EXPERIENCE AND NEW THEORY FROM THE 1980s ONWARD

The policy debates of the three decades following World War II influenced and were influenced by the analytical frameworks developed to understand the impact of trade and investment openness on inequality.

The experience of inequality in the United States (and other developed economies) is interesting because of the possible light it can shed on the predictions of the standard H-O model. The simple model has the powerful prediction that opening up will narrow the returns between labor and capital in countries with a relatively low capital-to-labor ratio, or between skilled and unskilled labor in countries with a relatively low skilled-to- unskilled labor ratio. The observance of these trends in East Asia was read as support for the model. The flip side of this same prediction is that the gap between these returns should widen in countries with relatively high ratios of capital to labor and of skilled labor to unskilled labor. This did not happen in the United States in the 1960s and 1970s, but it has been happening since the 1980s. Now, it can be argued that given the relative size of the U.S.

There is, however, the issue of how much of the rising inequality in the United States can be attributed to trade, and how much to other factors, specifically to technology. The overview by Pavcnik (2011) captures the recent consensus:

A large body of research on this topic finds little support that international trade in final goods driven by relative factor endowment differences can account for much of the observed increase in skill premiums in developed and developing countries.... First, the Stolper-Samuelson mecha­nism suggests that increased relative demand for skilled labour in countries abundant in skilled labour occurs as a result of shifts in the relative demand for skilled labour across industries.... However, the employment shifts across industries have not been sufficiently large to account for the large increase in wage inequality. Most of the observed increase in demand for educated labour in countries such as the United States is driven by increased relative demand for skilled labour within industries. (p. 242)

There is significant debate on the relative role of trade. Although Krugman (2008) argues against his own earlier view that trade was a relatively small factor in explaining the rise of inequality compared to technology, there are also criticisms of the “small role of trade” view by Irwin (2008), Katz (2008), and Autor (2010). It would be fair to say that skill- biased technical change is considered to be a major driving force, if not necessarily the dominant force, behind rising inequality.^[333] This empirical and policy debate has in turn

fed into an emerging literature that goes beyond simple H—O/Stolper—Samuelson for­mulations to consider within-industry wage differentials between heterogeneous firms and how these could be affected by trade.

The H-O predictions on trade and inequality could be argued to have been con­firmed by the experience of rising trade and falling inequality in East Asia in the 1960s and 1970s. They could equally be argued to have been confirmed by the experi­ence of the rising trade and rising inequality in the United States from the 1990s onward, although there is consensus that the forces of technology provide stronger explanation. However, the difficulty for the H-O model is that, contrary to its prediction, and con­trary to the experience of East Asia in the 1960s and 1970s, from the 1980s onward, the experience of Asian economies and that of Latin America until the 2000s has been one of rising trade and rising inequality. As the comprehensive review by Goldberg and Pavcnik (2007) concludes:

The survey of the evidence confirms Wood (1999), who noted that inequality increased in several middle-income Latin American countries that liberalized their trade regimes in the 1980s and 1990s. It further suggests that this positive relationship holds in the cases of India, China and Hong Kong. As noted previously by Wood (1999), the experience of developing countries that globalized during the 1980s and 1990s contrast with the experiences of several Southeast Asian countries (South Korea, Taiwan, Singapore) that underwent trade reforms in the 1960s and 1970s. The latter underwent a decline in inequality as they opened up their economies to foreign markets. (p. 54)

A number of comments are in order before we proceed to discuss the implications of these facts for the H-O model. First, although the economies of Latin America liberalized during the 1980s and 1990s, this was also a period of painful macroeconomic adjustments and slow downs, and this could confound attribution of the causes of inequality. Second, note that the simple Lewis-Kuznets model discussed in the last section could indeed still predict an increase in inequality with opening up. Finally, two major further stylized facts have been established since the Goldberg and Pavcnik (2007) survey.

