's comparatively poor beneficiaries when it attained maturity. When the social utility function is less concave, the fall in the mean social value of additional private income is apparently strong enough to cause a reduction in the project's net present social value. For similar reasons, the two methods of valuing private savings do not make much of a difference to the project's profitability. For the case in which nonfarm labor is fully employed, the project's net present social value increases with ,. The reason for this is that in the face of full employment of nonfarm labor, the additional income accruing to nonfarm households is more modest, so that the ratio of additional farm to nonfarm income generated by the project is greater than in the previous case. In turn, this implies that the social value of additional household income is also greater, for the distribution of income generated by the project is more favorable to farm households when nonfarm labor is fully employed. Moreover, the higher is 9q, the stronger is this effect. CONCLUSIONS It is clear that the analytical approach adopted here rests explicitly on the use of a simple general equilibrium model of the economy to determine a set of endogenous variables with the prices of all goods, domestic and foreign alike, parametrically given. Starting from an initial equilibrium for the economy, as characterized by the conditions set out earlier in this chapter, a "project" takes the form of a sequence of changes in the economy's exogenous variables and Semi-Input-Output Framework: The Muda Project 273 Table 11.4. Social Profits from the Project: Unemployed Nonfarm Labor (in thousands of 1972 U.S.dollars) Stream of Undiscounted Social Profits nri 1 n =2 Year n =0O A B A B 1965 - 4,812 - 5,332 - 5,388 - 5,556 - 5,590 1966 - 9,725 - 11,857 - 12,071 - 12,717 - 12,843 1967 - 21,088 - 27,482 - 28,087 - 29,910 - 30,268 1968 - 23,756 - 32,006 - 32,770 - 35,074 - 35,526 1969 - 21,334 - 27,082 - 27,637 - 29,309 - 29,637 1970 13,390 15,016 13,444 26,625 24,439 1971 15,1.35 18,971 16,546 33,862 30,707 1972 47,229 52,730 49,452 70,904 66,779 1973 77,296 77,528 73,867 92,290 88,109 1974 118,707 112,048 108,002 123,378 119,161 1975 124,630 114,731 110,686 126,061 121,844 Net Present Social Value of the Project n-i1 n -2 Accounting Rate nc 0O of Interest A B A B 5% 1,260,000 1,160,000 1,110,000 1,300,000 1,250,000 10% 547,375 494,632 469,906 568,036 541,805 20% 132,719 111,152 101,915 136,891 127,167 Note: The value n is the elasticity of the marginal social utility of private consumption. In case (A),the social weights 2 for private savings (0x) are drawn from table 11. 1, using the rule tk = I1 + Xk(,q)]/ . In case (B),a uniform weight of 0.7 applies to all private savings. underlying technology. Within each time period, we assume that the system adjusts fully to the changes in the exogenouLs variables. The time path followed by the system's variables may be forged out of the chain cf comparative static equilibria thus derived. This yields, among other things, the streams of ccinsumption, savings, and government revenues, the changes in which are the arguments of thet social welfare function. The criterion for accepting the project is that the net present social value of the stream of changes in national income, as given by equation ( 11.15), should be positi.ve. 274 Multipliers and SAM-Based Models Table 11.5. Social Profits from the Project: Fully Employed Nonfarm Labor (in thousands of 1972 U.S. dollars) Stream of Undiscounted Social Profits 2~~~~ Year n -O A B A B 1965 - 6,534 - 6,796 - 6,803 - 6,927 - 6,922 1966 - 16,130 - 17,453 - 17,481 - 17,956 - 17,936 1967 - 39,705 - 43,292 - 43,374 - 44,714 - 44,660 1968 - 47,282 - 51,984 - 52,087 - 53,780 - 53,712 1969 - 38,408 - 41,583 - 41,658 - 42,887 - 42,837 1970 -3,556 559 -704 13,423 11,428 1971 -7,370 826 -1,246 17,815 14,819 1972 19,267 30,897 28,016 52,011 48,014 1973 42,948 52,773 49,508 71,887 67,718 1974 78,069 84,371 80,721 101,484 97,144 1975 83,918 86,980 83,330 104,094 99,754 Net Present Social Value of the Project I n =2 Accounting Rate n - 0 of Interest A B A B 5% 734,979 779,666 737,101 992,354 940.839 10% 267,500 289,562 268,779 396,027 370,706 20% 13,801 19,780 12,625 57,499 48,728 Note: The value i is the elasticity of the marginal social utility of