WF§S 13H-S POLICY RESEARCH WORKING PAPER 1348 Quality Change and Other The long-nn rise in prices and in terms of trade for Influences on Measures expor of manvfactured goods has been of Export Prices overestimated because of Manufactured Goods measures of export prices for manuflaaured products. incduding the U:N. export unit value index, are biased RobertE. Lipsey upward. In particular, they fail to account adequately for improvements in the quality of exported manufactured goods. The World Bank nemational Economics Deparmnent International Trade Division a August 1994-- - - POLICY RESEARCH WORKING PAPER 1348 Summary findings Measures of long-term trends in world export prices ior accounted for in the price indexes reduces the rate of manufactured goods, and in the terms of trade between increase tor thosc products by about one percentage manufactured goods and primary products, are sensitive point a year, and that adjustment for only those products to many choices in methods for weighting indexes, basc reduces the estimated rare of incrcase in prices for all periods, and (most important) changes in quality. For manufactures by about half a percentage point a year. example: Conservative estimates of the bias in thc most - Weighting products by their importance in exports to commonly used measure of export prices of developing countries, rather than by their importance in manufactured products - the U.N. cxport unit value exports to all countries, reduces the estimated rate of index for manufactures - suggest that this measure increase in prices for manufactured goods by about 0.1 or overstates the long-run rise in prices for manufactured 0.2 percentage points a year. goods by more than half a percenrage point a year, - A shift in weights from those of an early year (1963) probably one percentage point or more. If so, there has to those of a recent year (1986) reduces the rate of been no long-term trend toward the prices of increase in prices by about a third of a percentage point a manufactured goods rising faster than prices for primary year. products. = Export price indexes with weights of Japanese However, no conceivabUc estimate of bias in measures exports grow about 0.2 to OA percentage points a year of prices for manufactured goods would reverse rhe less than one weighted by the U.S. export composition, picture of declining relative prices for primary products with the larger difference for indexes based on 1963 in the 1980s. weiglhts. a Adjusting the price index for exports of machinery and transport equipment for quality changes not This paper - a product of the International Trade Division, International Economics Department - is part of a larger effort in the department to analyze the role of commodities in the development process. Copies of the paper are available free from the 'World Bank, 1818 H Strcet NW, Washington, DC 20433. Please contact Jennifer Ngaine, room R2-052, extension 37947 (26 pages). August 1994. The Policy Resech Working Paper Series disseninates the findings of uwork in progres to encourage rhe excbantge of ideas about developmnent isss An objective of the seres is to genthe findigs our quwckly, even if the presentations are less than fully polished. The papers crny the names oftkhe athors and should be used and cted accordingly The rmdb%4 interpretations, and conclusions are the authors'own and shoud 5not hE attributed to the World Bank its Executive Board of Directom or any of its member countria Produced y the Policy Research Dissemination Center Quality Change and Other Influences on Measures of Export Prices of Manufactured Goods Robert E. Lipsey ABSTRACT Measures of long term trends in world export prices of manufactured goods and in the terms of trade between manufactured goods and primary products are sensitive to many choices in methods of constructing the indexes weighting, base periods, and, most important of all, the treatment of quality change. For example, weighting products by their importance in exports to developing countries, rather than in exports to all countries, reduces the estimated rate of increase in manufactured goods prices by about .1 or .2 per cent per year. A shift in weights from those of an early year (1963), to those of a recent year (1986), reduces the rate of increase in prices by about a third of a per cent per year. An export price index with the weights of Japanese exports grows by less than one weighted by the U.S. export composition, by between .2 and .4 per cent per year, the larger difference for indexes based on 1963 weights. Adjusting the price index for exports of machinery and transport equipment for quality changes not taken account of in the price indexes reduces the rate of increase for those products by about one per cent per year and that adjustment for only those products reduces the estimated rate of increase in prices of all manufactures by about one half per cent per year. Conservative estimates of the bias in the most commonly used measure of export prices of manufactured products, the United Nations export unit value index for manufactures, suggest that this measure overstates the long-run rise in manufactured goods prices by more than half of one per cent per year, probably one per cent or more. If this is the case, there has been no long term trend toward rising prices of manufactures relative to primary products. However, no conceivable estimate of bias in measures of manufactured goods prices would reverse the picture of declining relative primary product prices during the 1980s. Robert E. Lipsey Queens College and The Graduate Center, CUNY and National Buxeau of Economic Research 269 Mercer Street, 8th floor New York, N.Y. 10003 Introduction Data on export prices of manufactured goods for the industrial countries have long been used throughout the World Bank to assess and project trends in the dollar costs of imports needed in Bank-financed projects in the developing countries. These data also help to shape the Bank's view of trends in the barter terms of trade between commodities and manufactures in world markets, i.e., trends in "real" commodity prices. In recent years there have been growing concerns that trends in these price data are biased upward. This paper addresses these concerns. The issue of long-term changes in the terms of trade between primary products and manufactures has a long history, going back at least to Torrens (1821), Ricardo (1832), and Mill (1848). It engaged the attention at various times, of Jevons (1866) and Keynes (1912), and of Singer (1950), Prebisch, in United Nations (1950), Lewis (1952), Kindleberger (1956), Haberler (1959, 1961) and Viner (1953), among more recent commentators. The classical views were based mostly on theoretical considerations of the inevitability of increasing costs, while the later debate has been centered around empirical questions. Many improvements have taken place in the measures of prices of primary products, but the manufactures side of the debate has rested on a collection of inappropriate or low-quality series stitched together in a variety of ways by different analysts. The purpose of this paper is to examine the implications for this discussion of some efforts to construct better price series for exports of manufactures by developed countries for the period since 1953. Among the issues that are examined here with the new price data are: (1) The effects of shifting the dates of the export weights used; (2) The effect of changing the country on whose export pattern the weighting is based; (3) The difference between price movements for exports to developing countries and those for exports to all countries, and (4) Most important, the effects of making some adjustments for changes in the "quality" or characteristics, of manufactured goods, not taken into account in standard price measures. It has been said that the importance of the quality improvement issue to the terms of trade debate has been exaggerated because there have been quality improvements in primary products as well as in manufactures. For example, Spraos (1983) mentions that in Kenya, the proportion of coffee beans harvested that was of the highest quality rose from .2 per cent in 1955-58 to 16 per cent in 1964-65 and that the proportion of iron ore exports that had an iron content of more than 60 per cent rose from one third in 1960 to more than half in 1964 (p. 58). It is not clear how that information is relevant to the question of-bias in price measurement unless it is shown, as it has not been so far, that the primary product prices used for price indexes fail to take account of such quality improvement. These changes in the average quality of primary products 6ould certainly produce an upward bias in primary product prices if these prices were measured by unit values that took no account of quality distinctions. However, most price indexes for primary products and most calculations of the terms of trade between primary products and manufactures use specification prices for primary commodities, a procedure that should eliminate that source of bias. On the other hand, if commodities sold on organized markets, the source of specification prices, are a small part of total commodity trade, the reported prices may be biased as new and improved versions of commodities gradually replace the traditional specifications. -3- How large is the alleged movement in the net barter terms of trade? Spraos (1980) calculated many varieties, and these can be used as one reference point. For periods going back to the 1870s and ending before World War 1I the annual deterioration f-r primary product terms of trade reported ranged from a quarter of one per cent to one per cent per year, according to his calculations, clustering roughly around one half of one per cent. For periods after 1950, to 1975 or so, and excluding petroleum from primary products, the rates of deterioration range up to over 2 per cent per year. For the totality of primary products, Spraos thought the evidence for the long run, extending through the 1970s, did not show any statistically significant trends. Extending the data through the 1980s would probably restore that negative coefficient for relative primary product prices but would not bring it outside the range mentioned above. A paper by Cuddington (1990) examining terms of trade for 26 individual commodities from 1900 to 1983 lists time coefficients ranging fr6m +2 per cent per year to -1.2 per cent per year in trend-stationary models, with 5 out of 6 negative coefficients smaller than 1 per cent in absolute value. Of 10 negative time coefficients for commodities for which difference - stationary models are used, only 2 are above 1 per cent per year. Grilli and Yang (1988) calculated that the trend rate of decline in primary commodity prices relative to manufactures prices from 1900 through 1986 was .5 per cent per year, or .6 per cent per year if fuels were excluded from primary products. A. Measures of Prices of Manufactured Goods1 Measures of manufactured goods prices have long baza-.x a weak link in all 1This section is taken mainly from Lipsey, Molinari, and Kravis (1991). -4- kinds of empirical studies of international merchandise trade. The deficiencies of these indexes affect studies of international competitiveness, of real exchange rates, of income, price, and substitution elasticities, of. export supply, and of the terms of trade. What are these deficiencies? Most countries do not survey export and import prices at all, although the number doing so has increased over the last twenty years. Most trade studies rely on indexes of export and import unit values, despite a long history of adverse appraisals of their accuracy as price measures and despite the use of unit values in some studies to measure quality differences or quality changes, uses that imply their inaccuracy for measuring price changes- Not only are the unit values inaccurate as measures of the prices of individual products, but in different countries they are combined with different weights and using different index-number formulas. They are relied on because they are almost universally available from customs records and have been for a long time. An alternative to the unit-value indexes is the use of domestic wholesale or producer-price indexes. These are collected with a much higher degree of quality control than is applied to the unit-value indexes, but the prices themselves, and their weighting in the indexes, do not purport to apply to external trade. The producer-price indexes are computed with different weighting, coverage, and index-number formulas in different countries, as are the unit-value indexes. A more general problem with both sources of data, and also with existing export and import price indexes, is that new products are underrepresented for some period