World Batik Reprint Series: Number 431 F. Desmond McCarthy, Lance Taylor and Cyrus Talati Trade Patterns in Developing Countries, 1964-82 Reprinted by permission of the Jotrnal of Development Economics, Vol. 27 (1987), pp. 5-39. Published by the Elsevier Science Publishers, B.V. (North-Holland), Amsterdam. Journal of Development Economics 27 (1987) 5-39. North-Holland TRADE PATTERNS IN DEVELOPING COUNTRIES, 1964-82* F. Desmond McCARTHY, Lance TAYLOR and Cyrus TALATI Massachusetts Institute of Technology, Cambridge, MA 02139, USA Results are presented from a regression analysis of import and export shares of GDP, for 14 commodity and service categories in 55 developing countries over the period 1964-82. Explanatory variables include GNP per capita, population, domestic capacity utilization, the effective exchange rate, and OECD real GNP (for exports). Among other findings it is shown that countries' trade patterns are not closely related to their growth performance, and that developing countries on average run surpluses of exports over non-capital goods imports on merchandise trade. This 'current' commodity surplus is required to offset imports of capital goods and a structural deficit on factor and non-factor service trade. 1. Introduction Carlos Diaz covered the international waterfront on behalf of the develop- ing countries. From the whorls of geometric trade theory through clever empirical studies to the political economy of bargains with transnationals and the intricacies of debt, he was always on the scene with something recent and relevant to say. He used to talk of boyhood visits to Havana harbor and wishing to be a trader on a ship. He didn't take that route, but a better one, showing how trade and capital movements give with one hand, take with the other, and strongly affect possibilities for economic change throughout the Third World. Carlos's judgments at midpassage about trade appear in his 1978 paper. There, he partly associated himself with an old strand of thought more appealing to development than trade specialists, noting that '...a country's participation in foreign trade ... as measured by, say, the ratio of imports to GNP, has more to do with its per capita income, population size, and natural resource endowment than with its social system or domestic income distribution. One is tempted to talk about iron coefficients' (p. 119). Charac- teristically, he then went on to reasons why in certain policy situations, the *Revised version of World Bank Staff Working Paper no. 642. Comments by a referee and many seminar listeners are appreciated, especially those by Richard Cooper, Albert Fishlow, Gerry Helleiner, Moshe Syrquin and Martin Williams. The authors would like to thank K. Meyers and J. Rozanski for their help and M. Landivar who prepared the final manuscript. The rese.;ch was supported by the World Bank, but the views and interpretations herein are those of the authors and should not be attributed to the World Bank, to its affiliated organizations, or to any individual acting on their behalf. 0304-3878/87/$3.50 ( 1987, Elsevier Science Publishers B.V. (North-Holland) 6 F.D. McCarthy et al., Trade patterns in developing countries, 1964-82 coefficients might not be so die cast after all. Nonetheless, he thought that by and large the 'patterns' analysis of Kuznets, Chenery, and others applied to trade. Whether it does or not is the question we take up here. Countries at similar lcvels of income and size have long been known to share patterns of economic structure and resource allocation - this is the central message of the patterns studies.' Pace Carlos, however, how well this generalization applies to foreign trade has remained an open question. Largely because of lack of data, the relevant tests have not been run at a disaggregated level, especially for developing countries. More important, perhaps, is the widely held view that trade policy (and presumably observed trading behavior) should make a difference for country 'performance'.2 If this is the case, then uniform patterns may not be observed - faster-growing countries, for example, may have different levels or composition of trade than their slower neighbors. In this study, we examine these somewhat contrasting views, as well as other plausible explanations for trade patterns, e.g., time period, regional effects, oil importer or exporter status, level of capacity utilization, OECD growth, and countries' real effective exchange rates. Three principal questions are addressed. How do levels and composition of trade respond to the 'structural' variables per capita income and population? Are trade patterns related to the need on the part of developing countries to raise net foreign exchange resources to pay for non-substitutible imports of capital goods? How does trade respond to more policy-related variables like the real effective exchange rate and OECD growth? In this initial attempt to examine these issues, we use the simple statistical technique of ordinary least squares regression - multiple correlation in another guise. More sophisticated analysis requires accepted, prior theory, but this is not available in the area of trade. There are clearly multiple directions of causality among trade and other variables, and the number of competing theoretical explanations for trade patterns is embarrassingly large.3 For these reasons, we rely on panel regressions and sample splits to 'For summary statements, see Kuznets (1966) and Chenery and Syrquin (1975). Kuznets (1964) goes into more detail on trade issues and Strout (1985) presents results complementary to ours. 2Balassa (1983, 1985) presents the case with vigor. 3Of popular theories, 'old' ones like Heckscher-Ohlin focus on factor endowments and by extension the importance of natural resources and the national reserve of skills in determining trade patterns; 'middle-aged' models like the product cycle stress aspects of technology. Recent work by Amsden (1986) and Havrylyshyn (1983) has provided empirical evidence to account for differences in the direction of developing country trade while Balassa (1983) has examined the effects of policies on trade between developed and developing countries. Only the middle-aged Linder (1961) theory and newer proposals by Krugman (1980), Lancaster (1980) and others explicitly bring in market size and income-related tastes. As we will see shortly, such considerations go some way toward explaining trade patterns, especially for imports. Stewart (1982) gives a useful survey of the various trade theories. F.D. McCarth) et al., Trade patterns in derrv.,pitig countries, 1964-82 7 illustrate potentially important effects. The data base pools a sample of 55 developing countries, 14 categories of goods and service trade (plus partial aggregations) for the period 1961-82. Some statistical imprecision may result from pooling time series and cross section data. Better estimation in principle could be achieved by adopting a precise specification of covariances of error terms by country, commodity, and year. But again, there is no clear theoretical guidance about how this should be done. Specifications for import and export regressions are slightly different. The first is Si, = xci + fli In GNPPC, +;',i In POP, + 6 In (*)+ SSIn EER, + ri,, ( I) where t = 1964 to 1982, Si,t= share of sector i imports in GDP in period t, In GNPPC, = log of real gross national product per capita in period t, In POP, = log of population in period t, In = log of capacity utilization in period t, In EERt = log of the real effective exchange rate in period t, 'it = error term associated with the ith sector in period t. The export regressions wvere based on the following equation: Si, = xi + f3i In GNPPCt + ',i In POP, + ci In (OECD-GDP), + iln * + jlnEERr+ ci, (2) in which Si, now stands for an export share of GDP and ln(OECD-GDP), is the log of real OECD gross domestic product at time t. The trade shares are current U.S. dollar imports or exports divided by current dollar GDP. Potential real GDP is estimated as a linear envelope of actual levels over tinme, and capacity utilization (Y' Y*) is the ratio of actual to potential output.4 4Commodity trade information comes from the United Nations data base. Services trade and overall balance of payments numbers are from the International Monetary Fund. Non-factor service trade was calculated as a residual to bring these two sets of data into consistency. It includes errors due to the CIF to FOB conversion (IMF imports are given FOB), timing and residual discrepancies. GDP, GNP. and population data are from the World Bank, and real OECD GDP from the OECD Economic Outlo&k no. 34, December, 1983. Constant dollar data are from 1974, except the real effective exchange rate, which is 1976-78 based. Real effective exchange rates are import and exported weighted. and were calculated for all countries but Mali which had insufTicient data. 8 F.D. McCarthiy et al., Trade patterns in developing countries, 1964-82 T he commodity and service trade categories for which these equations were estimated appear in table 1. Categories 1 through 4 are referred to jointly in what follows as primary products, 5 through 10 as manufactures, 9 and 10 as capital goods, and 14 throughi 16 as servic.es. Note that category 14 (non-factor services) also includes discrepancies between data from the United Nations (the source for data on merchandise trade) and the Inter- national Monetary Fund (services and total current account). The terms in (1) and (2) for GNPPC and POP are standard in patterns studies such as Chenery and Syrquin (1975). Additional variables were included to capture specific forces affecting trade, Imports and exports are often thought to respond to capacity utilization - that variable is also a proxy for cyclical effects. The real effective exchange rate is supposed to affect all trading decisions, and OECD economy activity slhould stimulate devel- oping countries' export demand. The equations are best viewed as reduced forms, but certain coefficients can be given demand or supply interpretations. The specification assures that predicted trade shares from the equations add up to predicted totals (so long as all equations are estimated for the same data set) and the partial elasticity of a trade share with respect to an Table I Commodity and service groups. M1erchandlcise 1. Food: food and live animals, beverages, excluding cereals and cereal preparations. 