Causes of wage inequality: trade or new technologies?

The relative wage of high-skilled to low-skilled workers began to increase in US manufacturing industries in the late 1980s, and this phenomenon was also observed in Mexico. At the same time, the US–Mexican trade as a percentage of US GDP increased dramatically during the same period. Does this mean that trade is behind the observed pattern or is this an example of “correlation does not imply causation”?

Kurokawa (2014) ¹ examines the available evidence to draw a plausible conclusion. First he identifies the reasons against this hypothesis:

1. According to the standard Heckscher–Ohlin model (H-O) used to study international trade, after trade liberalization the relative wage of high skilled to low-skilled workers should increase in the high-skill abundant US, as production shifts toward high-skill intensive goods. In Mexico, however, production shifts toward low-skill intensive goods, and should increase the salaries in the sectors producing them and, consequently, decrease the wage inequality. If the H-O model is correct, the fact that the predictions for Mexico are not observed means that trade cannot be the reason for the inequality. This is called the “trade-wage inequality anomaly”.

2. In the H-O model, an increase in the relative wage of high-skilled to low-skilled workers should be driven by an increase in the relative price of high-skill to low-skill intensive goods in the high-skill abundant US. However, this is not observed, In the same vein as in point 1, this is called the “price-wage anomaly”.

3. A third criticism is based on the volume of trade: Krugman (1995) ² provides numerical examples to explain why the small volume of trade in the US makes it unlikely that trade can account for the change in wages. However, a point missed is that the volume is big in terms of the Mexican economy.

In view of these and similar observations for other countries, the primary explanations for the increase in wage inequality have been based on skill-biased technological change.

Krusell et al. (2000) ³ consider a standard theoretical model of aggregate production in which they allow for differences in capital and labor skill, and calibrate its parameters using US time series data during the period 1963-92. In the model they show how changes in factor inputs can account for most of the variation in the wages. According to their model the mechanism at work is as follows: a sharp decrease in equipment prices in the 1980s led to an increase in the demand for high-skilled workers, who could use this equipment, and a decrease in the demand for low-skilled workers, who are substituted by the new technology. This trend occurs both in the US and Mexico and is consistent with the “price anomaly”. This implies that the low-skilled workers in the US face competition, not from Mexican or other foreign workers, but from a combination of high-tech equipment and high-skilled workers in their own country.

Among many other studies relating skill-biased technological change to wage inequality we find Lindquist (2005) ⁴, who presents a Swedish version of Krusell et al’s work, and Goldin and Katz (1998) ⁵, who document the importance of capital-skill complementarity during the period 1909–1929, when the economy also experienced both significant technological change and a substantial increase in wage inequality.

It seems that the case should be closed and that trade-based models do not explain the data. However, the increase in international trade and recent studies suggest that trade can play a role even if technology is still the main reason for the increase in wage inequality.

On of the first attempts to resolve the anomalies of trade-based models is given by Feenstra and Hanson (1996) ⁶, who allow for a richer variety of skills and show that intermediate-skill activities that shift from the North to the South may be seen as relatively low-skill by the North standards, but high-skill by the South’s. Thus the skill intensity of production would rise in both the North and the South. However, Kremer and Maskin (2006) ⁷ note that their model shares with the other standard trade models the prediction that trade should be greatest between countries with the most different factor endowments; a prediction that is not observed. Other attempts to explain the anomalies faced similar problems, predicting facts that were not found in the data.

Kurokawa (2011a) ⁸ provides an alternative resolution of the anomalies based on Ethier (1982) ⁹. The variety trade in differentiated intermediate goods increases the variety of intermediate goods used by the final good in both countries. The increased variety of intermediate goods then can increase the variety of tasks to be handled and thus corresponds to higher demand for high-skilled workers. Through this variety-skill complementarity, the relative demand and thus the relative wage of high-skilled to low-skilled workers – the skill premium– rises in both countries. This can occur without a rise in the relative price of high-skill to low-skill intensive goods. Using the data from the period prior to the implementation of NAFTA, he also provides several numerical examples that illustrate that small amounts of variety trade can produce a significant increase in the relative wage.

This model is compatible with the data about the change in goods that the US an Mexico import from each other which show how each country started importing goods that it had not imported previously or had only imported in small quantities, thus indicating that the variety of manufactured imports increased in each country. A quantitative analysis of the hypothesis is done in Atolia and Kurokawa (2012) ¹⁰. The results indicate that the growth in the extensive margin of manufactured imports can account for up to approximately 15% of the change in the Mexican skill premium over 1987–1994.

Another alternative resolution is presented in Kurokawa (2011b) ¹¹, this time it is based on a modification of the H-O model assuming a “skill intensity reversal”. The idea is that the goods that US exports are produced using high-skill labor in the US, but low skill-labor in Mexico. If this is the case, liberalization of trade will increase the demand for high-skill workers in the US and will decrease the demand for low-skill workers in Mexico. Kurokawa (2011b) points out that US net exports to Mexico of electronics products increased from 1994 to 2000. US net imports from Mexico of non-electronics products also increased from 1994 to 2000. Non-electronics products were relatively low-skill intensive in the US but relatively high-skill intensive in Mexico both in 1994 and in 2000. The skill premium also increased in both countries. Thus the two-good H–O model with the reversal of relative skill intensities can be compatible with the data.

It is true that the H–O model is confronted with the anomalies, but variations of the model are successful in resolving the anomalies, weakening the criticisms of trade-based explanations of wage inequality. An increasing number of economists now argue that the effect of trade, though relatively small compared to that of technological change, is more significant than generally believed.