after their introduction, and complex products are permanently underrepresented. A related problem is that price indexes for some manufactured -5- goods suffer from upward bias due to the neglect of quality change, an issue that has been discussed extensively in connection with the domestic producer and consumer price indexes. The two issues are related because it is the complexity and rapid change in specifications that lead to the omission of many products from most countries' price iindexes. Computers and computer accessories may be excluded completely, although they have become an increasingly important part of manufactured-goods trade. To deal with these problems of measurement of manufactured goods prices in international trade, the National Bureau embarked on a program of international price measurement research beginning over 30 years ago. The present BLS export and import price indexes for the United States were to some extent an outgrcwth of those studies. The first phase of the National Bureau's research was summarized in Kravis and Lipsey (1971), and the latest publication is Lipsey, Molinari, and Kravis (1991). This research has produced new measures of the prices relevant to the explanation of trade flows. These are now being broadened with respect to the country coverage of the indexes and by experiments with methods of dealing with quality change. Some of the characteristics of these measures are: 1. Weighting. Two types of indexes of manufactured goods prices are constructed. The one discussed here, which is used, for example, in studying world trends in terms of trade, is a "world" or developed-country index of manufactured-goods prices, using as weights aggregate developed-country exports or developed-country exports to developing countries. Indexes for individual countries are also constructed based on these two sets of weights. A second type of index is intended mainly for studies of competition in international trade. It is calculated for the United States and for the aggregate of many of its major -6- competitors, using the export weights of the United States and weighting at a fine level of commodity detail, such as the 4-digit SITC. Indexes of domestic prices with own-country export weights are also calculated for each of the three countries, to examine the possibility of divergences between export and domestic price movements and the implications for changes in margins of profitability of export and domestic sales. 2. Missing prices. In place of the assumption that prices for products not covered move in the same way as those covered in the same country, in the aggregate indexes for all developed countries, a method is used that incorporates both country and commodity effects in estimating missing prices. The method is an adaptation of the country-product dummy (CPD) method developed by Sumers (1973) for the estimation of country price levels. It involves fitting an equation to each block of country and commodity price change observations for a given year. The block is defined by the full list of commodities and countries in, say, a 2-digit SITC class. The equation contains dummy variables for both country and commodity and therefore permits the data to determine the degree to which each influences the estimate of the missing price. 3. Quality corrections. There have been almost no studies examining the relation of changes in product quality to the measurements of export and import prices, although there have been studies in which attempts were made to measure changes or differences in product quality by changes or differences in unit values (for example, Ohlsson, 1980; Aw and Roberts, 1988; Boorstein and Feenstra, 1991; and Rodrik, 1988). There have been a number of studies attempting to correct domestic U.S. prices for quality change, mostly by the use of hedonic price measures. Many of these are summarized in Gordon (1990). In the indexes used here, the effect on the U.S. export price indexes of introducing quality- - 7 - adjusted price measures for a few commodities La calculated, as well as the off -t on world price indexes, under two possible assumptions about the relation between quality and price change in the U.S. and price change in other countries' exports. B. Data Collection For these studies, export and domestic price data have been collected at the most detailed level available, some from price data on computer tape or diskettes, but most keypunched from published or unpublished price compendia. These price data, which go back to 1953 for most countries included, are arranged according to the 4-digit SITC Revision 1. The countries included, some mov:e extensively than others, are the U.S., the U.K., Germany, Japan, Canada, The Netherlands, Sweden, and France. We have also begun to collect data for Korea, Hong Kong, Singapore, and Taiwan (China), but they are not included in the indexes discussed hero. The data are classified into the SITC Revision 1 at the four-digit level and then aggregated to 3-digit, 2-digit, and 1-digit levels for SITC 5-8. and to total manufactures, defined as SITC 5 through 8. The methods are explained most recently in Lipsey, Molinari, and Kravis (1991). Factors Affecting Measures of Exnort Prices of Manufactured Goods One advantage of the NBER price indexes used here over the UN unit value indexes and other representations of manufactured goods prices is that the price indexes here are transparent; the ingredients and how they were put together are known. The UN unit value indexes, on the other hand, are derived from a variety of national indexes, put together by national statistical offices in many different ways, and some of which, in fact, are not even based on unit values. The estimates here of the influence of various factors, such as weighting and quality change, on measures of manufactured goods prices are, therefore, derived from the NBER indexes. The influence of each factor is measured in two ways. One is by its effect on the change from the initial year of the indexes, 1953, to the final year, expressed also as an annual rate of change. Since this method carries the risk that it may be excessively influenced by the circumstances of the first and last years, a linear logarithmic trend is also fitted to each of the series and the estimated coefficient of time used as the average rate of change. Excorts to the World vs. Excorts to DeveloRing Countries The