Trade and Income Distribution: The Debate and New Evidence
by William R. Cline, Peterson Institute for International Economics
William R. Cline, a senior fellow at the Institute for International Economics, is currently on leave as Deputy Managing Director at the Institute of International Finance. This Policy Brief summarizes the main findings and recommendations of his Trade and Income Distribution, published by the Institute for International Economics in 1997, which are of central importance to the current debate over trade policy and globalization.
© Institute for International Economics. All rights reserved.
For most of the past 25 years, the distribution of wages in the United States has grown more unequal. The rising inequality is closely linked to educational and skill levels. Thus, the ratio of wages for workers with at least some college education to those for workers with high school education or less rose by 18 percent from 1973 to 1993 (figure 1).
Rising wage inequality might arguably not be a problem if it stemmed from a more rapid increase at the top of the distribution accompanied by some rise at the bottom as well. However, real wages have instead fallen in the lower part of the distribution. For most educational categories up through and including workers with some college, real hourly wages fell significantly from 1973 through 1993, with the greatest declines for the least educated (figure 2). Only the constant upgrading of the labor force, and the increased incidence of two-worker households, prevented real median household earnings from falling over this period.
Trade policy in particular has at times
been hostage to the fears of some
(including Ross Perot) that international trade
and globalization are reducing living standards
for American workers.
Although sustained growth and falling unemployment in the late 1990s have somewhat ameliorated the potential social pressures associated with widening inequality, and of falling real wages for many in the bottom half of the distribution, it is important to understand the sources of these trends. Trade policy in particular has at times been hostage to the fears of some (including Ross Perot) that international trade and globalization are reducing living standards for American workers.
There is in fact a respectable basis in economic theory for the proposition that free trade will undermine real wages of those toward the bottom of the distribution. Economists Eli Heckscher and Bertil Ohlin developed in the 1930s the theory that a country exports goods that are intensive in the use of its “abundant” factor (for the United States, skilled labor) and imports goods that intensively use its “scarce” factor (unskilled labor). Economists Wolfgang Stolper and Paul Samuelson further developed this theory in the 1940s to show that it implied trade liberalization would reduce the real wages of the scarce factor and increase those of the abundant factor. Essentially, more export opportunities would bid up wages of those primarily producing export goods while increased competition from imports would tend to bid down wages of workers producing substitutable goods domestically.
Many other factors than trade, however, have played a role in trends for skilled and unskilled wages, notably including rapid technological change, deunionization, and an eroding real minimum wage. In the mid-1990s a large economics literature developed seeking to disentangle the relative contributions of trade and other influences. Implicitly or explicitly the economists contributing to this literature mostly shared the analytical framework shown by figure 3.
This framework views wages of skilled workers relative to those of unskilled workers (vertical axis) as the outcome of the interaction between supply and demand in the labor market. The quantity of skilled labor available relative to that of unskilled labor is shown on the horizontal axis. For example, as college education rises so does the ratio of skilled to unskilled workers available, as illustrated by the shift of the supply curve from S0 to S1. If the demand curve (initially at D0D0) does not move outward as rapidly, then the equilibrium skilled/ unskilled wage ratio will fall (from W0 to W1), as indeed happened in the early 1970s with the explosion in the number of college graduates. If, however, the supply of skilled relative to unskilled workers does not keep up with growing relative demand for skilled workers, then the equilibrium wage ratio rises over time, for example from W2 to W3.
The economic debate about the causes of widening wage inequality may be seen as primarily reflecting differing views on what influences are driving outward the demand curve for skilled relative to unskilled labor. Those who blame trade invoke such an outward shift from trade liberalization and falling transport and communications costs. Those who instead emphasize technology argue that “skill biased technical change” has caused a rapid outward shift in the demand curve for skilled relative to unskilled labor. For its part, immigration of unskilled labor can shift the relative-skills supply curve backward to the left (by increasing the number of unskilled workers in the denominator), causing the equilibrium skilled/ unskilled wage ratio to rise. In large part there has been a “trade versus technology” division among economists that has spanned the subdisciplines of labor and trade economics.
Labor Economists 1 — The first wave of empirical research was by labor economists who stressed technological change and found little if any influence from trade. As a leading example, Berman, Bound, and Griliches (1994) used sophisticated statistical techniques to show that changes in the sectoral composition of US production contributed little to the rising relative demand for skilled labor. Inferring that trade must therefore not be responsible, they concluded that instead technological change must be the driving force. However, neither they nor most others directly measured the impact of technical change (just as this influence in the earlier “sources of growth” literature typically defied measurement and was left as a “residual”).
Labor Economists 2 — In contrast, another group of labor economists found larger international effects using a supply-demand framework. A prominent study by Borjas, Freeman, and Katz (1992) calculated the “labor embodied in trade,” treating imports as analogous to an addition to domestic labor supply and exports as a subtraction. Applying the resulting implied changes in wages by skill type (considering existing estimates of wage response to labor supply), they estimated that up to 20 percent of the rise in the skilled/ unskilled wage ratio could be attributed to increased trade and immigration, and that the effect was even greater for those with below high-school education.
