Copyright New York State School of Industrial & Labor Relations 1999

Headnote:

During the 1980s in Mexico the wage gap between skilled and unskilled workers widened. The authors assess the extent to which this increased wage inequality was associated with Mexico's sweeping trade reform in 1985. Examining data on 2,354 Mexican manufacturing plants for 1984-90 and Mexican Industrial Census data for 1965-88, they find that the reduction in tariff protection in 1985 disproportionately affected low-skilled industries. Goods from that sector, the authors suggest, may have fallen in price because of increased competition from economies with reserves of cheap unskilled labor larger than Mexico's. The consequent increase in the relative price of skill-intensive goods could explain the increase in wage inequality.

During the 1980s, Mexico experienced a dramatic increase in wage inequality. The wages of more-educated, moreexperienced workers rose relative to those of less-educated, less-experienced workers. While such events are interesting in their own right, what makes the change in Mexico's wage structure particularly noteworthy is that it coincided with a sweeping liberalization of trade. In 1985, Mexico announced that it was joining the General Agreement on Trade and Tariffs (GATT), bringing an end to four decades of importsubstitution industrialization. The government proceeded to drastically reduce most trade barriers in the following three years. Itwas in 1985 that wage inequality in Mexico began to rise.

In this paper, we analyze data on 2,354 Mexican manufacturing plants for 1984-90 together with Mexican Industrial Census data for 1965-88 to assess the extent to which the increase in the skilled-unskilled wage gap in Mexico was associated with the opening of the Mexican economy. The motivation for studying the Mexican case is to understand the apparent global trend toward greater wage inequality. Since the 1970s, the wages of skilled workers have increased relative to those of unskilled workers in the United States and in Great Britain.1 Several recent studies link the rise in wage inequality to the increased openness of the U.S. economy, arguing that competition from low-wage countries has reduced the relative demand for unskilled workers and caused their wages to fall relative to those of skilled workers (Leamer 1993,1998; Wood 1994; Feenstra and Hanson 1996a). Other studies instead associate rising wage inequalitywith technological change (Davis and Haltiwanger 1991; Bound and Johnson 1992; Lawrence and Slaughter 1993; Berman, Bound, and Griliches 1994). The reasoning is that the advent of computer technology has made skilled workers increasingly important in the workplace.

The focus of the literature has so far been on developed economies. Wage changes in middle- and low-income countries have received little attention.2 This is unfortunate. If trade is contributing to wage changes in developed countries, then we should observe the opposite wage movements in developing-country relative wages. If global skill-biased technical change is the cause of relative-wage changes, then we should observe similar relative wage movements in high-wage and low-wage countries. Given Mexico's proximity to the United States and its recent opening to trade, the country is an ideal candidate in which to look for such changes.

The Stolper-Samuelson Theorem

The link that standard trade theory identifies between trade and wages is embodied in the Stolper-Samuelson (1941) theorem and its generalizations (Ethier 1984). The Stolper-Samuelson logic is that trade affects relative factor rewards by changing relative prices. To explain wage changes in Mexico with this logic, we would need a succession of events such as the following. Trade liberalization causes the prices of skill-intensive goods to rise relative to those of non-skill-intensive goods. The price changes reduce the demand for labor in non-skill-intensive industries and increase the demand for labor in skill-intensive industries. The resulting shift in employment toward skill-intensive industries contributes to an increase in the relative demand for skilled workers, which causes their wages to increase relative to those of unskilled workers.

This story is consistent with either of two hypotheses. The first is that Mexico has a wealth of skilled labor and a dearth of unskilled labor relative to the rest of the world. In its reserves of skilled labor, Mexico is far behind the United States, of course, but it may have a decisive edge over lowincome countries, such as China. The second hypothesis is that under import substitution Mexico extended trade protection preferentially to industries that make relatively intensive use of unskilled labor. Trade liberalization would then have a disproportionately large impact on non-skill-intensive sectors. Although such a policy would seem at odds with Mexico's presumed comparative advantage in low-skill activities, political considerations may have led the government to protect these industries. In either case, the Stolper-Samuelson explanation for the observed wage changes implies that (1) the relative prices of skillintensive goods have increased, and (2) there has been a shift in employment toward skill-intensive sectors.

Recent literature on Mexican labor markets provides important insights into changes in the country's wage structure, but has yet to fully identify the channels through which trade affects wage inequality. Feliciano (1993) and Cragg and Epelbaum (1996), both using householdlevel data, found that the return to schooling increased in Mexico during the late 1980s. Revenga (1997), using plant-level data, found that blue-collar wages and employment were more responsive to changes in trade protection than those of whitecollar workers, which she attributed to the fact that blue-collar workers were relatively concentrated in industries that underwent the largest reduction in protection levels. Feenstra and Hanson (1997) found that the Mexican regions in which the relative demand for skilled labor rose most in the 1980s were those in which foreign investment was most concentrated, suggesting that foreign capital inflows may have contributed to rising wage inequality. Finally, Bell (1997) studied the impact of minimum wages on employment in Mexican manufacturing plants over the period 198490. Though real minimum wages fell substantially over the sample period, she found that minimum wages had no impact on manufacturing labor demand, which she attributed to the fact that the minimum wage was not binding for most manufacturing plants during the sample period. In 1984 minimum wages were 42% of the average blue-collar Mexican manufacturing wage; in 1990 the figure was 31%.3 Bell's results are important, for they suggest that the decline in Mexican minimum wages during the late 1980s cannot account for the increase in wage inequality.