composition of exports by developed countries (DCs) to developing coun.tries (LDCs) is somewhat different from that of their exports to developed countries and to the world as a whole. An indication of the effect of that difference on price measures is given by comparisons of changes in the price indexes for the two destinations between 1953 and 1991 (Table 1): Table 1 Comparison of Prices of Developed-Country Exports of Manufactures to the World and to Developing Countries 1988/1953 LDC DC exports to Relative to Weights of World LDCs World 1963 514 484 94.2 1975 485 454 93.5 1986 450 428 95.1 Source: Appendix Tables 1 and 2. The prices of developed country exports of manufactures to developing countries -9- declined by about 5 to 6 1/2 per cent over this period, depending on the base period used, relative to prices of exports to all destinations, or an average of between .13 and .18 per cent per year. If the average decline in prices for exports to developing countries relative to exports to the world is calculated from fitted equations in logs, the rate of decline is somewhat larger -- between .16 and .21 per cent per year, with the later weights indicating larger declines in the relative prices of exports to developing countries. Thus, an estimate of the terms of trade of developing countries based on data for developed-country exports to all countries would be biased downward slightly on this account. The Effect of Changing Base Years Price indexes have been calculated using several different base years, 1963, 1975, and 1986. The composition of trade in manufactures has, of course, changed substantially over this period. One way it has changed is that world trade has moved steadily toward more high-tech products, in the sense of products of industries in which R & D expenditures are important relative to value added or sales (Table 2). Table 2 Distribution of Market Economy Exports by Technology Class Low Medium High Technology Technology Technologv Total 1966 48.4 37.2 14.4 100.0 1977 43.2 39.0 17.8 100.0 1982 41.2 39.4 19.4 100.0 1986 34.3 41.2 24.5 100.0 1990 34.2 39.8 26.1 100.0 Source: Lipsey and Kravis (1992) and later UN Trade Tapes. Whether because of the sh5 ft over time to high-tech products or the more - 10 - general tendency of consumers to shift to products declining in price, the export price indexes based on 1975 weights rise less than those based on 1963 weights aid the indexes based on 1986 weights rise still less (Table 3): Table 3 Effect on Export Price Index, 1991/1953, of Changing Weight Base Years Index on Later Weights as Percent of Index on Earlier Weights 1975 Weights 1986 Weightc 1986 Weights 1963 Weights 1975 Weights 1963 Weights Developed Country Exports to World 94.47 92.84 87.71 Developed-Country Exports to Developing Countries 93.75 94.44 88.54 Source: Appendix Tables 1 and 2. The shift in commodity and country weights from 1963 to 1975 reduced the estimate of the price increase for manufactured goods by about 5 or 6 per cent, or about .15 per cent per year. The shift to a 1986 base reduced the rate of price increase still further, by about .19 per cent per year. Thus, the difference between the growth in prices calculated on 1963 weights and that based on 1986 weights is about a third of a per cent per year. The estimates from fitting logarithmic time trends are that the change from 1963 to 1975 weights reduces the growth in prices of exports to the world by about .11 per cent per year, and the change from 1975 to 1986 weights reduces it by about another .17 per cent or more per year, a total reduction by a little over a quarter of a per cent per year. For exports to developing countries the effect of the 1963 to 1975 weight shift is about .14 per cent per year and of the 1975 to 1986 shift, another .19 per cent, a total reduction of about a third of a per cent per year. - 11 - Within machinery and transport equipment, SITC 7, the effect of shifting weights was much larger. For exports to the world (Appendix Table 2), the index on a 1975 base rose more slowly than that on a 1963 base by more than a quarter of a per cent per year, as measured from a logarithmic regression, and the index on a 1986 base increased by about another third of a per cent less than that on the 1975 base. Thus, the shift of 20 plus years in the base reduced the rate of growth in prices of machinery and transport equipment by roughly .6 per cent each year. Countries apparently shifted their imports toward the products that were declining relatively in price much more extensively in machinery and transport equipment than in manufacturing as a whole. New and Rauidly Growing Countries Two of the major changes that have taken place in the world market for exports of manufactures have been, first and most important, the rapid growth of Japanese production and exports, and second, the rapid growth of exports by the Asian NICs. That growth could affect the markets and world prices for manufactures in several ways, depending on the way in which entry into or expansion in the share of world markets was accomplished. The most readily observable method of entry would be entry into world markets at high prices and low market shares, followed by relative declines in the newcomers' prices as their productivity grew relative to others. Indexes with later country weights would decline in comparison with indexes based on earlier weights, because they would give more importance to the new countries with declining prices. A second method of entry would be that the new countries enter world markets by selling at low prices, gradually raising their market shares as buyers learn about and shift to the new sources of supply, but with no change in relative prices offered - 12 - by old and new suppliers. Bath of these assume that the "law of one price" does not hold. A third method would be, if the law of one price does operate, that the new entrants influence the market by driving down the price for all suppliers, forcing less efficient producers out of the market. The various methods of market entry would show up in different ways in measures of price competition and of world price changes. The first scenario, with declining relative prices of new entrants, is exactly what is envisaged in measures of relative prices and price competitiveness, which would therefore be suitable for explaining shifts in market shares. However, as long-as separate export price