Institutions and unemployment: Do interactions matter?

Isolated effects of labor and product market institutions as well as the interaction between both aforementioned categories on unemployment have been extensively discussed in the empirical literature. However, interaction effects between individual labor market institutions have been widely neglected, mainly due to the infeasibility to correctly specify the model. In this paper, a model averaging approach is adopted to show that considering institutional interactions can improve the explanatory power of macroeconomic models explaining unemployment. The approach permits to tackle model specification problems directly related to the inclusion of a large number of interactions. Using a panel data set for 17 OECD countries from 1982 to 2005, 22 robust and significant interactions can be identified. Furthermore, country-specific marginal effects of institutional changes are calculated and their economic significance is analyzed for selected countries.