As an example of this “factor-content” type of study, Keith Maskus (1985) attempted to ascertain the implied factor endowments in the United States by examining the net exports and net imports of the services of five broad categories of factors of production. For the year 1958, he determined that the factor-abundance rankings were (1) scientists and engineers (most abundant because largest net export), (2) nonproduction workers except scientists and engineers, (3) human capital (reflecting mainly education), (4) production labor, and (5) physical capital (least abundant because largest net import). For 1972, the rankings for the first three factors were the same, but physical capital and production labor switched positions. Along the same line, a very ambitious study of factor abundance and net export of factor services was carried out in 1987 by Harry Bowen, Edward Leamer, and Leo Sveikauskas. This study examined 12 different factors of production in 27 different countries to predict the implied factor abundances (and hence the implied Heckscher-Ohlin trade flows). Their general qualitative results for six countries are given in Table 1. A plus sign indicates that, through trade, the country was a net exporter of the services of that factor and thus was “revealed” to be abundant in that factor (because more of the factor’s services were supplied in production than were demanded by the country through its consumption pattern); a minus sign indicates that the country was a net importer of a factor service and therefore that the country was relatively scarce in that factor (because more of the services of the factor were demanded through the country’s consumption pattern than were supplied in domestic production). For the United States in 1967 (the test year), the Leontief paradox did not seem to exist when allowance was made for factors other than capital and labor. The United States exported the services of capital (as general intuition would expect), as well as the services of professional/technical workers, agricultural workers, and arable land. Eight other services were imported and thus scarce in the United States. With respect to other countries, Table 1 indicates that Canada exported the services of capital, agricultural workers, and various kinds of land while importing different kinds of labor services. Germany and Japan imported capital and land services while exporting the services of professional/technical workers and managerial workers. (The latter two worker categories had the largest page 160net exports for Germany and Japan in the underlying data for Table 1.) For the developing countries in the table, Mexico was a net exporter of land-based factor services and various labor services while a net importer of the services of capital (these results are not surprising and thus are not paradoxical), while the Philippines imported capital and skilled-labor services and exported some “lower-skilled” labor services (again, results that are not surprising). A relatively recent addition to the list of factors that may serve as a source of comparative advantage is the financial sector. Svaleryd and Vlachos (2005) propose that underlying technological and organizational differences among industries cause them to differ in their need for external financing. If the assumption is made that the services provided by the financial sector are relatively immobile across national boundaries, patterns of industrial specialization should be influenced by the relative endowments of financial development. Using a broad group of developed countries for their empirical analysis, Svaleryd and Vlachos find that countries with well-functioning financial systems tend to specialize in industries highly dependent on external financing. In fact, their results show that differences in financial systems are more important than differences in human capital. Their results support the Heckscher-Ohlin-Vanek model. In a more recent study of comparative advantage in Brazil based on the H-O model, Muriel and Terra (2009) estimated the sources of comparative advantage revealed by its international trade. Two time periods were used in the analysis, the first (1980–1985) just prior to trade liberalization and the second (1990–1995) following trade liberalization. The authors employed estimates of both the relative factor intensities associated with net exports, as well as direct estimates of the factor content of international trade. The results led to the conclusion that Brazilian trade revealed comparative advantages in the use of unskilled labor, capital, and land, but not in goods relatively intensive in skilled labor. Given the actual relative abundance of especially unskilled labor and land, these results tend to support the basic H-O model. However, even the seemingly more friendly results for the Heckscher-Ohlin theorem (such as the Bowen-Leamer-Sveikauskas type) have been called into question. Remember the nature of factor-content tests—they are calculating whether a factor’s services, on balance, are being exported or imported by a country, and, if exported (imported), the conclusion is drawn that the country is abundant (scarce) in that factor. Then a judgment is rendered as to whether this abundance or scarcity fits with general intuition and thus whether H-O generally seems to have been validated. But a point of attack with respect to these more recent studies is that the calculated relative abundance (scarcity) of a factor may not match the actual relative abundance (scarcity) that can be ascertained by the use of different, independent data. In other words, for example, a factor might be calculated as being “relatively abundant” in a country because there is positive net export of the factor’s services, but independent data on actual endowments of the factor in that country and other countries