The Great Leveler, page 49
This central trend across different countries has been employed as a proxy of change over time in support of the idea that income inequality first rises and then falls with intensive economic growth. Thus inequality in an ideal-typical economy undergoing economic development would be expected to track this inverted U-curve as it matures.4
Yet this approach suffers from multiple and very serious problems. Data quality is a concern: surveys that draw on large numbers of observations from different parts of the world are feasible only if they accommodate evidence of questionable precision and reliability. Robust findings require data to be fully compatible across countries, which is not always the case. Moreover, and even more damagingly, it has increasingly become clear that across-country panels are effectively invalidated by massive macroregional idiosyncrasies. Thus the appearance of an inverted U-curve in such panels is largely a function of exceptionally high levels of inequality in middle-income countries in two different parts of the globe, Latin America and southern Africa. According to a survey of income Gini coefficients in 135 countries in or close to the year 2005, Latin American countries are very heavily concentrated at the top of the inequality spectrum. At that time, the income share of the richest 10 percent averaged 41.8 percent in Latin America compared to a mean of 29.5 percent in the rest of the world. If Latin America and a few highly unequal countries in southern Africa (South Africa, Namibia, and Botswana) are excluded or replaced by regional dummy variables, the inverted U-shape simply disappears from across-country charts. This holds true regardless of whether Gini coefficients or top income deciles are used to measure inequality. In most of the world, countries having dramatically different per capita incomes, from low-income countries in Sub-Saharan Africa and South Asia to middle-income countries in Asia and Eastern Europe to high-income developed countries, now often cluster in an income Gini range of about 0.35 to 0.45. There is no systematic income-dependent inequality curve. Inequality outcomes relative to per capita GDP are fairly heterogeneous in general, and especially at the high end, which the high-inequality United States shares with low-inequality Japan and parts of Europe.5
Figure 13.1Gross National Income and Gini coefficients in different countries, 2010
EAP: East Asia and Pacific, ECA: Eastern Europe and Central Asia, LAC: Latin America and Caribbean; MENA: Middle East and North Africa, SA: South Asia, SSA: Sub-Saharan Africa.
Within-country analysis is therefore the only reliable way of documenting change over the course of per capita growth. A pioneering study of longitudinal data undertaken in 1998 found no support for the Kuznets thesis. In forty out of forty-nine countries under review, no significant inverted U-shaped relationship between per capita GDP and inequality emerged as these economies developed over time. In four of the nine remaining cases, the data supported the opposite scenario of a U-shaped distribution that seems to turn the model on its head. Only five of the forty-nine countries showed a significant inverted U-shaped pattern, although two of them suffered from data anomalies that cast doubt on this finding. This leaves us with all of three countries having a significant Kuznetian correlation between economic development and inequality—one of which, Trinidad and Tobago, is rather small. (Mexico and the Philippines are the other two.) Although it must be noted that the time frame of this survey may have been too short to produce more robust observations, these findings fail to inspire much confidence in the Kuznets thesis.6
Since then, longer-term within-country surveys have likewise produced little tangible support for the hypothesized correlation. The best example currently available appears to be Spain, where the Gini coefficient of income first rose and then fell between 1850 and 2000. If we are prepared to discount sharp short-term fluctuations in the 1940s and 1950s in the wake of the Spanish Civil War and the establishment of the Franco regime discussed in chapter 6, we can observe a secular increase in income inequality from a Gini value of around 0.3 in the 1860s, when per capita GDP was about $1,200 (expressed in 1990 International Dollars), to a peak in the low 0.5s in the late 1910s, when per capita GDP was about $2,000, and a subsequent overall decline to the mid-0.3s by 1960, when per capita GDP had reached $3,000—all of this, arguably, as the result of a gradual shift from farming into industry. Conversely, as we will see, long-term time series from Latin American countries generally fail to show an overall inverted U-curve pattern related to economic development. More importantly, early industrializers likewise fail to reach an inflection point in inequality trends associated with a per capita GDP of $2,000: Britain reached that level around 1800, the United States around 1850, and France and Germany twenty years later, and in none of these countries did income (or wealth) inequality begin to decline—nor had it visibly fallen to lower levels by the times these economies reached $3,000, as they did between 1865 and 1907.7
Another study more recently focused on the relationship between the relative share of the agricultural population and inequality in order to test Kuznets’ original bisectorial model. Once again, the predicted correlation is not borne out by the evidence: it does not show up across countries and is not significant within individual countries. Finally, little support for a regular linkage of economic output and inequality emerges when we compare multiple within-country series via nonparametric regressions. This approach shows that developments in different countries vary a lot even at comparable levels of per capita GDP: both developing and developed countries betray considerable variation in the timing and direction of inequality trends relative to economic development. All in all, despite continuing efforts to identify inverted U-patterns and the existence of a few supporting cases, the preponderance of the evidence fails to support the idea of a systematic relationship between economic growth and income inequality as first envisioned by Kuznets sixty years ago.8
Is there a predictable connection between economic development and inequality? The answer depends on our frame of reference. We have to entertain the possibility that there may be multiple Kuznetian cycles, or at least swings whose presence interferes with tests designed to look for a single curve. In the broadest terms, there can be little doubt that economic transitions promote inequality—not only from agrarian to industrial systems but already from the foraging to the agrarian mode and, in the present, from an industrial to a postindustrial service economy. But what about leveling? As I argue in the appendix, effective inequality—relative to the maximum theoretically possible degree of income concentration in a given society—need not always decline as economies grow richer. Conventional measures of nominal inequality do not offer much support to the notion that at certain stages of development, economic advances predict an attenuation of inequality. The main alternative—that in the absence of violent shocks, transitional increases in inequality are unlikely to be reversed—is far more consistent with evidence across the long run of history.