Relative Wages and Employment in Mexico

Data are available from two sources. We have annual data on 2,354 Mexican manufacturing plants for the period 1984-90 from the Secretariat of Trade and Industrial Promotion (SECOFI). The SECOFI sample is the only plant-level data source available in Mexico. It is fortunate for our purposes that the sample period spans the implementation of trade reform.

One cause for concern is that by design the SECOFI sample covers only medium and large plants.4 In 1986, there was an average of 321 workers per establishment in the SECOFI sample, compared to 67 across all manufacturing establishments. To ensure that the empirical regularities we identify in the SECOFI sample are representative of Mexican industry as a whole, we also use data on manufacturing establishments from the Mexico Industrial Census. We have Census data on employment, number of establishments, and total payroll by state and two-digit (ISIC) manufacturing industry for the period 1965-88, at roughly five-year intervals. To a first approximation, the SECOFI sample is representative of the overall mix of industrial activity in Mexico. Unreported results show that the distribution of employment across two-digit (ISIC) industries in the SECOFI sample and in the Industrial Census are nearly identical.

A final issue we need to address before turning to the data is how to measure wages. To identify the effect of trade on relative wages, we must be able to distinguish workers by skill level. The SECOFI sample and the Industrial Census classify workers in two categories: obreros, who are equivalent to blue-collar workers, and empleados, who are equivalent to white-collar workers. The activities of blue-collar workers include machine operation, production supervision, repair, maintenance, and cleaning; those of white-collar workers include management, product development, administration, and general office tasks. We identify white-collar workers as skilled labor and blue-collar workers as unskilled labor. We measure earnings as the average annual salary or average hourly wage for each type of worker in a given plant. The whitecollar-blue-collar distinction has obvious limitations, but the substantial wage differences between the two types of workers suggest that the division is a useful one.5

There are no data for Mexico that provide a more detailed plant-level breakdown of employment by type and by industry.6

(Table Omitted)

Captioned as: Table 1.

The Liberalization of Trade

To frame the discussion, we begin by considering the dimensions of trade reform in Mexico. Mexico's economy was largely closed to trade from the 1950s until the mid-1980s. The government initiated a conscious policy of trade protection in the late 1940s, when it raised tariffs and instituted a system of import licenses. Successive administrations expanded trade barriers, mainly by increasing the range of goods covered by import licenses. These licenses effectively gave the government the discretion to impose import quotas at will. The government also used export controls to direct production toward the domestic market.

The government decided to open the economy to trade in 1985. It moved swiftly, drastically lowering most trade barriers within three years. In mid-1985, the national average tariff was 23.5%, and importlicense requirements covered 92.2% of national production. By December 1987, import-license coverage had been reduced to 25.4% of national production and the average tariff had been reduced to 11.8%, with a maximum rate of 20%. Concurrent with reform, the government abolished export controls and devalued the nominal exchange rate.

(Table Omitted)

Captioned as: Table 2.

Table 1 shows annual productionweighted-average tariffs and import-license coverage by two-digit (ISIC) industry for the period 1984-90.' In 1984, the average tariff ranged from 13.6% in basic metals to 47.3% in wood products; import licenses were required for over 85% of products in all two-digit industries. The government first cut import-license requirements, reducing average import-license coverage to below 4% by 1988 in all industries except food products and metal products. It then proceeded to reduce tariffs; by 1990, the maximum tariff rate in any industry group was 18.4%, for "other industries." The only industry that continued to enjoy relatively high levels of protection was metal products, due to import restrictions on automobiles.

Concomitant with trade reform, the Mexican government removed many barriers to foreign investment, including limits on the foreign share of equity ownership in a Mexican firm and requirements that foreign firms obtain government approval for technology transfer from abroad and other activities (Feenstra and Hanson 1997). In 1983, following the onset of the Mexican debt crisis, there was a de facto relaxation of restrictions on foreign investment. These changes were codified into law in 1989.

In the mid-1980s, the Mexican govern ment also launched a program to privatize state-owned firms (Lopez-de-Silanes 1997). Within manufacturing, two sectors, basic metals (iron and steel) and petroleum products, were subject to wide-scale state ownership. Privatization affected a far larger share of production in these industries than in other two-digit manufacturing industries.8 Further, in iron and steel, conflicts between unions and new owners over the interpretation of pre-existing (that is, government-negotiated) collective bargaining agreements led to the virtual collapse of production in several large plants. To distinguish the effects of trade reform from those of privatization, we present some of our results excluding petroleum products, iron, and steel from the sample of manufacturing industries.9

(Graph Omitted)

Captioned as: Figure 1.

Wages and Employment

Since 1984, there has been a dramatic increase in white-collar-blue-collar wage inequality among manufacturing workers in Mexico. Table 2 shows average real wages and average relative wages for the SECOFI sample of plants.10 Between 1984 and 1990, the ratio of average hourly whitecollar and blue-collar wages increased from 1.93 to 2.55; the ratio of average annual white-collar and blue-collar earnings show a similar change, increasing from 1.91 to 2.47. The rise in the skilled-unskilled wage differential was due to a combination of real-wage increases for white-collar workers and real-wage decreases for blue-collar workers. Between 1984 and 1990, average real hourly wages for white-collar workers increased by 13.4%, while those for bluecollar workers decreased by 14%."

