Comparable Indicators of Inequality Across Countries
produced for a number of countries (7 OECD countries - US, Canada, UK, Germany, Italy, Spain and
Sweden — together with Russia and Mexico) underpinning the papers published in a special issue of the
Review of Economic Dynamics journal in January 2010 (vol. 13, issue 1).
c. Time-series analysis of income inequality with wide coverage of OECD countries over the shorter pe-
riod from the 1980s can be undertaken with more reliable data on summary inequality measures from
LIS (Gini, Atkinson, percentile ratios in LIS Key Figures) or for a somewhat larger set of countries us-
ing data from the OECD’s Growing Unequal (Gini, S80/S20, mean log deviation, standardised coefficient
of variation), but only for certain years at 5-year intervals rather than annually. These sources sometimes
differ from other sources for certain countries, in which case alternative estimates could be tested.
d. For time-series analysis with no more than a maximum of 15 observations, for 14 EU countries only,
one can use the annual figures for the Gini and the S20/S80 inequality measures produced by Eurostat
from ECHP (1995-2001) and EU-SILC (2003 or 2004 to 2008 or 2009); these are the key series relied
on by the EU and of significant interest for that reason, but the reliability of the trends over time is
compromised by attrition in the ECHP and the break in the series between the ECHP and EU-SILC
which introduced significant non-comparabilities.
Pooled time-series/cross section analysis
a. Pooled time-series + cross-section analysis with significant coverage of OECD countries from the
1980s can be undertaken with micro-data from LIS for certain years at 5-year intervals, though with only
a limited number of non-income variables available for analysis and some unavoidable non-comparabil-
ity given harmonization is applied only after the event to national data.
b. From 1995 to 2001, this could be applied to micro-data for 14 “old” EU member states from ECHP,
with a much wider set of non-income variables available in the dataset and a high degree of harmoniza-
tion of definitions and variables since a uniform questionnaire/survey methodology was used (though
panel attrition remains a source of concern).
c. For a short period of time from the mid-2000s, this type of analysis can be carried out with a much wid-
er set of EU countries using data from EU-SILC, which also has a very substantial set of non-income
variables; the “output-harmonised” nature of EU-SILC has to be taken into account in assessing the
comparability of variables across countries.
Page • 35