Another popular perspective centers on what is known as the “race between education and technology.” Technological change shapes demand for particular skills: if supply lags behind demand, income differentials or “skill premiums” increase; if supply catches up with demand or overshoots, premiums decline. Yet important caveats apply. This relationship is primarily germane to labor income but is less likely to affect gains from capital. In societies having high levels of inequality of income from wealth, this is bound to mute the effects of the interplay between the demand and supply of specific types of labor on overall inequality. Moreover, in earlier periods, constraints on labor income other than skills could play an important role: slavery and other forms of coerced or semidependent labor might have distorted income differentials.9
Factors such as these may help explain why in premodern societies skill premiums and inequality were not systematically related. For parts of Europe, time trends have been traced back to the fourteenth century. Skill premiums collapsed in response to the Black Death as real wages of unskilled workers rose, a process I discussed in chapter 10. In Central and Southern Europe, they rose again once population recovered, whereas they remained low and quite stable in Western Europe until the end of the nineteenth century. The latter outcome is unusual and appears to have been made possible in part by flexible supply of skilled labor and in part by productivity growth in the agricultural sector that helped sustain unskilled wages, both of which benefited from improved labor market integration. However, although the late medieval fall in skill premiums went hand in hand with a general leveling of income inequality, the relationship between these two variables was far less straightforward later on: stable skill premiums in Western Europe from 1400 to 1900 did not translate to stable inequality.10
The more advanced an economy has become and the better its labor markets function, the more skill premiums can be expected to contribute to overall income inequality. We must ask to what extent mechanisms that regulate the supply of skills, foremost education, are themselves shaped by underlying factors. Mass schooling has been an outgrowth of modern Western state formation, a process associated with economic growth but also driven by interstate competition. More specifically, the interplay between demand for and supply of education was sensitive to one-time violent shocks. This is well illustrated by the evolution of skill premiums in the United States since the end of the nineteenth century. Skill ratios in the manual trades were much lower in 1929 than they had been in 1907. Yet most of this decline was concentrated in the late 1910s: in four of the five occupations for which we have data, the entire net reduction within this twenty-two-year period took place between 1916 and 1920. At that time, World War I raised relative demand for unskilled workers and reshaped the distribution of manual labor wages. Wartime inflation and an abatement of immigration flows precipitated by the conflict also contributed to this sudden and powerful equalizing change. The ratio of white-collar to blue-collar earnings followed the same pattern: once again, the entire net decline between 1890 and 1940 occurred over the course of just a few years between 1915 and the early 1920s.11
A second compression of wage dispersion is documented for the 1940s. World War II created renewed strong demand for unskilled labor, inflation, and growing state intervention in labor markets. This led to a narrowing of the ratio of top to bottom wage shares for all male workers and reduced the earnings gap between workers with high school and college education. Returns to education experienced a dramatic fall between 1939 and 1949, both for workers with nine years of schooling compared to high school graduates and for high school graduates compared to those with a college education. Although the war-related GI Bill subsequently contributed to this equalizing pressure, even increased access to college could not prevent a partial recovery already under way in the 1950s. The sharp downturns of the late 1910s and the 1940s are the only such changes of this magnitude on record. Thus even though ongoing increases in the supply of educational opportunities were instrumental in constraining wage skill-based differentials until they finally surged in the 1980s, actual leveling was almost entirely limited to relatively short periods in which the country went through violent shocks caused by warfare.12
”IF YOU COMBINE INTELLECTUAL AND PROFESSIONAL CAPACITY WITH A SOCIAL CONSCIENCE, YOU CAN CHANGE THINGS”: LEVELING WITHOUT SHOCKS?