Trade Economists 1 — These estimates provoked a sharp response from some trade economists. In particular, Lawrence and Slaughter (1993) argued that the “Labor-2” group had misunderstood basic trade theory, and that once properly analyzed, trade could be shown to have had no effect at all on relative wages. They emphasized that Hechsher-Ohlin theory requires two conditions for trade to be reducing unskilled wages relative to skilled. First, trade must reduce the relative price of the imported goods using unskilled labor. Second, there should be a general shift across all industries toward greater use of unskilled labor and lesser use of skilled labor in each product line, to free up skilled workers needed for expansion of exports and absorb unskilled workers displaced by imports. They argued that the evidence contradicted both of these conditions: import prices had not fallen relatively; and rather than shifting in the required direction, the combination of skilled with unskilled labor had systematically shifted in the opposite direction across industries, toward greater skill intensity. The latter point, however, did not address the fact that the entire US economy had experienced a sharp rise in the relative availability of skilled labor. (In terms of figure 3, trade could have shifted the demand curve outward without provoking a generalized shift toward greater unskilled-intensity in each product, because of the simultaneous offsetting outward shift in the supply curve.)
Trade Economists 2 — Some other trade economists in turn reacted sharply to the first group of trade economists, to argue that trade had in fact major effects. Leamer (1993, 1994) reshuffled trade and industry data in a way that seemed to reveal the presence of falling relative prices for imports, and argued that factor combinations (skilled-unskilled intensities) did not have to change for trade effects to be present (citing the possibility of “fixed coefficient” production conditions). Wood (1994) made the most dramatic calculations of all. He adjusted skill intensities to account for low-skilled goods that had been totally transferred to developing countries, and applied some rough multipliers for effects in services as well as induced technological change, to estimate that 100 percent of the rise in the skilled/ unskilled wage ratio could be attributed to increased North-South trade.
After having been a prominent member of the “Trade 1” group, Krugman (1995) effectively switched camps. Using an illustrative “general equilibrium” model, he calculated that rising North-South trade might have raised the skilled/ unskilled wage ratio by about 3 percent, or one-sixth of the total, so the “Labor 2” estimates had been about right after all. He pointed out that by trade theory the relative wage change is considerably larger than the relative product price change, so that the corresponding decline in relative prices of imported goods might have been too small to distinguish from the statistical noise. At the same time, his general equilibrium formulation highlighted why the textbook “small country” case with domestic wages driven by the “international margin” of foreign wages under free trade (the driving force in the much larger estimates by Leamer) was misleading. In reality, for the US and other OECD economies, the home country is large and trade is small, and the supply-demand approach of the “Labor 2” economists provides a good approximation of reality.
. . . it does not follow that increased
trade protection would benefit
I agree with Krugman's revised view that the Labor-2 group had come closest to the mark. A new general equilibrium model developed in Cline (1997) finds trade and immigration effects comparable to or somewhat greater than that school, but also shows that nonetheless they are unlikely to have been the dominant forces in rising wage inequality.
The Trade and Income Distribution Equilibrium (TIDE) model divides the world into 13 “countries” or regions. Using data for 1973, 1984 and 1993, it estimates production as a function of factors available in each country: skilled labor, unskilled labor, and capital (based on World Bank data). The model assumes falling transportation and communication costs over time, and similarly applies falling protection levels reflecting the successive rounds of trade negotiations. Each economy is divided into five sectors: three tradable (skilled-, unskilled-, and capital-intensive) and two nontradable (skilled- and unskilled-intensive). Sectoral exports are allowed to be up to half of domestic production, and imports up to half of domestic consumption. Trade must be balanced within 1 percent of GNP.
The market process is then simulated as follows. The model is run to determine the optimal allocation of each country's factors across products, and optimal levels of trade, to maximize the (“welfare”) value of consumption in each “country.” The value of one additional unit of skilled labor, unskilled labor, or capital in the optimal solution (“shadow price”) then provides the basis for examining the ratio of skilled to unskilled wages in the market solution. Simulation shocks to the model then provide a basis for examining what would have happened to relative wage levels if trade and immigration had been different from their actual paths. The difference from the baseline path then tells the contribution of trade and immigration to changing skilled and unskilled wage levels over time.
The underlying structure of the model and the data developed immediately suggest two opposing underlying forces at work. There is an income-concentrating force from integration with the world economy – the traditional Stolper-Samuelson effect whereby the scarce factor (unskilled labor in the United States) tends to lose from increased trade. However, there is also an income-equalizing dynamic factor change effect. The reason is that skilled labor, starting from a very low base, has been building up much more rapidly in developing countries than in industrial countries, so the relative abundance of skilled labor in the United States (and relative scarcity of unskilled labor) has been diminishing over time.