The Industrial Census shows similar movements in the white-collar-blue-collar wage gap. Figure 1 plots the ratio of average annual white-collar wages to average annual blue-collar wages in Mexican manufacturing for the period 1965-88. Between 1985 and 1988, the white-collar-blue-collar wage ratio rose from 1.84 to 2.16, which matches the movements in the SECOFI sample over the same time period. Of greater significance, the rise in the wage gap after 1985 appears to have halted a twodecade trend toward decreasing wage inequality.

The wages we report are average levels and do not control for changes in the composition of the labor force or for changes in the distribution of skill. In general, one should exercise caution in speculating about individual-level wage changes based on average wage changes. Nevertheless, the changes in relative wages in Table 2 are so large and occur over such a short period of time that it is extremely unlikely that they could be accounted for by compositional changes. The wage changes we report are consistent with household-level data on wage changes in Mexico over the same period reported in Feliciano (1993), Cragg and Epelbaum (1996), and Bell (1997) .

One source of relative-wage changes is shifts in the demand for different skill categories. An increase in the relative demand for skilled labor would cause an increase in both the wages and employment levels of skilled workers relative to those of unskilled workers.'2 Table 3 shows the ratio of white-collar to blue-collar employment for the SECOFI sample. Between 1984 and 1990, there was virtually no aggregate change in relative employment. The ratio of white-collar to blue-collar employment increased from 0.431 to 0.433; the ratio of white-collar to blue-collar hours worked decreased from 0.427 to 0.421. Industrylevel data also fail to indicate substantial changes in relative employment. Unreported figures show that four (of nine) twodigit industries experienced an increase in the relative employment of white-collar labor, while five two-digit industries experienced a decline. No industry showed large changes in either direction.'3

The Industrial Census shows somewhat larger changes in relative employment. Figure 2 plots the ratio of white-collar employment to blue-collar employment for the period 1965-88. The ratio fell from 0.346 in 1985 to 0.328 in 1988. The relativeemployment movements after 1985 arrested a two-decade trend toward greater relative employment of white-collar workers. We also calculate relative employment by twodigit industry for the Industrial Census in 1985 and 1988 (not shown). Seven (of nine) industries show a decline in relative white-collar employment, while two show an increase.

One possible explanation for the rising skill premium and the declining use of skilled labor is that the supply of skilled labor shifted to other sectors or migrated abroad. This could account for both a reduction in the share of skilled employment and an increase in the skilled wage. There is no evidence suggesting that the composition of Mexican emigrants changed over the 1980s. Borjas (1994) found that the average education level of Mexican immigrants in the United States arriving between 1980 and 1990 was the same as that of pre-1980 arrivals. The average educational level of Mexican immigrants to the United States is lower than that of immigrants from any other major source country. Borjas also found that during the 1980s the wages of Mexican immigrants in the United States declined relative to those of ethnically similar U.S. natives. This suggests that, if anything, there was an increase in the relative emigration of unskilled workers from Mexico during the 1980s. There is also little evidence of a shift of employment out of manufacturing into other sectors. The manufacturing share of the total labor force increased during the 1980s, rising from 16.7% in 1980 to 19.9% in 1988.

The Heckscher-Ohlin Model: Prices, Industry Composition, and Wages

The data show a substantial increase in the skilled-unskilled wage gap in Mexico following the liberalization of trade, but little change in the relative employment of skilled labor. One explanation for this pattern, consistent with the StolperSamuelson theorem, would be that trade increased the relative price of skill-intensive products. In this section, we search for evidence of Stolper-Samuelson effects.

Relative prices and skill intensity. To determine if relative price changes correspond to the observed relative-wage changes, we consider the correlation between relativeprice changes and relative skill intensity. In a many-good, many-factor world, there is no general definition of relative factor intensity. The crude measure of skill intensity that we use is the log ratio of whitecollar to blue-collar employment. If relative price increases in skill-intensive sectors account for increasing returns to skill, then we should see a positive relationship between skill-intensity and price increases in the 1980s.

Table 4 reports regressions of the log change in output prices on the log ratio of white-collar to blue-collar employment over the 1984-90 period. The unit of observation is the four-digit manufacturing industry.'4 We measure employment as the average annual number of workers by skill type. Since we are looking at discrete, rather than infinitesimal, price changes, we average log relative white-collar employment over the first and last period. We use two measures of prices: gross output prices, measured using a four-digit producer price index, and value-added prices, calculated using the output price index, input price indices, and input cost shares.'5 We also redo the analysis excluding petroleum products, iron, and steel industries, where stateowned firms initially played a large role and where substantial privatization occurred. To be consistent with the four-digit definitions of output prices, tariffs, and importlicense coverage rates (all of which the Mexican government constructs by apply ing output weights to more disaggregated data), we weight all regressions by the industry share of total manufacturing output; results using employment weights are similar.

We find no significant correlation, positive or negative, between price changes and relative white-collar employment using either gross output prices or value-added prices. Excluding petroleum products, iron, and steel leaves the results unaffected.

(Table Omitted)

Captioned as: Table 3.

(Graph Omitted)

Captioned as: Figure 2.

(Table Omitted)

Captioned as: Table 4.

(Table Omitted)

Captioned as: Table 5.

