THE WORLD BANK ECONOMIC REVIEW Volume 4 May 1990 Number 2 Toward Equitable and Sustainable Rural Water Supplies: A Contingent Valuation Study in Brazil John Briscoe, Paulo Furtado de Castro, Charles Griffin, James North, and Orjan Olsen On the Accuracy of Economic Observations: Do Sub-Saharan Trade Statistics Mean Anything? Alexander J. Yeats The Impact of the International Coffee Agreement on Producing Countries Takamasa Akiyama and Panayotis N. Varangis Second-Best Foreign Exchange Policy in the Presence of Domestic Price Controls and Export Subsidies David Tarr Import Dependency and Structural Adjustment in Sub-Saharan Africa Ramon E. L6pez and Vinod Thomas Voluntary Export Restraints and Resource Allocation in Exporting Countries Jaime de Melo and L. Alan Winters THE WORLD BANK ECONOMIC REVI]EW EDITOR Ravi Kanbur ASSISTANT EDITOR Clara L. Else EDITORIAL BOARD Carlos Rodriguez, John Holsen Centro de Estudios Macroeconomicos, Buenos Aires Ravi Kanbur T. N. Srinivasan, Yale University Sarath Rajapatirana Joseph Stiglitz, Stanford University Marcelo Selowsky Dennis N. de Tray The World Bank Economic Review is a professional journal for the dissemination of World 13ank-sponsored research which informs policy analyses and choices. It is issued three times a year, with an occasional additional issue devoted to a special area of current research or policy interest. The Review is directed to an international readership among economists and social scientists in government, business, and international agencies, as well as in universities and development research institutions. The Review emphasizes policy relevance and operational aspects of economics, rather than primarily theoretical and methodological issues. It is intended for readers familiar with economic theory and analysis but not necessarily proficient in advanced mathematical or econometric techniques. Articles will illustrate how professional research can shed light on policy choices without becoming essays in advocacy. Inconsistency with Bank policy will not be grounds for rejection of an article. Articles will be drawn primarily from work conducted by World Bank staff and consuhtants. Before being accepted for publication by the Editorial Board, all articles are reviewed by two referees who are not members of the Bank's staff; articles must also be recommended by two out of the three external members of the Editorial Board. The Review may on occasion publish articles on specified topics by non-Bank contributors. Any reader interested in preparing such an article is invited to submit a proposal of not more than two pages in length to the Editor. The views and interpretations in articles published are those of the authors and do not necessarily represent the views and policies of the World Bank or of its executive directors or the countries they represent. When maps are used, the designations employed are solely for the convenience of the reader and do not imply the expression of any opinion wharsoever on the part of the World Bank or its affiliates concerning the legal status of any country, territory, city, or area, or concerning the delimitations of its boundaries or national affiliation. Comments or brief notes responding to Review articles are welcome and will be considered for publication to the extent that space permits. The Editorial Office welcomes inquiries and requests to translate or reprint material from the Review. Please direct all editorial correspondence to the Editor, The World Bank Economic Review, The World Bank, Washington, D.C. 20433, U.S.A. Subscriptions to The World Bank Economic Review are available without charge to readers with mailing addresses in countries that are not members of the Organisation for Economic Co-operation and Development (OECD) but are members of the World Bank. Written request is required every three years for renewal of such subscriptions. See the insert at the back of this issue for subscription charges and additional information. ©1990 The International Bank for Reconstruction and Development / THE WORLD BANK All rights reserved Manufactured in the United States of America ISSN 0258-6770 Material in this journal is copyrighted. Requests for permnission to reproduce articles should be addressed to the Director, Publications Department, The World Bank, Washington, D.C. 20433, U.S.A. The World Bank encourages dissemination of its work and will normally give permission promptly and, when the intended reproduction is for noncommercial purposes, without asking a fee. Permission is not required to make photocopies for classroom and similar immediate use. This journal is indexed regularly in Current Contents / Social & Behavioral Sciences, the Social Sciences Citation Index"', the Journal of Economic Literature, and the Public Affairs Information Service. THE WORLD BANK ECONOMIC REVIEW Volume 4 May 1990 Number 2 Toward Equitable and Sustainable Rural Water Supplies: 115 A Contingent Valuation Study in Brazil John Briscoe, Paulo Furtado de Castro, Charles Griffin, James North, and Orjan Olsen On the Accuracy of Economic Observations: Do Sub-Saharan 135 Trade Statistics Mean Anything? Alexander J. Yeats The Impact of the International Coffee Agreement 157 on Producing Countries Takamasa Akiyama and Panayotis N. Varangis Second-Best Foreign Exchange Policy in the Presence 175 of Domestic Price Controls and Export Subsidies David Tarr Import Dependency and Structural Adjustment 195 in Sub-Saharan Africa Ram6n E. L6pez and Vinod Thomas Voluntary Export Restraints and Resource Allocation 209 in Exporting Countries Jaime de Melo and L. Alan Winters THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2: 1 1 5-1 34 Toward Equitable and Sustainable Rural Water Supplies: A Contingent Valuation Study in Brazil John Briscoe, Paulo Furtado de Castro, Charles Griffin, James North, and Orjan Olsen Because many rural people are poor, it is usually assumed that rural water supplies must be financed by government agencies. It is now widely recognized, however, that many rural people can and will pay for improved water supplies, and that sustaining and extending services depends on mobilizing this willingness to pay. This article describes a study of willingness to pay for water in Brazil. The study shows that surveys of actual and hypothetical water-use practices can provide policy-relevant information on willingness to pay, which is shown to vary according to household socioeconomic characteristics and the characteristics of the existing and new supplies of water. In rural Brazil, tariffs for yard taps can be increased substantially before significant numbers of households would choose not to connect to an improved system, whereas provision of free water at public taps can protect the poor without jeopardizing the financial viability of the scheme. Billions of people in developing countries face daily problems in obtaining water for drinking, cooking, bathing, and washing. More than 1,500 million people-30 percent of the world's population-are estimated to be without access to uncontaminated water; and an unknown but large proportion have to spend hours daily to collect water (Briscoe and de Ferranti 1988, Churchill 1987). Because the adverse consequences for productivity, health, and quality of life are so obvious and so widespread, extensive efforts have been mounted John Briscoe is a unit chief in the Water and Sanitation Division, Infrastructure Department, the World Bank. Paulo Furtado de Castro is an economist at the Institute of Social and Economic Planning, Brasilia, Brazil. Charles Griffin is an associate professor in the Department of Economics, University of Oregon, Eugene, Oregon. James North is a graduate assistant in the Department of Economics, University of Oregon. Orjan Olsen is the research director of the Brazilian Institute of Public Opinion and Statistics (IBOPE), Sao Paulo, Brazil. The IBOPE team, comprising Marcia Cavalcanti Nunes, Olival No'boa Leme, Helio Gastaldi Filho, Fredimar Alex Vasconcelos, and Lidio Shimizu, made major contributions to the design and execution of this study. The authors have benefited from numerous discussions with colleagues including Wilton Bussab, Emilio Rodriguez, V. Kerry Smith, and Dale Whittington. Financial support was provided by the United Nations Development Programme, the Swiss Development Corporation, the Norwegian Agency for International Development, the World Bank, and the government of Brazil. © 1990 The International Bank for Reconstruction and Development / THE WORLD BANK. 115 116 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 to correct this problem, with increased emphasis since the inauguration of the United Nations' International Drinking Water and Sanitation Decade in 1981. Most of these efforts have been based on the observation that the unserved are mostly poor people and on the assumption that these people cannot afford to pay for improved services. Accordingly, it has been concluded that services can be sustained and coverage increased only by mobilizing more public re- sources (from external support agencies and governments), by stretching the limited resources as far as possible by providing only a low, "basic-need" level of service, and by minimizing the cost of providing a given level of service. The result of this sector philosophy has been that these water supply systems provide a low level of service (usually through public taps or hand pumps); they are heavily dependent on (often unreliable) government investiment financ- ing and transfers for operations and maintenance expenses; and the quantity and quality of service are unreliable. The results are predictable. These "improved" systems often do not function: it is estimated that one in four systems is not working at any one time, and that the number of systems being abandoned is approximately equal to the number of systems being commissioned. And even if they do function, often they are not used. In C6te d'Ivoire and Kenya, for instance, surveys have shown that only one-third of the population reported to have access to improved facilities actually used them. The result is that the wealthier invest in individual supplies to secure a reasonable quality and reliability of service, while the poor are thrown back on traditional sources which often entail substlantial time, health, and even monetary costs. In recent years many have realized that precisely because the benefits of improved water supplies are so great, many people in developing countries can and will pay for improved services. They will do so, however, only if they are provided with services which, in their eyes, constitute significant improvements over their existing supplies. Now the challenge is to identify, under a range of socioeconomic and environmental conditions, the level of service that people want and for which they are willing to pay. 1. WHY DEMAND STUDIES FROM INDUSTRIAL COUNTRIES ARE OF LIMITED RELEVANCE There is a large literature on water demand in industrial countries, with most studies pertaining to single-family residences in the United States (for example, Jones and others 1984). The focus of this literature is on the estimation of income and price elasticities of demand. The majority of investigations of water demand in developing countries have been modeled on this industrial-country literature (for example, Inter-American Development Bank 1985a, 1985b, 1985c, 1985d; Katzman 1977; and Hubbell 1977). The economics of water utility management in an industrial country is a relatively simple matter. All potential users will connect to the system, and all Briscoe, de Castro, Griffin, North, and Olsen 117 will have multiple taps in their yards and houses. Because the quantity of water used is relatively inelastic with respect to price, future needs and revenues for a given tariff can be projected with some confidence. In a developing country the situation is considerably more complicated. The number of potential users who will choose to connect to a system is heavily dependent on exogenous factors (such as the family's socioeconomic situation, and the cost and perceived qual- ity of their existing sources, including accessibility, reliability, and aesthetic characteristics), as well as on factors controlled by the utility (such as the level of service offered, the connection cost, and the tariff charged). There are obvious dangers inherent in designing rural water supply systems without reasonable information on what services people want and for what they are willing to pay. On the one hand, in many cases facilities are built for which a community would never pay. This is the case, for instance, in the rural communities in Zimbabwe that participated in this multicountry study (Robin- son 1988). Here protected wells were perceived as being little more than a marginal improvement over the traditional open wells, and, given the many alternative uses for their money, on the average families indicated that they would pay less than 0.5 percent of their income for the improvements. On the other hand, in rural communities in the Indian state of Kerala (Singh and Ramasubban 1989) the existing level of service-public taps-was much too low. Many families were prepared to pay high tariffs for a reliable yard tap supply. This article describes a study undertaken in three rural areas of Brazil. The study addresses three basic questions: * Are people's responses to willingness-to-pay questions believable? * How much are people willing to pay for water? * Is it possible to raise tariffs and increase revenues while protecting the poor? Il. MODELING WATER DEMAND: SPECIFICATION AND DATA Specifying the Model In rural settings, the principal decision of interest to a water supply planner is the proportion of families that will connect for a given level of service and given prices. Thus what is needed is a model that describes the probability that a particular family will choose to use a new water source. First it is assumed that a family chooses between two sources (j) based on maximizing two conditional indirect utility functions, the first of which de- scribes the utility gained from using the new source (j = 1), and the second the utility derived from use of the current, old water source (j = 2). In each case, the utility, U, for a particular family, i, of a particular source depends on time and monetary costs of obtaining water from that source, Cj, the perceived quality of its water, Vj, household income, Y', and a set of socioeconomic variables used as proxies for the families' tastes, Z,. The researcher does not 118 THE WORLD BANK ECONOMIC REVIEW, VOL. 4. NO. 2 observe all components of utility, so an error term, ej, is added to the utility function (McFadden 1974). The utility function for each household is thus: Uij = aeo + aICj + a2V. + U3Y, + a,Zi + Ey The probability that family i will decide to use the new rather than the old source is the probability that the conditional indirect utility function for the new source, Ui,, is greater than the conditional indirect utility function for the old source, U,2. Letting Pr = Ui, - Uj2, the probability that family i will connect is the probability that Pr > 0. Assuming that the distribution of the resulting error terms is normal, the parameters of Z can be estimated using the probit model (Judge and others 1980). Generating Data for Estimating the Model There are two basic procedures for generating data to assess what water- supply services people want and what amount they are willing to pay for these. First consider a situation in which a new water-supply service has been installed and in which each family has decided whether to switch to the new source. For each family information is collected on the choice of source (the dependent variable), the characteristics of the old and new sources, and family character- istics (the independent variables). A discrete choice model (such as the probit or logit) is then used to assess the effects of the independent variables on the probability of connection. The great attraction of this so-called indirect method is that iniferences are based on actual behavior rather than the actions that survey respondents say they would take. For some variables of interest (such as family characteristics) this method works well, because there is typically significant cross-sectional variation in these characteristics. The great drawback is that in cross-sectional studies across a city or region, the most policy-relevant variables, such as connection cost, tariffs, and levels of service, are usually constant across all respondents. Without variation in these variables it is not possible to assess household response to them. In longitudinal studies it is similarly difficult because complex relationships among the "independent" variables can give rise to spurious correlations. For instance, studies have often noted that increases in tariffs have been followed by increases in numbers of connections and per capita consumption. This is not attributable to some perverse price elasticity but rather to the fact that in many developing countries potential demand has been suppressed by supply constraints, and to the fact that it is frequently politically feasible to raise tariffs only after a utility has been able to effect service improvements. The second, alternative, procedure for assessing what services people want and what they are willing to pay for these involves the simple, obvious, and direct procedure of presenting people with hypothetical options (both in terms of quality of service and prices) and asking them to indicate what choice they Briscoe, de Castro, Griffin, North, and Olsen 119 would make. Surveys of this sort have widely been considered unreliable "due to the pervasive feeling that interrogated responses by individuals to hypothet- ical propositions must be, at best, inferior to 'hard' market data or, at worst, off-the-cuff attitudinal indications which might also be expected to reflect ef- forts by individuals to manipulate the survey to their selfish ends" (Cummings and others 1985, p. 50). In the specific case of rural water supplies, in the late 1970s a World Bank review of water-demand studies using such hypothetical questions concluded that the approach had been shown to be "virtually useless" (Saunders and Warford 1977). The major, obvious, problem with this method is that biases may arise, for three related reasons. First, individuals may not understand or perceive cor- rectly the characteristics of the good or service being described by the inter- viewer (hypothetical bias). Second, the respondent may think that he can influ- ence the provision of services in his favor by not answering the questions truthfully (strategic bias). And third, the respondent may give answers which are influenced by his desire to please the interviewer (compliance bias). Over the last decade, however, primarily in response to the difficult problem of valuing nonmarketed environmental goods, there have been significant con- ceptual and empirical advances in methodologies for conducting such "contin- gent valuation" surveys. Despite great initial misgivings about the usefulness of the method, it is now generally acknowledged that the major sources of bias can be addressed. First, where the hypothetical service to be introduced is not well known to the community (seldom a problem with water supplies) infor- mation can be provided through pictures, films, and discussions so that the nature of the service is clear. Second, it is possible to assess the magnitude of strategic biases through the use of a variety of techniques, such as choosing settings which would encourage or discourage strategic behavior, and compar- ing the effects of, say, family characteristics derived from contingent valuation and "indirect" methods. And third, through careful recruitment and training of interviewers, compliance biases can be minimized. As a consequence of these advances, the contingent valuation method has become a widely accepted method for evaluating the benefits of environmental improvements in industrial countries (Cummings and others 1985). III. A CASE STUDY: WILLINGNESS TO PAY FOR WATER IN BRAZIL In 1987 the World Bank initiated a multicountry study of willingness to pay for water. One objective of the study was to assess whether the contingent valuation method was reliable for assessing demand for public goods in devel- oping countries. The empirical studies were designed to assess how the propor- tion of families using a new system was affected by characteristics of the family and of the old and new water supply systems. The study was also designed to suggest what might constitute an appropriate water supply system both techni- cally and financially in different environmental and socioeconomic settings. 120 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 The studies were carried out jointly by a World Bank team and collaborating institutions in Brazil, India, Nigeria, Pakistan, Tanzania, and Zimbabwe. This article describes the Brazil study briefly; preliminary results from other country studies have been presented elsewhere (Robinson 1988; Singh and Ramasubban 1989; Altaf and others 1989; Whittington and others 1990). Site Selection Three areas were chosen for study in Brazil: one in the interior of the relatively prosperous, well-watered southern state of Parana; and two in poor, dry areas of the Northeast, one in northeastern Minas Gerais, the other in CearA. The study design called for identifying areas where improved services were available but where not all families had chosen to connect (A sites), and others where improved services were not yet available (B sites). The aim was to interview about 200 families at each site. Summary information on the survey sites is shown in table 1. The A villages already have an improved water source available. In all three cases the water- supply system comprised a piped distribution network, to which households could connect. To encourage households to use the new systems, connections were provided free, with the only charge being a flat monthly tariff. The Al and A2 designations represent houses that have chosen to install yard taps and Table 1. Summary Description of the Community Data Sets, Brazil State, region Hooked Level of Monthly tariff Sample size 3id range and village up? service (cruzados) (households) (cruzados) Ceara (Northeast) Al Yes Yard tap 41 100 50-2C0 A2 No Yard tap 41 100 0-40 B3 - Not set Not set 200 0-10 for public taps 15-100 for yard taps Minas Gerais (Northeast) Al Yes Yard tap 41 340 50-200 Parana (South) Al Yes Yard tap 41 140 50-200 A2 No Yard tap 41 52 0-40 Bi - Yard tap 41 100 10-200 B2 - Yard tap Not set 100 10-200 B3 - Not set Not set 100 0-10 for public taps 15-100 for yard taps -Not applicable. Note: A = improved services available; B = improved services not available; Al = yard taps installed; A2 = chose not to hook up; B1 = yard tap and prevailing prices to be offered; B2 = yard tap offered, price not determined; B3 = neither tap nor tariff system determined. At the time of the surveys, U.S. dollar = 25 cruzados. Source: World Bank data available upon written request to John Briscoe. Briscoe, de Castro, Griffin, North, and Olsen 121 those that have chosen to not hook up, respectively, within the A villages. The B villages do not yet have improved water sources, but B1 villagers will be offered yard taps at the prevailing price in the system, while B2 villagers know that they will get a yard-tap-based system but that the tariff has not yet been determined. For B3 villagers, neither the level of service (yard tap or public tap) nor the tariff structure has yet been determined. The Survey For all families information was collected on socioeconomic characteristics including measures of income, assets, employment, education, and family size and structure; and characteristics of the old water source(s), including level of service, distance from the house, reliability, and perceived quality of the supply. Bidding games were administered to all families. The basic form of the bidding game was to ask: "If you were required to pay X cruzados per month for a connection, would you choose to connect to the system or would you prefer to use the alternative source?" Each family was asked this question for a range of monthly tariffs. Based on experience in industrial countries (Cum- mings and others 1985), the sequence of the bids was to start at extremes (the lowest or highest value to be asked) and converge inward. Thus, for instance, if the prescribed values were 50, 100, 150, and 200 cruzados, the order would be (for a low starting point): 50, 200, 100, 150. For the families which had connected at the current charge of 41 cruzados (the Al households), the bids ranged from 50 to 200 cruzados. For families which had chosen not to connect at 41 cruzados (A2), the bids ranged down- ward to zero. In the B villages the range extended from 10 to 200 cruzados for yard taps and from 0 to 10 cruzados for public taps. The bidding games are summarized in table 1. Because the Ceara and Minas Gerais families were interviewed after the Parana' families, their prices were deflated to account for the effect of inflation. Analysis The probability of connecting to an improved system. Household character- istics are hypothesized to affect both tastes and the opportunity cost of time. It is expected that a family will be more likely to connect if it is relatively well-off and if the family head is relatively well educated, is employed in the formal sector, and owns significant assets. Using cluster analytic techniques, Bussab (1988) showed that the best measure of accumulated assets for the study families was the number of domestic electric appliances that the household owns. Our measure of wealth is a dummy variable equal to 1 if the number of electrical appliances owned by the household was more than one standard deviation above the mean. In villages where the type and tariff of future systems is already set, B1, households might be expected to overstate their maximum willingness to pay, 122 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 because they know that this would not affect the type of service or the tariff in their villages, but might think high bids could positively influence the agency to initiate service in their area. Households in B2 and B3 villages, however, might be expected to understate their actual willingness to pay, because they might hope that a low bid would influence the utility to provide service at a relatively low tariff. Families would be more likely to connect at low than at high tariffs, all other things being equal, and more likely to connect to a high level of service (yard tap) than a low level of service (public tap). Where the distance to the existing source is great, families also are expected to be willing to pay more for a yard tap. Two wrinkles in the estimation procedures are worth mentioning. The first relates to the fact that in the sample, the proportion of households that did not connect to the water system was larger than their share in the general popula- tion in the A villages. To obtain consistent parameter estimates when using nonlinear estimation procedures, such as the probit used here, the estimates take into account the nonrandom sampling (Maddala 1983). Second, the obvious way of entering the data (eight data points entered for a respondent who gives answers to eight specific bids) introduces correlation among the errors for each household. This procedure gives unbiased parameter estimates, but inflated estimates of the t-statistics and therefore of precision. This problem was dealt with through a straightforward procedure which is consistent with the recent "bootstrapping" literature, in which computer time is substituted for analysts' time in calculating test statistics for problems that are intractable from a deductive standpoint. Two types of runs were made. First, the point estimates of the parameters were estimated from a model that used all the data. But to estimate the true standard errors of these estimates, one observation was picked at random from each group of observations for each family, thus reducing the sample to the number of respondents. This procedure was followed ten times; the reported values for the t-statistics are based on the average values of the standard errors from these ten runs. The analysis was further complicated by a loss of the data on distance to the water source for the Ceara families. The following procedure was devised to analyze the effect of distance and to use the full sample to obtain relatively precise estimates of the effects of the other independent variables. First the parameters of the full model were estimated using the Parana and Minas Gerais data sets (run 1). Then the parameters of the full model were estimated using the same data sets but excluding the "distance to alternative source" variable (run 2). This run showed that all of the coefficients that are significant in the fully specified model are significant in the misspecified model, with the probit coefficients for these variables virtually identical in the two cases. Also, the misspecified model, excluding distance to the alternative source, was estimated for the complete data set for the three regions (run 3). The parameter estimates were consistent with those of the second run. Briscoe, de Castro, Griffin, North, and Olsen 123 Because probit coefficients are difficult to interpret, the effects of the right- hand-side variables on the probability of connecting are presented in two other ways. For continuous independent variables, the elasticities are the meaningful measure: the percent changes in the probability of connecting induced by 1 percent changes in the income, price, and distance variables. For discrete inde- pendent variables elasticities are not meaningful because small changes in the values are not possible, and so the marginal probabilities are presented: the changes in the probability of connecting when the values of the independent variables are changed from zero to one. The results for the full sample (run 3) are presented in table 2. In all cases the signs of the parameters are as expected, and in all but two cases the estimates are highly statistically significant. On the basis of the full model, including distance from source but excluding the Ceara data (run 1), the elas- ticity of the connection choice with respect to the distance of the old source was calculated as 0.03. The estimate was significant at the 90 percent level in a one-tailed test. What about the magnitude of the effects? One way of thinking about these is first to conceive of the intercept as the probability of hooking up for a (counterfactual) family with an uneducated head and no income, that owns few appliances, has a close current source, can have a yard tap for free, and lives in a village in Ceara with an improved system. From table 2, the probability of such a family connecting to the improved system would be 45 percent. Educa- tion would have a large effect: a family identical except that the household head had one to four years of education would have a probability 7 percent higher, and the effect of completing primary education would be to raise the probability by 20 percent. Being employed in the formal sector would raise the probability by 7 percent, and owning a significant number (beyond one stan- dard deviation above the mean) of electrical appliances would increase the probability by 17 percent. If the family lived in a B village the probability of connecting would be 5 percent higher; if the family lived in a B2 or B3 village the probability would be 29 percent (-0.34 + 0.05) lower. For the continuous variables, elasticities are computed at the sample means. For the average family, the probability of connecting would increase by 15 percent if income were doubled (to about $320 per month) and decrease by 68 percent if the tariff were doubled (to about $7 per month). The probability would increase by 3 percent if distance to the existing source were doubled to about 100 meters (from run 1, not shown in table 2). (The estimated elasticities are, of course, not valid for such large changes in the independent variables. The large numbers, however, give a sense of the magnitudes involved.) Another way of looking at the same data set is to model the effects of characteristics of the family and the existing source on maximum willingness to pay. Because the range of willingness-to-pay bids was arbitrarily truncated (at zero and 200 cruzados) a modification of the ordinary least squares proce- dure was used (the tobit procedure; Judge and others 1980). The results for 124 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 Table 2. Determinants of the Probability of Connecting to the Improved System (Full Sample, 1,164 Observations) Marginal Elasticity probability (for Probit Mean (for discrete continuous Independent variable coefficients value' variables) variables) Constant 1.17 +0.45 (8.11) Family characteristics Monthly household income (cruzados) 0.00004 4,336 +0.15 (3.74) Major appliance ownership 0.44 0.26 +0.17 (4.54) Formal sector employment 0.17 0.55 +0.07 (2.13) Head's education, 1-4 years 0.18 0.50 +0.07 (2.35) Head's education, >4 years 0.53 0.10 +0.20 (2.84) Characteristics of new source Real price of yard tap (cruzados per month) -0.01 87 -0.68 (15.33) Site B village 0.13b 0.42 +0.05 (0.57) B2 or B3 village -0.90 0.33 -0.34 (4.80) Parana state -0.26 0.41 -0.10 (2.35) Minas Gerais state -0.15b 0.25 -0.06 (0.84) Note: The mean value of the dependent variable, the probability of connecting, is 0.58. T-statistics are in parentheses. All coefficient estimates are statistically significant at the 95 percent level (one-tailed test) except as noted. a. Proportion of sample except for household income and yard tap price. b. Significance level is less than 90 percent (one-tailed test). Source: Authors' calculations, based on World Bank data available upon written request to John Briscoe. the full sample (excluding the distance to the alternative source) are presented in table 3. As expected, the results are consistent with those of the probit estimations-the signs of the parameters are consistent with a priori expecta- tions, and the estimates that were significant in the connection-probability model are significant in the willingness-to-pay model, with only minor excep- tion in the significance of site characteristics. Choosing between a yard tap and a public tap. Throughout the developing world increasing attention is being given to recovering part of the costs of rural Briscoe, de Castro, Griffin, North, and Olsen 125 Table 3. Determinants of Willingness to Pay for a Yard Tap Tobit Mean Independent variable coefficients value' Constant 89.8 (15.0) Family characteristics Monthly household income (cruzados) 0.0017 4,462 (4.43) Major appliance ownership 34.26 0.27 (6.65) Formal sector employment 12.9 0.56 (3.02) Head's education, 1-4 years 14.92 0.49 (3.31) Head's education, >4 years 27.49 0.11 (3.64) Site B village 14.59 0.43 (1.68) B2 or B3 village -75.95 0.34 (8.71) ParanA state 3.67b 0.40 (0.71) Minas Gerais state -2.23b 0.24 (0.37) Note: Based on full sample for the three regions, excluding distance to the alternate source (1,082 observations). The mean value of the dependent variable, the willingness to pay, is 104 cruzados per month. T-statistics are in parentheses. All coefficient estimates are statistically significant at the 95 percent level (one-tailed test) except as noted. a. Proportion of sample except for household income. b. Significance level is less than 90 percent (one-tailed test). Source: Authors' calculations, based on World Bank data available upon written request to John Briscoe. water supply systems from the users. There are, however, legitimate concerns that increased tariffs will force many people (and particularly the poor) to return to alternative, and often polluted, traditional sources. One response has been to suggest that where yard taps are the standard level of service (in much of Latin America), increased tariffs for yard taps might be accompanied by the provision of free water at public taps. Although this proposition offers a partial solution to the equity concern, it raises a concern that the availability of free public taps might act as an incentive for people not to connect to the system and may thus compromise the financial viability of the system. To address this concern, additional data were collected at those sites in Ceara and Parana where no decision had been made on either the level of service to be provided or the monthly tariff to be charged (B3 villages). In the bidding games, tariffs for both a public tap and a yard tap were presented to the 126 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 respondent. The tariff for using the public tap was initially zero and the tariff for the yard tap initially 40 cruzados. Depending on the initial choice made by the respondent, he or she was led through a series of tariff combinations in which the public tap tariff was increased to a maximum of 20 cruzados per month, and the yard tap tariff was increased to 100 cruzados per month. At each price combination, respondents were asked whether they would use their original source, a public tap or a yard tap. The specification of the choice model is as before, except that now there are three (alternative source, public tap, or yard tap), rather than the original two possibilities (alternative source or yard tap) for the dependent variable. This requires use of an estimation procedure-the multinomial logit (Judge and others 1980)-which is conceptually similar to the probit model, but which allows for more than two possibilities for the dependent variable. It is expected that (as in the probit model) the probability of using a yard tap would increase with income, whereas the current source and public tap are probably inferior goods. Both formal sector employment and higher education increase the opportunity cost of collecting water from outside of the house, increasing the probability of choosing a yard tap and reducing the probability of using the current source. With regard to the public tap, however, the higher opportunity cost of collecting water from a public tap would cause the coeffi- cient to be negative, but increased knowledge of the benefits of using treated water would lead to a positive coefficient. The net effect is