would show the factor to be relatively scarce in that country. The independent data in these comparisons would consist of measures such as a country’s share of the world’s actual endowments of capital and labor (such as used in the Gourdon study discussed on pages 157–58). Indeed, Maskus (1985) found that, in tests comparing the United States with other countries, the actual U.S. relative abundances and scarcities matched the predicted abundances and scarcities two-thirds of the time in one test, one-third of the time in another test, and only one-sixth of the time in a third test. These results are hardly reassuring. Bowen, Leamer, and Sveikauskas also examined their 12 factors in 27 countries to see if the predicted relative abundances and scarcities matched the actual data. Only in 4 of the 12 factors was there a predictive success rate of 70 percent or more across the countries, and only 7 of the 12 factors were successfully predicted for the United States.page 161

Technology, Productivity, and “Home Bias”

In other work, Daniel Trefler (1995) examined, for the year 1983, data for 33 countries that accounted, at the time, for 76 percent of world exports and 79 percent of world GNP. Nine factors of production were considered: (1) capital; (2) cropland; (3) pasture land; (4) professional and technical workers; (5) clerical workers; (6) sales workers; (7) service workers; (8) agricultural workers; and (9) production, transport, and unskilled workers. His first test of the data with respect to whether net factor flows through trade matched the expectation from actual endowments produced disappointing results. However, he noted that, while the Heckscher-Ohlin theorem and the usual H-O tests assume that technology/productivity in any given industry is identical across countries, this assumption is very unrealistic. In fact, he found that, across countries, there tended to be systematic differences in productivity levels. For example, Panama’s industries tended to be about 28 percent as productive as U.S. industries, and Finland’s about 65 percent as productive as U.S. industries. Thus the United States had close to four times as much “effective” labor (1/0.28 = 3.6) in comparison with Panama as standard labor force measures would indicate [or, alternatively, Panama had only about one-quarter (0.28) as much “effective” labor relative to the United States as standard labor force measures would indicate]. Hence, Trefler adjusted his data to reflect these differences. Panama’s factor endowments, when compared with those of the United States, were thus only 28 percent of the actual level, Finland’s only 65 percent of the actual level, and so on. These adjusted factor endowments were then used in the comparison of factor endowments with the flows of factor services through trade. This type of adjustment was also accompanied by another adjustment—Trefler felt that, for whatever reason, consumers in any given country had a preference for home goods over foreign goods, and this preference needed to be incorporated into the examination of trade flows. (This adjustment was necessitated in Trefler’s view because standard Heckscher-Ohlin tests seemed to predict a much larger volume of trade than actually occurred—Trefler’s “home bias” was designed to account for this difference.) With both of these adjustments in place, Trefler more satisfactorily “explained” the existing trade patterns than he had done without the adjustments and than had been done with regard to explaining trade in the various previous studies. The thrust of his work, then, is that actual trade differs from Heckscher-Ohlin-predicted trade because technology/productivity levels differ across countries and because consumers have a general preference for home goods. These two factors should be taken account of in considering the determination of trade flows, because H-O by itself does not do a very good job of explaining the flows. In 2002, Conway confirmed Trefler’s conclusion regarding these mysteries but suggested alternate explanations. Further, Feenstra and Hanson (2003) indicate that it is standard in Heckscher-Ohlin models to assume that exports are produced entirely by combining domestic factors of production with domestically produced intermediate inputs. They suggested that in a globalized world this assumption is wrong and that accounting for trade in intermediate inputs can help resolve the mystery of the missing trade. In a survey of the relevant literature, Helpman (1999) strongly questioned the home-bias assumption of Trefler. However, Helpman and others have increasingly explored the notion of productivity/technology differences across countries, and the view that such differences are important is becoming more generally accepted. Helpman (1999, p. 133) noted that recent work suggests that “allowing for differences in techniques of production can dramatically improve the fit of factor content equations. Now economists need to identify the forces that induce countries to choose different techniques of production.” Further, Reeve (1998, cited in Helpman, 1999) and Davis, Weinstein, Bradford, and Shimpo (1997) have found, in different settings, that factor endowments do a decent job of predicting the location of particular types of industries, if not of predicting trade itself. That is, while a country that is relatively abundant in a particular factor may not have been found by page 162existing estimation techniques to be exporting the services of that factor, the country’s production structure emphasizes goods that are relatively intensive in that factor. This finding would be consistent with the Heckscher-Ohlin general notion that a country will specialize in the production of goods utilizing the country’s abundant factors, even if the next step of linking production to trade has not been empirically established. 

