Globalization and Inequality

Date April 4, 2014
Speaker Elhanan HELPMAN(Galen L. Stone Professor of International Trade, Harvard University)
Moderator WAKASUGI Ryuhei(Senior Research Adviser and Program Director, RIETI / Professor, Gakushuin University / Adjunct Professor, Yokohama National University / Professor Emeritus, Kyoto University)


Elhanan HELPMAN's PhotoElhanan HELPMAN

Trade has evolved over a long period of time as defined by the ratio of exports plus imports to the gross domestic product (GDP). There have been two big waves of globalization. The first started in the 19th century and ended essentially with World War I. The interwar period saw a decline in globalization. After World War II, globalization expanded again, and, although the figures here end in 1992, the growth of trade relative to income has continued at a pace roughly twice as fast as that of income. This long-term development actually is related to the growth rate of the world economy.

From 1820 until 1992 and up to the present, globalization and faster growth have been accompanied by increased personal income inequality in the world economy, treating every individual with equal weight. Inequality has increased continuously from the beginning of the 19th century up to the present. Interestingly, the sources of this inequality have changed dramatically over time. Both inequality within countries and differences in per capita income across countries have grown since the early 19th century, although the growth in inequality across countries was more significant. Inequality within countries as a contributor to total inequality in the world economy has declined, whereas inequality in the world economy has continued to rise, although not at an equal pace. This was due primarily to the rise of inequality across countries.

The percentage of people categorized as "poor" in the world economy steadily declined during this period. The World Bank tracks the percentage of people living on less than $1.25 and those on less than $2 a day. By these metrics, poverty has declined, and this accelerated after 1990 primarily due to the growth of India and China. Some argue that inequality hasn't increased over this long period of time because the income per capita of countries and the average income in the lowest quintile lie almost on a 45-degree line; growth in per capita income overall and in the lowest quintile has been about the same. However, much of the rise in inequality was at the upper end of the income distribution. In the United States, for example, the income share of the top 1% doubled during this period, and, in many countries, the top 1% gained much more than other parts of the income distribution.

We have very good data on inequality in the United States. Let's look at the college wage premium relative to the supply of college graduates. The college wage premium is the compensation received by college graduates relative to that of high school graduates. Starting in the early 1960s, the college wage premium was about 40%. A college graduate in the early 1960s earned about 40% more than a high school graduate. At the end of the period, in 1995, it was close to 70%. This was a major source of increased inequality. We then can compare this to the supply of college graduates relative to demand. The supply of college graduates increased during this entire period. Puzzlingly, despite the growth in supply, the relative price of these more numerous college graduates rose. This doesn't comport with our understanding of supply and demand. This generates polarization in compensation and earnings in the United States despite the rapid increase in the number of college graduates relative to non-college graduates.

Inequality in the United States is in a U-shape in the long term. It's well known that inequality began declining at the beginning of the 20th century, when it was extremely high, and then began increasing at an accelerating rate in the 1960s, 1970s, and 1980s. The decline in the early 20th century was driven primarily by a decline in income from capital relative to income from labor. The rise in inequality in later years results primarily from rising labor income among top earners relative to other income earners. In very recent years, capital also has started to play a significant role, at least in the United States.

Today, we need to focus on different parts of the income distribution; we shouldn't only look at indices of inequality covering the entire range of incomes, because very different shifts in inequality have occurred over time in different parts of the income distribution. This is a graph of two measures of inequality: one at the upper part of the distribution and one at the lower part. Two U-shaped lines appear over the long term. One represents the 50%-20% ratio of income distribution and the other represents the 80%-50% ratio. In the 1990s, there was a shift in inequality in the United States. In the upper part, inequality keeps rising but moderately. At the bottom, it declines. There is a prolonged period of decay in which the trends at the bottom and at the top go in opposite directions. This is very significant because people in the middle of the income distribution have lost relative to both the people at the top and the people at the bottom. This is known as the hollowing out of the middle class. A similar phenomenon is found in some other countries.

Looking at the trend of rising inequality over time, it does not rise to the same extent in different parts of the income distribution. Internationally, it's much richer than in the United States.

