Statistical Analysis of Labor Market Integration: A Mixture Regression Approach

In this paper we investigate the labor market integration of young males in Finland. The study is based on individual-level registry data which contains information about working, studying, sickness and maternity benefits. Using data from 2005 to 2013, we have studied the labor market attachment of the birth cohort of 1987, which totals 29,383 individuals. The statistical methodology is a multivariate version of mixture modeling applied to longitudinal data. We have modeled the information about labor market outcomes as a binary dependent variable, and thus we apply a multivariate logistic mixture regression model. We found that there are ten main clusters or groups that lead to different labor market outcomes. Our results suggest that our mixture regression approach applied to register-based population can reveal new information that may remain hidden in more formal, census-based labor market statistics. To our knowledge, application of finite mixture modeling methods have not yet been applied to extensive and comprehensive pension data of this sort. Our analysis also provides valuable information for policy-makers and good statistical tools for corresponding analyzes of register data.