Multistate Models for Clustered Duration Data—an Application to Workplace Effects on Individual Sickness Absenteeism

Abstract

Sickness absenteeism figures show a relatively large amount of variation across firms and organizations, indicating substantial within-firm correlations between absenteeism records of individual workers. To study the role of firm-specific circumstances and workforce composition, we specify three-state, multicycle duration models of work, sickness, and job separation, with workplace-specific fixed effects to account for unobserved differences between firms. In the most flexible specification, these fixed effects are separate, nonparametric, baseline hazards for each firm and each type of transition. Alternative estimation methods are discussed and applied to individual absenteeism histories of primary-school teachers.