difficult to predict a priori. One would expect standard own- and cross-price effects for the tap tariffs. It is also expected that families in Ceara (compared with those in Parana) would be more likely, all else being equal, to use both public taps and yard taps given the scarcity of water in Ceara. The results of the multinomial logit estimations are derived in three forms: the log odds ratios, the marginal probabilities, and the elasticities. As is the case for all regression procedures with qualitative dependent variables, the log odds ratios are difficult to interpret directly. A more meaningful presentation is of the marginal probabilities for the discrete independent variables and elas- ticities for the continuous dependent variables. Table 4 shows that the effects of income, price, and site are generally large and significant, and the direction of the effect is as predicted. Overall the correspondence between expected and actual signs is strong. Nineteen of twenty-one of the signs of the parameters are as expected: the two exceptions are signs we were not confident in predicting, namely the negative effect of one to four years of education on choosing a yard tap, and the positive effect of more than four years of education on choosing the current source. The implications of the results for rural water supply policy in Brazil. Consider the revenue effects of increasing the tariff charged for a yard tap. As expected, this would cause some families to switch to public taps and some to current sources. Table 4 shows that the direct price elasticity of the private tap at the mean is -0.47. That is, a 10 percent rise in the price of a yard tap Briscoe, de Castro, Griffin, North, and Olsen 127 Table 4. Determinants of the Probability of Using a Particular Source of Water Marginal probabilities (for discrete variables) and elasticities (for continuous variables) Current Public Yard Independent variable source tap tap Constant -0.09 +0.20* +0.30** (1.21) (1.43) (2.1) Family characteristics Monthly household incomea -0.42* -O.57** +0.24** (1.44) (1.80) (2.67) Major appliance ownership -0.14 -0.04 +0.18 (1.03) (0.41) (1.26) Formal sector employment -0.03 -0.06 +0.09 (0.24) (0.79) (0.31) Head's education, 1-4 years +0.03 +0.03 -0.06 (0.44) (1.06) (1.07) Head's education, >4 years + 0.03 -0.03 +0.01 (0.07) (0.05) (0.13) Characteristics of new source Real price of yard tapa +1.01## +0.81 -0.47** (1.94) (0.91) (2.33) Real price of public tapa +0.11 -0.36** +0.04 (0.74) (1.75) (0.81) Site CearA -0.39** +0.11** +0.28** (5.77) (1.91) (3.64) ##indicates statistical significance at the 95 percent level or above (one-tailed test). *indicates statistical significance at the 90 percent level or above (one-tailed test). Note: T-statistics are in parentheses. a. This is a continuous variable; therefore, the reported parameter value is an elasticity. Source: Authors' calculations, based on World Bank data available upon written request to John Briscoe. would induce a 5 percent decrease in the use of yard taps. Thus increasing tariffs above the mean (about 70 cruzados for these bidding games) would raise revenues, because the increased revenues from those staying connected would more than offset the revenues lost from those disconnecting. As expected, increasing the price of the public taps would reduce the number of households using them, although the decrease should be slight. The associated positive effect on revenues would not, however, be material to public tap policy, the goal of which is to provide a basic level of water supply for the poor, not to generate revenues. To get a feel for the magnitude of likely responses, it is informative to present these results graphically. Figure 1 shows how little water-use patterns in Parana' change as the tariff for a public tap is changed. But yard tap demand is much more price-responsive: as yard tap tariffs increase, connection probabilities 128 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 decline, and those who disconnect tend to revert to the alternative source rather than to using a public tap (see figure 2). In dry Ceara, at current tariffs the proportion using yard taps is high, but again water-use patterns are little affected by the public tap charge (figure 3). Because current water sources are less satisfactory in Ceara than in ParanA, as tariffs for yard taps increase a substantial proportion of the population would switch and use the free public taps (figure 4). Implications: Credibility, Willingness to Pay, and Protecting the Poor Credibility of the responses. What evidence is available to ascertain the extent of bias in the responses to the bidding games? There are several perspec- tives on this question. First, the field team comprised five highly experienced field researchers from the Brazilian Institute of Public Opinion and Statistics (IBOPE). They were initially skeptical of the notion of the bidding games but, after a few days of field testing, were convinced not only that the logic of the procedure was understood by respondents but that the respondents gave serious and thoughtful answers. The interviewers were confident that thiere was no hypothetical bias (because the product was well known to all respondents). The interviewers did, however, feel that there was some strategic behavior, particularly in the villages in which no decision had yet been made on whether a system was to be installed, or what level of service was to be provided or tariff charged (the B3 sites). The lower values of the willingness-to-pay param- eters for the B2 and B3 villages relative to otherwise comparable communities are consistent with these impressions. Second, the B villages were chosen so that there were different implicit incentives for strategic behavior. In villages in which the villagers were certain that a system was to be installed, and knew that the level of service and tariff were predetermined, B1, there appeared to be little incentive for strategic behavior. The marginal probability parameter on the "B village" variable in table 2 is not significantly different from zero, indicating that there was no discernible strategic behavior in these villages. In villages in which decisions had not yet been made on tariffs, B2, and level of service and tariffs, B3, there were substantially stronger a priori reasons to expect that respondents would underestimate their willingness to pay for a yard tap. The large and significant negative coefficient and marginal probability of connecting for these villages support this (table 2). A resident in a B2 or B3 village would indicate a maximum willingness to pay of 60 cruzados less on average than a comparable resident in other villages (table 3). Presumably the intention of such strategic behavior is to signal to decisionmakers that a lower tariff must be set for their region or they will not participate. From one perspective, this strategic bias is large (because average maximum willingness to pay is about 100 cruzados). Even in the B2 and B3 sites, how- ever, where there is evidence of considerable strategic behavior, the average Briscoe, de Castro, Griffin, North, and Olsen 129 Figure 1. Effect of Public Tap Tariff on Choice of Water Source, Parana' 1.0 0.8 St . 0.6 u 0.4 0 o 0.2 0.2 0.0 - 0 5 10 15 20 Public tap tariff (cruzados per month) Key: - current source; -- public tap; yard tap. Note: Results for villages in which neither the type nor tariff of the prospective water system has been determined. Figure 2. Effect of Yard Tap Tariff on Choice of Water Source, Parana' 8Current ariff ............ St' 0.6 f 0.4 0 t:02 0 -0 - - - . - - - --- - 0.0 - ~ ~ ~~~I III 0 50 100 150 200 Yard tap tariff (cruzados per month) Key: ._. _current source; - public tap; - yard tap. Note: Results for villages in which neither the type nor tariff of the prospective water system has been determined. 130 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 Figure 3. Effect of Public Tap Tariff on Choice of Water Source, Ceara 1.0 Q 0.8.- O 0.6 0 < 0.4 - ° 0.2 0.0 - I I 0 5 10 15 20 Public tap tariff (cruzados per month) Key: . current source; -- public tap; - yard tap. Note: Results for villages in which neither the type nor tariff of the prospective water system has been determined. Figure 4. Effect of Yard Tap Tariff on Choice of Water Source, Ceara 1.0 [ u . 0.8- ton Current tariff 0. . -0 0.4- 02 0.0 ...... T 50 100 150 200 Yard tap tariff (cruzados per month) Key: . current source; -- public tap; yard tap. Note: Results for villages in which neither the type nor tariff of the prospective water system has been determined. Briscoe, de Castro, Griffin, North, and Olsen 131 maximum willingness to pay was 74 and 56 cruzados, respectively. From the perspective of policy, the effect of strategic bias in this particular case is largely of academic interest, because the tariffs that would be justified on the basis of these "lower bound" estimates would still represent large increases over existing tariffs of only 41 cruzados. Third, the effects of virtually all variables were as expected on theoretical grounds. The effects of the economic variables (income, assets, tariffs for yard taps, and tariffs for public taps) were uniformly strong and sensible. What services do people want and for what are they willing to pay? For this sample, the average stated maximum willingness to pay for a yard tap was about 100 cruzados, 2.5 times the monthly tariff at the time of the survey, and about 2.3 percent of average reported family income for the sample. Because 35 percent of the sample appeared to behave strategically, this should correctly be regarded as a "lower bound" estimate of the average willingness to pay. The conclusion is that a majority of people are prepared to pay much higher tariffs than those currently being charged for yard taps in rural Brazil. As expected, willingness to pay for a yard tap is positively affected by income, assets, education, and formal sector occupation. It does not follow, however, that people in the poorer northeast Minas Gerais and Ceara are willing to pay less than relatively well-off villagers in the southeast Parana. This is because the alternative water source in Parana' is reliable dug wells dose to most homes, which is much more attractive than the alternative in Ceara- unreliable surface sources relatively far from most houses. The response to increased tariffs is different in the two regions. The own- price elasticity of demand for a yard tap connection is substantial in Parana (figure 2) where reliable dug wells are generally readily available. Despite the fact that families in CearA are much poorer, the impact of yard tap tariffs is less in Ceara (figure 4) than in ParanA, again presumably because alternative sources are much less satisfactory in Ceara. In relatively wealthy, well-watered Parana, families choose to use existing wells, with very few choosing to use public taps. In relatively poor and relatively dry Ceara, at high yard tap tariffs substantial numbers of families choose to use both public taps and alternative sources. Is it possible to have a financially viable tariff while protecting the poor? Households in Parana and Ceara were asked both about yard taps and public taps. The results suggest that households do not see these as close substitutes. Because prices for the two systems were varied, there was less movement between those two choices than there was back to the original choice. Moreover, families with more income and assets were more likely to choose a yard tap over a public tap. Assuming that the marginal cost of public taps is negligible, they offer a straightforward method to reach households that are unwilling or unable to pay for a yard tap. Given the low cross-price 132 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 Figure 5. Total Revenues under Different Tariffs for Yard Taps and Public Taps - 50 40 30 0 2 20 10 0~~~4 Yard tap tariff (cruzados per month) Key: z public tap tariff = 0; J public tap tariff = 20. elasticity between yard and public taps (table 4), and assuming that the water utility puts the public taps nearer to areas of poorer households, relatively few households would switch from paying for a yard tap to using a free public tap. What about the financial viability of a water utility under the alternative tariff structures? Figure 5 also shows that even when tariffs are 2.5 times current levels (which is the average maximum willingness to pay), total reve- nues rise as tariffs increase. But what about equity concerns? First, as implied by the positive income elasticity of demand for connections to yard taps (tables 2 and 3), it is the relatively better-off families in these rural communities who are choosing to install yard taps and who are, therefore, the principal beneficiaries of the subsidies implicit in this service. Second, as shown by the negative income elasticity of demand for use of public taps (table 4), it is the poor who would be the principal beneficiaries of free public taps. It would appear possible, then, to improve the equity of these rural water supply systems by effectively cross- subsidizing: by charging high tariffs for the yard taps used mainly by the better- off and providing free public taps that are used mainly by the poor. But is it possible to do this without threatening the financial viability of the rural water system? Figure 5 shows that the price charged for using water from a public tap has little effect on the revenues of the utility. Because of the need to protect the poor and because it is difficult to devise mechanisms for collecting payments for water use from public taps, the clear conclusion is that water should be provided free from public taps. Briscoe, de Castro, Griffin, North, and Olsen 133 IV. CONCLUSIONS The principal methodological question addressed by the study can be an- swered affirmatively. Well-designed and carefully administered surveys of ac- tual and hypothetical water-use practices can provide consistent, sensible, and believable information on willingness to pay for improved water supply ser- vices. The empirical results show that tariffs for yard taps can be increased very substantially before significant numbers of households would choose not to connect to an improved system. Major increases in tariffs for yard taps would both improve the financial viability of rural water supply schemes and reduce the subsidies that the better-off receive through heavily subsidized rates. The study also shows that the poor can be protected by providing free water at public taps, without jeopardizing the financial viability of the scheme. It should also be emphasized that these empirical results are specific to Brazil. Results of a similar study in Nigeria (Whittington and others 1990) are signifi- cantly different. The final publications of this multicountry study of water demand and pricing will document the methodological and policy implications from all six of the country studies. REFERENCES Altaf, M. A., H. Jamal, J. L. Liu, V. K. Smith, and D. Whittington. 1989. "Who Connects to Public Water Systems in Developing Countries: A Case Study of the Punjab, Pakistan." World Bank Infrastructure and Urban Development Department Discussion Paper. Washington, D.C. Processed. Briscoe, John, and David de Ferranti. 1988. Water for Rural Communities: Helping People Help Themselves. Washington, D.C.: World Bank, Bussab, W. 1988. "Disposicao a pagar por agua nas areas rurais." University of Sao Paulo. Sao Paulo, Brazil. Processed. Churchill, Anthony. 1987. Rural Water Supply and Sanitation: Time for a Change. World Bank Discussion Paper 18. Washington, D.C. Cummings, R. G., D. S. Brookshire, and W. D. Schulze. 1985. Experimental Methods for Assessing Environmental Benefits: Volume 1A: Valuing Environmental Goods: A State of the Art Assessment of the Contingent Valuation Method. Office of Policy Analysis, Planning, and Evaluation, U.S. Environmental Protection Agency. Wash- ington, D.C. Hubbell, L. K. 1977. "The Residential Demand for Water and Sewerage Services in Developing Countries: A Case Study in Nairobi." Urban and Regional Report 77-14. World Bank Development Economics Department. Washington, D.C. Processed. Inter-American Development Bank. 1985a. "Chile: Programs de Agua Potable Rural: Project Report." Washington, D.C. . 1985b. "El Salvador: Rural Water Supply Program: Project Report." Washing- ton, D.C. 134 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 -. 1985c. "Haiti: Second Stage of the Community Health Posts and Rural Drink- ing Water Supply Program: Project Report." Washington, D.C. -. 1985d. "Honduras: Rural Water Supply Program: Project Report." Washing- ton, D.C. Jones, C. V., J. J. Boland, J. E. Crews, C. F. DeKay, and J. R. Morris. 1984. Municipal Water Demand: Statistical and Management Issues. Boulder, Colo.: Westview. Judge, G. G., W. E. Griffiths, R. C. Hill, and T-C. Lee. 1980. The Theory and Practice of Econometrics. New York: Wiley. Katzman, Martin 1977. "Income and Price Elasticities of Demand for Water in Devel- oping Countries." Water Resources Bulletin 13: 47-55. Maddala, G. S. 1983. Limited Dependent and Qualitative Variables in Econometrics. Cambridge, England: Cambridge University Press. McFadden, Daniel 1974. "Conditional Logit Analysis of Qualitative Choice Behavior." In P. Zarembka, ed., Frontiers of Econometrics. New York: Academic P'ress. Robinson, P. R. 1988. "Willingness to Pay for Rural Water: The Zimbabwe Case Study." Harare, Zimbabwe: Zimconsult. Saunders, Robert, and Jeremy J. Warford. 1977. Village Water Supply. Baltimore, Md.: Johns Hopkins University Press. Singh, B., and R. Ramasubban. 1989. "Managing Improved Water Supply: Signals from Northern Kerala." Bombay, India: Centre for Social and Technological Change. Whittington, Dale, A. Okorafor, A. Okore, and A. McPhail. 1990. "Cost Recovery in the Rural Water Sector in Nigeria." World Bank Policy, Research, and External Affairs Working Paper 369. Washington, D.C. Processed. THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2: 1 35-1 56 On the Accuracy of Economic Observations: Do Sub-Saharan Trade Statistics Mean Anything? Alexander J. Yeats African governments are being urged to promote commodity exports, yet without reliable trade statistics it is difficult to formulate appropriate policies to achieve this goal. This article assesses the accuracy of U.N. trade statistics by comparing the declared value of African exports, plus a transport and insurance cost factor, with partner countries' reported import values. The results show that major discrepancies often exist between the two, with false invoicing and smuggling apparently responsible for much of the difference. Although major disparities exist in data on trade with developed countries, the average differences in intra-African trade statistics are substantially larger. Statistical tests show that these data cannot be relied on to indicate the level, composition, or even direction and trends in African trade. For more than thirty years economists have been aware of and attempted to correct discrepancies in developed countries' trade data observed in matched export and import statistics (Allen and Ely 1953; Ely 1961; Morgenstern 1963; U.N. Economic and Social Council 1974; Yeats 1978; OECD 1985). Much less attention has been given to the quality of developing countries' trade data because the potential for such analysis was limited by the lack of comparable disaggregated time-series information. It is clear, however, that there are im- portant reasons why such investigations should be undertaken. For example, efforts have been made to increase trade among developing countries through regional arrangements, such as the Andean Group, or through the recent plan for a Global System of Trade Preferences (GSTP) under which tariff concessions could be exchanged among all developing countries. The design and evaluation of these integration efforts require accurate and up-to-date information on participating countries' trade. On a broader scale, errors in developing country trade data could adversely influence government policies relating to investment, balance of payments, initiatives for the liberalization of trade barriers, ex- The author is an economist in the International Economics Department of the World Bank. He greatly benefited from an early discussion with Wolfgang F. Stolper on the methodological approach to be used in and the need for this study. The author would like to thank Jong-Goo Park, Paul Meo, Alfred Tovias, and Bela Balassa for comments and suggestions. ©D1990 The International Bank for Reconstruction and Development / THE WORLD BANK. 135 136 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 change rate policy, and a host of other factors that affect a nation's industrial- ization. In an attempt to evaluate the quality of statistics for one important group of developing countries, this study applies the trade reconciliation and evaluation procedures used on statistics from the Organisation for Economic Co-operation and Development (OECD) (Ely 1961; OECD 1985; Yeats 1978) to reported data on trade among Sub-Saharan African countries and between them and devel- oped countries. This study draws on previous research which identified eco- nomic and statistical factors that contribute to discrepancies in partner coun- tries' trade data. Bhagwati's pioneering studies (1964, 1967) showed that subsidies may encourage exporters to "overinvoice" shipments, whereas high tariffs create an incentive to underreport imports. The result may be that reported exports exceed matched imports (see also Sheikh 1974, 'Wulf 1981, and Gulati 1987). Overvalued exchange rates and foreign exchange controls may have the opposite effect. When the domestic currency is overvalued and exporters must turn in foreign exchange at the official (low) rate, there is an incentive to "underinvoice" and sell the unreported currency balance on the black market. Similarly, when the importing country has an overvalued ex- change rate and foreign exchange controls, the importer may overinvoice to obtain excess foreign currency and sell the balance on the black market. In this case, reported export values and quantities may be much smaller than reported imports. Higher import values than matched exports may also be the result of capital flight. Restrictions on private holdings of foreign assets may be adopted to induce domestic capital investment, and to reduce the demand for scarce for- eign exchange (sometimes in response to erosion of the value of the domestic currency). If an exporter is able to make necessary arrangements overseas, by underreporting foreign exchange earnings the excess can be placed in accounts or assets abroad. Similarly, by overinvoicing imports, the excess "paid" to the supplier can be deposited in accounts abroad. Although false invoicing often can be detected in matched trade data, a disquieting fact is that combinations of incentives may actually be self-disguis- ing. Trade data may show few discrepancies if an exporter earns subsidies while an importer is trying to accrue "extra" foreign exchange; both face incentives to overreport the value of shipments. Or, if an exporter wants to avoid export taxes while the importer faces import tariffs, both may underreport transac- tions. If the partners recognize their mutual interests in such false reporting and collude in it, the data may look quite consistent. A further important point is that all inconsistencies in trade data should not be attributed to illicit activities; a range of legitimate factors can cause differ- ences. Shipping costs, diversion en route, re-export of goods, differential time lags in reporting, multiple exchange rates, and differences between countries in commodity classification and valuation procedures all may cause discrepancies between matched trade data. Because exports are reported "free on board" Yeats 137 (f.o.b.), whereas imports normally include "cost, insurance, and freight" (c.i.f.), imports should exceed exports by the value of transport and insurance charges. In cases in which importers pay in advance (providing credit for delivery), the exporter may deduct finance charges from reported (f.o.b.) values, although these costs may be included in the c.i.f. import value. This would further widen the margin resulting from f.o.b. (export) and c.i.f. (import) reporting practices. Variations in exchange rates may cause trade data discrepancies. Exports and imports are first recorded in their respective national currencies, and if different official exchange rates are used to convert them to a uniform currency (say U.S. dollars), or if exchange rates change over the period of reporting, a disparity in the export and import figures will result. Another problem is that declared invoice values may be adjusted by customs authorities ("up-lifted") for assessing import duties and other taxes. Because these adjusted values are used in the importing country's official statistics, they may not correspond with those recorded by the exporting country. A problem that seems to be particularly prevalent in the African countries studied here is reporting discrepancies for transshipments in which goods are routed through countries bordering the exporter or importer. In these cases, the country of origin may inaccurately list a routing country as the importer, or the country of final destination may report the routing country as the exporter. A range of discrepancies may thus appear between the three (or more) parties to the transactions. Finally, the U.N. Statistical Office has followed some procedures that cause discrepancies in the African data to be underestimated. In several cases, the exact figure reported as imports by a partner country was inserted in the African country's export records and then designated as an estimate of trade. In these cases, there would be no differences between partner country trade data because the matched reported import figures were being used for both exports and imports. It was also clear that, in some instances, reported bilateral African trade for a given year was merely a reproduction of records relating to a different year. For example, the 1982 and 1983 records on Zimbabwe's imports from Malawi were generally identical down to the three-digit level of the Standard International Trade Classification (SITC). I. DATA AVAILABILITY: COVERAGE OF PERIOD AND PRODUCT The exclusive focus of this analysis is U.N. trade statistics, including the U.N. Series D Commodity Trade Tapes. They are the sole sources of developing and developed country export and import statistics which use a common clas- sification system-SITC-and are the most widely used source of data on South- South trade. Although national government publications may be available, they use varying product classifications that preclude accurate comparisons across countries. The United Nations reclassifies the government data to the SITC system using available concordances. Many policy or research studies are shaped 138 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 as much by the period and level of aggregation of the data as by the specific theoretical or empirical question being addressed. In addition, studies are con- strained by the interval or time span for which national trade statistics are available, particularly if trends or long-term changes in the commodity struc- ture of trade are being examined. Trade statistics for the developed countries with market economies are normally available from U.N. sources with a one- year lag. A continuous time series for these countries is available back to 1962 or 1963, with a product breakdown to the five-digit SITC level. In contrast to the situation for the OECD countries, Sub-Saharan African trade data as of mid-1989 generally extended back to 1962, but there are important gaps in the historical record (see table 1). In three cases, Botswarna, Lesotho, and Swaziland, no U.N. trade data exist because the trade of these countries is included in U.N. records for the South African Customs Union. In addition, the African countries' records show a much greater reporting lag: only six of the thirty-nine countries' records extend beyond 1983. Table 1. Trade Statistics for Sub-Saharan African Countries Available from U.N. Records as of July 1989 Years Years Region and country available Region and country available Customs and Economic Union Economic Community of the of Central Africa Great Lakes Countries Cameroon 1962-83 Burundi 1968-83 Central African Rep. 1962-83 Rwanda 1963-83 Chad 1962-83 Zaire 1962-83 Congo 1962-85 Other Africa Gabon 1962-83 Botswana Economic Community of West Djibouti 1969-83 African States Ethiopia 1962-85 Benin 1962-83 Kenya 1962-83 Burkina Faso 1962-83 Lesotho C6te d'Ivoire 1962-85 Madagascar 1962-86 Gambia 1962-83 Malawi 1964-83 Ghana 1962-83 Mauritius 1962-83 Guinea 1979-83 Mozambique 1962-83 Liberia 1962-84 Seychelles 1967-86 Mauritania 1962-83 Somalia 1962-83 Mali 1962-83 Sudan 1962-83 Niger 1962-83 Swaziland Nigeria 1962-83 Tanzania 1962-83 Senegal 1962-83 Uganda 1962-83 Sierra Leone 1962-83 Zambia 1964-83 Togo 1962-83 Zimbabwe 1979-83 -Not available: classified under South African Customs Union. Note: Records are not available for the following years: Burundi, 1979 (1965 data available); C6te d'Ivoire, 1984; Gambia, 1967, 1978; Mauritania, 1976-78; Rwanda, 1977; Senegal, 1977-78; Sey- chelles, 1969-70; Sierra Leone, 1977-78; Zaire, 1971; and Zimbabwe, 1966-78 (1963-65 data available). Data for six of the countries extended beyond 1983. Source: Compiled from U.N. Series D Commodity Trade Tapes. Yeats 139 The problems reflected in table 1 for the African countries do not exist for all, or even most, developing countries, although there have been persistent problems with some, such as India and Indonesia. Most Latin American coun- tries have had data available with a two- or three-year time lag (Mexico and Venezuela are important exceptions), whereas most of the Asian, newly indus- trializing countries have records that are as current as those of the developed countries. In addition, these countries' records normally extend back into the 1960s down to the four- and five-digit SITC level. The level of product detail in official trade statistics is often important in trade and commodity studies. Trade data for developed countries with market economies and many developing countries are compiled at very detailed levels. Many African countries' trade statistics lack this degree of precision and incom- pletely cover total trade even at the three- and four-digit level. In these cases the U.N. trade tapes allocate all trade possible given the available data, but this leaves some (possibly large) portion of trade unallocated. As shown in table 2, most African countries have relatively complete cover- age of total trade down to the three-digit SITC level. Mali is an exception: about one-quarter of its total exports are not reflected in the two-digit classification. But only eleven of the thirty-six countries retain full coverage at the four-digit level. The three-digit product groups are often too aggregated for product- specific studies, such as analyses of the influence of trade barriers (particularly tariff and nontariff barriers), or for investigations which require export and import unit values for fairly homogeneous products. Some problems in the U.N. trade tapes are clearly internal to the compilation process itself. For example, trade reported in component three- or four-digit SITC groups some- times exceeded the total reported at a higher level (see Ethiopia, table 2). II. THE AccuRAcY OF AFRICAN TRADE STATISTICS To assess the quality of the African trade statistics, African export data were compared with corresponding partner import statistics. Total exports of the African countries were tabulated for 1982-83, the last period for which data were available for all countries (see table 1). A two-year period was used to reduce the influence of time lags in recording trade flows and irregularities associated with a single year's statistics. Reported imports for the African countries' trading partners were matched with the corresponding African ex- port statistics. The percentage differences, P, between the matched data were computed as: (1) P = [(I - Ej,) . EJ,] x 100 where Eji are reported f.o.b. export values of African country j to destination i, and Iij are reported c.i.f. imports of destination i from j. 140 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 African Exports to All Partners The total reported exports of the thirty-six African countries and the per- centage difference between the corresponding reported partner countries' im- ports are summarized in table 3. The measured disparities are often large and Table 2. Level of Product Detail in African Countries' Trade Statistics, 1983 Total trade Proportion of total trade recorded (percent) (millions of dollars) Two-digit SITC Three-digit SITC Four-digit SITC Country Imports Exports Imports Exports Imports Exports Imports Exports Benin 348.5 79.4 100.0 100.0 99.6 99.5 - - Burkina Faso 287.5 57.0 100.0 100.0 100.0 100.0 100.0 100.0 Burundi 116.1 99.4 100.0 95.1 99.0 95.0 0.1 0.0 Cameroon 1,187.6 1,836.8 100.0 99.8 97.6 99.2 0.1 0.0 Central African Rep. 71.1 109.4 100.0 99.5 99.5 99.5 - 0.0 Chad 70.6 131.6 100.0 100.0 99.6 100.0 - 0.0 Congo 629.0 639.9 100.0 100.0 100.0 100.0 100.0 100.0 C6te d'lvoire 1,813.5 2,067.7 100.0 100.0 100.0 100.0 100.0 100.0 Djibouti 252.4 33.6 100.0 100.0 99.4 97.2 - 0.0 Ethiopia 587.0 422.6 100.0 98.7 139.9 98.2 0.2 0.0 Gabon 685.6 1,475.4 100.0 100.0 100.0 100.0 100.0 100.0 Gambia 79.4 45.0 100.0 97.4 99.7 97.4 0.1 0.0 Ghana 599.9 512.5 100.0 99.8 98.1 99.6 0.4 0.0 Guinea 252.6 420.5 100.0 99.4 99.3 99.3 0.1 - Kenya 1,379.1 947.3 100.0 100.0 100.0 100.0 100.0 100.0 Liberia 411.6 422.6 100.0 100.0 100.0 100.0 100.0 100.0 Madagascar 411.5 310.3 100.0 100.0 100.0 100.0 100.0 100.0 Malawi 310.5 239.2 100.0 100.0 100.0 100.0 100.0 100.0 Mali 303.8 98.2 100.0 76.9 99.5 76.3 - - Mauritania 350.5 290.7 100.0 100.0 99.8 100.0 0.1 0.0 Mauritius 441.6 360.8 100.0 100.0 100.0 100.0 100.0 100.0 Mozambique 500.1 239.8 100.0 100.0 99.8 99.9 0.1 0.0 Niger 209.3 261.6 100.0 100.0 99.7 100.9 0.1 0.0 Nigeria 7,008.4 12,381.8 100.0 100.0 97.9 99.9 0.4 - Rwanda 148.1 96.9 100.0 99.7 99.2 99.7 0.1 0.0 Senegal 790.1 440.8 100.0 99.7 98.9 98.8 0.1 - Seychelles 87.8 3.7 100.0 100.0 100.0 100.0 100.0 100.0 Sierra Leone 165.7 90.7 100.0 100.0 100.0 100.0 100.0 100.0 Somalia 352.3 149.9 100.0 100.0 99.5 99.8 0.3 0.0 Sudan 1,424.0 601.1 100.0 100.0 99.5 98.9 0.1 0.0 Tanzania 537.7 425.0 100.0 99.6 98.2 99.3 0.4 0.0 Togo 479.9 225.6 100.0 99.3 99.6 99.2 - 0.0 Uganda 257.6 360.1 100.0 99.9 99.6 99.9 0.1 0.0 Zaire 841.7 1,387.9 100.0 98.4 98.5 98.1 0.2 0.0 Zambia 560.8 825.4 100.0 100.0 99.4 99.9 0.2 - Zimbabwe 449.6 672.2 100.0 100.0 100.0 100.0 - - -Negligible (less than 0.5 percent of total trade). Note: All countries reported 100 percent of total trade at the one-digit SITC level. Trade data are not available for Botswana, Lesotho, and Swaziland, which therefore are omitted from this and subsequent analyses. Source: Author's calculations, based on U.N. Series D Commodity Trade Tapes. Yeats 141 far exceed the average 3-6 percent differences observed for trade between developed countries (see Yeats 1978; OECD 1985). Differences of 100 percent or more are frequently observed, and disparities of more than 600 percent are calculated on the exports of the Gambia, Liberia, Niger, and Seychelles. Some proportion of the difference between f.o.b. export and c.i.f. import prices is accounted for by the costs of transport. Because the most accurate and comprehensive data for calculation of transport correction factors are available from U.S. Customs invoice data, the factors and analysis here are based on African exports to the United States (see the appendix). In general, U.S.-based transport margins should be between 5 and 15 percent of the value of f.o.b. exports (the margins for the Gambia, Guinea and Somalia are higher). But, as table 3 demonstrates, there are numerous bilateral trade flows in which the differences greatly exceed these transport cost margins. For example, the high- est recorded nominal freight rate for Gabon's exports to the United States over 1982-87 was about 9 percent, yet table 3 reports a difference in matched f.o.b.-c.i.f. partner data of 75 percent. Niger is the most extreme case: ship- ments to the United States have a maximum freight factor of 9 percent, but the difference between the matched trade data exceeds 300 percent. Given the extremely limited foreign exchange reserves of most African coun- tries, the fact that the reported value of their exports in importing countries exceeds domestic reported values by more than $100 million in several cases takes on special importance. For example, CBte d'Ivoire reported exports of $2.3 billion (billion = 1,000 million) to the EEC, whereas the latter reported imports valued at more than $500 million higher. Discrepancies of more than $500 million also occur on several other bilateral trade flows (for example, Cameroon-EEC, the Congo-the United States, Gabon-the United States, Ni- geria-EEC). Matched export-import data were compiled at the three-digit SITC level for every total bilateral trade flow reported in table 3 that showed a difference of at least $20 million. When possible, matched quantity and unit value statistics were also computed for each of the partner countries. This procedure was adopted to identify specific product groups that generate the overall discrepan- cies, and to indicate if price or quantity differences might cause them. Table 4 summarizes the results of this analysis for twenty-five bilateral trade flows. As indicated, several common factors appear responsible for many of the statistical discrepancies. For oil-exporting countries such as Cameroon, the Congo, and Gabon, the data suggest purposeful underreporting of export quan- tities and values of shipments, possibly to conceal noncompliance with inter- national agreements on production and export quotas. Similarly, quantity and value discrepancies in the coffee and cocoa shipments of CBte d'Ivoire, Ghana, Kenya, and Madagascar may result from attempts to evade quotas established under international commodity agreements. These situations may also reflect, however, false invoicing by exporters to evade foreign exchange controls. The discrepancy for Burundi and the Central African Republic is almost entirely accounted for by precious stones, items that can easily be smuggled out of a Table 3. Reported Exports from African Countries and Their Relation to Matched Partner-Country Imports, 1982-83 Exports to (millions of dollars): Percentage difference between reported imports and exports' All Sub- All Sub- developed United Saharan developed United Saharan Exporting country countries Canada EECb EFTA' Japan States Africa countries Canada EEC" EFTA' Japan States Africa All Sub-Saharan Africa 50,482.1 405.7 27,216.1 1,538.0 1,522.2 17,497.0 3,016.2 14.4 -2.9 15.2 17.5 22.3 11.6 13.1 4a Benin 91.1 - 50.4 7.5 3.5 28.7 29.2 13.0' 154.5' 20.1* 0.4 5.1 5.1 -96.2 9 Burkina Faso 47.3 0.6 39.4 1.0 5.4 0.1 30.8 82.7 -100.0 63.4 141.1 238.8 242.2 -60.2 Burundi 197.0 0.0 113.1 21.7 8.8 47.4 2.2 16.7 0.0 29.0* -2.5 0.0 0.0 52.9 Cameroon 2,607.5 1.2 1,547.3 15.0 40.6 947.8 91.6 39.8: -1.9 35.9* 55.8't 80.1* 43.3' 36.6 Central African Rep. 135.1 0.0 95.3 2.4 18.5 8.6 2.0 49.3' 0.0 67.8' -29.1 6.2 1.3 11.0 Chad 146.7 0.0 39.4 1.3 3.0 70.6 21.7 3.8 0.0 9.6 0.0 0.0 0.0 -50.3 Congo 1,623.5 0.0 483.1 3.4 12.0 995.2 5.5 46.7" 0.0 36.4': 37.1 * 47.8* 54.2* 85.0 Cote d'lvoire 3,176.9 15.0 2,251.2 20.5 100.3 578.5 727.9 25.5* 23.2* 23.8' 207.3* 12.5 20.4' -3.5 Djibouti 10.7 0.0 9.8 0.7 0.0 0.1 38.7 1.3 0.0 1.4 0.0 d 0.0 -81.2 Ethiopia 541.6 2.3 272.0 11.2 50.8 198.2 74.7 11.3 57.2 7.8 50.6' 23.4' 2.8 47.5 Gabon 2,475.7 72.5 1,401.7 64.7 12.8 752.5 55.6 27.2' -88.3 6.5 -30.1 58.6* 75.4' 75.2 Gambia 54.8 0.1 34.1 14.2 0.0 0.4 1.0 19.3 0.0 30.9 0.0 0.0 0.0 1,138.2 Ghana 1,005.9 2.6 516.3 41.1 122.9 280.9 19.8 31.0' 116.5* 4.8 79.6' 15.8' 76.4* -43.1 Guinea 758.1 34.9 281.3 14.6 0.3 304.1 37.3 1.6 0.0 2.5 -1.2 0.0 0.0 -34.4 Kenya 941.7 15.2 710.3 65.1 18.8 115.4 466.2 28.0' 33.1" 24.3' 52.1' 43.8'* 27.2* 12.3 Liberia 842.1 2.3 625.2 1.2 9.6 158.7 18.3 82.4' -95.2 47.8' 1,024.4' 2,011.8* 39.0 137.9 Madagascar 457.9 0.5 275.4 3.1 49.2 118.0 5.7 19.2': 62.0.' 