LSIV1/21 Living Standards Measurement Study Working Paper No. 121 Infrastructure and Poverty in Viet Nam Dominique van de Walle LSMS Working Papers No. 49 Scott and Amenuvegbe, Sample Designs for the Living Standards Surveys in Ghana and Mauritania/Plans de sondage pour les enquetes sur le niveau de vie au Ghana et en Mauritanie No. 50 Laraki, Food Subsidies: A Case Study of Price Reform in Morocco (also in French, 50F) No. 51 Strauss and Mehra, Child Anthropometry in Cbte d'lvoire: Estimates from Two Surveys, 1985 and 1986 No. 52 van der Gaag, Stelcner, and Vijverberg, Public-Private Sector Wage Comparisons and Moonlighting in Developing Countries: Evidence from Cate d'Ivoire and Peru No. 53 Ainsworth, Socioeconomic Determinants of Fertility in C6te d'Ivoire No. 54 Gertler and Glewwe, The Willingness to Pay for Education in Developing Countries: Evidence from Rural Peru No. 55 Levy and Newman, Rigidit6 des salaires: Donnees microeconomiques et macro&conomiques sur l'ajustement du marche du travail dans le secteur moderne (in French only) No. 56 Glewwe and de Tray, The Poor in Latin America during Adjustment: A Case Study of Peru No. 57 Alderman and Gertler, The Substitutability of Public and Private Health Care for the Treatment of Children in Pakistan No. 58 Rosenhouse, Identifying the Poor: Is "Headship" a Useful Concept? No. 59 Vijverberg, Labor Market Performance as a Determinant of Migration No. 60 Jimenez and Cox, The Relative Effectiveness of Private and Public Schools: Evidence from Two Developing Countries No. 61 Kakwani, Large Sample Distribution of Several Inequality Measures: With Application to C6te d'lvoire No. 62 Kakwani, Testing for Significance of Poverty Differences: With Application to C6te d'Ivoire No. 63 Kakwani, Poverty and Economic Growth: With Application to C6te d'Ivoire No. 64 Moock, Musgrove, and Stelcner, Education and Earnings in Peru's Informal Nonfarm Family Enterprises No. 65 Alderman and Kozel, Formal and informal Sector Wage Determination in Urban Low-Income Neighborhoods in Pakistan No. 66 Vijverberg and van der Gaag, Testing for Labor Market Duality: The Private Wage Sector in C6te d'Ivoire No. 67 King, Does Education Pay in the Labor Market? The Labor Force Participation, Occupation, and Earnings of Peruvian Women No. 68 Kozel, The Composition and Distribution of Income in C6te d'Ivoire No. 69 Deaton, Price Elasticities from Survey Data: Extensions and Indonesian Results No. 70 Glewwe, Efficient Allocation of Transfers to the Poor: The Problem of Unobserved Household Income No. 71 Glewwe, Investigating the Determinants of Household Welfare in C6te d'Ivoire No. 72 Pitt and Rosenzweig, The Selectivity of Fertility and the Determinants of Human Capital Investments: Parametric and Semiparametric Estimates No. 73 Jacoby, Shadow Wages and Peasant Family Labor Supply: An Econometric Application to the Peruvian Sierra No. 74 Behrman, The Action of Human Resources and Poverty on One Another: What We Have Yet to Learn No. 75 Glewwe and Twum-Baah, The Distribution of Welfare in Ghana, 1987-88 No. 76 Glewwe, Schooling, Skills, and the Returns to Government Investment in Education: An Exploration Using Data from Ghana No. 77 Newman, Jorgensen, and Pradhan, Workers' Benefits from Bolivia's Emergency Social Fund No. 78 Vijverberg, Dual Selection Criteria with Multiple Alternatives: Migration, Work Status, and Wages No. 79 Thomas, Gender Differences in Household Resource Allocations No. 80 Grosh, The Household Survey as a Tool for Policy Change: Lessons from the Jamaican Survey of Living Conditions No. 81 Deaton and Paxson, Patterns of Aging in Thailand and Cte d'ivoire No. 82 Ravallion, Does Undernutrition Respond to Incomes and Prices? Dominance Tests for Indonesia No. 83 Ravallion and Datt, Growth and Redistribution Components of Changes in Poverty Measure: A (List continues on the inside back cover) Infrastructure and Poverty in Viet Nam The Living Standards Measurement Study The Living Standards Measurement Study (LSMS) was established by the World Bank in 1980 to explore ways of improving the type and quality of household data collected by statistical offices in developing countries. Its goal is to foster increased use of household data as a basis for policy decisionmaking. Specifically, the LSMS is working to develop new methods to monitor progress in raising levels of living, to identify the consequences for households of past and proposed government policies, and to improve communications between survey statisticians, analysts, and policymakers. The LSMS Working Paper series was started to disseminate intermediate prod- ucts from the LSMs. Publications in the series include critical surveys covering dif- ferent aspects of the LSMs data collection program and reports on improved methodologies for using Living Standards Survey (Lss) data. More recent publica- tions recommend specific survey, questionnaire, and data processing designs and demonstrate the breadth of policy analysis that can be carried out using LSS data. LSMS Working Paper Number 121 Infrastructure and Poverty in Viet Nam Dominique van de Walle The World Bank Washington, D.C. Copyright @ 1996 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 February 1996 To present the results of the Living Standards Measurement Study with the least possible delay, the typescript of this paper has not been prepared in accordance with the procedures appropriate to formal printed texts, and the World Bank accepts no responsibility for errors. Some sources cited in this paper may be informal documents that are not readily available. The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s) and should not be attributed in any manner to the World Bank, to its affiliated organizations, or to members of its Board of Executive Directors or the countries they represent. The World Bank does not guarantee the accuracy of the data included in this publication and accepts no responsibility whatsoever for any consequence of their use. The boundaries, colors, denominations, and other information shown on any map in this volume do not imply on the part of the World Bank Group any judgment on the legal status of any territory or the endorsement or acceptance of such boundaries. The material in this publication is copyrighted. Requests for permission to reproduce portions of it should be sent to the Office of the Publisher at the address shown in the copyright notice above. The World Bank encourages dissemination of its work and will normally give permission promptly and, when the repro- duction is for noncommercial purposes, without asking a fee. Permission to copy portions for classroom use is granted through the Copyright Clearance Center, Inc., Suite 910, 222 Rosewood Drive, Danvers, Massachusetts 01923, U.S.A. The complete backlist of publications from the World Bank is shown in the annual Index of Publications, which contains an alphabetical title list (with full ordering information) and indexes of subjects, authors, and countries and regions. The latest edition is available free of charge from the Distribution Unit, Office of the Publisher, The World Bank, 1818 H Street, N.W., Washington, D.C. 20433, U.S.A., or from Publications, The World Bank, 66, avenue d'I6na, 75116 Paris, France. ISSN: 0253-4517 Dominique van de Walle is an economist in the World Bank's Policy Research Department. Library of Congress Cataloging-in-Publication Data Van de Walle, Dominique. Infrastructure and poverty in Viet Nam / Dominique van de Walle. p. cm. - (LSMS working paper: 121) Includes bibliographical references. ISBN 0-8213-3544-8 1. Infrastructure (Economics)-in Vietnam. 2. Poverty-Vietnam. 3. Vietnam-Economic conditions-Regional disparities. 4. Household surveys-Vietnam. I. World Bank. II. Title. III. Series. HC444.Z9C38 1996 363'.09597-dc2O 95-52159 CIP Contents Foreword ............................................. vii Abstract . ............................................ ix Acknowledgments ....................................... xi 1. Introduction ........................................... 1 2. Poverty and Infrastructure in Viet Nam, 1992-93 .................... 2 2.1 Availability of Physical Infrastructure in Rural Viet Nam ... . . 3 2.2 DrinkingW ater ................................ 5 2.3 Sewerage andSanitation........................... 10 2.4 Access to Irrigation.............................. 13 2.5 Sources of Energy .............................. 15 2.6 Roads ...................................... 18 2.7 Summary andlmplications ......................... 18 3. Explaining Crop Income ...................................20 3.1 Determinants of Crop Income ....................... 20 3.2 The Benefits from Irrigation: Policy Simulations ............ 29 3.3 The Cost of HouseholdLabor ....................... 38 3.4 The Cost of Irrigation Expansion ..................... 39 4. Conclusions ...........................................43 References ................................................45 Tables Table 1: Rural Infrastructure and Poverty in Viet Nam .................... 3 Table 2: Rural Infrastructure in North and South Viet Nam ................. 4 Table 3: Source of Drinking Water in Rural and Urban Areas of North and South Viet Nam(%)........................... 7 Table 4: Source of Drinking Water by Region (%) ....................... 9 Table 5: Toilet Facilities in Rural and Urban Areas of North and South Viet Nam (%) . 12 Table 6: Toilet Facilities by Region(%) ............................. 13 Table 7: Average per Capita Square Meters of Irrigated, Non-irrigated, Other and TotalLand ................................. 14 Table 8: Average per Capita Square Meters of Irrigated, Non-irrigated, Other and Total Land byRegion .......................... 15 Table 9: Lighting Source in Rural and Urban Areas of North and South Viet Nam (%) 16 Table 10: Cooking Fuel in Rural and Urban Areas of North and South Viet Nam (%) 16 Table 11: Variable Definitions and Summary Data ....................... 22 Table 12: Regression Results: Crop Incomes ........................... 26 V Table 13: Marginal Effect on Net Crop Income Allowing for Interaction Effects ..... 28 Table 14: National Distribution of Impacts of Irrigation Under Different Scenarios .... 31 Table 15: R egional D istribution of Per C apita Im pacts of Sim ulation 1 . . . . .. . . . . . 32 Table 16: Regional Distribution of Per Capita Impacts of Simulation 2 ........... 33 Table 17: Regional Distribution of Per Capita Impacts of Simulation 3 ........... 34 Table 18: Regional Distribution of Per Capita Impacts of Simulation 4 ........... 35 Table 19: Regression Results: Family Labor Costs ....................... 40 Table 20: Marginal Effect on Family Labor Costs Allowing for Interaction Effects .... 42 Figures Figure 1: Safe Water Sources in Rural Viet Nam ........................ 8 Figure 2: Sources of Safe Water in VietNam .......................... 8 Figure 3: Sanitation Facilities in Rural Viet Nam ........................ 11 Figure 4: Sanitation Facilities in Urban Viet Nam ........................ 12 Figure 5: Total and Irrigated Annual Land Distribution in Viet Nam, 1992-93 ...... 17 vi Foreword Viet Nam is poor both in terms of household living standards and physical infrastructure. How important are future infrastructural investments likely to be in promoting pro-poor economic growth in Viet Nam? This is an important question for the government and donors alike, as Viet Nam moves through its transition to a market economy. This study uses the Viet Nam Living Standards Survey of 1992-93 to examine the association between household living standards and the level of access to various infrastructural services. It also explores in depth the distributional impact of an expansion in irrigation infrastructure. The paper is part of a larger effort in the Policy Research Department to understand how public spending policies affect household welfare. Lyn Squire, Director Policy Research Department vii Abstract Viet Nam has poor physical infrastructure and high levels of income poverty. What role might better infrastructure play in reducing poverty in Viet Nam? This paper explores the link between poverty and lack of infrastructure using the 1992-93 Viet Nam Living Standards Survey. The household data indicate that although there are some regional and urban-rural imbalances, in general access to infrastructure is not very different between poor and non- poor-infrastructure tends to be bad for everyone. Simulations of the potential benefits from an expansion of irrigation infrastructure and under certain assumptions about how it would be distributed, suggest that the policy would be redistributive, representing proportionately greater gains to the poor. The most pro-poor impacts would occur in Viet Nam's poorest regions and under a policy which targeted irrigation expansion to small per capita landholding households. The average annual economic rate of return of the irrigation investments considered would be at least 20 percent. The paper also finds evidence that various constraints over and above that presented by lack of irrigation appear to diminish the benefits of irrigation to poor and non-poor alike. ix � �� ,.°.э ,оэ � � л � ►гi К� ���� � �� ��� � . �� �� ����.� �� я р��.о�' D э с �а � � л � ���о � Э _� � °' �_ �. ° � О+ � �� � о � qg ' �:v'7d � �' � С � � � �� р' _� О � D• �, � � � х �; � �� �� � ил � о � оOOо �' � v � �. �. о° � tл ;✓ �'' н О 5• й• �' А� Г� А� �G 'д "'� CL '''. и n. о ё л n. � ,.�, •��•' С о 5�' а w�►, си ��•�•s• -..г�' � С с� с� � �а" о `У� � �з � < �b ° � С� 5'°- � '� �ц� �'� � �' R'��@'�°,� �.�������СΡ!'� o�,����гz�5 �ао��иL�й '.в "�f OD :± О. � С .Т ~ �. у • ~ � ;'� �D А �' � � '� б ' • г* G• о �' F+ � � ]�рΡг ►у �i С _ .-. 'v � � 'v ~ О `С ьrj � С �у, п � ...� fD р "�q и А� (9 й � С �. '� `с � � � .ё � ..• � -�i fD � л� � b С н, � �' � �. �ьуΡ „�у,� � ср О � � °' �' � � �' "д �-t Э '."'' � � '��' � � � �j и 3 r � � `t D О '" у � '"� й � � ^оо' �Vj ?.: с, � � � С' �i ~ьrуΡ.{ v' � �, � �• wл � � 5' �рΡ� i� � S с�о � а � �' � r dj' л п b V� i� о ��• (j о < Q' �D �'. � �'�' aQ F• уΡ� � г� � � $ о Pi, Е. � � й• "��. �' о �• р' `"' � � �г, � �• С• й п д'�Q 'r,�' • �'1 5• G 3 '� � р0о �г� �"� �. �'л'�� � � f�D °�й �к:°,С ° � � 5 G �С ��о � 04 о �. о л � �' С. �. � W и < С 5. 5 5''д '"". � v' � С � = а "• �• а, 5' '* с� ��* �� �.���``�� ��'л"��'�s��•5 �5 ��!� � �g� � � � g ,, �, � �' •д а, ,�о ё а, � � � � � $. � � и .,, о � й. � � •, � � о- о �.-. �. � й о� и �, � � � и у .r.�. Ф � ,-• �- � ��р В � � � �.а.� � � 5•С � с.� � �;С � ° °'S•д��,`� с � 3 ��,��Q � ��i ��'туΡ�3`"'�t� о 'b�•�л� �<(рΡ ои w W �' й �, v ^� ь.1� и ^ � р� � � N �� � ГА � � �Q i/� � `.. г* � � 1� �. ... я � � � < � С�� s�a� `"��' ��•� g" ��' я с��25•о 5'�� й 3� и �--� д.е g � � � о � � � �'� � о`� оC � ~ G° ~ bp � и � ом, � � $, � "�'" %'� � �гл OG• �• v'' �• j� � � f�D g � � •�у ` • � '•с7' О CD А� Ф F'° � „ � � а �, с� � о �- '� ,.