WHICH GLOBAL DISTRIBUTION OF INCOME?

Our starting point must be to clarify what we mean by the global distribution of income. Following Milanovic (2005) and Anand and Segal (2008), we can define four concepts of the global distribution of income and their associated levels of inequality, distinguished by the population unit and the income concept (which may be a measure of consumption expenditure) to which they refer.

Our first concept of the global distribution of income, denoted concept 0, is the dis­tribution of global income by country. In other words, the “population unit” is the coun­try and the “income concept” is the (total) national income of the country. Thus India and Canada, both with GDPs of US$1.8 trillion in 2012, count as equal, despite the fact that India has a population of 1237 million and Canada only 35 million. It is this concept 0 global distribution that is most relevant for questions of geopolitics and market access. In international negotiations over trade rules and macroeconomic policies it is a country’s total economic size that tends to determine its bargaining power. For such questions, it is a country’s weight in international markets—its command over internationally traded goods and services, or financial assets—that matters, and hence income at market exchange rates is likely to be relevant.

Next, the concept 1 global distribution again takes the country as the population unit, but now the income concept is the national income per capita of the country, not its total national income. This is the concept typically used in analyses of economic growth, and in particular of economic convergence, where the question is how the set of character­istics and policies associated with a given country affects its per capita income growth rate. Because it is real output that is of interest in this case, income levels will be measured at PPP exchange rates.

In the concept 2 global distribution, the population unit is the individual, and the income concept is again national (household) income per capita. (This is equivalent to taking the country as the population unit, as in concept 1, but weighting each country by the size of its population.) It is not obvious why this concept of global inequality would be intrinsically interesting, but some older studies have analyzed its evolution over time, mainly because of the ready availability of data on national income or GDP per capita (Boltho and Toniolo, 1999; Firebaugh, 1999, 2003; Melchior et al., 2000). Concept 2 is of instrumental interest, however, through its relationship with concept 3, which is the focus of this chapter.

The concept 3 distribution also takes the individual as the population unit, but the income concept is the per capita income of the household to which the individual belongs—under the assumption of equal sharing of household income (or consumption expenditure). It is the global analogue of the type of distribution typically used to calculate inequality within

It follows that this concept 1 distribution—unlike the concept 0 distribution—is not a “distribution of (total) global income among countries.”

Table 11.1 Concepts of global income distribution and inequality

countries.^[656] Henceforth we use the terms “global distribution of income” and “global inequality” without further qualification to refer to their concept 3 counterparts.

Concept 2 global inequality can be seen as the between-country component of con­cept 3 global inequality. Concept 2 inequality tells us what concept 3 inequality would be if there were no inequality within countries and each person in a country received the national (household) income per capita of that country. For decomposable measures of inequality, concept 3 inequality will then be equal to concept 2 inequality plus a weighted average of inequality within countries (the within-country component of concept 3 inequality). We will discuss these distinctions further when we present our calculations later in this chapter.

Table 11.1 summarizes the four global income distributions defined in terms of unit of analysis (population unit), associated ranking variable (income concept), and numeraire. It is important to emphasize that the four different concepts of global inequality can move in different directions. It should be immediate from the decomposition just mentioned that concepts 2 and 3 can move in different directions: a modest fall in between-country (i.e., concept 2) inequality may coexist with a rise in concept 3 global inequality if within- country inequality increases sufficiently.

Moreover, the same changes in national income may have opposite effects on different concepts of inequality. For example, China is the second-largest economy in the world, both in PPP$ and in current US$, but its per capita GDP in 2012 was PPP$7960, below both the unweighted mean GDP per capita across countries of PPP$12,300 and the population-weighted mean across countries of PPP$10,260.^[657] The fact that China’s above­average total national income has been growing much faster than the world average there­fore implies that China is a disequalizing force for concept 0 inequality.

Consider now the notion of “convergence” in the literature on economic growth, which is the closest that many economists get to thinking about global inequality. There are two commonly used definitions of “convergence,” namely beta and sigma conver­gence. Beta convergence means that when a country’s growth rate is regressed on its national income per capita, the coefficient on income is negative and significant.^[658] Thus, on average, countries with higher per capita national income (where per capita GDP is the measure typically used in these studies) have lower growth rates. Sigma convergence means that the dispersion across countries of per capita national income declines over time, often measured by the standard deviation of the logarithm of per capita national income. Both therefore refer to the concept 1 global distribution with the country as the population unit and per capita national income as the income concept.

In their survey of growth econometrics, Durlaufet al. (2009, p. 1098) state that sigma convergence has “a natural connection to debates on whether inequality across countries is widening or diminishing.” If “inequality across countries” refers to concept 1 inequal­ity, then it is a tautology that sigma convergence will measure “whether inequality across countries is widening or diminishing.”^[659] However, sigma convergence or divergence has no necessary connection to any other concept ofinequality across countries (e.g., concept 0 inequality as seen in the China example) or to global inequality across a different pop­ulation unit (e.g., individuals). A rise in the dispersion across countries of per capita national income (i.e., concept 1 inequality) may be associated with a fall in concepts 2 and 3 global inequality, as the following example demonstrates.

The Philippines has a population of 97 million people, and a per capita GDP of PPP$3800. Its per capita GDP is below both the unweighted and the population- weighted world means of PPP$12,300 and PPP$10,260, respectively, noted earlier. There are 35 countries that both have populations below 5 million and also, like the Philippines, have per capita GDP below the respective unweighted and population- weighted world means. For the purposes of sigma convergence, each of these 35 small countries has the same weight as the Philippines, yet their combined populations amount to 57 million, below that of the Philippines.^[660] Now imagine that there is sigma diver­gence, where all other countries are growing at a common rate, but the Philippines is growing faster and the 35 small countries are growing slower than this common rate. Global inequality is increasing according to concept 1 because while one country (the Philippines) whose per capita GDP is below the unweighted world mean is converging to the world mean, 35 other countries whose per capita GDP is below the world mean are diverging from it. But global inequality may be decreasing according to concept 2 because the convergence of the Philippines’ large population toward the weighted world mean outweighs the divergence of the populations of the 35 small countries away from the weighted world mean. Assuming that inequality within countries is unchanged, global inequality may therefore also be decreasing according to concept 3.

We conclude that global inequality tout court is an underspecified concept, and esti­mates of different definitions of global inequality can move in different directions—as we find in our empirical estimates in Sections 11.5 and 11.6.

11.4.