In another study of the issue of productivity/technology differences across countries, Schott (2003) looked beyond the traditional industry-based data to examine the actual subsets of goods produced in a country. His contention was that too many traditional attempts to find empirical support for the Heckscher-Ohlin idea that a country’s endowments determine production and trade have used the overly restrictive assumption that all countries produce a given category of goods using the same technology. Schott argued that within the category of electronics, for example, the more labor-abundant Philippines may be producing portable radios while the more capital-abundant Japan is manufacturing semiconductors and satellites. The Schott approach permits a given sector’s output to vary with a country’s endowments, allowing countries to move in and out of sectors as they develop. The estimation focuses on finding the capital per labor cutoffs, where changes in the subset of output produced take place and using this to group countries according to the subset of goods they produce. Schott’s technique highlights the potential differences in the choices of output within industries across countries and suggests that moving beyond the standard, industry-level data is necessary to test Heckscher-Ohlin hypotheses. By designating products within the category of manufacturing by the relative capital intensity of their production, Schott found strong support for the idea that country product mix varies with relative endowments. In addition to pointing out the need to move beyond industry-level data to test international specialization, this analysis suggests that the technique can be used to explore violations of Heckscher-Ohlin assumptions assigned to home bias in trade (e.g., Trefler, 1995). A newer approach to empirical verification of the Heckscher-Ohlin hypothesis suggests that the quality of available data has been a problem in the testing process. Fisher and Marshall (2007) found overwhelming support for the Heckscher-Ohlin paradigm using Organization for Economic Cooperation and Development (OECD) input-output data on the technology of production across 33 countries in the year 2000. Fisher and Marshall used this rich data set to create a technology matrix for each country. By observing a specific output vector for a country, they defined the country’s “virtual endowment” as the factor services that would be needed to produce that output in view of the implied level of technology of the country. While their innovative empirical approach is new, their results suggested that a more careful treatment of technological differences between countries can explain a substantial portion of Trefler’s missing trade. Fisher and Marshall hope that the availability of new and better data will allow economists to reexamine past studies that questioned the validity of the Heckscher-Ohlin theorem using differences in technology and restore the profession’s confidence in this important paradigm. Finally, some new work has attempted to test the relative strength of the Heckscher-Ohlin analysis and other approaches in explaining actual trade patterns. For a summary of this work, see the “In the Real World” box on pages 163–64. In overview, while the Heckscher-Ohlin theorem is logical, straightforward, and seemingly a commonsense hypothesis, there have been difficulties in demonstrating the validity of the theorem in practice. As empirical work continues, however, we are beginning to get a better picture of what the analysis does and does not seem to explain. Supplementary forces besides factor endowments and factor intensities increasingly need to be considered, however, and some of these factors are dealt with in the next chapter.page 163 This chapter began with Leamer and Levinsohn’s call for a reorientation of empirical work in international trade. Their call was summarized in the statement “Estimate, don’t test.” In other words, draw on relevant empirical evidence to improve understanding of a phenomenon rather than trying to devise a perfect, yes-or-no test of a theorem explaining the phenomenon. A review of empirical work on the H-O-V model by Davis and Weinstein (1996) suggested that the accumulation of results is more influential than any individual study. Each study sheds new light on the circumstances under which a particular theory is useful. Davis and Weinstein suggested that researchers should both estimate and test. Their suggested criterion: Does the test narrow the range of sensible application of the theory? An affirmative reply indicates that the test is useful.

Heckscher-Ohlin and Comparative Advantage

Two 2010 articles in the international economics literature focused on examining the contribution of a Heckscher-Ohlin explanation of country specialization compared to other explanations of the pattern of any given country’s trade. These recent additions to the long stream of Heckscher-Ohlin tests indeed suggest that the H-O model is relevant to the real world, but they also point out that the H-O analysis does not produce the sole explanation of the structure of trade and specialization. David Chor (2010) of Singapore Management University, in seeking to identify the underlying sources of comparative advantages of countries, looked at the patterns of specialization in bilateral trade of any given country with its various trading partners. The dependent variable in his regressions was the value of exports of a country to a specific trading partner in a specific industry. He employed a sample of 83 countries (and thus each country had 82 potential trading partners) and 20 manufacturing industries for the year 1990. Not all partners engaged in trade in all the industries, but Chor still ended up with over 45,000 data points. For explanatory variables (independent variables) of the amount of bilateral trade, he utilized distance and geographic variables (such as whether there was a common border, a common language, and so on), the Heckscher-Ohlin variables of human capital per worker and physical capital per worker, and institutional variables representing the level of a country’s financial development, the country’s dependence on external sources of finance (foreign investment or borrowing), the flexibility in a country’s labor market, and various aspects of the country’s legal system. Important results that Chor obtained were that, predictably, distance was significant—other things being equal, a halving of distance between partners led to more than a doubling of their bilateral trade. Pertaining directly to the focus of this chapter, there was verification of the importance of Heckscher-Ohlin, in that relatively skilled-labor-abundant countries had greater exports in the more skilled-labor-intensive industries. Also, if a country had a relatively greater endowment of physical capital per worker, it tended to have greater exports in