6.7 139.5' 72.1 18.7' -48.4 Malawi 366.3 2.3 242.5 18.4 20.7 32.5 53.1 6.5 160.6* 8.1 15.5 5.4 51.3* 46.6 Mali 129.3 0.0 106.8 2.2 11.2 0.7 140.1 -12.0 0.0 -15.1 -5.3 4.4 166.1 * -87.4 Mauritania 469.5 - 290.7 0.1 132.4 1.3 39.6 14.2 0.0 22.9 7.4 0.0 0.0 -4.5 Mauritius 698.6 8.9 613.1 10.1 0.1 60.7 3.2 7.1 -14.8 8.2 -23.5 -309.1' 10.8 -41.2 Mozambique 292.7 1.0 107.5 3.9 38.2 87.4 56.6 3.3 0.0 7.3 37.3 0.0 0.0 93.3 Niger 532.4 0.1 527.3 - 2.5 1.2 21.8 -3.8 -100.0 -5.3 -100.0 120.6* 327.2* 641.5 Nigeria 25,542.0 208.8 12,285.0 935.5 14.7 11,156.2 624.8 4.1 0.0 7.4 0.1 0.0 0.0 -13.7 Rwanda 154.2 0.0 70.4 14.7 4.1 64.8 8.3 3.6 0.0 7.5 0.7 0.0 0.0 12.1 Senegal 844.4 1.8 663.0 10.0 82.5 21.5 256.1 -19.1 -38.7 -10.2 -68.6 -67.6 -84.1 45.0 Seychelles 1.1 - 0.3 - 0.7 - 0.2 3,806.2- e 8,043.6V 600.0* 125.8* f 89.4 Sierra Leone 182.2 0.0 136.1 5.1 0.8 39.5 3.8 42.1* 40.4* -54.2 59.3* 52.1 * 134.9 Somalia 51.3 0.3 49.5 - 0.1 1.3 2.6 7.3 0.0 7.6 3.8 -1.3 -5.1 50.7 Sudan 365.4 3.0 271.7 10.6 41.5 30.3 3.8 37.7x -58.6 23.3* 50.3* 122.3; 24.9 * -17.8 Tanzania 598.4 3.7 487.3 35.5 29.3 33.0 47.6 0.3 4.4 -10.0 26.2* 24.2* 45.3': 177.2 Togo 256.5 0.1 201.3 10.2 4.1 30.9 72.2 12.2 -26.6 14.8 -1.5 5.9 2.8 65.0 Uganda 669.0 1.0 280.5 3.6 42.4 277.2 5.2 3.5 0.0 1.8 430.6* 0.0 0.0 11.9 Zaire 1,805.8 25.8 746.9 31.5 155.4 802.1 26.1 51.9 0.1 124.6: 2.0 0.0 -0.1 -33.7 Zambia 1,308.4 1.6 728.9 97.2 394.0 84.4 26.6 4.2 0.0 6.9 1.4 0.0 0.0 12.0 Zimbabwe 1,100.3 0.0 661.8 61.0 127.0 166.3 65.6 7.7 h 6.0 11.8 0.0 0.0 -16.1 -Negligible. Note: An asterisk (*) indicates that the partner country trade difference exceeded by at least 5 percentage points the maximum recorded nominal freight rate for exports of the African country to the United States at any time between 1982 and 1987. For the African-European data, both a shorter distance and generally larger shipment volumes to Europe suggest that freight costs should be lower than on shipments to the United States, and thus this underestimates the number of cases in which disparities 4a could not reasonably be attributed to transport costs (see the appendix). a. The difference between the value of total reported imports and exports divided by the value of exports, times 100. b. European Economic Community (ten member countries). c. European Free Trade Association. d. Djibouti reported no exports to Japan during 1982-83; Japan reported imports from Djibouti. e. Seychelles reported $6,000 in exports to Canada; Canada reported $33,863,000 in imports from Seychelles. f. Seychelles reported $7,000 in exports to the United States; the United States reported imports of $3,348,000 from Seychelles. g. Sierra Leone reported no exports to Canada; Canada reported $56,000 in imports from Sierra Leone. h. Zimbabwe reported no exports to Canada; Canada reported $7,817,000 in imports from Zimbabwe. Source: Author's calculations, based on U.N. Series D Commodity Trade Tapes. Table 4. Analysis of Differences of More than $20 Million between Matched Reported Import and Export Values, 1982-83 Reported difference~' Value (millions Exporter-importer Major commodities traded (percent share) Observations Percent of dollars) Burkina Faso-European Oilseeds (32), cotton (16), hides (10) Roughly half the discrepancy is in oilseeds, for 63.4 25.0 Economic which unit values differ by more than 40 Community (EEC) percent Burundi-EEc Coffee (61), precious stones (16), natural Precious stones imports exceeded exports by 29.0 32.8 abrasives(14) $22.2 million Cameroon-EEc Petroleum (39), cocoa (19), coffee (15) Discrepancy almost entirely due to 35.9 554.8 underreporting petroleum shipment volume Cameroon-U.S. Petroleum (89), petroleum products (4) More than 80 percent of discrepancy caused by 43.3 409.9 underreporting crude petroleum shipments Central African Precious stones (42), coffee (40), cotton (7) Precious stones imports exceeded exports by 67.8 64.6 Republic-EEC $98.2 million Congo-EEC Petroleum (63), wood (9), precious stones (7) Difference caused by underreporting volume of 36.4 175.6 petroleum exports Congo-U.S. Crude petroleum (94), petroleum products (4) Petroleum imports exceeded exports by $494 54.2 540.0 million C6te d'lvoire-EEC Coffee (26), cocoa (26), wood (16) Coffee import unit value exceeded exports by 24 23.8 536.6 percent Cote d'lvoire-European Cocoa (41), coffee (17), fruit (16) Cocoa imports exceeded exports by $45 million. 644.1 74.6 Free Trade Differences of $5 million to $10 million in Association (EFTA) coffee and fruit trade Gabon-U.S. Petroleum (99) Underreported petroleum exports 75.4 567.9 Ghana-U.S. Aluminum (73), cocoa (11), petroleum (7) Aluminum imports exceeded exports by $159 76.4 214.7 million. Major differences in reported quantities traded Ghana-EFTA Cocoa (93), nonferrous ore (5) Cocoa imports exceeded exports by $31.4 94.8 34.6 million. EFTA omits quantities so source of error could not be determined Kenya-EEC Coffee (39), tea (27), fruit (19) Tea and coffee account for about $80 million of 24.3 172.8 the total discrepancy Kenya-U.S. Coffee (51), crude vegetable material (16), Tea imports exceeded exports by $11 million 21.7 31.4 tea (19) Kenya-EFTA Coffee (85), fresh fruit (2) Coffee imports exceeded exports by $34 million. 68.2 41.7 EFTA omits quantities so source of error could not be determined Liberia-Japan Special transactions (88) No "special transactions" (SITC 931) exports 2,001.8 193.9 reported. Japan's imports were $179 million for this item Liberia-EEC Iron ore (66), precious stones (11) Iron ore import unit value exceeds exports by 47.8 298.6 more than 40 percent Liberia-EFTA Ships and boats (98) Liberia failed to report quantities so source of 1,024.4 127.0 error could not be determined Madagascar-Japan Fish (60), coffee (22), spices (7) Over half of the discrepancy is due to fish: 72.1 35.4 exports unit value more than 40 percent below import unit value Madagascar-U.S. Spices (58), coffee (28) A discrepancy of $15 million exists in the 18.7 22.1 reported coffee trade 4 Mauritania-EEC Iron ore (80), fresh fish (15) Difference caused almost entirely by iron ore. 22.9 66.6 Quantity information was not reported so unit values could not be computed Seychelles-Canada Sugar and honey (98) Sugar imports exceeded exports by $33 million b 33.9 Sierra Leone-EEC Nonferrous metals (28), pearls and precious Approximately 44 percent of the total 40.4 55.1 stones (27) discrepancy is accounted for by pearls and precious stones Sudan-Japan Cotton (61), oilseeds (21) Cotton imports exceeded exports by 80 percent. 122.3 50.7 Oilseeds imports exceeded exports by 460 percent Sudan-EEC Crude vegetable material (21), cotton (20), Oilseeds imports exceeded exports by $31 23.3 63.4 oilseeds (15) million. Unit values could not be computed because quantity information was not available a. Value reflects the amount by which reported imports exceed reported exports. Percent expresses this difference relative to reported exports. b. Canada reported $33.8 million in imports; Seychelles reported $6,000 in exports to Canada. Source: Author's calculations, based on U.N. Series D Commodity Trade Tapes. 146 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 country in order to evade taxes and secure foreign currencies. Rationales for the discrepancies for most other commodities are less obvious, with the excep- tion of one or two products which constitute special situations. Differences between official and black market dollar exchange rates were compiled for as many countries listed in table 3 as possible. The differences in exchange rates were found not to be significantly correlated with discrepancies in trade values, possibly because many smuggled goods were not reported in either export or import statistics. If smuggled goods were accurately reported in one of the partners' data, the correlations might achieve statistical signifi- cance. Trade among African Countries Because the discrepancies in data on intra-African trade were generally found to be considerably larger than those on trade with developed countries, more detailed partner country statistics were compiled and analyzed for these ex- changes. Table 5 shows the total reported value of each country's exports to all Sub-Saharan Africa and to its largest trading partner and gives the partner's reported imports. Similar information is also shown for African trade in man- ufactured goods. Major discrepancies in the African data are apparent from table 5. An average difference of 64 percent occurs in the matched total trade data of thirty-five of the countries (109 percent if the Gambia is included); the differ- ences with each exporter's largest single trading partner average 61 percent. A peculiarity revealed by the table is the fact that reported imports are less than reported exports for the total trade of eighteen of the thirty-six countries, and for the largest trading partner of twenty-one of the countries. These findings are unexpected because intra-African transport and insurance costs, which would be excluded from exporter f.o.b. values but included in importer c.i.f. import values, have often been found to reach 50 percent or more of a product's export value (Livingstone 1986). Table 5 shows the African statistics may be of limited use for identifying directions of trade or even trading partners. For example, Benin reported 1982-83 exports of $19 million to its major trading partner, Nigeria, but the latter reported no trade between them. A similar situation occurred for Dji- bouti-Somalia ($30 million in exports not accounted for) and for Zaire-Togo ($12 million). More broadly, only six of the thirty-six countries' export data reveal disparities of less than 50 percent in all of the categories analyzed. Benin, the Congo, Djibouti, the Gambia, Mozambique, Niger, Seychelles, Sierra Leone, and Tanzania showed discrepancies of greater than 50 percent in all three export categories. Information on the composition of trade among African countries may be needed for purposes such as formulating plans for regional or international trade agreements. Matched partner trade data could be used to verify infor- mation on goods exchanged down to the three-digit SITC level (most countries Table 5. Discrepancies in Reported Partner Country Statistics on Trade between African Countries, 1982-83 Manufactured exports to all All commodities (millions of dollars) Sub-Saharan Africa (millions of All Sub-Saharan Africa Largest African import market' dollars) Difference Different Difference Exporting country Exports Imports (percent) Partner Exports Imports (percent) Exports Imports (percent) Benin 29.2 1.1 -96.2 Nigeria 19.0 0.0 -100.0 20.6 0.5 -97.6 Burkina Faso 30.8 12.2 -60.2 C6te d'lvoire 16.9 2.0 -88.2 9.1 5.0 -45.1 Burundi 2.2 3.3 52.9 Kenya 0.7 0.8 14.3 0.2 - -100.0 Cameroon 91.6 125.2 36.6 Chad 20.3 40.5 99.5 45.3 58.2 28.5 Central African Rep. 2.0 1.8 -11.0 C6te d'lvoire 1.0 0.5 -50.0 0.8 0.6 -25.0 Chad 21.7 10.8 -50.2 Cameroon 21.1 10.6 -49.8 19.1 9.4 -50.8 Congo 5.5 10.1 85.0 Zaire 1.8 3.8 111.1 2.8 5.6 100.0 C6te d'lvoire 727.9 702.2 -3.5 Somalia 187.4 145.8 -22.3 290.0 225.4 -22.3 Djibouti 38.7 7.2 -81.2 Djibouti 30.2 0.0 -100.0 14.8 1.3 -91.2 Ethiopia 74.7 110.2 47.5 Nigeria 37.8 75.0 98.4 25.9 32.0 23.6 Gabon 55.6 97.4 75.1 Senegal 29.7 16.6 -44.1 9.3 15.4 65.6 Gambia 1.0 12.6 1,138.2 Togo 0.8 0.1 -87.5 0.1 - -100.0 Ghana 19.8 11.3 -43.1 Cameroon 14.6 0.7 -95.2 1.7 1.2 -29.4 Guinea 37.3 24.5 -34.4 Burkina Faso 36.6 18.5 -49.5 0.1 23.4 b Kenya 466.2 523.6 12.3 Uganda 189.1 200.9 6.2 139.5 185.2 32.8 Liberia 18.3 43.6 137.9 Nigeria 7.8 8.9 14.1 2.8 32.1 1,046.6 Madagascar 5.7 2.9 -48.4 Mauritius 2.1 1.6 -23.8 2.1 1.0 -52.4 Malawi 53.1 77.9 46.6 Zimbabwe 29.7 44.3 49.2 12.0 36.2 201.7 Mali 140.1 17.6 -87.4 C6te d'lvoire 91.7 4.9 -94.7 7.7 5.7 -26.0 Mauritania 39.6 37.8 -4.6 Cote d'lvoire 36.9 36.9 0.0 2.2 - -100.0 Mauritius 3.2 1.8 -41.2 Seychelles 2.0 2.7 35.0 1.8 2.4 33.3 (Table continues on the following page.) Table 5 (Continued) Manufactured exports to all All commodities (millions of dollars) Sub-Saharan Africa (millions of All Sub-Saharan Africa Largest African import market' dollars) Difference Different Difference Exporting country Exports Imports (percent) Partner Exports Imports (percent) Exports Imports (percent) Mozambique 56.6 109.4 93.2 Kenya 37.3 18.6 -50.1 12.8 4.4 -65.6 Niger 21.9 162.2 641.5 Burkina Faso 10.7 0.6 -94.4 1.0 11.3 1,030.0 Nigeria 624.7 539.0 -13.7 Ghana 229.2 205.4 -10.4 58.5 15.0 -74.4 Rwanda 8.3 9.3 12.1 Kenya 8.2 9.2 12.1 - - n.a. Senegal 256.1 371.2 45.0 Burkina Faso 104.9 8.2 -92.2 149.0 120.1 -19.4 Seychelles 0.4 0.2 -89.4 Liberia 0.2 0.3 50.0 - 0.1 b Sierra Leone 9.1 3.9 -134.9 Mauritius 2.9 8.5 193.1 0.4 0.7 75.0 a Somalia 2.6 3.9 50.7 Tanzania 2.6 2.8 7.7 2.4 0.9 -62.5 Sudan 3.8 3.1 -17.8 Ethiopia 2.1 0.4 -80.9 1.3 0.6 -53.8 Tanzania 47.6 132.0 177.2 Kenya 11.3 3.0 -73.4 21.0 47.3 125.2 Togo 72.2 119.1 65.0 C6te d'lvoire 36.4 38.7 6.3 54.5 80.7 48.1 Uganda 5.2 5.8 11.9 Kenya 4.9 3.8 -22.4 0.4 0.2 -50.0 Zaire 26.1 17.3 -33.7 Togo 12.1 - -100.0 20.2 8.7 -56.9 Zambia 26.6 29.8 12.0 Malawi 10.3 19.8 92.2 6.5 5.8 -10.8 Zimbabwe 64.5 54.1 -16.1 Malawi 42.2 35.8 -15.2 45.7 39.5 -13.6 -Negligible. n.a. Not applicable. Note: The difference between the value of total reported imports and exports divided by the value of exports, times 100. a. C6te d'lvoire is the largest single destination for all other Sub-Saharan African exports. b. Because no (or very few) exports were reported, the percentage difference between reported imports and exports could not be computed. Source: Author's calculations, based on U.N. Series D Commodity Trade Tapes. Yeats 149 do not report data at lower levels of aggregation). Reported values of intra- African trade for the leading fifty manufactured and agricultural exports are shown in tables 6 and 7. The rank of each product in the value of total intra- African trade (imports and exports) is also shown. Although in both product groups there is a positive rank correlation for export and import statistics that is significant at the 90 percent confidence level, variations for specific products could create important biases in analytical studies. As an illustration, differences of over 18 million dollars exist in the trade of woven textile fabrics (the difference is about 160 percent of reported imports); this item ranks fifth in manufactured exports but only twenty-second in imports. Footwear, road motor vehicles and parts, iron and steel bars, and ships and boats are other manufactured goods for which major differences occur in the rank and value of the partner country data (table 6). In the Table 6. Commodity Composition of Reported Trade in Manufactures among African Countries, 1982-83 Reported value (thousands of Rank in value of dollars) Difference reported tradea SITC Description Exports Imports (percent) Imports Exports 661 Lime and cement 135,736 162,832 20.0 1 1 652 Woven cotton fabrics 113,776 143,432 26.1 2 2 554 Soaps and cleaning preparations 45,338 40,883 -9.8 13 3 851 Footwear 32,243 19,154 -40.6 17 4 653 Woven textile fabrics 29,594 11,289 -61.8 22 5 711 Power-generating machinery 27,735 31,170 12.4 6 6 684 Aluminum 27,544 32,903 19.5 3 7 541 Medicinal products 27,389 19,117 30.2 14 8 561 Fertilizers, manufactured 27,003 30,324 12.3 5 9 599 Chemicals 27,001 28,602 5.9 9 10 631 Plywood and veneers 26,704 30,866 15.6 8 11 642 Articles of paper 24,681 24,088 -2.4 10 12 732 Road motor vehicles and parts 23,777 32,425 36.4 4 13 673 Iron and steel bars 22,502 14,487 -35.6 24 14 893 Articles of plastic 18,481 15,502 -16.1 20 15 718 Machines for special industries 17,667 17,612 0.3 15 16 735 Ships and boats 16,431 29,188 77.6 7 17 719 Machinery and appliances 16,078 22,841 42.1 11 18 729 Other electrical machinery 16,037 17,391 8.4 16 19 651 Textile yarn and thread 15,568 18,215 17.0 23 20 629 Articles of rubber 13,785 9,849 -28.5 28 21 581 Plastics and resins 12,192 13,928 14.2 18 22 733 Road vehicles other than motor vehicles 10,776 8,288 -23.1 25 23 691 Finished structural parts 8,938 7,195 -19.5 27 24 678 Iron and steel tubes and pipes 8,128 6,746 -17.0 29 25 a. Rank by value in 103 three-digit SITC manufactured products (excluding U.N. special codes for which no data are available). Source: Author's calculations, based on U.N. Series D Commodity Trade Tapes. 150 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 agricultural products group, a difference of more than 100 million dollars occurs for manufactured tobacco-it is first in imports but twelfth in exports. There are also major differences in the ranks and reported values for regener- ated fibers, textile waste, and unmilled maize. For many of these products, additional information on the underlying bilat- eral trade flows indicates that transshipment through neighboring countries is a source of major data errors. Whereas the country of final destination may simply list the routing country as the original exporter, the routing country may not list the products transported across its borders either as imports (when entering) or as exports (when exiting). As an example, in 1983, Sudan reports 69 million dollars in tobacco imports from Tanzania, although Tanzania re- ported no exports to Sudan. Because Tanzania only reported 58 million dollars in tobacco imports from all sources, it is likely that Sudan listed Tanzania as the origin of shipments that were transported through Tanzania but originated elsewhere. Similarly, Mali reported 1982 cotton exports to Cote d'lvoire, which Table 7. Commodity Composition of Reported Trade in Food and Agricultural Raw Materials among African Countries, 1982-83 Reported value (thousands of Rank in value of dollars) Difference reported tradea SITC Description Exports Imports (percent) Imports Exports 044 Maize, unmilled 116,271 23,705 -79.6 15 1 001 Live animals 103,320 132,305 28.1 2 2 031 Fresh fish 98,533 116,293 18.0 3 3 071 Coffee 64,755 69,597 7.5 4 4 263 Cotton 54,453 26,382 -51.5 11 5 074 Tea and mat6 47,252 29,087 -38.4 9 6 061 Sugar and honey 43,595 33,809 -22.4 7 7 121 Tobacco, unmanufactured 39,834 30,885 -22.5 8 8 422 Other vegetable oils 33,091 62,530 89.0 5 9 051 Fresh fruit and nuts 31,835 25,004 -21.5 13 10 122 Tobacco manufactures 30,135 138,267 358.8 1 11 292 Crude vegetable material 29,672 34,537 16.4 6 12 099 Food preparations 26,067 25,873 -0.7 12 13 054 Fresh or frozen vegetables 22,437 27,878 24.2 10 14 048 Cereal preparations 21,936 22,195 1.2 16 15 062 Sugar confectionery 16,977 9,241 -45.6 20 16 267 Textile waste 12,448 2,043 -83.6 40 17 243 Shaped wood 10,653 24,987 134.6 14 18 221 Oilseeds and nuts 10,021 4,092 -59.2 28 19 045 Cereals, unmilled 9,831 3,426 -65.2 31 20 112 Alcoholic beverages 8,753 7,208 -17.7 23 21 421 Fixed vegetable oils, soft 8,366 19,348 131.3 17 22 042 Rice 8,273 11,604 40.3 18 23 266 Synthetic and regenerated fibers 5,823 232 -96.0 53 24 242 Wood in the rough 3,980 9,742 144.8 19 25 a. Rank by value in 53 three-digit SITC food and agricultural raw material products (excluding U.N. special codes, for which no data are available). Source: Author's calculations, based on U.N. Series D Commodity Trade Tapes. Yeats 151 reported no imports from Mali. Again, the available evidence suggests that Mali's cotton was transshipped through Cote d'Ivoire to other destinations. Numerous improperly reported transshipments appear to be causing large er- rors in the U.N. statistics. Table 8. Trends in Intra-African Trade: Reported Partner Country Export and Import Statistics All Sub-Sabaran Percentage change in country exports and reported Africa, 1983 imports from the country (thousands of dollars) 1979-83 1980-83 1981-83 Country All exports All imports Exports Imports Exports Imports Exports Imports Benin 3,701 514 -54.1 -96.1 -55.1 -88.5 -54.3 -94.6 Burkina Faso 13,303 4,837 -64.4 -11.2 -64.8 -87.3 -52.4 -63.4 Burundi 396 2,863 -74.2 -16.5 -40.0 561.2 -76.9 296.5 Cameroon 5,007 58,951 -92.4 76.2 -84.9 32.0 -80.9 -15.1 Central African Rep. 571 174 -56.3 -84.5 -71.3 -89.5 -65.4 -81.0 Chad 10,695 238 62.7 -96.4 -2.9 -98.7 -2.5 31.9 Congo 3,994 2,815 3.2 -65.5 -25.3 -60.3 47.8 -51.3 Cote d'lvoire 351,298 332,761 49.2 -87.9 116.3 16.8 3.9 5.7 Djibouti 18,118 1,542 742.7 -90.2 -6.6 -92.6 0.2 -91.4 Ethiopia 35,144 45,991 10.9 2.7 -35.9 -33.4 56.0 -16.0 Gabon 24,322 53,650 1,386.7 8.8 32,767.6 69.7 -51.5 1.3 Gambia 400 92 426.3 -62.9 30.3 -77.0 -33.2 -99.1 Ghana 7,955 2,760 18.2 -79.9 133.3 -79.8 -26.5 -71.3 Guinea 18,665 1,161 46.8 -90.9 6.9 -94.2 0.0 209.6 Kenya 237,772 279,853 18.0 277.2 -22.5 -18.3 -22.2 -18.9 Liberia 8,679 32,512 -29.9 406.3 -22.4 125.5 -18.6 102.0 Madagascar 3,279 2,012 -39.5 -52.4 80.4 -18.0 -22.5 25.1 Malawi 51,792 39,190 193.7 88.8 43.3 8.3 30.8 -8.8 Mali 9,278 8,148 -63.9 -46.0 -48.7 -70.7 35.7 -11.6 Mauritania 20,881 20,280 5,120.3 2,419.3 1,858.8 2,913.3 382.0 412.0 Mauritius 1,462 2,332 -0.8 42.6 -26.5 -48.9 38.7 64.7 Mozambique 27,219 78,133 156.3 458.6 -59.5 16.8 62.5 480.5 Niger 9,094 81,439 -77.0 1,975.9 -87.4 17.1 -88.8 -1.3 Nigeria 200,434 190,111 -28.9 -38.9 -48.2 -63.6 -29.7 -62.5 Rwanda 4,153 5,163 83,060.0 -41.3 -41.5 -27.3 -49.7 -37.4 Senegal 180,514 174,769 97.4 68.4 40.5 34.3 1.2 -10.4 Seychelles 50 155 -86.8 -54.1 -89.5 -82.9 -69.3 -84.7 Sierra Leone 1,665 6,400 -34.2 122.1 -11.7 378.3 -61.1 55.3 Somalia 904 1,238 59.4 177.6 -39.2 -24.4 -44.3 -10.7 Sudan 521 1,163 -11.8 32.3 -64.1 45.9 -74.2 9.7 Tanzania 18,941 70,013 -67.7 315.5 66.4 8.7 -41.8 63.6 Togo 35,939 45,244 31.8 292.9 -59.1 -52.4 9.4 -24.5 Uganda 2,562 3,258 9,388.9 33.7 -74.2 -66.5 9.0 38.1 Zaire 8,508 13,195 21.0 28.3 -46.6 -9.9 -57.0 25.2 Zambia 11,703 15,737 -62.1 -57.4 25.4 -57.7 147.8 -44.8 Zimbabwe 34,057 26,273 344.8 330.8 -35.8 34.2 16.6 -12.6 n.a. Not applicable. Source: Author's calculations, based on U.N. Series D Commodity Trade Tapes. 152 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 Product supply and demand and broader trade policy analyses also require correct identification of trends or changes in the level of trade. Measuring long- term trends is not easily accomplished because of the major gaps in the histor- ical records (see table 1), but data for all Sub-Saharan African countries were available for the 1979-83 period. Table 8 shows intra-African export and import totals for 1983, their percentage change, and the proportion of coun- tries for which partners' trade data show opposing changes in the direction of intra-African trade. This would occur, for example, if a country's reported exports rose while the matched reported imports of its trading partners de- clined. The information in table 8 strongly suggests that these data are unreliable as an indicator of changes in the level of trade. The data showed conflicting changes in direction in more than half of the thirty-six countries for 1981-83 and more than 40 percent of the countries over the 1979-83 period. In addi- tion, even when the data showed the same direction of change in trade, there were often major differences in the magnitude. As an indication, African im- porters of Togo's exports reported increases nine times larger than those shown in Togo's export data. Correlations between changes in partner countries' ex- port and import statistics were not statistically significant for any of the three time periods reported in table 8. It was not possible to determine if some of these data discrepancies arose because of attempts by other Sub-Saharan Africa countries to conceal trade with South Africa. From 1979 to 1982 the South African Cust:oms Union reported no exports to any of the Sub-Saharan countries, although twenty-two out of the thirty-six countries reported imports from South Africa that totaled about $900 million. In 1983, the U.N. trade tapes show $116 million in South African exports to Malawi, $9 million to Seychelles, $3 million to Kenya, and minor exports (under $1 million) to six other African countries. Coal and petroleum accounted for about one-third of these exports. South Africa re- ported little trade in those products in which major differences exist in Sub- Saharan partner country data (see tables 6 and 7). III. THE ANALYTICAL AND POLICY IMPLICATIONS The key question that emerges from this study concerns the utility of African trade data for research and policy studies on intra-African trade. Five general findings bear directly on this question. First, the data cannot be used to assess the overall level of trade among African countries; the average discrepancy between matched export and import values is more than 60 percent for thirty- five of the countries (and more than 100 percent when the Gambia is included). Second, the data are probably useless for assessing the direction of intra-African trade because countries listed by the exporters as the largest markets for exports often fail to report any corresponding imports. Third, the data appear to be Yeats 153 equally deficient for determining the composition of trade because major dis- crepancies are revealed between partner country statistics at greater levels of detail. Fourth, there are large and persistent differences in the trends in both the magnitude and direction of intra-African trade as reflected in reported exports and matched imports. Fifth, the fact that reported f.o.b. exports fre- quently exceed matched reported c.i.f. imports suggests that smuggling is wide- spread in trade among African countries or that importers are intentionally underinvoicing to avoid high tariffs or quotas. Given these points, it is difficult to see how any confidence could be placed in the official U.N. data or the underlying national data upon which they are based. Similar conclusions about the reliability and utility of African trade data emerge from a comparison of these statistics with matched OECD data. Discrep- ancies between the African export and OECD import data are far greater than differences in statistics on trade among developed countries. Analyses of under- lying quantity and unit values indicate several factors are responsible. First, discrepancies in reported quantities traded of products such as petroleum, coffee, and cocoa suggest that exporters have intentionally been underreporting shipments in order to circumvent international commodity agreement quotas. Second, for high-value, low-volume products like pearls and precious stones, reported imports greatly exceed reported exports, suggesting that smuggling is occurring on a large scale. Third, large differences in the reported unit values for some products, particularly oilseeds and iron ore, suggest that exporters are purposefully underinvoicing (possibly to avoid government foreign exchange controls or restrictions on foreign asset holdings), or are not receiving full value for these items. Because most of the data errors appear to originate on the export side, this study suggests that any North-South analyses of African trade should primarily rely on OECD data. Because export subsidies and similar incentives are not widely used in the subject countries, the excess of reported exports over imports is consistent with underinvoicing by importers or smuggling on a fairly massive scale. The very high import tariffs in most African countries provide a strong incentive for such activities. Without further analysis it would be difficult to estimate the magni- tude of smuggling in African trade from data drawn from partner countries because there is no way to determine quantities and values that are not reported by either the exporter or importer as opposed to (smuggled) trade that is recorded by one of the countries involved. On a more general level, the results of this study accent the need for more information about the basic quality of official U. N. trade statistics. It would appear useful, for example, to extend the general approach employed in this analysis to other groups of developed and developing countries and make the findings of such investigations generally available. Until this is done, a strong possibility exists that basic research could be seriously biased, and inappropri- ate policy decisions made because of substantial errors in official trade data. 154 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 APPENDIX. THE TRANSPORT COSTS CORRECTION FACTOR One factor which produces a discrepancy between partner country trade data is international transport and insurance costs. Exporters typically value ship- ments on an f.o.b. basis, whereas imports are normally tabulated on a c.i.f. basis. As such, a key question is how large a difference should be expected between partner country statistics as a result of freight and insurance costs. Because the United States now tabulates international transport costs actually paid for all imports directly from customs vouchers, these data can be used to derive transport correction factors for partner country trade data with the United States (Finger and Yeats 1976; Brodsky and Sampson 1979; Yeats 1981, 1990). The U.S. Customs information reflects freight charges actually paid, including all discounts and surcharges. (Studies have often had to rely on published liner conference charges. The conferences, organized by unincorpor- ated associations of ocean liner owners, formally establish freight rates, sailing schedules, and regulations that affect competition among members. These are often unreliable as guides to actual transport payments.) Transport and insurance costs for each African country's exports were com- piled and converted to nominal equivalents. The formula used for estimating these nominal freight costs, which serve as "correction factors" for partner country data (Ci,), was: (A-1) CUJ= VfU where f represents freight and insurance charges and VfU the U.S. f.o.b. value of total imports from each African country. The resulting statistics show the percentage by which c.i.f. imports should exceed partner country f.o.b. data (see appendix table 1). For total trade the data suggest that most African countries' f.o.b.-c.i.f. correction factor would be in the range of S to 10 percent, although there are several exceptions. Guinea, whose exports are highly concentrated in low-grade metallic ores, has a nominal freight rate that ranges from 21 to 37 percent; the corresponding freight factors for Liberia and Somalia reach 25 and 28 percent. Although the data reported in appendix table 1 are based solely on U. S. statistics, they should be useful for assessing African-European partner country trade figures. Specifically, numerous studies of maritime transport costs show that freight rates are generally positively associated with distance, and for commodities like wood, ores, or petroleum which involve economies of scale in transport, freight rates are inversely related to volumes shipped. Because distance is smaller to European ports, and quantities transported are generally larger, the f.o.b.-c.i.f. correction factors reflected in appendix table 1 would probably serve as upper limits for those that should be applied to most Afri- can-European partner country trade data. Equation A-1 will likely understate the correction factor for landlocked African countries, however, because the Yeats 155 Appendix Table 1. Import Values and Related Nominal Transport Costs for African Countries' Total Exports to the United States 1983 U.S. import value (millions of dollars) Nominal freight costs (percent) Exporting country f.o.b. c.i.f. 1982 1983 1984 1985 1986 1987 Benin 26.9 28.7 7.1 6.7 n.a. n.a. n.a. 5.3 Burkina Faso 0.1 0.1 n.a. n.a. n.a. n.a. n.a. n.a. Burundi 2.8 3.0 7.5 7.1 5.3 9.0 6.0 6.4 Cameroon 515.0 535.9 4.1 4.1 4.1 4.5 9.3 7.0 Central African Rep. 3.5 3.6 4.1 2.9 n.a. n.a. 8.1 4.2 Chad 67.6 70.6 n.a. 4.4 n.a. 33.3 n.a. n.a. Congo 820.8 859.4 3.6 4.7 5.4 5.9 9.3 6.1 C6te d'Ivoire 342.7 371.3 7.4 8.3 6.4 6.3 6.7 8.5 Djibouti n.a. n.a. n.a. n.a. n.a. n.a. 50.0 n.a. Ethiopia 86.8 93.9 7.4 8.2 7.9 7.7 4.2 6.0 Gabon 657.1 685.1 4.1 4.3 4.0 4.3 8.9 6.0 Gambia 0.2 0.2 n.a. n.a. 16.6 33.3 20.0 n.a. Ghana 119.8 125.3 2.3 4.6 7.8 6.0 5.2 4.3 Guinea 104.4 138.6 36.7 32.7 25.7 21.4 25.8 27.7 Kenya 65.0 70.2 7.4 8.0 8.0 7.8 6.5 7.4 Liberia 90.5 107.5 25.2 18.8 20.1 2S.0 19.3 14.5 Madagascar 70.7 74.2 5.6 5.0 4.1 4.2 2.9 3.4 Malawi 14.5 15.5 9.8 6.9 8.2 9.0 10.6 8.0 Mali 0.7 0.7 9.1 n.a. 9.1 8.5 9.6 4.3 Mauritania 0.8 0.8 n.a. n.a. 10.0 25.0 10.5 12.4 Mauritius 31.5 33.9 10.2 7.6 8.1 9.3 9.6 10.2 Mozambique 28.5 31.0 10.8 8.8 9.2 8.8 6.4 6.0 Niger 4.2 4.3 n.a. 2.4 n.a. 8.8 5.1 5.6 Nigeria 3,736.0 3,882.7 3.2 3.9 3.9 3.5 5.9 5.4 Rwanda 28.4 29.9 6.1 5.3 4.7 6.3 5.6 6.4 Senegal 1.9 2.1 18.2 10.S 20.8 12.S 7.S S.7 Seychelles 2.9 3.1 4.9 5.2 5.6 4.2 3.6 5.3 Sierra Leone 21.5 22.8 6.2 6.1 5.1 6.8 7.5 4.4 Somalia 0.1 0.2 22.2 n.a. 28.5 22.2 n.a. n.a. Sudan 19.0 20.4 10.2 7.4 11.5 8.1 3.6 3.5 Tanzania 14.3 16.2 9.7 13.3 10.4 11.5 8.5 5.9 Togo 19.9 21.0 6.9 5.5 5.1 7.0 12.4 13.1 Uganda 103.9 110.6 7.2 6.4 5.4 5.7 4.9 5.8 Zaire 366.3 378.2 4.0 3.2 3.2 3.5 4.1 8.0 Zambia 52.1 53.7 3.7 3.1 3.6 2.1 2.4 3.2 Zimbabwe 73.8 79.5 7.1 7.7 6.6 6.3 5.8 6.5 n.a. Not applicable (no exports or imports recorded in that year). Note: The figures in this table are used to calculate correction factors for exporter (f.o.b.) and importer (c.i.f.) trade values. Source: Author's calculations, based on U.S. reported trade data. U.S. transport data do not account for shipping costs from the border of the landlocked country to the ocean port of export. The ratios could also be higher for such countries as Guinea, Liberia, and Sudan. 156 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 REFERENCES Allen, R. G. D., and J. Edward Ely. 1953. International Trade Statistics. New York: Wiley. Bhagwati, Jagdish. 1964. "On the Underinvoicing of Imports." Oxford Bulletin of Economics and Statistics 26, no. 4 (November): 389-97. . 1967. "Fiscal Policies, the Faking of Foreign Trade Declarations, and the Balance of Payments." Oxford Bulletin of Economics and Statistics (February). Brodsky, David, and Gary Sampson. 1979. "International Transport and Latin Ameri- can Exports to the U.S." International Journal of Transport Economics 7, no. 4 (December): 279-92. Ely, J. E. 1961. "Variations between U.S. and Its Trading Partner Import and Export Statistics." American Statistician 27, no. 2 (April): 51-72. Finger, J. Michael, and Alexander Yeats. 1976. "Effective Protection by Transportation Costs and Tariffs: A Comparison of Magnitudes." Quarterly Journal of Economics 90, no. 1 (February): 169-76. Gulati, Sunil. 1987. "A Note on Trade Misinvoicing." In Donald Lessard and John Williamson, eds., Capital Flight and Third World Debt. Washington, D.C.: Institute for International Economics. Livingstone, Ian. 1986. International Transport Costs and Industrial Development in the Least Developed Countries. U.N. Industrial Development Organization Report IS.616. Vienna. Morgenstern, Oskar. 1963. On the Accuracy of Economic Observations. Princeton, N.J.: Princeton University Press. OECD (Organisation for Economic Co-operation and Development). 1985. Discrepan- cies between Imports and Exports in OECD Foreign Trade Statistics. Paris: OECD Department of Economics and Statistics. Sheikh, Munir. 1974. "Underinvoicing of Imports in Pakistan." Oxford Bulletin of Economics and Statistics 36, no. 4 (November): 287-95. United Nations Economic and Social Council. 1974. International Trade Reconciliation Study. (E/CN.3/454) Geneva: United Nations. Wulf, Luc. 1981. "Customs Valuation and the Faking of Invoices." Economia Interna- zionale 34, no. 1 (February): 3-23. Yeats, Alexander. 1978. "On the Accuracy of Partner Country Trade Statistics." Oxford Bulletin of Economics and Statistics 40, no. 1 (November): 341-61. . 1981. Trade and Economic Policies: Leading Issues for the 1980s. London: Macmillan. - . 1990. "Do African Countries Pay More for Imports? Yes." World Bank Eco- nomic Review 4, no. 1 (January): 1-20. THE WORLD BANK ECONOMIC REVIEW, VOL . 