Although the findings do not support Stolper-Samuelson effects, there are several possible explanations. One problem is that changes in product prices reflect many changes in the economy in addition to trade reform, such as privatization and deregulation. Industry-wide price changes may thus be a poor measure of changes in trade policy. A more serious problem is that industry price changes may be a poor measure of actual price changes. Price indices capture changes in the prices of goods firms actually produce, rather than price changes for a fixed set of goods. If trade causes firms to alter the mix of goods they manufacture (Feenstra and Hanson 1996b), then our measure of product price changes will confound pure price changes with compositional effects. One indication of problems in using product price data is that correlations between price changes and skill intensity for the United States are sensitive to changes in either the time period or the sample of industries (Lawrence and Slaughter 1993; Krueger 1995; Schmitt and Mishel 1996).

An alternative price-based measure of changes in trade policy is tariffs. Indeed, for a small open economy, which is a reasonable description of post-trade reform Mexico, changes in tariffs fully capture trade policy-induced changes in product prices. We examine whether the pattern of tariffs and import licenses varied across industries according to the skill intensity of production. Table 5 presents raw correlations of industry tariffs and import-license coverage rates in 1984, changes in these variables over the 1984-90 period, and the log ratio of white-collar to blue-collar employment for four-digit manufacturing industries.

In light of the wide range of trade protection measures used in the literature, it is worth describing our measures in detail. The tariff for a four-digit industry is the production-weighted average administrative ad valorem tariff across all products within the industry. The import-license coverage rate for a four-digit industry is the fraction of output within the industry that is subject to import-license requirements. Both tariffs and coverage rates are thus percentage values. We calculate changes in these variables as the time difference over the 1984-90 period. Import-license coverage rates are somewhat difficult to interpret, as they are an indirect indication of quantity restrictions on trade. While import licenses give the government the discretion to impose quotas, we have no information on where quotas were actually binding.

Pre-trade reform tariff levels appear to have been lower in skill-intensive industries than in non-skill-intensive industries. Relative white-collar employment is negatively correlated with 1984 tariffs. This correlation is not statistically significant for all manufacturing industries, but it is significant once petroleum products, iron, and steel are excluded. More important, it also appears that tariff reductions were smaller in skill-intensive industries. There is a positive correlation between relative white-collar employment and the change in tariffs over 1984-90. This correlation is statistically significant at the 10% level for all industries and significant at the 5% level when we exclude petroleum products, iron, and steel.'6 In both samples of industries, relative white-collar employment is negatively correlated with import-license coverage rates in 1984 and with the change in coverage rates over 1984-90, but neither correlation is statistically significant.

The correlations between tariffs and skill intensity in Mexico are supportive of Stolper-Samuelson effects. Skill-intensive sectors were less protected and consequently had smaller reductions in tariff levels. Why Mexico protected low-skill industries is a puzzle, given the country's presumed comparative advantage in these sectors. Similar patterns of protection have been reported for other developing countries. Currie and Harrison (1997) found that in Morocco protection was significantly higher in sectors with a higher share of unskilled workers, such as textiles and clothing. They also found these sectors to be the most export-intensive. Paradoxically, the Moroccan government gave greater protection to the sectors in which it had the highest comparative advantage, if comparative advantage can be measured by export orientation. For Mexico, Revenga (1994, 1995) also found that the pattern of protection was skewed toward export-intensive sectors. The evidence suggests that developing countries often protect sectors in which they are likely to have a comparative advantage, such as the sectors with a high share of unskilled workers. In this light, it is not surprising that increasing wage inequality is observed in developing countries undergoing trade reforms.

The results presented in Tables 4 and 5 suggest that while changes in trade policy are consistent with Stolper-Samuelson effects, changes in product prices are not. Since Mexico historically protected sectors with more unskilled workers, tariffs fell less in sectors with more skilled workers. In a Stolper-Samuelson framework, this translates into an increase in the relative wage of skilled workers. Below, we explore the extent to which price changes and commercial policy were correlated during the 1980s.

Correlations between price changes and trade policy changes. We present correlations between product price changes and commercial policy in Table 6. We again use two measures of product prices, gross output prices and value added prices. We include an interaction term between tariffs and import-license coverage rates to account for the fact that the application of tariffs and quotas may be correlated across industries. Since the interaction term makes it difficult to evaluate the impact of an individual policy using the reported coefficient estimates, we report the net impact of tariffs and import licenses on product price changes at sample means for tariffs and licenses. To control for possible nonlinearities in how import licenses translate into trade protection, we include two additional indicator variables. In regressions using initial protection levels as regressors, we include a dummy variable for whether import licenses covered 100% of goods produced in the industry, which may indicate cases in which binding quotas were more likely. In regressions using the change in protection levels as regressors, we include a dummyvariable for whether import licenses were completely eliminated in the industry, which may indicate cases in which the reduction in quotas was most significant.

First, consider the correlation between price changes and initial levels of trade protection. For either measure of prices and for either sample of industries, the coefficients on tariffs and import-license coverage rates are negative and statistically significant, while the coefficient on the interaction term between tariffs and import licenses is positive and statistically significant. The net impact of policy changes shows that for gross-output prices (column 1) the impact of initial tariffs on price changes is negative, suggesting that prices fell more in sectors with higher initial tariffs. We obtain a similar result excluding petroleum products, iron, and steel (column 5). When we use value-added prices instead of gross-output prices (columns 3 and 7), the net impact of initial tariffs is reversed. Price changes appear to be higher in industries with higher initial tariffs. It is important to note, however, that in all cases the net impact of initial tariffs on price changes is very small relative to mean price changes for the sample period.

