believe that labor market rigidities are to blame, at least to some degree, and that these rigidities
manifest themselves through the use of monopsonistic bargaining by unions, overly generous benefits
that prolong unemployment by removing job search incentives or high labor income taxes. These
ideas have provided the cornerstone of OECD orthodoxy and lead to strong policy recommendations.
However, academic economists have been rather more reluctant to make such bold claims without
at least some qualifications. Glyn et. al. (2003) provide a masterly review of the existing literature
and conclude that the statistical evidence is in fact at best ambiguous and OECD studies have
not been successfully replicated in the academic world, which seems to have become inextricably
entangled in debates over particular measures and econometric estimation methods.
We propose here to expand the institutionalist approach in two directions. First, our claim is that
if institutions matter, we need to look beyond labor market institutions alone. Hence we propose a
full political economy approach that takes political and social institutions into account, in addition
to labor market institutions. Second, our claim is that in order to determine how these institutions
affect unemployment, we need to look at institutional complementarities. A complementarity exists
when "the presence (or efficiency) of one institution increases the returns from (or efficiency of)
the other" (Hall and Soskice, 2001). Because we are interested in unemployment, we build on
current macroeconomic and labor theory and examine the way labor institutions are embedded and
influenced by other institutions. The institutional complementarities we examine are hence focused
on our central question of the determinants of unemployment. In Section 2 we review the relevance
of some of the existing political science literature on institutionalism and varieties of capitalism to
the study of unemployment. We then construct a uniquely comprehensive dataset that builds on
those used by Glyn et. al. (2003), Nickell (1997), Nickell et. al. (2002) and Blanchard and Wolfers
(2000), by expanding the available dimensions to include a large set of other institutional indicators
that cover diverse institutions rooted in the political, social, cultural, financial, international or
technological fabric of society. We perform a detailed statistical analysis of this dataset in Section
3, and show uncover the complex patterns through which institutional complementarities come into
play in determining the effect of labor market institutions on unemployment.
In Section 4 we extend the statistical analysis using quantile analysis, which allows us to depart
2
Convention