2. Cereals: cereals and cereal preparations. 3. Non-food agriculture: tobacco, hides and skins, crude and synthetic rubber, lumber, pulp waste paper, and textile fibres. 4. Metals, minerals and fertilizers: crude fertilizer, minerals, and non-ferrous metals. 5. Intermediate manufactures: chemicals, and basic manufacturing. 6. Textiles and apparel: textiles, yarn fabric, clothing and footwear. 7. Automobiles: passenger motor vehicles (excluding buses), passenger motor vehicle chassis, motor vehicle parts, motorcyles and motorcycle parts, and motorized invalid carriages. 8. Other consumer goods: medical products, cos.rnetics, cleaning products, electrical and electroniic appliances and other miscellaneous mranufactured goods. 9. Transport equipment: buses, railway wagons and equipment, boats, ships, aircraft, and associated parts. 10. Electrical and mechanical goods: metal tanks, boxes, industrial equipment, central heating equipment, medical instruments, meters, counters, and other measuring equipment. 11. Oil and other fuels: mineral fuels, etc. 12. Non-oil total: sum of I to 10 above. 13. Total merchandise: sum of I to I1l above. Services 14. Non-factor services: freight, insurance, carriage insurance, passenger, tourism and other travel. 15. Factor income: investment income and workers' remittances. 16. Unrequited transfers: private and official transfers. 17. Total services: total of 14 and 15 above. 18. Grand total: total of merchandise, service sectors and transfers. F.D. McCarthy et al., Trade patterns in developing couintries, 1964-82 9 explanatory variable can be computed as its regression coefficient divided by the sample mean value of the share. One plus the share elasticity with respect to per capita GDP gives an approximation to the elasticity for the level of trade. In what follows, we discuss in section 2 the results from estimating (1) and (2) for the full sample. Thereafter, sections 3 through 7 take up five splits of the sample, used to illustrate differentials in trade patterns over time and place. The splits are: (1) Time period. Three periods - 1964-73, 1974-77, and 1978-82 -- are separated. These correspond to before and after the first oil shock and after the second shock. (2) 'Performance'. There is a positive correlation between the rate of CiDP growth of sample countries and their level of per capita GNP. Those lying above the regression line of growth on level of GNP per capita are classified as 'high' performers and those below the line as 'low'. Trade patterns are examined for these two groups as well as a further split by per capita GDP level (above and below $1,000 in 1982). (3) Oil exporting vs. importing countries. (4) Population size. Two divisions by countries with populations above and below 20 million in 1981. (5) Region. The sample is split into four groups by region - Latin America and the Caribbean, sub-Salharan Africa, Europe, Middle East and North Africa, and the rest of Asia. The detailed classification for each split appears in the appendix. Conclu- sions are presented in section 8. 2. Results from the pooled sample Results from the pooled regressions for import and export shares appear in table 2. Three general observations should be made about goodness of fit. First, most coefficients are well-determined with high t-values. The right- hand side variables are all significant determinants of trade. Secondly, values of R2 - the usual measure of precision in regression - cluster in the range 0.1 and 0.4. These numbers simply mean that trade slhares vary over a certain range across countries. Regressions for trade levels would give higher R2 values, but we thought it more illuminating to concentrate on the share results. Thirdly, standard errors of the regressions typically are between 50 and 100 percent of the relelant mean shares. According to this measure of goodness of fit, most countries' predicted shares lie within a band of width + 100 percent of the mean around the regression line. Great dispersion in trade patterns is not observed; in a broad sense Diaz-Alejandro's iron coefficients apply. Table 2 Mean Constant InGNPPC InPOP In(Y,'Y*) In EER Sector °0 GDP (t) (t) (t) (t) (t) RSQ F SE Least squares estimates oj inmport trade for !he pooled samnple MI Food 2.11 2.19 -0.00 -0.59 -1.99 0.26 0.234 71.19 1.52 (2.37) (-0.07) (-15.11) (-4.03) (1.37) M2 Cereal 1.26 4.72 -0.19 -0.15 -1.65 -0,45 0.068 16.93 1.22 (6.39) (-4.18) (-4.91) (-4.19) (-2.93) M3 Nfoodagr 0.82 -0.93 0.14 0.01 -0.27 0.19 0.037 8.94 0.72 (-2.11) (5.13) (0.31) (-1.14) (2.09) M4 Rawmat 0.34 -0.97 0.12 0.07 0.20 0.09 0.176 49.52 0.30 (-5.30) (10.94) (9.14) (2.10) (2.29) MS Intermed 4.61 1.04 0.07 -0.97 0.32 1.22 0.296 97.76 2.07 (0.82) (0.85) (- 18.35) (0.47) (4.62) M6 Textiles 1.52 5.29 -0.48 -0.66 -1.22 0.14 0.479 213.88 0.99 (8.82) (-12.78) (-26.35) (-3.81) (1.14) M7 Autos 0.89 0.36 -0.01 -0.22 0.43 0.25 0.247 71.81 0.53 (1.09) (-0.33) (-16.24) (2.48) (3.72) M8 Consumer 2.13 2.20 -0.11 -0.80 -0.02 0.56 0.491 224.67 1.09 (3.31) (-2.75) (-28.92) (-0.06) (4.05) M9 Transport 1.36 -0.02 -0.11 -0.30 0.49 0.61 0.130 34.64 1.10 (-0.02) (-2.54) (-10.60) (1.36) (4.37) MIO Elecmech 4.06 -2.11 0.33 -0.56 2.59 1.25 0.185 52.69 1.90 (-1.82) (4.62) (-11.52) (4.21) (5.17) Mll Oii 2.93 -11.04 0.98 -0.72 -3.46 2.10 0.136 31.> 4.10 (-4.42) (6.28) (-6.91) (-2.60) (4.03) M12 Noiltot 19.04 11.64 -0.22 -4.15 -1.06 4.11 0.325 112.21 8.14 (2.35) (-0.72) (-20.01) (-0.40) (3.98) M13 Totmerch 21.97 0.59 0.75 -4.88 -4.52 6.21 0.321 109.82 10.15 (0.10) (1.96) (-18.84) (-1.37) (4.82) M14 NFS 5.35 -1.91 -1.65 -1.75 -3.13 4.74 0.292 66.00 4.54 (-0.46) (-7.97) (-12.37) (-1.82) (5.32) MI5 Facinc 3.25 -3.18 0.21 -0.59 -1.33 1.44 0.176 34.17 2.11 (-1.67) (2.17) (-8.99) (-1.66) (3.47) M16 Transfer 0.96 -0.29 -0.57 -0.40 -0.10 1.26 0.242 51.08 1.32 (-0.24) (-9.37) (-9.78) (-0.21) (4.85) M17 Totservs 9.56 -5.38 -2.01 -2.75 -4.56 7.43 0.332 79.50 6.25 (-0.95) (-7.04) (-14.09) (-1.92) (6.07) M18 Grandtot 31.58 6.63 -2.17 -7.57 -11.55 12.35 0.421 116.52 12.91 (0.57) (-3.68) (- 18.79) (-2.36) (4.88) Mean Constant InGNPPC InPOP In OECD-GDP ln(Y/Y*) InEER Sector a GDP (1) (t) (t) (t) (t) (t) RSQ F SE Least squares estimates of export tradefor the pooled sample XI Food 6.16 -6.22 -1.35 -1.93 1.22 6.92 -0.22 0.256 64.24 4.49 (-0.34) (-7.91) (-16.68) (1.38) (4.71) (-0.36) X2 Cereal 0.31 2.42 0.01 0.08 -0.09 0.60 -0.10 0.030 5.64 0.78 (0.75) (0.39) (3.95) (-0.55) (2.34) (-0.97) X3 Nfoodagr 2.56 44.57 -0.83 -0.40 (-1.67 0.96 0.17 0.073 14.72 3.40 (3.22) (-6.44) (-4.56) (-2.51) (0.87) (0.36) X4 Rawmat 2.06 -50.63 0.13 -0.60 2.91 -2.00 -2.37 0.044 8.57 4.84 (-2.55) (0.70) (-4.79) (3.05) (-1.26) (-3.61) XS Intermed 1.35 - 15.74 0.31 -0.43 0.62 -3.37 0.53 0.172 38.74 1.86 (-2.08) (4.39) (-8.91) (1.70) (-5.54) (2.09) X6 Textiles 0.89 -33.49 0.17 0.11 1.47 -0.43 0.18 0.069 13.82 1.42 (-5.79) (3.11) (3.14) (5.28) (-0.94) (0.95) X7 Autos 0.05 -0.65 0.04 0.01 0.01 -0.26 0.02 0.187 34.29 0.10 (- 1.30) (9.76) (2.59) (0.54) (-6.61) (1.31) X8 Consumer 0.79 -35.24 0.37 -0.06 1.51 -0.86 0.14 0.129 27.71 1.32 (-6.54) (7.39) (-1.79) (5.84) (- 1.99) (0.78) X9 Transport 0.10 -2.06 0.08 0.02 0.05 -0.04 0.10 0.099 18.88 0.24 (-1.97) (8.09) (3.44) (1.05) (-0.46) (2.46) Table 2 (continued) Mean Constant InGNPPC InPOP In OECD-GDP In(Y/Y*) InEER Sector °. GDP (t) (1) (t) (t) (t) (t) RSQ F SE XJO Elecmech 0.28 -12.19 0.19 0.02 0.49 0.11 0.10 0.130 27.39 0.54 (-5.51) (9. 10) (1.44) (4.63) (0.59) (1.35) Xli Oil 3.23 - 120.18 2.54 -0.62 4.73 5.26 1.26 0.092 18.11 8.64 (-3.28) (7.52) (-2.72) (2.68) (1.83) (1.01) X12 Noiltot 14.49 - 108.45 -0.84 -3.16 6.50 1.67 -1.53 0.219 52.41 7.74 (-3.44) (-2.85) (-15.84) (4.28) (0.66) (-1.46) X13 Totmerch 17.64 -251.61 1.59 -3.84 12.27 7.41 -0.05 0.268 65.16 9.44 (-6.28) (4.30) (-.15.43) (6.37) (2.36) (-0.04) X14 NFS 4.74 -77.20 -0.39 --1.53 3.34 -0.85 3.21 0.244 43.27 3.99 (-3.06) (-2.15) (-12.46) (2.77) (-0.59) (3.93) XIs Facinc 1.58 -3.68 -0.14 -0.65 0.36 -3.50 -0.09 0.219 37.45 1.90 (-0.31) (-1.59) (- 11.12) (0.62) (-5.08) (-0.24) X16 Transfer 4.26 -145.25 -2.05 -1.45 6.81 -0.56 3.47 0.130 19.96 7.18 (-3.20) (-6.31.) (-6.57) (3.14) (-0.21) (2.37) X17 Totservs 10.58 -226.13 -2.57 -3.62 10.51 -4.91 6.58 0.215 36.64 10.83 (-3.31) (- 5.26) (-10.90) (3.21) (-1.25) (2.98) X18 Grandtot 27.31 505.49 -2.22 -6.87 24.22 -2.22 6.66 0.416 95.40 11.77 (-6.80) (-4.17) (- 19.04) (6.81) (-0.52) (2.77) F.D. McCarthy et al., Trade patterns in developing countries, 1964-82 13 Some trade categories show more uniformity across countries than others. Among the imports, groups of products close to the consumer (textiles and other consumer goods) have thc best explained shares. Equations for goods traded among producers (cereals, non-food agriculture, and capital goods) are less precise. Intermediate manufactures occupy a position in the middle of this scale. The results suggest that aside from obvious target sectors for substitution, a country's import patterns are likely to be affected by its history of industrial growth, and the particular structure of its interindustry relationships. Among exports, the best-fitting equations are for food agriculture, inter- mediates, automobiles, consumer and capital goods, and non-factor services. One can hypothesize that both resource availabilities and characteristics of partner countries will influence export patterns, and these variables are excluded from the analysis. The usual patterns-of-growth assumptions that uniform evolution of tastes and technology explain structural shifts as the economy develops apply more directly to imports. Table 2 suggests that the share elasticity of cereal trade falls with output per capita while that of non- food agriculture rises, confirming the declining importance of staples as income rises. By contrast, the share elasticity is strongly positive for the electromechanicals category. Now we take Up the roles of the explanatory variables in affecting trade, beginning with per capita