How long workers stay in their jobs is of central importance for individuals’ work histories, employers’ personnel policies and the functioning of the labor market. Within and across industries, companies, institutional regimes and groups of workers, employment spells differ vastly in duration. Using German linked employer-employee-data we analyse job durations from both sides of the employment relationship. We address two main questions. First, what is the relative importance of worker characteristics, firm-level variables and institutions for job duration, and how are they related to each other? Is there evidence for segmentation within companies, such that a core workforce is protected against job losses by the employment of a marginal workforce with short tenure on average, or do observed differences in tenure mainly stem from different tenure levels at different firms? A second question is whether effects differ between different exit states namely between job-to-job and job-to-unemployment changes. Kaplan–Meier survivor functions show that more than fifty percent of all new employment relationships end after in the first two years. The estimation results indicate huge differences between individuals but also between establishments with different characteristics. We conclude that persistence in individual mobility behavior is much reduced when firm-level heterogeneity is accounted for. Blue collar workers tend to select into long-tenure firms as compared to white collar workers. Positive effects of works councils and further training on job duration are high for blue collar skilled or semi-skilled workers but non-existent for white collar employees. Competing risks analysis shows that mobility to another job and exit to unemployment follow strikingly different processes. For instance having a university degree lowers the unemployment hazard but increases the likelihood of a job-to-job change

Keywords

Job durations, job exit, tenure, linked employer-employee data