Trade Policy, Growth, and Income Distribution

by Sebastian Edwards
Trade Policy, Growth, and Income Distribution
Sebastian Edwards
The American Economic Review
Start Page: 
End Page: 
Select license: 
Select License

Trade Policy, Growth, and Income Distribution

For well over a century scholars, social commentators, and pamphletists have debated the merits of free trade. This controversy con- tinues today, even as the world is experiencing an unprecedented wave of trade liberalization. This debate has traditionally revolved around two questions: (i) Does freer trade result in faster economic growth? And (ii) how does freer trade affect social conditions, including income distribution? Two main theoretical problems have influenced this discussion: first, the difficulty, until recently, of theoretically linking trade policy to equilibrium growth; and second, the fact that the distributive effects of trade policy are very sensitive to assumptions on factor mobility, cones of diversification, and factor market distortions. The discussion on trade policy and economic performance has also been seriously affected by data limita- tions. It has been particularly difficult to obtain satisfactory comparative data on trade orien- tation, total factor productivity growth, and so- cial conditions (including poverty and income distribution) for a large group of countries.

In this paper I deal with the relationship be- tween trade policy and income distribution. First I discuss data and measurement prob- lems. Then, I address two questions: Is there evidence that open developing economies have a more unequal distribution of income than more protectionist ones? And, is there ev- idence that countries that have liberalized in- ternational trade have experienced an increase in inequality? These questions are germane to the development policy debate. Critics of the World Bank and the IMF, for example, have argued that these institutions' policy recom- mendations (including trade liberalization) have resulted in heightened inequality in the developing countries.

* Anderson Graduate School of Management, Univer- sity of California, Los Angeles, CA 90095, and National Bureau of Economic Research.

I. Measurement Issues

The empirical literature on trade and economic performance has been seriously hindered by measurement problems. The co- existence of tariffs and quantitative restrictions (QR's) has made the construction of satisfac- tory summary indicators of trade policy at dif- ferent points in time particularly difficult. Many early cross-country comparative studies relied on readily available indicators, includ- ing trade-dependency ratios and the rate of growth of exports, as proxies for trade policy (Bela Balassa, 1982). The main limitation of these measures is that they are not necessarily related to policy; a country can distort trade heavily and still have a high trade-dependency ratio. A large number of studies have used what I have called "subjective" criteria to classify trade orientation. The Heritage Foun- dation index of trade policy, which classifies countries into five categories according to per- ceived distortions, is perhaps the best example of this approach (Bryan Johnson and Thomas Sheehy, 1996). Other studies have relied on some (fragmentary) information to construct quasi-subjective indexes of trade orientation. Anne Krueger ( 1978) used dummy variables to classify trade regimes, and Michael Michaely et al. ( 1991) constructed a "subjec- tive index" of trade liberalization for their cross-country regressions. The World Bank's 1987 World Development Reportconstructed an ''outward-orientation' ' index for 4 1 coun- tries at two points in time. The classification of Korea as a "strongly outward-oriented7' country both in the 1963- 1973 and 1973

1985 periods has elicited considerable criti- cism, however, since during the earlier years the Korean trade regime was considerably more restrictive (Edwards, 1993 ) .Jeffrey Sachs and Andrew Warner (1995) used a series of trade-related indicators to construct a composite indicator of openness. Al- though this index is an improvement over previous ones, it still provides only a binary


classification: a country is either open or closed. Also, most of the underlying data used to construct this index are only available at one point in time.

In an effort to avoid subjectivity, some au- thors have used observed values of variables associated with trade restrictiveness as indi- cators of trade orientation. The most popular variables are tariff averages, average coverage of QR's, the black-market premium, and collected-tariff ratios (CTR7s), defined as the ratio of tariff revenues to imports. These in- dicators allow for intermediate situations where countries are neither totally open nor totally closed. A limitation, however, is that they are not always related among themselves and frequently provide a misleading picture of trade policy. Lant Pritchett ( 1991) found that some of these proxies were only weakly cor- related, and Pritchett and Geeta Sethi (1994) found that collected tariff rates misrepresent true protection. James Anderson and Peter Neary ( 1994) used ''welfare equivalence" to define a trade-orientation index. Anderson ( 1994) calculated this indicator for a group of 23 countries and found that the weighted av- erage tariff tended to underestimate the "true7' degree of trade restriction.

Some authors have used a regression-based approach to construct trade indexes. Edward Leamer ( 1988), for example, used an empir- ical Hecksher-Ohlin model with nine factors to estimate net trade flows and trade-intensity ratios for 183 commodities at the three-digit SITC level for 53 countries. He interpreted the deviations of actual trade intensity ratios from predicted values as indicators of trade barriers. Holger Wolf (1993) extended Learner's ap- proach, using a larger set of factors of produc- tion, a more disaggregated set of commodities, and three alternative base years ( 1975, 1980, and 1985).