�г►, �. < � со л � 5' �е С+ �г��� р� �,��'и�..,а,�.о �оCD �.�D �5;�'�<�cr С ,., н , � � о о $, R. < �у с о г.. � 5• �. и � ., �q � о �р �р � г� ������� а�.� �� 5•� .��'° о ��: ц С �� �•?� �� �,'л' о и� а� а� ��.� �� � 5 � �•о � .о`� � � М � �.� � 5' С� ° с<е � "'• С � р а yt" ]" � �р �. С �. . С, � й. � � � � 5'�мS:сR � � sQ � "� ��' � 3' й ГХ' й aq �� 0�0 й С�� Г,' �`е �* 5 �рΡ � S'�* (р л S� ~ о ►.� о л и �-.��,•ы' й b i5 D с� L7 •, а � � °о � b' л �, а: х � � �' � а а < .о �• � � $ о о � с� 1ro 1?' � С. р�• р и„�р,у � v' G3. �' (р р� у� �" й N� ьу '� �'� � �� с Э G �'� а`с� �• �.�'д �, �о' ��о � � �,5 v' � �• � �'' b .-г a�v� � `r �� ° оаго �•и �•�' �.�,,�� � 1 � �"гQj � °•'" � � � � � и а �.. �Q- `гJ � О А � � � � � -о, �'.о�.. `"'�'�''�',� -�i R'� �''Р'�' �'�G.��' ��`' � f''л < �i f"л .� а ►-• �,��уа,сР< �а�оскл �S�оЭfDмS. °°' � � "'�о 0�0��_b w й O �° �с' ►�+�� � ��•о �•С5 к'b W � р � � ~•о� м�а ~ � � �рΡ й• �'j� п < .Q• 5• п� �-�i � � ~ � � � z '.�Т "��' �3. v' � ь�" � � �О � � �.*• R' `л • С (9 о � v' F ��ii ��с�о . о й � OdSQ й ��" 'Г''..г'...�.�.N�o � (D рΡ�� со С• . О. -, � v, � м.. cD fD и и� � Ф� � 5 (р � G 5' � 2. Poverty and Infrastructure in Viet Nam, 1992-93 Except where noted, the analysis is based on the nationally representative 1992-93 Viet Nam Living Standards (VNLSS) survey. The survey covers 4800 households (23,790 persons) of which 3840 (19,094 persons) are rural, and includes a separate questionnaire on the communes in which sampled rural households are found.' Collected information covers a wide spectrum of aspects of living standards. The household survey touches upon access to and usage of infrastructure facilities in the context of its focus on household members' activities, income sources, health, education, housing and so on. 'Me community survey provides detailed information on the availability of infrastructure services in each rural household's commune of residence. It does not cover urban areas. For certain types of infiastructure, the community survey is the sole source of information in the VNLSS. For others, details are also provided at the household level. However, the latter are often conditional on the household's usage and so tend to provide a skewed view of "access". For example, for households who do not report an illness or whose member's illness was not externally treated, the survey reveals nothing about the household's accessibility to health facilities. The commune level data must also be treated with caution. Because communes vary in size and distances differ, the figures do not reveal all that we would ideally like to know about household access to infrastructure services. These data were supplemented by a number of field trips to rural areas of the North, Center and South of Viet Nam during 1993 and 1994. Throughout, the paper uses household consumption expenditure per person as the welfare indicator. Since prices vary spatially, each household's expenditure is deflated by the region specific poverty line relative to the national poverty line.' This provides a measure of household per capita expenditures at what can be termed "all Viet Nam prices". All monetary units are also converted into real values in this way. The analysis is thus based on real expenditure values representing purchasing power parity throughout the country. For the distributional analysis in section 3 and the figures, individuals are ranked by the converted household per capita consumption expenditures and placed into 14 class intervals defined on per capita expenditures. In the following sections, the paper first briefly looks at the general availability of physical infi-astructure in the rural communes to which households belong. It then turns to access to drinking water, sewerage and sanitation, irrigation, energy sources and roads for each household in both urban and rural areas. 2. Expansion factors are not needed as the survey is self-weighted. The community questionnaire relies on interviews of village leaders, health care workers, teachers and local government officials. I Regional poverty lines am estimated based on the Ocost of basic needs' methodology (Ravallion 1994), and detailed in World Bank, 1994c. Deflating by region specific poverty lines is an alternative to using a regional price index. Because the weighing diagram used in deriving poverty lines tends to be more appropriate to the poor than that typically used in spatial price indices, their use is often preferred for investigations concerning poverty. 2 2.1 Availability of Physical Infrastructure in Rural Viet Nam Tables 1 and 2 combine information on household living standards from the VNLSS household level survey with information on infrastructure facilities in each household's commune of residence from the community schedule. As mentioned, this is possible only for rural households. Table 1 shows availability across various household groups for all of rural Viet Nam while Table 2 desegregates this information across North and South household groups. For example, (from first row, Table 1) 70.2% of the population as a whole are estimated to live in communes which have a passable road, while this is true of 74.7% of people living in "non- poor" households and of 67.3% of those living in "poor" households (with consumption per person above and below the poverty line, respectively). Using a lower poverty line (arbitrarily set at close to two thirds of the national poverty line) 72.8 and 62.5 % of those living in non- poor and poor households respectively live in communes with a passable road. Table 1: Rural Infrastructure and Poverty in Viet Nam Percent of rural p,pulation living in communes Headcount with this infrastructure Index of Poverty (% poor High Poverty Line Low Poverty Line among those with this infrastructure) High Poverty Low INFRASTRUCTURE Total Non-Poor Poor Non-Poor Poor Line Poverty Line Passable road 70.2 74.7 67.3 72.8 62.5 58.6 22.3 Passenger transport 52.3 56.2 49.8 54.0 47.3 58.2 22.7 Electricity 43.1 47.2 40.6 45.8 35.3 57.6 20.6 Pipe-borne water 5.2 7.2 3.9 5.5 4.2 45.8 20.3 Post office 34.4 36.2 33.2 34.9 32.9 59.0 24.0 Lower secondary school 87.9 87.7 88.0 88.7 85.4 61.2 24.4 Upper secondary school 9.7 10.7 9.1 10.2 8.3 57.3 21.5 Clinic 93.3 93.6 93.1 93.7 92.1 61.0 24.8 Total 61.1 25.1 Note: The table combines data from the household and community questionnaires. Poor defined by higher poverty line are those with yearly per capita expenditures deflated by regional poverty line which are less than the national poverty line of 1,209,300 Dongs. Poor defined by lower line are those with per capita expenditures deflated by regional poverty line which are less than (0.65)*national poverty line. Electricity is defined as most households in commune have it; pipe-borne water is defined as at least some households have it. Source: 1993 VNLSS. 3 Tabe 2: Rural Infrastructure in Noith and South Vet Nam Percent of rural population living in communes Headcount Index of Poverty with this infrastructure (% poor among those with this High Poverty Line Low Poverty Line infrastructure) High Low INFRASTRUCTURE Total Non-Poor Poor Non-Poor Poor Poverty Line Poverty Line RURAL NORTH Passable road 76.8 89.6 70.4 82.5 62.5 61.1 23.1 Passenger transport 47.2 53.4 44.1 50.0 40.1 62.3 24.1 Electricity 55.9 68.1 49.8 61.1 42.6 59.4 21.6 Pipe-borne water 3.5 6.2 2.2 4.1 2.1 41.9 17.0 Post office 27.7 29.5 26.7 28.4 25.9 64.3 26.6 Lower secondary school 90.6 93.2 89.3 92.5 85.9 65.7 26.9 Upper secondary school 9.3 9.6 9.2 9.4 9.1 66.0 27.8 Clinic 93.9 97.1 92.3 95.3 90.4 65.6 27.3 Total 66.7 28.4 RURAL SOUTH Passable road 58.3 56.5 60.0 57.3 62.4 52.4 20.3 Passenger transport 61.5 59.7 63.2 60.2 66.8 52.3 20.6 Electricity 20.2 21.7 18.8 21.3 15.5 47.4 14.6 Pipe-borne water 8.1 8.4 7.8 7.8 9.7 49.0 22.8 Post office 46.5 44.4 48.6 45.2 52.1 53.2 21.3 Lower secondary school 83.0 81.0 84.9 82.7 84.2 52.1 19.3 Upper secondary school 10.5 12.0 9.1 11.5 6.3 44.1 11.4 Clinic 92.2 89.3 95.0 91.2 96.7 52.4 19.9 Total 50.9 19.0 Note: The table combines data from the household and community questionnaires. Poor defined by higher poverty line are those with per capita expenditures deflated by regional poverty line which are less than national poverty line of 1,209,300 Dongs. Poor defined by lower line are those with per capita expenditures deflated by regional poverty line which are less than (0.65)*national poverty line. Electricity is defined as most households in commune have it; pipe-borne water is defined as at least some households have it. Source: 1993 VNLSS. Infrastructure for social services-schools and clinics-is much more widely accessible than other physical infrastructure such as electricity and water (Tables I and 2). There are clinics in communes accounting for 93% of the total rural population, lower secondary schools in communes covering 88% and primary schools (not reported) exist in every sampled rural commune. Facilities tend to be somewhat more prevalent in the North. Differences between 4 poor and non-poor are not large. Thus, according to the VNLSS, communes tend to be quite well-provisioned in at least basic social services. However, the data also remind us that the quality of social services may leave a lot to be desired. For example, although all surveyed rural communes report having a primary school, 20% of children not attending school say this is because the school is too far; and 64% of communes complain of poor material conditions as the number one problem facing their commune's primary school. Forty-three percent of the rural population live in communes in which "most" households have electricity.! The variation from North (56%) to South (only 20%) is striking. Pipe-borne water is even less frequently present in communes. Only 5% of the rural population reside in communes where at least some households have piped water. This percentage is somewhat higher in the South. The availability of electricity and piped water is also related to living standards, with the poor less likely to make their home in communes where these are obtainable. Particularly for water in the North, headcount indices for households in communes with this infrastructure are considerably lower than for the population at large. Some of these data must be interpreted carefully. For example (as noted), the survey indicates that 70% of the rural population are found in communes serviced by a road which is passable year round. Two caveats should be mentioned. First, in the South, coastal areas and parts of the North, canals and waterways are widely used to transport goods and passengers, so that roads may not be the relevant entity. Second, the survey gives little indication of the quality of the roads or how it defines "passable". Based on casual observation during my field work in rural Viet Nam, it seems likely that being passable by a motorcycle or bicycle may have been sufficient to qualify as "passable". For these reasons, the availability of passenger transport may be a more informative indicator of accessibility. Tables 1 and 2 thus include this variable as a proxy for the presence of a serviceable road or waterway. Around half the population are in communes in which some kind of passenger transport is available. Transport is more frequently found in the South probably reflecting widespread use of boats as well as road vehicles there. There is also a more pronounced difference between poor and non-poor in terms of access to a passable road and transport in the North, indicating the remoteness of some of the poorest households in the North. In the rest of section 2 the household questionnaire is used to further examine access to specific infrastructure services at the household level in both urban and rural areas. 2.2 Drinking Water Over half the population of Viet Nam (52%) secure their drinking water primarily from wells which are not equipped with a pump. A further 20% obtain it mostly from rivers, lakes and other water bodies, while another 11 % rely on rainwater. There are sharp regional and 4. The questionnaire asks whether 'most" or 'just a few' have electricity. 5 urban-rural differences, as well as some variation related to living standards. Table 3 presents percentages of various population subgroups according to their drinking water source. In rural areas the pattern closely follows the national one: wells without pumps are most common, followed by water bodies and rainwater. However, wells are much more prevalent in the North where they are by far the most common source of drinking water (71 %). The population in the rural South rely somewhat more on surface water (41 %) than wells (33%). Public standpipes and private taps, whether inside or outside the residence, are almost non-existent in rural areas. Comparable data for Pakistan, Ghana, Tanzania and Peru indicate that 14 percent or more of these countries' rural populations have access to piped water compared to none in Viet Nam.s Differences between rural poor and non-poor are not large, though the poor almost always have less access to the more desirable sources of drinking water. Figure 1 which plots how use of a water drinking source varies across expenditure per capita groups reinforces this conclusion. Though the slopes show a tendency to slightly incline or decline as living standards rise, on the whole the variation across expenditure groups is not dramatic. Differentials between rich and poor are more pronounced in urban than in rural areas. This is illustrated by Figure 2 which adds the urban pattern to the rural to give the national equivalent of Figure 1. Access to water taps exhibits a steady rise beyond the seventh class interval while use of wells steadily drops from about the sixth. Table 3 shows that the better off are considerably more likely to have access to an inside tap, outside tap or public standpipe than the poor. Still, although piped water systems are limited to urban areas, less than half the total urban population have access to tap water (public or private). Again, this compares poorly with the urban areas of our 4 previous comparator countries, where the lowest access is found in Pakistan at 57% of urban households.' The data clearly show how inadequate these networks continue to be. There are also distinct North-South differences between urban areas. Private indoor taps are more prevalent in the South (44 versus 24%) while publicly provided standpipes are more standard in the North (16 versus 4%). Wells with pumps are also more frequent in the South (10 versus 2%). As in rural areas, wells without pumps are the most common source of drinking water for the North's urban populations (43%). It is striking to note that as much as 14% of the South's urban non-poor population relies on assorted water bodies, while 30% of the poor do so. 5. These countries are chosen as comparators because they have Living Standard Measurement Surveys which follow the same methodologies and ask similar questions. The data on access to piped water in rural areas is the following: Ghana: 13.5 percent of households (1991/92 Ghana Living Standards Survey); Tanzania: more than 21.5% of households (1993/4 Human Resources Development Survey) though this number excludes. obtaining water from a neighbor's piped water supply; Pakistan: 15% of households (Pakistan Integrated Household Survey 1991/2); and Peru: 43% of the rural Sierra population (1991 Living Standard Measurement Survey). 