GNP and population. The first point to observe is that shares of merchandise trade rise with countries' real incomes. However, the shares of non-oil merchandise and service trade (both imports and exports) decline. Among the commodity categories on the import side, capital goods (groups 9 and 10) have a large and generally increasing shnare. Shares of the primary products are stable, with food products going down slightly and non-food up. The major sectors for import substitution are textilcs and non-factor services; on the other hand factor income payments to the rest of the world rise with per capita GNP. Among exports, shares of both food and non-food agriculture decline with per capita income -- this is a classic pattern. In parallel fiashion, shares of manufactured products go up. There is a sharp reduction in receipts from transfers as per capita GNP is hligher. The negative responses of service and non-oil merchandise trade are strong enough to make total import and export shares drop as countries grow richer. When population is used as a proxy for 'market size', it appears that in the recent period trade expansion (including services) has not substantially outpaced income growth in the Third World.5 5An alternative correction for market size is total GNP. In that specification, coefficients for per capita GNP would equal the difference between the coefficients on In GNPPC and In POP in table 2. On this correction trade shares do rise with income, but one could argue that population is a better measure of a country's long-term structural prospects. Relative population magnitudes do not change very fast. 14 F.D. McCartlhy et al., Trade patterns in deteloping countries, 1964-82 Finance of capital goods imports is a crucial factor affecting growth in developing economies, emphasized in models of the two-gap tradition from Chenery and Bruno (1962) to Bacha (1984). Table 3 uses predicted trade shares from the '.able 2 regressions to shed light on the process at different levels of per capita income (other right-hand variables are held at their mean levels). The increasing GDP share of capital goods imports with rising income (line 13) has already been noted. How are they paid for? TalUe 3 Capital goods import finance at various levels of per capita GNP (predicted shares). $250 $850 $3,000 linports 1. Non-oil merch. 19.19 18.90 18.60 2. Oil 2.39 3.59 4.83 3. Serv. exc. trans. 9.63 7.86 6.05 4. Total 31.21 30.35 29.48 Exports 5. Non-oil merch. 14.85 13.77 12.66 6. Oil 1.54 4.65 7.85 7. Serv. exc. trans. 6.65 6.00 5.33 8. Total 23.04 24.42 25.84 9. C.a. gap (exc. trans.) 8.17 5.93 3.64 Noan-cap. suirp. 10. Merch 0.03 1.41 2.85 11. Serv. -2.98 - 1.86 -0.72 12. Total - 2.95 -0.45 2.13 13. Cap. goods imps. 5.22 5.48 5.77 14. Net transfer inflows 3.91 2.10 0.24 To address this question, it is convenient to rearrange entries in the current account as follows: Capital goods imports-=current account gap (excluding transfers) + (merchandise exports - non-capital merchandise imports)+surplus on factor and non-factor services. (3) The predicted shares of these components of the current account in table 3 show that at low income levels, the current account gap (excluding net transfers) exceeds capital goods imports. Transfers plus financial inflows used to cover the gap in principle provide the foreign exchange resources required F.D, McCarthy et al., Trade patternis in developing countries, 1964-82 15 for imports of capital goods. However, as line I I shows, the pattern is for developing countries to run deficits on non-factor and factor service trade. Hence, they rely on large and growing suipluses of merchandise exports over non-capital merchandise imports (line 10) to finance capital goods imports, especially at relatively high income levels where net transfers are small (line 14). As discussed below, this pattern (with variations) reappears in most sample splits, and illustrates an important aspect of underdevelopment - when service trade is taken into account, financial capital inflows are not large cnough to pay for physical imports of capital goods. Elasticities of major trade groupings (as a share of GDP) with respect to per capita GNP are as follows: Imports Exports Primary prod. 0.02 -(.18 Manufactures -0.02 0.34 Oil 0.33 0.79 Merch. trade 0.03 0.09 Fac. and non-fac. serv. -0.17 -0.08 The elasticity of the lerel of merchandise imports (constant U.S. dollars) with respect to GNP per capita is about 1.03, on the basis of the above estimates. Since the average annual growth in GNP per capita for the sample countries over the period 1964-82 was about three percent, this implies that volume imports grew at about the same rate over this period. During the period, GNP per capita in the industrialized countries grew about two percent per year. The imports of the developing countries served as a source of demand growth for the industrialized world. Passing from per capita GNP to other explanatory variables, observe that increased population size r, i-ces most import shares, in line with the usual expectation. Partial elasticities for botlh merchandise and service trade are -0.22, so a country with twice the population of another is predicted to have import shares more than 40 percent lower. For service exports, the service elasticity is -0.34. For merchandise, it is only -0.22, reflecting the fact that export shares for cereals, textiles, automobiles, and capital goods rise with population. A reasonable hypothesis is that early import substi- tution in these categories in large countries provides the base for subsequent exports. Responses of import and export shares to the capacity utilization variables take both signs. Merchandise and service import totals decline, but raw material, intermediate, automobile, and capital goods shares go up, presum- ably to meet higher aggregate demand (especially for producers' goods). The merchandise export share is strongly related to capacity utilization, with an 16 F.D. McCarthy et ail., Trade patterns in developing countries, 1964-82 elasticity of 0.42. The categories with a positive association are agricultural products (1 through 3) and oil. Except for an insignificant coefficient in category 10, the remaining merchandise export shares other than oil fall as capacity utilization goes up. Causal links are likely to run from the primary commodity exports (including oil) to aggregate demand. In the other categories, higher capacity utilization represents greater domestic demand and diverts sales from abroad to home. For similar reasons, factor service exports (e.g., emigrant worker remittances) and transfers have a negative association with capacity utilization. Almost all export categories (non-food agriculture and cereals excepted) are positively related to real GDP in OECD countries, as might be expected. Elasticities of merchandise and service shares are, respectively, 0.7 and 0.99, with an average of about 0.89. The level of real GNP in the sample countries has an elasticity with respect to OECD GDP of 1.6 (roughly estimated as the ratio of logarithmic growth rates of the two groups over the sample period). Hence, the elasticity of the level of dollar exports is about 2.5. This estimate is consistent with those of other, more macroeconomically oriented studies, e.g., Dornbusch (1984). The final explanatory variable is the real effective exchange rate, given by EER, where Home country exchange rate index / Partner exchange rate index EER=- -, Home country consumer price index / Partner consumer price index' where partners are weighterd by overall trade, oil exports are excluded, and the index is set to 100 in 1978-79. From price effects alone, a higher real rate should increase trade shares in GDP. The results support this view except for imports of cereals, and exports of food prod&tcts, cereals and (especially) raw materials. Causality may run from high shares of primary product exports to a low real exchange rate, in line with the literature on mineral exporter syndromes. Oil exports, which might be expected to have the same effect, have a positive but insignificant coeficient. A second observation is that a higher real exchange rate increases import shares of GDP more than export shares (even ignoring raw materials). Shares of oil and service imports rise sharply, probably because prices of these goods are fixed in dollars. The same may be true for service exports as well. For imports, the elasticity of the share with respect to EER minus one gives an approximation to a quantity elasticity, on the assumption that internal prices of importables have a unit elasticity with respect to the exchange rate. 'Quantity elasticity' values by category are: F.D. McCarthy et al., Trade patterns in developing countries, 1964-82 17 Imports - Quantity elasticity with respect to EER Food -0.88 Transport -0.55 Cereals -1.36 Elec. & mech. -0.69 Non-food ag. -0.77 Oil -0.28 Raw mat. -0.74 Non-oil total -0.78 Intermed. -0.74 Merch. total -0.72 Textiles -0.91 Non-fac. serv. -0.11 Autos -0.72 Fac. serv. -0.56 Consumer -0.74 Transfers 0.31 Elasticities of such magnitude are consistent with conventional views about price-responsiveness in trade, e.g., Khan (1974), Williamson (1983) and Bond (1985) on the developing countries. Goldstein and Khan (1985) give similar magnitudes for industrial economies. Price responsiveness for exports is less clearcut in our developing country sample. As indicated above, share coefficients with respect to the log of EER are negative for primary products (except non-food agriculture, which is not significant). Estimated share elasticities for non-Ricardian categories are: Exports - Share elasticity with respect to EER Intermed. 0.39 Oil 0.39 Textiles 0.20 Non-oil total (cat. 5-10) 0.31 Autos 0.40 Merch. total (cat. 5-11) 0.35 Consumer 0.18 Non-fac. serv. 0.68 Transport 1.00 Fac. serv. -0.06 Elec. & mech. 0.36 Transfers 0.81 If these values approximate quantity export demand elasticities for non- primary products, then the standard Marshall-Lerner stability condition (that the sum of absolute import and export demand elasticities exceed one) would barely be satisfied for merchandise trade (the situation is a bit more favorable for services). For manufactured goods, a 35 percent real devalua- tion (which after inflationary effects would require a nominal depreciation of 50-100 percent) would raise their predicted GDP share by 10 percent, from 0.04 to 0.044 at a per capita income level of $850. Export-led growth acceleration would not obviously result; indeed, modest export pessimism seems to be built into our results. 