private consumption. In case (A),the social weights for privatesavings (t) are drawn from table 11.1, using the rule tk = [1 + Xk(X)]/2.In case (B), a uniform weightof 0.7 applies to all private savings. The implicit assumption of instantaneous and ful adjustment to the exogenous disturbances of the period in question, and the neglect of all such disturbances in the future, which entails myopic expectations, are certainly very strong assumptions. Yet the "dual" route of using shadow prices is tarred with the same brush (Srinivasan, 1982). Perhaps nothing better can be done where practical analysis is concerned, but this does not exonerate the analyst of the responsibility to be clear about what he is doing. Even if a truly dynamic structure were available, It is far from obvious that the costs of such intrinsically difficult refinements would be justified by the improvements in practical decisions they would bring about. Still, one is left feeling Semi-Input-Output The Muda Project Framework: 275 uncomfortable that the results should rest on such foundations-unless one seeks refuge in the convention that projects are small enough for these considerations not to matter. A more particular point to emerge from the analysis is that the only interindustry linkages that have an induced-as opposed to a direct-effect on the determination of the endogenous variables are those expressed by the technologies for producing nontradables. This follows immediately from the semi-input-output formulation, of course, but it highlights the fact that capturing multiplier effects in cost-benefit analysis requires, in principle, that the estimates of the relevant parameters pertain to the region in question. As developing countries are usually regionaUly heterogeneous, thLisrequirement cannot be dismissed as a theoretical nicety. However, the analyst usually considers himself fortunate if he has access to a fairly recent national input- output table. As for the issues in social cost-benefit analysis, in this particular case study, variations in the elasticity of the marginal social utility of private consumption have relatively little effect on the project's present social value, at least within the range of values considered as normal in this literature. The project's main beneficiaries are poor and hence socially deserving. However, even when all incomes are given the same social weight the social rate of return exceeds 20 percent,. It will not have escaped bhe reader's attention that for all the care lavished on incorporating all of the project's effects in a static framework, the treatment of the intertemporal aspects of resource allocation has been rough and ready. Thus, in estimating the social value of private savings, we have resorted to one shortcut method recommended in the literature. As it turns out, the social profitability is robust to changes in this parameter. Were this not so, a detailed treatment of the dynamics of the economy would have been necessary. Once the model has been built and estimated, it can be used to evaluate any project. Compu- tationally, nothing more than matrix inversion is needed. Even so, this invites the retort that valuing the direct inputs and outputs of a project at their (appropriate) shadow prices is more straightforward still, assuming that a set of such shadow prices is available. That, however, raises an important question: What are the shadow prices that correspond to the primal model laid out here? Although we have addressed this question at length elsewhere (Bell and Deva- raj an, 1983), it should be noted that the Muda project was sufficiently large to affect the supply price of labor, so that the shadow prices for the economy at that point would not have been stationary. Hence, the use of shadow prices based on a constant supply price of labor, which is advocated (and usually valid) for small projects, would be incorrect in this case. Moreover, estimating the shadow prices appropriate to this case would involve the same steps followed in the primal route taken. here, so that there would be no saving in effort. References The word "processed" is used in this list to indicate that a work has been reproduced by some process other than conventional typesetting and printing, such as photocopying or mimeo- graphing. Adelman, I., and