Unconditional and conditional institutional effects on unemployment: a Bayesian model averaging approach

Abstract

Labor and product market institutions are regarded as central factors for explaining differences in the evolution of unemployment across countries. It is generally assumed that reforms which serve to reduce institutional rigidities are a valuable approach to reduce unemployment. However, both the theoretical as well as the empirical literature provide inconclusive results to the question of which institutions matter for unemployment, and which reforms help to lower the unemployment rate. A large number of potentially relevant institutional factors makes it difficult to specify the econometric model correctly. Additionally, possible interdependencies between institutions enlarge the set of explanatory factors considerably, which further complicates the derivation of valid estimates. This thesis clearly identifies six institutional factors which contribute significantly to the explanation of unemployment rates in 17 OECD countries between 1982 and 2005. The main innovation is the application of a Bayesian model averaging approach which allows the clear-cut identification of significant institutions when the number of explanatory factors is high and observations are limited. Furthermore, this approach also proves to be well-suited for the incorporation of institutional interdependencies which can be estimated systematically for the first time. 22 bivariate interactions are identified as significantly related to the unemployment rate. The outcome of this analysis allows the determination of potential reform effects for various institutions for all 17 OECD countries separately, what is a substantial improvement over the existing empirical as well as theoretical literature.