indexes were constructed for each exporting country, the part of the price decline that consisted of a shift toward sources with declining relative prices would be missed in a fixed-weight world price index, although it would show up as a decline in indexes with later country weights relative to those with earlier weights. The second method of entry would be difficult to observe without measures of relative price levels, as well as of price changes, as in Kravis and Lipsey (1971). In both the present indexes and the standard unit value or price indexes, the effective decline in price would be missed, because both export unit values and export prices are compiled separately for each country, in effect treating each country's export of a product as a separate product. These effects could be caught in import price or unit value indexes but their only mark on exuort indexes would be in the rise in market shares of the low-price countries. Finally, in the third case, where new entrants bring down their rivals' prices there would be an indication of the change in country shares even in conventional export price indexes. A world export price index on Japanese weights, for example, would decline relative to an index based on world or other - 13 - countries' weights because the Japanese weights would give greater importance to products that are declining relatively in price. Some evidence for the existence of the third sequence of events is provided by a comparison of annual rates of growth of world manufactures export price indexes based on world weights with world export price indexes based on U.S., German, and Japanese weights. Table 4 Effect on Price Indexes for Developed-Country Exports of Manufactured Goods of Weighting by Export Patterns of the U.S., Germany, and Japan Average Annual Rate of Growth (Per Cent) of Price Index, 1953-1991 Weirhts of 1963 1975 1986 WeiEhts of Developed Countries 4.40 4.24 4.04 U.S. 4.41 4.26 4.03 Germany 4.39 4.18 4.20 Japan 4.00 3.95 3.80 Sources: Appendix Tables 2 and 3 World export price indexes for manufactures weighted by Japanese export weights rose less rapidly than indexes based on world, or U.S. or German, export weights, no matter what year's export composition is used, although Japan's advantage in this respect is smaller with the later year's weights. In other words, Japanese exports were more concentrated than, for example, U.S. exports on products for which prices were rising less r..pidly than average or were even declining. Either Japan's exports were more heavily concentrated on products for which productivity was rising relatively fast everywhere or its rapid productivity growth was driving down world prices of the products in which its exports were concentrated, whether or not other countries' productivity was growing rapidly in these products. - 14 - That phenomenon of the impact of new suppliers appears even more strongly if we compare developed-country export price indexes for 1991 relative to 1981 based on developed-country weights with developed-country price indexes based on the weights of Hong Kong, Korea, Singapore, and Taiwan (China). ..Table 5 Effect on Price.Indexes for Developed-Country Exports of Manufactured Goods of Weighting by Export Patterns of Taiwan (China), Hong Kong, Korea, and Singapore Average Annual Rate of Growth (Per Cent) of Price Index with 1986 Weights, 1981-1991 Developed Countries 4.42 Hong Kong 3.68 Korea- 3.54 Taiwan (China) 3.97 Singapore 3.37 Source: Appendix Table 9 Even over this short period, developed-country export price indexes based on the weights of the fast-growing Asian exporters declined substantially relative to an index based on the- developed countries' own export weights. The differences ranged from a half per cent per year to more than one per cent. Ouality Change Our attempt to adjust manufactured goods price indexes for changes in quality is based on the idea that neither unit value indexes nor price indexes based on specification pricing have captured the effects of changes in the quality of manufactured goods, particularly those that are complex or one-of-a- kind. Many such products are simply omitted from most price indexes and, when they are included, the prices tend to be list, rather than transactions prices, - 15 - because the actual products sold differ substantially from one transaction to another. There are virtually no studies of the relation of quality change to price measures that refer specifically to exported or imported goods aside from the studies mentioned above attempting to measure quality change from unit value data. A partial exception is a set of hedonic price indexes in Kravis and Lipsey (1971) as part of a much broader attempt to measure prices in world trade, but even there, most of the data for the hedonic price indexes actually referred to domestic markets. For years since 1982, there are also recent calculations by the Bureau of Economic Analysis (BEA) of the U.S. Department of Commerce for exports and imports of computers. Most of the literature on the adjustment of the price quotations for quality change involves the calculation of hedonic price indexes to replace the standard specification-based price measures. There are other types of corrections, such as the cost of improvement calculations used for automobiles by the U.S. Bureau of Labor Statistics, calculations based on prices in second hand markets, and corrections for the bias resulting from the slow introduction of new products into price indexes. However, what is used here are hedonic price indexes that have been calculated in various studies of domestic technological change and prices. The specific series we have used are listed in the notes to Appendix Table 5. What we did was to collect studies of quality change that had been carried out by others. All referred to domestic prices in the United States, and all the products were in SITC Division 7, Machinery and Transport Equipment. Within Division 7, they were confined to 9 four digit subgroups and 2 three-digit groups and drew heavily on Cordon (1990) and two or three papers on individual products. - 16 - We had to assume, in effect, that the quality adjustments applied equally to domestically used and exported products. Since only some products were covered, and since most of the quality corrections did not go beyond 1983, we felt that the adjustments we made were minimal even for SITC Division 