B. Assumptions behind income distribution estimates
Estimating income inequality based on national distribution estimates is no easy
task. In an ideal world, there would be cross-nationally comparable household
surveys for all countries over time with mean income estimates for different
population deciles or quintiles derived from those surveys. In reality, however,
household surveys are based on various methodologies, ranging from
consumption (with and without transfers) and expenditure to earnings (gross and
net) and income (monetary and taxable or disposable and gross). Moreover,
household studies are not carried out on a regular basis in most countries, with
methods often changing between studies. Since existing national income
estimates must be converted from national currencies in order to be compared,
there is also the issue of the most appropriate exchange rate (see Box 1 for
discussion on using market versus PPP-adjusted exchange rates).
To date, the World Bank’s PovcalNet offers the best attempt to create the ideal
income distribution database (Note: distribution estimates published in the World
Bank’s World Development Indicators—referred as World Bank 2011 in this
paper—are derived from PovcalNet). Using nearly 700 household surveys across
116 developing countries, it contains income/consumption distribution
information along with mean per capita income/consumption estimates based on
the latest PPP exchange rates (2005). Regrettably, however, PovcalNet does not
include information for any developed countries and is further characterized by
large data gaps over time.
Given our objective to understand global inequality trends since 1990, we
sacrifice the quality assurance of income/consumption estimates offered by
PovalNet in favor of an expanded sample of countries and across more time. We
do so by complementing PovcalNet’s estimates with income/consumption
distribution data compiled by UNU-WIDER as well as by Eurostat. The UNU-
WIDER and Eurostat data suffer from differing calculation methods and do not
offer mean per capita income/consumption estimates based on household
surveys. We further acknowledge the statistical inaccuracies in comparing
distribution estimates from three unique data sources. Comparability
shortcomings aside, combining these sources enables us to carry out rough
approximations of income/consumption distribution estimates in a sample of
136 countries between 1990 and 2007.
The expanded sample requires us to use a less preferable income gauge: GDP.
Using GDP as an income metric is inherently flawed given that investment and
government spending are assumed to be distributed in the same way as
household consumption (or disposable income). As a result, while GDP includes
items which may have something to do with future welfare, it is not an accurate
measure of current income (e.g. consumption expenditure in China is less than
40% of GDP as of 2009). Despite the measurement inadequacies, our intent is to
show the general evolution of income distribution over time, and our calculations
assume that the distribution of total household income/consumption and total
GDP are equally proportionate. Adopting the GDP metric further allows us to