Worlds Apart: Measuring International and Global Inequality.

Minoiu, Camelia

Worlds Apart: Measuring International and Global Inequality, Branko Milanovic (Princeton: Princeton University Press, 2005), 240 pp., $29.95 cloth.

The extent and trend of world inequality are generally taken to be important for assessing the justice of the contemporary model of global economic interaction. Worlds Apart: Measuring International and Global Inequality represents one of the more satisfying attempts to date to describe the world distribution of income and how it has changed over time. Branko Milanovic's approach to assessing world inequality is in many respects superior to previous studies. The book's strengths lie in its use of individual-level data from household surveys for a large number of countries (unit data), rather than exclusively relying on income profiles derived from income shares (tabulated data); its analysis of interpersonal global inequality using data from 1988, 1993, and 1998--those years for which a significant share of the world's population is covered; and its rich discussion of the consequences of various assumptions for global inequality estimates.

Milanovic defines and discusses three working concepts of inequality. First, in unweighted international inequality, the units among which inequality is assessed are countries and each country is attributed an income corresponding to its per capita GDP. Second, in population-weighted inequality, the units are individuals assumed to possess an income level equal to the per capita GDP of the country in which they live. Third, in global inequality, the units are individuals, and each individual is assigned an income level estimated from representative household surveys. The more appropriate measure for world inequality is identified through the third concept, since the first assumes that countries are of equal size and have perfectly egalitarian distributions of income, whereas the second concept addresses the former shortcoming, yet retains the latter. In subsequent parts of the book, the author computes a wealth of inequality indicators, focusing in particular on the Gini coefficient and the Theil index, in order to assess the respective extent and trend of global inequality these three working concepts capture. Finally, he discusses his inequality estimates in historical perspective and argues for the policy relevance of undertaking global inequality assessments.

One of the main conclusions of the book is that global inequality (defined under the third concept) stood at a high level in 1988, as illustrated by a Gini coefficient of 0.62. It rose between 1988 and 1993 to a Gini coefficient of 0.65 and fell to a Gini coefficient of 0.64 by 1998. If one thinks of the world as a single country, these findings suggest that the world is as unequal as some of the most unequal societies--such as Brazil and South Africa, whose Gini coefficients in 1990 were 0.61 and 0.63, respectively (see UNU/WIDER, "World Income Inequality Database V 2.0a," June 2005). The analysis also reveals that "the greatest contribution to the world Gini coefficient are large countries that are at the two poles of the income distribution spectrum: among the poor countries, China and India ... and among the rich countries, the United States, Japan, Germany, France, and the United Kingdom" (p. 111). Finally, Milanovic attributes the decrease in global inequality between 1993 and 1998 to the high growth rates of average incomes in rural areas in China and India, which have "subtracted a full Gini point from overall inequality" between the two years (p. 114).

Milanovic's analysis has some limitations, however, that raise questions about the validity and robustness of his conclusions. The first issue regards the use of purchasing power parity conversion factors (PPPs) to transform the income (and consumption) of individuals in different countries and different years in "equivalent" U.S. dollars. PPP transformations are aimed at translating the incomes of countries (or individuals), originally expressed in local currency units, into the currency of a "base" country and at the prices of that country in a base year. As such, the PPP-adjusted income (consumption) estimates are intended to capture "equal" purchasing power over a basket of goods. This data, however, is used in the book despite the fact that the way in which PPPs are constructed renders them inappropriate for poverty and inequality assessments. For example, a central problem associated with currently available PPPs is that PPPs calculated for poor countries are likely to be understated. (For a critique, see Sanjay Reddy and Thomas Pogge, "How Not to Count the Poor" (2005); available at www.socialanalysis.org.)

The second issue regards the choice of estimates of countries' true average income. Milanovic opts for the National Accounts-based estimates to compute population-weighted inequality, and for survey-based estimates to compute global inequality. The primary reason underlying his second choice are the widely documented discrepancies between survey-based and National Accounts-based estimates of the level and growth rates of income and consumption. It is difficult to assess theoretically the magnitude and the direction of biases that may result from choosing either of the two estimates; however, it is noteworthy that this choice will not be inconsequential for the accuracy of world inequality figures.

A third issue concerns the use of consumption and income as the relevant indicators of country and individual welfare. The information provided by income and consumption may fail to quantify satisfactorily either welfare or living standards. This is because factors such as access to health and education are important in determining the level of advantage of countries and individuals--hence, they should be taken into account when evaluating welfare.

A fourth issue concerns the use of unadjusted per capita consumption/income instead of adult equivalent consumption/ income, although the latter reflects better differences in household structure. Equivalence scales are adjustments to consumption patterns that reflect the gender and age composition of households. They are parameters of relevance in inequality assessments because adults have different consumption requirements than children. The use of equivalence scales, however, is difficult because of the lack of estimates for a large number of countries. Moreover, in Milanovic's analysis, the use of tabulated instead of unit data for several countries makes the use of equivalence scales practically impossible.

A fifth issue, which Milanovic discusses to a lesser extent despite its significant implications for the accuracy of inequality estimates, regards the use of tabulated instead of unit data to measure inequality for a number of countries, such as China, India, Indonesia, and Bangladesh (even though large-scale sample surveys are available for India through the National Sample Survey Organization). In particular, inequality in these countries is estimated from income profiles of five to twenty "observations," each observation representing the average income of large groups of individuals. These groups are obtained by ordering the income levels of a country's individuals in ascending order and by dividing the population into a few equal-sized groups (the number of such groups is commonly between five and twenty). For example, average incomes for groups comprised of 180 million individuals are the basis for estimating rural China's income distribution. Naturally, this data structure requires the analyst to assume that there is no within-group inequality--an assumption that gives rise to downward biases in national and global inequality estimates. More generally, tabulated data represent a weak informational basis for describing features (of which inequality is one example) of a country's entire income distribution.

The main implication of these observations is that definite conclusions regarding the extent and trend of global inequality cannot be drawn before a serious pause. It is conceivable that an estimation exercise of global inequality, which uses superior PPPs, better estimates of average incomes, and adult equivalence scales, and relies entirely on unit rather than tabulated data, may lead to a markedly different outlook on global inequality (such analysis is used to measure global poverty in Sanjay Reddy and Camelia Minoiu's "Has World Poverty Really Fallen in the 1990s?" (2005); available at www. columbia.edu/~sr793/sensitivityanalysis.pdf). With this call for caution in mind, it can be concluded that Worlds Apart offers a thorough description of relative inequalities in the world, and does so by setting research quality standards to which future studies should be held.

--CAMELIA MINOIU

Columbia University