Comparable Indicators of Inequality Across Countries
research literatures, but also because research on earnings has generally wanted to distinguish between trends at
different points in the distribution, notably towards the bottom versus the top, rather than relying on one single
summary measure. Research on earnings dispersion thus tends to be heavily reliant on percentile ratios, including
the ratio of the 90th and the 10th percentiles to each other and to the median. Single summary measures such as
the coefficient of variation are also commonly used, often together with percentile-based indicators, and variance
decomposition methods employed.
There is particular interest in capturing and studying low pay, and the common practice in doing so is to use
thresholds derived as proportions of median or average earnings, such as half the mean or two-thirds of the median,
and derive the proportion of low-paid employees on that basis. Note that this differs markedly from the US concept
of poverty wages, which are based on the (poverty of the) receiving household based and not on the earnings dis-
tribution itself - essentially in the same vein as poverty definitions differ between the US and Europe.
Sources
Data on the distribution of earnings are generally drawn from administrative sources (associated with collec-
tion of tax and social insurance contributions), or from surveys of employees and/or employers. Each has advan-
tages and disadvantages. There is considerably scope for reporting error both in earnings and in hours of work
- which will affect the calculation of hourly earnings - and there may be systematic differences in the reliability of
information gathered from establishments versus individuals. One might expect employers to have more accurate
data on earnings and hours from payroll records, but individuals may have more accurate information on their own
characteristics including education levels. Establishment surveys often have restrictions in terms of coverage not
shared by household surveys, notably for small firm sizes and certain sectors of the economy (such as agriculture
or the public sector), and the data produced from administrative sources may reflect the specific requirements of
the tax and social insurance system rather than the concept/measure desired from an economic perspective.
While many country-specific studies on the distribution of earnings and skills differentials have been carried
out using national sources, in recent years comparisons of low pay and earnings dispersion across countries often
rely on the database brought together by the OECD (see OECD 1996). This is an extremely valuable compilation,
regularly updated, which allows both up-to-date comparisons and trends over time to be studied. However, the
OECD make clear that the nature of the exercise - bringing together data from different national sources - means
that there are potentially important differences in definition and coverage across countries, which are extremely
well documented. The figures for the most part refer to full-time employees, but the period over which earnings are
measured, how “full-time” is defined and measured, and the sectors covered may vary across countries. Thus for
Page • 29