In the early 1990s, labor economists observed the rise in the college wage premium and the resulting inequality despite the increased supply of college graduates, and they first attributed this to globalization. The standard neoclassical theory prevailed, which held that the increasing integration of many less-developed countries into the world economy supplied many low-skill-intensive products, also called labor-intensive products, reducing the relative prices of these products. As the relative prices of these goods declined, there was a related decline in low-skilled workers in industrial countries. This generated the rising income gap between high- and low-skill workers and increased the college wage premium. The rise in the college wage premium was more extreme in the United States, but it also happened in other countries. An alternative explanation proposed that the rise in the college wage premium was due to a skill bias caused by technological change. This complementary relationship between skills and technology was the cause of the increase in the college wage premium.

Two of my colleagues, Claudia Goldin and Larry Katz, wrote a famous book titled, "The Race between Education and Technology," which provides a very detailed analysis of the revolution in education and compensation in the U.S. economy going back to the 19th century. They pointed out that technological change didn't always increase the demand for skills. Technological change in the late 19th/early 20th century was biased toward unskilled workers. This partly explains the attendant decline in the college wage premium. Later, technological change increased the demand for skills. Research by labor economists who have studied the composition of occupations in detail and the characteristics required for workers in these occupations supports this view.

The globalization vs. technology debate was settled 20 years ago in favor of technology. Empirical studies done at the time attempted to assess how much of the rise in the relative wages of skilled vs. unskilled workers could be attributed to globalization, and it was found to be a relatively small fraction; in some studies about 20%. Assessing changes in relative prices and how they translate into changes in relative wages quantitatively requires estimates of how much the relative prices of labor intensive goods change vis-à-vis those of skill intensive goods. This must then be translated into relative wages, which requires elasticity of substitution between skilled and unskilled workers. If the estimates of what happened to relative wages and the estimates of elasticity of substitution between skilled and unskilled workers at the time are aggregated, a very small fraction of the rise in the college wage premium can be explained through this mechanism. Additional globalization-related explanations were sought, such as foreign direct investment (FDI). One example was an extensive study of FDI between the United States and Mexico. U.S. companies shifted particularly labor intensive production-related activities to Mexico. Mexico is much more unskilled labor intensive compared to the United States. However, if you look at the labor intensity of the shifted activities, they were labor intensive in the United States, but compared to activities in Mexico, they were actually skill intensive. This helps explain the rise in the college wage premium in both Mexico and the United States, but it cannot close the gap. The contribution of offshoring is another question. Offshoring can be done by FDI or at arm's length, but its impact on relative wages in principle will be similar quantitatively. Taking this into account, you get an increase in the contribution of globalization, but the gap is still too big to be explained. Finally, if you look at what happened to the composition of employment, it's actually much more consistent with the technology explanation than with the trade explanation.

People who looked at what happened in different sectors of the economies of various industrialized countries and middle-income countries found that the ratio of skilled workers to unskilled workers has increased sector by sector. This is inconsistent with the globalization explanation: if the driver is the change in the relative prices of labor-intensive products, and this raises the relative price of skilled workers, then the increase in the relative wages of the skilled workers should reduce the use of skilled workers relative to unskilled workers in every sector of the economy. The opposite was true, which put the final nail in the coffin of globalization as the explanation of the rise in the college wage premium.

There recently has been renewed interest in the topic and a renewed attempt to determine whether other aspects of globalization might contribute to inequality of the type that we see. There are additional mechanisms which people were not aware of 20 years ago but which prove to be quite powerful in understanding the evolution of inequality and its relationship to international trade.

The shift in inequality internationally is not uniform in two important ways. The first is that it's not the same in every country and the other is that it's not the same at the upper and lower edges of the distribution, varying much across countries. Data from the Organisation for Economic Co-operation and Development (OECD), starting in 2000, give the ratios of the deciles: the fifth to the first, nine to five, and one each for the years 2000 and 2007. Even during these seven years, inequality increased both at the top and bottom of the distribution in some countries using this measure, including Ireland, Japan, Korea, Norway, and the United States. There are exceptions such as France, where inequality declined both at the top and bottom. In many countries, inequality at the bottom and at the top went in opposite directions: Canada and the United Kingdom are examples. This is the hollowing out of the middle class. There are countries such as Sweden and Germany where it moved in opposite directions. In countries for which we have detailed data on distributions, very rich movements can be seen in different segments which cannot be explained by technology alone. Understanding the extent to which the technology explanation needs to be supplemented by other mechanisms requires study not only of rough measures of inequality but also of details of the distribution in different parts of the income structure.