7 for the Unitad States, and must surely understate what a comprehensive quality correction would show. For the export prices of the United States, based on the weights of developed-country exports of manufactures to developing countries, the effect of the quality adjustments is as follows, for all manufactures and for SITC 7 (Table 6): Table 6 Effect of Quality Corrections on Price Indexes for Manufactured Exports from the United States, 1991/1953 1975, and 1986 Weights of Developed-Country Exports to Developing Countries WeiEhts of 1975 1986 SITC 7 Unadjusted Index 4.353 3.516 Adjusted Index 3.134 2.545 Ratio: Adjusted/Unadjusted .720 .724 All Manufactures Unadjusted Index 4.165 3.700 Adjusted Index 3.511 3.079 Ratio: Adjusted/Unadjusted .843 .832 Source: Appendix Tables 4 and 5 For SITC 7, the group in which the quality adjustments were carried out, the rise in U.S. export prices was considerably less with the quality adjustment, by about one quarter, or .86 per cent per year- And that quality adjustment in SITC 7 alone results in a reduction in the estimated price increase for all U.S. - 17 - manufactured exports, assuming no omitted quality change in other groups, of almost a half of one per cent per year. In other words, the estimated upward bias from the omission of quality change is about a third for U.S. exports of machinery and transport equipment and almost 20 per cent for U.S. exports of all manufactures. If we estimate the average quality correction from an equation rather than from the change from beginning to end years, the relative decline in the adjusted price index for the U.S. is around .4 per cent per year for all manufactures and .7 or .8 per cent per year for SITC 7. There are several possible ways to calculate the implication of these quality adjustments for world price changes, of which we show two here. The most conservative is to adjust only the U.S. price index and combine the adjusted U.S. price index with the conventional indexes for the other countries. That calculation, referred to below as "adjusted for U.S. only," must understate the quality change bias in our conventional indexes. The other adjustment is performed by assuming, in effect, that for these products the law of one price holds in time-to-time form. It involves substituting the U.S. hedonic price index for a subgroup for whatever is reported by each country and also using that U.S. hedonic price index where no conventional price index is reported. The assumption flies in the face of the strong evidence that the law of one price is frequently violated. But it is more reasonable than the alternative assumption that the prices of these goods move as indicated by conventional indexes or, since most of these products are not covered by most countries' price indexes, the assumption that these prices move in the same way as those of all covered products. Some support for our assumption of common worldwide price changes is that the price index for U.S. - 18 - imports of computers and parts falls by about as much as the domestic and export prices. The effects of these quality change adjustments over the 1953 to 1991 period as a whole, calculated from indexes with 1975 weights, Ls shown in Table 7, for SITC 7 and for all manufactures. Table 7 Effect of Quality Corrections on Price Indexes for Total Exports of Manufactures by Developed Countries to the World and to Developing Countries, 1991/1953, 1975 Weights World Developing Countries SITC 7 Unadjusted Index 488.0 484.6 Adjusted for U.S. only 446.4 377.2 Adjusted for all countries 340.8 264.9 All Manufactures Unadjusted Index 485.2 453.6 Adjusted for U.S. only 458.4 407.8 AdJxsted for all countries 397.4 332.9 Source: Appendix Tables 1, 2, 6, and 7. The most conservative adjustment, assuming no quality change outside the U.S.., indicates an upward bias in the price indexes for developed-country exports to the world of about .15 per cent per year for all manufactures and .23 per cent per year for machinery and transport equipment (SITC 7) alone. If we adopt the assumption that in the 11 products with quality adjustments prices in other countries followed those of the U.S., the adjustment is much larger: over half of one per cent per year for all manufactures and almost one per cent per year for machinery and transport equipment. The adjustment is even larger for price indexes of exports to developing countries. The average declines in the adjusted indexes relative to the unadjusted - 19 - ones for exports to the world estimated from equations are, when the adjustment is made to all countries' indexes, more than a half per cent per year for all manufactures and a little over 1 per cent per year for SITC 7, slightly higher than the estimates from beginning to end years. The same relationship holds for exports to developing countries. Given the rough nature of these adjustments and the incompleteness of coverage, these estimates must be subject to a considerable range of error. However, it seems as if quality adjustments could well be large enough to offset the supposed long-run deterioration in the terms of trade of primary products, as it has been estimated by supporters of the thesis that there has been such deterioration. It is worth mentioning that these quality adjustments are much too gradual to offset the large short-term or medium term cycles in terms of trade, where the year-to-year movements are far greater than any conceivable offsetting changes in quality on the manufactures side. Thus they do little to alter the picture of deteriorating terms of trade for primary products during the 1980s. Comnarison with UN Unit Value Index for, Exorts of Hanufactures If we want to get some idea of what difference the use of these manufactures price indexes would make in comparison with the movement of commodity prices over the same period, we can start by comparing the UN unit value index to the price index for all manufactured exports to the world by developed countries, without the adjustment for quality change. That difference - 20 - 1291/1953 UN Unit Value Index 4.804 NBER Index for Exports of Manufacturss by Developed Countries, Unadjusted for Quality Change, 1986 Weights 4.505 NBER/UN .938 Source: Appendix Tables 2 and 8. of almost .4 per cent per year measured from beginning to end years, or .35 per cent estimated from an equation, might be the result of many factors, some possibly offsetting, but in