Distinct settings of labor market institutions like the employment protection or the unemployment benefitt system have attracted considerable attention as a potential explanation for differences in the unemployment rates of industrialized countries over the last two decades. A plethora of theoretical and empirical studies have dealt with the identification and quantification of direct labor market effects of institutional reforms. However, while theory predicts that the interplay between individual labor market institutions is as well important to determine the impact of institutional reforms, empirical studies have widely neglected such interdependencies so far.

The main problem in empirical studies is that macroeconomic labor market models quickly become very large if interactions are taken into account. Hence, the estimation of a model considering a set of institutional interactions requires either exact and comprehensive theoretical predictions on which interactions to include or a large number of observations to receive reliable results. Unfortunately, theoretical studies mainly focus on broad concepts of institutions like the bargaining power or the firing costs, and empirical data-based models cannot be directly derived from theory. The low number of available observations requires the subjective selection of some interactions, what is also not an appropriate solution because neglecting potentially relevant information can severely bias the outcomes.

In this study, I use a bayesian model averaging framework to estimate reliable parameters for all available bivariate interaction terms. Using data on 14 institutional indicators of 5 institutional categories (product market regulation, employment protection, unemployment benefit system, labor tax system, bargaining system), 91 bivariate interactions are analyzed concerning the question whether these interactions can significantly contribute to the explanation of unemployment.

On the basis of the model averaging approach, I identify 22 robust and significant bivariate interaction terms. The empirical evidence emphasizes the importance of institutional interactions for the determination of unemployment. More concretely, taking interactions into account significantly improves the explanatory power of the empirical model. The calculation of country-specific marginal effects of institutions sheds light on the question why institutional reforms might result in different outcomes in different countries in terms of unemployment. Furthermore, the results can give advice how reform-packages implemented to tackle labor market rigidities should be designed in order to decrease unemployment.

Sachs, Andreas (2011), Institutions and Unemployment: Do Interactions Matter?, ZEW Discussion Paper No. 11-057, Mannheim. Download


Unemployment, Institutions, Labor and Product Markets, Model Averaging, Institutional Interactions, Institutional Design