4, NO. 2: 1 57-173 The Impact of the International Coffee Agreement on Producing Countries Takamasa Akiyama and Panayotis N. Varangis Simulations of a global coffee model incorporating a vintage capital approach to production are run. Over the recent period of operation of the International Coffee Agreement's export quota system, the authors find that the quota system had a stabilizing effect on world coffee prices. The quotas reduced real export revenues for most small exporting countries, but large producers gained. Most small countries gained, however, in terms of risk reduction. If a brief suspension of the quota occurs from time to time, caused, for example, by adverse weather which results in a shortfall in world supply, the quota system works like a buffer stock scheme; on average, producing countries as a whole lose transfer benefits but gain risk benefits. The International Coffee Agreement (ICA), which utilizes an export quota sys- tem, has had an important influence on the world coffee market in recent years. The export quota scheme succeeded in stabilizing world coffee prices in its most recent period of operation (October 1980-June 1989) in spite of wide fluctuations in world coffee production. Because of disagreements among mem- bers over economic clauses that were introduced into the ICA in October 1989, however, the quota system was suspended in July 1989. World coffee prices fell by more than 40 percent following the suspension of the quotas, which led to large declines in producers' incomes and in export and government revenues in most coffee-exporting countries. In spite of continuing negotiations, current prospects for the reintroduction of the quota system are bleak. The main objective of this article is to analyze the impact of the ICA export quota system on the world coffee market, focusing on increases in real export revenues (transfer benefits) and reductions in income variability (risk benefits) in each exporting country. In pursuit of this objective, we make extensive use of a new global model of the coffee economy. We begin with a brief description of recent developments in the world coffee market in section I, followed by a review of previous studies of the world coffee market in section II. In section III, a description of the model and its validation The authors are economists in the International Economics Department of the World Bank. They would like to thank Ron Duncan for his valuable contributions to this article and the anonymous referees for their comments on an earlier draft. ©1990 The International Bank for Reconstruction and Development / THE WORLD BANK. 157 158 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 are given. Ex-post simulation results from the model are presented in section IV, and section V gives our conclusions. 1. RECENT DEVELOPMENTS IN THE WORLD COFFEE MARKET AND THE ICA The first International Coffee Agreement came into force in October 1963. It aimed to halt a declining price trend and stabilize prices above their free market level. That exporters supported the ICA should not be surprising, but most importers are also signatories to the ICA. Their compliance with ICA provisions designed to increase prices is more difficult to explain, since the support of prices above the free market level is a direct cost to importers. Fisher (1972), Krasner (1973), and Gordon-Ashworth (1984) argue that the consum- ing countries' participation in the ICA can be explained only by political mo- tives, although a number of roasters in the United States and Western Europe said that they preferred stable prices even if at somewhat higher average levels. The United States, recognizing the strategic importance of Latin America, considered it necessary to raise and stabilize world coffee prices to promote political and economic stability in the region. The European Community had similar objectives with regard to Africa. Whatever their motivation, importers' membership in the ICA is strictly voluntary; there are no mechanisms to keep consuming countries in the agreement or to punish those who leave. In the early 1980s New Zealand and Israel did, in fact, withdraw from the ICA. The ICA'S main market regulatory instrument was an export quota system. The quota system was abandoned in 1973, however, because producing and consuming countries could not agree on the level of the support price and the level and allocation of quotas. World coffee prices were at historically low levels in the early 1970s, but when a serious frost hit Brazil in 1975, prices as measured by the ICA "Other Milds" indicator price skyrocketed to a peak of 317.68 U.S. cents per pound in April 1977. [This indicator price is the weighted average of major Central American fully washed arabica coffee ex-dock in New York (75 percent) and Bremen-Hamburg (25 percent)]. After 1977 prices de- clined sharply, prodding producing and consuming countries to negotiate a new agreement, which again contained an export quota system as its main economic provision. The mechanics of this latest ICA export quota system are described in detail in Gilbert (1987); its key features were as follows. The global export quota, which was the total quantity sold by exporting members (covering more than 98 percent of world exports) to importing members (covering 85-90 percent of world imports), was adjusted to keep world coffee prices within an agreed- upon range. World coffee prices were proxied by the average of arabica and robusta prices, ex-dock in New York and Bremen-Hamburg and ex-dock in New York and Le Havre-Marseilles respectively. The target price range for the period October 1980-June 1989 was 120-140 U.S. cents per pound. In gen- Akiyama and Varangis 159 eral, when prices rose above this range for a sustained period, quotas were increased; when prices fell below the target range, quotas were reduced. Initial quotas for each exporting member were based on past export volumes and were supposed to be adjusted periodically to reflect production capabilities. In practice, few such adjustments were made. To enforce the quotas, the Interna- tional Coffee Organization issued "export stamps" on a quarterly basis to each exporting member. Importing members agreed to import only coffee covered by these stamps. Exporting members were essentially free to sell any quantity of coffee to nonmember importing countries. This nonquota market consisted of New Zealand, the U.S.S.R., and other centrally planned economies of Eastern Europe except Yugoslavia, and all developing countries except Greece and Portugal. As Bohman and Jarvis (1989) point out, exporting members whose average production was considerably larger than the sum of export quotas and domestic demand had an incentive to export to the nonquota market even at large discounts. Although reliable data on prices in this market are unavailable, they were reported to be at a 30-50 percent discount relative to quota market prices. Exports to the nonquota market would have depressed prices in the quota market to the extent that this coffee was re-exported to the quota market. No reliable estimates exist on the size of this "tourist coffee" trade, but it was considered to be relatively small-at most 3-4 percent of sales to the quota market. The new export quota system became effective in 1980 and was successful in stabilizing prices from October 1980 until February 1986, when coffee prices sharply increased, triggering suspension of the quotas. The price increase was caused by a sizable reduction in Brazil's 1986-87 crop as a consequence of a severe drought in 1985. Prices declined steadily after the spring of 1986, leading to prolonged discussions among ICA members and the eventual rein- statement of the quota system in October 1987. Because the ICA was set to expire in September 1989, intensive negotiations were carried out among the ICA members in 1988-89 concerning the economic clauses of the new agreement. Two key issues went unresolved, resulting in the suspension of the economic (quota) clauses of the ICA in July 1989. The issues were, first, the allocation of quotas, especially among mild arabica-, unwashed arabica-, and robusta-producing countries, in the face of shifting world demand in favor of mild arabicas, and second, the large discount sales made by export- ing members to nonmember importing countries. The ICA without economic clauses will continue until September 1991, providing a forum for discussion and negotiation of the economic clauses. Reintroduction of the quota system hinges on resolution of the quota allo- cation and discount sales issues. Recent discussions demonstrate the sensitivity of quota allocation as it affects each exporter's financial gain from membership in the ICA. Several exporting members have said that they will withdraw from the ICA if their quotas are reduced. The issue of discounted sales to nonmem- 160 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 bers is discussed by Bohman and Jarvis (1989), who claim that it will be difficult to resolve because many exporting countries derive significant benefits from these nonquota exports. Given these obstacles, prospects for the reintrod- uction of the quota system in the near future are poor. II. PREVIOus ANALYSES OF THE WORLD COFFEE MARKET Despite the existence of a number of models of the world coffee market, the long-term effects of the quotas on export revenues in individual exporting countries and on coffee prices have not been analyzed. Models of the coffee market either do not include the export quota system or, if they (1o, focus on the short-term effects of the export quota and/or treat the coffee producers and consumers as broad aggregates. Ford (1978) built a world coffee model to evaluate the effect of different stabilization policies on the world coffee market. These policies included coffee tax variations, changes in the Brazilian diversifi- cation program, and buffer stock schemes. The majority of Ford's analysis concentrated on the size, cost, and effects of buffer stocks. Another set of models of the coffee market concentrates on the political aspects of the ICA. Such models attempt to explain the distribution of quotas within the agreements (Lien and Bates 1987) and the participation of govern- ments in the making and enforcement of international trade agreements (Bates and Contreras 1988). Several attempts have also been made to evaluate the welfare effects of the nonquota market sales of coffee. Bohman and Jarvis (1989) calculate the welfare effects on major exporting countries from partici- pation in the nonquota market, whereas Herrmann (1986) estimates the wel- fare gains of nonmember importing countries resulting from the quota policy. Both analyses are based on aggregated, short-run models. Palm and Vogelvang (1988) examine the effects of policies designed to reduce production. Policies are analyzed in a scenario without export quotas and in a scenario in which export quotas are introduced when the spot market price drops below a trigger level. As the authors point out, they use a short-run model in which production is predetermined. Finally, Akiyama and Duncan (1982) and de Vries (1975) do not model the export quota system. Even when other commodities are considered, models of the impact of export quota schemes are available only at theoretical levels (Maizels 1982) or are targeted to individual countries (Dick and others 1982). III. DESCRIPTION OF THE MODEL AND ITS VALIDATION The new model used in the study consists of a large number of equations estimated econometrically on annual data. Data were mostly obtained from the International Coffee Organization, the U.S. Department of Agriculture, and the International Monetary Fund (various years). Detailed data for new plant- ings, age of trees, and yields for a number of producing countries were obtained Akiyama and Varangis 161 from informal unpublished country-specific sources. Export supply is modeled for thirty-one countries or regions, and import demand is modeled for twenty- one ICA member importing countries and two nonquota markets. Demand, Supply, and Price Determination Price. The need to explicitly model the effects of the quota system on world prices precludes the use of a price equation linking prices to stocks. This is evident in the case in which world prices are increased through reduction of the global quota. As the quota is reduced, prices rise but stocks held in produc- ing countries (which usually account for the majority of world stocks) also rise. In this case the correlation between prices and world stocks would be positive; this is the opposite of the relationship assumed in price equations with stocks as an explanatory variable.' We concluded that the only satisfactory way to determine the world price when a quota scheme is operating is to equate import demand and export supply in the quota market. When the quota system is not operating, price will be determined by equating world import demand and world export supply. Demand. For each importing country, demand is specified in a conventional manner, that is, on a per capita basis with real income per capita, population, taste, and real retail prices as explanatory variables. Real retail prices, in turn, are a function of exchange rates, inflation, and world prices. A time trend is used as a proxy for changes in tastes. The twenty-one countries modeled as ICA importing countries are Australia, Austria, Belgium, Canada, Denmark, Fin- land, France, Federal Republic of Germany, Greece, Ireland, Italy, Japan, Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, United Kingdom, United States, and Yugoslavia. In addition, two nonmember market regions are modeled: noncoffee producing developing countries and Eastern Europe (including the U.S.S.R.). Production and exports. Specification of production is based on a version of the vintage capital model (Akiyama and Trivedi 1987) in which supply is determined in two stages-at planting and at harvesting. Planting decisions in the case of a perennial crop like coffee are investment decisions which will affect production capacity in current and future years. Harvesting decisions are short-term and will depend in part on current and past planting decisions and the producer prices prevailing just before and/or during the harvest period. The advantage of this specification is that it allows us to empirically distinguish between short-run and long-run supply responses to the quota system. New plantings (the long-run supply decision) are determined by recent real farmgate prices, which in turn are a function of exchange rates, inflation, and 1. As discussed by Labys (1973) and Ghosh and others (1987), in many commodity models price equations are inverted stock demand equations, and hence the stock variable should have a negative coefficient. 162 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 Figure 1. Price Determination in the Model D_ D~~~~D Price p ___,/ Q Export supply and import demand the world price of coffee. The total number of trees, represented by new and past plantings, together with their yield determines production capacity. Actual production (the short-run supply decision) is then a function of production capacity, real farmgate prices, and other variables such as weather and the biennial production cycle. The amount of output available for export is defined as the sum of produc- tion and carryover stocks net of domestic consumption. This output will be allocated among total exports-the sum of exports to the quota market and exports to the nonquota market-and additions to stocks. When quotas are in force, we assume that demand in the quota market is sufficient to ensure that exporters are able to sell their entire allotment in this market, so that exports to the quota market are exogenous and equal to the quota. Exports to the nonquota market will depend on the world price of coffee. Any residual output which is not exported to either the quota or nonquota markets is allocated to year-end stocks. Price in the quota market, PQ in figure 1, is determined where the amount of the world quota, Q, equals the import demand of member countries, DmDm. Note that we are assuming no interaction (re-exports) be- tween the quota and nonquota markets. This assumption is reasonable if the amount of "tourist coffee" is small, as reported. When the quota system is not in force, total exports are a function of output available for export, world price, exchange rates, and inflation. E]xports to Akiyama and Varangis 163 nonmember countries are modeled as before, and exports to member countries are calculated as the difference between total exports minus exports to non- members. In figure 1, export supply, SS., is an increasing function of price, and world import demand is the sum of quota market and nonquota market demand, D.DW. Price is thus determined at P when quotas are not operative. Because export supply is a function of world price here, existing stocks act to stabilize prices to some extent, even when quotas are not in force. Estimated Elasticities in the Model A large number of parameters and elasticities was estimated and used in the model. Those for supply and import demand are discussed here. Supply is modeled for thirty-one countries or regions, but the degree of detail in the supply specification for different countries varies according to the avail- ability and reliability of data. We distinguish four categories of supply specifi- cation in the model: (a) For Colombia, equations describing new plantings, stumping, production capacity of old trees, and production were estimated. More detailed analysis was possible for Colombia because of the availability of reliable data from the National Coffee Federation on new plantings, stumping, and stock of old trees. (b) For sixteen countries (Brazil, Costa Rica, Cote d'Ivoire, Dominican Republic, Ecuador, El Salvador, Guatemala, Honduras, India, Indonesia, Kenya, Madagascar, Mexico, Papua New Guinea, Philip- pines, and Rwanda), equations describing net new plantings and production were estimated. (c) For nine countries and regions (rest of Asia, Burundi, Cameroon, Ethiopia, Nicaragua, Peru, rest of South America, Tanzania, and Zaire) simple supply equations were specified because of lack of success in the estimation of new planting equations or the unavailability of tree stock data. (d) For five countries and regions (rest of Africa, Angola, rest of Central America, Uganda, and Venezuela) production was taken to be exogenous be- cause of the unsatisfactory results obtained from attempts to estimate supply equations. Because these countries play a small role in the world market, however, we do not believe that this reduces the effectiveness of the model. The countries for which the vintage capital approach was used (countries which fall into categories a and b above) account for about 70 percent of world production. Table 1 presents the elasticities of new plantings with respect to output price for the countries in categories a and b. For countries in category c, new plantings were not estimated. For these countries, it was assumed that tree stocks or production capacity change with time, and prices affect production only in the short term. Short-, medium- and long-term supply elasticities are given in table 2 for most of the countries in categories a, b, and c.2 These elasticities were derived from model simulations. As discussed by Akiyama and Trivedi (1987), the price elasticity of supply is not time-invariant but instead increases over time as 2. Prices were not significant in the new plantings and supply equations for Honduras and Mexico. 164 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 Table 1. Price Elasticities of New Plantings Country t-1 t-2 t-3 t-4 Brazil 1.02 2.34 (2.06) (7.72) Colombia 1.68 (3.37) Costa Rica 2.03 (1.93) C6te d'lvoire 4.19 (2.52) Dominican Republic 1.49 0.73 (2.47) (2.18) Ecuador 1.63 (3.22) El Salvador 2.29 (3.47) Guatemala 2.88 (3.18) Honduras 0.49 (3.17) India 2.59 (2.95) Indonesia 0.56 (2.67) Kenya 1.72 1.56 (4.1) (3.82) Madagascar 4.48 (2.93) Mexico 1.23 (3.78) Papua New Guinea 1.20 0.43 (2.36) (1.84) Philippines 2.19 1.30 0.45 (5.50) (4.05) (3.08) Rwanda 2.67 0.98 (3.27) (1.99) Note: t-statistics are in parentheses; only significant coefficients are reported. Column heads t - i= 1, . . . , 4 refer to the elasticity of new plantings with respect to P, - i, that is, price at time t - i. Source: Authors' calculations, based on World Bank data available from the authors on written request. producers adjust their planted acreage with farmgate prices. Estimated elastic- ities tend to be high in countries where general economic and coffee policies have been stable and where data are reliable. The short-term (within the first year) price elasticity of supply for countries in categories a-c taken together is found to be 0.04, and the short-term price elasticity of export supply is found to be equal to 0.06. Behrman (1978) assumed the short-term price elasticity of supply to be zero. Herrmann (1986) found it to be slightly less than the short-term export supply elasticity, which he calculated to be 0.04, whereas Akiyama and Duncan (1982) estimated a Akiyama and Varangis 165 Table 2. Elasticities of Supply in Selected Countries Years after price change Country Two years Five years Ten years Brazil 0.03 0.10 0.36 Burundi 0.03 0.47 0.95 Cameroon 0.04 0.14 0.16 Colombia 0.16 0.44 0.74 Costa Rica 0.11 0.15 0.41 C6te d'lvoire 0.55 0.68 0.84 Dominican Republic 0.19 0.34 0.78 Ecuador 0.11 0.13 0.14 El Salvador 0.13 0.15 0.16 Ethiopia 0.06 0.15 0.19 Guatemala 0.13 0.13 0.20 Honduras 0.13 0.15 0.20 India 0.19 0.10 0.15 Indonesia 0.14 0.17 0.25 Kenya 0.04 0.14 0.45 Mexico 0.02 0.06 0.13 Papua New Guinea 0.07 0.18 0.18 Philippines 0.06 0.18 0.20 Zaire 0.02 0.15 0.17 Source: Authors' calculations, based on World Bank data available from the authors on written request. short-term supply elasticity of 0.12, somewhat higher than in other studies. The results from all of these studies confirm Ford's (1978) perception that coffee supply is very price inelastic in the short run. Average estimated income and price elasticities of demand in importing coun- tries and selected producing countries for 1968-863 are given in table 3.4 The price and income elasticities of world coffee demand for that period are esti- mated to be equal to -0.33 and 0.6 respectively. Behrman (1978) found a price elasticity of demand of -0.2; Herrmann (1986) estimated a value of -0.27; and Akiyama and Duncan (1982) obtained a value of -0.186. The somewhat higher price elasticity here might be the result of the fact that we used deflated retail prices of each individual consuming country while other studies used world prices. The income elasticity of demand is similar to that obtained by de Vries (1975), Akiyama and Duncan (1982), and Herrmann (1986). 3. The "crop year" used in producing countries varies from one country to another in terms of starting date. For example, Colombia's crop year is the same as the international year, which starts October 1, but Brazil's crop year starts July 1. In this paper the production year refers to the ending year unless otherwise specified; for example, Brazil's production for the 1987-88 crop year is referred to as Brazil's 1988 crop. All exports are on the international coffee year basis, thus exports for the period October 1987-September 1988 are referred to here as exports of 1988. 4. Income elasticities for some countries, such as Ireland and Japan, were found to be very high because of the low levels of per capita consumption in these countries in the 1960s and early 1970s. 166 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 Table 3. Demand Elasticities Country Income Price Importing Austria 1.30 --0.54 Australia 1.72 --0.37 Belgium 0.36 --0.28 Canada 0.28 --0.13 Denmark 0.58* - 0.43 Finland 0.34 - 0.08* France 0.68 - 0.13 Germany, Fed. Rep. 0.98 -0.17 Greece 0.52 -0.49 Ireland 2.89 -0.34 Italy 0.92 -0.18 Japan 2.03 -0.31 Netherlands 0.89 -0.34 New Zealand 1.28 -0.13 Norway 0.26 -0.14 Portugal 0.62 -0.28 Spain 1.07 -0.07* Switzerland 0.56* -0.24 Sweden 0.70** -0.29 United Kingdom 1.26 -0.51 United States 0.50** -0.46 Yugoslavia 0.12** -0.15## Nonmember importing Noncoffee-producing developing 0.68 -0.13 Eastern Europe (including U.S.S.R.) - -0.22 Selected producing' Brazil 0.50** -0.09 Colombia 0.41 -0.14 Dominican Republic 0.20** -0.08 Ecuador 0.40 -0.08 India 0.24 0.08** Indonesia 0.18** -0.07 Mexico 0.3S** -0.14 -Not available. Note: *Significant at 10 percent level of significance. ' Significant at 25 percent level of significance or below. All others without asterisks significant at 5 percent level of significance. a. Because of the unavailability of retail price data for these countries, the international coffee price in terms of local currencies and deflated by the local consumer price index was used. Therefore, the price elasticities presented here should generally underestimate the demand response to changes in retail prices. Source: Authors' calculations, based on World Bank data available from the authors on written request. Akiyama and Varangis 167 Table 4. Descriptive Statistics for the Ex-post Simulation Run ICA other milds World Total exports to Statistic indicator price production the quota market Mean percent absolute error 5.4 1.1 0.9 Root mean squared percentage error 4.3 1.4 1.4 IV. EX-POST SIMULATION RESULTS WITH AND WITHOUT THE ICA The model was run for the period 1974-86 and the results compared with the actual values for prices, output, and exports. Some statistics from this comparison are given in table 4, and indicate that the simulation is fairly accurate in predicting price and especially production and quota market ex- ports. In order to evaluate the effects of the ICA, the model was run with the quotas (factual) and without the quotas (counterfactual) for the period 1981-86. (Projections for the 1987-2000 period based on this model are presented in Akiyama and Varangis 1989.) The simulation results for the world price and exports are given in figures 2 and 3. To allow examination of the market stabilizing effects of the ICA, coefficients of variation around the mean for prices, exports, and export revenues are given in table 5. The results show significant price-stabilizing effects of the ICA during the period 1981-85. An interesting result is the ICA'S price stabilizing impact in 1986 when the quota system was not in place. If there had been no quota scheme during the period 1981-85, world coffee prices would have been 24 percent higher in 1986. The explanation for this is that when the quota system was in operation during the period 1981-85, many producing countries were forced to accumulate stocks. When the quotas were lifted in 1986, these stocks were released, dampening the rise in price caused by the drought in Brazil. The simulation results show that had the quotas not been in force for the period 1981-85, total stocks held in producing countries at the end of 1985 would have been 22 million bags instead of the actual 33 million bags, and exports to the quota market would have been 55 million bags in 1986 instead of the actual 62.5 million bags. Table S also shows that the quota system led to significant reduction in varia- tion of export revenues but increased the variation of export quantities. This implies that under the quota system total exports adjusted to stabilize prices resulting in stabilization of export revenues. The benefits of the quota system for each producing country were also estimated. Transfer and risk benefits were calculated in terms of real export revenues (nominal U.S. dollar exports deflated by the World Bank's export unit value of manufactures). Export revenues are sums of export revenues derived from exporting to the quota and nonquota markets. It is assumed here that coffee prices in the nonquota market were 30 percent lower than in the quota market when the quota system was in operation, while the two prices are Figure 2. Factual versus Counterfactual Simulations 260 z 240 - 220 200 ZD 180 - u~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 160 X 60 140 52 ; -8 SU5 11120 100 0 16 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 Key: - counterfactual; -- -- factual. Figure 3. Factual versus Counterfactual Simulations 62- 60- £ 58 56- 54 2 52 -%J 50 48 46- 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 Key: - couriterfactual; ---factual. 168 Akiyama and Varangis 169 Table 5. Coefficient of Variation of Key Variables, with and without the Export Total export quantity to Total real Real world prices the quota market export revenue' Period With quota Without quota With quota Without quota With quota Without quota 1981-85 3.89 10.29 4.86 1.73 9.18 12.67 1981-86 13.79 30.42 9.57 2.82 24.23 30.29 a. Deflated by World Bank's export unit value of manufactures (Muv). assumed to be the same when there were no quotas. Following Newbery and Stiglitz (1981, p. 93), the benefits can be calculated as: B _AT RAU2 (1) Y Y 2 where B = total benefits of the quota system, Y = average real export revenue under the without-quota scenario, AY = the difference in real export revenues with and without quotas, R = coefficient of relative risk aversion, and y = coefficient of variation of real export revenue. The first and second terms of the right-hand side of equation 1 are transfer and risk benefits, respectively. We assumed R = 1, that is, the producers are only somewhat risk averse (see Newbery and Stiglitz 1981 for development of the theory with empirical application based on Binswanger 1980; see also Kanbur 1984 for an empirical review). Before one can calculate risk benefits, instability must be defined. After examining several possibilities we concluded that deviation of export revenues from their three-year moving average would be appropriate, since policymakers in many coffee-exporting countries often use average export revenues of the preceding two to three years as expected export revenue for the current year. Export revenues for 1981-85 and 1981-86 (in 1985 constant dollars), and transfer and risk benefits of the quota system for all ICA exporters and for individual countries with and without quotas are given in table 6. Total transfer benefits from the quota system for the period 1981-85 are negligible. This is partly because of the fact that when the quota was operating, prices received from sales to the nonquota markets were considerably lower than when there was no quota. If 1986 is included, however, the total transfer benefits would have been 4.7 percent higher if there had been no quota system during the period 1981-85. This is because world prices would have been much higher in 1986 had there been no quotas in the period 1981-85.5 5. Another interpretation is that the quota system played the role of a buffer stock in 1986. In this case, as suggested by Newbery and Stiglitz, the transfer benefit from the producers' point of view is negative as long as the absolute value of the price elasticity of demand is constant (as assumed here) and less than unity. See equation 6-61 in Newbery and Stiglitz (1981, p. 95). Table 6. Total Real Export Revenues and Benefits of the Quota System, 1981-85 and 1981-86 Revenues (millions of constant 1985 Benefits (percentage of real U.S. dollars) export revenue) With Without Country Date quota quota Transfer' Risk Total World total 1981-85 54,869 55,115 -0.5 0.4 -0.1 1981-86 69,828 73,087 -4.7 1.6 -3.1 Brazil 1981-85 15,568 15,191 2.4 1.7 4.1 1981-86 17,979 18,327 -1.9 1.8 -0.1 Burundi 1981-85 427 459 -7.3 0.3 -7.0 1981-86 604 599 0.7 0.2 0.9 Cameroon 1981-85 1,397 1,353 3.1 1.9 5.0 1981-86 1,809 1,867 -3.2 5.6 2.4 Colombia 1981-85 8,271 8,114 1.9 0.2 2.1 1981-86 10,876 11,358 -4.4 -1.6 -6.0 Costa Rica 1981-85 1,445 1,524 -5.5 1.4 -4.1 1981-86 1,781 1,942 -9.1 4.4 -4.7 C6te d'lvoire 1981-85 3,726 3,659 1.8 -0.1 1.7 1981-86 4,712 4,887 -3.7 4.5 0.8 Dominican Republic 1981-85 485 581 -5.0 -0.2 -5.2 1981-86 773 772 0.2 -0.6 -0.4 Ecuador 1981-85 1,198 1,316 -9.8 1.4 -1.9 1981-86 1,661 1,757 -5.8 3.9 -1.9 El Salvador 1981-85 2,230 2,184 2.1 -0.4 1.7 1981-86 2,831 2,932 -3.6 5.7 2.1 Ethiopia 1981-85 1,232 1,223 0.7 0.4 1.1 1981-86 1,508 1,567 -3.9 -1.0 -4.9 Guatemala 1981-85 1,947 1,978 -1.6 -1.4 -3.0 1981-86 2,468 2,627 -6.4 5.4 -1.0 Honduras 1981-85 942 969 -2.9 0.7 -2.2 1981-86 1,278 1,387 -8.5 3.9 -4.6 India 1981-85 1,050 1,135 -8.1 0.5 -7.6 1981-86 1,433 1,613 -12.6 4.5 -8.1 Indonesia 1981-85 3,327 3,588 -7.8 -0.5 -8.3 1981-86 4,550 5,035 -10.7 11.5 0.8 Kenya 1981-85 1,305 1,300 0.4 0.5 0.9 1981-86 1,769 1,878 -6.2 2.7 -3.5 Madagascar 1981-85 776 783 -0.9 1.2 0.3 1981-86 991 1,041 -5.1 2.0 -3.1 Mexico 1981-85 2,166 2,261 -4.4 -0.1 -4.5 1981-86 3,001 3,302 -10.0 -6.3 -3.7 Nicaragua 1981-85 721 737 -2.3 1.2 -1.1 1981-86 859 909 -5.9 -1.5 -4.4 Papua New Guinea 1981-85 660 662 -0.2 2.1 -1.6 1981-86 854 914 -7.0 5.4 -1.6 Peru 1981-85 748 859 -4.8 4.1 -0.7 1981-86 1,091 1,118 -2.5 1.5 -1.0 Philippines 1981-85 394 457 -6.2 -1.1 -7.3 1981-86 595 586 1.6 -2.4 1.9 Rwanda 1981-85 446 452 -1.4 0.6 -0.8 1981-86 638 663 -4.0 3.5 -0.5 Tanzania 1981-85 776 809 -4.2 -0.4 -4.6 1981-86 973 1,029 -5.8 2.5 -3.3 Uganda 1981-85 2,172 2,081 4.2 1.0 4.3 1981-86 2,712 2,75s -1.6 1.8 0.2 Zaire 1981-85 1,018 976 4.1 0.5 4.5 1981-86 1,514 1,593 -5.3 7.9 2.6 a. A negative number indicates lower total export revenues in the with-quota case. Source: Authors' calculations, based on World Bank dara available from the authors on written request. 170 Akiyama and Varangis 171 The total risk benefit is small (0.4 percent) for the period 1981-85 but increases to 1.6 percent of total revenue when 1986 is included. This again reflects the significant stabilizing effect that the 1981-85 quotas had on export revenues and on prices in 1986. For individual countries it is interesting to note that transfer benefits are negative for most of the small exporters for the period 1981-86, so that these exporters would have been better off if there had been no quota system for that period. But large exporters such as Brazil, Colombia, and Cote d'Ivoire are among the few countries that gained in terms of transfer benefits from the quota system for 1981-85. This is because when large countries increase their exports, world prices decline-often to the extent that marginal export reve- nues are small or even zero. This result is also verified by the fact that the risk benefits are significant for many small exporting countries but are small for Brazil and Colombia because their export quantity is negatively correlated with world prices. The effect of price stabilization on income stability is greater for countries whose export quantities are positively correlated with prices than those for which the correlation is negative. In fact, price stabilization could reduce income stability for the latter countries. Hence, in general, the quota system benefited large countries in terms of transfer benefits and small countries in terms of risk benefits. A qualification should be made about the interpretation of these results. When the counterfactual runs (that is, without-quota runs) were made, possible effects of changes in risk on supply were not taken into account, and it was assumed that there would be no changes in government policies affecting pro- duction. In many producing countries, high export taxes on coffee are report- edly used to suppress production so that large stocks will not accumulate under the quota system. If this is the case, then in the counterfactual scenario some of these countries could have had lower export taxes and consequently larger production and exports than what the simulation results indicate-and there- fore world prices could have been lower. V. CONCLUSIONS The results in this paper reveal several interesting findings which models focused on short-run effects would not. The ex-post simulation results show that the export quota system had an important stabilizing impact on world coffee prices over the period 1981-85. They also show that coffee prices in 1986, the year prices increased sharply because of the drought in Brazil in 1985, would have been much higher had the quota system not operated during the period 1981-85. This is because producing countries accumulated stocks during the period 1981-85, which were released into the market when quotas were lifted in 1986. In this case, the quota system worked like a buffer stock scheme, that is, it prevented a large increase in coffee prices, which otherwise would have resulted from a significant production shortfall. Symmetrically, the 172 THE WORLD BANK ECONOMIC REVIEW, VOL.4, NO. 2 quotas could have prevented a large drop in prices had a large Brazilian crop been realized. The impact of the quota system in terms of gains in real export revenues was estimated to be rather small. For the period 1981-85, both transfer and risk benefits were calculated to be quite small overall but increased when 1986 was included. The increase in risk benefits when including 1986 shows the signifi- cant stabilizing effect quotas had on prices, although this was at the expense of revenues. The quotas led to decreased real export revenues for most countries, except for the large exporters such as Brazil and Colombia. These countries gained because they face very small or even zero marginal export revenues from increased exports because of their large market shares and the price inelasticity of demand for coffee. However, the risk benefits of the quota system to large exporters are small while they were found to be large for most of the small exporters. To evaluate the total benefits of the quota system, exporting countries should weigh the transfer and risk benefits against the cost of holding additional stocks, for example, interest and warehouse costs. Judging from the fact that exporting countries show great interest in the quota system, risk benefits might be considerably higher than calculated here, especially to the policymakers in these countries. In other words, they might be more risk averse than is assumed here, which implies that the coefficient of relative risk aversion (R) exceeds unity. The distribution of benefits of the quota system for the period 1981-85 favored exporters that were large and/or were traditionally assigned high quo- tas. In such a system, countries with potential for expansion were penalized. Negotiations for a new quota system in the 1990s are likely to include, among other issues, proposals for a redistribution of quotas in favor of these countries. REFERENCES Akiyama, Takamasa, and R. C. Duncan. 1982. Analysis of the World Coffee Market. World Bank Staff Commodity Working Paper 7. Washington, D.C. Akiyama, Takamasa, and P. K. Trivedi. 1987. "Vintage Production Approach to Per- ennial Crop Supply." Journal of Econometrics 36: 133-61. Akiyama, Takamasa, and Panayotis N. Varangis. 1989. "Impact of the International Coffee Agreement's Export Quota System on the World's Coffee Market." World Bank PPR Working Paper 148. Washington, D.C.: Processed. Bates, R. H., and C. Contreras. 