any case, hard to identify. The difference between unit values and specification prices is there, as well as differences in weights, in country coverage, and in index number formulas. It is impossible to explain the differences between these price indexes and the unit value series reliably because the commodity composition of the individual-country unit value series is not known. The country indexes are combined into a world index using Paasche (current-year) weights, and many of the country indexes are of the same type. However, some important countries use other types of index numbers such as Laspeyres (United Kingdom) or Fisher (The Netherlands and, until recently, the U.S.). In addition, for eight countries, price indexes, rather than unit value indexes, are used to produce the world unit value indexes (United Nations, 1991, p. 42). As can be seen in the most recent description of the UN indexes (United Nations, l991a), the unit value indexes cover more countries than those presented here. They include 25 developed countries, instead of the 8 in the price indexes here. However, these 8 countries included here accounted for about three quarters of the total exports by the 25 in 1989. To judge by the unit value data, the use of the 8 countries in our list - 21 - produces an index that is biased upward to some degree, as compared with that for the full list of developed countries, at least during the 1980s. For 1980-89, a unit value index with 1989 export value weights for the 8 countries rose by 2 per cent, or about .2 per cent per year relative to the 25-country unit value series. Conclusions It seems likely that certain characteristics of the UN export unit value index for manufactures cause it to be biased upward as a measure of the prices of manufactured goods exported by developed countries to developing countries. It also appears that these biases could be large enough to offset the long-term deterioration in export prices of commodities relative to the UN unit value index for manufactured exports that has been calculated by proponents of the belief t-hat there has been such a deterioration. Thus, the case is weak that there has been a long-term deterioration in commodity prices in general relative to a price index, adjusted for quality change, of manufactured exports from developed countries to developing countries. The main ingredient in this conclusion is the adjustment for quality change, which we estimate to be at least .5 per cent per year, just about enough by itself to offset the reported deterioration in relative coamodity prices. We find an additional "bias" of about .35 or .40 per cent per year from the difference between our unadjusted index for exports by developed countries to all countries and the UN Unit Value Index, possibly, but not.necessarily, due to the use by the UN of unit value data rather than specification price data. The country selection for our index might bias it upward relative to a more comprehensive index by about another' .2 per cent per year, although we have - 22 - evidence for this only for the 1980s and the direction of the effect could be opposite, if it even exists, in earlier years. The use of an index weighted by exports to developing countries rather than to all countries mounts to about another .1 per cent or more per year. The sum of these items implies that the true rate of growth of manufactured goods prices may be lower than that in the UN unit value index by as much as 1 per cent per year. Although there is certainly room for disagreement with the assumption used here, there are grounds for judging that the adjustments are conservative. In particular, the assumption is made, in effect, that there are no overlooked failures to correct for quality improvements outside of the eleven groups for which hedonic measures are used here. While some of the obvious cases are covered in the adjustments included here, there are other likely candidates for adjustment. Even within SITC 7, there are products for which quality adjustments may be desirable, such as aircraft engines, agricultural machinery other than tractors, office machinery other than computers, such as copiers and fax machines, all, non-electrical machinery for special industries, radio and television receivers, electrical apparatus for medical purposes, such as diagnostic equipment, motor vehicles and parts except passenger chrs, and ships. In groups other than SITC 7, where we made no quality adjustments, pharmaceuticals price indexes are almost certainly biased upward, and to a substantial degree, and prices of plastics may also be subject to quality adjustments. This listing is not proof that there are any quality improvement overlooked in specification price indexes, but it suggests that we have not exhausted the possibilities for such adjustment. A further reason for considering our estimates of quality adjustment conservative is that the ones we have used, with the exception of two, are not - 23 - available after 1983. We are therefore, in effect, assuming that there was no further unaccounted for quality change outside of computers and semiconductors after 1983, an unlikely story. The conclusion here is that the estimate of long tern increases in world prices of manufactured goods derived from the UN unit value index is too high by at. least one per cent per year. Furthermore, the overestimate could be considerably greater if the quality corrections made here have counterparts in other manufactured products, even if there would be few, if any, on the scale of the corrections for computers and semiconductors. - 24 - References Aw, Bee Yan, and Mark J. Roberts (1986), "Estimating Quality Change in Quota- Constrained Import Markets: The Case of U.S. Footwear," Journal of International. Economics, Vol. 21, pp. 45-60 Boorstein, Randi, and Robert C. Feenstra (1991), "Quality Upgrading and its Welfare Cost in U.S. Steel Imports, 1969-1974," in Elhanon Helpman and Assaf Razin, eds.,. International Trade and Trade Policy, Cambridge, MIT Press, pp. 167-186 Cartwright, David W. (1986), "Improved Deflation of Purchases of Computers," Survey of Current Business, Vol. 66, No. 3, March Cartwright, David W., and Scott D. Smith (1988), "Deflators for Purchases of Computers in GNP: Revised and Extended Estimates, 1983-88," Survay of Current Business, Vol. 68, No. 11, November Cuddington, John T. (1990), 'Long-Run Trends in 26 Primary