3. Trade by time period Now we turn to discussion of several splits of the sample panel. The aim is heuristic - to suggest how different trade patterns may arise in broad categories of countries. No attempt is made at statistical testing, since with a sample of several hundred observations the standard Chow test will say that almost any split is 'significant' by not rejecting the null hypothesis of 18 F.D. McCarthy et al., Trade patterns in developing countries, 1964-82 different coefficients across groups. It should come as no surprise that nations or time periods differ; the real question is the economic importance of the differences across relatively homogenous groups. In large samples, simple statistical tests do not address this issue. The first split is by time periods: 1964-73 before the first oil shock, 1974- 77 between the two shocks, and 1978-82 after the second one. Over the tumultuous two decades in the sample, trade patterns remained quite stable. There were significant changes in direction of trade but this analysis does not include any appropriate variable for this. For broad commodity groups, regression equations for the different periods fit more or less well in the same way as in table 2, and coefficients at least for per capita GNP and population change typically occur only in the second figure after the decimal. One way of looking at effects of sample splits is through deviations of predicted values given by the subsample equations from predictions given by the pool. Table 4 gives such a breakdown for time periods at different income levels (with the other explanatory variables at their mean levels). For example, the predicted share of food imports from the pool regression when per capita GNP is $250, is 2.08 percent. The predicted share from the 1964- 73 equation is 0.11 percent less (i.e., 1.97 percent) and so on. Two points stand out in the results of table 4. The first is unsurprising - the oil import and export shares in trade rose by between two to three percent between the first and third period. Secondly, non-oil trade shares for both merchandise and services increased. Both price and quantity effects are no doubt involved in these changes. Among imports, the largest share increases were for intermediates, capital goods, and non-factor and factor services. Price effects may have been important, especially for the manufactures, since most developing countries had rising incremental capital-output ratios and an adverse shift in the terms of trade after the mid-1970s. On the side of exports, primary commodity shares declined at most income levels in the final period, reflecting the shift in the terms of trade, On the other hand, deviations of the share of all manufactured exports were as follows: Manufactured exports Deviation from pooled sample Real GNP Pooled per cap. sample 1964-73 1974-77 1978-82 $250 2.55 -0.26 0.44 1.32 $850 3.97 -1.12 0.93 2.10 $3000 5.43 -2.00 1.43 2.88 The increases in the final period outweigh reductions in the total primary Table 4 Imports. Deviation from pooled sample Per capita Pooled 1964- 1974- 1978- Sector GNP (1982) sample 1973 1977 1982 Food US$ 250 2.08 -0.11 -0.07 0.20 US$ 850 2.08 -0.10 0.11 0.17 US$ 3000 2.07 -0.09 0.16 0.14 Cereal US$ 250 1.40 -0.26 0.37 0.25 US$ 850 1.17 -0.20 0.20 0.13 US$ 3000 0.93 -0.14 0.03 -0,00 Nfoodagr US$ 250 0.74 -0.01 0.00 -0.03 US$ 850 0.92 -0.04 0.05 -0.02 USS 3000 1.09 -0.06 0.10 -0.00 Rawmat US$ 250 0.25 0.01 0.04 0.05 US$ 850 0.40 0.00 0.06 0.05 US$ 3000 0.55 -0.01 0.09 0.05 Intermed USS 250 4.61 -0.66 0.62 0,84 US$ 850 4.69 -0.74 0.58 0.56 US$ 3000 4.78 -0.83 0.54 0.27 Textiles US$ 250 1.79 0.22 -0.25 -0.27 US$ 850 1.20 0.06 -0.02 -0.03 US$ 3000 0.60 -0.10 0.22 0.20 Autos US$ 250 0.88 -0.03 0.11 0.10 US$ 850 0.87 -0.01 0.03 0.15 US$ 3000 0.86 0.00 -0.06 0.20 Consumer US$ 250 2.22 -0.10 -0.02 0.14 US$ 850 2.09 -0.19 -0.04 0.18 US$ 3000 1.95 -0.29 -0.06 0.21 Transport US$ 250 1.38 -0.12 0.23 0.23 US$ 850 1.24 -0.20 0.30 0.31 US$ 3000 1.11 -0.29 0.38 0.40 Elecmech US$ 250 3.84 -0.50 0,42 0.73 US$ 850 4.24 -0.56 0.42 0.67 US$ 3000 4.66 -0.62 0.42 0,61 Oil US$ 250 2.39 -1.12 0.64 2.24 USS 850 3.59 - 1.48 1.01 1.63 US$ 3000 4.83 -1.84 1.38 0.99 Noiltot US$ 250 19.20 -1.68 1.70 2.27 US$ 850 18.93 -2.15 1.80 2.15 US$ 3000 18.65 -2.63 1.90 2.02 Totmerch US$ 250 21.50 -2.78 2.47 4.51 US$ 850 22.42 -3.59 2.95 3.79 US$ 3000 23.36 -4.42 3.44 3.05 NFS US$ 250 6.45 -1.55 1.12 0.89 US$ 850 4.43 -1.30 0.63 1.31 US$ 3000 2.35 -1.05 0.13 1.74 Facinc US$ 250 3.18 -0.39 -0.31 0.81 US$ 850 3.43 -0.56 -0.38 0.97 US$ 3000 3.70 -0.74 -0.46 1.14 Transfer USS 250 1.31 -0.09 0.15 -0.04 US$ 850 0.61 -0.20 0.11 0.15 US$ 3000 -0.11 -0.31 0.07 0.34 Totservs USS 250 10.86 -1.92 1.03 1,74 US$ 850 8.40 -1.96 0.43 2,50 US$ 3000 5.87 -2.00 -0.19 3.28 Grandtot US$ 250 33.25 -5.39 1.99 5.57 US$ 850 30.60 -5.98 1.99 6.48 US$ 3000 27.86 -6.58 1.99 7.41 Table 4 (continued) Exports. Deviation from pooled sample Per capita Pooled 1964- 1974- 1978- Sector GNP (1982) sample 1973 1977 1982 Food US$ 250 7.04 -0.03 0.46 -0.36 US$ 850 5.38 -0.12 0.27 -0.47 US$ 3000 3.68 -0.21 0,08 -0.58 Cereal US$ 250 0.19 0.06 0.03 0.04 US$ 850 0.21 -0.00 0.05 0.09 US$ 3000 0.22 -0.07 0.06 0.14 Nfoodagr US$ 250 3.07 0.22 0.00 -0.62 US$ 850 2.05 0.12 -0.07 -0.35 US$ 3000 1.01 0.02 -0.14 -0.08 Rawmat US$ 250 2.00 -0.58 0.68 0.92 USS 850 2.16 0.19 0.37 -0.33 US$ 3000 2.32 0.99 0.06 -1.62 Intermed USS 250 1.20 -0.07 -0.20 0.28 US$ 850 1.58 -0.37 0.13 0.37 US$ 3000 1.97 '-0.67 0.47 0.46 Textiles US$ 250 0.83 -0.19 0.19 0.42 US$ 850 1.04 -0.35 0.22 0.54 USS 3000 1.25 -0.51 0.26 0.66 Autos US$ 250 -0.08 -0.02 -0.02 -0.04 US$ 850 -0.03 -0.04 -0.00 0.01 US$ 3000 0.02 -0.07 0.01 0.06 Consumer US$ 250 0.51 -0.04 0.23 0.46 US$ 850 0.96 -0.26 0.30 0.72 US$ 3000 1.43 -0.49 0.38 0.98 Transport US$ 250 -0.02 0.08 0.07 0.14 US$ 850 0.08 0.04 0.08 0,19 US$ 3000 0.18 0.00 0.09 0.24 Elecmech US$ 250 0.11 -0.02 0.17 0.06 US$ 850 0.34 -0.14 0.20 0.27 US$ 3000 0.58 -0.26 0.22 0.48 Oil USS 250 1.54 -0.83 0.81 0.66 US$ 850 4.65 -1.05 1.67 1.15 US$ 3000 7.85 - 1.28 2.55 1.65 Noiltot US$ 250 15.09 -0.86 1.05 0.69 US$ 850 14.07 - 1.22 1.04 0.52 USS 3000 13.01 - 1.58 1.03 0.34 Totmerch USS 250 16.70 - 1.59 2.27 1.50 US$ 850 18.64 -2.03 2.90 1.64 US$ 3000 20,65 -2.48 3.54 1.79 NFS US$ 250 4.90 -0.35 0.48 0.74 USS 850 4.42 -0.50 0.12 1.27 US$ 3000 3.93 -0.65 -0.26 1.81 Facinc US$ 250 1.75 0.03 -0.06 -0.21 USS 850 1.58 -0.19 0.09 0.06 USS 3000 1.40 -0.42 0.24 0.33 Transfer US$ 250 5.22 - 1.44 1.17 1.74 USS 850 2.71 -0.98 0.85 2.15 USS 3000 0.13 -0.50 0.52 2.56 Totservs USS 250 11.91 -1.97 1.33 2.38 USS 850 8.76 -1.87 0.77 3.55 US$ 3000 5.52 - 1.77 0.21 4.76 Grandtot US$ 250 27.95 -3.34 3.19 4.34 US$ 850 25.24 - 4.03 3.18 5.80 US$ 3000 22.44 -4.73 3.17 7.30 F.D. McCarthy et al., Trade patterns in developing countries, 1964-82 21 share (-0.02, -1.06 and -2.14 at the three income levels) and reflect country trade diversification in response to the oil shocks and the decline in terms of trade.6 Despite these shifts in some categories, the overall changes were on the order of one percent of GDP - in that sense trade patterns were stable. The same is true for response parameters with respect to the structural variables - per capita GNP and population. Table 5 illustrates this observation with partial elasticities for non-oil merchandise and non-factor service trade. The estimates for GNPPC and POP in the pool and sub-period vary in a narrow range; the same is less true for the other explanatory variables. Trade patterns maintained themselves but responses to policy-related variables fluctuated over the 19-year sample period. Table 5 Elasticities with respect to explanatory variables by sub-period -- non-oil merchandise and non-factor service trade shares. Pool 1964-73 1974-77 1978-82 Non-oil merch. Imports GNPPC -0.01 -0.03 -0.01 -0.01 POP -0.22 -0.23 -0.21 -0.27 Y/Y* -0.06 0.05 -0.32 0.52 EER 0.22 0.04 0.13 0.59 Non-factor service Imports GNPPC -0.31 -0.36 -0.32 -0.21 POP -0.33 --0.29 - 0.40 - 0.37 Y/Y* -0.58 -0.58 -0.99 0.86 EER 0.89 0.77 0.76 -0.01 Non-oil merch. Exports GNPPC -0.06 -0.08 --0.05 -0.07 POP -0.22 -0.23 -0.23 -0.18 OECD-GDP 0.45 0.16 0.66 -0.85 Y/Y* 0.12 0.93 0.56 -0.69 EER -0.11 -0.07 -0.33 0.11 Non-factor service Exports GNPPC -0.08 -0.12 -0.14 0.01 POP -0.32 -0.31 -0.31 -0.37 OECD-GDP 0.70 0.01 0.82 4.41 YJY* -0.18 0.45 -0.67 0.49 EER 0.68 0.88 0.90 0.50 6For detailed country analysis, see Helleiner (1985). 