Multiple unemployment spells duration in Poland

I study multiple unemployment spells. I refer to Poland, as this country experiences high incidence of the long-term unemployment and long mean incomplete unemployment duration. I estimate conditional risk set model for multiple spells, and Cox model for 1st to 5th spell, separately. I use time-to-event data (complete spells) for almost 435000 individuals in the 2007-2014 period from five public employment offices. Almost 99% of the individuals have up to 5 spells, but ''only'' 60% of them experience one complete spell.I find that subsequent unemployment spells are on average shorter, but slightly more individuals experience longer subsequent spells. Thus, those who exit either learn how to seek job or decide to leave the pool. Younger males with primary education are better off at leaving the pool, but the impact of sex and education diminishes in subsequent spells. Tertiary education, since 2nd spell, also entails higher hazard of leaving the pool. Results indicate heterogeneity among job seekers and mismatch bet