For import licenses, the findings are more consistent. For either price measure, the net impact of initial import licenses on price changes is positive (columns 1, 3, 5, 7), suggesting that relative price changes were higher in sectors with higher initial license coverage. As with tariffs, the net impact of import licenses is very small relative to mean price changes for the period. Despite the statistical significance of the coefficients, initial levels of trade protection appear to say little about the magnitude of price changes over the sample period.

Next, consider the correlation between price changes and changes in protection (columns 2, 4, 6, 8). In all regressions, price changes are positively correlated with the change in tariffs and the change in import-license coverage rates but negatively correlated with the interaction between tariffs and import licenses. These correlations, however, are statistically significant at the 10% level only for gross-output prices, excluding petroleum products, iron, and steel (column 6). Despite the lack of statistical significance, it is worth considering the net impact of changes in trade policy on product price changes. In all regressions, the net impact of either the change in tariffs or the change in import licenses on price changes is positive, providing weak evidence that relative prices rose in sectors that experienced the smallest reductions in trade protection.17 It is again the case that the net impact of changes in trade policy on product prices is very small relative to mean values for price changes over the 1984-90 period.

Regressions of product price changes on tariffs and import licenses provide weak support for the Stolper-Samuelson theorem. We do find that relative price increases were higher in sectors with smaller reductions in trade barriers, which, coupled with our finding that skill-intensive sectors had the lowest reductions in tariffs, would be consistent with Stolper-Samuelson, but this correlation is neither statistically significant nor economically significant. The weak correlation between price changes and changes in trade policy again suggests that aggregate price indices may be poor measures of actual product price changes.

Reallocation of employment and skill intensity. Strictly speaking, Stolper-Samuelson does not require that trade reallocate labor across industries. Relative price changes alone may generate changes in relative factor awards. Nevertheless, if the relative prices of skill-intensive goods rose, we would expect reallocation of labor toward those sectors.

Despite the magnitude of trade reform, there was little employment reallocation across industries. In unreported results, we construct industry shares of national employment by type of worker for 1984 and 1990. The largest relative-employment declines occurred in textiles and apparel (ISIC 32), whose share of total manufacturing employment fell from 0.128 to 0.112, and iron and steel (ISIC 37), whose employment share fell from 0.075 to 0.061; these two industries also experienced a decline in absolute employment over the period. The largest relative employment increases occurred in food products (ISIC 31), whose employment share increased from 0.195 to 0.214, and metal products (ISIC 38), whose employment share increased from 0.266 to 0.280; these two industries experienced absolute employment growth over the period. The changes in industry employment shares resulted almost entirely from changes in rclative bluecollar employment; the distribution of white-collar employment across industries stayed nearly constant over the period.

(Table Omitted)

Captioned as: Table 6.

(Table Omitted)

Captioned as: Figure 3.

Even though changes in industrial composition following trade reform were small, we want to know if the industries that experienced the most employment growth are those that are relatively intensive in the use of skilled labor. To see if there was a shift in employment toward skill-intensive sectors, we ask whether employment growth was higher in sectors that employed a relatively high share of white-collar workers. Figure 3 plots the change in log employment of white-collar workers between 1984 and 1990 against the average log ratio of white-collar to blue-collar employment in 1984 and 1990 by four-digit industry. The relationship between skill intensity and employment growth is positive and statistically significant, suggesting employment growth was higher in skill-intensive sectors.

Figure 4 shows that there is also a positive, though weaker, correlation between bluecollar employment growth and skill intensity. Hence, there is evidence that employment growth was relatively high in skillintensive sectors.

The role of foreign direct investment. We have found that skill-intensive sectors were less protected than non-skill-intensive sectors and thus had smaller reductions in trade barriers (Table 5), which suggests that trade reform may have contributed to an increase in the white-collar-blue-collar wage gap. While there is weak evidence linking changes in trade protection to changes in product prices (Table 6), there is no direct evidence linking relative price increases to skill-intensive sectors (Table 4). One possibility is that changes in commercial policy are a more precise measure of how trade reform affected product prices in Mexico than are changes in aggregate price indices.

A second possibility is that changes in trade policy are correlated with other changes in the Mexican economy, which affected skill-intensive sectors positively. Feenstra and Hanson (1997) suggested one such possibility. They examined whether foreign outsourcing influences the demand for skilled and unskilled labor in Mexico. Outsourcing by U.S. firms to Mexico increased dramatically during the 1980s, following Mexico's relaxation of restrictions on foreign direct investment. At the level of regional manufacturing industries, they found that the relative demand for skilled labor was positively correlated with the change in the number of foreign off-shore assembly plants, which suggests that foreign direct investment may have contributed to increasing wage inequality in Mexico.

To pursue this issue, we examine whether the correlation that we find between trade policy and skill intensity could be proxying for the correlation between foreign investment and skill intensity. In Table 7, we report regressions of the change in the stock of foreign capital on relative whitecollar employment and either the level of or change in trade protection in Mexico.

(Graph Omitted)

Captioned as: Figure 4.