22 F.D. McCarthy et al., Trade patterns in developing counatries, 1964-82 4. Growth performance Relationships between trade and growth have long been debated in development economics. In this section, we approach the question in a slightly novel fashion, asking if country growth rate differences have any impac+ on their observed patterns of trade. Many studies run the regression the other way, i.e., from export orientation to growth. Our procedure - using sample splits by economic 'performance' to investigate trade patterns - is no less justified in terms of causality, and allows greater disaggregation. We use it for that reason. First, a word on 'performance'. As fig. 1 illustrates, there is a positive relationship between 1982 per capita GNP levels and 1964-82 growth rates. Growth over the period itself is not enough to explain the association. The data, therefore, suggest that richer developing countries in the recent past have grown faster than poor ones. The regression equation using end-of- period income to 'explain' growth is GDP growth rate = 0.17 + 0.67 In GNPPC(82), R2 =0.119, (0.10) (2.68) with t-values in parentheses. Results are not strikingly different if income in earlier years is used on the right-hand side. We thus classify countries as high (low) performers according to whether their actual growth rate lay above (below) the regression line. A further split was made by per capita income level, with $1,000 as the dividing line. Growth rates of GDP in the different groups (annual percentage rate) were as follows: Low performers High performers All 3.48 All 7.00 Low income 3.51 Low income 7.07 High income 3.62 High income 6.98 The growth rate for the sample as a whole (from a semi-log regression on time) was 5.51 percent. Given such large differences in growth rates between the two groups (a 3.5 percent growth rate differential will double the income ratio between two countries in about 20 years), it is striking how similar their trade patterns are. Table 6 illustrates this point with respect to financing imports of capital goods. Note first from line 15 that when all right-hand side variables in the regression are set at their (sub-) sample mean values except for per capita income, low performers have a higher capital goods import share at all income levels. They run bigger current account gaps (line 11) but have lower surpluses on factor and non-factor service trade. Hence, merchandise trade surpluses in the two groups are of similar size. G D 0 Jordan P 9- G R R 0 Korea 0 T 0 Indonesia 0 Ecuador R 7 o Tnailand Malaysia 0 0 Brazil 0 Tunisia 0 Mexico T 0 Kenya Domin. Rep. 0 0Algeria E r 6- ~Nigeria o C.asto og % ° Malawi PhlipiEgyp Rcsa 0 0 0 Cyolo.bia g 9 P Moroaisa Mrcco ° P0ual°Cpu R TanzaniaO OSri Lanka 0 Venezuela OTrinidad & Tobago A 0ua HondurasS N 4- Mai0Togo ° ZambiaS NO India u o El Salvador m 0 Etiopia0 Peru M- a Ethiopia 0 Burkina Faso 1 0 Bangladesh 0 Argentina 9 OSenegal 0 Uruguay 0 Nicaragua 6 OCent. Afr. Rep. 0 Chile 4 2 o Niger OMadagascar 0 Jamaica 2 0 Ghana N 100 1000 10000 GNP PER CAPITA-1982 US$ Fig. 1. GDP growth rate v. GNP per capita. Table 6 Capital goods import finance in countries grouped by performance (deviations from pool regression). $250 $850 $3000 Pool Low High Pool Low High Pool Low High GNP per capita pred. perf. perf. pred. perf. perf. pred. perf. perf. Imports z 1. Prim. prod. 4.47 0.64 -0.21 4.57 0.69 -0.08 4.64 0.76 0.05 2. Total manuf. 14.72 2.61 -0.72 14.33 1.94 -0.25 13.96 1.27 0.23 3. Oil 2.39 0.20 0.00 3.59 -0.92 0.33 4.83 -2.08 0.67 4. Serv. exc. trans. 9.63 0.99 -0.33 7.86 2.88 - 1.79 6.05 4.82 -3.29 5. Total 31.21 4.44 -1.26 30.35 4.59 -1.79 29.48 4.77 -2.34 Exports 6. Prim. prod. 12.30 0.54 0.34 9.80 0.97 0.10 7.23 1.43 -0.15 7. Total manuf. 2.55 0.27 0.34 3.97 0.61 0.23 5.43 0.94 0.11 8. Oil 1.54 1.51 -1.08 4.65 -0.91 0.12 7.85 -3.41 1.36 9. Serv. exc. trans. 6.65 1.09 -1.51 6.00 1.45 -1.26 5.33 1.81 -1.01 10. Total 23.04 3.41 -1.91 24.42 2.12 -0.81 25.84 0.77 -0.31 11. C.a. gap (exc. transfers 8.17 1.03 0.65 5.93 2.47 -0.98 3.64 4.00 -2.65 Non-cap. surplus 12. Non-oil merch. 0.03 -0.05 -0.04 1.41 -0.22 -0.08 2.85 -0.43 0.20 13. Serv. exc. trans. -2.98 0.10 -1.18 -1.86 -1.43 0.53 -0.72 -0.43 2.28 14. Total -2.95 -0.05 -1.22 -0.45 -1.65 0.61 2.13 -3.44 2.48 lo 15. Cap. goods imps. 5.22 1.08 -0.57 5.48 0.82 -0.37 5.77 0.56 -0.17 16. Net transfer inflows 3.91 -0.26 -0.94 2.10 1.02 -0.71 0.24 2.34 -0.45 F.D. McCarthv et al., Trade patterns in developing countries, 1964-82 25 The low performers' economies are more open that those of the high performers. with higher total import and export shares (lines 5 and 10). They import more primary products, manufactures and oil (lines 1-3) and export more primary products and manufactures but - except at low income levels - less oil (lines 6-8). The slow growers also receive a higher transfer share of GDP (line 16). The deviations in all these categories are a few percent at most - well within the range of standard errors from table 1. Table 7 goes into more detail on deviations from the pool regression for different trade categories. The more open nature of the low performing economies shows up in most, e.g., higher imports (or less substitution) of food, raw materials, intermediates, and electrical and mechanical equip- ment; exports of food and non-food agriculture, and consumer goods. The major deviations by category are in service trade, where the low performers are again more open, and have a greater excess of imports over exports than do the high performers. Finally, shares of non-oil merchandise exports by group are: $250 $850 $3000 Pool pred. value 15.09 14.07 13.01 Low perf. deviation 0.59 1.28 2.00 High perf. deviation 0.25 -0.06 -0.38 Non-oil commodity exports do not seem to 'lead' growth in any obvious sense - their GDP share declines in all countries, but more in the high performers. Relative specialization in the slow growers runs more toward primary products (especially agricultural exports). For commodity trade, that appears to be the main distinction between the two sets of countries. For factor and non-factor services, the high performers have lower shares of both imports and exports, especially the former. The story on trade elasticities is similar. The high performers have a slightly higher elasticity of their non-oil export share with respect to OECD GDP - the value is 0.38 as opposed to 0.27 for the low performers. Other elasticity differences between the two major groups are of similar magnitude or smaller. When the high and low performers are split into sub-samples by income level, the higher export elasticities with respect to OECD income carry over to the smaller groups. Elasticities with respect to per capita GNP for major export categories are: Low performers High performers Low High Low High Pool All inc. inc. All inc. inc. Primary -0.18 -0.20 0.63 - 1.16 -0.15 0.02 0.06 Manufactures 0.34 0.30 0.36 -0.28 0.42 -0.08 0.0 Services -0.08 -0.07 -0.49 -0.74 -0.04 0.34 0.07 Table 7 Deviations by country performance groups of predicted trade shares from pooled sample predictions. Imports Exports Per capita Pooled Low High Pooled Low High Sector GNP (1982) sample perf. perf. sample perf. perf. Food US$ 250 2.08 0.37 0.07 7.04 0.55 -0.13 US$ 850 2.08 0.45 0.11 5.38 1.46 -0.35 US$ 3000 2.07 0.53 0.15 3.68 2.41 -0.57 Cereal US$ 250 1.40 0.18 -0.13 0.19 0.07 -0.10 US$ 850 1.17 0.12 -0.06 0.21 -0.33 0.07 US$ 3000 0.93 0.06 0.02 0.22 -0.75 0.24 Nfoodagr US$ 250 0.74 0.05 -0.08 3.07 0.88 -0.66 US$ 850 0.92 0.01 -0.05 2.05 0.23 -0.29 USS 3000 1.09 -0.02 -0.03 1.01 -0.44 0.09 Rawmat US$ 250 0.25 0.04 -0.07 2.00 -0.96 1.23 US$ 850 0.40 0.11 -0.08 2.16 -0.39 0.67 US$ 3000 0.55 0.19 -0.09 2.32 0.21 0.09 Intermed USS 250 4.61 0.79 -0.40 1.20 0.00 0.20 US$ 850 4.69 0.48 -0.24 1.58 0.03 0.20 US$ 3000 4.78 0.17 -0.08 1.97 0.05 Q.20 Textiles US$ 250 1.79 0.25 0.15 0.83 0.04 -0.20 US$ 850 1.20 0.29 0.17 1.04 -0.21 -0.11 US$ 3000 0.60 0.33 0.20 1.25 -0.48 -0.02 Autos US$ 250 0.88 0.13 003 -0.08 0.12 0.13 US$ 850 0.87 0.04 0.08 -0.03 0.14 0.13 US$ 3000 0.86 -0.05 0.13 0.02 0.17 0.13 Consumer US$ 250 2.22 0.36 0.07 0.51 0.01 -0.09 US$ 850 2.09 0.31 0.11 0.96 0.18 -0.15 US$ 3000 1.95 0.26 0.15 1.43 0.36 -0.21 Transport US$ 250 1.38 0.50 -0.17 -0,02 0.05 0.21 US$ 850 1.24 0.21 0.00 0.08 0.16 0.18 US$ 3000 1.11 -0.09 0.18 0.18 0.27 0.14 Elecmech US$ 250 3.84 0.58 -0.40 0.11 0.05 0.09 US$ 850 4.24 0.61 -0.37 0.34 0.31 -0.02 US$ 3000 4.66 0.65 -0.35 0.58 0.57 -0.13 Oil US$ 250 2.39 0.20 -0.00 1.54 1.51 -1.08 US$ 850 3.59 -0.92 0.33 4.65 -0.91 0.12 US$ 3000 4.83 -2.08 0.67 7.85 -3.41 1.36 Noiltot US$ 250 19.20 3.21 -1.03 15.09 0.59 0.25 US$ 850 18.93 2.62 -0.46 14.07 1.28 -0.06 US$ 3000 18.65 2.01 0.12 13.01 2.00 -0.38 Totmerch US$ 250 21.50 3.50 -0.91 16.70 2.15 -0.62 US$ 850 ?Z2.42 1.80 -0.01 18.64 0.55 0.23 USS 3000 23,36 0.04 0.93 20.65 -1.10 1.12 NFS US$ 250 6.45 1.62 0.59 4.90 0.79 -0.88 US$ 850 4.43 2.70 -1.14 4.42 0.65 -1.27 US$ 3000 2.35 3.81 -2.92 3.93 0.49 -1.68 Facinc US$ 250 3.18 -0.63 -0.92 1.75 0.30 -0.63 US$ 850 3.43 0.18 -0.65 1.58 0.80 0.01 USS 3000 3.70 1.01 -0.37 1.40 1.32 0.67 Transfer USS 250 1.31 -0.23 -0.31 5.22 -0.49 -1.25 USS 850 0.61 0.40 -0.02 2.71 1.42 -0.73 US$ 3000 -0.11 1.04 0.27 0.13 3.38 -0.18 Totservs US$ 250 10.86 -1.14 -2.66 11.91 0.73 -2.75 US$ 850 8.40 1.92 -2.62 8.76 2.97 -1.98 US$ 3000 5.87 5.07 -2.59 5.52 5.28 -1.18 Grandtot US$ 250 33.25 1.98 -3.90 27.95 4.27 -2.42 US$ 850 30.60 3.98 -2.28 25.24 6.29 0.51 US$ 3000 27.86 6.05 -0.60 22.44 8.37 3.54 F.D. McCarthy et al., Trade patterns in developing countries, 1964-82 27 Export-led growth in the form of high share elasticities, even for manu- factures among high-income, high-performance countries, is not apparent. 5. Oil exporters and importers In the contemporary world, there is obvious interest in the different economic performances of oil exporting and importing countries. There are 10 oil exporters' in our sample; the difference between this group and the other 45 countries with respect to the rinancing of capital goods imports is illustrated in table 8. Over the sample period, total GDP grew at 7.0 percent per year in the exporter, and 5.0 percent in the importer group. Note in line 16 that the exporting group is predicted to have higher capital goods imports than the oil purchasers; this difference is probably related to their faster growth. The oil exporters' economies are also more open than those of the overall sample - imports (line 5), exports (line 10) and their current account gaps (line 11) generally exceed the pool's. At both income levels above $250, they import more of all the broad commodity groups (including oil), but their non-oil trade deficits are substantially higher. Lines 12 and 14 show net deficits of about six percent of GDP each for both non- oil merchandise and services at the $850 income level and around ten percent at $3,000. The petro-economies' lack of export diversification shows up sharply, despite their overall high trade shares. Imports and exports (in1 total, and for most categories) of the oil importers lie below those of the pool - by about one pzrcent for the totals at $250, three percent at $850, and five percent at $3,000. On the other hand, the importers have higher net surpluses on non-oil merchandise trade (excluding capital goods imports) in line 12, and services in line 14. Their capital goods imports lie a bit below the pool's average. The petroleum importing countries have to run surpluses of several percent of GDP in goods and services apart from oil and capital goods to pay for imports of these last two categories. Oil exporters don't bear such a burden, and the differences in trading pattetns of the two groups of countries derive from this fact. 6. Country size So far, we have seen that trade patterns are not notably affected by time period and country performance, while oil exporters and importers differ in plausible fashion. Contrasts are a bit sharper in two other sample splits - by population size and region. We take up the first here, separating countries with populations above and below 20 million in 1982. 