The measurement problems summarized here suggest two things: first, additional ef- forts to improve existing indexes are needed; and second, robust empirical analyses should rely on more than one measure of openness. Measurement and data problems have not been restricted to trade orientation, but have also affected data on productivity growth, wages, and income distribution. Albert Fishlow ( 1996), for example, reports that it is extremely difficult to make comparisons of income inequality (and, in particular, of Gini coefficients) through time. For this reason, most comparative studies on income distri- bution have relied on cross-country analyses at one point in time. The World Bank, how- ever, has recently assembled a massive data set on the evolution of income distribution through time for a group of 77 countries (Klaus Deininger and Lyn Squire, 1996). This data set provides information on Ginis and income quintiles, and on data quality, and should be useful for future empirical analyses.

11. Trade Policy, Growth, and Income Distribution: Preliminary Analysis

Even though the connection between trade and income distribution will tend to be case- specific, international comparative studies can still be useful in providing some notion of regularities across nations. In this section, I present results from a preliminary analysis on the relationship between trade policy and income distribution in a cross section of countries. I ask whether there is evidence that more open countries have more unequal in- come distribution than protectionist ones. I use the new World Bank data set on income distribution, and I rely on six alternative mea- sures of trade orientation, four of which have data for two moments in time, roughly the 1970's and 1980's. The following trade- orientation indicators are used: average tar- iffs (TAR); average QR coverage; the World Bank index of outward orientation (OUT);average collected tariff ratio (PRO) ; Wolf's index bf import outward orientation (CLOSE);and average black-market premia (BL). In all cases a higher value of the index represents higher protection. Two variables that measure the change in income inequality between the 1970's and the 1980's were used:

the change in the average Gini coefficient for each of the two decades (AGINI = Gini80s -Gini70s; a positive number of this index denotes an increase in inequality);and
the decade-to-decade change in the av- erage share of income accruing to the poorest 20 percent of the population (APOOR; a pos- itive number for this indicator represents a

decline in inequality). The data set comprises all countries with available data on income distribution and trade-policy indexes BL and CLOSE for the 1970's and 1980's. The data set includes 44 countries, 27 of which are de- veloping (see Edwards [I997 1 for data sources).

In Figures 1 and 2, I present data, for the LDC's, on the relationship between the initial value (approximately 1970's) of the six trade- policy indicators and the Gini coefficient in the 1980's. As can be seen, in no case is it possible to detect a negative relationship. Quite on the contrary, these figures suggest that countries with more distorted trade regimes have had a more unequal distribution of income; TAR no causal relationship, however, is claimed. These results are very robust to the trade and inequality indexes used.

Table I deals with trade liberalization and changes in income distribution. A country is defined as a trade reformer if, according to a specific trade-orientation measure, there has been a relaxation in the degree of trade restric- tiveness. Since four indicators have data at two points in time (CLOSE, PRO, BL, and OUT), I used four different criteria to classify coun- tries into reformers and nonreformers. For both groups, Table 1 contains the distributions of AGINI and APOOR, as well as chi-square tests for the equality of the distributions. As can be seen, the majority of countries, both reformers and nonreformers, experienced a re- duction in income inequality during this pe- riod. Moreover, there appear to be no major differences in the distributions of AGINI and APOOR for reformers and nonreformers. For- mal chi-square tests for the equality of distri- butions, in fact, confirm that there are no statistical differences between the changes in inequality across reformers and nonreformers (see the last column in Table 1 ).

In order to explore this issue further I esti- mated a series of cross-country regressions that included other explanatory variables, in addition to trade indicators. The first additional regressor is the change in the coverage in sec- ondary education between the 1960's and the 1980's (A secondary). A second group of re- gressors captures macroeconomics conditions: GNP per capita in 1980 (GNPC80), average growth, and inflation. Due to the small number




First Third Variable quartile Median quartile X'


CLOSE Reformers -4.8 -1.2 4.8 1.0 Nonreformers -7.8 -1.8 0.6

PRO Reformers -7.9 -2.0 1.3 1.8 Nonreformers -9.3 -3.4 -1.5

BL Reformers -7.7 -2.0 -0.5 1.2 Nonreformers -5.3 -1.0 3.8

OUT Reformers -5.5 -1.9 5.1 4.4 Nonreformers -9.3 -1.0 3.8


CLOSE Reformers -0.55 0.23 0.75 1.1 Nonreformers -0.07 0.38 0.99

PRO Reformers 0.17 0.71 1.10 0.7 Nonreformers -0.55 -0.14 0.18

BL Reformers -0.55 0.04 1.80 1.3 Nonreformers -0.14 0.29 0.72

OUT Reformers -0.54 0.16 0.86 0.8 Nonreformers -0.39 0.43 0.71

Notes: AGINI is the decade-to-decade (1970's vs. 1980's) change in the average Gini coefficient. APOOR is the decade-to-decade change in the average share of income going to the poorest 20 percent of the population. The following four trade-orientation indicators are used to classify countries as reformers or nonreformers: Wolfs index of import outward orientation (CLOSE); average collected tariff ratio (PRO); average black-market premia (BL); and World Bank index of outward orientation (OUT). X2is the chi-square statistic from a test for equal- ity of distributions for reformers and nonreformers. In all cases, it is not possible to reject the hypothesis of equality of distributions.