physical-capital-intensive industries, other things being equal. For the institutional variables, one result was that countries with higher levels of financial development were more successful in exporting goods from industries that had greater amounts of external funding. Of further interest, Chor also attempted to focus on welfare and found that, as a general rule, if the human capital variables’ separate influences on specialization were taken away, country welfare on average would fall by 3.1 percent; for physical capital by itself, the fall in welfare would be 2.8 percent. These welfare declines were larger than would occur, for example, with the institutional variables of the country’s financial development and its labor market flexibility. Peter Morrow (2010) of the University of Toronto studied the role of the Heckscher-Ohlin theorem in explaining trade patterns in a somewhat different manner. He noted that, in theory, the Ricardian model (of Chapters 2–4 in this text) indicates that countries export goods from industries that have the greatest relative productivity (or, in modern terminology, the greatest relative total factor productivity [TFP]) compared to trading partners. The Heckscher-Ohlin analysis, on the other hand, ignores productivity differences across countries by assuming that the production function in any given industry is the same everywhere; the analysis then attributes comparative advantage to relative factor endowments and relative factor intensities. In the real world, page 164though, suppose that an exporting industry indeed uses the relatively abundant factor in a relatively intensive way in the production process but that the factor is also relatively more productive compared to the trading partner countries—in this mixed kind of setting, do you consider comparative advantage to be a result of the Heckscher-Ohlin model or of the Ricardian model? To get at this question of the relative importance of the Ricardian and Heckscher-Ohlin models, Morrow empirically tested trade patterns and their causes with a sample of 20 countries (both developed countries and developing countries), 24 manufacturing industries, and 11 years (1985–1995). The formulation of his theoretical model and of his precise empirical tests is beyond the scope of this text, but the essence is that the variability of the relationship between productivity (TFP) and skill intensity across industries was decomposed into the variation due to relative factor abundance and the variation due to Ricardian-type international relative productivity differences. Important Morrow results were that countries with a relative abundance of skilled labor specialized in the production of skilled-labor-intensive goods and that relative productivity in any given industry was uncorrelated with the relative skill intensity employed in that industry. In other words, although relative productivity in an industry is certainly, in and of itself, a source of comparative advantage, that result is not being intermingled with Heckscher-Ohlin relative skill intensity when empirical tests are run. Therefore, separation of the Ricardian model from the Heckscher-Ohlin model is possible, and Morrow concluded that, empirically, both the Ricardian and the Heckscher-Ohlin models are useful in explaining specialization patterns of countries. Finally, the H-O explanation of comparative advantage was somewhat more important than the Ricardian explanation, in that a given variation (one standard deviation) in relative factor abundance was judged by Morrow to be 1.6 to 2.3 times as powerful as a variation of one standard deviation in Ricardian TFP in explaining patterns of production in a country. Sources: David Chor, “Unpacking Sources of Comparative Advantage: A Quantitative Approach,” Journal of International Economics 82, no. 2 (November 2010), pp. 152–67; Peter M. Morrow, “Ricardian-Heckscher-Ohlin Comparative Advantage: Theory and Evidence,” Journal of International Economics 82, no. 2 (November 2010), pp. 137–51. 

Heckscher-Ohlin and Income Inequality

In recent years, a debate has been taking place in the United States and in western Europe over a phenomenon that is associated with the Heckscher-Ohlin analysis. While the debate is not always couched in H-O terms (the average person on the street, unlike you, is not an expert on Heckscher-Ohlin!), it involves an important implication of that analysis and has also been the subject of empirical tests. The phenomenon is the growing income inequality that has been occurring in the developed countries.3 It is clear that income inequality in the United States has been increasing in recent years. For example, U.S. Census Bureau4 data indicate that the share of household income (in constant 2013 dollars) received by the lowest 20 percent of households fell from a high of 4.3 percent in 1976 to 3.2 percent in 2013, while the top 20 percent of households experienced an increase in their share of income from 43.7 percent in 1976 to 51.0 percent in 2013. During the same period, the middle 20 percent (40–60) of households experienced a drop from 17.0 percent of income to 14.4 percent, while the second quintile page 165(20–40) faced a decline from 10.3 percent to 8.4 percent and the share of the fourth quintile (60–80) dropped from 24.7 percent to 23.2 percent. With respect to average real household income by household in dollar terms from 1976 to 1913, the changes were as follows: lowest 20 percent—minus 0.7 percent; second lowest 20 percent—plus 6.2 percent; middle 20 percent—plus 10.7 percent; second highest 20 percent—plus 21.7 percent; and top 20 percent—plus 52.5 percent. (Above that level, the Census Bureau also gives the change for the top 5 percent of households—plus 74.2 percent.) Thus, over this long period, the second lowest quintile’s income grew faster than the first’s, the third’s faster than the second’s, and similarly on up the scale. There has been a general widening of the income distribution; it is not correct, looking at these particular data, to say that it’s only the very top income earners who have gained at the expense of everyone else; the only group that lost was the lowest group, but the distribution widened all the way up the scale. This same general broad widening of the entire income distribution also occurs if we look at the Census Bureau real household income data pertaining to the “Great Recession” of recent years and the beginnings of recovery. In those years, all quintiles experienced a decrease in household income, but the decreases were smaller as one goes up the income scale. The changes from 2007 to 2013 were as follows: lowest 20 percent of households—decrease of 10.2 percent; second lowest 20 percent—decrease of 7.8 percent; middle 20 percent—decrease of 6.8 percent; second highest 20 percent—decrease of 6.0 percent; highest 20 percent—decrease of 1.9 percent; and top 5 percent—decrease of 0.1 percent.