6. Results on access to piped water in urban areas are the following: Ghana:73.5 percent of households (1991/92 Ghana Living Standards Survey); Tanzania: more than 56% of households (1993/4 Human Resources Development Survey) though this number excludes obtaining water from a neighbor's piped water supply; Pakistan: 57% of households (Pakistan Integrated Household Survey 1991/2); and Peru: 91 % of the rural Sierra population (1991 Living Standard Measurement Survey). 6 Table 3: Source of Drinking Water in Rural and Urban Areas of North and South Viet Nam (%) Rural North Rural South Non-poor Poor Total Non-poor Poor Total Private Tap 2.2 0.1 0.8 0.4 0.0 0.2 Public Standpipe 1.3 0.1 0.5 0.1 0.5 0.3 Well w/ pump 2.0 1.4 1.6 10.6 7.6 9.0 Well no pump 68.2 72.2 70.9 31.8 34.1 33.0 River/lake/pond 7.1 15.2 12.5 38.6 43.1 40.9 Rainwater 17.7 9.3 12.1 15.5 11.7 13.6 Other 1.4 1.6 1.5 3.0 3.0 3.0 Urban North Urban South Non-poor Poor Total Non-poor Poor Total Inside.tap 33.2 9.0 24.2 52.9 13.0 44.2 Outside tap 7.9 3.9 6.4 4.1 1.2 3.5 Public standpipe 18.6 11.8 16.1 3.9 2.6 3.6 Wel w/pump 2.4 1.1 1.9 11.3 5.4 10.0 Well no pump 31.1 61.9 42.6 10.3 39.5 16.6 River/lake/pond 1.2 4.0 2.3 13.6 30.3 17.3 Rainwater 4.8 5.3 5.0 3.9 8.0 4.8 Total 100 100 100 100 100 100 Note: The figures are % of persons in each subgroup according to their household's primary source of drinking water. Totals may not add up to 100-remainder is attributable to "other". Private inside and outside taps are aggregated for rural areas. Bottled water is one of the options though it is rare. Source: 1993 VNLSS. 7 figure 1: Safe Water Sources in Rural Viet Nam Population % 70 60 50 40 30 20 10 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Per Capita Expenditure Group 1Wellw/Pump *Wellw/oPump ASurface 4Rain figure 2: Sources of Safe Water in Viet Nam Population % 80 70 60 50 40 30 20 10 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Per Capita Expenditure Group WPvt. Tap 4-Well 0:Surface 44Rain 8 Table 4 presents some of the same information disaggregated across urban and rural areas of the seven geographical regions. Among rural areas, the deltas-particularly the Mekong-stand out. In contrast to all other rural regions where populations overwhelmingly rely on wells without pumps, in the Mekong use of water from rivers and lakes, rain and wells with pumps is higher than elsewhere. In both deltas, over 20% depend on rainwater. A large majority of the urban populations of the Northern Uplands and the North Coast also procure their drinking water from wells (no pumps). It is noteworthy that piped water is found primarily in the Red River, Central Coast and Mekong Delta regions reflecting the better provisioned urban centers of Hanoi, Danang, and Ho Chi Minh City respectively. Table 4: Source of Drinking Water by Region (%) Northern Red North Central Central South Mekong RURAL Uplands River Coast Coast Highlands East Delta Total Well w/pump 0.44 3.5 0.5 0.8 0.0 7.0 11.4 4.3 Well no pump 72.8 54.7 85.5 83.1 79.4 74.2 9.2 57.4 River/lake/pond 18.3 11.5 6.7 13.5 19.2 5.2 58.3 22.6 Rainwater 5.5 25.7 7.0 0.0 0.0 1.0 20.7 12.6 Other 3.0 1.3 0.2 1.6 0.5 12.0 0.0 2.1 URBAN Inside tap 0.0 35.9 0.0 36.4 -- 64.8 15.5 33.8 Outside tap 0.0 12.1 0.0 7.3 -- 2.3 5.2 5.0 Public standpipe 0.0 38.3 0.0 11.1 -- 5.0 1.8 10.1 Well w/pump 1.3 0.8 0.0 3.9 -- 4.4 17.9 5.8 Well no pump 80.0 0.8 100.0 39.0 -- 23.4 7.2 30.1 River/lake/pond 6.5 1.4 0.0 1.0 -- 0.0 41.3 9.5 Rainwater 8.3 9.7 0.0 0.1 -- 0.2 11.1 4.9 Total 100 100 100 100 100 100 100 100 Note: The table gives the percent of persons in each subgroup according to their household's primary source of drinking water. Source: 1993 VNLSS. The VNLSS data also suggest that up to 80% of the population lives within 100 meters of their source of non-piped drinking water and almost all (98%) within 1 kilometer. Close to 79% of all households obtain bath and laundry water from the same source. Almost all the rest also live within a kilometer of their household's source for non-drinking water needs. Though there may be seasonal variations not captured by the VNLSS-the extended dry season may require collection from surface water at greater distances (World Bank 1990)-these data do not 9 lend support to the claim that water collection presents a severe burden on women and children (UNICEF 1994, NEDECO 1993). Finally, it is important to point out that the data tells us nothing about the quality or safety of the obtained drinking water. Piped water is reputed to often go untreated. Wells are for the most part shallow and the water prone to contamination. Indeed, one study found that up to 80% of wells were of unacceptable standard and harbored harmful bacteria though households still preferred them to alternative sources (UNICEF 1994). Given sewerage and sanitation conditions (see below), the presumption must be that surface water is rarely safe for drinking. Evidence on the country's health profile and high incidence of water-related diseases implies severe water and sanitation problems. Viet Nam's National Programme of Action for Children (NPA) reckons that 21% of the rural population have a hygienic and ample water supply, though UNICEF cautions that this is on the high side of most estimates (UNICEF 1994). 2.3 Sewerage and Sanitation Untreated industrial and residential waste waters tend to be dumped straight into existing sewerage systems and waterways. Waste from flush toilets enters sewers directly or via septic tanks and is eventually discharged into rivers and other water bodies. Waste water systems generally manage both flood and waste waters. Other than that provided through septic tanks, there is little treatment (World Bank 1990). For the most part, sewer systems were built pre- 1954 in the North and pre-1975 in the South. Coverage is limited and performance of existing networks is poor. A fraction of the population is serviced by conventional sewerage systems. The VNLSS indicates the style of toilet used by each household. The survey lists flush, double vault composting latrines (DVCL), pit latrines, other, and no toilet as choices. The first three are considered hygienic and desirable relative to other methods by the Ministry of Health. "Other" which includes bucket and fishpond latrines, toilets hanging over water bodies, and animal and human waste manure tanks, are of lower standard and not officially sanctioned. Human wastes are used extensively in agriculture and aquaculture and are a major concern for rural sanitation. A 1989 national study revealed extremely high rates of parasitism-as high as 90-95%-in villages where DVCL are common and excreta used as fertilizer, and a much lower rate-40%-in similar villages where fishpond latrines are prevalent (World Bank 1990). Although DVCLs are hygienic when operated correctly-including allowing sufficient composting time in a sufficiently large vault-in practice, proper use appears to be rare. The NPA's figure for the percent of the rural population with access to adequate sanitation facilities is 13% (UNICEF 1994). Tables 5 and 6 show population subgroups according to the type of toilet used by the household. Nationally, 26% of the population reports having no toilet. This indicates a worse situation than that found in Peru (17%), Tanzania (5.3% of households) and Ghana (25% of 10 households).' A further 33% of the population use pit latrines, 22% use "other" and the rest flush and double vault composting toilets. A tiny percentage have access to flush toilets in rural areas. While not commonly in use anywhere in the South, DVCLs are employed by 13% of the rural North's population and by 21% of its rural non-poor. Pit latrines are by far the most common toilet type in the rural North while other and no toilet are most prevalent in the South. The urban areas of the Central Coast and the South East account for most flush toilets. However, the former also betrays one of the worst waste management situations in the country. A household without access to a flush toilet in the urban Central Coast is most likely to have access to no toilet at all. In its rural hinterland, the percentage of the population without toilet facilities of any kind reaches a national high of 55 %. How household waste management varies across expenditure groups can be seen in Figure 3 for rural areas and in Figure 4 for urban areas. Figure 3 indicates a clear and steady decline in the proportion of the population without recourse to any kind of toilet facility as living standards rise. The use of DVCLs starts at a very low rate and rises with welfare levels, as does the use of flush toilets. Interestingly, pit latrines exhibit more of a flat inverted u shape. These are used primarily in the mountainous areas of the North. The graph reflects the fact that the poorest in these regions probably have worse access than the less poor, while less of the higher expenditure population is found in these areas. In urban Viet Nam, patterns are less clear except in the case of flush toilets which exhibit a steady increase as per capita expenditures rise. Figure 3: Sanitation Facilities in Rural Viet Nam Population % 60 50 40 30 20 10 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Per Capita Expenditure Group jmFlush *DVCL *Pit -None 7. Sources as detailed in footnote 5. 11 Figure 4: Sanitation Facilities in Urban Viet Nam Population % 100 80 60 40 20 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Per Capita Expenditure Group UFlush #DVCL -&Pit -None Table 5: Toilet Facilities in Rural and Urban Areas of North and South Viet Nam (%) Rural North Rural South Toilet Type Non-poor Poor Total Non-poor Poor Total Flush 4.3 0.5 1.8 3.8 0.8 2.2 DVCL 21.4 8.2 12.6 0.7 0.2 0.4 Pit latrine 47.2 44.8 45.6 21.3 18.4 19.8 Other 7.9 16.8 13.8 51.3 39.7 45.4 No toilet 19.2 29.7 26.2 22.9 41.0 321 Urban North Urban South Toilet type Non-poor Poor Total Non-poor Poor Total Flush 48.1 21.6 38.2 67.7 20.0 57.3 DVCL 11.1 12.4 11.6 1.9 2.0 1.9 Pit latrine 23.6 30.8 26.3 6.3 32.3 12.0 Other 0.8 3.3 1.7 13.3 29.9 16.9 No toilet 16.3 31.9 22.1 10.7 15.8 11.8 Total 100 100 100 100 100 100 Note: The table gives the percent of persons in each subgroup according to the toilet facility used by their household. DVCL is a double vault compost latrine. Source: 1993 VNLSS. 12 Table 6: Toilet Facilities by Region (%) Northern Red North Central Central South Mekong RURAL Uplands River Coast Coast Highlands East Delta Total Flush 0.1 1.0 0.8 7.3 2.1 5.9 0.9 1.9 DVCL 11.4 13.3 12.0 13.7 2.2 0.2 0.2 8.3 Pit latrine 58.8 54.6 34.3 20.2 53.2 40.2 6.2 36.4 Other 6.5 16.2 25.4 4.2 2.0 18.3 63.3 25.1 No toilet 23.2 14.8 27.4 54.6 40.4 35.4 29.4 28.3 URBAN Flush 2.0 50.4 3.8 60.7 -- 82.8 21.9 47.4 DVCL 10.5 8.1 42.4 6.0 -- 0.00 4.6 6.9 Pit latrine 64.8 19.7 31.1 6.4 -- 14.4 8.7 19.4 Other 0.9 0.6 11.0 0.5 -- 0.00 40.5 9.1 No toilet 21.8 21.2 11.7 26.4 -- 2.9 24.3 17.2 Total 100 100 100 100 100 100 100 100 Note: The table gives % of persons in each subgroup according to the toilet facility used by their household. DVCL is a double vault compost latrine. Source: 1993 VNLSS. 2.4 Access to Irrigation Around half of Viet Nam's annual crop cultivation area is currently irrigated. Irrigation needs are largely supplied by surface water (Vu and Taillard 1993). Irrigation includes both large-scale (networked investments) and small-scale (wells, bore-holes) systems. The two major river deltas are characterized by complex hydraulic systems dating back hundreds of years which incorporate navigation, flood control, drainage, saline intrusion control and irrigation functions. The latter rely on a system of canals with pumping stations and on-farm water control arrangements. These networks are in a state of severe disrepair. It is estimated that by rehabilitation of the existing infrastructure alone, there is the potential to expand irrigation to some 700,000 hectares assuming there is sufficient water (SPC et al. 1989). Outside the deltas, irrigation-like other physical infrastructure-is less well developed, though there are expansion possibilities. In some areas, this requires storage dam construction and gravity irrigation systems. In others the development of small scale irrigation systems such as based on small electric pumps drawing water from reservoirs and natural water bodies is more feasible. 13 Some features of the distribution of irrigation infrastructure based on information from the VNLSS can be seen in Tables 7 and 8, and in Figure 5.8 The tables present mean per capita square meters of total land and its various components, including irrigated and non- irrigated annual crop land and perennial land, across poor and non-poor groups (Table 7) and across Viet Nam's seven geographical regions (Table 8). Differences between the rural North and South are the most striking feature of Table 7. First, overall per person land amounts are quite a bit higher in the South as are all components with the exception of "other" land. But, the most notable difference is in the variation between poor and non-poor land endowments. These are relatively equitable in the North, though the poor have less irrigated land. However, these differences are dwarfed by the disparities in the South. On average, the Southern poor have access to less than half the amount of land that the non-poor control. Table 8 reveals some of the geographical variation-attributable in large part to environmental terrain and variable population densities. Average total land cultivated per person varies from 702 m2 in the Red River Delta to 1977 m2 in the Mekong Delta, while irrigated land ranges from 17 in the Central Highlands to 713 M2 in the Mekong River Delta. Figure 5 combines the geographical and distributional aspects of land allocation. Total land and total irrigated annual land are graphed separately (both in square meters per capita) by per capita expenditure groups for all of rural Viet Nam and individually for the 7 regions. Past land reform has ensured relatively low intra- regional inequality of access to land (Vu and Taillard 1993). Significant variation in farm size and land quality can be found between regions. Landholdings are more correlated with living standards in some regions. In particular, all three regions of the South reveal a more pronounced positive association of land size with per capita expenditure levels. In general the distribution of irrigation appears to be more equitable or about the same as that of total land. Based on the tables and Figure 5, there appears to be considerable room for expanding irrigation. Table 7. Average per Capita Square Meters of Irrigated, Non-Irrigated, Other and Total Land Rural North Rural South National Non-poor Poor Total Non-poor Poor Total Non-poor Poor Total Irrigated annual land 414.8 333.4 360.1 825.9 346.0 584.2 590.3 336.8 434.0 Non-irrigated annual 288.9 378.2 348.9 1149.8 660.9 903.5 656.6 454.4 531.9 land Perennial land 55.0 50.8 52.2 373.9 212.8 292.7 191.2 94.4 131.5 Other land 173.8 126.1 141.8 156.0 29.7 92.4 166.2 100.1 125.5 Total land 932.4 888.6 902.9 2505.7 1249.4 1872.8 1604.3 985.8 1222.9 Note: Per capita m' of land are calculated over the rural farm population. Other land includes forest, water surface, and "other" as defined in footnote 10. Source: 1993 VNLSS. 