1988. "The Economics of Politics in International Trade Policy: Lessons from the International Coffee Agreement." Duke University Program in International Political Economy, Working Paper Number 38, March. Behrman, J. R. 1978. Development, the International Economics Order, and Commod- ity Agreements. Reading, Mass.: Addison-Wesley. Binswanger, Hans P. 1980. "Attitudes Towards Risk: Experimental Measurement Evi- dence in Rural India." American Journal of Agricultural Economics 62 (August): 395-407. Akiyama and Varangis 173 Bohman, M., and L. Jarvis. 1989. "The International Coffee Agreement: Economics of the Nonmember Market." Paper presented at the Canadian Agricultural Economics and Farm Management Society Meeting, Montreal, July 1989. Dick, H., S. Gupta, T. Mayer, and D. Vincent. 1982. "Indexation of UNCTAD Core Commodities Prices by Buffer Stocks or Export Quotas? A Comparison of the Bene- fits for Two Developing Economies." Journal of Development Economics 11: 379- 401. Fisher, B. S. 1972. The International Coffee Agreement: A Study in Coffee Diplomacy. New York: Praeger. Ford, D. J. 1978. "Simulation Analysis of Stabilization Policies in the International Coffee Market." In F. G. Adams and J. R. Behrmann, eds., Econometric Modeling of World Commodity Policy. Lexington, Mass.: Heath. Ghosh, S., C. L. Gilbert, and A. J. Hughes-Hallett. 1987. Stabilizing Speculative Commodity Markets. Oxford: Clarendon Press. Gilbert, C. L. 1987. "International Commodity Agreements: Design and Performance." World Development 15: 591-616. Gordon-Ashworth, F. 1984. International Commodity Control: A Contemporary His- tory and Appraisal. London: Groom Helm. Herrmann, R. 1986. "Free Riders and the Redistributive Effects of International Com- modity Agreements: The Case of Coffee." Journal of Policy Modeling 8, no. 2: 1-25. International Monetary Fund. Various years. International Financial Statistics. Wash- ington, D.C. International Coffee Organization. Various years. Quarterly Statistical Bulletin. Lon- don. Kanbur, S. M. R. 1984. "How to Analyze Commodity Price Stabilization? A Review Article." Oxford Economic Papers 36, no. 3: 336-58. Krasner, S. D. 1973. "Manipulating International Coffee Markets: Brazilian Coffee Policy 1906 to 1962." Public Policy 21, no. 4: 493-523. Labys, W. C. 1973. Dynamic Commodity Models: Specification, Estimation, and Sim- ulation. Lexington, Mass.: Heath. Lien, D-H. D., and R. H. Bates. 1987. "Political Behavior in the Coffee Agreement." Economic Development and Cultural Change 35 (April): 629-36. Maizels, A. 1982. Selected Issues in the Negotiation of International Commodity Agree- ments: An Economic Analysis. Geneva: U.N. Conference on Trade and Development. Newbery, D. M. G., and J. E. Stiglitz. 1981. The Theory of Commodity Price Stabili- zation. Oxford: Clarendon Press. Palm, F. C., and E. Vogelvang. 1988. "Policy Simulations Using a Quarterly Rational Expectations Model for the International Coffee Market." Paper presented at the Twenty-Fifth International Conference of the Applied Econometric Association, In- ternational Commodity Market Modeling, World Bank, Washington, D.C., October 24-26. U.S. Department of Agriculture. Various years. Foreign Agricultural Service: Coffee. Washington, D.C. Vries, J. de. 1975. Structure and Prospects of the World Coffee Economy. World Bank Staff Working Paper 208. Washington, D.C. THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2: 1 75-1 93 Second-Best Foreign Exchange Policy in the Presence of Domestic Price Controls and Export Subsidies David Tarr Poland, like many developing countries, has required its exporters to surrender a share of their foreign exchange earnings to the government at an overvalued exchange rate. During the late 1980s, it progressively increased the share which exporters were allowed to retain (the retention ratio), but other distortions to the trade regime remained. A model developed here estimates the effects of these policies on welfare under different foreign exchange elasticities, export and import subsidies, official exchange rates, and policies on exporter retention of foreign exchange earnings. The retention ratios in effect in early 1989 were equivalent to a 51 percent tax on exports or an import tariff of 130 percent. As economic theory would suggest, maximum social benefit would derive from removal of the full range of distortions. Full retention of foreign exchange by exporters in the absence of other distortions would provide social benefits equivalent to 8 percent of gross domestic product. But the net effect of the other policies together is a bias toward tradables, so that a policy of somewhat less than full retention of foreign exchange is optimal in this second-best world. Foreign exchange restraints are not uncommon in developing countries, and they are typical in centrally planned economies. Exporting enterprises are re- quired to remit to the government all or part of their foreign exchange earnings, and they receive payment at the official (overvalued) exchange rate. In this article I develop a model to analyze the welfare consequences of partial or full liberalization of foreign exchange restraints, in which enterprises are allowed to retain a greater share or all of their foreign exchange earnings. In the empirical application of the model, I estimate the welfare costs of foreign exchange restraints in Poland, incorporating all significant distortions to the trade regime. As of early 1989, that involved export subsidies and-a novelty-implicit subsidies to imports because of domestic price controls. In January of 1990, as part of its "shock" stabilization program, Poland massively devalued its exchange rate, unifying its official and parallel market rates (see Lipton and Sachs 1990). Thus the findings here can be viewed as estimates of The author is an economist in the Technical Department, Europe, Middle East, and North Africa Region, the World Bank. He would like to thank Bohdan Wyznikiewicz for helpful comments. K 1990 The International Bank for Reconstruction and Development / THE WORLD BANK. 175 176 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 the benefits Poland will likely receive from the foreign exchange reforms of 1990. The estimates confirm the conventional wisdom that if there were no other trade distortions, full foreign exchange liberalization would be optimal, or that if other distortions were removed, full foreign exchange liberalization would be the best policy. In Poland, however, foreign exchange restraints are an offset- ting distortion to the export subsidies and implicit subsidies to imports, so that full liberalization of foreign exchange is not optimal. Thus the optimal share of foreign exchange to be retained by the firm (the retention ratio), from the government's perspective, is estimated as a function of the level of other distor- tions. The model is also used to calculate the anti-export bias of less than full retention of foreign exchange. Moreover, because a tax on exports is equivalent to a tax on imports (in that both will lower the equilibrium level of foreign exchange use), this calculation is also made in terms of the equivalent level of tariff protection. The Polish trade and exchange rate regime, as it existed in 1989, is first described. The model is then presented and the benefits to Poland from various retention ratios are estimated, and the optimal retention ratio is calculated, given various assumptions regarding the extent of subsidies to imports. I have also estimated the benefits to Poland of foreign exchange liberalization, with subsidies to imports and exports removed. In the concluding section, most of the important simulations in the article are consolidated in the two summary figures. I. THE TRADE AND EXCHANGE RATE REGIME IN POLAND: 1989 In 1989 Poland embarked on a dramatic trade liberalization effort. Princi- pally this occurred through three institutional changes: the state monopoly on importing was greatly liberalized, foreign exchange was made available at market rates, and foreign exchange obtained through the market was expected to finance the majority of imports. Previously, importing and exporting were restricted to state foreign trade organizations, which are organized along product lines. Exporters turned over to the government their foreign exchange, and the government allocated it to those imports it desired. The first liberalization in this area came with the introduction of accounts for exporters that allowed them to retain a share of their foreign exchange earnings (known as ROD accounts). For 1989, in addition, the government intended to auction a substantial share of the foreign exchange, so that only about one-third of it would be obtained through central allocation. The major- ity of imports thus were to be financed out of ROD accounts or from foreign exchange purchased at market rates. The combination of these two reforms Tarr 177 meant that importers could purchase foreign exchange at market rates and import most goods through competitive trade organizations. Given these reforms, the principal (but not only) distortions which remained in the trade regime involved foreign exchange, for which there were multiple exchange rates. The three most important were the official rate for centrally allocated foreign exchange, the parallel market rate for nonrestricted and illegal transactions, and the rate received by exporters for the dollar value of their exports. First, those institutions that received centrally allocated foreign exchange paid the official price. This included in 1989 the import demands of the health and education sectors, which had little income. The official exchange rate, which was frequently subject to devaluation, was 239 zloty to one U.S. dollar as of March 1987; 400 in March of 1988; and slightly less than 550 as of March 10, 1989. Second, there is the parallel or black market exchange rate, which in the three years before July 1988 was between 3.3 and 4.6 times the official ex- change rate. The parallel market rate has been very volatile: it began to rise in August 1988 and by December 1988 reached 3,380 zloty to the dollar, or 6.7 times the official rate. By March 1989 it was about 2,900, and it rose by about 1,000 zloty per dollar in April 1989. In early 1989 there were a variety of auctions at which foreign exchange could be purchased, yielding a variety of parallel exchange rates. Enterprises could sell their unrestricted foreign exchange from ROD accounts at the Export Development Bank auctions. Early in 1989 these rates closely approximated the parallel market rate, and it was expected that $200 million would be auctioned there in 1989. Foreign exchange to import any one of fifty-two important intermediate products was sold at the Bank Handlowy auctions. It was expected that between $1.5 and $2 billion (billion = 1,000 million) would be auctioned through Bank Handlowy. At the first auction, prices varied from 100 zloty above the official rate to 3,100 above it. Generally, lower prices were paid for larger contracts and for contracts further in the future. Lower prices for the future contracts were efficient because the buyer had to pay immedi- ately, forwent interest, and bore the risk of exchange rate changes. Lower prices for the larger contracts were the result of lack of competition. As of April 1, 1989, individuals have been able to purchase foreign exchange at exchange counters established for this purpose. In addition, about $60 million was expected to be auctioned for the import of consumer goods, for which mostly wholesale enterprises were to be the buyers. These auctions were new and a considerable amount of inefficiency in the markets existed, most notably at Bank Handlowy, so that one price for similar transactions did not emerge. The third important exchange rate is the rate received by exporters for their export earnings. Exporters retain a share of their foreign exchange and must sell the balance to the government at the official exchange rate. The ratio that 178 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 they are allowed to retain varies by product but is generally between 10 and 50 percent and averages about 30 percent overall. This means that the exchange rate received by the exporter is a weighted average of the official exchange rate and the parallel market exchange rate. Given the average retention ratio of 30 percent, in March 1989 exporters were receiving on average an exchange rate of 1,255 zloty per dollar, compared with the 2,900 parallel market rate. II. FOREIGN CURRENCY MARKETS: THE MODEL The model estimated here is explained in detail in Harberger (1988) and discussed in other international trade texts, such as Salvatore (1987) and Dorn- busch and Helmers (1988). This model has been widely used to estimate the effects on the real exchange rate of policy changes generally, and specifically of foreign exchange restraints (Bhagwati 1978). I focus here on the hard currency market and convert all hard currencies to U.S. dollars, so that we can use the term "dollars" interchangeably with "hard currency foreign exchange." The model formulates a system of supply and demand for foreign exchange. The supply of dollars comes principally from the sale of exports. Exporters in each industry have a supply function that expresses their willingness to export as a function of their real exchange rate received, E,. Nominal Polish exchange rates deflated by the ratio of Polish to world prices are real exchange rates. Exchange rates used in the model are real unless otherwise specified. If units of each commodity are chosen so that one unit yields one dollar at world prices, then the aggregate of the supply functions of all of the exporters of the economy is the total supply function of dollars (Harberger 1988). Ex- porters are willing to supply more dollars the higher the exchange rate that they receive. Thus the supply of foreign exchange, Fs, is written as a linear function of the real exchange rate received: (1) F, = a + bE,(l + s) In equation 1, Er is the real exchange rate received (it is derived by deflating the ratio of Polish to world prices), and s is the rate of subsidies on exports. If exporting firms receive subsidies at a positive rate, they will be willing to supply foreign currency (dollars) at a lower exchange rate. Equation 1 is depicted in figure 1; with no subsidy (s = 0), foreign exchange supplied is curve SS. Given a positive subsidy rate, the curve shifts down and to the right and is labeled S(1 + s). (Other sources of foreign exchange that are sensitive to the real exchange rate are discussed in the appendix.) The exchange rate received is a weighted average of the official exchange rate, E., and the parallel market exchange rate, EP: (2) E, = REP + (1 - R)Eo where R is the firm's foreign exchange retention ratio. Note that when R = 1, E, = EP. Tarr 179 Figure 1. Model of Real Exchange Rate Determination Real exchange rate (zloty per dollar) 2,900 1,255(1 + s) -,Sl… 1,255 x \ 550 …-I- … S S(1 + s) tsg Foreign currency D Note: D = foreign exchange demand; D'D = foreign exchange demand given a positive import tariff; D"D = foreign exchange demand given a positive implicit import "subsidy equivalent"; S = foreign exchange supply; s = export subsidy. The demand for foreign exchange derives from a number of sources, but first from the demand for imports. Real exchange rate depreciation raises the num- ber of zloty required per dollar and reduces the quantity of imports demanded. Demand (and supply) in individual sectors will also be affected by nontariff barriers. The aggregate of these sectoral demand functions represents the im- port demand for dollars, Fd, which is assumed to be a linear function and is written as (3) Fd = c-dEP(1 + t) where t, most easily thought of as the tariff rate, is a parameter that measures the policy-induced bias against imports. Payments for external debt would be incorporated in the parameter c. Equation 3 is depicted in figure 1 as the curve DD, and is drawn with t = 0. With a positive tariff rate, equation 3 would shift in and to the left, as is depicted in figure 1 as the dotted line D'D'. Finally, we require equilibrium between demand and supply of foreign ex- change: (4) Fd = Fs In most economies we think of the parameter t as positive, reflecting a tariff rate that discourages purchase of imports. This does not appear appropriate for Poland, because tariffs average less than 10 percent and price controls and 180 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 shortages produce implicit import subsidies. This occurs because price controls generate widespread shortages, that is, the quantity supplied is less than the quantity demanded, given Polish incomes. Given that imports could be pur- chased if dollars were obtained, the excess demand for domestic goods spills over into increased demand for imported consumer goods. Thus, the demand for imports would decrease if the shortages in the economy were reduced through the removal of price controls. That is, the price controls induce a bias toward import purchases that we think of as an implicit "subsidy equivalent," creating a negative value for t in equation 3 (D"D" in figure 1). In related analyses (Tarr 1990a), I have estimated this implicit subsidy equiv- alent for imports in automobiles and color televisions, two industries which have been subject to severe shortages and price controls. It was found that the elimination of the price controls in autos and color televisions would result in a sizable reduction in imports in the long run. Although obviously the govern- ment of Poland did not wish to subsidize the import of autos or color televi- sions, and there was no financial payment for their import, the impact on import demand of price controls on the domestically produced products was equivalent to a subsidy for the import of autos of 43 percent and imports of color televisions of 22 percent. Similar general results were found in the Polish butter market (see Tarr 1990b). One cannot, however, generalize these partial equilibrium results of the television and auto industries to all goods simultaneously. If only price controls on autos are removed, producers can attract labor and capital from other sectors of the economy. If price controls are removed in the aggregate, when all sectors attempt to increase supply, the limited factors of production will result in a bidding up of the prices of those factors, and not all sectors will be able to expand. After a period of adjustment, however, one would expect that liberalization of price controls would result in a more efficient allocation of resources across industries, producing some increase in the value of Polish gross domestic product (GDP). On the basis of these considerations, it is assumed (given best-guess esti- mates) that the import-increasing effect of the price controls is moderately stronger than the disincentive from the tariff, shifting the demand for imports out and to the right. The parameter t in equation 3 is interpreted as the difference between the tariff rate and the import subsidy equivalent of the price controls. That is, if t' is the tariff rate and p is the rate at which imports are encouraged as a result of price controls, then the net ad valorem effect on import demand is: t = t' - p. Given the uncertainty regarding its value, how- ever, I simulate the effects of a number of positive and negative values of t. These equations constitute the model. In a full general equilibrium model, a firm's export supply function depends on all its output and factor prices, not just the relative price of exports to domestic sales. This model, however, focuses on the relative price of exports, including taxes and subsidies, because this is the most important of the variables that affects export supply. In com- Tarr 181 parison with literally dozens of variants of general equilibrium models, this simple model has always reliably predicted the qualitative change in the real exchange rate found in the full general equilibrium models (including versions which incorporate labor-leisure choice). (These comparisons are presented in detail in de Melo and Tarr, forthcoming.) The welfare analysis is based on the concept of net consumers' and producers' surplus, using measurement of Harberger triangles. Welfare estimates based on consumers' surplus have been shown to be good proxies for exact welfare estimates because measures of consumers' surplus are found to lie between the two "exact" measures: the Hicksian equivalent and compensating variations (Willig 1976). Hausman (1981) has taken a critical view of consumers' surplus; he has managed to construct examples in which the Hicksian equivalent and compensating variations differ by a significant amount. But to the extent that consumers' surplus differs from the equivalent variation, it will be closer to the compensating variation, and the latter is as good a welfare measure in most instances. Moreover, in hundreds of simulations using a general equilibrium model we found values based on the Hicksian equivalent and compensating variation measures to be within 1 percent of each other (de Melo and Tarr, forthcoming). 111. LIBERALIZATION OF FOREIGN EXCHANGE POLICY: THE SIMULATION RESULTS Impact of Full Retention of Foreign Exchange Earnings The model described above is used to simulate an increase in the firms' retention ratio to 100 percent and to simulate the effects on the real exchange rate, welfare, and foreign exchange earnings. Six initial simulations are per- formed assuming: (1) high elasticities; (2) low elasticities; (3) high elasticities but a (high) initial parallel market exchange rate of 4,000 zloty to the dollar; and for (4), (5), and (6), high elasticities and net implicit subsidies to imports of 5, 20, and 50 percent, respectively. Aside from these varying parameter values, the values of the macroeconomic variables are based on best estimates for March 1989. High elasticities. For this simulation we begin with the official exchange rate at 550, the parallel market exchange rate in the initial equilibrium at 2,900, the initial retention ratio at 0.3, net tariffs at zero, net subsidies at 0.121, initial foreign exchange supplied at $9.644 billion, and the elasticities of foreign exchange supply and demand at 1.16 and -1.0, respectively. The effects of increasing the retention ratio to 1 are presented in figure 2. With full retention, Polish exporters received real exchange rate increases, inducing additional ex- ports and thus a greater supply of foreign exchange. The additional supply of foreign exchange drives down the price of the dollar in the parallel market until 182 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 Figure 2. Simulation of Welfare Effects of Foreign Exchange Liberalization Assuming High Elasticity of Foreign Exchange Supply and Demand Real exchange rate (zloty per dollar) 2,900 S(1 + s) 1,908 - 1,702 ------- 1,407 9.6 13.1 13.6 Foreign currency (billions of U.S. dollars) Note: D = foreign exchange demand; S = foreign exchange supplied; s = export subsidy rate. demand and supply equalize and the market is in equilibrium at a parallel market exchange rate of 1,702 zloty. Note that, excluding subsidies, exporters were receiving a weighted average rate of exchange of 1,255 in the initial equilibrium as a result of partial retention, so that 1,702 is an increase of only 477 zloty; that is, one should not consider 1,702 an increase to exporters of 1,152 from the official exchange rate of 550. Including the average 12.1 percent export subsidy, exporters received 1,407 zloty for every dollar of export earnings in the initial equilibrium, 1,255 zloty from foreign exchange earnings, and 152 zloty from a government subsidy. The increased real ex- change rate raises foreign exchange earnings by 41 percent-an additional $3.984 billion dollars-for total earnings of $13.628 billion. What are the welfare costs and benefits of these shifts? The demand curve DD represents at any point the marginal value of an additional unit of foreign exchange to the Polish economy. The supply curve SS represents the marginal opportunity costs to the Polish economy of providing the goods and services that yield the foreign exchange. As long as DD exceeds SS, the economy benefits from raising the firms' retention ratios to increase the supply of foreign exchange. But 100 percent retention is not optimal, because of the presence of the export subsidies. The large triangle, A, represents the benefits of increasing retention ratios. To the left of $13.145 billion, DD exceeds SS, so that the Tarr 183 Table 1. Welfare and Exchange Rate Effects of Full Retention of Foreign Exchange Earnings No import subsidies High elasticity High elasticity and import and parallel subsidies of: High Low exchange 5 20 S0 Macroeconomic elasticity elasticity rate' percent percent percent variable (1) (2) (3) (4) (5) (6) Welfare In billions of zloty 2,564 733 4,247 2,106 730 -2,019 In percentage of GDP 9.1 2.6 15.0 7.4 3.9 -7.1 Shadow exchange rate (zloty per dollar) 1,847 1,895 2,392 1,819 1,720 1,427 Equilibrium exchange rate (zloty per dollar) 1,702 1,752 2,199 1,702 1,702 1,702 Optimal retention ratio 0.85 0.85 0.85 0.79 0.60 0.35 Increased foreign exchange earnings at optimal retention ratio (billions of dollars) 3.501 1.003 3.876 3.227 2.493 0.156 Increased foreign exchange earnings at 100 percent retention (billions of dollars) 3.984 1.145 4.340 3.984 3.984 3.984 Note: The estimates assume existing export subsidies (at 0.121). Column numbers refer to the respective simulations. a. 4,000 zloty to the U.S. dollar. Source: Author's calculations, based on data sources described in the appendix. value of an additional dollar exceeds the cost of supplying it. The exchange rate is 1,847 at this "optimal" quantity of foreign exchange. Because subsidies encourage exports, however, the economy will supply more foreign exchange than is optimal once full retention is granted. This will drive down the exchange rate to 1,702, which is slightly appreciated relative to the optimal rate (or shadow price of foreign exchange). The small triangle, B, represents the cost to the economy of supplying too much foreign exchange. The net benefits of providing full retention, given the export subsidies, are equal to the value of area A less that of B-2,564 billion zloty, which is about 9 percent of the estimated GDP for 1988. These results are summarized in column 1 of table 1. We calculate the optimal retention ratio of foreign exchange to be 85 percent, given the official exchange rate of 550 and the continuation of export subsidies. Once full retention of export earnings is achieved, there is no anti-export bias in the system, and there would be no value in continuing with the export 184 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 subsidies, which are justified as a means of countering the net anti-export bias. Thus, having a retention ratio of 100 percent and simultaneously removing the export subsidies would yield optimal results; this would increase welfare by a further 49 billion zloty (or 0.1 percent of GDP). If we consider only the foreign exchange market, this model indicates that a retention ratio of 85 percent and a 12 percent export subsidy yield the same benefits as 100 percent retention ratios without subsidies. But both subsidies and the multiple exchange rate regime create administrative and lobbying costs and incentives to seek rents. When lobbying or administrative failures result in uneven application of subsidies or exchange rates across sectors, this will also distort intersectoral allocation and production. Thus 100 percent retention rates without subsidies is the preferred policy. Low elasticities. The elasticities of supply and demand are now assumed to be 0.3 and -0.3, respectively; otherwise, parameter selection is identical to the high-elasticity case. Because of the lower elasticities, the increase from 30 percent to full retention brings forth a much smaller rise in foreign exchange earnings--only 1.145 billion dollars. The equilibrium exchange rate (given the subsidies) falls to 1,752, but the shadow exchange rate is 1,895; these are not dramatically different from the high-elasticity case. Only if the difference in the supply elasticities in the two cases is much larger or smaller than the difference be- tween the demand elasticities will significantly different exchange rates be gen- erated. There is, however, a considerable decrease in the estimate of the benefits to the economy from full retention. The benefit is equal to 733 billion zloty; this is equal to 2.6 percent of estimated GDP. Because the economy makes a smaller adjustment to the change in retention policy, there is less to be gained from liberalization. Again, there is an overadjustment because of the subsidies, and the optimal retention ratio is 85 percent, given the continuation of subsidies. A policy of 100 percent foreign exchange retention and elimination of the subsi- dies would also yield the optimal result. These results are summarized in column 2 of table 1. High initial parallel exchange rate and high elasticities. Inasmuch as the parallel exchange rate is somewhat volatile, and depreciated during April 1989 by more than 1,000 zloty relative to the value that we simulated for March 1989, we have also simulated the results with an initial parallel exchange rate of 4,000 zloty to the U.S. dollar. With a more depreciated parallel exchange rate, the tax on exports from less than full retention of foreign exchange earnings is greater. Thus, as shown in column 3 of table 1, the benefits of full retention reach 15 percent of GDP in this case. Simulations 4, 5, and 6-high elasticities and subsidies to imports. Empirical evidence suggests and the final set of simulations assumes that the Tarr 18S bias to import purchases as a result of price controls is stronger than the import taxes (that is, in equation 3, t < 0). Otherwise, parameter values are set as in the high-elasticity case. Compared with the high-elasticity case without import subsidies, the optimal increase in foreign exchange supplied is smaller and the optimal retention ratio lower for all levels of import subsidy (see table 1). A retention ratio of 100 percent will now lead to greater excess foreign exchange supplied, so that net benefits are reduced. At an import subsidy of 50 percent, the bias to importing is so great that full retention of foreign exchange reduces welfare by 7 percent of GDP. As discussed above, our best assessment of the actual distortions in the Polish economy is that net implicit subsidies to imports in Poland are about 5 percent. Thus, the benefits of full retention are best characterized by the results of simulation 4, shown in column 4 in table 1. The Anti-Export Bias and the Level of Protection Calculating the anti-export bias. The system of less than full retention of foreign exchange earnings creates a disincentive to export. The model is now used to estimate how high the tax rate on exports would have to be to produce the initial (lower) level of exports if full retention of foreign exchange were in effect, assuming high elasticities and no import subsidy. In order to offset the 3.984 billion dollar increase in exports which would result from a 100 percent retention ratio, it would be necessary to impose a net export tax of 51.5 percent without the existing 12.1 percent export subsidy, or 63.6 percent if the subsi- dies remained in effect. That is, it would require an average net export tax of 51.5 percent just to prevent an increase in export earnings if full retention of export earnings were allowed. Calculating the equivalent level of protection. It is intuitive to think of less than full retention of foreign exchange as a tax on exports. But a basic theorem of international economics (the Lerner symmetry theorem, first established in the 1930s) states that a tax on imports is equivalent to a tax on exports. Applied in this situation it means that the same effects of less than full retention of exchange earnings would result from full retention and imposing a tax on imports. This results because a tax on exports reduces foreign exchange earn- ings and, because the quantity of foreign exchange demanded and supplied must be equal, demand must be reduced to reestablish equilibrium. This occurs as the excess demand for foreign currency drives up its price, raising import prices and lowering import quantities. Thus giving full retention would increase the amount of imports in the economy, not just increase the amount of exports. By identical reasoning, a subsidy to imports is equivalent to a subsidy to exports. In particular, to the extent that domestic price controls induce addi- tional imports, they depreciate the real exchange rate and encourage additional exports. An import tax of 131 percent would be necessary to reduce imports to their 186 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 original level if 100 percent retention rates were granted. Thus the overvalued average exchange rate at which exporters were paid for the surrendered foreign exchange was equivalent to providing import protection of 131 percent. Be- cause the majority of Latin American and Asian countries have had average tariff rates of between 20 and 40 percent, the system of foreign exchange surrender meant that Poland was highly protected in early 1989. Varying the Retention Ratio and Trade Distortions No implicit net subsidies to imports; existing subsidies to exports. I first assume that the average 12.1 percent export subsidy remains in place and there are no net subsidies to imports. That is, I assume that the implicit subsidy to imports from price controls is 8.1 percent, which exactly offsets the 8.1 percent import tariff. In table 2, welfare estimates and the exchange rates vary with alternate retention ratios under these assumptions. The greatest welfare in- crease occurs at a retention ratio of 85 percent; at ratios greater than that, the cost of foreign exchange supplied is greater than the benefits, because of the export subsidy. The Polish government increased retention rates by about 5 percent between 1985 and 1989. The first column in table 2 shows the costs of a 5 percent reduction of retention ratios from the 30 percent base. Thus the absolute value of the welfare changes shown there could be interpreted as the benefits Poland has gained from its policy of increasing the retention ratio in the four years before 1989. Implicit net subsidies to imports of 5 or 20 percent; existing export subsidies retained. As shown in tables 3 and 4, with implicit net import subsidies of 5 and 20 percent, the optimal retention ratios (which yield the maximum gain in welfare) are reduced to 79 and 60 percent, and the benefits to the economy of full retention are reduced to 7.4 and 2.6 percent of GDP, respectively. As suggested above, given macroeconomic resource constraints, it is unlikely that the true net import subsidy is above 5 percent. Thus, the results of tables 2 and 3 are most relevant for the Polish case. Table 2. Simulated Effects of Different Foreign Exchange Retention Ratios with No Implicit Import Subsidies Retention ratio Macroeconomic variable 0.25 0.5 0.75 0.85 1.0 Welfare changes In billions of zloty -1,020 1,990 2,584 2,613 2,564 As a percentage of GDP (1988) -3.6 7.0 9.1 9.2 9.1 Shadow exchange rate 1,847 1,847 1,847 1,847 1,847 Equilibrium parallel exchange rate 3,089 2,361 1,958 1,847 1,702 Note: Assumes 12.1 percent export subsidy and initial retention ratio of 30 percent. Source: Author's calculations, based on data sources described in the appendix. Tarr 187 Table 3. Simulated Effects of Different Foreign Exchange Retention Ratios with 5 Percent Implicit Import Subsidies Retention ratio Macroeconomic variable 0.25 0.5 0.75 0.79 1.0 Welfare changes In billions of zloty 926 1,754 2,204 2,209 2,106 As a percentage of GDP (1988) -3.3 6.2 7.8 7.8 7.4 Shadow exchange rate 1,819 1,819 1,819 1,819 1,819 Equilibrium parallel exchange rate 3,089 2,361 1,958 1,910 1,702 Note: Assumes 12.1 percent export subsidy and initial retention ratio of 30 percent. Source: Author's calculations, based on data sources described in the appendix. Implicit net subsidies to imports of 50 percent; existing export subsidies retained. Under this simulation (not shown), any retention ratio in excess of 32 percent has a negative impact on welfare. The optimum retention ratio is about a 2 percent increase from the assumed existing retention ratio, to 32 percent. Thus, if existing implicit import subsidies are this high, the import incentives are so great that a liberalization of the foreign exchange regime almost immediately becomes counterproductive. No net subsidies to imports or subsidies to exports. In this experiment, initially, export subsidies are removed and net subsidies to imports are set at zero; subsequently the effects of different retention ratios are simulated and reported. Because liberalization of the other trade distortions occurs first, in- creasing the retention ratio is unambiguously beneficial (table S). The most striking result is that a reduction of 5 percent in the retention ratio (to 25 percent) results in a loss of 9.1 percent of GDP, which is considerably greater than the losses from a reduction in the retention ratio to 25 percent in the previous simulations. This occurs because at the initial equilibrium (given the 30 percent retention ratio) there is already too little foreign exchange supplied. Reduction of the export subsidies in addition to the reduction in the retention ratio further reduces foreign exchange supplied, exacerbating losses. Table 4. Simulated Effects of Different Foreign Exchange Retention Ratios with 20 Percent Implicit Import Subsidies Retention ratio Macroeconomic variable 0.25 0.5 0.6 0.7 1.0 Welfare changes In billions of zloty -645 1,048 1,137 1,106 730 As a percentage of GDP (1988) -2.3 3.7 4.0 3.9 2.6 Shadow exchange rate 1,720 1,720 1,720 1,720 1,720 Equilibrium parallel exchange rate 3,089 2,362 2,175 2,024 1,701 Note: Assumes 12.1 percent export subsidy and initial retention ratio of 30 percent. Source: Author's calculations, based on data sources described in the appendix. 