In the 1990s, new and much more detailed data sets about the structure of the global economy became available, enabling more detailed study of wages and the role of business firms and technology in the process. The data sets indicated that 1) industry-by-industry, in many countries for which data are available, only a small fraction of firms export, 2) exporters are typically larger and more productive than non-exporters, and 3) exporters pay higher wages. This is true in the United States, Japan, France, and many other countries.

Understanding what's generating these patterns has implications for inequality. Various labor economics studies have found that much inequality is driven by what's known as residual inequality. Wages vary from person to person based on varying levels of education, skills, experience, etc. However, the return based on these various characteristics proves to be a mere fraction of the variance. There is a large residual amount which economists call residual inequality, which has increased greatly over time.

Two lines of research have evolved in response to these phenomena and to the availability of these new insights. One looks at the contribution of international trade or globalization to the rise in residual inequality. In a variety of countries for which there were studies, you find very similar phenomena. The other line of inquiry that has evolved looks at the contribution of trade inequality in different segments of the income distribution. There are few empirical studies at the moment, but quite a rich variation in shifting inequality can be explained with this type of theory.

Generally speaking, we find that a lot of the wage differences are within groups rather than across them. It almost doesn't matter how you cut the data. Within occupations as opposed to across occupations, 80% of the variation is there. About 90% of the growth over time is within occupations rather than between them. This is true in Brazil. If you look instead within sectors, again, 83% of the inequality is within sectors rather than across them. In terms of change over time, more than 70% occurs within the sectors. Cutting the data even finer, looking at occupations within sectors, you still get 67% of the inequality within sector occupation cells and about 1/3 across these cells, and again, you get a big contribution of what happens within these cells over time. This is not unique to Brazil; we just happen to have their data available. It's similar though not identical in countries such as Sweden.

Turning to residual inequality, the residual wage in Brazil was close to 60%, 88% of which is within sector occupation cells. The amount of variation in wages seen in narrowly defined groups of workers after accounting for their education, experience, gender, and type of sector, is stunning. A great deal of the inequality within sectors can be attributed to differences in firm characteristics. International trade and globalization are relevant here because of what we call an "export wage premium"; firms that engage in foreign trade pay significantly higher wages than other firms accounting for their size. Labor economists uncovered a wage premium for firm size many years ago, but there is an additional premium for exporting firms.

The current theories have tried to develop analytical methods to explain this data and to provide a system that can be studied statistically. They emphasize the contribution of individual firms to foreign trade and globalization rather than looking at trade at a sectoral level. Because this viewpoint treats firms as heterogeneous, it can be expanded to account for different wage payments by these firms. Analytical models can be constructed to account for heterogeneous firms, heterogeneous workers, and frictions in the labor market interacting to generate this distribution of wages. The question of how globalization shifts this distribution of wages can then be asked. The emphases are on heterogeneity of firms, heterogeneity of workers, or what you might call ability beyond observables. The important thing is that earnings vary across firms. Two broad mechanisms can generate this type of variation. The first is that even where two people are identical on the observables, they may have different hidden abilities. A statistician can't account for hidden abilities which firms are able to identify through the resources they invest in human resource management. This leads to variation across firms as larger, more productive firms hire better workers with similar observables, creating a wage gap between similar workers who work in different firms. Second, if labor markets are not fully competitive and have some inherent frictions, firms incur costs in hiring workers. A worker within a firm therefore is not equivalent to one outside the firm with the same characteristics. This gives workers within a firm leverage in wage negotiations which can generate differences in wages across firms depending on how substitutable the former is.

Inequality in different segments of the income distribution has not been studied empirically as intensively, but we have more reliable theories on this issue at present, and all point to a significant impact of globalization on inequality. The interplay between theory and empirical analysis in this field has been very important, but there remains a constant race between scholars and events.


Q1: I have three questions. First, is the distinction between the non-tradable sector and the tradable sector still relevant? Second, Japan has seen a lot of discussion about the lack of inbound FDI, although we have a lot of outbound FDI. When you talk about globalization, you probably mean both, but is there any difference between them? The third relates to policy. In Japan, we are discussing corporate tax policies which means globalization not only influences opportunities for a company but also the policy itself, the tax system, and other things which may have implications for changing inequality. Please let us know if you have any comments on that.

