Measuring unemployment composition over the cycle

This paper proposes a strategy to measure, in a unified setting, how the job finding probability and the job separation probability conditional on observable and unobservable individual characteristics varies over the business cycle. Recent papers by Shimer and Hall point out how new measurements of the job finding probability explain unemployment volatility significantly better than the separation rate. Accepting the above measurements at face value, it is still to be determined why this is the case. A natural step is analyzing in micro detail the dynamics of the job finding and separation probabilities and trying to discover if its cyclical variation is linked to worker characteristics. This is important for at least two motives: 1) In guiding macro modelling, in so far as it can tell if heterogeneity should be an important element in theoretical models and 2) for policy analysis, for instance targeting specific kinds of unemployed workers (or not targeting them) during recessions. The econometric model proposed is semi-structural and it specifically accounts for unobserved heterogeneity that generates from both basic exogenous "initial conditions" and from endogenous match specific quality or experience