It should be noted, however, that measurements of the amount of growth in income inequality in the United States over the last several decades depend to a considerable degree on the estimates/measurements of income. The above Census Bureau measures of income include money income received by households—including income from work, interest, dividends, and cash transfer payments of various kinds such as unemployment compensation and welfare payments (although not non-cash transfers such as food stamps), but not capital gains. Another set of data by other researchers, for example, concentrates on the top 1 percent of income earners in the United States, and measures income before deduction for income taxes and prior to any transfer payments and including capital gains, and finds that the percentage of income received by the top 1 percent rose from 9 percent in 1976 to 20 percent in 2011.5 This result seems hard to reconcile with the Census Bureau figures that had the top 5 percent earning 22.2 percent in 2013 (and 22.3 percent in 2011). The treatment of taxes, transfers, and capital gains obviously makes a difference. Nevertheless, by any measure, and many studies have been conducted, there has been a clear rise in the degree of income inequality in the United States during the past several decades. Finally, figures pertaining to wealth (i.e., the net worth of households = the value of all assets minus the value of all liabilities) rather than income show increased inequality as well. For example, the Pew Research Center in 2011 indicated that in 2009 the median net worth of the typical household headed by an older individual (65 years and older) was $170,494, while the median net worth of a household headed by a younger individual (under 35 years) was $3,662. Although the older household would clearly be expected to have greater net worth because of a lifetime of accumulation and earnings, the 2009 ratio of 47 to 1 ($170,494 ÷ $3,662 = 47) of the two typical households had been a ratio of 10 to 1 in 1984. To put this difference in clearer perspective, the net worth of the median older household had risen (in real terms) by 42 percent since 1984, while the net worth of the median younger household had decreased by 68 percent from the 1984 level.6page 166 Taking a longer view, Wojciech Kopczuk suggested that there are four different ways of measuring wealth distribution in the United States and the share held by top groups—by carrying out a sample survey weighted toward wealthy individuals, by looking at estate tax records, by estimating wealth from the flow of capital income of individuals, and by examining lists of the wealthiest individuals such as provided by Forbes magazine.7 Kopczuk is able to assemble data over the long haul on capital income (1913–2012) and the estate tax (1916–2000). He concludes from this long-run information that the share of U.S. wealth held by the top 10 percent of individuals has fluctuated between 65 and 85 percent of total wealth, the share of the top 1 percent between 20 and 45 percent, and the share of the top 0.1 percent from a little below 10 percent to 25 percent. Importantly, after 1980, one series (estimating wealth from capital income) showed a dramatic increase for these top groups through 2012 while the other series (estimating wealth from estate tax collections) stayed relatively constant through 2000. Note, though, that because of estate tax changes and increases in the size of estate needed in order to be required to pay federal estate tax, that series was not really able to be calculated usefully after 2000. On the basis of this work and on other studies, however, it is fair to say that wealth as well as income inequality has increased in recent decades in the United States.8 Meanwhile, in western Europe, where wage rates are less flexible than in the United States due to institutional factors such as strong labor legislation and prominent unions, the increased inequality has registered itself not so much through widening wage differentials as through increased unemployment rates (with consequent loss of income). In 1973, the unemployment rate for developed countries in Europe was 2.9 percent, but unemployment averaged 9.3 percent from 1983 to 1991 (Freeman, 1995, p. 18) and in mid-1999 was at double-digit levels in Belgium (12.7 percent), France (14.2 percent), Germany (10.5 percent), Italy (12.0 percent), and Spain (16.1 percent) (The Economist, September 11, 1999, p. 114). For the same five countries, the mid-2015 figures were 8.6 percent, 10.3 percent, 6.4 percent, 12.4 percent, and 22.5 percent, respectively (The Economist, July 11, 2015, p. 80). This phenomenon of rising inequality clearly has generated considerable tension and dissatisfaction. It was an important factor in the emergence of the Occupy Wall Street movement in the United States in 2011, although other matters lay behind that movement as well. To many observers, a disturbing factor about this rise in inequality is that it has been occurring at the same time that the United States and the world as a whole have been becoming more open to international trade. In 1970, the ratio of U.S. exports to U.S. gross domestic product (GDP) was 5.5 percent, while that of imports to GDP was 5.4 percent; by 1980 these ratios had reached 10.0 percent for exports and 10.6 percent for imports; and in 2014 the figures were 13.5 percent and 16.6 percent, respectively (Economic Report of the President, February 1999, pp. 326–27; U.S. Department of Commerce, Bureau of Economic Analysis, Survey of Current Business, October 2015, Table 1.1.5). And, in particular, rapid growth has been occurring in imports into the United States and western Europe from developing countries. These imports into the United States were 14 percent page 167 of all imports in 1970 but rose to 35 percent by 1990, 49 percent in 2000, and 54 percent in 2013. For the European Union countries, the figures were 5 percent of all imports in 1970, 18 percent in 1990, 24 percent in 2000, and 29 percent in 2013.9 These kinds of increases in trade in general and especially in imports from developing countries suggest that there may be a link between them and the rising inequality. Clearly, the Heckscher-Ohlin and Stolper-Samuelson theorems can provide such a link. As was developed in Chapter 8, the H-O theorem postulates that a country will export goods intensive in the country’s relatively abundant factor of production and will import goods intensive in the country’s relatively scarce factor of production. Extending this pattern of trade to income distribution considerations, the Stolper-Samuelson theorem indicates that, with trade, the real return to the abundant factor rises and the real return to the country’s scarce factor falls. In the context of an expanded H-O framework for the United States where labor is divided into relatively skilled labor and relatively unskilled labor, such as the framework utilized in empirical tests discussed earlier in this chapter, the implication is that the real incomes of highly skilled workers (who tend to be in the upper portions of the income distribution) will increase with expanded trade and the real incomes of less skilled workers (who tend to be in the lower portions) will decrease. Probing further into the matter, there is an increasing body of information that supports the idea that increased levels of education have a positive relative effect on workers’ incomes. In the United States, full-time working college graduates aged 25–32 on average annually earn about $17,500 more than peers who have only a high school diploma. The actual amount, of course, depends on the degree pursued and the characteristics of the individual involved.10 Other research carried out by the Organization for Economic Development suggests that, across 25 OECD countries, the net long-term returns to having a tertiary degree instead of an upper secondary degree are greater than $175,000 for men and more than $110,000 for women.11 Other studies showed that greater levels of higher education benefit the public sector as well through higher levels of productivity, tax revenues, and social contributions that exceed public investment costs. Finally, additional work has suggested that the return to higher education in developing countries is greater than in higher-income countries.12 The critical question facing trade economists concerns the extent to which the rising imports are the cause of the increased wage inequality.13 Most studies of the relationship have found trade to be a factor accounting for the increased inequality but not a major page 168factor. For example, Borjas, Freeman, and Katz (1992, discussed in Burtless, 1995, p. 808) calculated that from 8 to 15 percent of the 1980–1988 rise in the wage differential between college and high school graduates in the United States was attributable to the combined effects of trade and immigration into the United States, with most of this 8 to 15 percent due to the trade component. Other studies also found modest effects, and Richard Freeman (1995, p. 25) summarizes by stating that “factor content analysis studies indicate that trade can account for 10–20 percent of the overall fall in demand for unskilled labor needed to explain rising wage differentials in the United States or rising joblessness in Europe.”