Commodity Prices: A Disaggregated Look at the Prebisch-Singer Hypothesis," Economics Department, Georgetown University Dulberger, Ellen R. (1993), "Sources of Price Decline in Computer Processors," in Murray F. Foss, Marilyn E. Manser, and Allan H. Young, Eds-., Price Measurements and Their Uses, NBER Studies in Income and Wealth, Vol. 57, Chicago, University of Chicago Press. Gordon, RLbert J. (1990), The Measurement of Durable Goods Prices, Chicago, University of Chicago Press Grilli, Enzo, and Maw Cheng Yang (1988), "Primary Commodity Prices, Manufactured Goods Prices, and the Terms of Trade of Developing Countries: What the Long Run Shows," The World Bank Economic Review, Vol. 2, No. 1, January 1988 - 25 - Haberler, Gottf:.ied (1959), International Trade and Economic Develooment, Cairo, National Bank of Egypt ______ ,(1961), "Terms of Trade and Economic Development," in Howard S. Ellis, Ed., Economic DeveloRment for Latin AMgerica, London, Macmillan Jevons, W. Stanley (1866), The Coal Question, London, Macmillan Kravis, Irving B., and Robert E. Lipsey (1971), Price Competitiveness in World Trade, New York, National Bureau of Economic Research Lewis, W. Arthur (1952), "World Production, Prices, and Trade, 1870-1960," Manchester School of Economic and Social Sciences, 20, No. 2, -May, 105-138 Lipsey, Robert E., Linda Molinari, and Irving B. Kravis (1991), "Measures of Prices and Price Competitiveness in International Trade in Manufactured Goods, in Peter Hooper and J. David Richardson, Eds., International Economic Transactions: Issues in Measurement and Empirical Research, Chicago, University of Chicago Press Lipsey, Robert E., and Irving B. Kravis (1992), "Sources of Competitiveness of the United States and of its Multinational Firms," Review of Economics and Statistics, Vol. LXXIV, No. 2, May, pp. 193-201 Mill, John Stuart (1848), Principles of Political Economy, New York, D- Appleton & Co., 1909, from the Fifth London Edition. Norsworthy, J.R., and Show-Ling Jang (1993), "Cost Function Estimation of Quality Change in Semiconductors," in Murray F. Foss, Marilyn E.-Manser, and Allan H. Young, Eds., Price Measurements and Their Uses, NBER Studies in Income and Wealth, Vol. 57, Chicago, University of Chicago Press. Ohlsson, Lennart A. (1980), Engineering Trade Specialization of Sweden and Other Industrial Countries, Studies in International Economics, No. 6, Amsterdam, North-Holland. - 26 - Ricardo, David (1832), Princinles of Political Economy Rodrik, Dani (1994), "Industrial Organization and Product Quality, Evidence from South Korean and Tainanese Exports," in Paul R. Krugman and M. Alasdair Smith, eds., Emnirical Studies of Strategic Trade Policy, Chicago and London, University of Chicago Press, forthcoming. Spraos, John (1980), "The Statistical Debate on the Net Barter Terms of Trade Between Primary Commodities and Manufactures," The Economic Journal, 90 (March), pp. 107-128 Summers, Robert (1973), "International Price Comparisons based upon Incomplete Data," Review of Income and Wealth, Series 9, No. 1, March Torrens, Robert (1821), An Essay on the Production of Wealth, London, Langman, Hurst, Rees, Orme, and Brown. Reprinted by Augustus 14. Kelley, New York, 1965 United Nations (1991a), Methods Used in Compiling the United Nations Price Indexes for External Trade, Vol. II, Statistical Papers, Series M, No.82 ____ (1991b), 1989 International Trade Statistics Yearbook, New York, United Nations. (1992), 1990 International Tr:ade Statistics Yearbook, New York, United Nations. (1993), Monthly Bulletin of Statistics, June, New York, United Nations. Viner, Jacob (1953), International Trade and Economic Development, Oxford, Oxford University Press. Policy Research Working Paper Series Contact Title Author Date for paper WPS1331 The Myth of Monopoly: A New View Annette N. Brown August 1994 M. Berg of Industrial Structure in Russia Barry W. Ickes 36969 Randi Ryiermnan WPS1 332 Poverty and Household Size Peter Lanjouw August 1994 P. Cook Marin Ravallion 33902 WPS1 333 A Test of the Intemational Norman V. Loayza August 1994 R. Martin Convergence Hypothesis Using 39026 Panel Data WPS1334 Taxation, Public Services, and the Juan Braun August 1994 R. Martin Informal Sector in a Model of Norman V. Loayza 39026 Endogenous Growth WPS1335 Labor Regulations and the Informal Norman V. Loayza August 1994 R. Martin Economy 39026 WPS1336 Modernizing Payment Systems in Robert Listfield August 1994 F. Montes-Negret Emerging Economies Femando Montes-Negret 37832 WPS1337 The Countrywide Effects of Aid Howard White August 1994 E. Khine Joke Luttik 37471 WPS1338 Commodity Exports and the Adding- Maurice Schiff August 1994 A. Kim Up Problem in Developing Countries: 33715 Trade, Investment, and Lending Policy WPS1339 China's Emergence: PRospects, Andres Boltho August 1994 J. Queen Opportunities, and Challenges UrnDadush 33740 Dong He Shigeru Otsubo WPS1340 Opportunity Cost and Prudentiality: Herbert L Baer August 1994 P. Sintim-Aboagye An Analysis of Fulures Clearinghouse Virginia G. France 38526 Behavior James T. Moser WPS1341 Explaining Pakistan's High Growth Sadiq Ahmed August 1994 A. Bhalla Performance Over the Past Two 84440 Decades: Can It Be Sustained? WPS1 342 Winners and Losers in Transition: Peler F. Orazem August 1994 Retums to Education, Experience, Milan Vodopivec and Gender in Slovenia WPS1343 Strategic Interdependence in the Watik Grais August 1994 K Zheng East-West Gas Trade: A Hierarchical Kangbin Zheng 36974 Stackelberg Game Approach Policy Research Working Paper Series Contact Title Author Date for paper WPS1 344 Which Foreign Investors Worry About Eric Bond August 1994 A. Estache Foreign Exchange Risk in South Anlonio Estache 81442 Asia and Why? WPS1345 The Decentralization of Public Jacques Cremer August 1994 A. Estache Services: Lessons from the Theory Antonio Estache 81442 of the Firm Paul Seabright WPS1346 Unking Competiton and Trade Bernard M. Hoekman August 1994 F. Hatab Policiles in Central and Eastern Petros C. Mavroidis 35835 European Countries WPS1347 Antitrust-Based Remedies and Bemard M. Hoekman August 1994 F. Hatab Dumping in Intemational Trade Petros C. Mavraidis 35835 WPS1348 Quality Change and Other Influences Robert E. Lipsey August 1994 J. Ngaino on Measures of Export Prices of 37947 Manufactured Goods WPS1349 The New Regionalism and the Threat Andrew Hughes Hallett August 1994 A. Kim of Protectionism Carls A. Primo Braga 33715