This paper develops and estimates a split population model for the duration of temporary layoffs in the German labour market; the population being split according to whether a layoff is temporary or permanent. A flexible piecewise constant hazard is employed and the recall propensity is used to condition the duration of a temporary layoff. The distinction between factors that influence the recall propensity and factors that influence the duration of a temporary layoff is shown to be of empirical significance. The results of the paper are used to evaluate implicit contract and search theories of temporary layoffs. The paper suggests that implicit contracts can explain the behaviour of unemployed workers with high worker-firm attachment whilst search can explain the behaviour of workers with low worker-firm attachment. Copyright (C) 2004 John Wiley Sons, Ltd.