Trade and Income Inequality in a Less Developed Country: The Case of Mozambique

As less developed countries are influenced by the policy prescriptions of the International Monetary Fund and adopt trade strategies that favor export trade, the impact of these policies on the countries needs to be analyzed. A considerable amount of effort has been devoted to examining the link between trade and income inequality at the national level in developed and developing countries. To this point, much less attention has been paid to regional differentiation within the countries. Julie A. Silva (2007) uses Mozambique as a test case for examining the regional differences in inequality. Mozambique is an interesting case because there is a long history of uneven development. The area to the south of the Zambezi River is relatively more developed, while the north is more isolated. The two regions are governed by the same economic development policies but differ in terms of climate, infrastructure, culture, and levels of development. Most of the population lives in rural areas, and 90 percent of rural Mozambicans are involved in agriculture. Households can produce cash (export) crops, domestically traded vegetable crops, or both. Using data from government censuses conducted between 1996 and 2000, cross-sectional analysis was possible. There are significant differences in the trade orientation between the two regions. In the northern districts, 31 percent of agricultural households produce cash (export) crops, while only 13 percent of southern households do. Both regions are similar (18 percent in the north and 13 percent in the south) in terms of growing domestically traded vegetable crops. The primary focus of the analysis is on the impact of the trade variables on income inequality. In the south, Silva’s results suggest that only the domestic trade orientation has a significant impact, and it increases inequality. The coefficient of the export orientation variable is not significant, but this is not surprising because the southern households have a long tradition of resistance to growing cash crops. In addition, the local markets for domestically traded vegetable crops are well developed in the south. The indication that higher levels of trade would increase the income inequality contradicts the standard H-O predictions. In the northern districts, her cross-section analysis suggests that the export orientation has a negative and statistically significant impact on inequality. The domestic trade orientation coefficient was positive (contributing to an increase in inequality as in the south) but was not significant. In both regions, the variables measuring physical and human capital were significant. In addition, the percentage of households that were female-headed was also significant but also curious. In the south, a higher percentage of female-headed households contributed to greater income inequality. In the north, the coefficient on female-headed households was negative and significant. While this result contradicts the expectation that economic marginalization of women in sub-Saharan Africa would lead to higher inequality, the explanation may lie in the matrilineal social system of three of the largest tribes in northern Mozambique. In these tribes, wealth and land tenure pass through the female line. Overall, this study suggests the need to move beyond the traditional H-O framework focusing on the national level as we study income inequality in developing countries. The framework seems to be too narrow to capture the dynamics within and across regions in the less developed countries. The case of Mozambique suggests that differences in history, culture, and capital across regions may be as critical as the economic forces in explaining income inequality. Source: Julie A. Silva, “Trade and Income Inequality in a Less Developed Country: The Case of Mozambique,” Economic Geography 83, no. 2 (April 2007), pp. 111–36. page 169 The findings on this relatively minor role for trade have been disputed by other economists.14 The most ardent advocate of the view that the increased trade with the developing countries has led to the increased income inequality in developed countries has been Adrian Wood of the University of Sussex (see Wood, 1991, 1994), who contends that the usual estimates of the decrease in demand for unskilled labor in the developed countries are significant underestimates. In essence, he claims that replacing labor-intensive imports from developing countries with developed-country production would require considerably more low-skilled labor than is generally thought to be the case. As a response to these arguments, economists have usually made several major points, which we summarize here: A major consideration brought out in the discussion is that if trade is operating in accordance with the Stolper-Samuelson theorem to generate the increased inequality, then the prices of low-skill-intensive goods would also be falling. This follows because factor prices in the Heckscher-Ohlin analysis move in the same direction as the prices of the goods that the factors are used to produce. However, studies of relative goods’ price movements in recent years do not find a pronounced decline in the prices of unskilled-labor-intensive goods relative to skilled-labor-intensive goods. Thus, the trade explanation for the increased inequality lacks a mechanism that is consistent with trade theory. The rise in the demand for skilled labor relative to unskilled labor in the developed countries has not been confined to the traded goods industries—indeed, it has occurred across almost all industries. If the increased inequality were purely a trade phenomenon, the fall in the relative price of unskilled labor would cause the nontraded goods industries to substitute toward the use of relatively more unskilled labor, which is the opposite of what has happened. Rather, the use of skilled labor relative to unskilled labor has risen across industries, whether the industries are producing traded goods or nontraded goods. Consequently, the general rise in demand for