8. According to the interviewer's instruction manual, irrigated land in the VNLSS includes all land which is irrigated either through a system of canals or by means of electric or petrol pumps which prevent flood and drought. 14 Table 8: Average per Capita Square Meters of Irrigated, Non-Irrigated, Other and Total Land by Region Northern Red North Central Central South Mekong Rural Uplands River Coast Coast Highlands East Delta Total Irrigated 229.3 521.2 307.9 325.9 17.1 484.5 713.1 434.0 annual land Non-irrigated 697.2 78.7 349.3 321.2 1015.5 823.1 905.9 531.9 annual land Perennial land 76.7 31.8 58.9 42.8 398.3 354.9 256.6 131.5 Other land 311.0 69.8 112.8 50.3 27.2 101.3 101.5 125.5 Total Land 1314.2 701.5 828.8 740.2 1458.0 1763.7 1977.2 1222.9 Note: Per capita m' of land are calculated over the rural farm population. Other land includes forest, water surface, and "other" as defined in footnote 10. Source: 1993 VNLSS. 2.5 Sources of Energy The North is generally better endowed in energy, with coal reserves and hydroelectric power plants, while the South is well known for its electricity shortages. Throughout Viet Nam, electricity is also available through diesel operated generators generally run by local authorities. Population percentages by lighting source are given in Table 9. Two methods dominate: electricity is used by 49% of the national population while 50% utilize gas, oil and kerosene lamps. There are pronounced differences between the North and South and across expenditure groups. Electricity networks are better developed and reputed to be more reliable in North Viet Nam and this is reflected in household usage. Practically half of the North's rural population rely on electricity compared to less than a quarter (22%) in the rural South. For the non-poor, this rises to 64% in the North and 27% in the South. The rest make use of gas, oil and kerosene accounting for 61 and 80% of the poor and 34 and 68% of non-poor, in the North and South respectively. In urban Viet Nam a large majority of the population depends on electric lighting though the proportions are smaller in the South than in the North and among the less well off than the better-off. Again, comparison with countries for which similar data exists can help place the Viet Nam numbers in context. On the whole, household access to electricity in Viet Nam compares favorably to that in both Ghana (69% of urban and 9% rural households) and Tanzania (35% and 1 %) but less well to Peru (95% of the total population). Electricity is rarely used for cooking. Table 10 indicates that wood and leaves predominate in the rural areas of the North and wood dominates in the South's rural sector, while coal and kerosene also take on importance in the urban household sector. The differences between poor and non-poor rural groups are small. In the urban areas of both parts of the country however, the poor are more likely to use wood and the better off to use coal or kerosene. 15 Table 9: Lighting Source in Rural and Urban Areas of North and South Viet Nam (%) Rural North Rural South Non-poor Poor Total Non-poor Poor Total Electricity 63.9 38.0 46.6 27.4 16.4 21.8 Battery 1.0 0.4 0.6 4.9 0.6 2.7 Gas/oil/kerosene 34.2 60.9 52.0 67.7 80.3 74.1 Other 1.0 0.7 0.8 0.0 2.7 1.4 Urban North Urban South Non-poor Poor Total Non-poor Poor Total Electricity 97.0 80.9 91.0 91.0 64.9 85.3 Gas/oil/kerosene 3.0 19.1 9.0 9.0 35.1 14.7 Total 100 100 100 100 100 100 Note: The table gives % of persons in each subgroup according to their household's source of lighting. Source: 1993 VNLSS. Table 10: Cooking Fuel in Rural and Urban Areas of North and South Viet Nam (%) Rural North Rural South Non-poor Poor Total Non-poor Poor Total Wood 39.3 46.0 43.8 86.7 88.3 87.5 Leaves 51.1 52.7 52.1 9.3 11.3 10.3 Coal 8.9 1.3 3.8 3.7 0.4 2.0 Urban North Urban South Non-poor Poor Total Non-poor Poor Total Wood 34.7 50.7 40.7 57.9 94.6 65.9 Leaves 5.0 16.1 9.1 2.2 3.2 2.5 Coal 48.7 31.7 42.3 23.8 0.8 18.7 Electricity 3.9 0.4 2.6 1.9 0.0 1.5 Kerosene 7.7 1.1 5.2 13.7 1.4 11.0 Total 100 100 100 100 100 100 Note: The table gives % of persons in each subgroup according to their household's cooking fuel. Totals may not add up to 100. The remainder is attributable to "other" and kerosene and electricity in rural areas, and to other and bottled gas in urban areas. Source: 1993 VNLSS. 16 Figure 5: Total and Irrigated Annual Land Distribution in Viet Namn, 1992-93 (&l/person) Northern Uplands Red River Delta South East North Central Central Coast National ii taiå Pe CaiaEpndtr ru 2.6 Roads Other than the information (from the community schedule) of whether a road passes through the commune, there is little information in the VNLSS to illustrate the poor state of the country's transportation sector. The public road network consists of around 105,100 km of roads: 11,400 of national roads, 14,200 of provincial level roads, 25,300 of district level roads, 46,200 of village roads and 2,600 urban roads and 5,400 special roads (World Bank 1994b). In 1992, around 12 percent of Viet Nam's existing road network was paved compared to 30% of India's in 1985 and 48% of Indonesia's in 1990 (World Bank 1994b). Average road density is quite low at 0.32 km per sq km of land area and 1.6 km per 1000 inhabitants. Not unexpectedly, densities are highest in the two deltas and lowest in the more mountainous regions. Together with the rest of the infrastructure stock in Viet Nam, the road network-dating largely from before the 1970s in the South and pre-1954 in the North-is old and in severe disrepair. This is also true of other transport infrastructure including inland waterways (a 40,000 km network), ports, and the railway system (World Bank 1994b). 2.7 Summary and Implications The current state of physical infrastructure in Viet Nam is clearly poor by most standards. Nearly a third of the rural population live in communes without a passable road. Nearly half do not have access to passenger transport. More than half do not have electricity. Barely half of annual crop land is irrigated. All but 5% of the rural population live in communes where no one has access to piped water. There are also marked differences in access to infrastructure between urban and rural areas with urban areas generally favored, as well as considerable regional imbalances. The poor tend to have worse access to infrastructure than the non-poor. However, for many types of infrastructure the poor in rural Viet Nam do not have appreciably worse access than do the non-poor: many types of basic infrastructure are equally bad for both. Basic infrastructure ventures will not automatically be redistributive. It cannot be argued that the non- poor already have adequate basic infrastructure and the poor have none such that new investments will necessarily benefit the poor. What do these data imply for the distributional impacts of future investments in infrastructure? The answer will depend on the marginal benefits from infrastructure investments. For example, take irrigation. If a household's land is fully irrigated then clearly the marginal benefits to that household from expanding total irrigated area are zero. Those who benefit from a general expansion of irrigation, say, will be those who have non-irrigated land. If it were true that the rich have fully irrigated land while the poor don't, then the benefits would go to the poor. However, if anything it is the non-poor who have more of both non-irrigated and irrigated land (Table 7 and Figure 5). Looking at Figure 5 one would not think that undifferentiated irrigation infrastructure would be an important redistributive instrument as such; the poor would 18 benefit but probably less so than the non-poor.' There are, however, a number of other factors which are correlated with the marginal benefits from irrigation. It is often argued for example that smaller farms are more productive so that marginal benefits may be higher for the poor. Or it might be argued that marginal benefits will tend to be higher for those with more human capital, likely to be the not-so-poor. On balance, it is not clear what the outcome would be. Inferring the potential gains from irrigation using cross-sectional data thus requires controls for these other factors. To properly address such questions we need to go beyond the simply descriptive analysis and investigate marginal impacts. 9. It may be a different question if one could adequately target irrigation to the poor, but targeting can often be difficult and costly (van de Walle 1995a). 19 3. Explaining Crop Income This section attempts a detailed assessment of the likely distributional impacts of an expansion in the irrigated land area. The data set contains detailed information on household land assets and farm incomes. The vast majority of Viet Nam's population derives its livelihoods from farming (van de Walle 1995b). The extent to which annual crop land is irrigated and water adequately managed are widely recognized as key factors in the productivity and success of agriculture at the local level. Although the aggregates hide considerable regional variation, around one half of agricultural annual crop land is currently under irrigation.'o For these reasons, it is possible to look closely at the impact of irrigation infrastructure on livelihoods in rural Viet Nam. How large would the gains in incomes from further irrigation investments be? Would those gains be pro-poor? Such questions are of considerable interest given the multiple policy choices faced by policy makers in a country setting where budgets are heavily constrained and simultaneously challenged by a plethora of real investment and consumption needs. 3.1 Determinants of Crop Income In attempting to throw some light on these questions the paper looks at the determinants of net farm crop income and the role played by irrigation. The size of the difference in marginal returns between irrigated and non-irrigated land determines the income gains from irrigating a unit of land. To quantify the gains from irrigation the paper posits the profit function (p, L N L', z) giving the farm-household's maximum profit conditional on a vector of prevailing output and input prices (p), amounts of non-irrigated (LN) and irrigated (L') annual crop land, and a vector of other fixed factors (z) including other types of agricultural production land, human capital, location specific agro-ecological variables and other constraints arising from market imperfections (such as the underdeveloped state of labor markets and supervision costs of labor). In specifying z a wider range of variables are allowed than one would normally posit in a profit function, recognizing that this is a transition economy in which markets are still underdeveloped. For example, in many parts of Viet Nam household demographic factors can be an important constraint on production arising out of labor-market imperfections and institutionalized non-market modes of factor allocation. 10. In addition to annual crop land, households derive agricultural incomes from perennial land (used for perennial tree crops), forest land (natural forest or reforested areas used for inseminating of young plants and growing of forest tree crops), water surface land (for raising water products) and what I will call other land. The latter includes various other land categories listed in the survey: vacant lots and bald hills (land managed by household but not cultivated for at least 12 months); virgin land (burnt and fallow land); and other (area of road and dike sides, river banks, etc). 20 Output and variable input prices are assumed to vary between but not within communes, so these can be represented by a vector of commune dummy variables. For the j household, the profit function is assumed to be linearized as follows: (1) i = (p, LjN Lj, z) = a + [ PL + yL ' z +8d +ej where the marginal returns to non-irrigated and irrigated land are given by (2) N = + N d and respectively, and where d is a vector of regional dummy variables. The error term in (1) is assumed to be independently and identically normally distributed. The regression includes a full set of commune dummy variables which are meant to pick up prices as well as any spatial variations in other omitted or fixed factors, and should also capture the influences of community level characteristics, including geographical and infrastructural variations. These spatial effects are compressed into the 7 regional entities in their effects on marginal returns, though a full set of commune dummies is allowed in the intercept of the profit function. Profit is measured by net farm crop income, net of variable costs." Table 11 lists the right hand-side variables and provides a description and summary statistics. 11. Total revenue from agricultural land production includes all crops evaluated at harvest prices (missing values are replaced by average community prices); the value of crop byproducts consumed or sold (such as thatch, straw, trunks of cassava, maize, jute, etc.); land incomes (rents from other households and government assistance); and income from leasing out farm production equipment. The total costs of production are then subtracted. These include hired labor expenses; costs of seeds and young plants; fertilizer, manure and insecticide costs; animal rental, transport, packaging and storage, equipment rental, repair and maintenance fees, fuel oil and electricity costs; an accounting depreciation charge for owned farming equipment (5%); taxes and fees to cooperative (such as for irrigation, crop protection, plowing); land payments such as rent for land leased in and land taxes paid to government or cooperative. Transformation of home grown crops (such as producing cured tobacco, peanut oil, rice noodles) is not included, but treated as family off-farm enterprises using farm inputs. Note also that the opportunity costs of household farm labor are not included. Section 3.3 provides a test of whether the results are sensitive to this assumption. Also note that profits from livestock raising are not part of net crop income. 