In the baseline run, with no skill-biased technological change incorporated into production, the buildup of domestic skilled labor relative to unskilled labor supply in the United States from 1973 to 1993 has a potentially dramatic effect in reducing the relative wage of skilled workers. By the measure used (primarily, some college and above) skilled workers rose from 37 percent of the US labor force to 56 percent over this period. This should have meant that the relative wage of skilled workers would have fallen by 40 percent, rather than rising by nearly 20 percent. This paradox immediately suggests that there were strong offsetting unequalizing forces at work, and again skill biased technological change is one of the most likely candidates.
The counterfactual simulations then provide a basis for measuring the impact of trade. The model is rerun with transportation and protection costs held constant at their 1973 level, and the resulting solution shows that the ratio of US skilled to unskilled wages would have been 7 percent lower in 1993 than in the baseline. Another counterfactual adjusts US skilled and unskilled labor supply in 1993 to remove the contribution to each from immigration during the previous 20 years. In this experiment the optimal solution places the US skilled/ unskilled wage ratio 2 percent lower than in the baseline. These results are moderately higher than in the estimates of the “Labor 2” school. However, the model's finding of much larger other forces toward inequality means that these findings remain qualitatively within the same family of results finding that trade and immigration do not account for the bulk of rising inequality.
Table 1 reports an illustrative synthesis of the new estimates in Cline (1997) along with other estimates from the literature on other forces affecting US wage inequality. The table first shows that the equalizing force of rising relative skilled labor supply would have reduced the skilled/ unskilled wage ratio by 40 percent from 1973 to 1993 in the absence of other influences. The next section reports the contribution of each of several unequalizing forces to rising inequality. Altogether these forces must have amounted to amost a 100 percent rise in the initial skilled/ unskilled wage ratio for the net effect to have been an 18 percent rise (specifically: 1.97x0.6=1.18).
The table attributes a 6 percent rise in the skilled/ unskilled wage ratio to the impact of trade, based on the TIDE model and other model results in Cline (1997). It uses the TIDE estimate of 2 percent for the impact of immigration. The principal estimates in the literature are used for the central estimates shown for the impact of the falling real minimum wage (5 percent increased wage ratio) and deunionization (3 percent). There is still a large missing unequalizing force, and the table arbitrarily allocates half of this (rather than the entirety as often done) to skill biased technical change, and the remaining half to other unexplained influences.
Including a possible further impact of outsourcing, trade and immigration together thus account for a 9 percent rise in the relative wage over this period, or about half of the net increase. Nonetheless, trade and immigration remain far from dominant as the cause of rising inequality, because together they contribute only about one-tenth of the gross (total) unequalizing forces at work over this period.
Despite the finding that trade probably has had some unequalizing impact, it does not follow that increased trade protection would benefit American workers. For one thing, probably about half of the impact of increased trade stems not from trade policy but from technological changes that have reduced transportation and communication costs. Moreover, simulations of the TIDE model and findings from earlier work (including on the distributional impact of textile protection costs) strongly suggest that while higher protection might reduce real skilled wages, it would do little to raise real unskilled wages. The efficiency losses (including dynamic effects) and the tendency for protection costs to be disproportionately concentrated on lower-income households offset the traditional Stolper-Samuelson absolute protection gains for the unskilled workers. This analysis also carries over to potential further trade liberalization. Little if anything would be gained for real unskilled wages through the avoidance of such liberalization, while its potential overall economic gains would be foregone if liberalization were to be halted. Moreover, the TIDE model projections for the next two decades show minimal adverse impact for unskilled wages from a scenario of complete removal of remaining protection.
What the findings do imply, however, is that the market process alone does not necessarily assure that the gains from trade are evenly distributed. This suggests in turn that a wide array of other equity-oriented policies should appropriately accompany a free-market (and free-trade) strategy. More narrowly, the results also suggest the potential importance of training and education in assuring that rising relative demand for skilled workers translates into broad advances for all workers rather than leaving behind a large cadre of unskilled. As for immigration, the findings suggest that it may be appropriate to take another look at the historic shift in recent decades toward family ties rather than skills as the criterion for eligibility.
The views expressed in this publication are those of the author. This publication is part of the overall program of the Institute, as endorsed by its Board of Directors, but does not necessarily reflect the views of individual members of the Board or the Advisory Committee
Figure 1: Skilled/unskilled wage ratio, 1961-93
Figure 2: Hourly wages by education, 1973-93
Figure 3: Interaction of Supply and Demand for Skilled/Unskilled Labor
Table 1: Illustrative Sources of Increase in the Ratio of Skilled to Unskilled Wages: United States, 1973-93 (percent)a
|A. Equalizing forces
Increase in stock of skilled relative to unskilled labor
|B. Unequalizing forces
Trade: Stolper-Samuelson effects
Lower transport and communication costs
Trade: outsourcing, Hicksian convergence
Falling minimum wage
Skill-biased technological change
|C. Net effect||18|
a. Percentages for unequalizing forces must be chained, not simply added, to equal total unequalizing impact. Similarly, “A” and “B” must be chained to calculate “C”.