I now turn to my second strategy of identifying equalizing economic forces by looking for examples of inequality attenuation in countries that had not directly been subjected to the violent shocks of 1914 to 1945 and their fallout during the following generation and that had also been spared revolutionary transformations. For most of the world, this approach yields little solid evidence for leveling by peaceful means. Since the 1980s, Western countries generally have not registered more than highly temporary declines in income inequality. Drops in the Gini coefficient of market income in Portugal and Switzerland in the 1990s conflict with information regarding top income shares. Post-Soviet countries have partly recovered from the post-1989 or post-1991 surge in inequality that had been caused by huge increases in poverty. Very large countries such as China and India have witnessed rising inequality, as have other populous countries, such as Pakistan and Vietnam. Those four countries alone account for about 40 percent of the world population. Offsets in that part of the world, such as in Thailand, have been few and far between. In the Middle East, Egypt reportedly experienced inequality declines in the 1980s and again in the 2000s, but the most recent studies stress the shortcomings of the data. Moderate fluctuation since reform-driven attenuation of inequality in the 1950s and 1960s (discussed herein, in the section on land reform in chapter 12) may be the most plausible scenario for this country. Other examples include Iran in the 1990s and especially the 2000s as well as Turkey in the 2000s. Israeli disposable income inequality has been rising even as market income inequality has remained fairly stable, a puzzling pattern indicative of regressive redistribution.13
Sub-Saharan Africa is sometimes regarded as a beneficiary of peaceful income equalization during the first decade of this century. Yet this impression rests on shaky foundations: for all but one of the twenty-eight countries for which standardized income Gini coefficients are available for that period, the underlying data are poor and margins of uncertainty generally very wide. In the one case that has produced high-quality information, South Africa, inequality remained fairly flat—at a very high level. No significant trend could be observed in thirteen of the twenty-seven other countries, and in five more, inequality actually grew. Only ten of the twenty-eight countries registered a decline, and they account for only a fifth of the population of the overall sample. What is more, confidence intervals for the relevant Gini coefficients tend to be very wide: at the 95 percent confidence level, they average about 12 percentage points, clustering mainly between 9 and 13 points. (The mean is roughly the same for countries with declining inequality and for all others.) In many cases, these margins exceed the scale of the implied changes in inequality. Under these circumstances, it is difficult, if not impossible, to identify an overall trend. Yet even were we prepared to take these results at face value, they would not point to a consistent process of inequality attenuation. Although some countries in the region may well have enjoyed a measure of peaceful leveling in recent years, there is simply not enough reliable evidence on which to base more general conclusions about the nature, extent, and sustainability of such developments.14
This leaves us with the biggest and best-documented case—that of Latin America. Most of the countries in the region for which we have data have shown a significant reduction of income disparities since the beginning of this century. There is a good reason for considering developments in Latin America in greater detail. In terms of the violent leveling forces discussed in the previous chapters, the entire region provides the closest—albeit in many ways not particularly close—counterfactual to much of the Old World and North America that we can find on the planet. With only the rarest of exceptions untouched by intense violent shocks such as mass mobilization warfare and transformative revolutions, Latin America allows us to explore the evolution of inequality in a more sheltered environment.15
Some series of proxy data and creative modern reconstructions reach back several centuries. Reliable income Gini coefficients are often available only from the 1970s onward, when more states began to conduct surveys, and have greatly improved in quality since the 1990s. Findings for earlier periods thus need to be taken with a grain of salt. Even so, it has become possible to trace the long-tern evolution of Latin American income inequality, at least in broad outlines. The first age of globalization sustained export-led economic growth from the 1870s into the 1920s, driven by exports of organic and mineral commodities to the industrializing Western world. This process proved disproportionately beneficial to elites and raised inequality.16
Export-driven development first slowed in the wake of World War I, which dampened European demand, and it ground to a halt when the Great Depression hit the United States in 1929. World War II further curtailed at least some forms of trade. The years from 1914 to 1945 have been characterized as a period of transition and decelerating growth. In six documented countries, income inequality continued to rise during this period, from 0.377 in 1913 to 0.428 in 1938, weighted for population. Although spared direct involvement in the wars, Latin America was nonetheless very much exposed to the fallout from violent and macroeconomic shocks that occurred outside the region. Interruptions of trade and an inflow of changing ideas were among the most consequential consequences. These shocks ushered in the end of the first phase of globalization, a decline in economic liberalism, and a turn toward increasing state intervention.17
In the following decades, Latin American governments adapted to this global trend by more heavily promoting industrial capacity, primarily aimed at domestic markets, and by relying on protectionist measures to facilitate this development. This eventually revived economic growth and left its mark on the distribution of incomes. Outcomes varied greatly across the region. In the more developed economies, growth boosted the middle class, the urban sector, and the share of white-collar workers in the waged labor force. These changes were on occasion accompanied and reinforced by more welfare-oriented and redistributive policies. External influences played a significant role, as Britain’s 1942 Beveridge Report on social insurance and other Western postwar programs inspired social security schemes in southern South America. Inequality was affected in different ways. Income disparities sometimes weakened, as in Argentina and possibly also in Chile; sometimes they grew, most notably in Brazil; and in other countries they increased at first and declined later, as in Mexico, Peru, Colombia, and Venezuela, where large reservoirs of unskilled surplus labor and high demand for skilled workers boosted inequality until these pressures subsided in the 1960s and 1970s.18