S. Robinson (1978) Income Distribution Policy in Developing Countries: A Case Study of Korea, London: Oxford University Press. Aghevli, B. B. (1977) "Money, Prices and the Balance of Payments: Indonesia 1968-73," Journal of Development Studies, vol. 13, January, pp. 37-57. Aghevli, B.B.,and M. Khan (1977) "Inflationary Finance and the Dynamics of Inflation: Indo- nesia 1951-72," American Economic Review, vol. 67, no. 3, June, pp. 390-403. Ahluwalia, M., and H. B. Chenery (1974) "A Model of Distribution and Growth," ch. 11 in H. B. Chenery and others, Redistribution with Growth, London: Oxford University Press. Ahluwalia, M., and F. Lysy (1979) "Welfare Effects of Demand Management Policies: Impact Multipliers under Alternative Model Structures," Journal of Policy Modeling, vol. 1, no. 3, September, pp. 317-42. A.ltinixr, 0. (1975) "Income Distribution Estimates from Household Surveys and Population Census in Latin America: An Assessment of Reliability," paper prepared for the Economic Commission for Latin America and the Development Research Center, Washington, D.C.:World Bank (processed). Anand, S. (1977) "A Comparison of the IBRD and Little-Mirrlees Appraisal of a Highway Project in Malaysia," Washington, D.C.:World Bank (processed). Atkinson, A. B. (1970) "On the Measurement of Inequality," Journal of Economic Theory, vol. 2, September, pp. 244-63. Bacharach, M. 0. L. (1970) Biproportionate Matrices and Input-Output Change, Cambridge: Cambridge University Press. Balassa, B., and associates (1971) The Structure of Protection in Developing Countries, Balti- more, Md.: Johns Hopkins University Press. Barkay, R. M. (1975) "National Accounting as a Planning Tool in Less Developed Countries: Lessons of Experience," Review of Income and Wealth, series 21, no. 4, pp. 349-70. Barna, T. (1975) "Quesnay's Tableau in Modern Guise," Economic Journal, vol. 85, no. 339, September, pp. 485-96. Bell, C. L. G., and S. Devarajan (1979) "Towards a Synthesis of Semi-Input-Output and Little- Mirrlees: A Social Cost-Benefit Analysis with Multiplier Effects of an Irrigation Project in Northwest Malaysia," Pakistan Development Review, vol. 18, no. 2, pp. 165-85. (1980) "Semi Input-Output and Shadow Prices: A Critical Note," Bulletin of the Oxford Institute of Economics and Statistics, vol. 42, no. 3, pp. 251-56. (1983) "Shadow Prices for Project Evaluation under Alternative Macroeconomic Regimes," Quarterly Journal of Economics, vol. 98, no. 3, pp. 457-78. Bell, C. L. G., and P. B. R. Hazell (1980) "Measuring the Indirect Effects of an Agricultural Investment Project on its Surrounding Region," American Journal of Agricultural Economics, vol. 62, no. 1 (February), pp. 75-86. Bell, C. L. G., P. Hazell, and R. Slade (1982) Project Evaluation in Regional Perspective, Baltimore, Md.:Johns Hopkins University Press. 276 References 277 Bhat, M. D. (1972) "Forecasti,ng in Financial Sector," presented at the European meeting of the Econometric Society, Buda:pest, September. Bhatia, R. J., and P. Engstrom. (1972) "Nigeria's Second National Development Plan: A Financial Analysis," IMF Staff Papers, vol. 19, no. 1, March, pp. 145-73. Bhatt, V. V. (1969) "Savings and Flow of Fu,nds Analysis: A Tool for Financial Planning," Reserve Bank of India Bulletin, vol. 22, no. 9, September, pp. 1387-1401. Blitzer, C. R., P. Clark, and L. Taylor (1975) Economy-Wide Models and Development Planning, London: Oxford University Press. Botswana, Ministry of Finance and Development Planning (1973) Manpower and Employment, Gaborone: Central Statistical Office. (1976a) Rural Income Distribution Survey, 1974/5, Gaborone: Central Statistical Office. (1976b) National Accounts of Botswana 1974/5, Gaborone: Central Statistical Office. Brainard, W. C., and J. Tobin (1968) "Pitfalls in Financial Model-Building," American Economic Review, vol. 58, no. 2, May, pp. 99-122. Byron, R. (1978) "The Estimation of Large Social Accounting Matrices," Journal of the Royal Statistical Society, series A, vol. 141, part 3, pp. 359-67. Cambridge University, Department of Applied Economics (1962-74) A Programme for Growth, vols. 