The Haves and the Have-Nots

This way of looking at income distribution through the prism of social classes did not change much with the key turning point in the history of economics, the replacement of classical "political economy" by the "marginalist revolution" that started around 1870 and focused on individual optimization rather than on broad economic evolution of social classes, nor did it change later with the synthesis of the two strands (classical and marginalist) under the title of "neoclassical Marshallian economics" (from the Cambridge economist Alfred Marshall) and its establishment in the mainstream position. It was only in the early 1900s that the distribution of income among individuals (not among classes) attracted the attention of Vilfredo Pareto, a Franco-Italian economist who taught at the University of Lausanne in Switzerland. (His contribution is highlighted in Vignette 1.10.)

It was around the same time that the data on personal income distribution became available for the first time. This went hand in hand with economic development (countries becoming richer) and a broader fiscal role of the state. The early statistical information about income distribution emerged because of nation-states' need to collect direct taxes in a "fairer" way—that is, according to income—and to increase total tax intake, to spend it for public education, workers' invalidity, and, above all, war. The ideological change that saw all individuals as equal before the law, and thus the rich due to contribute more in accordance with their greater wealth and income, was important, too. Taxes had to be more tightly linked to income, and this required better information on incomes and their distribution among households. It is thus not surprising that the data used by Pareto to study income distribution among persons all came from the European late-nineteenth-century fiscal sources. At that point, our topic had been born.

Economists and social scientists are concerned with inequality in three ways. The first type of question they ask is: What determines inequality among individuals within a single nation? Are there certain regularities that make inequality behave in a particular way as societies develop? Does inequality increase as the economy expands—that is, is it pro- or anticyclical? In these types of questions, inequality is something that ought to be explained. It is a dependent variable. In the second type of question, inequality enters as a variable that explains other economic phenomena. Is high or low inequality good for economic growth, for better governance, for attracting foreign investments, for spreading education among the population, and so on? In these instances, we look at inequality in a purely instrumental sense: We are interested in whether it furthers or hampers some particular desirable economic outcome. The third way inequality enters within the social scientists' purview is when they address ethical issues linked with it. In these cases, they are concerned with the justice of social arrangements that exhibit different amounts of inequality. Is increasing inequality acceptable only if it raises the absolute incomes of the poor? Should inequality due to one's better family circumstances be treated differently than inequality due to superior work and effort?

It wasn't until 1955 that Simon Kuznets, a Russian-American economist and statistician, proposed the first real theory of what propels change in income distribution. (He is profiled, together with Pareto, in Vignette 1.10.) He argued—having had access to not many more data points than Pareto (although the data were of a different kind, household, not fiscal, surveys)—that inequality among people is not the same regardless of the type of society but varies predictably as society develops. Inequality in very poor societies must be low because the income of the vast majority of the population is just around subsistence, and there is little economic distinction among people. Then, as an economy develops and people move from agriculture into industry, Kuznets posited, a gap emerges in average earnings between industrial workers (richer) and farmers (poorer). The industrial sector also sees more differentiation in incomes between individual workers than is the case among farmers simply because tasks required by modern industry are more diversified. Therefore, income inequality increases both because of the growing gap in average earnings between industry and agriculture and because of rising inequality among industrial workers. Finally, in even more advanced societies, the state begins to play a redistributive role (see Vignette 1.7), education becomes more widespread, and inequality goes down (see Vignettes 1.1 and 1.2). Thus was formulated the famous "Kuznets' hypothesis" of an inverted U curve charted by income inequality in the course of economic development: Inequality must first increase before it goes down.

The idea, however, was not entirely novel. It was expressed some 120 years earlier by French social scientist and politician Alexis de Tocqueville and is worth quoting in full:

But, of course, Tocqueville, not being an economist like Kuznets, did not say anything beyond this, in particular about the mechanism whereby this inverted U shape would be brought into existence.

Kuznets' hypothesis has been empirically tested and retested by economists ever since it was first published in 1955. The ever-greater availability of national household surveys of income and consumption, the key source of information on income distribution, has greatly advanced the empirical exploration of Kuznets' hypothesis. In principle, the hypothesis should work best when we study the evolution of inequality in a single country, as the country undergoes radical transformation from agricultural to industrial and eventually to service-oriented economy. But in that context, its performance has been mixed: Some countries (and over some time periods) exhibited an inverted U pattern, while others did not.

