The orthodox explanation of unemployment that argues that institutions matter20 has been
subject to fairly extensive econometric testing, and in recent years, the validity of the
empirical results supporting this view has been called into question. It has proved
difficult to estimate a set of cross-country panel unemployment regressions that contain a
lagged unemployment rate and a full set of year and country dummies and show that any
of the labour market rigidity variables work. This is a crucial test. This is the first main
similarity between European labour markets: labour market institutions do not tend to
cause unemployment. The major exception is changes in the replacement rate, which, in
some specifications, do appear to be negatively correlated with changes in the
unemployment rate. Blanchard and Wolfers (2000) have argued that “the interaction of
shocks and institutions does a good statistical job of fitting the evolution of
unemployment both over time and across countries.” This result is questionable because
it is obtained in an over-fitted model — few data points and lots of variables —and the
results appear to be driven by the cross-section variation rather than by any time series
changes.21 There are only eight time series data points as they use five year averages
from 1960-1995.