The impact of foreign trade on the employment of unskilled U.S. workers: some new evidence.

• Author: Pryor, Frederic L.

1. Introduction

   At first glance, it seems unlikely that foreign trade should have a significant impact on the employment of unskilled U.S. workers. For example, imports from non-OPEC low-wage nations amount to less than 2% of the GDP.(1) Furthermore, as shown elsewhere (Pryor and Schaffer 1998), the number of jobs for prime-age workers (ages 25-49) in the United States that requires a high school diploma or less has increased at a faster rate than the number of workers in this age bracket with the corresponding educational credentials.

   However, does the foreign trade sector really have such a benign impact? Over the last two decades, the relative importance of foreign trade in the U.S. economy has roughly doubled, and imports from developing countries have grown even faster. Simultaneously, the rates of employment of prime-age men with only a high school education or less have fallen, real wages among employed but less educated workers have fallen, and wage differences between workers with different levels of education have increased.

   Attention to the linkage between foreign trade and labor market trends is part of a broader debate on whether the falling wages and employment rates of less educated male workers in the United States are due to skill-biased technical change or to foreign trade. In the fields of both labor economics and international economics, various economists place quite different weights on these two factors.(2)

   For example, in this debate among foreign trade economists, some, such as Lawrence and Slaughter (1993), find little empirical linkage between these foreign trade and labor market trends, Others, such as Bhagwati and Dehijia (1994), question on theoretical grounds the empirical relevance of the factor content arguments linking trade and domestic employment in manufacturing. Moreover, if labor markets function properly, foreign trade should have little overall employment effect, although the shift of unskilled workers from import-competing industries to other industries might place a downward pressure on their wages.

   By way of contrast, Wood (1994, 1995) finds considerable employment and wage effects due in large measure to the increasing imports of noncompeting goods that are highly unskilled-labor intensive. He further shows that among nations of the OECD, a strong inverse relationship exists between the decline in their share of manufacturing workers in the labor force and the increase in net imports of manufacturers from developing nations as a ratio of their GDP. Theoretical arguments about employment effects of foreign trade on unskilled workers focus on imperfections in the labor market. For example, import penetration places downward pressures on U.S. prices of goods, which are unskilled-labor intensive and which compete with imports from low-wage nations. Such pressures are in turn transmitted to wages of the U.S. workers in these industries. If the U.S. wages in these industries are sticky or if their equilibrium levels sink below the minimum wage, U.S. producers can no longer profitably employ their labor force. Alternatively, if the wages of U.S. unskilled workers in affected industries sink below their reservation wage, they would voluntarily leave their employers and perhaps the labor market. In either case, if wages in other U.S. industries are also downwardly sticky, such displaced workers might never find alternative employment.

   It is difficult to sort through these various competing theoretical and empirical claims regarding employment and wages of unskilled U.S. workers and to decide which tests are decisive. In this essay, I focus mainly on employment trends, as the causal forces underlying employment are easier to disentangle than those for wages. Further, a more exclusive focus on employment issues allows some crucial new empirical evidence to be introduced. The major conclusion is that foreign trade has relatively little impact on the employment of less educated workers, even in import-competing industries. I reach this result from the exploration of four hypotheses linking trade and the employment of unskilled U.S. workers:

   (i) Over the long run, net imports of the United States have increased, especially in those industrial branches that are characterized by a high percentage of unskilled workers in their labor force. Although this has occurred in certain well-known cases, I find no evidence to support the notion that this has been a general tendency for the manufacturing sector as a whole.

   (ii) Over the long run, prices in unskilled-labor-intensive branches of industry that are vulnerable to import penetration have risen less than in other branches of industries once domestic shocks such as differential changes in productivity are taken into account. I also find no empirical support for this hypothesis.

   (iii) Over the long run, productivity and investment in unskilled-labor-intensive branches of industries have increased more rapidly than other branches to meet potential competition from imports. This hypothesis receives no empirical confirmation.

   (iv) Imports are increasing most rapidly at the "low end of the market" and for products whose import prices are generally increasing less rapidly than prices of domestic production for the same product category. The low end of the market consists of those goods in a given product line that are most unskilled-labor intensive. Although the data on these matters leave much to be desired, the empirical evidence does not appear promising.

   If trade does not account for much of the fall of employment rates of unskilled male workers, then what does? The final section of this paper briefly surveys several other possible culprits.

2. The Data and a Methodological Issue

   Like others working on this topic, I use the NBER databases for trade and production, recently updated by Bartelsman and Gray (1996) and Feenstra (1997). Most investigators have used as a proxy of unskilled to skilled labor in a given industry the ratio of production workers to nonproduction workers. Many of those skeptical of the relationship between trade and unemployment find this proxy an imperfect solution, especially because many production workers are considerably more skilled and educated than office clerks classified as nonproduction workers. In the tables in this paper, I use both this proxy and a more direct measure, namely, the average years of formal education of all workers in an occupation or an industry.(3) I also carry out regression experiments using average wages as another proxy for skill. Because I reached the same conclusions with the wage variable as with the other two proxies, I do not discuss these results further. Several additional methodological issues also require our immediate attention.

   First, should we calculate only the skills directly embodied in various domestically produced goods, or should we use some type of input-output table and calculate both direct and indirectly embodied skills? On this topic, Feenstra and Hanson (1996a, b) contend strongly that focusing only on the skills directly embodied in various domestically produced goods misses the impact of increased outsourcing abroad. This trend, which was especially important in the period 1979-1990, allows a firm to focus its domestic efforts only on skilled-labor intensive parts of their production and to import unskilled-labor intensive inputs, a strategy that allows them to stave off import penetration. From this we can infer that import penetration should be inversely related to skill intensity of production of domestic production. More specifically, those industries that have upgraded their production, either by changing their production methods or by outsourcing unskilled-labor-intensive parts of their production process, have experienced less import penetration than other industries. In Table 1, I specifically test this hypothesis and find no empirical evidence to suggest that this effect is statistically significant.(4) Moreover, when I [TABULAR DATA FOR TABLE 1 OMITTED] examine the entire structure of industry in a quite different way in Table 5, I also find no such effect.

   Second, how can we interpret changes in total (direct and indirect) requirements of skilled and unskilled labor? This issue comes sharply into focus when we compare direct and indirect requirements of good A for a dollar's worth of production of good B over a period of more than a decade and a half.(5) The changes have been enormous, and, combined with the important shifts in the structure of imports, we have a problem in interpreting the relevant regression coefficients. That is, we do not know what is the key causal force, whether it is changes in production methods in the industry under examination, in the domestic industries supplying the inputs, or in the outsourcing abroad. If we focus only on direct requirements of skilled and unskilled labor, the causal relation is clear, and the impact of outsourcing abroad is picked up by directly examining net import penetration of those products used as inputs in the production of other goods.

   Third, what trade variable should be explained? In this paper, I focus on the change of net exports (exports - imports) to domestic consumption (domestic shipments + imports - exports) of the various product branches. Such an approach seems appropriate because it captures most fully the basic idea of the Heckscher-Ohlin approach toward trade, namely, that it is the contrast between the factor endowment of domestic production versus potential imports that determines whether net exports in that industry are positive or negative. If, by way of contrast, we focus on the gross levels of trade, we introduce additional issues of interindustry trade that, insofar as the goods are relatively similar, take us far from the factor-proportion hypotheses that we are trying to test.(6) For the empirical results of any such alternative approach to allow...