21 Table 11: Variable Definitions and Summary Data Variable Definitions cropinc Net household crop income, 1993 Dongs sick Dummy for household member being sick in last year sexhhh Gender of household head hhsvgs Initial stock of household savings, 1993 Dongs hhsize Size of the household propO6 Proportion of household members who are 6 years and younger prop716 Proportion of household members who are 7 to 16 years, inclusive pfadlt Proportion of household members who are female adults (17 +) pmadlt Proportion of household members that are male adults (17 +) hedl Years of primary education of household head hed2 Years of post-primary education of household head oed1 Years of primary education of other adult household members (17 +) oed2 Years of post-primary education of other adult household members (17 +) irrigated Irrigated annual crop land area (mi) nonirrigated Non-irrigated annual crop land area (mi) perennial Perennial land area (mi) forest Forest land area (mi) waterland Water surface land area (mi) otherland Other land area (m) proplt Proportion of annual land which is long-term propauct Proportion of annual land which is auctioned proppriv Proportion of annual land which is private propshare Proportion of annual land which is sharecropped/rented propall Proportion of annual land which is allocated urban Dummy variable for urban residence nu Dummy variable for the Northern Uplands region rr Dummy variable for the Red River Delta region nc Dummy variable for the North Coast region cc Dummy variable for the Central Coast region ch Dummy variable for the Central Highlands region mk Dummy variable for the Mekong River Delta region 22 Table 11 (continued) Summary Data Variable Obs Mean Std. Dev. Min Max cropinc 3049 2282069 2391173 -6061975 2.63e+07 sick 3049 0.933 0.250 0 1 sexhhh 3049 0.809 0.393 0 1 bhsvgs 3049 939223 8990056 0 4.66e+08 hhsize 3049 5.033 1.992 1 15 prop06 3049 0.156 0.176 0 .75 prop716 3049 0.213 0.204 0 .80 pfadlt 3049 0.327 0.169 0 1 pmadlt 3049 0.282 0.160 0 1 hed1 3049 4.379 1.114 0 5 hed2 3049 2.513 2.842 0 16 oedl 3049 6.872 5.372 0 42 oed2 3049 4.111 5.287 0 51 irrigated 3049 2267.58 3997.50 0 80000 nonirrigated 3049 2605.92 5632.34 0 97150 perennial 3049 678.43 2169.50 0 50000 forest 3049 279.22 1970.98 0 50000 waterland 3049 122.89 1203.53 0 50000 otherland 3049 217.50 2106.11 0 70000 proplt 3049 0.20 0.380 0 1 propauct 3049 0.023 0.092 0 1 proppriv 3049 0.227 0.341 0 1 propshare 3049 0.043 0.165 0 1 propall 3049 0.507 0.431 0 1 urban 3049 0.057 0.231 0 1 nu 3049 0.183 0.387 0 1 rr 3049 0.275 0.447 0 1 nc 3049 0.178 0.383 0 1 cc 3049 0.090 0.286 0 1 ch 3049 0.020 0.139 0 1 mk 3049 0.20 0.40 0 1 23 Explanatory variables aiming to capture the influence of household characteristics include household size and composition; gender of the household head; years of primary school education (0 to 5) and of any additional education of the household head; ditto for all other adult household members (aged over 17)12; access to various kinds of land"; proportions of annual crop land in various forms of ownership"; the stock of household savings, and dummy variables for urban residence and whether a household member was sick in the last year. For the most part, land is not allocated to households through the market mechanism in Viet Nam. Thus the usual concerns about regressing outputs on inputs chosen by the household (and hence endogenous) do not arise in this setting with respect to land. And although there may be endogeneity of placement regionally for irrigation, the existence of irrigation in an area can be treated as exogenous at the household level given that there has been little to no mobility in the country. However that does not mean that there are not other, more subtle, forms of endogeneity. Over a long period, land has traditionally been allocated by administrative fiat. There may be some omitted variable in the model which also influences the amount of non- irrigated and irrigated land allocated to the household. Then land will be correlated with the error term in the regression giving a bias. There is nothing that can be done about this more subtle potential form of endogeneity problem in a cross-section data set such as the VNLSS. A number of functional form specifications were tested including linear, semi-log, and double log forms, with and without quadratics in land and education. Explanatory variables were also tested logged and unlogged. OLS is used on a regression sample consisting of all farm households (including some urban households who farm) for which the data are complete (3049 households). Results are reported in Table 12. Two regressions are presented: one, referred to as the unrestricted model, contains all variables, while the restricted model is the outcome of pruning variables with t-ratios below 1 in the unrestricted model and following 12. The education of school age children is omitted to avoid possible endogeneity problems. The latter could result if, for example, households with unobserved factors contributing to higher farm profits are more likely to pull their children out of school. 13. See footnote 10. 14. It may be important to distinguish between land ownership rights. During survey implementation and before the new land law of late 1993, land was classified into 5 types: i) Allocated: (applicable in the North) land from the cooperative's land fund which was distributed to households according to number of workers. ii) Auctioned: (applicable in the North) around 5 to 10 % of cooperative's land which was reserved for bidding by those who wanted more land. This land was more expensive and had a 3 to 5 year tenure depending on the region. iii) Long term use: the South's equivalent of the North's allocated land. iv) Private: land used by the household as a garden area. Often of lower quality, this land required no payment. v) Sharecropped or rented. 15. In addition, cash income received from the Government Social Fund, dummy variables for the household head's ethnicity, age, religion, language, and whether born in present residence were tried and found to be insignificant and to have no effect on the other regression coefficients. 24 iterations. Note however, that in an effort to make Table 12 more wieldy, variables with t-ratios below 1 and the 119 commune dummy variables are not reported.'" Because of the many interaction effects, the impact of individual variables is difficult to assess directly from the regression coefficients. Table 13 presents the calculated total marginal effects and t-statistics of such variables evaluated at the mean points. Turning to this table first, a few things are worth noting. Annual land, both irrigated and non-irrigated, and perennial land all have high significant positive overall effects on crop income. The highest is from irrigated land, an impact which is more than twice as large as that of non-irrigated land. There are also high returns to perennial land (higher than to non-irrigated annual land). Returns to forest, water surface, and other land are much lower, though only water surface land in the restricted model is statistically significant. All education variables are found to have pronounced and significant positive impacts on crop income. In particular, one extra year of primary education for the head of household increases crop income by an amount equal to about 8 percent of mean crop income. There are decreasing returns to the education of the household head (the coefficient on hedl is much greater than that on hed2), but not to that of the rest of the household (coefficients on oedl and oed2 are roughly similar). Finally, larger households have higher crop incomes. This implies that family labor endowments matter in agricultural production probably because labor markets are underdeveloped. The marginal effects of demographic composition variables are not statistically significant. The regression shows strong, though diminishing, impacts of annual crop land-both irrigated and non-irrigated-on crop income (Table 12). A higher share of allocated or auctioned annual land significantly increases annual crop profits. Household size matters, though composition effects do not appear to be of consequence independently of the interaction with land. However, many of the interaction effects are significant and of interest. For example, household size appears important in its interactions with nearly all land variables (all are significant in the restricted model), as does the share of females in household adults. These effects are positive in most cases and demonstrate the importance of family labor inputs, and particularly female ones." They suggest a dependency on own household labor, probably indicating the presence of labor market imperfections and the inability of many households to hire labor in or out. In contrast, interacting household size and the female adult share with irrigated land, and the female adult share with non-irrigated land, results in a significant negative impact. A careful investigation finds that when the sample is partitioned across geographical regions, this effect holds only in the South, and particularly in the Mekong delta, though it is clearly strong enough to influence estimation results for the national model. 16. Full regression details are available from the author. 17. Women play a major role in agriculture in much of Viet Nam. The VNLSS indicates that women averaged the equivalent of 182.5 8 hour days work on the family farm and men 159.4 days. 25 Table 12: Regression Results: Crop Incomes Unrestricted Model Restricted Model cropinc Coefficient t-ratio Coefficient t-ratio urban 1093640 1.08 837371 1.25 sick -318465.4 2.59 -317534.3 2.61 hhsize 81451.7 2.12 67405.6 2.61 prop06 -586514.9 1.18 -429757.1 2.08 hedl -468646.3 2.69 -527229.5 3.12 hedl*hed1 67862.6 2.66 76644.0 3.13 oedl*oed1 -1932.7 2.48 -1566.7 3.22 oed2 21023.8 1.24 27026.3 2.85 irrigated 352.40 4.27 362.59 4.78 irr*irr -0.0030 4.42 -0.0030 4.82 nonirrigated 238.40 3.81 206.72 8.20 nonirrig*nonirrig -0.0036 9.60 -0.0034 10.63 perennial -277.04 1.73 -238.17 2.11 perennial*perennial -0.0097 6.43 -0.0099 7.10 forest -372.88 1.35 -80.33 1.02 forest*forest -0.0026 1.38 -0.0022 2.0 waterland*waterland -0.0401 3.80 -0.0042 5.12 otherland -611.27 1.22 -426.07 2.16 otherland*otherland -0.0024 1.32 -0.0016 1.19 propauct 1116555 2.54 1048419 2.62 proppriv 325505.5 1.49 215568 1.50 propall 470198.4 2.10 337486.1 2.07 hedl*irrigated 47.87 6.06 49.80 6.93 hedl*otherland -113.39 2.58 -111.97 3.40 hed2*irrigated -6.46 1.53 -5.10 1.46 hed2*perennial 21.90 2.53 25.85 4.09 hed2*forest 23.01 1.62 26.73 3.23 hed2*waterland 72.61 1.58 30.70 2.12 hed2*otherland 33.45 1.51 22.67 1.14 oedl*irrigated 20.74 8.03 20.74 8.58 oedl*nonirrigated 7.27 3.38 5.66 3.51 oedl*perennial 5.42 1.21 5.10 1.20 oedl*forest -21.37 1.72 -12.80 3.20 oedl*otherland -49.20 4.66 -39.95 4.48 oed2*irrigated -4.179 2.18 -4.57 2.52 oed2*nonirrigated 1.741 1.04 1.990 1.34 oed2*perennial -10.914 2.65 -10.694 2.76 oed2*otherland 33.814 4.0 25.259 3.60 hhsize*irrigated -35.991 6.94 -35.865 7.51 hhsize*nonirrigated 4.639 1.13 7.243 2.22 26 Table 12 (continued) Unresnicted Model Restricted Model cropinc Coefficient t-ratio Coefficient t-ratio hhsize*perennial 52.933 4.62 51.712 4.79 hhsize*forest 37.473 1.68 28.892 2.38 hhsize*otherland 79.081 2.62 64.675 2.28 pfadlt*irrigated -176.63 2.16 -189.17 2.55 pfadlt*nonirrigated -137.02 2.07 -115.10 2.30 pfadlt*perennial 610.10 3.33 628.69 4.12 pfadlt*otherland 1941.44 4.18 1751.47 4.48 pmadlt*irrigated -162.40 1.71 -142.17 1.70 pmadlt*perennial 289.39 1.92 290.70 2.63 prop716*irrigated 155.85 2.03 132.86 1.94 rr*irrigated 271.75 4.06 260.97 4.17 rr*forest 135.35 1.03 74.85 1.19 mk*irrigated -67.71 1.94 -83.64 2.65 mk*perennial -158.92 2.94 -147.73 2.95 nu*irrigated 255.81 3.47 241.34 3.43 nu*perennial -199.58 2.27 -215.25 2.77 nu*otherland 434.86 1.20 361.41 5.02 nc*perennial -218.53 3.02 -205.53 2.96 nc*otherland 528.01 1.39 480.14 3.09 cc*irrigated -203.38 3.63 -211.93 3.97 cc*nonirrigated -152.25 2.62 -147.69 3.03 cc*perennial -228.68 1.26 -226.31 1.26 ch*irrigated -973.79 1.66 -1051.07 1.81 ch*nonirrigated -134.37 2.65 -130.90 3.29 ch*perennial 310.57 4.37 326.09 4.96 ch*waterland 5195.78 1.31 4078.46 2.70 Number of obs = 3049 Number of obs = 3049 F(233, 2815) = 19.06 F(183, 2865) = 24.04 Prob > F = 0.0000 Prob > F = 0.0000 R-square = 0.6120 R-square = 0.6057 Adj R-square = 0.5799 Adj R-square = 0.5805 Root MSE = 1.5e+06 Root MSE = 1.5e+06 Note: The restricted model results from the pruning of all variables with t-ratios less than I in the unrestricted model. The unrestricted model also contained the following variables: demographic composition variables, pnum716, pfadit, pmadlt and interactions with land variables; education variables: hed2, hed22, oedl, oed22 and interactions with land; land: waterland and interactions between types of land and regions; propit, propshare. 27 Table 13: Marginal Effect on Net Crop Income Allowing for Interaction Effects Unrestricted model Restricted Model Marginal effect on net crop Marginal effect on Variable income t-ratio net crop income t-ratio irrigated annual land Dongs/100m 48,571.5 16.1 48,226.3 17.9 non-irrigated annual Dongs/100m 19,994.0 8.1 21,876.2 16.3 land perennial land Dongs/100m 21,269.1 4.1 23,385.0 6.7 forest land Dongs/100m2 8,722.1 1.9 6,325.0 1.8 water surface land Dongs/100m -86,491.2 0.1 15,729.2 3.3 other land Dongs/100m' 10,422.2 1.2 2,346.8 0.4 household size Dongs/person 59,065.9 2.0 62,154.9 2.8 prop female adults Dongs/% -2,366.5 0.1 78,456.6 0.4 point prop male adults Dongs/% -1,165.1 0.2 -125,228.9 0.6 point prop aged 7-16 Dongs/% 1,041.6 0.2 301,322.2 1.9 point primary ed (head) Dongs/year 191,875.8 3.0 232,762.7 4.1 ed > primary (head) Dongs/year 38,584.9 2.0 22,132.4 2.3 primary ed (other Dongslyear 35,094.1 2.5 31,466.8 4.7 adults) ed > primary (other Dongs/year 22,195.7 1.8 20,094.3 2.6 adults) mean crop income 2,282,069 2,282,069 Note: Marginal effects are evaluated at mean points. The latter results can be interpreted to indicate that the market labor constraint does not bite as much for irrigated land in the South. For households with larger amounts of irrigated land, family labor becomes less of a constraint. It is the way in which household labor influences crop income which is important here. If the household could buy or sell as much labor time as it required then one would not expect household demographics to be significant in the crop income equation. The fact that they are significant can then be taken as an implication of labor market failure. Family labor becomes an input to production but the extent to which this matters depends on how much market conditions apply to each household. The results indicate that family labor is generally a constraining factor in farm production in the 28 North, but less so in the South and particularly less so for households with lots of irrigated land in the Mekong delta." A test of the linear restriction that the overall influence of household size is zero when evaluated at mean sample values is not rejected for the South (though the number is positive), but it is found to be positive and significant in the North. (The same is found for the share of female adults.) Thus, the importance of the labor market constraint varies from household to household and from region to region. Education is found to be of considerable importance to agricultural productivity. The primary schooling of the household head is important on its own and is found to be convex in its impact on crop incomes, implying increasing returns to schooling. Interaction effects between education variables and land are generally positive. Notable exceptions include a negative effect of both primary education variables interacted with other land and post-primary years of education of adults other than the household head interacted with both irrigated annual and perennial land. Interestingly, the results imply that primary education interacts strongly with irrigated land to increase crop income while post-primary education does not. Finally, almost half of the 119 commune dummies are significant at the 5% significance level. There are also non-negligible spatial differences in the effects of both irrigated and non- irrigated land, and other land types. These effects are all relative to land impacts on crop incomes in the South East (left out of the regression) and show expected signs and magnitudes. 