There are no published data on the aggregate stock of foreign capital by industry in Mexico. We use the SECOFI data to generate such a measure. For each plant, the data indicate the share of equity ownership by foreign agents, as well as the value of fixed capital at year-end replacement cost. For a given four-digit industry, we measure the change in the stock of foreign capital as the change in the value of fixed capital in plants that have a foreign equity ownership share of at least 5%.'

One problem with this measure is that the share of foreign equity ownership is only available in 1990, while the value of fixed capital is available in all sample years. Our measure is thus the change in the capital stock over 1984-90 in plants that were at least 5% foreign-owned in 1990.'9

Since foreign investment in Mexican manufacturing increased over the sample period, our measure likely over-estimates the industry change in the foreign capital stock (since more plants were foreign-owned at the end of the period than at the beginning). To examine whether upward bias in the change in the foreign capital stock contaminates the regression results, we tried alternative minimum foreign equity ownership shares (10%, 20%, 30%, 50%) to calculate the change in the foreign capital stock. Results using these alternative measures are very similar to those reported in Table 7.

It does not appear that foreign investment can account for the observed correlation between skill intensity and trade protection. In Table 7, the correlation between the change in the log foreign capital stock and log relative white-collar employment, while positive in all cases, is statistically significant at the.10% level in only one regression. There in also no statistically significant correlation between changes in the foreign capital stock and either the initial level of or the change in tariffs or import-license coverage rates. The only statistically significant correlation between foreign investment and trade protection is a negative relationship between growth in foreign capital stock and the complete elimination of import licenses in a sector (columns 2 and 4). This provides support for tariff-jumping explanations of foreign investment. Foreign firms seem to have avoided sectors from which all quota protection was removed.

(Table Omitted)

Captioned as: Table 7.

There is some contrast between our results and those in Feenstra and Hanson (1997), but this may be attributable to the different definitions of foreign capital used in the two papers. Feenstra and Hanson measured changes in the foreign capital stock by the change in the number of foreign off-shore assembly plants, which follows from their emphasis on foreign outsourcing. The SECOFI sample, however, excludes these plants by design (see notes 4 and 19). Thus, we find a weak positive correlation between foreign investment and skill intensity, excluding plants that are dedicated exclusively to foreign outsourcing.

To summarize our findings, over the period 1984-90 Mexico experienced an increase in the white-collar-blue-collar wage gap; little change in the ratio of whitecollar to blue-collar employment; a fall in trade barriers, with evidence that initial tariff levels and reductions in tariffs were lower in skill-intensive sectors than in nonskill-intensive sectors; higher employment growth in skill-intensive sectors; and higher foreign investment in sectors that maintained import quotas than in sectors that dropped them. We do not find that relative prices rose in skill-intensive sectors, although industry price indices may be noisy measures of product-price changes.

Conclusion

We have examined changes in the relative wages and relative employment of skilled and unskilled workers in Mexico that followed the country's 1985 trade reform. Since the enactment of trade reform, there has been a dramatic increase in the skilled-unskilled wage gap. The relative-wage changes have occurred without large changes in relative employment. One of our primary concerns has been to search for the effects of trade reform on relative prices and the sectoral distribution of labor that could explain the relative-wage changes.

While we do not find a positive correlation between skill intensity and relative product price changes during the 1980s, we do find that Mexico offered relatively high trade protection to low-skill industries. The reduction in trade barriers was most dramatic in these sectors, which suggests that trade reform affected unskilled labor disproportionately. This finding is consistent with a world in which Mexico has an intermediate abundance of skilled labor. The exposure of Mexico to competition from China and other countries that have abundant unskilled labor appears to have contributed to a decrease in the relative wages of unskilled workers.

Footnote:

1On wage dispersion in the United States, see Davis and Haltiwanger (1991), Bound and Johnson (1992), and Katz and Murphy (1992); for international evidence, see Davis (1992).

Footnote:

20ne notable exception is Robbins (1995), who found mixed evidence that trade contributes to increased wage inequality in developing countries. In Chile, Colombia, Costa Rica, and the Philippines, the relative wage of skilled workers rose following trade liberalization, while in Argentina and Malaysia wage inequality was stable or declining during periods of increased openness to trade.

Footnote:

sThe evidence on manufacturing wages does not imply that the minimum wage was nonbinding in all sectors, as manufacturing workers tend to be relatively highly paid. Using household-level earnings data for 1988, Bell found that a substantial number of individuals, most of whom were women working in the informal sector (outside of manufacturing), earned less than the legal minimum wage, which suggests that there is a high degree of non-compliance with minimum wage law in Mexico. Widespread non-compliance is an additional reason that the fall in real minimum wages may not have had substantial labor market consequences.

Footnote:

4In 1989, the SECOFI plants accounted for 29% of total manufacturing employment, as measured by the Mexico Industrial Census. One additional problem with the SECOFI sample of plants is that it contains a balanced panel-we do not observe entry or exit. This is unfortunate, for it means that we cannot explore the relationship between plant mortality and the relative employment of skilled workers.

Footnote:

While the SECOFI sample is representative of Mexican manufacturing overall, there is one important sector that is missing in the data. The sample by design excludes off-shore assembly plants, the socalled maquiladoras. Maquiladoras import from abroad virtually all of the inputs they use to assemble final goods. The plants in the SECOFI sample, in contrast, are more vertically integrated establishments.