7See table A.1 for listing and derinition. 00 Table 8 Capital goods imports finance in oil exporting and importing countries (deviations from prediction of pooled regression). $250 $850 $3000 Pool Oil Oil Pool Oil Oil Pool Oil Oil GNP per capita pred. exp. imp. pred. exp. imp. pred. exp. imp. Imports 1. Prim. prod. 4.47 0.91 -0.20 4.57 0.25 -0.23 4.64 -0.43 -0.26 2. Total manuf. 14.72 -0.21 -0.26 14.33 1.85 -0.68 13.96 3.99 - 1.10 3. Oil 2.39 -1.94 0.23 3.59 1.02 -0.59 4.83 4.07 -1.44 4. Serv. exc. trans. 9.63 2.35 -0.84 7.86 5.33 -1.40 6.05 8.39 - 1.98 5. Total 31.21 1.11 -1.07 30.35 8.45 -2.90 29.48 16.02 -4.78 Exports 6. Prim. prod. 12.30 0.61 -0.09 9.80 -2.86 0.63 7.23 -6.43 1.37 7. Total manuE. 2.55 0.19 0.35 3.97 - 1.58 0.66 5.43 -3.38 0.97 8. Oil 1.54 0.98 -1.07 4.65 13.15 -4.23 7.85 25.70 -7.50 9. Serv. exc. trans. 6.65 -0.61 -0.03 6.00 -1.09 0.12 5.33 -1.60 -0.26 10. Total 23.04 1.17 -0.84 24.42 7.62 -2.82 25.84 14.29 -4.90 11. C.a. gap (exc. transfers 8.17 -0.06 -0.23 5.93 0.83 -0.08 3.64 1.73 0.12 Non-cap. surplus 12. Non-oil merch. 0.88 0.33 0.56 0.35 -5.31 1.80 -0.17 -11.10 3.05 13. Oil -0.85 2.92 - 1.30 1.06 12.13 -3.64 3.02 21.63 -6.06 14. Serv. exc. trans. -2.98 -2.96 0.81 - 1.86 -6.42 1.52 -0.72 -9.99 2.24 15. Total -2.95 0.29 0.07 -0.45 -0.40 -0.32 2.13 0.54 -0.77 16. Cap. goods imps. 5.22 0.23 -0.16 5.48 1.23 -0.40 5.77 2.27 -0.65 17. Net transfer inflows 3.91 -2.18 0.07 2.10 -1.08 0.25 0.24 0.04 0.42 F.D. McCarthy et al., Trade patterns int developitng counttries, 196442 29 Mean trade shares for the countries with population above and below 20 million in 1982: Large Small Merch. imports 15.12 25.79 Serv. imports 5.10 9.48 Merch. exports 11.95 20.94 Serv. exports 4.20 6.57 It is clear that the lager economies are on average much more closed to trade than small ones. The, 'lso grew faster - 5.7 percent per year as opposed to 4.8 percent for small countries. Table 9 gives deviations from the pooled regressions for the two size groups, as well as the regions discussed in the next section. On the import side, note that the merchandise share in GDP declines with income in big countries, while it rises in small ones. The same is true for most import categories - oil and raw materials have the only rising shares in large economies, cereals and textiles the only falling ones in the small economies. For textiles and non-factor services, shares are low in large economies at low income levels - import substitution occurs early. The decline in shares of these categories takes place at higher income levels in less populous countries. For merchandise exports, the GDP share of the total again goes down with income in the populous countries, and rises in the others. Both groups show declining shares for primary products and rising ones for manufactures (except textiles in big countries), but the transition takes place at higher income levels in the smaller economies. There are differences on the order of several percent of GDP in patterns of export specialization in food and non- food agriculture, and raw materials. At a per capita GNP level of $850, the share of these three categories is predicted to be 8.14 percent higher in small countries. Share differences in manufactured exports and all imports are spread much more evenly across categories. The large countries have higher export shares for cereals and most manufactures. Predicted patterns of capital goods import finance appear in table 10. Large countries have smaller deficits on non-factor and factor service trade and import fewer capital goods. Except at low income levels, they rely less on a merchandise surplus net of capital imports to finance their foreign purchases of investment goods. Consistent with their more open economies, the smaller countries have larger current account gaps (net of transfers) overall. Their non-capital goods merchandise surpluses increase with income (import substitution and export promotion come 'later' for small countries) while those of the large countries decline. Table 9 Deviations by population size and region of predicted trade shares from pooled sample predictions - Imports. Per capita Pooled High Low Eur. Mid. Lat. Am. Sector GNP (1982) sample pop. pop. Africa Asia E. & N.A. & Carib. Food US$ 250 2.08 -0.74 0.44 0.15 -0.33 2.71 -0.86 US$ 850 2.08 -0.99 0.59 0.93 -0.46 1.15 -0,47 US$ 3000 2.07 - 1.25 0.74 1.73 -0.59 -0.46 -0.08 Cereal US$ 250 1.40 -0.22 0.10 -0.27 0.49 1.72 -0.43 US$ 850 1.17 -0.45 0.21 0.23 -0.65 0.47 -0.23 US$ 3000 0.93 -0.69 0.32 0.75 - 1.82 -0.82 -0.03 Nfoodjagr US$ 250 0.74 0.18 -0.08 -0.23 0.33 0.72 -0.43 US$ 850 0.92 -0.04 -0.03 -0.66 1.37 0.50 -0.41 US$ 3000 1.09 -0.27 0.02 -1.10 2.44 0.27 -0.38 Rawmat US$ 250 0.25 0.09 -0.06 -0.09 0.31 0.05 -0.13 US$ 850 0.40 0.13 -0.10 -0.27 0.98 0.17 -0.18 US$ 3000 0.55 0.17 -0.14 -0.46 1.67 0.29 -0.23 Intermed US$ 250 4.61 - 1.09 0.48 -0.15 -0.60 1.49 1.20 USS 850 4.69 - 1.60 0.85 0.74 0.58 0.61 -0.07 IJSS 3000 4.78 -2.13 1.22 1.66 1.79 -0.29 -1.38 Textiles US$ 250 1.79 -0.99 0.51 0.49 -0.84 0.28 -0,19 US$ 850 1.20 -0.75 0.44 1.13 -0.24 0.32 -0.07 USS 3000 0.60 -0.49 0.36 1.79 0.37 0.36 0.06 Autos US$ 250 0.88 -0.26 0.19 0.17 -0.24 0.23 -0.29 US$ 850 0.87 -0.41 0.32 0.24 0.27 0.12 -0.05 US$ 3000 0.86 -0.56 0.44 0.32 0.78 0.01 0.19 Consumer USS 250 2.22 - 1.08 0.46 0.16 -1.07 0.53 0.70 US$ 850 2.09 -1.19 0.61 0.52 -0.07 0.24 0.09 US$ 3000 1.95 - 1.31 0.76 0.90 0.95 -0.06 -0.54 Transport US$ 250 1.38 -0.30 0.25 0.26 -0.41 0.41 --0.23 USS 850 1.24 -0.55 0.43 0.43 0.12 0.44 -0.16 USS 3000 1.11 -0.80 0.62 0.60 0.68 0.46 -0.10 Elecmech USS 250 3.84 -0.52 0.29 0.29 -0.13 1.23 -0.20 US$ 850 4.24 -0.99 0.59 1.16 2.67 0.58 -0.42 US$ 3000 4.66 - 1.47 0.89 2.06 5,56 -0.09 -0.65 Oil USS 250 2.39 -0.58 0.18 -0.18 0.52 -0.46 -3.54 US$ 850 3.59 - 1.51 0.97 - 1.15 1.41 -0.38 0.01 USS 3000 4,83 -2.47 1.77 -2.14 2.34 -0.29 3.67 Noiltot USS 250 19.20 -5.10 2.61 0.84 -2.48 9.32 -0.85 US$ 850 18.93 -6,96 3.89 4.66 4.58 4.55 -2.04 USS 3000 18.65 -8.88 5.22 8.59 11.86 -0.35 -3.26 Totmerch USS 250 21.50 -5.59 2.78 0.69 - 1.83 8.89 -4.36 US$ 850 22.42 -8.37 4.85 3.55 6.15 4.21 -1.99 US$ 3000 23.36 -11.23 6.98 6.48 14.38 -0.61 0.44 NFS USS 250 6.45 -3.63 1.73 3.18 -3.09 -0.42 -2.62 USS 850 4.43 -2.31 1.30 6.18 -0.65 -0.21 --0.41 US$ 3000 2.35 -0.95 0.86 9.27 1.85 0.00 1.87 Facinc USS 250 3.18 -0.88 0.40 0.58 - 1.23 3.47 -0.01 USS 850 3.43 -1.15 0.71 1.85 0.68 0.46 0.19 US$ 3000 3.70 -1.43 1,02 3.16 2.65 -2.65 0.39 Transfer USS 250 1.31 -0.96 0.56 1.38 - 1.07 10.16 -- 1.22 USS 850 0.61 -0.40 0.30 3.34 -0.06 (1.12 -0.29 US$ 3000 -0.11 0.18 0.03 5.36 0.99 0.08 0.67 Totservs USS 250 10.86 -5.45 2.68 5.21 -5.31 3.30 -3.78 US$ 850 8.40 -3.83 2.29 11.44 0.05 0.46 -0.44 LUSS 3000 5.87 -2.17 1 89 17.86 5.57 -2.47 3.00 Grandtot LJSS 250 33.25 -12.09 5.92 5,14 -7.78 17.49 -3.31 US$ 850 30.60 - 12.19 7.17 14.62 4.25 7.98 -3.38 USS 3000 27.86 -12.29 8.45 24,39 16.76 -1.82 -3.46 Table 9 (continued) Deviations by population size and region of predicted trade shares from pooled sample predictions - Exports. Per capita Pooled High Low Eur. Mid, Lat. Am. Sector GNP (1982) sample pop. pop. Africa Asia E. & N.A. & Carib. Food US$ 250 7.04 -3.54 1.61 0.96 -2.91 -3.45 8.54 US$ 850 5.38 -2.77 1.36 5.67 -2.02 -2.24 1 93 USS 3000 3.68 - 1.97 1.11 10.52 -1.10 -0.99 -4.88 Cereal USS 250 0.19 0.48 -0.10 0.01 0.52 0.20 -6.20 USS 850 0.21 0.44 -0.14 -0.09 0.22 -0.03 0.23 USS 3000 0.22 0.39 -0.17 -0.19 -0.08 -0.27 0.67 Nfoodagr USS 250 3.07 - 1.36 0.43 0.32 0.95 -0.89 -0.05 US$ 850 2.05 -1.16 0.56 1.22 6.52 -0.84 -0.48 USS 3000 1.01 -0.96 0.70 2.16 12,26 -0.79 -0.92 Rawmat USS 250 2.00 -1.12 0.55 1.70 -0.45 0.90 -0.72 US5 850 2.16 - 1.47 0.82 6.21 2.49 -0.60 -0.35 USS 3000 2.32 - 1.84 1.10 10.85 5.51 -2.16 0.03 Intermed USS 250 1.20 -0.59 0.18 0.05 -0.12 -0.45 -0.29 USS 850 1.58 -0.56 0.23 -0.21 1.29 -0.04 0.04 US$ 3000 1.97 -0.52 0.28 -0.47 2.74 0.37 0.38 Textiles US$ 250 0.83 0.58 -0.36 -0.63 1.11 0.44 0.08 USS 850 1.04 0.01 -0.04 -0.80 1.60 0.49 -0.56 USS 3000 1.25 -0.59 0.29 -0.98 2.10 0.54 - 1.22 Autos USS 250 -0.08 0.04 0.13 0.04 0.15 0.17 0.07 IJS$ 850 -0.03 0.10 0.10 0.01 0.17 0.28 0.04 USS 3000 0.02 0.17 0.08 -0.01 0.20 0.39 0.02 Consumer USS 250 0.51 0.19 0.04 -0.34 0.66 0.32 0.57 USS 850 0.96 0.02 0.10 -0.74 3.06 0.52 -0.28 USS 3000 1.43 -0.16 0.16 -1.14 5.53 0.72 -1.15 Transport USS 250 -0.02 0.07 0.11 0.14 0.09 -0.11 0.02 USS 850 0.08 0.14 0.05 0.09 0.47 0.06 0.02 USS 3000 0.I8 0.22 -0.01 0.04 0.86 0.23 0.02 Elecmech US5 250 0.11 -0.01 0.03 0.01 0.38 -0.11 0.07 llSS 850 0.34 0.07 - 0.02 -0.10 1.80 0.22 -0.15 USS 3000 0.58 0.14 -0.07 -0.22 3.26 0.56 -0.38 Oil USS 250 1.54 1.57 -0.91 0.94 0.47 2.91 -12.01 USS 850 4.65 -2.21 1.40 1.61 -0.97 -1.74 0.74 USS 3000 7.85 -6.11 3.79 2.30 -2.46 -6.53 13.88 Noiltot UJSS 250 15.09 -5.44 2.64 1.86 0.04 -3.41 7.92 US$ 850 14.07 -5.44 3.05 10.95 15.11 -2.69 0.25 LUSS 3000 13.01 -5.44 3.48 20.32 30.65 -1.94 -7.66 Totmercli USS 250 16.70 -3.91 1.67 2.51 0.33 -0.56 -3.84 USS 850 18.64 - 7.66 4.41 12.17 14.26 -4.35 1.01 US5 3000 20.65 - 11.54 7.23 22.12 28.61 -8.26 6.00 NFS USS 250 4.00 -2.39 1.11 1.40 -3.08 3.74 -0.98 US$ 850 4.42 -1.74 0.78 2.34 -2.02 1.88 -0.45 USS 3000 3.93 - 1.07 0.44 3.31 -0.94 -0.04 0.09 Facinc USS 250 1.75 -0.64 0.26 0.11 -0.08 1.04 -1.09 USS 850 1.58 -0.82 0.43 -0.45 1.28 0.87 -0.47 USS 3000 1.40 - 1.01 0.61 - 1.03 2.68 0.69 0.17 Transfer USS 250 5.22 -1.75 1.00 -0.41 -2.71 12.88 -3.40 USS 850 2.71 -0.81 0.15 -4.39 -5.41 5.97 -1.74 USS 3000 0.13 0.16 -0.72 - 8.48 -8.20 - 1.14 -0.02 Totservs USS 250 11.91 -4.87 2.56 1.02 -6.16 17.62 -5.77 USS 850 8.76 -3.48 1.54 -2.59 -6.46 8.67 -2.98 USS 3000 5.52 -2.06 0.50 -6.31 -6.78 -0.54 -0.10 Grandtot USS 250 27.95 -9.11 4.45 3.73 -5.74 17.13 -2.54 US$ 850 25.24 -9.52 5.02 10.04 10.41 6.11 -2.47 USS 3000 22.44 -9.93 5.61 16.53 27.04 -5.26 -2.41 Table 10 t-, Finance of imports of capital goods in large and small countries and by regions (predicted GDP shares). Merc. surp. Fac. & non- Surplus Net Total Total C.a. except cap. fac. serv. inet of cap. Cap. goods transfer Pool imports exports gap goods imps. surplus goods imps. imports inflows $250 31.21 23.04 8.17 0.03 -2.98 2.95 5.22 3.91 $850 30.35 24.42 5.93 1.41 - 1.86 --0.45 5.48 2.10 ) $3000 29.48 25.84 3.64 2.85 -0.72 2.13 5.77 0.24 Z Large countries $250 21.19 16.32 4.87 1.03 -1.50 -0.47 4.40 3.12 $850 18.54 14.47 4.07 0.83 -0.96 -0.13 3.94 1.69 $300() 15.83 12.53 3.30 0.62 -0.42 0.20 3.50 0.22 Small countries $250 36.10 26.12 9.98 -0.48 --3.74 -4.22 5.76 4.35 $850 37.36 30.05 7.31 1.85 -2.66 -0.81 6.50 1,95 $3000 38.36 34.29 4.07 4.76 - 1.55 3.21 7.28 -0.51 Africa $250 35.57 27.75 7.82 3.18 -5.23 - 2.05 5.77 2.12 $850 41.68 39.t8 2.50 -6.59 -8.00 1.41 7.07 -5.63 Asia $250 24.92 20.73 4.19 2.31 - 1.82 0.49 4.68 1.90 $850 36.36 38.31 -1.95 12.85 -2.63 10.22 8.27 0.14 EMENA $850 34.82 23.25 11.57 -5.71 0.64 --5.07 6.50 3.86 ol $3000 26.21 18.80 7.41 -3,85 2.58 -1.27 6.14 0.12 LAC $250 24.18 17.05 7.13 0.08 -2.42 -2.34 4.79 4.68 $850 28.17 24.68 3.49 3.97 -2.56 1.41 4.90 0.38 $3000 32.27 32.55 -0.28 8.02 -2.72 5.30 5.02 - 1.52 F.D. McCarthy et al., Trade patterns in developing countries, 1964-82 33 7. Regional effects There are characteristic regional differences in trade volume and compo- sition, as illustrated in tables 9 and 10 for sub-Saharan Africa, Asia outside the Middle East, developing countries in Europe, the Middle East and North Africa (EMENA), and Latin America and the Caribbean (LAC). Regional GDP growth rates, in percent for the period 1964-82, were Africa, 5.1; Asia, 5.6; EMENA, 5.2; LAC, 5.7. When countries are classified by growth performance, the results of section 4 reveal no substantial differences in trade patterns; the results here show that different regions grew at about the same rate, but differ from one another in trade. The trade/growth nexus is not strong. The general nature of regional trade patterns is illustrated in table 10. Import and export shares of GDP rise with per capita income in all regions but EMENA, with imports lower in LAC - the traditional bastion of import substitution. Countries in sub-Saharan Africa and EMENA are predicted to have larger trade gaps. Deficits on factor and non-factor service trade are larger in Africa than the other regions; as a consequence the African economies' predicted surplus on merchandise trade (apart from capital goods imports) is fairly high. EMENA countries run service surpluses and non- capital merchandise deficits - a pattern reversed from the rest of the developing world's. Africa and EMENA have relatively high shares of capital goods imports; Asia and LAC relatively low. Table 9 gives detail on trade patterns. At the relevant income levels, sub- Saharan African countries have high import - ires for food, textiles, and services; Asia for non-food agriculture, raw iaiaterials, and capital goods; EMENA for food. Among exports, Africa shows strong positive deviations for primary products and non-factor services, but is low on textiles and manufactured consumer goods. Asian countries are strong exporters of non- food agricuhural products, raw materials, and several categories of manu- factures. They are low on non-factor services and high on factor services. EMENA is low on agricultural exports, and high on services. LAC food and oil export shares are high (Venezuela and Trinidad and Tobago at the top of the region's income range). Elasticities of trade response appear in table 11. Sign differences for responses to changes in per capita GNP between Afiica and Asia on the one hand and EMENA and LAC on the other are striking - most shares rise with income in the first two groups and decline in the others. Elasticities of manufactured exports with regard to the real exchange rate are positive everywhere but Asia - are Africa, EMENA and LAC overvalued? Finally, non-oil exports respond positively to real OECD GDP in all regions but Africa. Because of the continent's extreme specialization in primary trade, its overall foreign exchange availability does not rise when thle industrial world grows faster. 34 F.D. McCarthy et al., Trade patterns in developing countries, 1964-82 Table 11 Elasticities with respect to explanatory variables by region. Pool Africa Asia EMENA LAC GNP per capita Manuf. imps. -0.02 0.14 0.43 -0.11 -0.12 Non-oil imps. -0.01 0,15 0.32 --0.18 -0.07 Merch. imps. 0.03 0.14 0.36 -0.12 0.13 Fac. and non-fac. serv. imps. -0.17 0.30 0.50 --0.64 -0.52 Manuf. exps. 0.34 0.20 1.25 0.42 0.01 Non-oil exps. -0,06 0.41 0.73 -0.02 -0.49 Merch. exps. 0.09 0.53 0.72 -0.11 0.28 Fac. and non-fac. serv. exps. -0.08 -0.15 0.01 -0.34 -0.93 Effective exch. rate Non-oil merch. imps. 0.22 -0.11 0.11 1.09 0.32 Manuf. exps. 0.31 1.18 -0.09 0.97 0.66 OECD GDP Non-oil exps. 0.45 -0.21 0.32 0,23 0.60 8. Conclusions The main conclusions from the pool regressions and sample splits are as follows: (1) The overall shares of imports and exports of goods and factor and non-factor services in GDP rise with the level of per capita GNP in the pool equations; however, non-oil merchandise and service shares decline. The shares of oil imports and exports consequently rise. Among exports, shares of primary commodities fall and manufactures rise. The share of capital goods imports in GDP goes up with per capita income. (2) In the pool regressions, the current account trade gap (net of transfers) less the predicted net deficit on factor and non-factor service trade is smaller than imports of capital goods. Hence, the predicted surplus of merchandise exports over non-capital goods imports is positive. Because of their net service trade deficits, most developing countries have to run surpluses on 'current' merchandise trade to pay for imports of capital goods. (3) Larger countries have lower shares of imports and experts overall. However, shares of most manufactured exports rise with population. Prior import substitution for a larger internal market may be involved. (4) In the pool regressions, coefficients of primary product and oil exports with respect to capacity utilization are positive - reverse causality may occur in that higher primary product sales abroad lead to greater domestic spending and output. Other export shares of GDP decrease with capacity utilization, i.e., low domestic demand opens a vent for industrial export sales abroad. When capacity use is high, GDP shares of intermediate and capital goods imports go up. F.D. McCarthy et al., Trade patterns in developing countries, 1964-82 35 (5) The elasticity of the overall export share with respect to real OECD income is about 0.9, consistent with estimates in other studies. (6) An increased real effective exchange rate leads to higher trade shares except for primary product exports. For the latter, reverse causality from a high share to a low exchange rate along 'Dutch disease' lines may be involved. The share elasticity of manufactured imports to the real exchange rate is about 0.3, implying a volume elasticity of about -0.7. The non- primary export share elasticity is about 0.3. The numbers imply a modicum of elasticity pessimism, especially for exports. (7) When the overall 1964-82 time period is split into the sub-periods 1964-73, 1974-77 and 1978-82 the regression equations remain stable, especially the coefficients for per capita GNP and population. The oil import and export shares rise by two or three percent (at different income levels) between the first and last period. Other trade shares typically go up as well, with a mix of price and quantity effects no doubt acting. Toward the end of the period, primary product export shares fall from those of the pool regression, reflecting the decline in the terms of trade that has occurred since the mid-1970s. (8) Over the sample period, there is a positive association between the GDP growth rate and the level of per capita GNP - richer developing countries grew faster. When the sample is split into high (low) performers with positive (negative) residuals in the regression of growth rate on per capita income, there is no great difference between the two groups in trade shares. The low performers have higher trade shares (by a percent or two overall) and larger gaps on net factor and non-factor service trade. Their capital goods import shares are also higher. The high performers do 7lot have higher shares of exports, and their export shares do not have higher elasticities with respect to per capita GNP. In either sense, export-led growth does not appear on average, though the low performance countries are a bit more specialized in primary products with low income elasticities of demand in industrialized countries. (9) When the sample is split into oil exporting and importing economies, the former have a less diversified export basket, have larger trade gaps and capital goods imports, and grow faster. Oil importers need a large surplus of merchandise exports over non-capital, non-oil imports to balance their current accounts, aind have trade patterns resembling those of the pool regression. (10) When the sample is split by population (at 20 million) overall trade shares rise with per capita GNP in small countries and decline in large ones. As noted above, large countries have higher shares of manufactured exports in GDP, despite their less open economies. They have lower capital goods import shares and smaller deficits on net service trade. Hence, large countries rely less than small on a 'current' surplus on commodity trade to pay for imports of investment goods. 