3.2 The Benefits from Irrigation: Policy Simulations Irrigation is in part a private good. Individual households can invest in a bore-hole or pump but require capital to do so and face capital market constraints. It is also a collective good where substantial resources are needed to set things up and benefits are distributed across many people. The combination of these two factors helps explain why there could be underinvestment in irrigation in Viet Nam. Some amount has to be publicly provided while another amount can be provided privately, provided credit is available for that purpose. What are the potential benefits from irrigating a unit of non-irrigated land, holding total cultivable land area constant? How would those benefits be distributed across expenditure groups? And how might they vary with other factors? Here, an attempt is made to quantify those benefits using the above regression model. 18. A number of things lend support to this interpretation. Commune level collected wage data provide some evidence that labor markets are better developed in the South. Wage rates were unavailable, and hence missing in the survey, for a larger proportion of surveyed households in the North than in the South for both agricultural and unskilled non-agricultural work. Salinger (1993) provides further corroboration for the underdeveloped state of labor markets in North relative to South Viet Nam. In addition, more so than elsewhere, the Mekong delta is characterized by large areas which are either irrigated or not irrigated. As the paper shows in section 3.3, irrigation appears to increase the labor input requirement. It may be surmised that labor markets are likely to have adapted and more fully developed in areas of the Mekong which have large irrigated farms. 29 It is assumed that non-irrigated annual land can feasibly be irrigated (though costs may vary widely across regions), and that perennial and other land types cannot. Conversion of annual land to irrigation may be through rehabilitation and expansion of existing irrigation networks or through new construction. The paper considers the distributional impact which would arise from irrigating around 10 percent of the annual crop land currently not under irrigation. The impacts are examined under four possible policy scenarios for how the irrigation expansion is distributed across farms. In each case the same total amount of land is brought under irrigation. A first simulation simply extends irrigation to all farm households who have non-irrigated land. Because some farm households have little or no non-irrigated land, in practice a policy of bringing 10% of non-irrigated annual land under irrigation allows up to a maximum 500 m2 of newly irrigated land to farm households to reach a mean of around 260 m' per household. The current distribution of the share of annual land under irrigation tends to be bimodal. There is a tendency for farm households to have all their annual land irrigated, or none at all. It may be argued that a policy of converting 10 percent of the country's non-irrigated land to irrigation would more realistically be implemented in areas where farm households presently have little irrigation. Simulation 2 limits the expansion of irrigation to farms currently lacking access to irrigated annual land. The increment is zero if the farm has any irrigated land or has no non-irrigated land. Finally, simulations 3 and 4 target smallholders. Section 2.4 showed that poor farm households tend to have less annual, as well as less irrigated land than non-poor households. It is therefore of interest to examine how the distributional effects of bringing 10 percent of the country's non-irrigated annual land under irrigation would differ if those improvements were targeted to households with low total annual crop land holdings. Simulation 3 distributes the irrigation on the basis of low total household annual landholdings, while simulation 4 targets on the basis of low per capita annual landholdings. Once again, the simulations hold total annual land constant. Given the existing distribution of irrigated and non-irrigated land across households, simulation 3 results in irrigating all the non-irrigated land of households who have less than 3250 m' total annual land and simulation 4, the non-irrigated land of all with less than 620 m2 per person. The expected marginal benefit from irrigation-the change in household crop income from irrigating one unit of non-irrigated land-can be estimated by 0 - 7,, where 0 and ON are estimated at each data point, using the parameter estimates for the relevant interaction effects in Table 12 applied to the household-specific values of the relevant variables. To simulate policy impacts one can then multiply the marginal benefit by the household specific increment. However, the marginal benefit function, 1 - ^Nis only a first-order approximation and strictly valid for small changes only. For estimating the gains from discrete changes, a more accurate methodology is to recalculate the value of the function after substituting constrained land changes into the profit function as follows: 30 (4) Asn = I(pj,L, -ALj,Lj+ AL,,zj) - % (pt, L, L1,z) where, for example in the case of simulation 1 (and as appropriate for the others) ALj = 0 if Lj = 0 = Lj' if Lj' s! 500 so that the exact amount of land shifted into irrigation is appropriate to each household's circumstances-zero for those who have no unirrigated land and up to 500 m2 for households who do. The results are termed the "simulated total impacts." The distribution of the impact in per capita Dongs and as a percent of per capita household expenditures across all farm households classified into expenditure groups, is shown in Table 14 for the four simulations. Tables 15 to 18 show the results disaggregated across regions (with the exception of the Central Highlands where there are too few observations for the breakdown). Table 14: National Distribution of Impacts of Irrigation Under Different Scenarios Total impact as % of household Simulated total impacts (per capita) expenditure Expenditure group % cffarm ('000 Dongs/personlyr) population 1 2 3 4 1 2 3 4 1 0-500 4.1 9099.3 6700.7 11878.1 19310.3 2.13 1.57 2.78 4.52 2 501-600 4.1 12645.b 15513.5 13267.7 15314.7 2.29 2.81 2.40 2.77 3 601-700 6.9 16059.4 13118.9 14217.7 14604.0 2.47 2.02 2.18 2.24 4 701-800 9.6 14132.9 12312.0 15185.0 16657.2 1.88 1.64 2.02 2.22 5 801-900 9.2 13750.1 11443.8 15974.4 10287.4 1.62 1.35 1.88 1.21 6 901-1000 9.2 14775.1 10704.2 14482.1 9108.2 1.56 1.13 1.53 0.96 7 1001-1100 9.1 12924.6 9183.1 11281.8 11328.6 1.23 0.88 1.08 1.08 8 1101-1250 11.5 11396.7 10261.8 11310.6 7977.2 0.97 0.87 0.96 0.68 9 1251-1400 8.0 15035.7 12469.1 14279.8 14383.4 1.14 0.94 1.08 1.09 10 1401-1550 7.3 14185.6 13287.7 13194.1 9476.3 0.96 0.90 0.90 0.64 11 1551-1800 7.2 10240.5 10226.4 6653.9 2836.9 0.62 0.62 0.40 0.17 12 1801-2200 6.9 9947.0 9328.1 10000.7 3602.2 0.50 0.47 0.51 0.18 13 2201-3000 5.0 10142.9 11421.5 5143.5 3433.6 0.41 0.46 0.21 0.14 14 3001-4500 2.1 10935.7 13185.7 5215.5 3900.1 0.28 0.34 0.13 0.10 Total 100 12844.6 11226.5 12221.3 10293.6 1.05 0.92 1.00 0.84 Note: Results are based on the unrestricted model. A conversion of 10% of non-irrigated annual land to irrigation is common to all simulations. Under simulation (1): irrigation is distributed to all households subject to feasibility; (2) irrigation is distributed only to households without irrigated land; (3) irrigation is targeted to households with low total annual landholdings; and (4) irrigation is distributed to households with low per capita annual landholdings. 31 Table 15: Regional Distribution of Per Capita Impacts of Simulation I Northern Uplands Red River Delta North Coast Central Coast South East Mekong Delta Total Simulated Total Simulated Total Simulated Total Total Total Exp. Simulated impact % total impact % total impact % total impact % Simulated impact % Simulated impact % group total impact of hh ep impact of hh exp impact of hh exp impact of hh exp total impact of hh exp total impact of hh exp 1 27985.5 6.53 8335.3 1.88 14788.3 3.35 1855.0 0.47 1216.7 0.30 741.1 0.17 2 31584.0 5.80 5188.6 0.93 18062.1 3.28 2989.3 0.54 12305.0 2.25 -699.6 -0.13 3 35212.4 5.37 9028.8 1.39 15553.4 2.40 2645.0 0.41 13792.3 2.08 6001.8 0.91 4 34651.3 4.62 5162.9 0.68 16729.4 2.23 2704.7 0.36 10336.9 1.38 4027.0 0.54 5 35820.9 4.22 6623.6 0.78 18529.7 2.19 5195.5 0.61 14442.1 1.69 3895.0 0.46 6 35807.7 3.81 10776.5 1.14 20855.1 2.19 5480.6 0.57 6382.4 0.67 4785.5 0.50 7 30698.2 2.93 5547.2 0.53 19847.3 1.89 6768.6 0.64 14822.4 1.41 3409.2 0.33 8 33514.2 2.83 11012.6 0.93 18230.6 1.54 5822.9 0.50 4831.1 0.41 4704.2 0.40 9 33915.7 2.58 12650.2 0.95 22188.9 1.68 7073.4 0.54 14254.1 1.06 4175.5 0.31 10 30031.7 2.05 9656.7 0.67 25121.2 1.70 4587.0 0.31 14492.9 0.96 8923.5 0.60 11 29949.4 1.82 8392.4 0.50 22120.2 1.35 3201.8 0.20 18261.2 1.10 7429.5 0.45 12 36663.7 1.86 10786.7 0.55 26043.7 1.32 6518.4 0.33 16186.4 0.83 7982.8 0.40 13 29741.7 1.23 11884.4 0.47 22954.4 0.93 9928.3 0.41 21402.4 0.86 7954.7 0.31 14 28907.6 0.84 8488.1 0.22 2900.7 0.08 10387.3 0.31 9820.4 0.20 11955.2 0.31 33211.4 3.00 8798.9 0.72 18767.2 1.96 5151.6 0.42 12805.3 0.86 5839.3 0.40 Note: Results are based on the unrestricted model. Under simulation I the conversion of 10% of non-irrigated annual land to irrigation is distributed to all households subject to feasibility. Table 16: Regional Distribution of Per Capita Impacts of Simulation 2 Northern Uplands Red River Delta North Coast Central Coast South East Mekong Delta Total Total Total Total Simulated Total Simulated Total Exp. Simulated impact % Simulated impact % Simulated impact % Simulated impact % total impact % total impact % group total impact of hh exp total impact of hh exp total impact of hh exp total impact of hh exp impact of hh exp impact of hh exp 1 26082.1 6.08 14515.3 3.28 14568.9 3.30 521.3 0.13 1302.0 0.33 488.8 0.11 2 41024.3 7.53 8358.8 1.49 18248.0 3.31 3641.7 0.66 15214.6 2.78 2187.2 0.39 3 32702.6 4.99 8242.8 1.27 8442.5 1.30 -1617.5 -0.25 23844.4 3.60 9507.0 1.45 4 32412.3 4.32 3453.1 0.46 12476.5 1.67 2698.4 0.36 11907.8 1.59 6280.3 0.84 5 34634.2 4.08 2869.5 0.34 14847.7 1.75 5327.0 0.63 13564.9 1.59 6345.0 0.75 6 27266.9 2.90 4734.2 0.50 16719.4 1.75 6941.2 0.73 7792.4 0.82 8003.3 0.84 7 18956.6 1.81 4392.2 0.42 14756.4 1.41 3512.5 0.33 25629.1 2.43 5471.4 0.52 8 34830.1 2.94 6677.5 0.57 18470.4 1.56 4240.4 0.36 7473.9 0.63 7758.6 0.66 9 28508.3 2.17 5480.9 0.41 22731.8 1.73 3667.2 0.28 16530.1 1.23 6941.2 0.52 10 33139.2 2.26 1548.1 0.11 24518.8 1.66 1473.0 0.10 21100.6 1.40 13724.1 0.93 11 36228.6 2.20 0.00 0.00 20050.8 1.22 1858.6 0.11 28771.1 1.74 12848.4 0.77 12 28678.7 1.46 1006.7 0.05 27010.3 1.37 4097.1 0.21 24500.3 1.26 13416.3 0.68 13 24897.8 1.03 4849.1 0.19 16223.0 0.66 13118.5 0.54 29843.5 1.20 13759.6 0.54 14 36872.1 1.07 4315.9 0.11 0.00 0.00 17885.7 0.53 7788.0 0.16 18823.8 0.49 Total 30499.2 2.76 4323.2 0.35 15755.8 1.65 4153.3 0.34 17769.4 1.19 9677.5 0.67 Note: Results are based on the unrestricted model. Under simulation 2 the conversion of 10% of non-irrigated annual land to irrigation is distributed only to households without irrigated land. Table 17: Regional Distribution of Per Capita Impacts of Simulation 3 Northern Uplands Red River Delta North Coast Central Coast South East Mekong Delta Total Total Total Total Simulated Total Exp. Simulated impact % Simulated impact % Simulated impact % Simulated impact % Simulated Total impact total impact % group total impact of hh erp total impact of hh exp total impact of hh exp total impact of hh ap total impact % of hh exp impact of hh ep 1 40040.0 9.34 0.00 0.00 16744.0 3.79 201.9 0.05 1287.6 0.32 1670.3 0.38 2 28688.4 5.26 8789.7 1.57 18282.8 3.32 1763.3 0.32 13122.9 2.40 1577.5 0.28 3 41771.8 6.37 2980.7 0.46 16166.5 2.49 -4232.3 -0.66 0.00 0.00 6011.3 0.92 4 32566.1 4.34 3814.9 0.50 22552.6 3.01 3308.1 0.44 23583.0 3.15 1821.5 0.24 5 48170.9 5.68 2630.6 0.31 30108.0 3.56 9489.9 1.12 15552.6 1.83 2026.1 0.24 6 30457.6 3.24 10523.6 1.12 18341.9 1.92 6740.2 0.71 2634.5 0.28 4037.7 0.43 7 22029.0 2.10 3662.2 0.35 21273.1 2.03 7292.6 0.69 16304.3 1.55 652.7 0.06 8 29176.5 2.46 9711.9 0.82 23350.0 1.97 7016.3 0.60 9025.0 0.76 -473.9 -0.04 9 31302.4 2.38 11247.7 0.85 17343.6 1.32 3902.9 0.30 27273.7 2.03 1390.3 0.10 10 36372.9 2.48 8862.0 0.60 31239.5 2.11 4978.5 0.33 4394.7 0.29 3578.1 0.24 11 21393.9 1.30 6116.1 0.37 24493.6 1.50 4465.9 0.27 10752.1 0.65 1530.2 0.09 12 30343.8 1.54 9582.8 0.49 41661.6 2.11 6564.8 0.33 11809.2 0.61 4639.1 0.23 13 44089.7 1.82 14057.4 0.56 22001.1 0.89 23895.0 0.98 7336.1 0.29 152.4 0.01 14 522.3 0.02 6134.8 0.16 0.00 0.00 9968.8 0.29 5342.3 0.11 5280.6 0.14 Total 32870.9 2.97 7061.2 0.58 21878.8 2.29 6166.9 0.98 12051.9 0.81 2144.9 0.14 Note: Results are based on the unrestricted model. Under simulation 3 the conversion of 10% of non-irrigated annual land to irrigation is targeted to households with low total annual landholdings. Table 18: Regional Distribution of Per Capita Impacts of Simulation 4 Northern Uplands Red River Delta North Coast Central Coast South East Mekong Delta Total Total Total Total Total Total Exp. Simulated impact % Simulated impact % of Simulated impact % of Simulated impact % of Simulated impact % of Simulated impact % group total impact of hh etp total impact hh exp total impact hh exp total impact hh exp total impact hh exp total impact