Footnote:

5Berman, Bound, and Griliches ( 1994) and Schmitt and Mishel (1996) provided evidence for the United States suggesting that the white-collar-blue-collar classification is a reasonable division of the labor force by skill.

Footnote:

fiA second problem with the blue-collar-whitecollar classification is that neither data source breaks down non-wage compensation by worker type. Our measure of earnings excludes non-wage payments. This omission does not appear to be egregious, as wages account for the majority of payments to workers. In the SECOFI sample, the wage share of total compensation was 0.71 in 1984 and 0.69 in 1990; in the Industrial Census, the wage share of total compensation was 0.73 in both 1985 and 1988.

Footnote:

7The tariff and import-license data we use are from unpublished records of SECOFI, available on request from the authors. These data are discussed extensively in Ten Kate (1992) and Ten Kate and de Mateo V. (1989).

Footnote:

sThe Mexican constitution places myriad restrictions on the ownership of subsoil resources. These restrictions prevented petroleum refineries from being privatized, but some petrochemical firms were transferred to private hands.

Footnote:

9Currie and Harrison (1997) suggested that parastatal firms behave quite differently from private firms.

Footnote:

"Real wages are calculated as nominal annual remuneration per worker or per hour worked, deflated by the June consumer price index in a given year.

Footnote:

I:The total change for the period masks large realwage swings that occurred within the period. Real white-collar wages declined by 16.1% between 1985 and 1988 and then increased sharply, surpassing their 1985 levels by 1989. What the real-wage swings may reflect is unanticipated inflation in 1986 and 1987. While unexpected inflation can perhaps account for a temporary decline in wages for both types of workers, it cannot account for the increase in the whitecollar-blue-collar wage gap over the period.

Footnote:

12There is considerable evidence that rising wage inequality in the United States is due to an increase in the relative demand for skilled labor (Bound and Johnson 1992; Katz and Murphy 1992).

"SOne explanation for the lack of relative employment changes is that the supply of skilled labor is inelastic in the short to medium run. While this idea is plausible, it would require a high degree of immo

Footnote:

bility between manufacturing and other sectors on the part of skilled labor. The secular increase in relative white-collar employment from 1965 to 1985, which is evident in Figure 2, suggests that skilled labor is relatively mobile.

Footnote:

"aThe four-digit industry classification code we use is that in the SECOFI data, which is based on Mexico's Industrial Census. It is similar to the SIC code used in the United States.

l5We can write the log change in the industry gross output price between 1984 and 1990 as

Footnote:

deltalogPPI = vashare * deltalogPVA + matshare * deltalogPMAT + energyshare * deltalogPENERGY,

Footnote:

where PPI is the industry producer price index, PVA is the price of value-added, PMAT is the price index for material inputs, PENERGY is the price index for energy, and A indicates the change over 1984-90. The shares of value-added, materials, and energy in total costs are given by vashare, matshare, and energyshare, where the shares are averaged over the first and last period. Given this equation, we can calculate the log change in the value-added price as

Footnote:

deltalog PVA = ( 1 / vashare) (deltalog PPI - matshare * deltalogPMAT -energyshare * deltalogPENERGY).

Footnote:

16At the plant level, we also find a positive and statistically significant correlation between relative white-collar employment and either the initial tariff level or the change in tariffs.

Footnote:

17Somewhat in contrast to these results, relative prices appear to have increased in sectors that completely eliminated import licenses.

Footnote:

181n 1990, 24.8% of the plants had a foreign equity ownership share of at least 5%. Aitken, Hanson, and Harrison (1997) showed that most foreign investment in Mexico is from the United States.

19Feenstra and Hanson (1997) measured the change in the foreign capital stock as the change in the number of off-shore assembly plants, which are the agents of foreign outsourcing in Mexico, at the region (and not industry) level. The SECOFI sample

Footnote:

of plants that we use excludes off-shore assembly plants by design (see note 4). Thus, we measure the change in the foreign capital stock using a nonoverlapping sample of plants (and we measure foreign capital at the industry rather than region level).

Reference:

REFERENCES

Reference:

Aitken, B., Gordon Hanson, and Ann Harrison. 1997. "Spillovers, Foreign Investment, and Export Behavior." Journal of International Economics, Vol. 43, No. 1/2, pp. 103-32.

Bell, Linda A. 1997. "The Impact of Minimum Wages in Mexico and Columbia." Journal of Labor Economics, Vol. 15, No. 3, pp.S102-S126.

Reference:

Berman, Eli, John Bound, and Zvi Griliches. 1994. *Changes in the Demand for Skilled Labor within U.S. Manufacturing: Evidence from the Annual Survey of Manufacturers." Quarterly Journal of Economics, Vol. 109, No. 2, pp. 367-98. Borjas, George J. 1994. "The Economics of Immigration." Journal of Economic Literature, Vol. 32, No. 4, pp. 1667-1717.

Bound,John, and George Johnson. 1992. "Changes in the Structure of Wages in the 1980s: An Evaluation of Alternative Explanations." American Economic Review, Vol. 82, No. 3 (June), pp. 371-92.

Reference:

Has Wagg, Michael 1., and Dispersion Grown in Mexico? Is It the Incidence of Reforms or the Dispersion Grown in Demand fo? Is It the Incidence of Reforms or the Growing Demand for Skills?, Journal of Development and Ann Harrison. 51997. "Sharing the

Labor in Morocco." Journal of Labor Economecs, pp. 99-117.

