We use microdata obtained from the Turkish Employment Agency (ISKUR) to shed light on the functioning of the unemployment insurance (UI) system over the period 2010-16. We examine the individual, institutional and macroeconomic determinants of exit rates to employment using a variant of the Cox-PH model. We rely on stratification that permits arbitrary baseline hazard functions that vary by strata. Effects of the explanatory variables on the hazard are assumed to be constant across strata and are estimated using the Stratified Partial Likelihood Estimator (Ridder and Tunali, 1999). Subsequently we are able to obtain non-parametric estimates of the cumulative hazard rate by stratum, and examine how the exit rate varies across strata. We initially stratified by PBD and found striking violations of the proportional hazards assumption for indicators of gender and "type" of termination, a classification based on the reason provided by the employer. We argue that these could serve as markers for differences in the re-employability of the unemployed individuals and stratify further, by PBDxgenderxtype of termination. We then exploit various features of the institutional set-up and use quasi-experiments - a DD formulation and two RD designs { that allow us to tease out causal effects of changes in UI benefits and statutory benefit duration (PBD). We conclude that the UI system - one of the least generous among comparator countries - does not suffer from major incentive problems.
Research Fellows
Insan Tunali
Associate Professor Emeritus, Koç University
Authors
Mustafa Ulus
Associate Professor of Economics, Galatasaray University