of hh exp 1 66093.4 15.42 0.00 0.00 23270.5 5.27 -12.8 -0.003 1287.6 0.32 156.1 0.04 2 51335.4 9.42 385.7 0.07 15460.6 2.81 12444.7 2.26 13122.9 2.40 1097.9 0.20 3 31943.0 4.87 2162.0 0.33 25377.8 3.92 -2818.3 -0.44 0.00 0.00 5435.2 0.83 4 43557.8 5.80 4634.6 0.61 21889.9 2.92 572.5 0.08 22153.2 2.96 -2719.8 -0.37 5 35629.3 4.20 2384.4 0.28 13680.9 1.62 3166.1 0.37 15552.6 1.83 649.7 0.08 6 26856.1 2.86 2441.6 0.26 21172.2 2.22 10350.7 1.08 2634.5 0.28 1183.4 0.12 7 27165.8 2.59 2837.3 0.27 21864.7 2.08 2006.5 0.19 16304.3 1.55 -2227.8 -0.21 8 26032.9 2.20 8312.0 0.71 21085.6 1.78 4904.6 0.42 10021.0 0.84 -473.9 -0.04 9 37681.2 2.87 11383.2 0.86 8878.4 0.67 5234.4 0.40 16070.5 1.20 2678.8 0.20 10 25623.7 1.75 3072.4 0.21 18073.4 1.22 6787.5 0.46 4394.7 0.29 1855.4 0.13 11 1761.8 0.11 4892.5 0.29 4787.3 0.29 4066.6 0.25 1635.3 0.10 1530.2 0.09 12 8570.6 0.43 5237.3 0.27 14804.9 0.75 6902.3 0.35 7262.8 0.37 2064.8 0.10 13 29762.1 1.23 2424.2 0.10 10597.2 0.43 25819.3 1.06 7336.1 0.29 152.4 0.01 14 522.3 0.02 6134.8 0.16 0.00 0.00 0.00 0.00 0.00 0.00 5280.6 0.14 Total 31234.6 2.83 4660.9 0.38 19199.2 2.01 5785.7 0.47 9284.6 0.62 906.9 0.06 Note: Results are based on the unrestricted model. Under simulation 4 the conversion of 10% of non-irrigated annual land to irrigation is distributed to households with low per capita annual landholdings. Converting 10 percent of non-irrigated land to irrigation produces an increase in crop incomes equal on average to around 1 percent of mean household expenditures. This implies an elasticity of 0.1. The elasticities vary only slightly across the simulations. However, the level and distribution of per capita impacts differs across national expenditure groups according to how the irrigation is distributed (Table 14). This reflects the method of allocating the irrigation expansion combined with the existing household distribution of irrigated and non- irrigated annual crop land and the influence of other household and community specific factors entering the marginal benefit of irrigation function such as education, household size and region. Under equal distribution to all households subject only to land constraints (simulation 1), impacts are smaller at the lower and upper ends of the distribution but otherwise relatively steady across expenditure groups. Simulation 2 tends to be more generous towards the upper end of the distribution and less so at the bottom end though it is not altogether that different from impacts under simulation 1. Targeting the irrigation expansion to smallholders results in larger absolute impacts at the lower end of the distribution which fall much more sharply when targeting is done on the basis of per capita than household annual landholdings. Impacts under all 4 simulations are certainly progressive-declining as a proportion of household expenditures as living standards rise-and so inequality reducing. Progressivity is most pronounced for simulation 4 which confers large benefits to the poorest groups (worth 4.5 percent of household expenditures for the poorest group and only 0.1 percent for the wealthiest expenditure group). Gzins are very concentrated regionally (Tables 15 to 18). The potential benefits of Jrrigation appear to be strongest in the Northern Uplands where simulated total impacts are lrgest for all simulations (mean impacts of up to 3 percent of mean household expenditures). h Tigation expansion is inequality reducing there, exceptionally so when irrigation expansion is lt rgeted to low per capita landholding farm households. However, the net absolute gains tend to be relatively steady across the distribution of per capita expenditures in all except simulation 4. The next most substantial impacts are found in the North Coast and South East regions. In the North Coast, impacts are generally inequality reducing on the whole, though the gradient is muc h lower than for the Northern Uplands. Absolute benefit levels tend to increase with expe..nditure class except under simulation 4 which, here too, is found to result in the most progiessive distribution of benefits. In the South East, total impacts tend to be larger for the bettet off (with the exception of simulation 4) and flat or only somewhat progressive when expre. ;sed as a proportion of household expenditures. The smallest total impacts are evidenced for tha, Central Coast and the Mekong River Delta. In both regions the benefits are also far from pi ogressive. Indeed, the simulated per capita total impacts, though small, tend to increase for high er expenditure groups. One interesting finding from the above is that concentration of benefits and progressivity appear to go hand in hand. Benefits tend to be higher where their distributi on is also more pro-poor. The results hint towards targeting irrigation expansion to the Northern Uplands and North Coast regions, where absolute benefits are not only higher but also well distributed. These are also Viet Nam's poorest regions (World Bank 1994c; Dollar and Glewwe 1995). The overall regional picture is quite robust across simulations. Nationally, there is not much of aii obvious tradeoff between the ways of distributing the irrigation across regions. 36 Interestingly, however, there is a distinct regional pattern to which simulation has the greatest impact on regional absolute benefit levels. This no doubt reflects characteristics of how annual land is distributed regionally. Simulation 1 produces the highest absolute gains for both the Northern Uplands and the Red River; simulation 3 for the North and Central Coasts; and simulation 2 for the South East and Mekong. In each case these are contiguous regions. Simulation 4 is distinguished not by producing the largest benefits in any region but by tending to favor the poor with larger absolute impacts and by producing the most progressive distribution of benefits almost universally across regions. At first sight, the simulation outcomes appear surprising for the Mekong River Delta. They also appear robust. This is the country's primary producer of rice with, as yet, only half its total cultivated area under irrigation. It is sometimes said that extending irrigation will enable double and triple cropping and boost production and incomes formidably in this ideal setting for paddy cultivation. However, the Mekong delta situation is complex. Various characteristics of the region's ecosystem and economy appear to provide credible explanations for the simulation results. Irrigation systems in the Mekong Delta have been plagued by problems of sea water intrusion and acid-sulphate soils. In recent years, as supplementary areas are brought under irrigation in upstream areas, the level and flow of the Mekong river has dwindled, resulting in salt water intrusion in previously productive irrigated fields downstream (NEDECO 1991)." This has meant that only one crop can be grown annually or, under the worse case, that continued rice cultivation is rendered impossible. In the latter case, the areas are often transferred to aquaculture activities such as the farming of brackish shrimp which can be very profitable but would be reflected in lower crop incomes. Furthermore, fully irrigated areas may also suffer from extensive flooding and water logging for long periods of the year. In such areas of the Mekong, what is needed is better water management and drainage control rather than irrigation as such. The issue seems to be that, because of the heterogeneity of irrigated land in the Mekong Delta, it is hard to generalize about the impact of irrigation infrastructure there. If the data allowed a separation between irrigated areas which suffer from salinity and acidity problems and other irrigated land, we would probably get strong impacts of additional irrigation investments in the Mekong River Delta. Irrigation can be very positive depending on whether it is upstream or not. The results indicate low marginal benefits on average, where they are being averaged over a considerable amount of heterogeneity. The profits from irrigation vary by region but there is also variation within region. 19. NEDECO 1991 quotes farmers in the center of the delta complaining about this. 37 3.3 The Cost of Household Labor The costs of household labor inputs on the family farm were ignored in defining net crop incomes.' This is entirely defensible if one is concerned solely with the impact of irrigation on family consumption (since the implicit payment for own-labor inputs is exactly matched by the receipts leaving consumption unchanged). However, to assess the gains to farm profits the family labor cost should be debited. And, like other costs of production, family labor inputs may well be related to whether land is irrigated or non-irrigated. It is not obvious that irrigation would save on family labor. However, one then faces the long-standing issue of what wage rate one should use for valuing family labor inputs; with surplus labor in rural areas and supervision costs, the opportunity cost of family labor may be well below the market wage rates for similar work. Here I try to assess the possible bias in the above results due to the omission of family labor input costs. The results may either over- or under-estimate the net returns to irrigation, depending on family labor requirements on irrigated versus non-irrigated annual land. The shadow wage of family labor is somewhere between zero and the agricultural wage rate. I assume that the shadow wage is a constant proportion (#) of the prevailing market wage. The cost of family inputs to own-farm production by household j is then given by (5) F.j 4'y-kWkFjk where wk is the wage rate for the k'th demographic-group (adult men and women, and children), and F is the labor time devoted to farm work by demographic group for household j. Information on wage rates are available in the community survey separately for men, women and children for a series of tasks (preparation, planting/transplanting, weeding, and harvesting)." However, the household survey does not include the time allocation for each member by those tasks. Furthermore, in practice the wage data are very incomplete reflecting the lack of labor markets in many of the communes. Thus demographic-specific commune mean agricultural wages are formed over all tasks for which wage rates are recorded and these are used to value the time on all farm tasks by each household member. Missing data at the commune level are then replaced by the regional means for males, females or children as appropriate.22 20. Recall that non-family labor costs are included. 21. Note that since the wage rates can only be obtained from the community schedule, they are not household specific. 22. Data are missing for 398, 430, and 2406 households on wage rates for men, women, and children respectively. 38 The parameter + is unknown. To get an upper bound estimate, family labor input costs are evaluated at agricultural market wage rates (so + = 1), and the family labor cost is regressed against the same right-hand-side variables used to explain crop incomes. The net marginal impact of irrigation over non-irrigation on the cost of the family labor input can then be compared to the previously calculated net marginal effect of irrigated over non-irrigated land on crop incomes. If there is no significant difference between irrigated and non-irrigated land, then we need not worry; for any value of 4 my results for crop income carry over to profits net of family labor. If there is a difference then we can ask if there is an admissible value of + which reverses the earlier conclusions. The regression results (given in Table 19) indicate that irrigation tends to increase work on the family farm. It is also notable that, other things being equal, bigger households and ones with a larger proportion of adults and teenagers tend to use more family labor. Table 20 presents the total marginal effects of the main variables on labor costs allowing for all interaction effects. The effect on the market value of family labor time of irrigation over non-irrigated land is estimated to be 7,279 Dongs per 100 m2. Subtracting this full amount (i.e., at 4 = 1) from the average net impact on crop incomes of converting 100 m2 of non-irrigated land to irrigation reduces the latter to 21,299 Dongs, a 25 percent decline. Of course this is an upper bound estimate which may considerably overestimate the costs. If the opportunity cost of family labor is half of the market wage (+ = 0.5), then the gain in farm profit from irrigating 100 m' of non- irrigated land is 24,939 Dongs, a 13 percent decline. In conclusion, the earlier results overestimated the marginal effect of irrigation on farm profits, though the difference is not prohibitively large, representing a maximum of 25% of the previous estimates of net income gains. 3.4 The Cost of Irrigation Expansion Information on the costs of irrigation expansion is hard to come by, and generalizations across regions and types of irrigation investments are risky. Nonetheless, even a rough sense of the cost-benefit appraisal can be useful. Irrigation project costs have been estimated by a number of agencies for various regions. Estimated average costs-including for a World Bank irrigation rehabilitation project in the Central Coast and for a large number of water resource development projects in the Mekong River Delta drawn up as part of the Mekong Delta Master Plan-fluctuate around 85,000 Dongs per 100 m2; these are averages over appraisals for multiple 39 Table 19: Regression Results: Family Labor Costs Unrestricted Model Restricted Model laborcost Coefficient t-ratio Coefficient t-ratio sexhh 158669.5 1.45 160862.6 1.52 hhsvgs -0.0231 3.58 -0.0239 3.78 hhsi2 246193.1 5.31 269214.9 6.93 prop716 1471817.0 2.29 1185507 4.24 pfadlt 1507441.0 2.20 1435968 4.09 pmadlt 1754666.0 2.50 1711478 5.2 oedl*oedl 2276.06 2.42 3135.26 5.03 oed2 -24043.93 1.17 -19321.13 1.19 oed2*oed2 -1079.62 1.53 -1059.71 1.68 irrigated 322.208 3.24 306.387 6.16 irrigated*irrigated -0.0021 2.52 -0.00258 4.66 nonirrigated -126.674 1.68 -58.595 1.20 nonirrig*nonirrig -0.00252 5.65 -0.00290 7.23 perennial -197.220 1.02 -283.297 1.88 perennial*perennial -0.00858 4.72 -0.00887 6.51 waterland 12564.61 2.82 13290.42 3.07 propauct 842682.1 1.60 939847.9 1.95 propall 488905.0 1.81 509353.9 2.80 hedl*perennial 46.6154 2.17 34.251 2.30 hed2*irrigated -8.9632 1.77 -12.528 3.30 hed2*nonirrigated -8.4584 1.97 -8.368 2.61 hed2*waterland 119.046 2.15 57.577 1.67 oedlI*irrigated 9.620 3.09 9.627 3.96 oedl*nonirrigated -5.151 1.99 -2.592 1.38 oedl*forest 25.395 1.70 12.208 2.06 oed2*irrigated -2.499 1.08 -2.863 1.32 oed2*perennial -7.088 1.43 -7.862 1.90 oed2*otherland 15.0682 1.48 6.371 1.91 hhsize*irrigated 13.962 2.24 13.641 2.70 hhsize*nonirrigated 17.257 3.48 15.766 3.55 hhsize*perennial 22.0446 1.60 27.969 3.14 pfadlt*irrigated -179.628 1.82 -178.334 2.60 pfadit*nonirrigated 182.699 2.29 131.447 2.00 pfadIt*perennial 273.755 1.23 327.912 1.90 pmadlt*nonirrigated 268.217 3.54 193.512 3.15 pmadlt*perennial 373.679 2.06 458.372 3.33 pmadk*forest -684.857 1.77 -399.353 2.17 prop7l6*irrigated -318.899 3.45 -326.617 4.48 prop7l6*perennial 203.644 1.14 300.683 1.91 rr*irrigated 85.881 1.065 99.855 1.34 rr*waterland -13158.55 2.99 -13419.56 3.10 mk*irrigated -125.50 2.99 -115.506 3.25 40 Table 19 (continued) Unrestricted Model Restricted Model laborcost Coefficient t-ratio Coefficient t-ratio mk*nonirrigated 51.107 1.25 66.613 2.39 mk*waterland -12923.81 2.93 -13184.76 3.05 nw*irrigated -144.161 1.63 -133.200 1.60 nw*nonirrigated 48.626 1.03 65.681 1.87 nw*waterland -13165.85 2.99 -13408.45 3.10 nc*irrigated -123.831 1.36 -118.265 1.37 nc*nonirrigated 88.077 1.58 106.054 2.35 nc*perennial -129.440 1.49 -104.656 1.36 nc*waterland -12856.66 2.91 -13153.33 3.03 cc*perennial -284.91 1.30 -266.521 1.27 cc*waterland -106753.7 1.42 -106546.7 1.42 ch*irrigated -798.508 1.13 -877.190 1.27 ch*perennial -260.649 3.05 -245.386 3.20 ch*otherland 789.428 1.58 782.188 3.02 ch*waterland -13286.87 2.79 -13973.44 2.98 Number of obs = 3025 Number of obs = 3025 F(232, 2792) = 16.32 F(232, 2792) = 16.32 Prob > F = 0.0000 Prob > F = 0.0000 R-square = 0.5756 R-square = 0.5756 Adj R-square = 0.5404 Adj R-square = 0.5404 Root MSE = 1.9e+06 Root MSE = 1.9e+06 Note: The restricted model results from the pruning of all variables with t-ratios less than 1 in the unrestricted model. The unrestricted model contained exactly the same variables as the crop income regression reported in table 12. projects, though the variance is low.23 For these costs, the estimated model indicates an annual gain in net crop income of around 28,600 Dongs per 100 m2, falling to a gain of 21,300 Dongs in farm profit at the maximum shadow wage for family labor. This represents a rate of return of at least 25 to 35% per year, assuming the project delivers such benefits indefinitely. But even under conservative assumptions of a project life of only 10 years and with the maximum shadow wage for family labor, the rate of return is about 20%.' 