'L-11, London: Chapman and Hall. (1975-) Cambridge Studies in Applied Econometrics, London: Chapman and Hall. Ceylon, Central Bank (1971) Annual Report, Colombo. Ceylon, Department of Census and Statistics (1971) "Preliminary Report on the Socioeconomic Survey of Ceylon, 1969-70," Colombo. (1973) National Accounts 1970, Colombo. Champernowne, D. G. (195'3) "A Model of Income Distribution," Economic Journal, vol. 63, no. 250, pp. 318-51. (1969) Estimation andUncertainty in Economics, vols. 1-3, Edinburgh: Oliver and Boyd. (1973) The Distribution of Income between Persons, Cambridge: Cambridge University Press. Chander, R., and others (1980) "Social Accounts and the Distribution of Income: The Malaysian Economy in 1970, Review of Income and Wealth, vol. 26, no. 1, March, pp. 67-85. Chenery, H. B., and others (:L974) Redistribution with Growth, London: Oxford University Press. Chenery, Hollis B., and Alan Strout (1966) "Foreign Assistance and Economic Development," American Economic Review, vol. 56, pp. 679-733. Clark, P. B. (1975) "Intersectoral Consistency and Macroeconomic Planning," ch. 5 in C. R., Blitzer, P. Clark, and L. .'aylor, Economy-Wide Models and Development Planning, London: Oxford University Press. Copeland, J. R., and E. Hen.ry (1975) Irish Input-Output Income Multipliers, 1964 and 1968, paper no. 82, Dublin: Economic and Social Research Institute. Corden., M. (1966) "The Structure of a Tariff System and the Effective Protection Rate," Journal of Political Economy, vol. 74, no. 3, June, pp. 221-37. Crawford, E., and E. Thorbecke (1978) "Employment, Income Distribution, Poverty Alleviation and Basic Needs in Kernya," report of an International Labour Office consulting mission (processed). Deaton, A. (1974) "A Reconsideration of the Empirical Implications of Additive Preferences," Economic Journal, vol. 84, no. 334, pp. 338-48. Deaton, A., and J. Muellbauer (1980) "An Almost Ideal Demand System," American Economic Review, vol. 70, no. 3, June, pp. 312-26. 278 References deMelo, M. (1979) "Agricultural Policies and Development: A Socioeconomic Investigation Applied to Sri Lanka," Journal of Policy Modelling, vol. 1, no. 2, May, pp. 217-34. Dervis, K., J. de Melo, and S. Robinson (1982) General Equilibrium Models for Development Policy, Cambridge: Cambridge University Press. Dervis, K., and S. Robinson (1978) The Foreign Exchange Gap, Growth and Industrial Strategy in Turkey: 1973-83, Staff Working Paper no. 306, Washington, D.C.:World Bank. Desai, P. (1961) "A Short-Term Planning Model for the Indian Economy," Review of Economics and Statistics, vol. 43, no. 2, pp. 193-200. Divatia, V. V. (1969) "An Operational Technique of Financial Planning," Reserve Bank of India Bulletin, vol. 23, no. 9, September, pp. 1402-13. Drud, A., W. Grais, and G. Pyatt (1983) "The TV-Approach: A Systematic Method of Defining Economy-wide Models Based on Social Accounting Matrices," paper presented at the Fourth IFAC/IFORS Conference on Modeling and Control of National Economies, Washington, D.C., June 17-19. Frank, C.R. (1967) "The Seers Modified Input-Output Table: Some Projection Techniques," paper no. 112, New Haven, Conn.: Yale University, Economic Growth Center. Goldman, R. N. (1975) Staple Food Self-Sufficiency and the Distributive Impact of Malaysian Price Policy, Minden, Penang: Universiti Sains, Malaysia, School of Comparative Social Sciences. Goldsmith, R. (1969) Financial Structure and Economic Development, New Haven, Conn.: Yale University Press. Goodwin, R. M. (1949) "The Multiplier as Matrix," Economic Journal, vol. 59, no. 236, pp. 537- 55. Grais, W. (1982) "The Transaction Values Approach to the Formulation and Implementation of Computable General Equilibrium Models," Washington, D.C.:World Bank (processed). Greenfield, C. C., and H. Fell (1979) "The Estimation of Price Effects in a Social Accounting Matrix," Review of Income and Wealth, series 25, no. 1, March, pp. 65-81. Hayden, C.,and J. I. Round (1982) "Developments in Social Accounting Methods as Applied to the Analysis of Income Distribution and Employment Issues," World Development, vol. 10, no. 6, pp. 451-65. Hicks, J. R. (1942) The Social Framework, Oxford: Clarendon Press (4th ed., 1971). International Labour Office (1970) Towards Full Employment: A Program for Colombia, Geneva. (1971) Matching Employment Opportunities and Expectations: A Program of Action for Ceylon, vols. 1 and 2, Geneva. (1973a) Employment and Income Policies for Iran, Geneva. (1973b) Strategies for Employment Promotion: An Evaluation of Four