Unemployment Duration Variance Decomposition a la ABS: Evidence from Spain

In a recent paper, Alvarez, Borovickova, and Shimer (2014) revisit the analysis of the determinants of unemployment duration by proposing a new method (the ABS method hereafter) that directly estimates the importance of each component and implementing it using precise information on unemployment spells from social security administrative data for Austria. In this paper, we apply the ABS method to social security administrative data for Spain with the objective of comparing these two very different labor markets as well as Spain along the business cycle. Administrative data have many advantages compared to Labor Force Survey data, but the incomplete nature of the data needs to be addressed in order to use the data for unemployment analysis (e.g., unemployed workers that runout of unemployment insurance have no labor market status in the data). The degree and nature of such incompleteness are country-specific and are particularly important in Spain. Following Lafuente (2018), we approach the matter of data incompleteness in a systematic way by using information from the Spanish LFS data as well as institutional information. We hope that our approach will provide a useful way to apply the ABS method in other countries. Our findings are as follows: (i) The aggregate component is clearly the most important one, followed by heterogeneity and duration dependence, which are roughly comparable. (ii) The relative importance of each component and, in particular, duration dependence is quite similar in Austria and Spain, especially when minimizing the effect of fixed-term contracts in Spain. Similarly, we do not find big differences in the relative contribution of the different components along the business cycle in Spain. (iii) These comparisons suggest that statistical discrimination due to dynamic sample selection does not seem to be the main driver of duration dependence.