The dissatisfaction with the performance and predictive capacity of Kuznets' hypothesis led to the addition of new elements that could better explain the behavior of income inequality. The revisions are known as the "augmented" Kuznets' hypothesis. Factors such as the "financial depth" of an economy, extent of government spending or state-sector employment, openness of the economy, and so forth now appear alongside income level as possible additional variables that explain the movement of inequality. Many economists argued that these additional elements could improve our understanding of the movement of inequality. For instance, the rationale was that a more efficient and broader financial sector would allow poor individuals to borrow to finance their own educations, and this would reduce inequality as the doors of educational advancement are thrown open to all and not reserved only for the rich. Government spending (as a share of gross domestic product, or GDP) or government employment (as a share of the total labor force) are supposed to have a dampening effect on inequality, first because it helps the poor, and second because it limits wage inequality. Greater openness to trade should, in poor countries, reduce inequality as it increases the demand for low-skill intensive products (say, textiles) in which these countries specialize; this would tend to raise the wages of unskilled workers compared to the wages of skilled workers or the profits of capitalists. In rich countries, openness to trade should produce the opposite effects since rich countries tend to export high-tech products. Their production requires highly skilled workers (say, computer scientists or engineers), so the earnings of college graduates increase relative to those with only a primary or secondary education. Thus, inequality goes up. Economists would test a typical Kuznets' hypothesis today by including all of these factors, and possibly quite a few others, in addition to income, often in an ad hoc fashion (e.g., adding age composition of the population or distribution of landownership). The results are better than when we use income level alone, but hardly spectacular.

More recently, French economist Thomas Piketty has produced a series of empirical studies, conducted jointly with a number of other economists (Emmanuel Saez, Anthony Atkinson, Abhijit Banerjee) and covering about a dozen countries, which have undermined both the Kuznets' and the augmented Kuznets' hypotheses. Piketty shows that after a long downward swing, inequality in Western nations has decisively increased in the past quarter of a century. Although these facts were well known before, Piketty has provided a "political" explanation, explaining them by governments' decisions to increase or decrease direct taxation of current income and inherited wealth as well as by the effects of wars (i.e., destruction of physical capital and reduced income of capitalists). This is arguably a political theory of income distribution where social attitudes (of what is just or unjust) and economic interests, reflected through voting and stances of political parties, as well as the war needs of economies, determine the path that inequality charts over time.

Piketty's studies, in order to explain what moves inequality over a long time period (the entire twentieth century), resorted to an old and rather discarded source of data: fiscal statistics. The reason fiscal statistics, first used by Pareto, have been replaced by household surveys is that fiscal data cover only a portion—the higher end—of income distribution since in most countries direct taxes are not paid by the poor. Household surveys, on the contrary, include everyone. The problem with the use of fiscal data is that conclusions drawn from them are valid only if the following two assumptions hold: (1) Taxable incomes are reasonable approximations of actual incomes of households (and the highest taxpayers are also the richest people), and (2) the evolution of overall inequality can be well approximated by the change in the share of the top income groups (say, the top 1 percent of taxpayers whom we believe to be also the richest 1 percent of households). Neither assumption is fully defensible. Taxable income used by Piketty and coauthors is called market (or pre-fisc) income, which excludes both taxes paid and government transfers.³ However, we are normally interested in what happens to inequality of disposable income, that is, income that belongs to households and individuals after they have paid taxes and received government transfers. Thus, if either taxes or transfers change, market and disposable income inequality can move in different directions. The problem with the second assumption is that inequality statistics should, in principle, include incomes of all people, not focus only on the rich. It could, for example, happen that the income share of the top increases while the income share of the poor goes up as well and both do so at the expense of the middle-class share. We may then not be able to say that the overall inequality went up, as we would tend to conclude solely from the rising income share of the top. Since Piketty-type studies depend on this particular assumption (which we know does not hold in all places and times), the interpretation of the results becomes problematic. Of course, if we had income or consumption surveys of the population for the periods sufficiently far in the past, the problem would be solved. We would not need to resort to the much less precise and fragmentary fiscal data. Unfortunately (as we shall see below), such surveys are available in rich countries generally only for the period after World War II and for many developing countries only for the past twenty or thirty years.

This is the situation of inequality studies today. It would be unfair or even impossible to summarize which of these different points of view has won the debate. Probably none. But it does lead to the question beyond simply measuring inequality or understanding how it evolves, and gets to the heart of whether inequality is necessary for an economy to grow and, if so, how high it should be.