Currie,Janet, and Ann Harrison. 1997. 'Sharing the Costs: The Impact of Trade Reform on Capital and Labor in Morocco." Journal of Labor Economics, Vol. 15, No. 3, pp. S44-S72.

Reference:

Davis, Stephen J. 1992. "Cross-Country Patterns of Change in Relative Wages. In Olivier Blanchard and Stanley Fischer, eds., NBER Macroeconomics Annual 1992. Cambridge, Mass.: MIT Press. Davis, Stephen J., and John Haltiwanger. 1991. "Wage Dispersion between and within U.S. Manufacturing Plants, 1963-86." Brookings Papers on Economic Activity: Microeconomics 1991. Washington, D.C.

Reference:

Ethier, Wilfred J. 1984. "Higher Dimensional Issues in Trade Theory." In Ronald W. Jones and Peter B. Kenen, eds., Handbook of International Economics. Amsterdam: North-Holland.

Feliciano, Zadia. 1993. "Workers and Trade Liberalization: The Impact of Trade Reforms in Mexico on Wages and Employment." Mimeo, Harvard University.

Reference:

Feenstra, Robert C., and Gordon H. Hanson. 1996a. "Globalization, Outsourcing, and Wage Inequality." American Economic Review Papers and Proceedings, Vol. 86, No. 2, pp. 240-45.

. 1996b. "Foreign Investment, Outsourcing, and Relative Wages." In Robert C. Feenstra, Gene M.

Reference:

Grossman, and Douglas A. Irwin, eds., Political Economy of Trade Policy: Essays in Honor of Jagdish Bhagwati. Cambridge: MIT Press, pp. 89-127. 1997. "Foreign Direct Investment and Relative Wages: Evidence from Mexico's Maquiladoras." Journal of International Economics, Vol. 42, No. 3/4, pp. 371-94.

Reference:

Katz, Lawrence, and Kevin M. Murphy. 1992. "Changes in Relative Wages, 1963-1987: Supply and Demand Factors." QuarterlyJournalofEconomics, Vol. 107, No. 1 (February), pp. 35-78. Krueger, Alan B. 1995. "Labor Market Shifts and the Price Puzzle Revisited." Mimeo, Princeton University.

Lawrence, Robert Z., and Mathew J. Slaughter. 1993. "Trade and U.S. Wages: Great Sucking Sound or Small Hiccup?" Brookings Papers on Economic Activity, Vol. 2, pp.161-227.

Reference:

Leamer, Edward. 1993. "Wage Effects of a U.S.Mexico Free Trade Agreement.' In Peter M. Garber, ed., The Mexico-U.S. Free Trade Agreement. Cambridge, Mass.: MIT Press.

Learner, Edward E. 1998. "In Search of StolperSamuelson Linkages between International Trade and Lower Wages." In Sharon M. Collins, ed., Imports, Exports, and the American Worker. Washington, D.C.: Brookings Institution Press, pp. 141-203. Lopez-de-Silanes, Florencio. 1997. "Determinants of Privatization Prices." Quarterly Journal of Economics, Vol. 112, No. 4, pp. 965-1026.

Reference:

Revenga, Ana. 1997. Employment and Wage Effects of Trade Liberalization: The Case of Mexican Manufacturing.' Journal of LaborEconomics, Vol. 15, No. 3, pp. S20-S43.

1995. "North American Integration and Factor Price Equalization: Is There Evidence of Wage Convergence between Mexico and the United States?" Mimeo, World Bank.

Robbins, Donald. 1995. "Trade, Trade Liberalization, and Inequality in Latin America and East Asia: Synthesis of Seven Country Studies." Mimeo, HIID.

Reference:

Schmitt, John, and Lawrence Mishel. I991. Did Trade Lower Less-Skilled Wages during the 1980s? Standard Trade Theory and Evidence." Mimeo, Economic Policy Institute.

Stolper, Wolfgang, and Paul A. Samuelson. 1941. "Protection and Real Wages." Review of Economic Studies, Vol. 9, No. 1 (November), pp. 51-68. Ten Kate, Adrian. 1992. "Trade Liberalization and Economic Stabilization in Mexico: Lessons of Experience." World Development, Vol. 20, No. 5 (May), pp.

Reference:

659-72.

Ten Kate, Adrian, and Fernando de Mateo Venturini. 1989. "Aperatura Comercial y Estructura de la Proteccion en Mexico: Un Analisis de la Relacion

Reference:

entre Ambas.' Comercio Exterior, Vol. 39, No. 6 (June), pp. 31-43.

Wood, Adrian. 1994. North-South Trade, Employment, and Inequality. Oxford: Clarendon.

Author Affiliation:

GORDON H. HANSON and ANN HARRISON*

Author Affiliation:

*Gordon H. Hanson is Associate Professor, Department of Economics and School of Business Administration, University of Michigan, and Faculty Research Fellow at the National Bureau of Economic Research. Ann Harrison is Assistant Professor, Graduate School of Business, Columbia University, and Faculty Research Fellow, NBER. This research was sponsored by the World Development Report 1995 and World Bank research project 678-29, *Technology Spillovers, Agglomeration, and Direct Foreign Investment." Hanson acknowledges financial support from the Center for the Study of Western Hemispheric Trade at the University of Texas.