Academic Commons


Improving Measurement of Latin American Inequality and Poverty with an Eye to Equitable Growth Policy

Berry, Albert

In Latin America the last 30 years have seen enormous strides in the measurement of inequality and poverty at the national (and subnational) level and in the feasibility of making meaningful comparisons across countries and over time in the same country. Latin America is around the norm for developing countries as a whole in this regard; the availability and accuracy of its data is better than in some other regions but weaker than in others. It now permits reasonable guesses as to the true level of inequality in most countries of the region, makes it clear that inequality is higher in some than in others and, usually with less confidence, allows some judgments on trends. The regional distinctions are basically more feasible than the over time judgments, because the observed variance is several times higher; thus the reported Gini coefficient in Brazil is usually around 0.60 and that in Uruguay is around 0.40,1 so the likelihood that non-comparability of estimates would negate the conclusion that inequality is much greater in Brazil than in Uruguay is extremely remote. Over time it is rare to see the reported Gini coefficient for a given country vary by more than 0.04-0.05, so if the error of estimate changed considerably over the period of that change, it is less assured that a true change did occur. Reported changes of up to about 0.05 in the Gini coefficient make it necessary to attach a probability statement to any surmise as to whether a change really occurred. Such caveats notwithstanding, our knowledge has increased greatly, both about the level of inequality per se, its components (e.g. how much is associated with inequality of labor income) and various of its correlates, and with respect to the nature of the remaining measurement errors. The fact that reported inequality tends to be very stable in most countries, in the absence of identifiable changes in measurement (e.g. in the nature of the questions asked in the surveys) suggests that true inequality is also stable. The main possible exception to this generalization is that overall measurement error could change over time in a way closely related to changes in true capital income distribution and/or true capital share; thus reported capital income and its impact on reported inequality might not change while true capital income and true overall distribution did. Such a change might result from financial liberalization, frequently argued to have contributed to rising inequality in developing countries (Cornia, 2004; Behrman et al, 2000).



  • thumnail for IPD_WP_Improving_Measurement_of_Latin_American_Inequality.pdf IPD_WP_Improving_Measurement_of_Latin_American_Inequality.pdf application/x-pdf 184 KB Download File

More About This Work

Academic Units
Initiative for Policy Dialogue
Initiative for Policy Dialogue
Initiative for Policy Dialogue Working Paper Series
Published Here
November 1, 2012