188 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 Table S. Simulated Effects of Foreign Exchange Restraints, Export Subsidies, and Implicit Import Subsidies Retention ratio Macroeconomic variable 0.25 0.5 0.75 1.0 Welfare changes In billions of zloty -2,584 1,060 2,030 2,209 As a percentage of GDP (1988) -9.1 3.7 7.2 7.8 Shadow exchange rate 1,819 1,819 1,819 1,819 Equilibrium parallel exchange rate 3,198 2,494 2,085 1,819 Note: Initial 12.1 percent export subsidies and S percent implicit net import subsidies are reduced to zero, and the initial retention ratio is 30 percent. Source: Author's calculations, based on data sources described in the appendix. IV. LIMITATIONS OF THE ANALYSIS There are two weaknesses in the methodology used here that should be noted. First, the model does not account for the demand or supply of dollars created by asset accumulation. People may wish to hold dollars as an asset, not as a means to purchase imports, especially during periods of domestic inflation. In the short run, these financial reasons may dominate the trade-related factors emphasized above. In the long run, however, people will not continue to acquire dollars as an asset, without those dollars entering the market for goods and services. Moreover, elasticities of trade-driven foreign exchange supply and demand are greater in the long than the short run, and should dominate in a longer time frame. As Caves and Jones (1985, p. 361) have stated: "We can hardly imagine commodity markets running permanently out of equilibrium with unexploited opportunities to arbitrage goods from one country to another. ... Therefore we expect that financial forces will determine the exchange rate in the short run following any disturbance to equilibrium, but trade flows must dominate the long run result." Thus this model appears to be a reasonable representation of the factors which influence the exchange rate in the long run. Second, the supply of and demand for dollars here is derived fromn the supply of exports and demand for imports. These underlying demand and supply schedules are dependent on domestic prices, including factor market prices. To the extent that the underlying prices are distorted, the demand and supply schedules may not represent the marginal social costs or benefits of the goods. In that case, the supply and demand for dollars similarly may not represent society's valuation of the benefits or costs of an additional dollar. This implies that our measure of welfare is subject to error. In a related paper (Tarr 1989, available upon written request), a model is constructed in which the price of foreign exchange differs from the marginal social benefits and costs. Those differences are characterized by random varia- bles which may be interpreted as reflecting measurement error from all the sources of uncertainty, but most notably from uncertainty about the difference Tarr 189 between the price of foreign exchange and marginal social benefits and costs. The paper shows that the variance of the net benefits of a foreign exchange retention policy increases as the retention ratio increases. Consequently, the risk-averse decisionmaker will desire a lower foreign exchange retention ratio than either a decisionmaker who is risk-neutral or a decisionmaker who oper- ates under less uncertainty as a result of less distortion in the underlying supply and demand curves. The qualitative impact of risk aversion under uncertainty is the same as the effect of export subsidies: a further departure from 100 percent retention of foreign exchange is optimal. V. CONCLUSIONS Using the model outlined above in an application to Poland in 1989, changes in welfare have been related to the share of foreign exchange earnings retained by exporters given various foreign exchange elasticities, degrees of overvalua- tion, and trade subsidy levels. On the basis of the results of the various simu- lations (tables 2-4), the change in welfare (as a percentage of GDP) is graphed against the foreign exchange retention ratio in figure 3. All curves have the export subsidy at its level in March 1989, and different levels of implicit import subsidies assumed in the initial equilibrium. The higher curves are associated with an assumed lower initial level of trade distortions. The subsidies to exports and implicit subsidies to imports (t) represent offsetting distortions to the foreign exchange restrictions. Less than 100 percent retention of foreign ex- change discourages exports and imports, but the net effect of the other trade distortions is to encourage exports and imports. Thus foreign exchange liber- alization produces greater benefits the lower are other trade distortions in the initial equilibrium. Without reduction of the other trade distortions, increasing foreign exchange retention ratios beyond 65-85 percent will reduce welfare. Other trade distortions must be reduced if retention ratios of less than 85 percent are to produce additional benefits. These are "second-best" experiments that simulate the effects of changing the foreign exchange restrictions, given the presence of other distortions to the trade regime. First-best policies are then simulated, assuming first the best estimate of the March 1989 level of trade distortions, then simulating their elimination and increases in foreign exchange retention ratios (table 5). Figure 4 shows in- creased benefits up to 100 percent retention of foreign exchange, which yields benefits equivalent to 7.8 percent of GDP. This is the estimated benefit from more efficient allocation of foreign exchange; reducing other distortions, such as domestic price controls, will produce a more efficient allocation of domestic resources and yield additional benefits that are not estimated by this model. In the application to Poland, less than full retention of foreign exchange is estimated to be equivalent to either a net export tax of 51.5 percent or an import tariff of 131 percent. Figure 3. Simulated Welfare Effects of Foreign Exchange Retention Ratios under Differing Trade Policies 10 9- a- t 0,' t =-0.06 = 2; s = 12.1 c- 12.1z=Os=1. *., a 6- 5 n 4- a 3- 0 ;_t 12.1 ;- . 0 -6 s -3 -8 - -56 -7 0.25 0.30 0.32 0.35 0.40 0.50 0.60 0.65 0.70 0.75 0.79 0.85 1.00 Retention ratio of foreign exchange Note: s = rate of subsidy as a percentage of export value; t= net policy (dis)incentive to imports as the difference between tariffs and the implicit subsidy equivalent resulting from price controls on domestic substitutes. Figure 4. Simulated Welfare Effects of Foreign Exchange Retention Ratios Given Liberalization of Trade Policy 8- 7 _- /;s 6- 0 4 3- 0. U~1-1 -2 Cd 3-3 Cd -5 -6 -- -7- -8 -9 0.2S 0.30 ~~~0.50 0.75 1.00 Retention ratio of foreign exchange Note: s rate of subsidy as a percentage of export value; t = net policy (dis)incentive to imports as the difference between tariffs and the implicit subsidy equivalent resulting from price controls on domestic substitutes. 190 Tarr 191 APPENDIX. THE DATA, ASSUMPTIONS, AND PARAMETER ESTIMATES The official exchange rate was slightly less than 550 during early March 1989, and the parallel rate was 2,900 at that time. The retention rate is taken to be 30 percent. Without loss of generality, indexes were chosen so that relative prices between Poland and the rest of the world are initially unity. Thus, the nominal exchange rates are initially identical to the real rates, but as relative prices change over time, the nominal and real exchange rates will differ. The subsidy rate to exports has two parts: fiscal incentives and the price equalization payments. Fiscal subsidies in 1988 were 7.1 percent of the value of exports. Data on subsidies and tariffs are taken from UNDP / World Bank (1989). Some enterprises were taxed on exports and some were subsidized, but on balance in 1987 the price equalization payments created an 8.2 percent subsidy to exports to hard currency areas. In an efficiency enhancing reform, payments to exporters under the price equalization fund were reduced over time, so that we take 5 percent as the price equalization subsidy in 1989. This yields a 12.1 percent subsidy from fiscal subsidies and price equalization pay- ments. Net customs duties were 4.1 percent in 1987. In addition, through the operation of the price equalization account, firms paid varying import taxes, but on balance additional import taxes of 5.9 percent were collected in 1987. As above, the import taxes under the price equalization account are assumed to be lower (4 percent) as a result of the reforms, so that in aggregate we take import duties to be 8.1 percent in 1989. There is an opposite bias on imports deriving from the price controls, however, which appears to be at least as strong as the tariff. The supply of foreign exchange is derived from data on the balance of payments in convertible currencies in 1988. The sources of foreign exchange in the current account were merchandise exports ($7.248 billion), private remit- tances ($1.433 billion), and sales of services ($0.963 billion), for a total supply of foreign exchange in the initial equilibrium of $9.644 billion. The demand for foreign exchange is based on merchandise imports ($6.307 billion), nonfac- tor services ($0.544 billion), and debt payments which are derived as the difference between foreign exchange supply and merchandise and service im- port demand so that foreign exchange demand and supply is equilibrated. There is also a supply and demand for foreign exchange deriving from capital movements and government import programs, respectively. The effects of mov- ing to greater retention of foreign exchange are completely independent of the initial value of foreign exchange supplied or demanded. The additional foreign exchange generated simply shifts the supply and demand curves out to the right, but it does not affect variables otherwise influenced by policy shifts. This is the one parameter in the model that does not affect the results. Thus the supply of foreign exchange deriving from capital movements is ignored. 192 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 The final parameters required are the elasticities of supply and demand with respect to the real exchange rate. Two estimates of the supply elasticity are available: an elasticity of 0.2, from annual data for three years endling in 1987, (UNDP / World Bank 1989) and, from monthly export data between December 1985 and December 1987, 1.16 during the first sixteen months and 0.19 during the last nine months (Marczewski 1988). Because the model here is long-run and elasticities are larger in the long run than the short run, and because recent institutional changes allow increased price response of exports, 1.16 is used as the best (high) estimate, but results for a low-elasticity estimate of 0.3 are also calculated. There is less information available on the elasticity of demand, but the results are estimated for a range of elasticities from 1.0 to 0.3, comparable to the supply elasticities. I solve for the unknown parameters in equations 1-4 (a, b, c, and d). Given the value of the supply and demand elasticities, I first determine the value of b in equation 1 and d in equation 3. Given the values of b and d, I solve equations 1-3 for the unknown parameters a and c. Given the values of these parameters, other parameters in the model are then changed to simulate policy experiments. REFERENCES Bhagwati, Jagdish. 1978. Anatomy and Consequences of Foreign Exchange Controls. Cambridge, Mass.: Ballinger. Caves, Richard, and Ronald Jones. 1985. World Trade and Payments. Boston: Little, Brown. de Melo, Jaime, and David Tarr. Forthcoming. A General Equilibrium Analysis of U.S. Trade Policy. Cambridge, Mass.: MIT Press. Dornbusch, Rudiger, and F. L. Helmers, eds. 1988. The Open Econonzy: Tools for Policymakers. New York: Oxford University Press. Harberger, Arnold. 1988. "Trade Policy and the Real Exchange Rate." World Bank Economic Development Institute. Processed. Hausman, Jerry. 1981. "Exact Consumer's Surplus and Deadweight Loss." American Economic Review 11: 662-76. Lipton, David, and Jeffrey Sachs. 1990. "Creating a Market Economy in Eastern Europe: The Case of Poland." Brookings Papers on Economic Activity, no. 1. Marczewski, K. 1988. "The Real Exchange Rate in Poland and Hungary in the 1980s." Handel Zagraniczny 33: 11-15. Salvatore, Dominick. 1987. International Trade. New York: Macmillan. Tarr, David. 1989. "The Implications of Distortion-Induced Uncertainty on Optimal Foreign Exchange Policy." Technical Department, EMENA Region, World Bank. Wash- ington, D.C. Processed. . 1990a. "A Disequilibrium Model of the Welfare Effects of Foreign Exchange Restraints and Price Controls: The Case of Automobiles and Color Televisions in Tarr 193 Poland." World Bank Europe, Middle East, and North Africa Region Working Paper IDPOO58. Processed. . 1990b. "Quantifying Second Best Effects in Grossly Distorted Markets: The Case of Butter in Poland." Journal of Comparative Economics 14: 105-09. UNDP (United Nations Development Programme) / World Bank. 1989. "Poland: Poli- cies for Trade Promotion." Trade Expansion Program, Country Report 1. New York. Processed. Willig, Robert. 1976. "Consumer Surplus without Apology." American Economic Re- view 66: 589-97. THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2: 1 95-207 Import Dependency and Structural Adjustment in Sub-Saharan Africa Ramon E. Lopez and Vinod Thomas One of the effects of structural adjustment programs in Sub-Saharan Africa has been the reduction of imports in the face of scarce foreign exchange. This article takes the analysis of import demand beyond the traditional income and price determinants to account for factors likely to be important to Sub-Saharan African countries in the 1990s. First, the effect of demand on imports is reflected by the level of absorption rather than the less direct income variable. Second, because adjustment programs may cut government consumption and, through increases in interest rates, reduce investment, these components of absorption are also considered independently to assess their differential effect on imports. Third, import barriers are often set in dollar terms to limit the use of foreign exchange. Because reliable and complete data for import restrictions are not available, the ratio of exports to debt is included as an indicator offoreign exchange availability to reflect its effect on trade barriers and thus imports. The findings suggest that this more comprehensive assessment of import demand will be needed if the size and even direction of changes in import demand in response to policy reform is to be understood and anticipated. In response to the crises in trade and foreign currency flows in the 1980s, most Sub-Saharan African countries adopted structural adjustment programs. Given the external sector constraints, substantial declines in imports result, at least in the short term. But traditional analysis has not been able to anticipate the extent-and sometimes even the direction-of changes in imports in response to the adoption of adjustment policies. We extend the coverage of earlier models of import determination to account for the factors that often lead to the adoption of adjustment programs, particularly changes in terms of trade and foreign exchange shortages, and the policy changes that are commonly included in them-reductions in the level of and shifts among the components of absorption, and devaluation of the exchange rate. Most empirical analyses of aggregate imports have used a static, single equa- tion specification in which the value of imports is determined by real income Ram6n E. L6pez is an economist and professor of agricultural and resource economics at the University of Maryland, College Park, and Vinod Thomas is division chief in the Country Economics Department of the World Bank. The authors thank Ann Harrison, Bela Balassa, John Holsen, and Paul Isenman for their useful comments on an earlier version of this article. © 1990 The International Bank for Reconstruction and Development / THE WORLD BANK. 195 196 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 and the relative price of imports. The price of imports is commonly defined as the nominal exchange rate times the international price of the import deflated by a domestic price index. Implicit in the conventional model are the assump- tions that there are no binding import quotas and that absorption (public and private investment and consumption), which is a more direct determinant of import demand than is income, varies proportionally with real income. But if the ratio of absorption to real income varies, as has been the case in the countries examined here, the model would yield an unstable coefficient of the real income variable and thus could not determine import values adequately. Furthermore, if there are binding import quotas, they prevent domestic prices of imports from being determined directly by the exchange rate and the inter- national price. Import quotas are often used not only to protect domestic industry but also to limit the use of scarce foreign exchange. Thus quotas are often set in value terms, and foreign exchange availability will affect both the exchange rate and the value of import quotas, neither of which is reflected in the traditional model. (Excellent discussions concerning the traditional import demand models are contained in Leamer and Stern 1970, Magee 1975, and Goldstein and Khan 1985; Agbonyitor 1986 and Pritchett 1987 estimate im- port demand in Sub-Saharan Africa using the traditional approach.) The model developed here explicitly accounts for the existence of binding quotas set in value terms, for the effect of foreign exchange availability on quota levels, for changes in the level and composition of absorption relative to real income, and for the conventional demand variables. Thus this model draws from and expands upon both the traditional import demand models that treat imports as demand-determined and the more recent models that assume that imports are determined by foreign exchange availability (Hemphill 1974; Sun- dararajan 1986; Winters 1987; and Moran 1989). We consider the role of international debt and export receipts as determinants of the degree of restrictiveness and scope of import restrictions. It is assumed that as the availability of foreign exchange decreases, governments gradually extend the coverage of import quotas and reduce their dollar value. By includ- ing the stock of debt as an explanatory variable in the import equations, we account for its effects on both foreign exchange availability and on net inter- national assets, which, as an important component of wealth, is a factor determining expenditure decisions and thus imports (see Sachs 1981, 1982; Dornbusch 1983). We further strengthen the analysis through the disaggrega- tion of absorption into government consumption, private consumption, and aggregate investments-changes in each of which can have a different impact on imports. An important consequence of the existence of value import quotas affecting a subset of the imports is that the effects of exchange rates and border prices on aggregate imports are not identical as usually assumed. This framework is used to estimate import demand for 1966-86 for seven Sub-Saharan African countries: C6te d'Ivoire, Kenya, Madagascar, Nigeria, L6pez and Thomas 197 Tanzania, Zaire, and Zambia. In these countries, the lack of external financing in the 1980s has limited import growth. Import levels have been compressed considerably, which has contributed to declines in growth rates of gross do- mestic product (GDP). The countries have, to varying degrees, cut absorption, devalued their real exchange rates, and reduced trade restrictions. Thus their situation, and the analytical framework that reflects it, are relevant to macro- economic policy reform in many developing countries in the 1990s. I. POLICY VARIABLES AND THE ECONOMETRIC SPECIFICATION Macroeconomic Policy Capital and final goods imports have a more direct relationship with real levels of absorption (consumption and investment) than with total income. Imported intermediate goods, as inputs into production, are related more di- rectly to real income than to absorption. During adjustment programs, it is likely that all components of absorption will decline: cuts in government expenditures and the contractionary effects of those cuts will decrease both public and private consumption and investment, whereas increased interest rates (common in adjustment programs) will inde- pendently discourage investment but may raise savings levels. Trade policy reform often includes exchange rate devaluation, which both encourages ex- ports and discourages imports. Thus absorption would grow at a slower pace than real income during adjustment, and the share of imports in overall growth would also decline, at least in the short term. When the import intensities of the components of total absorption are significantly different, changes in the structure of absorption also have important effects on imports. The changes in the relative shares of public and private investment and consumption that often result from adjustment programs may have a more significant effect than changes in the total level of absorption. Import Specification with Import Restrictions We consider both freely importable commodities and imports subject to quantitative restrictions (QRS) set in foreign value terms (dollars). The total value of transactions for each restricted import (ir) is set by the dollar-denomi- nated quota ceiling, Oi, and is the product of the border price, p*, times the quantity of real net domestic demand, Di: (1) oi = P,;* Di (Plrl .. I Pir ** PNrl Pf, Y-y;A 1, N Demand is determined by the domestic prices of the N quota-restricted imports, Prl . . . Pr the prices of unrestricted imports (the vector pf), GDP, income Y, and absorption, A. All domestic prices pi, pif Y, and A are deflated by a GDP deflator which includes export prices. 198 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 The solution of equation 1 gives the domestic prices of restricted imports: (2) p= i, Yf y .x p; . p;1 . . .r N In some of the countries studied, several quota-restricted imports are also subject to domestic price controls so that equation 1 would not be valid. However, the existence of parallel or black markets for goods subject to price controls is well documented in Africa. Thus, equation 1 could still be used to determine the relevant Pir for parallel rather than the official markets. Because these prices are endogenous and do not enter directly into the empirical esti- mation below, the lack of data on parallel market prices poses no problem. Under the commonly used assumption of demand substitutability, it can be shown that (3) aap@/ --< 0, apk 0, for all i, j, k. We also assume that the domestic prices of freely importable commodities are equal to their foreign currency prices, p* times the exchange rate, e, adjusted by the average tariff rate, t: (4) Pf = e(1 + t)p . In the aggregate, both demand for merchandise imports, m, and the average price of restricted imports, PI, are functions of domestic prices of iunrestricted and quota-restricted imports, income, and absorption: (5) m = m [P(Pf PI), )Y, (6) = PI ° yA where p(pf,p,) is a price index for aggregate imports, normalized by the price of exportables; 0 is the total value of imports subject to QRS; and p is an index of world prices of value-restricted imports. Equations 4-6 can be combined into the following aggregate import specifi- cation: (7) m, = bo + b, Y, + bY)- b3Le(l + t)p ] + b4( p-) + b5sm,l in which all variables are expressed in log form, all coefficients (except the intercept) are expected to be positive, and we include m, to allow for a lagged adjustment. Assuming that the world prices p* and p r move approximately proportionally, we can consider one world foreign currency price, p*. (Note that this does not imply that domestic prices pr and pf also move together.) Lopez and Thomas 199 Thus equation 7 can be rewritten as (8) bo ==bo - b3(1 + t) + b4O + b,Yt + - b3e, - (b3 + b4)p* + b5mt- Equation 8 yields a specification in which imports are simultaneously deter- mined by demand factors and foreign exchange constraints. In our initial estimation, we incorporated the terms b3(1 + t) and b40 into the constant term because we assume no change in total quota values of restricted imports and because African tariff rates changed little during the period of analysis: (9) mt = bo + biYt + b2() - b3e - (b3 + b4)p + b5mt, Later we relax the assumption that 0 is constant by using measures of foreign exchange availability as proxies for 0. It is important to note that in equation 9 the elasticity of demand with respect to the world price is greater than the elasticity with respect to the exchange rate, as would be suggested by equation 7. When import quotas are set in foreign currency values, devaluation (an increase in e) will have no direct effect on the real imports of quantity-restricted goods (equation 6). Devaluation will have only an indirect effect due to the increase in the domestic price of freely importable goods. An increase in p*, however, will reduce the physical quantities of imports subject to quotas as well as the imports of freely import- able commodities. The conventional specification that ignores the effect of absorption and assumes equal exchange rate and international price elasticities can be verified by testing the null hypothesis b2 = 0 and b4 = 0, respectively. Because we use real income and absorption as explanatory variables, equa- tion 9 may seem to come close to being an identity. But the left-hand side covers only merchandise imports, which have been less than 70 percent of all imports in most countries, and this share has fluctuated substantially during the period. Furthermore, exports are not exogenous and have also varied significantly. Also, we consider the share of absorption in income rather than simply absorption, which is likely to mitigate this potential problem. To account for the endogeneity of absorption, we estimated equation 9 using lagged values of absorption and fiscal deficits as instruments and obtained a pattern of results quite similar to those reported here, with some loss in effi- ciency. It appears that the variability in the absorption-GDP ratio in most African countries is largely related to changes in exogenous fiscal expenditures and monetary policy. This makes the problem of endogeneity of absorption less serious. Use of GDP less exports instead of absorption as an explanatory variable would reduce the likelihood of spurious correlation resulting from the endogeneity of explanatory variables. In many African countries, however, the 200 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 export data are particularly weak because of widespread smuggling and under- invoicing (see Yeats, this issue) and hence this procedure would have added considerable noise to the estimation. II. ESTIMATION OF AGGREGATE IMPORTS The Initial Specification We estimated import demand for the seven Sub-Saharan African countries using annual data for 1966-86 (see the appendix for data definitions and sources). Table 1 shows both the estimated long-run elasticities of imports with respect to each of the explanatory variables and the F-test for the hypothesis that the elasticities with respect to exchange rate and international import prices are the same, b4 = 0. The long-term import elasticities for 1966-86 in table 1 show the expected signs and are generally significant. The GDP coefficients, which serve as the estimates of real income elasticities, have an average value greater than 1. Similarly, however, with the exception of Madagascar and Zaire, the coeffi- cients of absorption are positive and highly significant. This therefore suggests that the value of the GDP coefficient is conditional on absorption remaining as Table 1. Estimated Long-run Aggregate Import Elasticities, 1966-86 Dollar import Equality of Real Absorption/ Exchange prices, _ effects of Country GDP, Y GDP, A/Y rate, e px R2 eandp*a C6te d'lvoire 0.97 1.01 -0.58 -0.61 0.94 Accept (2.88) (2.49) (-2.88) (-2.88) [0.40] Kenya 0.55 1.74 -0.51 -0.36 0.87 Accept (3.82) (6.65) (-3.02) (-3.05) [2.25] Madagascar 3.34 0.39 -0.74 -0.83 0.77 Accept (3.33) (0.24) (-2.11) (-2.97) [1.90] Nigeria 1.07 2.50 -0.59 -0.66 0.96 Reject (2.38) (5.27) (-2.52) (-1.92) [4.81] Tanzania 0.74 2.96 -0.21 -0.54 0.48 Reject (1.63) (3.25) (-0.81) (-2.06) [6.09] Zaire 2.79 -0.32 -0.24 -0.74 0.92 Reject (4.42) (-0.95) (-2.16) (-5.46) [28.10] Zambia -0.04 0.96 -0.27 -0.35 0.95 Reject (-0.04) (3.56) (-1.59) (-1.25) [7.30] Medianb 1.07 1.74 -0.51 -0.57 Note: t-statistics are in parentheses. Ordinary least squares estimates were calculated for Kenya, Tanzania, Zaire, and Zambia; others are first-order Cochrane-Orcutt. Variables are deflated by the GDP deflator, except the import price index, px. Using the h-statistic, the null hypothesis of no autocorrela- tion could not be rejected at a 5 percent significance level for every country except Madagascar. a. F-statistics for the restriction that the coefficients of the exchange rate and dollar import prices are equal; Ho: b4 = 0. b. Excludes those coefficients that have the "wrong" sign. Source: World Bank data (see appendix). L6pez and Thomas 201 a constant share of GDP. If the absorption-GDP ratio varies by even a small amount as GDP changes, the net impact on imports is likely to be strongly affected, as indicated by the generally high coefficient for the absorption-GDP variable. This would imply that specifications that ignore the absorption-GDP effect are likely to produce income elasticities that are highly unstable. Consistent with the predictions from the theoretical model discussed above, the import price elasticities are generally higher in absolute values than the exchange rate elasticities. Kenya is the one exception: the hypothesis of equal price and exchange rate elasticities cannot be rejected at the 1 percent signifi- cance level. In addition the hypothesis of identical exchange rate and interna- tional price elasticities is rejected in four countries (Nigeria, Tanzania, Zaire, and Zambia) at the 5 percent level of significance, whereas in Kenya it was rejected at the 10 percent level. The null hypothesis could not be rejected at any reasonable level of significance only for C6te d'Ivoire. This is probably because C6te d'Ivoire appears to have relied less on import restrictions than have most of the other countries (Halevi 1988). The complete conventional specification, which ignores absorption and the independent effects of the ex- change rate and world prices (that is, which imposes b2 = 0 and b4 = 0), cannot be refuted at the 10 percent level of significance only for Madagascar. For all other countries it appears that significant gains are attained by including absorption as an explanatory variable or by allowing for different price and exchange rate elasticities. Because absorption is determined'in part by GDP, exchange rates, and import prices, ideally the model should be closed with an absorption equation that should be jointly estimated with the import demand equations using simulta- neous equation methods. This would permit the calculation of the total effect of exchange rates, prices, and GDP, including their direct effects as well as the indirect effects via absorption changes. If an increase in the exchange rate reduces the absorption-GDP ratio, this would strengthen the effect of devalua- tion on imports. The results in table 1 show that government policies, particu- larly those affecting absorption and exchange rates, can have a powerful effect on imports. Accounting for the Structure of Absorption The import equations next were reestimated using disaggregated expendi- tures to examine the influence of the structure rather than only the level of total absorption. Expenditures were disaggregated into private and government consumption and investment, and the income and absorption-GDP ratio were excluded. The specification estimated is: (10) m, = CO ± c E + CcEp + c ce + c,pI + c +4m where Ep is real private consumption, Eg is real government consumption, I is real investment, and all are deflated by the domestic currency GDP deflator. All 202 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 other variables are as previously defined, and as in equation 8, all variables are in log form. We excluded real GDP from equation 9 after several runs showed that its coefficients were quite unstable, probably because of the high degree of collinearity between GDP and some of the absorption variables. This analysis covers only 1966-83 because of a lack of more recent data on disaggregated expenditures. Table 2 shows that the long-term import demand elasticities have the ex- pected signs, with the exception of private consumption in C6te d'Ivoire and Nigeria, and the exchange rate in Tanzania. Only in Cote d'Ivoire, however, is the unexpected sign significantly different from zero. Unfortunately, we see a marked increase in the number of coefficients that are not significant at the 5 percent level, rising from 36 percent in table 1 to 63 percent here. The relationship of the different categories of expenditures with imports varies widely across countries. Government expenditure is the most import- intensive of the three categories in Cote d'Ivoire, Madagascar, and Zaire, whereas private consumption is the least import-intensive in four of the seven. For these countries, this would imply that decreasing government consumption and increasing private consumption by the same amount could result in a net decrease of imports. The quantitative importance of the exchange rate is somehow lower when total absorption is disaggregated into its components. Thus the net effect of a package of trade reform with exchange rate adjustment on the level of imports is uncertain. If tariffs are substituted for import restrictions and the tariffs are reduced, import levels need not increase. A lowering of import restrictions and Table 2. Estimated Long-run Aggregate Import Elasticities Considering Disaggregated Absorption, 1966-83 Real private Real government Real Exchange Dollar import consumption, consumption, investment, rate, prices, Country E, Es I et p* R2 Cote d'lvoire -0.55 0.64 0.22 -0.83 -0.54 0.98 (2.11) (8.49) (3.04) (-14.65) (-3.61) Kenya 0.18 0.02 0.52 -1.03 -0.81 0.64 (0.13) (0.03) (2.76) (-1.86) (-1.00) Madagascar 0.75 0.96 0.10 -1.44 -1.33 0.77 (1.15) (1.90) (0.41) (-2.06) (-3.62) Nigeria -0.77 0.19 0.92 -0.63 -0.25 0.96 (-0.84) (0.71) (6.58) (-1.06) (-0.29) Tanzania 0.90 0.61 0.88 0.38 -0.54 0.81 (1.38) (2.92) (4.33) (0.69) (-0.99) Zaire 0.38 0.57 0.49 -0.16 -0.68 0.86 (0.81) (2.72) (2.62) (-0.53) (-2.34) Zambia 0.35 -0.03 0.53 -0.92 -0.98 0.88 (0.66) (-0.08) (2.80) (-1.04) (-1.69) Note: t-statistics are in parentheses. All estimates are ordinary least squares. Source: World Bank data (see appendix). L6pez and Thomas 203 tariffs will increase imports. But that increase may be offset by a depreciation of the real exchange rate, induced by macroeconomic policies. Trade reform often involves export expansion, but any significant increase in nontraditional exports may be expected to increase demand for imported inputs. The import- intensity of export production, however, is not independently assessed here. Foreign Exchange Variables In Sub-Saharan Africa, import quotas are often imposed in response to the increasing scarcity of foreign exchange. To assess the role of foreign exchange availability in the determination of imports, we estimate equation 9 allowing 0, the value of import quotas, to change over time. Because data for 0 are not available, we use data on foreign exchange availability as a proxy for 0. In analyzing the effect of foreign exchange constraints on imports, it is necessary to use both stock and flow variables: foreign reserves or net debt, plus exports, capital inflows, or changes in foreign exchange reserves. If one uses only a subset of flow variables, the estimates are likely to be unstable because their effect will be dependent on underlying stock variables. If all flow foreign exchange variables are used, however, then one is estimating something very close to an identity-imports are equal to the foreign exchange used for imports. The implication is that not all flow variables can be used, and that stock variables should be used to account for stock adjustment processes. We thus use exports as a share of debt to incorporate both flow and stock variables. If domestic and foreign-currency-denominated wealth are perfect substitutes, domestic currency could be freely used to pay for debt or imports, and consid- eration of foreign exchange constraints would be unnecessary. Many develop- ing countries have some possibilities of substitution; Latin American debt swap operations are an example. But most African countries cannot readily substitute domestic for foreign assets, and therefore we treat domestic and foreign ex- change wealth as differentiated assets. Under this assumption one can solve the following problem: maximize wel- fare as an increasing and concave function of imported goods subject to an intertemporal budget constraint. Real income and domestic prices are also included in the objective function on the assumption that they increase the effectiveness of imports in promoting welfare. The constraint restricts the pres- ent value of future trade deficits to be equal to the current stock of foreign exchange assets, defined as the level of foreign exchange reserves less the stock of net international debt. Solution of this problem gives the import level as a function of net foreign exchange assets, current and expected future exports, the terms of trade, real income or total real expenditures, and the real domestic price of imports. Table 3 presents long-term import elasticities estimated from this model, including the same variables as in equation 9, plus the ratio of exports to debt. The greater the value of that ratio, the greater the foreign exchange available and the smaller the pressure to further restrict imports. As this would suggest, 204 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 all the coefficients associated with this ratio are positive, although those for Nigeria and Zaire are not significant. The very low coefficient for Nigeria may be explained by Nigeria's large potential oil export earning, which would mean that Nigeria is not likely to have suffered serious foreign exchange restrictions during this period. The lack of significance of the coefficient for Zaire cannot so easily be explained. Because data on national debt are only available since 1970, the coefficients reported in tables 1 and 3 are not strictly comparable. Nonetheless, a review of the 1966-69 period for these