Elhanan HELPMAN:
There are very few, if any, studies on non-traded sectors. The studies I mentioned focus on manufacturing, basically, because this is the majority of what is traded internationally. Services trade has been increasing rapidly, but there are no comparable studies of the service sector. Many phenomena look very much the same for all sectors, but the mechanisms may not be the same and alternative mechanisms have not been studied extensively. The role of FDI in inequality hasn't been explored in as much detail as has foreign trade. There will be a difference between outbound and inbound FDI. The common denominator for the impact of FDI and trade on inequality is that larger and more productive firms engage in FDI. A globalization-related wage gap feeds into inequality. We don't know how much of the rise in inequality can be explained with FDI. With regard to policy, there is some work on policy issues, but it's very limited in scope. One seemingly robust observation is that tax competition across countries mutes marginal tax rates. This reduces the distributive power of governments because countries are concerned that raising their marginal tax rates will cause them to lose business to other countries. The extent of this phenomenon and how much it translates into global inequality has not really been studied. There has been a lot of progress in this field of globalization and inequality, but many open questions remain.

Q2: You mentioned the hollowing out. I recently read a book that states that top-wage workers will be replaced by machines in the long term. I'd like to hear your opinions about that and if that will affect global business activities.

Elhanan HELPMAN:
This has always been a problem. New machines always replace workers. However, this is not the issue as far as I'm concerned. The issue is whether workers can find useful employment even when new machines become available. One difficulty is that machines replace particular types of workers while simultaneously enhancing employment opportunities for other types of workers. It has a differential impact. It is not uniform across the entire labor force. This is why many people have argued that government should provide retraining programs to people who lose their jobs due to technological change to enable them to find new jobs. There is a delicate balance between the negatives of machines replacing workers and the positives of engaging in innovation.

Q3: I think firms operate a variety of businesses in a variety of sectors, so I'm not sure what the relationship is between firms and sectors in your statistics. With regard to policy, wage inequality is enlarged by globalization of the world market, and it may in turn cause further inequality in education or skill training and enlarge the productivity heterogeneity of firms and eventually inequality. How can we mitigate the vicious circle between the globalization of the economy and rising inequality?

Elhanan HELPMAN:
Typically, in the way data are classified, firms are classified according to their core operations. For many firms, this is reasonably satisfactory, although there are some exceptions where leakage will occur. On policy, we have to distinguish between the inequality due to economic forces leading to growth and that due to frictions within the system. When it arises due to frictions, the natural tendency is to formulate policies to reduce the frictions. When inequality arises due to growth, like innovation, there is a tradeoff. Each country needs to decide whether it will have more growth and more inequality or if it will give up some growth and pursue policies that limit growth and also inequality. It's not just an economic issue; it's a social issue. How much are we willing to pay for equality? The amount of concern over inequality also differs between countries.

Q4: You suggested that the matching of workers and firms in the labor market contributes to inequality or intensifies the effect of trade on inequality. My impression was that making the labor market better and more efficient intensifies the effect of trade on inequality. Just now, you mentioned better policies to remove friction. How can I understand the relationship between the removal of friction and inequality?

Elhanan HELPMAN:
Not every friction is alike. Some frictions in the labor market enable inequality mechanisms while others do not. If you remove those that enable inequality across firms, mostly within sectors, you reduce inequality. For others, it might increase. It's not a uniform phenomenon.

Q5: You mentioned that scholars are in a race to develop new theories and catch up with events. You have shown that scholars are eager to face new events and develop new theories to explain new phenomena. In Japan, we must relate everything to Abenomics. The "third arrow" of the growth strategy, investigating the relationship between globalization and technological development and growth and inequality, is essential for us. According to today's presentation, Japan's inequality is not so serious comparatively and is not growing much, but there is almost no growth. That's our problem. Do you have any suggestions? With low inequality, how can Japan grow?

Elhanan HELPMAN:
This is what we are currently researching. We are integrating the new view of trade and globalization and inequality with endogenous economic growth. Theoretically, what we find is that even if the growth mechanism leads to some convergence in growth rates, it can still leave a lot of inequality at different levels across countries.

*This summary was compiled by RIETI Editorial staff.