Interagency Employment Missions, Geneva. - (1976) Employment, Growth and Basic Needs: A One-World Problem, Geneva. Johansen, L. (1960) A Multi-Sectoral Study of Economic Growth, Amsterdam: North Holland. Kahn, R. F. (1931) "The Relation of Home Investment to Unemployment," Economic Journal, vol. 41, no. 162, pp. 173-98. Khan, M. S. (1974) "Experiments with a Monetary Model for the Venezuelan Economy," IMF Staff Papers, vol. 21, pp. 389-413. King, G. (1696) "Natural and Political Observations and Conclusions upon the State and Condi- tion of England," printed as an appendix to Sir George Chalmers, An Estimate of the Compar- ative Strength of Great Britain. Reprinted in Two Tracts by Gregory King, Baltimore, Md.: Johns Hopkins University Press, 1936. References 279 Kuyvenhoven, A. (1975) "The Integration of Project and Sector Analysis-Some Further Remarks," Bulletin of the Oxford Institute of Economics and Statistics, vol. 38, pp. 66-77. Lancaster, K. J. (1966) "A NEw Approach to Consumer Theory," Journal of Political Economy, vol. 74, no. 2, April, pp. 132-57. (1968) Mathematical Economics, London: Macmillan. Little, I. M. D., and J. Mirrlees (1974) Project Appraisal and Planning for Developing Countries, London: Heinemann. Malaysia, Department of Statistics (1974) External Trade Statistics, Kuala Lumpur. (1975) National Accounts of Peninsular Malaysia, 1960-71, Kuala Lumpur. Malaysia, Government of (1976) Third Malaysia Plan 1976-80, Kuala Lumpur. Maton, J., F. Paukert, and J. '3kolka (1978) Redistribution of Income Patterns of Consumption and Employment: A Case Study for the Philippines, Geneva: International Labour Office. Narapalasingam, S. (1970) "On the Construction and Implementation of a Planning Model for Ceylon," Ph.D. dissertation. University of Bristol. Ng,. L. S. (1974) "An Income Distribution and Employment Consistency Model of the Philip- pines," Ph.D. dissertation, Iowa State University. Nicholson, J. L. (1964) "Redistribution of Income in the United Kingdom in 1959, 1957, and 1953," Income and Wealth, Series X, London: Bowes and Bowes. Otani, I. (1975) "Inflation in an Open Economy: A Case Study of the Philippines," IMF Staff Papers, vol. 23, no. 1, March, pp. 164-99. Otani, I., and Y. C. Park (1976) "A Monetary Model of the Korean Economy," IMF Staff Papers, vol. 22, no. 3, November, rp. 750-74. Patrick, H. (1966) "FinanciaL Development and Economic Growth in Underdeveloped Countries," Economic Development and Cultural Change, vol. 14, no. 2, January, pp. 174-89. Phillips, A. (1955) "The Tatleau Economique as a Simple Leontief System," Quarterly Journal of Economics, vol. 69, no. 1, February, pp. 137-44. Ploeg, F. van der (1982) "Reliability and the Adjustment of Sequences of Large Social Accounting Matrices," Journal of the Royal Statistical Society, Series B, vol. 145, part 2, pp. 169-94. Polak, J. (1957) "Monetary Analysis of Income Formation and Payments Problems," IMF Staff Papers, vol. 6, no. 1, November, pp. 1-50. Pyatt, G., and others (1972) Employment and Income Policies for Iran: Methodology for Macro- economic Projections, Geneva: International Labour Office. Pyatt, G., A. R. Roe, and associates (1977) Social Accounting for Development Planning with Special Reference to Sri Lanka, Cambridge: Cambridge University Press. Pyatt, G., and J. I. Round (1977) "Social Accounting Matrices for Development Planning," Review of Income and Wealth, series 23, no. 4, December, pp. 339-64. (1978) "The Distribution of Income and Social Accounts: A Study for Malaysia in 1970," Washington, D.C.:World Bank, Development Research Center (processed). (1979) "Accounting and Fixed-Price Multipliers in a Social Accounting Matrix Frame- work," Economic Journal, vol. 89, no. 356, December, pp. 850-73. Pyatt, G., and E. Thorbecke (1976) Planning Techniques for a Better Future, Geneva: Interna- tional Labour Office. Quesnay, F. (1758) Tableau Economique, reproduced in facsimile with an introduction by H. Higgs by the British Economic Society, 1895. Revell, J., and others (1967) The Wealth of the Nation, Cambridge: Cambridge University Press. Roe, A. R. (1978) "Administered Interest Rates, Transaction Costs and Capital Market Imper- 280 References fections in Developing Countries," Washington, D.C.: World Bank, Development Economics Department (processed). Saito, K. W. (1977) "The