countries would not suggest any important changes in the macroeconomic relationships that we are analyzing. This is supported by the statistical significance of the estimates here; only thirteen of the thirty-five are not significant at the 5 percent level, which is almost the same share as in table 1. The inclusion of the foreign exchange availability does not appear to reduce the significance of the coefficients associated with GDP and the absorption-GDP level, even in countries in which the export-debt coef- ficient is significant. III. CONCLUSION We have examined here the effects of several primary macroeconomic factors on imports. Within them, two key characteristics commonly altered by struc- tural adjustment programs are real absorption and the exchange rate. In addi- Table 3. Estimated Long-run Import Elasticities Considering Foreign Exchange Constraints, 1970-86 Absorption/ Exchange Dollar import Export- Country Real GDP, Y GDP, A/Y rate, e price, pt debt ratio R2 C6te d'Ivoire 0.88 0.85 -0.72 -0.43 0.22 0.97 (3.53) (3.35) (-10.81) (-2.71) (2.82) Kenya 0.88 2.53 -0.69 -0.62 0.21 0.94 (4.78) (10.21) (-6.03) (-7.85) (5.75) Madagascar 1.78 2.37 -0.51 -0.56 0.09 0.82 (2.03) (2.58) (-1.67) (-2.85) (2.24) Nigeria 3.06 3.19 -0.86 -1.02 0.01 0.97 (14.51) (6.89) (-11.08) (-3.23) (0.00) Tanzania 1.58 3.09 -0.38 -0.58 0.23 0.67 (1.63) (3.40) (-1.46) (-2.45) (2.64) Zaire 1.92 1.09 -0.10 -0.71 0.06 0.95 (1.52) (0.97) (-0.55) (-5.29) (0.94) Zambia -3.00 2.30 0.54 0.46 0.56 0.95 (-1.48) (1.14) (1.12) (0.39) (2.25) Median' 1.23b 2.45 -0.51 -0.58 0.22 Note: t-statistics are in parentheses. All estimates are ordinary least squares except those for Mada- gascar, which used a Cochrane-Orcutt procedure to correct for autocorrelation. a. Excludes estimates with "wrong" signs. b. Excludes Nigeria because of the extremely high elasticity obtained. Source: World Bank data (see appendix). L6pez and Thomas 205 tion to assessing the relationship of the level of absorption and imports, the relationships when the composition of absorption varies were also estimated. Exchange rate adjustments were also included, first independently, and then using a proxy for foreign exchange availability to explicitly account for import restrictions that would limit the extent to which imports could directly respond to changes in the exchange rate. We estimated the determinants of imports for seven Sub-Saharan African countries over the 1966-86 period. Our findings correspond with those of previous works regarding the positive effect of income on imports. Policies often adopted in structural adjustment programs also are important, however. First, on average, a 1 percent reduction in the absorption-GDP ratio is associ- ated with a 2 percent decline in imports. The composition of absorption also turns out to be relevant. Although the import-intensity of the three components varies widely across the seven countries, a reduction in the share of government consumption can be seen to be significantly related to a decrease in imports in some countries. Second, although the exchange rate, as would be expected, is an important determinant of imports, it is less so than is the dollar import price. The reason is the presence of import quotas set in dollar terms that limit the total dollar value of a subset of imports. Thus the quantity of imports for the restricted goods can only rise if the dollar price falls. On average, a 1 percent depreciation of the exchange rate is associated with a 0.5 percent decline in imports, holding the absorption-GDP ratio constant. Third, we have seen evidence that foreign exchange availability is positively associated with the total import quota values set by some African countries. We thus used the ratio of exports to debt as an indicator of foreign exchange availability, and found it positively and significantly associated with imports. The demand-related variables (real GDP and absorption-GDP) remain significant when the foreign exchange availability variable is included. This suggests that aggregate imports are jointly determined by both foreign exchange and demand factors. These findings suggest that adjustment programs that include exchange rate depreciation and aggregate demand reduction are likely to reduce imports. Although import liberalization may raise import-GDP ratios, simultaneous re- duction in absorption and depreciation of the exchange rate may lower imports during the adjustment process. In Sub-Saharan Africa, as for most developing countries, adjustment programs must account for these interactions in addition to consideration of demand and foreign exchange constraints if the effects on imports are to be understood and anticipated. APPENDIX: DATA DEFINITIONS AND SOURCES Real merchandise imports: Nominal dollar merchandise imports deflated by a country-specific dollar import price index (1980 = 1). Merchandise imports 206 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 are taken from the World Bank ANDREX tapes (International Economics De- partment-IECSE), line "CP, IMP, TOTAL." This series follows very closely merchandise imports (Standard International Trade Classification--SITc-cate- gories 0-9) as reported by the United Nations Conference on Trade and Devel- opment (UNCTAD). However, as the latter series covers up to 1983 only, we used the IECSE series instead. Dollar import price: The country-specific dollar import price indexes are from the ANDREX tapes, line "PT, IMP, TOTAL." These price indexes are weighted averages of subindexes for five categories: manufactures, food, non- food agriculture, metals and minerals, and fuels, defined according to one-, two-, and three-digit SITC codes (see Moran and Park 1986). Real GDP: Nominal home-currency GDP deflated by the implicit home-cur- rency GDP deflator (1980 = 1). Both series are from the World Bank Economic and Social Data (BESD) tapes. Real absorption to GDP: Nominal home-currency absorption relative to nom- inal home-currency GDP deflated by home-currency GDP deflator (1980s). Ab- sorption is the sum of private consumption, public consumption, and invest- ment, and it is taken from BESD. Real exchange rate: A nominal exchange rate index (1980 = 1) deflated by the implicit GDP deflator (1980 = 1). The nominal exchange rate is a bilateral rate that indicates the number of domestic currency units traded per U.S. dollar. It is taken from the BESD and is the same as the period-average exchange rate in IMF (various years) line "rf." Ratio of exports to net debt: Nominal dollar exports of goods and services relative to nominal dollar net debt deflated by a country-specific dollar import price index (1980 = 1). Exports are taken from ANDREX, line "CR, EXP, GS." Net debt is defined as the stock of debt exclusive of foreign exchange reserves. The stock of debt is taken from BESD, and it refers to long-term public and publicly guaranteed disbursed outstanding debt. Foreign exchange reserves, including gold holdings, are taken from the data tapes for International Finan- cial Statistics (IMF various years) line 1..D. Real exports of goods and services: Nominal dollar exports of goods and services. Investment share in GDP: Nominal home-currency investment relative to nominal home-currency GDP. Investment incorporates fixed investment as well as changes in stocks. Both investment and GDP are from BESD. REFERENCES Agbonyitor, Alberto. 1986. "Import Elasticities of African Countries: Some Empirical Evidence." Economic Analysis and Projections Department, Country Analysis and Projections Division Working Paper 4, World Bank. Washington, D.C. Processed. Dornbusch, Rudiger. 1983. "Real Interest Rates, Home Goods, and Optiimal External Borrowing." Journal of Political Economy 11: 141-53. L6pez and Thomas 207 Goldstein, Morris, and Mohsin Khan. 1985. "Income and Price Effects in Foreign Trade." In R. W. Jones and P. B. Kenen, eds., Handbook of International Economics, vol. 2. Amsterdam: Elsevier Science Publishers. Halevi, Nadav. 1988. "Trade Liberalization and Adjustment Lending." World Bank Country Economics Department. Washington, D.C. Processed. Hemphill, William. 1974. "The Effects of Foreign Exchange Receipts on Imports of Less Developed Countries." International Monetary Fund Staff Papers 21: 637-77. International Monetary Fund. Various years. International Financial Statistics Year- book. Washington, D.C. Leamer, Edward, and Robert Stern. 1970. Quantitative International Economics. Bos- ton: Allyn and Bacon. L6pez, Ram6n, and Vinod Thomas. 1988. "Imports and Growth in Africa." World Bank Policy, Planning, and Research Working Paper 20, Washington, D.C. Processed. Magee, Stephen. 1975. "Prices, Income and Foreign Trade: A Survey of Recent Eco- nomic Studies." In P. Kenen, ed., International Trade and Finance: Frontiers for Research. Cambridge, England: Cambridge University Press. Moran, Cristian. 1989. "Imports under a Foreign Exchange Constraint." World Bank Economic Review 3, no. 1 (May): 279-95. Moran, Cristian, and J. G. Park. 1986. "Merchandise Trade Deflators for Developing Countries." World Bank Economic Analysis and Projections Department Working Paper 1986-7. Washington, D.C. Processed. Pritchett, Lant. 1987. "Import Demand Elasticities: Estimates and Determinants." World Bank Economic Analysis and Projections Department Working Paper No. 1987-4, Washington, D.C. Processed. Sachs, Jeffrey. 1981. "The Current Account and Macroeconomic Adjustment in the 1970s." Brookings Papers in Economic Activity 10: 201-68. . 1982. "The Current Account in the Macroeconomic Adjustment Process." Scandinavian Journal of Economics 84: 147-59. Sundararajan, V. 1986. "Exchange Rate Versus Credit Policy." Journal of Development Economics 20: 75-105. Winters, L. A. 1987. "An Empirical Intertemporal Model of Developing Countries' Imports." Weltwirtschaftliches Archiv 123: 58-80. THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2: 209-233 Voluntary Export Restraints and Resource Allocation in Exporting Countries Jaime de Melo and L. Alan Winters This article analyzes the resource implications of voluntary export restraints (vERS) for exporting countries. A simple analytical method is used to demonstrate that, by reducing the marginal revenue of its factors of production, a VER causes an industry in the exporting country to contract, and that the efficiency losses from a VER depend on the ease with which sales can be diverted from the restricted toward the unrestricted markets. The method is applied to test the effects of the U.S. Orderly Marketing Agreement (OmA) for producers of leather footwear in the Republic of Korea during the period 1977-81. We estimate that the marginal revenue product of factors employed in leatherfootwear declined by as much as 9 percent because of the OMA, an estimate that is corroborated by inspection of time series on output, employment, and wages of the Korean footwear sector. This implies that there was pressure on the Korean footwear industry to contract as a result of the OMA. Most previous investigators of the effects of voluntary export restraints have been concerned with the welfare costs of such restraints to consumers in im- porting countries. Examples of previous studies include: estimates of quality adjusted welfare costs of VERS on automobiles (Feenstra 1984; Dinopoulos and Kreinin 1988); the claim that, for commodities with little product differentia- tion and low start-up costs (for example, footwear and textiles), VERS are ineffective (Baldwin 1982; Bhagwati 1986); and simulations which show that terms of trade effects are likely to reduce substantially the costs of (gains from) VERS to importing (exporting) countries (Tarr 1987; Trela and Whalley 1988). None of these studies-even those which look into the implications of VERS for exporters-contain investigations into the effects of VERS on resource allocation in exporting countries. This article fills the gap. For the sake of simplicity, we analyze the effects of a VER in two steps. First, Jaime de Melo is an economist at the World Bank and an associate of the Centre for Economic Policy Research, London. L. Alan Winters is a professor of economics at the University of Birmingham and an associate of the Centre for Economic Policy Research. The authors thank three referees, Wendy Takacs, Paul Brenton, and Taeho Bark, participants at the Centro Interuniversitario di Studi Teorici per la Politica Economica-Centre for Economic Policy Research Conference in Venice, May 1988 and at the European Research Workshop in International Trade in Bergen in July 1989 for helpful comments on an earlier draft. They also thank Maria Ameal and Alexander Pfaff for logistical support. (C 1990 The International Bank for Reconstruction and Development / THE WORLD BANK. 209 210 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 we consider a very-short-run equilibrium in which the size of the exporting industry is fixed. Second, we consider the pressures at this equilibrium for the industry to contract. When the size of the exporting industry is fixed, firms respond to the VER by reallocating sales away from the restricted to some other (unrestricted) export market. Such diversions of exports are assumed to have increasing opportunity costs, which implies that the commodities; exported to the two markets are not perfect substitutes in production-perhaps because they have different characteristics. Two effects of a VER are distingLlished. First, the sales revenues of the exporting firms may rise or fall depending on the elasticities of demand in the restricted and unrestricted markets. Second, there is a misallocation of resources. The VER moves the composition of the export industry's output to a point on the production possibilities curve where its slope is no longer equal to the ratio of the two export prices; this constitutes alloca- tive inefficiency. In the second step of our analysis, the industry is allowed to shrink in response to the VER. The VER reduces the marginal revenue products of the exporting industry's factors of production, and the industry will contract by an amount dependent on the elasticities of factor supply. We proceed as follows. The impact of a VER on resource allocation in exporting countries is analyzed in section I. In section II, we present an estimate of the effect of the U.S. Orderly Marketing Agreement (OMA) for nonrubber footwear imports on the Korean leather footwear industry during the period 1977 to 1981. A model is developed to estimate econometrically the slope of the production possibilities curve for the footwear industry. This helps to determine the new point reached on the production possibilities curve as a consequence of the OMA. The difference between the slope of the curve at the new point and the new price ratio is used to estimate the fall in the marginal revenue product for factors employed in the footwear industry. We find that the marginal revenue product may have fallen by as much as 9 percent because of the OMA. (The econometric techniques used in section II are described more fully in appendix A.) In section III we present the results of a simulation exercise. Combining our econometric estimates with alternative assumptions about output demand and factor supply elasticities gives illustrative estimates of the likely welfare effects of the OMA. (The model used to derive the welfare effects is described in appendix B.) Some concluding remarks are made in section IV. I. REAL RESOURCE ALLOCATION IMPLICATIONS OF A VER In this section, we present an analysis of the sales revenue effects and the resource allocation implications for an exporting country entering a VER. We demonstrate under fairly representative conditions that a VER will reduce the marginal revenue product of factor inputs in the affected industry,, and hence de Melo and Winters 211 reduce the size of the industry. To simplify the exposition, we assume that all output is produced by firms which are identical and perfectly competitive in all product and factor markets. The output is sold in one of two foreign markets. A VER restricts exports to one of these markets while those to the second market remain unrestricted. In addition, while we assume that individual ex- porting firms are price takers in each export market, the industry as a whole faces downward-sloping demand curves in each market. The effects of a VER are presented in two steps. Initially it is assumed that industry size is fixed and that there are costs to diverting sales from the re- stricted toward the unrestricted market. With this assumption, it is a simple matter to analyze the revenue and distortionary implications of the VER. In the second step, we show that a VER will reduce the marginal revenue product of factors employed in the industry, which is likely to lead the industry to con- tract. This two-step discussion of the effects of a VER is expositionally convenient and corresponds to the two-stage assumption about firm decisions adopted in the econometric work of section II. We show in appendix B that the independ- ence of (weak separability between) input decisions and sales decisions is not fundamental for the results that are established here. The assumption that there are increasing marginal costs to the diversion of sales from the restricted toward the unrestricted market may be interpreted as indicating the short-run effects of a VER in an industry characterized by product differentiation. In the short run, where the quantities of factors employed in the industry are fixed, a shift in production between outputs which use factors in different proportions will result in increasing marginal costs. Moreover, if one of the outputs has a specific factor which is in relatively inelastic supply- for example, a particular labor skill-the mere fact that the outputs are differ- entiated is sufficient to produce an increasing marginal rate of transformation between the two products. The above assumptions allow us to represent our model diagramatically. In figure 1, foreign export demands in the restricted (A) and unrestricted (B) markets are represented in quadrants I and II, respectively, while quadrant III depicts the substitution possibilities facing the industry as it reallocates produc- tion between sales for market A and those for market B. It is obtained by the aggregation of the substitution possibilities facing each exporter. The bowed- out shape of the export transformation curve G(XA, XB) = X reflects the as- sumption of increasing costs of production associated with changing the prod- uct mix. Assuming fixed aggregate footwear production X, successive incre- ments in sales to market B impose increasing resource costs in terms of larger and larger decreases in export sales to market A. Quadrant IV is bisected by the 450 line which translates export sales to A from quadrant III to quadrant I. The unrestricted allocation, At, is represented by the price-quantity pairs (PA, XA) and (PB, X*), where superscript asterisks are used to denote the 212 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 Figure 1. Equilibrium Allocation of Sales P4 -t---4- …- T = PB | I XA fA(PA) XI f"(P11)~ ~ ~ ~ ~ ~~~I Xy =ft( PB) | I Il I X,~~~~~~~~ G(A X)= \ \, AA/P I \ P P KB ~ ~ ~ ~ ~ ~ ~ ~ ~ A"~ ~~X APB/PA de Melo and Winters 213 unrestricted equilibrium. The slope of the export transformation curve is given by - MRTA,B = - _. dXB GA where G, = aG/aXi > 0; i = A, B, indicate positive opportunity costs; and the bowed-out transformation curve reflects the fact that opportunity costs increase with output.' The equilibrium for a competitive industry requires that G-A PA (1 ) A~ = A B B and Pi = O*G, where 0* is the marginal cost of producing a unit of X. In figure 1, the equilibrium condition (equation 1) is represented by the tangency of the export transformation curve and the price line, P, at the unrestricted equilib- rium, A*. Now impose a VER which restricts exports to A to the level XA. The restricted allocation is represented by the new price-quantity pairs (iA, XA) and (P, XB) where superscript bars indicate the restricted equilibrium. We decompose the effects of the VER into two parts: a revenue effect, which reflects the way the VER affects the industry's revenue from exporting, and an allocative distortion, which arises as the economy is obliged to produce a non-optimal mix of outputs. Consider the revenue effect. We have assumed that the industry as a whole comprises a large number of atomistic firms and therefore cannot behave like a discriminating monopolist. The presence of unappropriated monopoly rents means that it is possible that the VER will raise industry revenues. Let EA and E, denote the elasticity of demand in the restricted and unrestricted markets, respectively, and assume EA, eB = 0. A departure from the free trade equilibrium will raise revenues if marginal sales revenue is higher in B (unrestricted) than in A (restricted), that is, if PB(1 - 1 /eB) > PA(1 - 1 /EA). Thus if the elasticity of demand is much greater in the unrestricted than in the restricted market, a VER may push sales allocation toward that which would be selected by a discrimi- nating monopolist.2 In the case described above, the VER has the same type of effect as an optimum export tax. That exporting governments do not preempt the imposi- tion of an OMA by imposing export taxes of their own volition probably reflects their uncertainty about the long-run elasticities of demand they face, a fear of 1. Increasing marginal costs requires GAA, GBB > 0 and GAAGBB - GA 2 0. 2. While we do not wish to stress the empirical relevance of this possibility, it has been pointed out in previous theoretical discussions of the effects of VERS in non-competitive markets. It is interesting to speculate that the two-tier quota allocation system used in Korea (and elsewhere) implies that greater sales toward non-restricted markets may have the objective of revenue maximization. For further analysis see Bark and de Melo (1988). 214 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 retaliation from importing governments, and a fear of reprimand from inter- national organizations. Of course, once importing governments have estab- lished OMAS, exporting governments are likely to try to exploit their potential market power through alternative quota allocation schemes. Note that if the exporting industry had initially acted as a discriminating monopolist or faced optimum export taxes, it would already have been maximizing its profits, and hence could only have suffered from the imposition of a VER. Consider next the allocative distortions caused by the VER. Because we have assumed at this stage that firms maximize revenue for a given level of X, which is equivalent to maximizing profits, the new allocation must lie on the same export transformation curve as the old, and given the exogenous value of XA, the chosen point will be R. At this new equilibrium, the equality of relative prices to the marginal rate of transformation no longer holds. The relative price of A has been forced up by the constraint on sales so that it is tangent to the tranformation curve at A, but the marginal rate of transformation (MRTAB) has fallen to R as the relative output of A has fallen (see figure 1). Hence - > LI, whereas G < 0. B B~G This violation entails a well-understood distortion cost, which is independent of whether total sales revenues have increased after the imposition of the VER. For producers to choose point R voluntarily, they would have to face the relative price line (PA'/PB), which equals the marginal rate of transformation at R. PA' is known as the virtual price and, as is clear from figure 1, it implies a relative price less than both the unconstrained price (P/IP*) and the actual price (PAIPB). We now turn to the second step of the discussion and show that the VER will create an incentive for the industry to contract. With production and allocation decisions separable, input mixes (which depend on factor prices) are indepen- dent of output mixes (which depend on output prices); hence, given fixed relative factor prices, we can construct a composite factor, Z, with wage W. This allows us to characterize production in terms of the aggregate output index as X = X(Z) and, assuming constant returns to scale, we can select units such that X -Z. Hence we can assess the effects of a VER on the size of the industry as measured by aggregate input Z. In an unrestricted equilibrium, sales are allocated between markets in such a way that the marginal revenue product of a factor devoted to producing goods for market A equals that of the factor if it were used to produce for market B. We may write the condition for an unconstrained equilibrium as follows: dRA dRB, dR (2) d ds= dR = W (2) ~~~~~dZ -dZ dZ where RA is the revenue derived in market A, and so forth, W is the cost of factor Z and, in our earlier notation, de Melo and Winters 215 dRi P-* dZ G<- To assess the effects of the VER on the size of the industry, we need to consider whether the marginal revenue product of Z is increased or decreased by the VER. This can be done entirely in terms of market B. Under free trade, the marginal revenue products are equal across markets, whereas under the binding VER, only market B can accommodate marginal sales. The marginal revenue product in B falls as more and more sales are switched to B, but while the VER is binding, the marginal revenue product in A is zero. Hence marginal product for a given Z falls. Constrained revenue maximization by the represen- tative firm implies that in the new equilibrium: dR3 PBdB - e (3) ~~dZ GB Since the VER redirects sales from market A to market B, it drives up the costs of producing for market B (it reduces the marginal product of Z), because GBB > 0. Thus, even if demand for B is perfectly elastic-that is, PB = PB-the value of the marginal product of Z is reduced by the VER. If demand is less than perfectly elastic, that is, PB < PB, the effect is exacerbated by the drop in price. Hence independent of whether the VER increases or decreases industry revenue, it always reduces the marginal revenue product schedule of its factor inputs. Unless factor supply is entirely inelastic, a falling MRP schedule will cause the industry to reduce both its output and input levels. The new long-term equilibrium could be represented in figure 1. The trans- formation curve shrinks inward and is tangential to a price ratio with slope between (PA/PB) and (PAX'5B) at a point with output (XA, X's) where X'B > X*. Point N represents one possibility. A more direct approach to analyzing the long run is directly in terms of the market for the composite factor. Figure 2 illustrates three cases of interest. The VER causes the marginal revenue product curve to fall, say from MRP' to MRP2. In the very short run, in which factor inputs cannot be changed, the input level remains at Z-, but there is a loss of factor rents and therefore welfare to factor owners equal to area AEFC. Thus there are no output losses but large distor- tion and resource costs. This, of course, is the implication derived by consid- ering the allocation part of the model alone. In the opposite case, if the foot- wear industry (in addition to each of its individual firms) faces an infinitely elastic supply of Z, then the new input level is given by Z; there are large output losses, but no distortionary resource costs because factors shift to other industries in which they are equally productive. Under these circumstances, 0 = 0*, and the shift in the marginal revenue product curve is accommodated by output contraction alone. In another case, the industry faces an upward- sloping supply curve for factors, the final input level is Z, which entails a smaller contraction but imposes welfare losses equal to the area AEDB. Now 216 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 Figure 2. Three Cases for the Market for the Composite Factor Z Where the VER Causes the MRP Curve to Shift Down Input price, W \ = W(Z) A E 4 W= W B-- - = MRP | I I ~~MRP' Z Z Z2- Aggregate input, Z Note: Z*is the short-run case where Z is fixed at Zn. Z is the case where there is an infinirely elastic supply of Z. Z is the case where there is an upward-sloping supply curve for factors. 0 < 0-, and there is an efficiency loss, but it is smaller than that implied in the very short run in which industry factor use is fixed. In a more complete exposition of the implications of a VER, the number of affected markets would be increased and the two-step approach relaxed. The empirical analysis in section II addresses the more general case in 'which sales are allocated to one restricted and two unrestricted markets, and we discuss below how the analysis would be modified to increase the number of unres- tricted markets. As for the two-step approach, de Melo and Winters (1990) show for the general case of a two-output, one-input technology that spillover to unrestricted markets and output contraction will occur unless there is a very strong positive relation between output destined for one market and the costs of producing for others. (Their findings are also discussed briefly in appendix B.) Because marginal production costs for each market are likely to show only small interdependencies, it is unlikely that the qualitative predictions of the above analysis would differ in a more general setup. Because we have only two markets, we have been able to establish the contractionary effect of a VER by considering the marginal revenue product in each market directly. With more markets, as in the empirical application be- low, it is more convenient to use an alternative approach, derived fiom Neary and Roberts (1980). These authors show that a constrained equilibrium can be expressed as an unconstrained equilibrium at a different set of prices. These de Melo and Winters 217 are the virtual prices, mentioned earlier, which are simply the set of prices at which, given the overall level of activity, actual quantities supplied in the constrained equilibrium would be supplied voluntarily. For unconstrained mar- kets, virtual prices are equal to actual prices. Referring back to figure 1, the quantities given by R would be voluntarily supplied at the set of prices (PX, PB3), whereas actual prices in the constrained case are represented by (PA, PB)- For any unconstrained equilibrium, the value of the marginal product arising from an extra unit of aggregate input Z, optimally allocated, is a positive function of the set of prices (P1, . . . , P.); where n - 1 is the number of unrestricted markets. For a constrained equilibrium, the Neary-Roberts results allow us to calculate the value of the marginal product schedule by evaluating the same function at virtual prices (PV). The effect of a binding VER in market A is to reduce Pv below its actual price. Because, in unconstrained markets, actual and virtual prices are equal, the reduction in PA means that at least some prices fall and none rise and that therefore the value of the marginal product falls. Although it is not possible, without knowledge of the factor supply curve, to determine equilibrium values for prices and output, this approach suggests that the VER puts downward pressure on output. This is the procedure used in section II to measure the shift in the marginal revenue product schedule; the equivalent of distance EF in figure 2. II. ESTIMATING THE REDUCTION IN FACTOR DEMAND: KOREAN LEATHER FOOTWEAR In this section, we present estimates of the effect of the United States's Orderly Marketing Agreement (OMA) on nonrubber footwear on the demand for the factors employed in Korean leather footwear production. We use the simple model described in the previous section: it is assumed that the Korean footwear industry produces an aggregate quantity of footwear using a single composite factor of production, and subsequently allocates this aggregate amount to one of three markets according to a constant elasticity of transfor- mation (CET) allocation function. This simple function allows us to analyze, albeit indirectly, the efficiency implications of the OMA without access to spe- cific data on the allocation of factor inputs to sales in each market. The crucial parameter in our model is the elasticity of transformation, which is a measure of the extent to which production may be shifted between outputs destined for different markets. In terms of figure 1, it measures responsiveness to changes in the output mix along a given export transformation curve. With this parameter value, it is possible to predict the actual output mix in the OMA period, and the difference between actual and virtual prices during the period. From the latter we can calculate the extent of the shift in the marginal revenue product schedule for the composite factor. 218 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 Using quarterly data for the period from the first quarter of 1975 (1) through the fourth quarter of 1986 (IV), we estimate the elasticity of transformation between supplies of leather footwear destined for three markets-the United States, the unconstrained European Community, (which comprises France, the Federal Republic of Germany, Italy, and the Netherlands), and the rest of the world. The United States imposed the OMA on Korean exports of nonrubber footwear between the third quarter of 1977 and the second quarter of 1981, inclusive (although the evidence suggests that the restrictions on Korea ceased to be binding by mid-1980; Aw and Roberts 1986). Hence, the observations corresponding to this period are not included in the estimation period. The unconstrained European Community group, according to Hamilton (1989), imposed no quantitative restrictions on Korean footwear exports over our sample period. Some of the countries in the rest of the world did have import restrictions on footwear, but they may reasonably be treated as unconstrained overall. Following the model described in section I, individual Korean exporters are assumed to be price takers. They seek to maximize revenues subject to a CET transformation function, which relates the quantities of each type of export footwear to an overall index of output (input); that is, max X p,X, subject to E aXj =y where Xi is exports to market i at price pi, X is the index of aggregate output, andy > 1. Writing p = 1 /(-y - 1) for the elasticity of transformation, standard manip- ulation allows us to express the share of market i in total exports as (see Hickman and Lau 1973): (4) si = al(p,/p)P; i = 1, 2, 3 where si is the share of i in the volume of exports, si = X,/E Xj, is a fixed-weight price index, and ai = a--. Although firms are price takers, the exporting industry is not, so the aggre- gate Korean export price is potentially endogenous to our model. Both this fact and the possibility of there being errors in variables suggest the need for more robust methods of estimation than are possible for nonlinear systerns of equa- tions with complex error structures. We decided, therefore, to linearize the model about a base period (see Hickman and Lau 1973). If we set prices to unity in the base period (second quarter of 1984), introduce a time-trend with de Melo and Winters 219 value zero in the base period, and add seasonal factors and a lagged dependent variable, we obtain (5) y, p cO (p, (Pit- p) + 7,t + E 6D,ik + Xyi,-, + ui, k where yi, =S - s ?to is the deviation of i's share from its base value, a', P- E czxp°, is a base-weighted price index, t is a time trend incremented by one per quarter, E oiDk are seasonal effects for quarter k, k = 1, 3, 4 where the dummy for the second quarter has been suppressed because the base period is a second quarter, X, represents dynamic effects on the share of market i of its own lagged values, and ui, are stochastic errors. Because yi, sums to zero over i in each time period, equation 5 for one of the three markets must be dropped in estimation. We dropped the equation for the unconstrained European Community. We then estimated the remaining equa- tions by a procedure described in appendix A. The final equation is given in table 1. The estimated elasticity of transfor- mation is perhaps a little low, given the anecdotal evidence that exists on the degree of competition and product substitution-homogeneity in world footwear markets; but it is a fairly robust result. Moreover, two other pieces of evidence suggest that Korean exports to different markets are imperfect substitutes. First, the unit values of Korean exports to different markets differ by up to 50 Table l. The Allocation Function for Korean Leather Footwear Exports Variable Estimated coefficient Standard error p 1.311 0.756 YR -0.0013 0.0009 ,Yu 0.0024 0.0010 3RI 0.074 0.022 3ul -0.076 0.021 3R3 0.060 0.019 6U3 -0.063 0.018 3R4 0.058 0.019 6U4 -0.060 0.020 Xl 0.401 0.102 0.137 R2 Rest of the world 0.80 United States 0.88 Unconstrained European Community 0.75 Long-run elasticity of transformation' 2.19 Note: The subscript R refers to the rest of the world and the U to the United States; r is a first-stage estimate of the autocorrelation parameter. a. Calculated as pl(1 - X,). Source: de Melo and Winters (1989, appendix). 