of observations, the regressions were esti- mated for the full sample; interactive dummy variables (D) for the developed countries were included, however. Due to space limitations I only report the results for the BL index of trade distortions. Results are similar, however, for other indexes (see Edwards, 1997). The num- bers in parentheses are t statistics:


= -1.3 1 -0.058 (A secondary) 
(-0.83) (-1.82) 

+ 0.41 1 (growth) + 0.026 (inflation)

(1.41) (2.26)

- 2.47 (REFORM) 

+ 1.09 (REFORM X D)


The coefficient on the initial level of trade distortions (BL70) is positive, while the re- form indicator has a negative and insignifi- cant coefficient. This suggests that, other things given, countries that initially had a more distorted external sector experienced an increase in inequality; trade reform, however, does not appear to have significantly affected changes in income distribution. The estima- tion also suggests that trade policy has had different effects on LDC's and advanced nations. This, of course. is reasonable given the differences in factor endowments across these two groups of countries. The regression also suggests that, other things given, coun- tries that improved their education system during the 1960's and early 1980's experi- enced a reduction in inequality. Inflation in- creased inequality, and perhaps surprisingly, countries that grew faster tended to experi- ence an increase in inequality; the coefficient in the latter case, however, is not significant.

111. Conclusions

My analysis of the relationship between trade policy and income distribution using new comparative data sets strongly suggests that, for the developing countries, there is no evi- dence linking openness or trade liberalization to increases in inequality. These results appear to be robust to the use of alternative trade- policy and income-distribution measures. However, these results are still subject to some measurement problems, and to well known shortcomings of cross-country analyses. They should not be considered, in any way, as sub- stitutes for careful and detailed historical stud- ies of specific countries' experiences with trade reform.


Anderson, James E. "Trade Restrictiveness Benchmarks." Unpublished manuscript, World Bank, 1994.

Anderson, James E. and Neary, Peter. "Measuring the Restrictiveness of Trade Policy." World Bank Economic Review, May 1994, 8(1), pp. 151-69.

Balassa, Bela. Development strategies in semi- industrial countries. Oxford: Oxford Uni- versity Press, 1982.

Deininger, Klaus and Squire, Lyn. "Measuring Income Inequality: A New Data Base." Working paper, World Bank, 1996.

Edwards, Sebastian. "Openness, Trade Lib- eralization, and Growth in Developing Countries." Journal of Economic Litera- ture, September 1993, 31(3), pp. 1358-


."Openness, Trade Liberalization and Income Distribution." Working paper, Na- tional Bureau of Economic Research, Cam- bridge, MA, 1997.

Fishlow, Albert. "Inequality, Poverty and Growth: Where Do We Stand?" in Michael Bruno and Boris Pleskovic, eds., Annual Bank conference on development econom- ics. Washington, DC: World Bank, 1996, pp. 25-39.

Johnson, Bryan and Sheehy, Thomas, eds. 1996 Index of economic freedom. Washington, DC: Heritage Foundation, 1996.

Krueger, Anne. Foreign trade regimes and eco- nomic development. Cambridge, MA: Bal- linger, 1978.

Leamer, Edward. "Measures of Openness," in

R. Baldwin, ed., Trade policy and empirical

analysis. Chicago: University of Chicago Press, 1988.

Michaely, Michael; Papageorgiou, Demetris and Choksi, Armeane, eds. Liberalizing foreign trade. Oxford: Blackwell, 199 1.

Pritchett, Lant. "Measuring Outward Orienta- tion in Developing Countries: Can It Be Done?" Unpublished manuscript, World Bank, 1991.

Pritchett, Lant and Sethi, Geeta. ''Tariff Rates, Tariff Revenue and Tariff Reform: Some New Facts." World Bank Economic Re- view, January 1994, 8( 1 ), pp. 1 -16.

Sachs, Jeffrey and Warner, Andrew. "Economic Reform and the Process of Global Integra- tion." Brookings Papers on Economic Ac- tivity, 1995, (1), pp. 1-118.

Wolf, Holger. "Trade Orientation: Measure- ment and Consequences." Estudios de Economia, June 1993,20(3), pp. 51-86.

World Bank. World development report 1987. Oxford: Oxford University Press, 1987.

  • Recommend Us