Working Paper 2015-382
In response to a growing interest in comparing inequality levels and trends across countries, a number of cross-national inequality databases are now available. These databases differ considerably in purpose, coverage, data sources, inclusion and exclusion criteria, and quality of documentation. A special issue of the Journal of Economic Inequality, which this paper introduces, is devoted to an assessment of the merits and shortcomings of eight such databases. Five of these sets are microdata-based: CEPALSTAT, Income Distribution Database (IDD), LIS, PovcalNet, and Socio-Economic Database for Latin America and the Caribbean (SEDLAC); two are based on secondary sources: “All the Ginis” (ATG) and the World Income Inequality Database (WIID); and one is generated entirely through multiple-imputation methods: the Standardized World Income Inequality Database (SWIID). Although there is much agreement across these databases, there is also a non-trivial share of country/year cells for which substantial discrepancies exist. In some cases, different databases would lead users to radically different conclusions about inequality dynamics in certain countries and periods. The methodological differences that lead to these discrepancies often appear to be driven by a fundamental trade-off between a wish for broader coverage on the one hand, and for greater comparability on the other. These differences across databases place considerable responsibility on both producers and users: on the former, to better document and explain their assumptions and procedures, and on the latter, to understand the data they are using, rather than merely taking them as true because available.
Authors: Francisco H. G. Ferreira, Nora Lustig, Daniel Teles.