36 F.D. McCarthy et al., Trade patterns in developing couintries, 1964-82 (11) When the sample is split regionally into sub-Saharan Africa, develop- ing countries in Europe, the Middle East and North Africa (EMENA), the rest of Asia, and Latin America and the Caribbean (LAC), all four areas grow at about the same rate but their trade patterns differ. EMENA has a net surplus on service trade and a deficit on 'current' merchandise, departing from the developing country norm. Africa has a large service deficit and specialization in primary exports. Asia is more specialized in manufactured exports; LAC has lower import shares. Elasticities of various tracle categories with respect to per capita GNP are typically higher in Asia and Africa than EMENA and LAC. In line with the pattern of specialization, manufactured exports are elastic with respect to the real exchange rate everywhere but Asia. If the regressions have a moral, it is that developing economies are constrained by commodity and service trade. But the fetters are not murderously tight; the R2 values are well under 100 percent. Room for maneuver exists around the common trade patterns; the policy problem is how to exploit the degrees of freedom that exist. These are conclusions which Carlos Diaz would have approved. He would have gone on to make prescient suggestions about which policy lines would make sense in a given historical and institutional context. No one else now seems able to draw the lessons that Carlos could. In that sense, his absence is his survivors' enormous loss. Table A.l Sample panel of countries.' = Grouping in analysis 1982 GNP per capita (tIS$) Growth capita (US_) 1982 pop. rate GDP Income Income! Performance/ Oil exposrters/ Debt Debt/ Country b in millions 1964-82' Population' Region' Performance' group' performanceh population' importers' status' oilt Algeria 2110 2200 20.3 6.5 H C H H HH HH X H X Argentina 1873 2520 28.6 2.8 H D L H HL HL M H M Bangladesh 12l 140 91.6 2.7 H B L L LL LH M L Brazil 2126 2310 122.7 7.0 H D H H HH HH M H M Burkina Faso 185 210 6.5 3.1 L A L L LL LL M L. United Republic of Cameroon 691 850 8.9 5.4 L A H L LH H L X L - Central African Republic 276 310 2.4 2.1 L A L L LL LL M L Chile 1921 2210 11.5 2.1 L D L H HL LL M H M Colombia 1418 1470 27.0 5.5 H D H H HH HH M L. Congo 1182 1180 1.7 6.0 L A H H HH HL M L Costa Rica 958 1260 2.4 5.5 L D H H HH HL M L Cyprus 3343 3740 0.6 5.1 L C L H HL LL M L Dominican Republic 1309 1400 5.8 6.5 L D H H HH HL M L Ecuador 1610 1210 8.9 7.6 L D H H HH HL X L. Egypt 777 680 44.3 5.8 H C H L LH HH X H X a4 El Salvador 711 740 4.8 3.4 L D L L LL LL M L. Ethiopia 111 140 32.6 3.2 H A L L. L LH M L Ghana 2621 350 12.2 0.9 L A L L L Li Li M L Greece 3960 4284 9.8 5.2 L C L H HiL LL M L. Guatemala 1118 1130 7.7 5.3 L D H H HH HiL M iL Honduras 658 670 4.0 4.2 L D L L L Li Li M L India 236 250 705.7 3.6 H B L L L L LH M H M ol Indonesia 572 580 152.6 7.7 H B H L LH HH X H X Ivory Coast 744 1140 8.8 6.0 L A H H HH HL M L . Jamaica 1390 1350 2.2 1.3 L D L H HL LL M L . Jordan 1226 1860 3.5 9.4 L C H H HH HL M L Kenya 330 390 18.1 6.5 L A H L LH HL M L Korea, Republic of 1801 1760 39.6 8.0 H B H H HH HH M H M Madagascar 299 320 9.2 1.4 L A L L L Li Li M L. Malawi 194 210 6.5 5.5 L A H L LH HL M L Malaysia 1718 1840 14.5 7.0 L B H H HH HL X Lw Mali 146 180 7.1 3.8 L A H L LH HL M L. Mexico 2164 2250 73.1 6.7 H D H H HH HH X H X Table A.l (continued) w 00 Grouping in analysis 1982 GNP per capita (US) (;rowth 1982 pop. rate GDP Income Income/ Performancei Oil exporters/ Debt Debt/ Country b in millions 1964-82' Population' Region' Perfottners' group' performance" population' impontersJ status' oil' Morocco 746 821) 21.6 5.2 H C H L LH H14 M H M Nicaragua 981 900 2.9 2.3 L D L L LL LL M L Niger 255 31(0 5.9 1.7 L A L L LL LL M L. Nigeria 833 821) 90.6 6.0 H A H L LH HH X L. Pakistan 374 38(0 87.0 5.2 H B H L LH HH M L - Peru 1198 1250 17.5 3.3 L D L H HL LL X L Philippines 774 820 50.7 5.6 H B H L LH HH M H M m Portugal 2207 2490 9.9 5.0 L C L H HL LL M L Senegal 409 441) 6.0 2.5 L A L L LL LL M L Spain 4714 5328 38.5 4.4 H C L H HL LH M L SriLanka 310 320 152 4.5 L B H L LH HI M L is Sudan 370 460 19.8 4.2 L A L L LL LL M L. United Republic of Tanzania 254 281) 19.3 4.5 L A H L LH HL M L 5a Thailand 735 800 48.5 7.1 H B H L LH HH M L -- is Togo 292 350 2.8 4.0 L A L L LL LL M L Zs Trinidad and Tobago 6466 6450 1.2 4.6 L D L H HL LL M L- Tunisia 1227 1370 6.7 6.8 L C H H HH HL M L: Turkey 1160 1331) 47.5 5.5 H C H H HH HH M H M Ulruguay 3089 2651) 2.9 2.6 L D L H H L LL M L . Venezuela 3971 32291 15.9 4.6 L D L H HL LL X H X Yugoslavia 2693 2840 22.7 5.6 H C H H HH HH M L Zambia 595 58(0 6.0 4.0 L A L L LL LL M L 'Source: Economic Analysis and Projections Department, Data Bank. 'GNP per capita using Atlas method. 'Except for Cyprus, Ecuador, Indonesia, Jordan and Trinidad and Tobago. is 'High and low (H aind L) correspond to countries witlt 1982 populations greater than 20 million and less than 20 million, respectively. 'The Four regions are Africa (A), Asia (B), Europe, the Middle-East and North Africa (C) and thte Caribbeats and South American countries (D). 'Based on regressing 1964-82 GDP growth rate on the log of per capita GNP for 1982. Observations below the regression line have been classified lowv growth performers (L), and those above it ,' highs growth performers (H). 'High and low (H and L) correspond to countries with 1982 per capita GNP greater than $1000 and less $1000. Atlas GNP was used for the country selections while GNP per capita at market exchange rates was used for the regression atialysis. hThe high and low performers were further subdivided according to their income levels, hiigh or low as in footnote g. The first letter indicates performance and the secorid letter income. 'The high and low performers were subdivided on the basis of their population size as in footnote b. The first letter indicates performance and the second letter population. 'The oil exporters group (denoted X) comprises those countries classilied as middle-income oil exporters in the 1984 World Derelopment Report. The remaining cDuntries are classified as oil importers (denoted M). 'The high debt countL:es comprise Algeria, Argentina, Brazil, Chile, Egypt, India, Indonesia, Korea, Mexico; Morocco, Philippines, Turkey and Venezuela, and are denoted H. T'he remaining countries are classiFied as low debt countries, denoted L. 'The high debt group was classiried according to its oil exporter/importer status, denoted X and M. F.D. McCarthy et al., Trade patterns in developing countries, 1964-82 39 References Amsden, Alice H., 1986, The direction of trade - past and present - and the 'learning effects' of exports to different directions, Journal of Development Economics 23, 249-274. Bacha, Edmar L., 1984, Growth with limited supplies of foreign exchange: A reappraisal of the two-gap model, in: Moshe Syrquin, Lance Taylor and Larry Westphal, eds., Economic structure and performance: Essays in honor of Hollis B. Chenery (Academic Press, New York). Balassa, Bela, 1983, Trends in international trade in manufactured goods and structural change in the industrial countries, 7th World Congress of the International Economic Association on Structural Change, Economic Interdependence and World Development, Madrid. Balassa, Bela, 1985, Exports, policy choices and economic growth in developing countries after the 1973 oil shock, Journal of Development Economics 18, 23-25. Bond, Marian E., 1985, Export demand and supply for groups of non-oil developing countries, International Monetary Fund Staff Papers 32, 56-77. Chenery, Hollis B. and Michael Bruno, 1962, Development alternatives in an open economy: The case of Israel, Economic Journal 72, 79-103. Chenery, Hollis B. and Moises Syrquin, 1975, Patterns of development, 1950-1970 (Oxford University Press, New York/London), Diaz-Alejandro, Carlos F., 1978, Delinking North and South: Unshackled or unhinged?, in: Albert Fishlow et al., Rich and poor nations in the world economy (McGraw-Hill, New York). Dornbusch, Rudiger, 1984, The effects of macroeconomic policies on non-oil LDC's: A review (Department of Economics, Massachusetts Institute of Technology, Cambridge, MA). Goldstein, Morris and Mohsin S. Khan, 1985, Income and price effects in foreign trade, in: Ronald Jones and Peter Kenen, eds., Handbook of international economics (North-Holland, Amsterdam). Havrylyshyn, O., 1983, The direction of developing country trade: Empirical evidence of differences between South-South and South-North trade, Conference on 'South-South or South-North?', Brussels. Helleiner, G.K., 1985, Balance of payment experience and growth prospects of developing countries: A synthesis, World Development, forthcoming. Khan, Mohson S., 1974, Import and export demand in developing countries, International MonetarTFund Staff Papers 21, 678-693. Krugman, Paul, 1980, Scale economies, product differentiation, and the pattern of trade, American Economic Review 70, 950-959. Kuznets, Simon S., 1964, Quantitative aspects of the economic growth of nations IX. Level and structure of foreign trade: Comparisons for recent years, Economic Development and Cultural Change 13, no. 1, Part 2, 1-106. Kuznets, Simon S., 1966, Modern economic growth: Rate, structure and spread (Yale University Press, New Haven, CT). Lancaster, K.J., 1980, Intra-industry trade under perfect monopolistic competition, Journal of International Economics 10, 151-175. Linder, S.B., 1961, An essay on trade and transformation (Almquist and Wicksell, Stockholm). Stewart, Frances, 1982, Recent theories of international trade: Some implications for the South, Mimeo. (Oxford University, Oxford). Strout, Alan, 1985, Structural determinants of South-South trade expansion (Departmient of Urban Studies and Planning, Massachusetts Institute of Technology, Cambridge, MA). Williamson, John, 1983, The open economy and the world economy (Basic Books, New York). THE WORLD BANK Headqttarters: 1818 H Street, N.W. Washington, D.C. 20433, U.S.A. Europenit Office: 66 avenue d'lena 75116 Paris, France bkiyo Office: Kokusai Building 1-1 Marunouchi 3-chome Chiyoda-ku, Tokyo 100, Japan The full range of World Bank publications, both free and for sale, is described in the WoradBatk CatalogofPutblicatiouis, and of tlecontinuingresearch program of the World Bank, in World Batk Researcth Programii: Abstracts of Current Stuidies. The most recent edition of each is available wvithout charge from: PUBLICATIONS UNIT THE WORLD BANK 1818 H STREET, N.W. WASHINGTON, D.C 20433 US.A.