220 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 Table 2. Estimated Effects of the OMA on Korean Leather Footwear Exports to the United States, 1977-80 Proportionate Actual minus difference between Change predicted actual and virtual in aggregate share price price index Year and quarter (1) (2) (3) 1977,3 -0.176 -0.121 -0.095 1977,4 -0.024 -0.017 -0.013 1978,1 -0.076 -0.050 -0.039 1978,2 -0.159 -0.103 -0.081 1978,3 -0.098 -0.056 -0.044 1978,4 -0.093 -0.051 -0.040 1979,1 -0.208 -0.096 -0.076 1979,2 -0.188 -0.081 -0.063 1979,3 -0.284 -0.110 -0.087 1979,4 -0.159 -0.065 -0.051 1980,1 -0.052 -0.023 -0.018 1980,2 -0.092 -0.040 -0.031 Source: table 1 and authors' calculations. percent, which suggests that there may indeed be genuine product heterogene- ity. Second, the estimates in table 1 display dramatically different seasonal patterns-with the allocation of shares between the United States and the rest of the world switching by over 10 percentage points with the season. Table 2 explores the effects of the omA on Korean exports to the United States more closely. On the basis of our estimates, we can predict the share of exports to the United States in the OMA period if quantities had been uncon- strained at the actual price. Column 1 reports the difference between the actual and predicted shares during the OMA period. It is consistently negative, which suggests that the omA was binding, although it shows signs of weakening during 1980. Column 2 approximates the proportionate difference between the virtual and actual prices of exports to the United States. Because the actual U.S. share falls short of that predicted by the export allocation model, the virtual price for the United States is below the actual price by as much as 12 percent (in third quarter of 1977). Thus the OMA may be seen to have had an effect equivalent to reducing the price of exports to the United States by 5-12 percent below actual levels with no compensating price rises in other markets. This makes it clear that the OMA put pressure on the Korean footwear industry to contract. The extent of the contractionary pressure can be calculated as the difference in the aggregate price index evaluated at actual and virtual prices. This calcu- lation, reported in column 3 of table 2, is a linear approximation to the change in the marginal return to the aggregate factor in the leather footwear sector. It shows that the marginal revenue product of the factors of production in the de Melo and Winters 221 Figure 3. Index of Employment and Output in Footwear Relative to All Manufacturing (1980 = 1) 1.8 A 1.7 - 1.6 - 1.5 / 1.4 - lOutpu 1.3 - 1.0 0.9 /O 0.8 Emp oyment 0.7- 0.6 0.5 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 leather footwear industry declined by as much as 9 percent because of the OMA.3 The econometric estimates strongly suggest that the Korean footwear indus- try would have contracted during the period of the OMA. This prediction is supported by time series data on output, employment, and wages of the Korean footwear sector. These data are illustrated in figure 3, which reports footwear output and employment relative to the corresponding series for the entire manufacturing sector: although Korean industry generally contracted over this period (see appendix table A-i), the footwear sector suffered more than aver- age. Output and employment in the footwear industry peaked in 1978, one year after the signing of the OMA agreement. This peak is later than predicted by our model, but it is not out of line with the detailed account of the OMA given by Yoffie (1983). He remarks that Korean producers went to considerable lengths to negotiate the OMA in a fashion that allowed extended periods of adjustment. Thus it is quite conceivable that output and employment remained high into 1978. 3. Comparisons with free trade rather than actual prices would reduce these estimates somewhat. In the absence of an estimated elasticity of demand, however, the former are not calculable, but the higher the elasticity of demand the smaller the difference. In our simulation results below, we use free trade prices. 222 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 One of the causes of the Korean slump was rising real wages over the period 1977-79, which might have harmed labor-intensive industries such as foot- wear. In fact, wages in footwear fell during 1978 relative to those in other labor-intensive sectors such as textiles, apparel, and leather products. Thus although it cannot be entirely ruled out that the time series on footwear reflect only economywide phenomena, there is some evidence of particular hardship in the footwear sector. Certainly we find nothing to refute our prediction that the OMA caused a contraction of Korea's footwear sector. III. ILLUSTRATIVE WELFARE CALCULATIONS The results above confirm our hypothesis that a VER leads to output contrac- tion and has adverse effects on efficiency if the factors employed in the industry are not available to it in perfectly elastic supply. However, as discussed in section I, a VER also results in revenue effects which may either reinforce or counteract the efficiency costs. (It was demonstrated in section I that the latter, in the long run, lead the industry to contract and to reduce factor incomes.) The combined long-run effect is captured in the calculation of the welfare effect, which is estimated by simulation techniques in this section. Welfare effects are calculated by adding the gains in or losses of profits to firms and the loss of income to factors. The simulation model is used to provide rough orders of magnitude of the potential welfare effects of a VER, using the U.S. OMA on Korean exports of leather footwear as a reference case. The calibrated simulations are based generally on the model presented in section I, with constant foreign price elasticities of demand; a CET function describing sales allocation; and a con- stant elasticity of factor supply. (Further details are given in appendix B.) For the reference-case calculations, we use our estimate of the elasticity of transformation from section II to measure the ease with which exporters may divert sales from restricted to unrestricted markets and complement it with rough estimates of factor supply elasticities and price elasticities of export demand. Our estimates of the price elasticity of export demand are consistent with the range of 0.5-1.0 reported by Pearson (1983, p. 78) and Goldstein and Khan (1985). Welfare is measured by the sum of profits and factor incomes, and the change in welfare is expressed as a proportion of variable factor (Z) income before the VER. The change in factor demand from the restricted industry affects the wages throughout the markets in which the factors are traded and the contraction of the restricted industry drives down factor rewards both for itself and for any other industry using the same factors. Hence a given wage reduction has a larger impact on welfare, the larger the stock of factors affected. This effect is represented crudely in the welfare calculations by L, the size of the total factor stock affected relative to the initial size of the affected industry, Z. A value of L = 1 implies that initially only the VER-affected industry employs the factor de Melo and Winters 223 concerned; a value of 5 implies that the VER industry originally employed 20 percent of the relevant factor stock. Because L will vary depending on the industry under consideration, we give calculations for cases where the market for the variable factor Z is 1, 5, or 10 times the initial allocation to the industry under the VER. Illustrative calculations for a range of elasticities are given in table 3. For all calculations, it is assumed that there is a 10 percent reduction in the volume of sales to the restricted market, where the initial share of exports to the restricted market is 42 percent of total exports (a figure corresponding to the leather footwear case). Before examining the results for the different elasticities shown as the five cases of the table, in which several elasticities are varied simultane- ously, we describe briefly the effects of varying elasticities one by one and compare the results with those in case 1, where all elasticities are unity. In the case of unitary export demand elasticities, there are no sales revenue effects, so it is easy to isolate the effects of varying supply elasticities. The more difficult it is to reallocate the existing volume of production, the higher the Table 3. Illustrative Welfare Calculations Case Item 1 2 3 4 5 Value of elasticity Type of elasticity Price elasticity of restricted demand, EA 1.0 0.3 0.5 1.0 2.0 Price elasticity of unrestricted demand, e, 1.0 0.6 1.0 2.0 4.0 Elasticity of transformation, p 1.0 0.5 1.5 1.5 3.0 Elasticity of factor supply, e, 1.0 0.5 2.0 2.0 3.0 Result of simulation Variable Outputb -5.0 -3.9 -4.3 -4.5 -4.1 Sales revenue' 0.0 11.6 4.6 0.0 -2.1 Factor wageb -4.0 -7.7 -2.2 -2.3 -1.4 Welfarec 1 6.2 19.7 10.2 5.5 2.7 5 -0.5 6.9 6.5 1.7 0.4 10 -8.9 -9.1 1.9 -3.0 -2.5 Note: Notation is given in appendix B, equation, B-2. Unrestricted equilibrium: XA = 100; XB 140; Pi = 1.00; Z = 100. a. Size of market for Z in relation to initial allocation of Z in industry subject to VER. b. Percentage change. c. Change in income expressed as a share of initial variable factor income (see appendix B, equation B-13). These changes are shown for three values of the size of the total factor stock affected relative to the initial size of the affected industry. Source: table 1 and authors' calculations. 224 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 efficiency cost of a given VER because the adjustment comes from output contraction rather than from sales reallocation. Likewise, as explained in sec- tion I, the higher the elasticity of factor supply, the lower the efficiency costs of a VER. Indeed, a similar variation (around unity) of the elasticity of factor supply has more of an effect on efficiency than will an equal variation of the elasticity of transformation. Cases 2 through 5 give estimates of the welfare effects of a VER for low, medium, and high sets of elasticities. The results in case 4 may be viewed as best guess calculations. In the case of a VER, there is a net loss if the market for Z is large. But if the market for Z is small (relative to the initial allocation of Z to the industry), there is a net gain from the VER despite the negative efficiency effects because of their relatively smaller weight in the welfare calcu- lation. The same is true for the case where all the elasticities are low (case 2). The larger efficiency costs are offset by the larger revenue gains. It is also noteworthy that the simulated decreases in the marginal revenue product of Z (represented by the factor wage row of table 3) are similar in magnitude to the range reported from the econometric estimates in column 3 of table 2. Finally, in case 5, with higher demand elasticities, the revenue effect becomes negative, which implies larger welfare losses. Thus, if demand elasticities are not too low and supply elasticities are not too high, a VER is likely to lead to a welfare loss.' Although the net effects of the VER are fairly small in table 3, the gross effects are significant. The VER increases profits for those with the right to export to the restricted markets but harms all other agents. In particular, labor is likely to lose from industrial-country protection, an ironic result in light of the fact that protection is often advocated as a means of protecting workers. If so, protection should be seen as a means of protecting rich industrial country workers at the expense of workers in poor developing countries. IV. CONCLUSIONS This article has presented a simple model to analyze the revenue and effi- ciency effects of a VER at the industry level. Motivated by the evidence that developing countries often have limited success in switching sales toward unres- tricted markets, we have analyzed the impact on the exporting industry. In order to do this, we have separated revenue effects arising from sales realloca- tion toward unrestricted markets from efficiency effects arising from output contraction. The analytical discussion of the effects of a VER was then corroborated with an application to the U.S. OMA agreement with Korean exporters of leather 4. The elasticity of factor supply (es) is not independent of the relative size of the industry in the factor market (L). For example, for an industry like footwear, Ef is likely to be in the range of 2-4 and L perhaps 10 or more, whereas in textiles, the corresponding pair would be e, in the range of O.S-2.0 and L around 5. de Melo and Winters 225 footwear. The econometric estimates indicate both a limited ability to switch sales toward unrestricted markets and a sharp fall in the marginal revenue product of factors employed in the Korean leather footwear industry during the period where the OMA was in effect. Combined with extraneous estimates of export demand and factor supply elasticities, illustrative welfare calculations suggest that the oMA may well have resulted in a welfare loss to exporting countries, especially if demand elasticities are relatively elastic and supply re- sponses relatively inelastic. APPENDIX A This appendix presents a detailed description of the econometric model of export allocation introduced in section II and an account of its estimation. We assume that exporters are price takers and that they seek to maximize their revenues subject to a CET transformation function relating the quantities of each type of footwear export to an overall index of output (input). Their objective is (A-1) max E piX, subject to l|aiXj = X xi i where X, is exports to market i, at price pi, X is the index of aggregate output, andy> 1. Standard manipulations (see Armington 1969) produce supply functions for the individual markets, (A-2) Xi = ai-P(pi/p*)P9 where p* is the dual CET price index of X given by: (A-3) p [* a1p(+P)11 and is the (negative of the) elasticity of transformation between exports for any pair of markets; p > 0. Further manipulation (see Hickman and Lau 1973) transforms equation A-2 into the more convenient form: Xi= aipi[ aU'P{' x or (A-4) Si= ai(pi/p)P + ui 226 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 where X -- X. is a simple aggregation of exports, si is the share of i in the volume of exports, p 01c,P'I is a fixed-weight price index, a, = a,-e, and ui is a stochastic component added at this stage for estimation purposes. To facilitate the treatment of simultaneity and errors in variables, equation A-4 is linearized about a base period. We used the second quarter of 1984 as the base because it lay well outside the period of possible rationing, and yet was relatively central to our sample of unrationed observations. Subsequent tests suggested that the choice of base period affects the results slightly, but not sufficiently to disturb the qualitative conclusions in the text. If we set prices to unity in the base period, introduce a time trend with value zero in the base, and add seasonal factors and dynamics, the linearization gives (A-5) Y = P aAA(p, - PI) + q,t + E SikDk + ±lyit-I + X44Yit-4 + u,a where y,, si, - ao is the deviation of i's share from its base value, ao, p, ae up X°P. is a base-weighted price index, t is a time trend incremented by one per quarter, E bikD, are seasonal effects for quarter k, k = 1, 3, 4 where the dummy for quarter two has been suppressed because the base period is a second quarter, and Xk, X4 represent dynamic effects on the share of market i, felt through lags of itself. Adding up requires that , c; = 1 and that , 6 k = S _y = j U = 0 for all k and t. The first condition is satisfied autom'atically; the latter, by dropping the equation for the unconstrained European Community. Usually with this procedure, the final estimates are invariant with respect to the equation dropped, but because of the methods required by the errors in the variables, that is the use of instrumental variables, this is not so. In practice, however, the choice made very little difference. Adding up also requires that, unless the errors are characterized by full vector autoregression, the dynamic structure implied by the lags X, and X4 must be common to all commodities. The lag X therefore appears in both equations as a cross-equation restriction. The use of lagged dependent variables may be justified on several grounds-for example, partial adjustment of price expecta- tions, as in Hickman and Lau (1973), or habit formation. For systems of sum- constrained equations, it represents by far the most convenient approach to dynamic generalization. The choice of lags 1 and 4 to capture the dynamics was made a priori on the basis of previous experience with quarterly data sets. Equation A-5 may be stacked over i and written in matrix form: de Melo and Winters 227 __ O(p,-p) t O D, O D, 0 D, 0 Ly, L 4y _ _ u (A-6) j 12 (A-6) Y2 = O(P2P) t O D, O D 0 D4 Ly2 L4y2 2l1 623 814 824 xi X24 where L is the lag operator and all the roman letters denote (n x 1) vectors, where n is the number of observations. Ignoring the errors in variables, equation A-6 may be simply estimated allow- ing for autocorrelation and cross-equation correlations. Following Parks (1967), we first estimate A-5 for each commodity separately, and calculate a single first-order autocorrelation coefficient. (The serial correlation adjustment factor must be common to all equations if the system is to add up). Transforming the data appropriately, we then reestimate by commodity to calculate E(fiiu->j), where the ui are the errors from the transformed equations. Finally, using these variances and covariances, we transform the data again to estimate equation A-6 by generalized least squares. To allow for the simultaneity and the errors in variables, we use instrumental variable estimation. The technique presumes that the instrumental variables are correlated with the true values of the variables in equation A-5 but not their errors of observation. If this is true, our estimates are consistent and asymptot- ically efficient. Instruments were drawn from both the importing countries (industrial production, the wholesale price index for manufactures, and the exchange rate vis-a-vis the dollar) in order to reflect demand factors, and from Korea (the per unit value of manufactured exports, the index of industrial production, and the dollar exchange rate) to reflect broad supply-side phenom- ena. Whenever Ly and L 4 are included in the equation, the instrumental varia- bles are also included in lagged form. Finally, the genuinely exogenous varia- bles in A-6-that is, Di and.t-are also included in the set of instruments. The estimation method is based on Aigner and others (1984). We assume that there exists a true relation equivalent to equation A-6 but without errors in variables, and which may be written as: (A-7) y = W + u where t represents the true value of X and y and u have their usual definitions. The relation between the true t and the observed X independent data is (A-8) X= t + V 228 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 There also exists a set of relations between the k true independent variables and the I indicator (instrumental) variables (Z). (A-9) Z = uF' + A. The error terms V and A are assumed to be independently normally distributed with zero means and also to be independent of t. The covariances of the rows of V and A (v, and at) are given by Q and 0 respectively and the variance of u by aJ2. The true independent variables are assumed to have an expected scaled cross-product matrix, K = Em-'t'S, where m = 2n is the number of rows in the matrixes y, X, Z, (, V, and A. Following Aigner and others, we can write the various covariance matrices ,J= Em-I'J, for I, J = X, y, Z as (A-lOa) E;= a2 + 13K3 (A-lOb) ;xy =Kg (A-10c) EZy = rKf (A-iOd) zz= K + Q (A-1Oe) Zzx = rK (A-lOf) EZZ = rKr' + 0 Equations (A-lOc) and (A-1Oe) yield Ezy = EzxO from which, multiplying both sides by ExE-1 and substituting sample values S,1 for population values EIJ we obtain (A-11) = zz = [X' Z (Z' Z)-')X' Z] (Z' Z)-1 Z' y. , is multivariately normally distributed with asymptotic variance (A-12) var (p3) = (a2 + 3' 0) (S'xSz-S,x)-1 which is the minimum variance bound that can be derived by linear methods. We approximate A-12 below by substituting , for j and using A-10 to express the first bracket in terms of observables. System A-10 presumes that the errors are independently and identically dis- tributed, but in our case we need to allow for the presence of autocorrelation and the fact that E(u1tu2,) : 0 where u, and u2 are subvectors of u referring to the first and second equations. In fact, however, these modifications make virtually no difference to the estimator. Taking the latter first, partitioning all variables in A-7 to A-9 conformably with u, and u2, versions of A-10 may be de Melo and Winters 229 derived for all combinations of yi, Xi and Zi, i = 1, 2. If the only change in assumption is that E(uiuj) = aij, i #- j, only A-lOa is changed; it becomes (A-1 Oa' ) = ± + )3K In all other equations, the partitioned covariances are the same as the unparti- tioned ones in A-10. This means that the same instrumental estimation method may be applied to a set of first-stage estimators to derive the a , which are then used to transform all the observable data into the form assumed in the main stage just described. Provided that the estimates of aij are consistent, the asymp- totic properties of the final estimates are unchanged. A similar approach is taken to the autocorrelation. The variance estimate (equation A-12) may be used to conduct statistical inference on the coefficients. The validity of a set of q linear constraints, Q,B = r, may be explored by means of the test statistic (Qf - r)' [Q Var (j)Q' 1-' (QB - r) which is distributed X2 under the null hypothesis (see Amemiya 1985). This test suggested that it was acceptable to set X4 = 0 in the final equation. (The data were collected and prepared by Taeho Bark and Paul Brenton from Korean Customs data publications. They are fully described in the appendix of de Melo and Winters 1989. In terms of the final classifications in table 2 of that source, leather footwear is defined here as headings 6402.1000- 6402.4900.) Appendix Table A-1. The Korean Footwear Sector, 1974-83 Footwear as a percentage of total Footwear manufacturing Year Employment' Outputb Wages' Employment OutpUtd Wages 1974 6,600 33 303 0.518 85.3 85.3 1975 11,000 53 364 0.788 114.5 77.9 1976 14,200 74 493 0.84 121.3 82.6 1977 19,800 108 657 1.046 147.1 85.1 1978 26,000 159 846 1.249 174.9 79.3 1979 22,000 107 1,136 1.055 105.0 81.1 1980 22,700 100 1,366 1.127 100.0 79.2 1981 26,000 111 1,538 1.293 97.9 74.8 1982 36,500 113 1,809 1.77 94.6 78.4 1983 40,500 122 2,025 1.86 87.8 80.2 a. Person-years. b. Index: 1980 = 100. c. In thousands of won per year. d. Ratio of index numbers: 1980 = 100. Source: U.N. Industrial Development Organization data. 230 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 APPENDIX B General Model and Welfare Calculations General model. Consider the general case of firms in perfect competition in which the allocation and input decisions are made jointly.5 In this case, weak separability is not imposed on the allocation and production decisions. Tech- nology is represented by a one-input, two-output production function. Let variable factor requirements, Z, allocated to the restricted (XA) and unrestricted (XB) markets be given by: (B-1) Z = G(XA, XB) where Z is the quantity used of the composite factor, Gi is aZIaX, > 0, and G is homothetic and homogeneous of degree r < 1. Under the assumption of profit maximization, de Melo and Winters (1990) show that the imposition of a VER on sales to A(XA < X*), leads to the following expressions for output (B-2) and for national welfare (B-3). 1 dZ GB;- GBB,+ GBGA]_ (B-2) G1A dXA C G GK, XGeBB Ze GB (B-3) dXA I -B [eB + N1] GA where eA, EB, 6N < 0 are respectively the elasticities of demand for A, B, and the variable factor Z with respect to the wage in other sectors using Z, E, > 0 is the elasticity of supply of Z. From B-2, it is clear that a VER in A will most likely lead the industry to contract if one assumes increasing marginal costs, that is, Gii > 0, and if one recognizes the constraints imposed by the second- order conditions for profit maximization. Only very strong (and implausible) interactions between A and B leading to a large positive value for GAB would lead the industry to expand. Hence, a VER is likely to lead the industry to contract. From B-3, the change in national welfare defined as the sum of industry profits and factor payments is determined by an allocation component which measures whether switching sales from A to B raises revenue, and a size com- ponent which measures whether switching factors across sectors is beneficial. Welfare calculations. The welfare calculations in section III come from a numerical application of the model presented in section 1, with constant elastic- ity of demand curves (equations B-4 and B-5); a CET function to allocate sales between the restricted and unrestricted markets A and B (equation B-6); and a 5. This appendix draws on de Melo and Winters (1990). de Melo and Winters 231 constant elasticity of supply function for the factor, Z (equation B-11). An unrestricted equilibrium is described by the following set of equations: (B-4) XA = AAPA A EA > O (B-5) XB =AB PB eB > 0 (B-6) X = AC(AAXA + UBXB) P = 1/7-1 ; 'y > 0 (B-7) PA 0 AO +lCp aA(XA/gX)/P (B-8) PB 0AC CB(XB1/ X) (B-9) X = XS (B-10) XS = ASZEZ (B-11) z = Az PlZ; e. > 0 (B-12) PZ = Aso- where Ai, (i e A, B, C, Z, S) are normalizing constants determined by calibra- tion, that is, constants calculated so that the set of equations describing the model is satisfied for the initial levels of prices and quantities. In the free-trade equilibrium, industry profits, 7r, are zero as sales revenue equals payments to Z, Pzz. With the VER, XA is fixed at XA < XA and the first-order condition for the allocation to the restricted market (B-7) is dropped. As explained in section I, as a result of the VER, 0 > 0* (unleSS e, = The welfare measure is: (B-13) AW = W, - W* = (Ar + APZL)/PZZ where L is a scalar indicating the size of the industry in the market for Z. The calculations in section III are obtained by solving the model represented by equations B-4 to B-12 for an unrestricted equilibrium and for a restricted equilibrium with XA = 0.9 XA. When elasticities are varied, the A, parameters are recalibrated so as to start from the same initial unrestricted values for prices and quantities. REFERENCES Aigner, D. J., C. Hsia, Arie Kapteyn, and Tom Wansbeek. 1984. "Latent Variable Models in Econometrics." In Zvi Griliches and M. D. Intriligator, eds., Handbook of Econometrics, vol. 2. Amsterdam: North-Holland. Amemiya, Takeshi 1985. Advanced Econometrics. Oxford: Blackwell. 232 THE WORLD BANK ECONOMIC REVIEW, VOL. 4, NO. 2 Armington, P. S. 1969. "A Theory of Demand for Products Distinguished by Place of Production." International Monetary Fund Staff Papers 16: 159-76. Aw, B. Y., and M. S. Roberts. 1986. "Measuring Quality Change in Quota-Constrained Import Markets." Journal of International Economics 21: 45-60. Baldwin, Robert 1982. "The Inefficacy of Trade Policy." Essays in International Finance no. 150. Princeton, N.J.: Princeton University. Processed. Bark, Taeho, and Jaime de Melo. 1988. "Export Quota Allocations, Export Earnings, and Market Diversification." World Bank Economic Review 2, no. 3 (September): 341-48. Bhagwati, Jagdish. 1986. "VERS, Quid Pro Quo DFIs, and VIEs: Political Economy Theoretic Analyses." International Economic Journal 1: 1-14. Dinopoulos, Elias, and Mordechaie Kreinin. 1988. "Effects of the U.S.-Japan Auto VER on European Prices and U.S. Welfare." Review of Economics and Statistics 70: 484- 91. Feenstra, Robert 1985. "Automobile Prices and Protection: The U.S.-Japan Trade Restraint." Journal of Policy Modelling 7: 49-68. Goldstein, Morris, and Mohsin Khan. 1985. "Income and Price Effects in Foreign Trade." In Ronald Jones and Peter Kenen, eds., Handbook of International Econo- mies, vol. 2. Amsterdam: North-Holland. Hamilton, Carl 1989. "The Political Economy of Transient 'New' Protectionism." Welt- wirtschaftliches Archiv 125: 522-46. Hickman, B. G., and L. J. Lau. 1973. "Elasticities of Substitution and Export Demands in a World Trade Model." European Economic Review 4: 347-80. Hufbauer, G. C., D. T. Berliner, and K. A. Elliot. 1986. Trade Protection in the United States: 31 Case Studies. Washington, D.C.: Institute for International Economics. Melo, Jaime de, and L. Alan Winters. 1989. "Price and Quality Effects of VERs Revis- ited: Case Study of Korean Footwear Exports." Policy, Planning, and Research Discussion Paper 216, World Bank. Washington, D.C. Processed. . 1990. "Do Exporters Gain from vERs?" Discussion Paper 383. London Center for Economic Policy Research. London, England. Processed. Neary, Peter, and K. Roberts. 1980. "The Theory of Household Behavior Under Rationing." European Economic Review 13: 25-42. Nogues, Julio, Andrzej Olechowski, and L. Alan Winters. 1986. "The Extent of Non- tariff Barriers to Industrial Countries' Imports." World Bank Economic Review 1, no. 1 (September): 181-99. Parks, R. W. 1967. "Efficient Estimation of a System of Regression Equations when Disturbances Are Both Serially and Contemporaneously Correlated." Journal of the American Statistical Association 62: 500-09. Pearson, C. 1983. Emergency Protection in the Footwear Industry. London: Trade Policy Research Center. Tarr, David G. 1987. "Effects of Restraining Steel Exports from the Republic of Korea and Other Countries to the United States and the European Economic Community." World Bank Economic Review 1, no. 3 (May): 397-418. Trela, Irene, and John Whalley. 1988. Do Developing Countries Lose from the MFA? National Bureau of Economic Research Working Paper 2618. Cambridge, Mass. de Melo and Winters 233 Winters, L. A., and P. A. Brenton. 1988. "Non-Tariff Barriers to International Trade: UK Restrictions on Imports of Leather Footwear from Eastern Europe." University College of North Wales, Bangor. Processed. Yoffie, David 1983. Power and Protectionism. New York: Columbia University Press. Subscription Coupon If you are not already a subscriber to The World Bank Economic Review, you may begin a subscription by completing and returning this coupon to World Bank Publications. You may also use this coupon to subscribe to The World Bank Research Observer. Written for noneconomists and students,the Observer provides over- views of key issues in development economics research and assumes no economics training on the part of the reader. It is also helpful to economists seeking concise surveys of work outside their own specialties. Subscriptions to both The Economic Review and The Research Observer are available on a complimentary basis to readers vwith mailing addresses in non-OECD countries. For readers with mailing addresses in OECD countries, discounts are available for three-year subscriptions and subscriptions to both journals. Indicate below the type of subscription desired. Please complete both sides of this subscription coupon. For Readers with Mailing Addresses in Non-OECD Countries: Please begin my complimentary subscription to [1 The World Bank Economic Review 2 The World Bank Research Observer (check either or both boxes). For Readers with Mailing Addresses in OECI) Countries: (Check appropriate box). One Year Three Years Individuals Institutions tnIdinduls Institutions The World BaikEconomc $25 0 $45 2 5 65 1: $125 O Review (three issues per year) The World BankResearch $20 2 $35 [ $ 50 0 $ 95 2 Observer (two issues per year) BothJournals $40 2 $70 2 $100 2 $175 C Check method ofpayment 2 Enclosed is my check for US$ , payabie to World Bank Publications. Chage my l VISA 2 MasterCard O American Express Credit Card Number Signature Expiration Date OPlease bill me. (Institutional orders only.) Purchase order number Please complete both sides of this subscription coupon. Send to: World Bank Publications Box 7247-8619 Philadelphia, PA 19170-8619, U.S.A. Pleaseprint clearly Name Title Organization name Organization address __ County Postal code Please check only one. [ 01 International, regional organization (including field offices), that is, wx Ho, UNODP [1 02 Central banks, finance and planning ministries U- 03 Other governmnent agencies (including embassies) E 04 Nongovernmental organizations (N(;os) and other associations U 05 Research institutions- U 06 University faculty Ul 07 Libraries U 08 Bookstores U 09 Commercial banks and other financial institutions (including savings and loan associations, security dealers, and so forth) U1 10 Business - manufacturing (goods) Where both apply check "goods" U1 11 Business - consulting (services) U 12 Student Cl 13 News media Ul RO General economics: theory, history, systems U RI Economic growth: development, planning, fluctuations U] R2 Economic statistics Ul R3 Monetary and fiscal theory and institutions U] R4 International economics U] R5 Administration, business finance, marketing, accounting U:1 R6 Industrial organization; technological change; industry studies U R7 Agriculture, natural resources, rural development Ul R8 Labor, population and human resources U R9 Welfare programs; consumer economics; urban and regional economics By exploring the effectiveness of rapid small-farm growtth in reducing p)overtV -- antd the scope 01 social action to improve tlhat effctiveness -- thlis splendid book shoiuld capture the attention of policynakers and researclhers. --Michael Lipton, Institute of Development Studies, University of Sussex The Great Ascent The Rural Poor in South Asia Inderjit Singh This new book challenges the widely held view that the rapid growth of small farms in South Asia has done the poor more harm than good. By critically reviewing recent academic literature and the World Bank's operational experience in Bangladesh, India, and Pakistan, Inderit Singh shows that agricultural growth has benefited all classes of the poor. He concludes that the "great ascent" from poverty to a more materially rewarding life has begun. The report examines in detail a variety of programs intended to help the poor. It evaluates agricultural research, extension, and training activities, as well as programs in dairying, poultry farming, commercial fishing, and forestry. Mr. Singh also examines the failure of land reforns to reduce poverty. Publishedfor the World Bank by The Johns Hopkins University Press 472 pages / ISBN 0-8018-3954-8 / US$39.95 / Order Stock #43954 Of Related Interest by Inderjit Singh Land and Labor in South Asia 244 pages / ISBN 0-8213-1129-8 / US$12.95 / Order Stock #20033 Small Farmers in South Asia Their Characteristics, Productivity, and Efficiency 82 pages / ISBN 0-8213-1090-9 / US$6.95 / Order Stock #20031 Tenancy in South Asia 128 pages / ISBN 0-8213-1091-7 / US$10.95 / Order Stock #20032 To order The Great Ascent or other publications by Inderjit Singh, list title, stock number, and price on the coupon that follows. Women in Development A Progress Report on the World Bank Initiative 40 pages / ISBN 0-8213-1623-0 / US$4.95 / Order Stock #11623 Agricultural Extension for Women Farmers in Africa Katrine A. Saito and C. Jean Weidemann 74 pages / ISBN 0-8213-1657-5 / US$6.95 / Order Stock #11657 Helping Women Improve Nutrition in the Developing World Beating the Zero Sum Game Judith S. McGuire ahd Barry M. Popkin 102 pages / ISBN 0-8213-1415-7 / US$7.95 / Order Stock #11415 Household Food Security and the Role of Women J. Price Gittinger, with the collaboration of Sidney Chernick, Nadine R. Horenstein, and Katrine Saito 54 pages / ISBN 0-8213-1627-3 / US$5.95 / Order Stock #11627 Bangladesh Strategies for Enhancing the Role of Women in Economic Development 182 pages / ISBN 0-8213-1630-3 / US$10.95 / Order Stock #11630 The milost con1ven1ien1t an1d user-friencdly software ever released by the World Bank World Bank Data on Diskette An increasing number of data collections published by the Bank are now available on diskette for use in personal computers. Using the Bank's newly developed Socioeconomic Time-series Access and Retrieval System (STARS), the data can be easily exported to spreadsheet and word processing programs. Each diskette package includes double-density 5 1/4" diskettes and 3 1/2" diskettes for use on PCs with a hard disk and MS-DOS versions 2.0 or higher. Also included is a User's Guide to the software and an English paperback edition of the corresponding book. World Tables 1989-90 Data on Diskette ISBN 0-8213-1531-5 / US$70.00 / Stock #11531 The most detailed collection of economic data published by the Bank. Social Indicators of Development 1989 Data on Diskette ISBN 0-8213-1530-7 / US$95.00 / Stock #11530 Reorganized and expanded to provide a clearer picture of the social effects of economic development. World Debt Tables 1989-90 Data on Diskette ISBN 0-8213-1533-1 / US$95.00 / Stock #11533 Compiled from information gathered through the World Bank's exclusive Debtor Reporting System. World Development Indicators 1990 Data on Diskette ISBN 0-8213-1532-3 / US$95.00 / Stock #11532 Drawn from World Development Report, the Bank's most widely circulated statistical tables on economic and social development. African Economic and Financial Data on Diskette ISBN 0-8213-1529-3 / US$95.00 / Stock #11529 Designed to track the performance of development programs and aid flows in the region. To order any of the data collections on diskette, list title, stock number, and price on the coupon that follows. Use this coupon to order any of the titles listed on the previous pages. Title Stock Number Price Subtotal US$ Shipping and handling US$ 3.50 Airmail surcharge outside USA (US$6.00 per copy) US$ Total US$ Check your method of payment. Orders from individuals must be accompanied by payment or credit card information. Ask local distributors about forms of payment acceptable to them. OI Enclosed is my check payable to World Bank Publications in U.S. dollars. Charge my OII VISA Lii MasterCard LI American Express These credit cards are accepted for orders addressed to Philadelphia. (Check with your local distributor about acceptance of credit cards in your country.) Credit card account nurnber Expiration date Signature Institutional customers only: FJ Send invoice and reference purchase order no. Please print clearly Name Title Firm Address City State Postal Code_ Country Telephone Customers in the United States: Detach this form and mail to World Bank Publications, P.O. Box 7247-8619, Philadelphia, PA 19170-8619, U.S.A. Customers outside the United States: Contact World Bank Publications, 1818 H Street, NW, Washington, DC 20433, U.S.A. for a list of World Bank publications distributors. Write to the local distributor in your area for information on prices and terms of payment. Coupons received in Washington from countries with authorized distributors will be returned to the customer. Customers in countries with no authorized distributor, return your coupon to World Bank Publications, P.O. Box 7247-8619, Philadelphia, PA 1917048619, U.S.A. Coming in the next issue of THE WORLD BANK ECONOMIC REVIEW September 1990 Volume 4, Number 3 A symposium on imperfect information and rural credit markets, including articles on ... * The costs of informal lending (with case material from Pakistan) by Irfan Aleem * Institutional and informal credit agencies (with case material from India) by Clive Bell * Public and private roles (with case material from Thailand) by Ammar Siamwalla and others * Rural credit as insurance (with case material from Nigeria) by Christopher Udry * Peer monitoring by Joseph E. Stiglitz Introduction by Karla Hoff and Joseph E. Stiglitz