23. The World Bank project average costs are estimated at about 83,150 Dongs per square meter when excluding consultant costs as well as physical and price contingencies. The Mekong Master Plan projects average 85,830 Dongs per m' (table 8.2), 87,650 Dongs (table 8.3), and 87,120 Dongs (table A2.1) all in NEDECO 1993. 24. These are internal rates of return which equate the present value of the stream of benefits over the chosen time period with initial costs. 41 Table 20: Marginal Effect on Family Labor Costs Allowing for Interadion Effects Unrestricted model Restricted Model Marginal effect on Marginal effect on Variable family labor cost t-ratio family labor cost t-ratio irrigated annual land Dongs/100m? 19,249.7 5.3 20,032.5 5.7 non-irrigated annual land Dongs/100md 11,970.8 4.0 10,864.3 6.3 perennial land Dongs/100m' 23,231.6 3.7 21,530.7 5.3 forest land Dongs/100m2 -3,777.0 0.7 -2,795.3 1.4 water surface land Dongs/100m -764,911.9 1.1 -755,668.6 1.1 other land Dongs/100m2 12,901.6 1.2 2,612.2 1.9 household size Dongs/person 342,305 9.5 360,211.8 11.2 prop female adults Dongs/% point 17,197.3 2.7 1596511 5.3 prop male adults Dongs/% point 25,086.1 3.9 2415231 7.9 prop aged 7-16 Dongs/% point 9,143.5 1.5 648723.4 2.8 primary ed (head) Dongs/year -31,351.0 0.4 23,235.7 2.3 ed > primary (head) Dongs/year -49,473.3 2.2 -43,136.9 3.6 primary ed (other adults) Dongs/year 78,568.7 4.7 61,565.8 7.9 ed > primary (other Dongs/year -41,642.1 2.8 -38,451.3 3.1 adults) mean family labor costs 3,034,006 3,034,006 Note: Marginal effects are evaluated at mean points. 42 4. Conclusions Viet Nam has poor infrastructure and high poverty. These two facts are intimately connected. However, the nature of those connections and their implications for the role of infrastructure investments in fighting poverty are complex to disentangle. This paper has focused on some aspects of the link between poverty and lack of infrastructure using the VNLSS. Access to infrastructure services tends to be poor for the majority of Vietnamese. Urban areas are better provisioned and some regions certainly fare worse than others. In particular, there are some distinct differences between the North and South of the country. Imbalances are also evidenced among infrastructure services. For example, the provision of social service facilities is generally superior than that of other physical infrastructure. Piped water provision and electricity reveal considerable disparities between poor and non-poor groups. But, by and large, the data indicate that basic infrastructure services are generally inadequate for all groups, though generally worst for the poor. As a result, it cannot be surmised that an expansion in investment in basic infrastructure will be well-targeted to the poor. Indeed, there is ample scope for the non-poor to capture the lion's share of the direct gains from infrastructure investment in Viet Nam. To assess the impacts on poverty it is necessary to examine the distribution of the marginal benefits of specific infrastructure investments. The paper focuses on irrigation investments to explore this issue in more depth. The cross-sectional variation is used to estimate the marginal impacts of converting non-irrigated annual crop land over to irrigation. In particular, a policy of irrigating 10 percent of currently non-irrigated annual land is simulated based on a regression model for crop income which includes irrigated and non-irrigated land as explanatory variables. The simulations allow for four different ways of distributing the irrigation expansion across households: in simulation (1): irrigation is distributed to all households subject to feasibility; in (2) it goes only to households currently without irrigated land; in (3) it is targeted to households with low total annual landholdings and in simulation (4) it is targeted to households with low per capita annual landholdings. In general, at the national level the absolute income gains across expenditure groups imply that an undifferentiated expansion of irrigation would be redistributive-having higher proportionate gains to poorer households. Targeting the irrigation expansion to households with small per capita landholdings produces the most progressive incidence of gains as well as the largest absolute benefits to the poor. The results under all simulations show the highest total impacts on net crop incomes would occur for Viet Nam's two poorest regions-the Northern Uplands and the North Coast, where the impacts also show the most pro-poor distribution. These substantial potential gains from irrigation from an equity point of view are likely to be accompanied by sizable average rates of return. Even under quite conservative assumptions-namely a project life of only 10 years and valuing family labor inputs at the market wage for similar work-the average annual rate of return implied by my estimates of the gains to farm profits, and recent estimates of the investment cost of irrigation, is about 20%. 43 An even larger impact may be possible with a more differentiated expansion of irrigation-emphasizing key regions such as the Northern Uplands and addressing the need for rehabilitation of existing irrigation infrastructure, to realize its full potential. Conversely, the rate of return will undoubtedly be lower in some areas where irrigation expansion is particularly costly. Lack of irrigation infrastructure is clearly not the only constraint to reducing rural poverty in Viet Nam. The quantity (in particular household size) and quality (education) of the family's human resources also matter greatly. And not only do other important constraints exist, but these are inextricably bound to the benefits which can ultimately be derived from irrigation infrastructure. The analysis uncovers important complementarities between education, particularly primary education, and the gains from irrigation. Demographics are also found to be key. Finally, one can conjecture that the current lack of other infrastructure such as roads, electricity, communications and so forth, must also conspire to reduce the impacts which can be garnered from irrigation alone. 44 References Binswanger, Hans P., Shahidur R. Khandker and Mark R. Rosenzweig. 1993. "How Infrastructure and Financial Institutions Affect Agricultural Output and Investment in India." Journal of Development Economics, 41(2): 337-366. Dollar, David and Paul Glewwe. 1995. "Poverty and Inequality in Viet Nam: The Current Situation." Mimeo, EA1CO and PRDPH, World Bank. Goldstein, Ellen. 1993. "The Impact of Rural Infrastructure on Rural Poverty." Mimeo, South Asia Region, World Bank. Howe, J. and P. Richards. 1984. Rural Roads and Poverty Alleviation. ILO, Intermediate Technology Publications Ltd., London, U.K. Jimenez, Emmanuel. 1995. "Human and Physical Infrastructure: Public Investment and Pricing Policies in Developing Countries." In Jere Behrman and T.N. Srinivasan, eds., Handbook of Development Economics, Volume 3, Amsterdam: North-Holland. Lipton, Michael and Martin Ravallion. 1995. "Poverty and Policy." In Jere Behrman and T.N. Srinivasan, eds., Handbook of Development Economics, Volume 3, Amsterdam: North- Holland. NEDECO. 1991. "Mekong Delta Master Plan: Inception Report." May 13, The Netherlands. . 1993. "Draft Master Plan for the Mekong Delta in Viet Nam: A perspective for Sustainable Development of Land and Water Resources." June, The Netherlands. Ravallion, Martin. 1994. Poverty Comparisons. Chur, Switzerland: Harwood Academic Press, Fundamentals in Pure and Applied Economics, Volume 56. Salinger, Lynn B. 1993. "Viet Nam's Agricultural Comparative Advantage and Export Potential." Associates for International Resources and Development, Cambridge, Mass. State Planning Committee, UNDP, FAO and World Bank. 1989. "Viet Nam Agricultural and Food Production Sector Review." Mission report. UNICEF. 1994. Situation Analysis of Women and Children in Viet Nam. Hanoi. van de Walle, Dominique. 1995a. "Targeting and Incidence: An Overview of Implications for Research and Policy." In D. van de Walle and K. Nead Public Spending and the Poor. Theory and Evidence. London and Baltimore: The Johns Hopkins University Press. 45 . 1995b. "Rural Poverty in an Emerging Market Economy: Is Diversification into Non Farm Activities in Rural Viet Nam the Solution?" Mimeo, PRDPE, World Bank. Vu, Tu Lap and Christian Taillard. 1993. Atlas du Vlet Nam. Montpellier et Paris: Reclus, La Documentation Francaise. World Bank. 1990. "Viet Nam: Water Supply and Sanitation Sector Study." April. . 1994a. World Development Report 1994. Infrastructure for Development. New York: Oxford University Press. . 1994b. "Viet Nam Transport Sector: Serving an Economy in Transition." Report No. 12778-VN, August. . 1994c. "Viet Nam: Poverty Assessment and Strategy." Report No. 13442 VN,September. 46 LSMS Working Papers (continued) Decomposition with Applications to Brazil and India in the 1980s No. 84 Vijverberg, Measuring Income from Family Enterprises with Household Surveys No. 85 Deaton and Grimard, Demand Analysis and Tax Reform in Pakistan No. 86 Glewwe and Hall, Poverty and Inequality during Unorthodox Adjustment: The Case of Peru, 1985-90 No. 87 Newman and Gertler, Family Productivity, Labor Supply, and Welfare in a Low-Income Country No. 88 Ravallion, Poverty Comparisons: A Guide to Concepts and Methods No. 89 Thomas, Lavy, and Strauss, Public Policy and Anthropometric Outcomes in C6te d'Ivoire No. 90 Ainsworth and others, Measuring the Impact of Fatal Adult Illness in Sub-Saharan Africa: An Annotated Household Questionnaire No. 91 Glewwe and Jacoby, Estimating the Determinants of Cognitive Achievement in Low-Income Countries: The Case of Ghana No. 92 Ainsworth, Economic Aspects of Child Fostering in C6te d'Ivoire No. 93 Lavy, Investment in Human Capital: Schooling Supply Constraints in Rural Ghana No. 94 Lavy and Quigley, Willingness to Pay for the Quality and Intensity of Medical Care: Low-Income Households in Ghana No. 95 Schultz and Tansel, Measurement of Returns to Adult Health: Morbidity Effects on Wage Rates in C6te d'Ivoire and Ghana No. 96 Louat, Grosh, and van der Gaag, Welfare Implications of Female Headship in Jamaican Households No. 97 Coulombe and Demery, Household Size in Cbte d'lvoire: Sampling Bias in the CILSS No. 98 Glewwe and Jacoby, Delayed Primary School Enrollment and Childhood Malnutrition in Ghana: An Economic Analysis No. 99 Baker and Grosh, Poverty Reduction through Geographic Targeting: How Well Does It Work? No. 100 Datt and Ravallion, Income Gains for the Poor from Public Works Employment: Evidence from Two Indian Villages No. 101 Kostermans, Assessing the Quality of Anthropometric Data: Background and Illustrated Guidelines for Survey Managers No. 102 van de Walle, Ravallion, and Gautam, How Well Does the Social Safety Net Work? The Incidence of Cash Benefits in Hungary, 1987-89 No. 103 Benefo and Schultz, Determinants of Fertility and Child Mortality in C6te d'lvoire and Ghana No. 104 Behrman and Lavy, Children's Health and Achievement in School No. 105 Lavy and Germain, Quality and Cost in Health Care Choice in Developing Countries No. 106 Lavy, Strauss, Thomas, and De Vreyer, The Impact of the Quality of Health Care on Children's Nutrition and Survival in Ghana No. 107 Hanushek and Lavy, School Quality, Achievement Bias, and Dropout Behavior in Egypt No. 108 Feyistan and Ainsworth, Contraceptive LIse and the Quality, Price, and Availability of Family Planning No. 109 Thomas and Maluccio, Contraceptive Choice, Fertility, and Public Policy in Zimbabwe No. 110 Ainsworth, Beegle, and Nyamete, The Impact of Female Schooling on Fertility and Contraceptive Use: A Study of Fourteen Sub-Saharan Countries No. 111 Oliver, Contraceptive Use in Ghana: The Role of Service Availability, Quality, and Price No. 112 Montgomery, KouamO, and Oliver, The Tradeoff between Number of Children and Child Schooling: Evidence from C6te d'Ivoire and Ghana No. 113 Pradhan, Sector Participation Decisions in Labor Supply Models No. 114 Beegle, The Quality and Availability of Family Planning Services and Contraceptive Use in Tanzania No. 115 Lavy, Spratt, and Leboucher, Changing Patterns of Illiteracy in Morocco: Assessment Methods Compared No. 116 Lavy, Palumbo, and Stern, Health Care in Jamaica: Quality, Outcomes, and Labor Supply No. 117 Glewwe and Hall, Who Is Most Vulnerable to Macroeconomic Shocks? Hypotheses Tests Using Panel Data from Peru No. 118 Grosh and Baker, Proxy Means Tests for Targeting Social Programs: Simulations and Speculation No. 119 Pitt, Women's Schooling, the Selectivity of Fertility, and Child Mortality in Sub-Saharan Africa No. 120 Grosh and Glewwe, A Guide to Living Standards Measurement Study Surveys and their Data Sets THE WORLD BANK CD Headquarters European Office Tokyo Office 13544 9 78082 335444 ISBN 0-8213-3544-8