The basic H-O/Stolper-Samuelson framework is foundational in the discourse on trade and inequality. However, questions about its validity have been raised by the find­ing that inequality in many developing countries has increased since the 1980s despite increases in trade. This disconnect between prediction and outcome has led to a fruitful search for alternative explanations of why an increase in trade may increase inequality, and some of the theories advanced have also been helpful in understanding the impact of trade on inequality in developed countries as well. In this section, we examine a range of such theories as illustration of the direction the literature is taking in light of the expe­riences of the last three decades.

In the wake of the failure of the basic two goods, two factors H-O model to predict co-movement of trade and inequality, a range of models were developed that vary the technology or number of factors and goods (including the introduction of nontraded goods) in order to derive predictions more consistent with the data. Thus, for example, Wood (1994) moves from the two-factor model with a skilled/unskilled labor division to consider a three-factor model with workers classified as skilled (high education)/ semi-skilled (basic education)∕unskilled (no education). Further, there are three types of production—skill-intensive manufacturing, semi-skilled intensive manufacturing, and agriculture. In this setting, for a country with comparative advantage in agriculture we get the standard prediction that opening up will reduce inequality. However, for countries with a relatively large number of semi-skilled workers, opening up will increase their wages relative to the wages of both high-skill and unskilled workers.

In the same spirit, Davis (1996) considers a two-factor (he calls them capital and labor), three-goods H-O model, with market imperfections that prevent factor price equalization and full diversification of production. The three goods differ in the capital intensity of production technology. With countries ranked by capital intensity of factor endowment, the least developed countries will export the least capital intensive com­modity and import the next most capital intensive. For these countries, the standard result will hold—opening up will narrow the gap in factor returns. However, for countries with intermediate levels of capital intensity of factor endowment, which will export the com­modity with intermediate capital intensity of production and import the commodity with highest capital intensity of production, opening up will have the opposite effect. Of course, for the most developed countries we again have the standard Stolper-Samuelson result. At least for developing economies at intermediate levels of capital intensity, then, this type of theorizing might explain co-movement of trade and inequality. Such coun­tries might, in principle, include East Asia from the 1980s onward and Latin America at the time of its opening up in the 1980s and 1990s.

The papers by Wood (1994) and Davis (1996) are examples of attempts to predict co-movements of trade and inequality within a recognizable H-O framework but with more disaggregated specification of commodities or factors. This trend has continued in the literature, with added complications such as capital-skill complementarity in production—to the point that the discourse of today cannot really be labeled as a H-O discourse.

Helpman et al. (2010) bring together several strands of the modern trade literature with a focus on firm and worker heterogeneity and derive predictions on trade and inequality that are consistent with many of the empirical findings of the last 30 years. Following Melitz (2003), the model supposes heterogeneous firms producing differen­tiated commodities. Firms can enter by paying a fixed cost, but discover their produc­tivity only after paying the sunk cost. The productivities are drawn from a Pareto distribution, an assumption that helps the tractability of the model. After productivity is revealed, firms decide whether and how much to produce for export, for the domestic market, or for both, or they exit altogether. Production involves a fixed cost, and output is a function of firm productivity, number of workers hired, and their average ability. A specific functional form is used for tractability, but the key aspect is that these three elements are complementary to one another.

Worker ability is also assumed to have a Pareto distribution—again for tractability. Search and matching frictions exist in the model, and firms can pay more to match with more workers. Further, among the workers the firm can screen for higher abilities above a cutoff by paying a cost (with a higher cutoff costing more), but it cannot distinguish abilities beyond this cutoff. Thus, all workers in a firm are paid the same wage. The wage is modeled as emerging from the outcome of a bargaining game between the firm and the average worker.

Fixed costs of production, and fixed costs of exporting, mean that firms with very low productivities do not produce at all, while firms with high productivities select exporting because of the existence of a cost of trading. Given costs of search and screening, it can also be shown that firms with higher productivity and revenue search more and use a higher ability cutoff, so that they have higher ability workers on average and thus higher wages. The key point is that exporting firms pay higher wages in equilibrium. Thus, if we start from autarky, where fixed costs of exporting are so high that nobody exports, and reduce these fixed costs in a comparative static manner so that some firms begin to export, wage inequality is introduced where none existed before. This applies to all coun­tries; thus, opening up can increase inequality in all countries—developed and developing—because of the selection effects of exporting.

