Brian Nolan, Ive Marx and Wiemer Salverda
nificantly affect trends over time (as US experience with the CPS illustrates), and it is essential to be aware when
this is being done.
Even where the definition and coverage of the income concept is the same, differences in the way income is
actually measured may give rise to variation across sources and datasets. The most obvious example is the differ-
ence between data drawn from administrative sources versus household surveys. Data from population registers
drawing inter alia on the records of the tax and social security systems have been available for many years for
Scandinavian countries, whereas many other countries have relied on household surveys. The ECHP was survey-
based, whereas the EU-SILC framework allows countries to incorporate data from different sources, including
administrative records. The difference the use of administrative versus survey-based data could make to measured
levels of inequality is unclear but needs to be kept in mind.4 Even where data are all taken from surveys, there may
be significant differences in the way specific income sources are captured. This is particularly pronounced in the
case of income from self-employment. The Canberra Group refers to the “net operating profit or loss accruing to
working owners of, or partners in, unincorporated enterprises” (page 118), while for EU-SILC purposes Eurostat
guidelines refer to the annual profit/loss from accounts, taxable income, and the yearly amount drawn out of the
business. There is considerable scope for differences in measurement practices to affect the income estimates, to
which the level of measured inequality in turn may be sensitive since self-employment incomes may represent a
significant proportion of cases at both the top and the bottom of the distribution. A related issue where practices
vary is the treatment of negative incomes, most often losses on self-employment income. Such negative incomes
may be amended or set to zero when the data are processed, or they may simply be reported. The ECHP, for exam-
ple, set negative incomes to zero, whereas they are reported in EU-SILC.
The Time Period
The time period over which income is measured is another significant issue, and one where practice once again
varies across sources. This matters because lengthening the reference period will generally smooth out transitory
fluctuations and thus reduce measured inequality. The Canberra Group recommended that a year be used as the ref-
erence period, on the basis that this was the natural accounting period for sources such as self-employment income
as well as for data derived from the income tax system. However the Group also noted that different periods may
4 One may venture that inequality of gross earnings may be better measured administratively while personal/household surveys may do
relatively better for net earnings.
Page • 14