Determinants of Savings Behaviour: A Survey of the Evidence," Studies in Domestic Finance no. 35, Washington, D.C.:World Bank, Development Economics Depart- ment. Scott, M. F. G. (1976) "Shadow Wages in Mauritius," in I. Little and M. Scott, eds., Using Shadow Prices, London: Heinemann. Seers, D. (1975) "The Structure, Content and Uses of the Revised SNA in the Light of the Requirements and Circumstances of Developing Countries," paper prepared for the Interre- gional Seminar on the Revised System of National Accounts, United Nations Statistical Office, New York (processed). Sen, A. K. (1973) On Economic Inequality, Oxford: Clarendon Press. Seton, F. (1973) "Shadow Prices for Chile," in R. Eckaus and P. Rosenstein-Rodin, eds., Analysis of Development Problems: Studies in the Chilean Economy, Amsterdam: North Holland. Shorrocks, A. F. (1976) "Income Mobility and the Markov Assumption," Economic Journal, vol. 86, no. 343, pp. 566-78. Squire, L., and H. van der Tak (1975) Economic Analysis of Projects, Baltimore, Md.: Johns Hopkins University Press. Srinivasan, T. N. (1982) "General Equilibrium Theory, Project Evaluation and Economic Devel- opment," Walras-Bowley lecture, Econometric Society, Bou7lder, Colorado, published in M. Gersovitz and others, eds., The Theory and Experience of Economic Development, London: Allen and Unwin, pp. 229-51. Stone, J. R. N. (1966) "The Social Accounts from a Consumer Point of View," Review of Income and Wealth, series 12, no. 1, March, pp. 1-33. Stone, J. R. N., D. Champernowne, and J. Meade (1942) "The Precision of National Income Estimates," Review of Economic Studies, vol. 9, pp. 111-25. Stone, J. R. N., and Giovanna Stone (1977) National Income and Expenditure, 10th ed. London: Bowes and Bowes. Taylor, L., and others (1980) Models of Growth and Redistribution for Brazil, London: Oxford University Press. Thorbecke, E., and J. Sengupta (1972) "A Consistency Framework for Employment Output and Income Distribution Projections Applied to Colombia," Washington, D.C.:World Bank, Devel- opment Research Center (processed). Tinbergen, J. (1966) "Some Refinements of the Semi-Input-Output Method," Pakistan Devel- opment Review, vol. 6, pp. 243-47. United Kingdom, Central Statistical Office (1952-) National Income and Expenditure, London: HMSO. (1953-) Economic Trends, London: Her Majesty's Stationery Office (HMSO). (1973) Input-Output Tables for the United Kingdom, 1968. studies in official statistics no. 22, London: HMSO. United Kingdom, Department of Employment (1961-) Family Expenditure Survey, London: HMSO. United Kingdom, Ministry of Overseas Development (1977) "Social Accounting Matrix, Botswana, 1974/5," London (processed). United Nations Industrial Development Organization (ITNIDO) (1972) Guidelines for Project Evaluation, New York: United Nations. References 281 United Nations Statistical Office (1953) A System of National Accounts and Supporting Tables, Studies in Methods, series F, no. 2, rev. 1, 1960; rev. 2, 1964, New York: United Nations. (1968) A System of National Accounts, series F, no. 2, rev. 3, New York: United Nations. (1972) "Draft System of Statistics on the Distribution of Income, Consumption and Accumulation," Statistical Commission agendum Paper, New York: United Nations. (1973) Input-Output Tables and Analysis, Studies in Methods, series F, no. 14, rev. 1, New York: United Nations. (1975) Towards a System of Social and Demographic Statistics, series F, no. 18, New York: United Nations. (1977a) "Provisional International Guidelines on the National and Sectoral Balance Sheet and Reconciliation Aecounts of the System of National Accounts," Statistical Papers, series M, no. 60, New York: United Nations. (1977b) Provisional Guidelines on Statistics of the Distribution of Income, Consumption and Accumulation of Households, Studies in Methods, series M, no. 61, New York: United Nations. Vandome, P. (1958) "Aspects of the dynamics of consumer behaviour," Bulletin of the Oxford Institute of Economics and Statistics, vol. 20, no. 1, pp. 65-105. Veitch, M. D. (1976) National ]?arameters for Project Appraisal in Malaysia, vol. 1, Kuala Lumpur: Economic Planning Unit. Venkatachalam, T. R., and Y. Sarma (1977) "Structure and Trends in the National Balance Sheet of India," Journal of Income and Wealth, New Delhi, vol. 