Verhoogen’s (2008) is another example of a similar model where selection effects can explain co-movement of trade and inequality. The idea here is that exporting requires the production of higher quality products and only the most productive will find it profitable to go into exporting. With a mechanism of higher wages in more productive firms, this leads to greater inequality with more openness. It should be noted that the Helpman et al. (2010) model also has an intriguing result at the other end of the spectrum where export­ing costs are so low that all firms export. Then, once again, wages are equal. In their model, inequality first increases and then decreases as opening up intensifies—an “inverted-U” relationship between inequality and openness. It is of course an empirical question as to whether the intensified globalization from the 1980s onward has now taken some countries to the point where the model would predict falling inequality. If this was the case for some countries, of course, the model could not explain the co-movement of trade and inequality for those countries, and other explanations would have to be considered.

A selection mechanism of a different sort is present in studies of outsourcing as exem­plified by Feenstra and Hanson (1996, 1997), which also relates to a broader literature in outsourcing and FDI in trade. They considered a scenario where the final output is pro­duced using intermediate inputs that are produced using different intensities of skilled and unskilled labor. Consider now two economies with different endowments of skilled and unskilled labor. For any given pattern of trade costs, the skilled labor abundant (devel­oped) economy will use the more skilled intensive production of intermediate inputs. When trade costs are lowered in a comparative static exercise, some of this production is relocated from the developed economy to the developing economy. However, the activity that is relocated is the least-skilled intensive in the developed economy and the most-skilled intensive in the developing economy. This increases skill intensity of production in both the developing and the developed economy and, hence, widens the wage gap between skilled and unskilled labor in both economies. Feenstra and Hanson (1997) show empirical support for this as explaining rising wage inequality in Mexico.

Feenstra and Hanson (1997) highlight an aspect of globalization that has come to the forefront in the last 30 years, namely, foreign direct investment (FDI). The issue ofport- folio and financial flows will be discussed in a subsequent section, but longer term FDI has also been important in the recent growth surges in developing countries. What are the implications of FDI for inequality?

The theory of FDI in the simple Lewis model discussed in the previous section sug­gests that as wages rise in a former surplus-labor economy, capitalists will look to invest­ment opportunities abroad, presumably in economies where wages are lower still. If these economies are themselves in a state of surplus labor then further investment will raise the share of capital and worsen the distribution of income for that reason. However, if the “Lewis turning point” has already been reached in the economy receiving FDI, this investment will raise wages further in that economy, and this could be a channel for reducing inequality.

Modern theories of the impact of FDI build on the H-O framework and then bring in firm and worker heterogeneity, as in the analysis of Feenstra and Hanson (1997). Overall, it would be fair to say that the theoretical conclusions are ambiguous, with some sugges­tion of FDI contributing to an increase in inequality in developing countries at the start of the process, with a possible turnaround in the later stages. For example, Figini and Gorg (1999) discuss the transition as domestic firms absorb the new technology of the FDI.

Inequalities may be created in the early stages, but are mitigated in later stages as the tran­sition proceeds—an inverted-U relationship has framed much of the empirical work in this area. The large and growing empirical literature also gives mixed results, with perhaps a greater weight to the conclusion that FDI is associated with rising inequality in earlier stages,^[336] but that there may be a turnaround, and that the impact is muted or even neg­ative at higher levels of income per capita.^[337]

Selection effects as the result of global integration are now central to the trade and FDI literature and, thus, to the attempts to explain co-movement of trade and inequality. They do appear to provide a coherent explanation of increases in inequality in both developed and developing countries, and for this reason, they merit close theoretical and empirical attention in the years to come.^{[338] [339]}

20.4.