Rotation group bias in European Union social indicators

As EU-SILC relies on a four-wave rotational panel
design, a new population sample is drawn every
year and selected respondents are interviewed
annually for up to 4 years. This means that a complete
EU-SILC cross-sectional data set contains data
from samples drawn independently in 4 different
years. In Chapter 17, Alessio Fusco, Giovanni Gallo
and Philippe Van Kerm apply influence function (IF)
regression methods to examine to what extent this
design has an impact on the estimates of four EU
social indicators (the AROPE AROP indicators, the
MSD indicators, introduced above, and the Gini
coefficient (7)), in other words if a rotation group
bias in these indicators is observed. Their analysis
highlights that estimates of income inequality and
poverty rates for newer rotation groups are often
higher than for older ones. Individuals interviewed
at least twice (and especially those who have been
in the sample for 3 or more years) tend to drag the
selected social indicators downwards, compared
with individuals interviewed for the first time. These
impacts remain significant even when accounting
for observable sociodemographic characteristics
of households; not all countries are affected by the
bias, however.