1, April, pp. 95-136. Visaria, P. (1979) "Incidence of Poverty and the Characteristics of the Poor in Peninsular Malaysia," Washington, D.C.:World Bank (processed). von Saleski, R. (1977) "A Network Approach to the Refinement of Social Accounting Matrices," Baltimore, Md.: Johns Hophins University, Department of Mathematical Sciences (processed). Wanhill, S. R. C. (1974) "A Note on the Integration of Project and Sector Analysis," Bulletin of the Oxford Institute of Economics and Statistics, vol. 36, pp. 109-13. World Bank (1979) Yemen Arab Republic, Development of a Traditional Economy, Washington, D.C.:World Bank. I I I I I I -E THE WORLD BANK In his contributionto this volume,Sir Richard Stone,the 1985Nobellaureate, notesthat "of all the interestingand useful things that could be done to improve the national accounts, the one most worthyof consideration is the disaggregation of the householdsector."This statisticalapproach can help in measuring how living standards for different groups change in the process of economic development, and thus it can helpin addressingquestionsabout the distributionof income,the scope of employmentopportunities,and the aleviation of poverty, whichin recent years have cometo the forefront of debateson economicpolicyin developing countries. The establishedtechniquefor capturing the detailsof disaggregatednationalaccountsis the social accountingmatrix, or SAM, in which data are displayedin a single-entrymatrix format, rather than in the traditionalform of double-entrybookkeeping. The data base embodiedin the SAMthen can serve as a statement of initial conditionsin an economyand as a starting point for theoretical analysis of the mechanicsof growth or the likely effects of policies.A particular feature of this approach is that the SAM structure graphicaly demonstrates the interconnectionbetween the distributionof livingstandards and the structure of productionin an economy. Althoughlimitationsand inadequaciesof data may always plague national accounting,a SAMis an invaluabletool for bringing together whatever data are available. The many examplesin this volume of SAMs created for developingcountries show that much can be done even in smaUl countrieswith limitedresources.Moreover,becausethe SAMapproach as illustratedhere is simpler than other systemsof national accounts, it is a potentialy lively and constructivelink between statisticiansand economicplanners-two groups that are all too often observedto have a muted relationship. The elevenchapters in this volumecover three broad areas of concernin socialaccounting.The first four describethe methodologyof SAMsas a disciplinewithineconomicstatisticsand includea comprehensive, nontechnicalintroductionto the subject. The next three recount the experienceof Sri Lanka, Swaziland,and Botswanain constructing,maintaining,and using SAMs.The final four chapters illustrate the step from data systemsto modelsin a SAMcontext.In additionto the editors and Sir Richard Stone, contributors includeCliveBell, ShantayananDevarajan, Coln Greenfield, BenjaminB. King, S. Narapalasingam,Alan R. Roe, Eric Thorbecke,and S. J. Webster. ALSOOF INTEREST Improving the Macroeconomic Data Base: A SAMfor Malaysia, 1970 Graham Pyatt and Jeffery I. Round, assisted by Jane Denes This 1984addition to the World Bank Staff WorkingPaper series (number 646) documentsthe constructionof a social accountingmatrix in Malaysia,with emphasison the statisticalissues that arose and their practicalresolution.The SAMdescribedis a large, disaggregated one, more similar to detailednational accountsthan to the aggregatedformats that are often adopted for macroeco- nomic modelsand analysis.Becausethe final versionof the matrix was reduced in sizefrom initial compilations, the study is able to explorethe implications of working at differentlevelsof statistical detail. The study draws on the wealth of economicdata availablefor Malaysia.It outlinesthe strengths and weaknesses of using market prices as the basis of evaluationof commoditybalancesand gives reasons,in formalalgebraicterms, for preferring this approachto other conventions. The studyalso addresses issuesthat arise in disaggregatinghouseholdaccounts, particularly the reconciliation of national accountsand household survey data, not leastwith regard to savings. ISBN 0-8213-0550-6