skilled labor in all industries is likely to have occurred because of the nature of technological change in this age of increased use of computers, robots, and so on. There are other reasons for the decline in the relative earnings of unskilled labor besides trade and the above-mentioned technological change. In regard to the United States, some such reasons are the increased immigration of relatively unskilled labor, the decline in the importance and influence of organized labor, and the fall in the real minimum wage (since the nominal minimum wage has not kept pace with the price level). Indeed, in an informal survey of economists attending a conference at the Federal Reserve Bank of New York, the average respondent attributed 45 percent of the rising wage inequality in the United States to technological change, 11 percent to trade, and less than 10 percent each to the decline in the real minimum wage, the decline in unionization, and the increased immigration of unskilled labor (with the remainder attributed to various other reasons). (See Burtless, 1995, p. 815, and Economic Report of the President, February 1997, p. 175.) Despite these strong points, however, the matter of the causes of the inequality is an unsettled one. For example, Wood has countered the technological-change argument by page 170strongly suggesting that the adoption of the unskilled-labor-saving type of new technology is occurring as a response to the threat of imports and thus that this reduction in the demand for unskilled labor should also be attributed to trade.15 Further, it could be said that the weakening of unions is also a result of new trade pressures. In addition, other potential causes for the increased inequality have been suggested. For example, Robert Feenstra and Gordon Hanson (1996) hypothesized that an important factor in reducing the demand for unskilled labor is the rise in “outsourcing” by U.S. firms. The point here is that firms are increasingly shipping abroad their component and intermediate-input production that is relatively unskilled-labor-intensive in nature, and this can also put downward pressure on the wages of U.S. low-skilled laborers. In Feenstra’s view (1998, p. 41), outsourcing and the shifting of activities abroad lead to the result that “the whole distinction between ‘trade’ versus ‘technology’ becomes suspect.” This muddling of trade and technology occurs because outsourcing can importantly be a response to technological change (e.g., improvements in communications, enhanced use of computers for inventory and monitoring purposes), and then trade responds to the outsourcing. Hence, in this view, the most important cause of increasing inequality is not to be identified as trade or technology—rather, both trade and technology are involved together in the increasing-wage-inequality process. To conclude this discussion, it is difficult in empirical analysis to sort out the specific impact on inequality of trade by itself in a complex, dynamic economy undergoing continuous structural change. Economists in general tend to doubt that trade is the dominant factor in the rising wage inequality, but needed work continues on this important issue. 

Outsourcing and Wage Inequality

Robert Feenstra and Gordon Hanson (1996) maintained that outsourcing has played an important causal role in the increasing wage inequality that has occurred in the United States in recent decades. To test this hypothesis, they constructed a measure of outsourcing and a measure of the trend in wage inequality for the years 1972–1990 for 435 U.S. industries and ran statistical tests to see if there was a significant association between the two constructed data series. Feenstra and Hanson measured outsourcing for an industry as the share of imported intermediate inputs in the purchases of total nonenergy materials by the industry. Hence, if $30 worth of inputs were imported and the industry’s total nonenergy input purchases were $1,000, outsourcing would be calculated for this industry as 0.03 (=$30/$1,000). Energy inputs generally cannot be outsourced since the geographical location of such supplies cannot be shifted, but this measure in effect treats all other imported inputs as fitting into the “outsourced” category. This is a very broad measure of outsourcing. To some observers, outsourcing (often called “offshoring”) intuitively implies something more narrow, as in Hummels, Rapoport, and Yi (1998, p. 82), who defined it as “the relocation of one or more stages of the production of a good from the home country”; or, as in current discussions in the United States, the sending of particular jobs abroad, such page 171as staffing call centers in India rather than in the United States (for more discussion of outsourcing, see Chapter 16). In contrast, the Feenstra-Hanson measure counts goods that have nothing to do with relocated production from a home firm. Nevertheless, using their measure (which they called SO), Feenstra and Hanson indicated increasing interdependence between the United States and other countries, since SO for the 435 industries in the aggregate doubled from 1972 (5.34 percent) to 1990 (11.61 percent). Feenstra and Hanson also examined another measure of interdependence (which they called SM), which was the share of imports in final U.S. consumption of the products of the various industries. This figure for the 435 industries as a whole also doubled from 1972 (5.02 percent) to 1990 (10.65 percent). Feenstra and Hanson (1996, p. 242) took the fact that SO and SM moved together over the period as “consistent with the idea that outsourcing is a response to import competition.” In other words, when final goods imports as a percentage of U.S. consumption rose, U.S. firms responded by seeking lower costs through obtaining intermediate inputs from foreign locations. As a measure of wage trends, Feenstra and Hanson calculated, for each of the 435 industries, the share of the industry’s wage bill that is paid to nonproduction workers. This is used as a proxy for the relative demand for skilled labor. As payments to nonproduction workers (e.g., executives, scientists, computer technicians) rise relative to payments to production workers, this measure (which they called SN) will rise. This rise is interpreted by Feenstra and Hanson to be a relative increase in the demand for skilled labor and hence an indication of greater wage inequality. Again, this measure is clearly a broad one, and it ignores differing skills and wage trends within the nonproduction-worker category, as well as within the production-worker category. With the SO, SM, and SN figures in hand for the 435 industries, Feenstra and Hanson then conducted statistical tests for the years 1972–1990. They divided the period into two parts (1972–1979 and 1979–1990), in recognition of the fact that the increasing-inequality phenomenon had basically begun only at about the end of the 1970s. Their results were that, for 1972–1979, the annual changes in SN were not related to the annual changes in SO (after allowing for other influences on the wage share besides the annual changes in SO); for 1979–1990, however, there was a highly significant positive association between SN and SO. Changes in SN were also positively related to changes in SM in a highly significant manner in the later period, whereas that had not been the case in the earlier period. In view of these statistical associations and the differing results for the later period as compared with the earlier period, Feenstra and Hanson (1996, p. 243) concluded that their research suggests, for the 1979–1990 period, “that outsourcing has contributed substantially to the increase in the relative demand for nonproduction labor.” Indeed, they estimated that the outsourcing could account for from 30.9 to 51.3 percent of the increase that had occurred in the share of the wage bill going to nonproduction workers. In a followup study, Feenstra and Hanson (2003) presented evidence of a direct link between trade and wage inequality. By using data on changes in industry behavior over time, they showed that foreign outsourcing or offshoring is associated with increases in the share of wages paid to skilled workers in the United States, Japan, Hong Kong, and Mexico. In several cases, outsourcing can account for half or more of the observed skill upgrading. In the case of the United States, Feenstra and Hanson present evidence that during the 1980s and 1990s outsourcing contributed to changes in industry productivity and product prices that, in turn, mandated increases in the relative wage of skilled labor. Building upon the Feenstra-Hanson work, David Hummels, Rasmus Jorgensen, Jakob Munch, and Chong Xiang (2014) examined the effects of offshoring by firms in Denmark on Danish workers’ wages. They note that the purchase of an input from a foreign supplier can clearly replace work previously done by a domestic laborer, hence potentially costing the domestic worker a job or resulting in a lower wage. On the other hand, the domestic firm, by buying cheaper inputs from abroad, can reduce the firm’s costs, increase the firm’s productivity, and perhaps thus lead to increased domestic output and employment and to a higher domestic wage. Hummels, Jorgensen, Munch, and Xiang investigated this research question by looking at the period 1995–2006 and at the individual firm, product, and worker level. A central conclusion was that offshoring increased the wages of Danish skilled labor and lowered the wages of Danish unskilled labor. Interesting other conclusions were that workers doing routine jobs lost wages, while workers in occupations that utilized knowledge in math, social science, and languages gained. Natural science and engineering workers were neither more nor less likely to be protected from the effects of offshoring on wages than was an average manufacturing worker. page 172 

Summary

The seemingly straightforward and intuitively appealing Heckscher-Ohlin theorem has been subjected to a large number of empirical tests. However, the theorem has not had a particularly high success rate. An early and extensive test for the United States resulted in the famous Leontief paradox. Various explanations for the occurrence of this paradox have been offered, but none of them has been judged to be entirely convincing. The most promising avenues for further testing of Heckscher-Ohlin seem to lie in the incorporation of a larger number of factors of production by disaggregating labor into different skill categories and by adding natural resources, as well as incorporation of technology differences across countries. Nevertheless, tests for other countries besides the United States have sometimes shown success in the standard two-factor framework. In addition, tests of a significant implication of the Heckscher-Ohlin analysis—that a country’s scarce factor loses from trade—have been conducted in the context of examining the importance of trade as a cause of rising income inequality (especially in the United States); these tests have also yielded mixed results. Particularly given the frustrations that have emerged with respect to successful verification of the Heckscher-Ohlin predictions on trade patterns, two questions pose themselves to economists: (1) Should we look for better ways of testing Heckscher-Ohlin? or (2) Should we search for other theoretical explanations of trade patterns and the composition of trade, as H-O has not been particularly successful? The literature has moved in both directions in response to these questions. The “better-testing” approach has focused importantly on extending the analysis to a greater number of factors than the original two, capital and labor. This approach runs the risk that, in disaggregating into more factors, we lose generality and meaningful understanding of the forces influencing a country’s trade pattern. As Paul Samuelson has suggested, we may end up concluding that Switzerland exports Swiss watches because Switzerland is well endowed with Swiss watchmakers! The second approach of searching for alternative trade theories to Heckscher-Ohlin has generated a great deal of activity in recent years. These newer theories are the subject of the next chapter.