Heterogeneity in the dynamics of disaggregate unemployment

This paper explores the role that unobserved heterogeneity within an observed category plays in the dynamics of disaggregate unemployment and in the cross-sectional differences across individuals of the duration of unemployment spells. The distribution of unobserved heterogeneity is characterized as a mixture of two distributions with each mean and weight determined by the inflows and outflows of workers with unobserved types H and L, which are identified based on the nonlinear state-space model of Ahn and Hamilton (2016). I found that the contribution of each factor to the dynamics of disaggregate unemployment differs by observed category. The inflow of type L workers is the most important factor in the majority of demographic groups in the business-cycle frequency. I identify permanent job loss to be the observable characteristic most closely associated with the type L attribute. A simple model of heterogeneity based on two unobserved types can account for explain more than 50 percent of the cross-sectional dispersion in completed-duration spells after the Great Recession, while observed heterogeneity makes only a minor contribution