The competitive model of the labor market predicts that the development of individual earnings over the life cycle follows the development of individual marginal productivity. Beside productivity related factors such as on-the-job learning and improvements in worker-firm matches over time, shocks to aggregate labor demand for instance due to a major recession – will also have an impact on wage rates. In a spot labor market, however, those temporary changes in labor demand are relately short lived and should not influence wages over prolonged periods of time. This view has been seriously challenged both by studies on cohort size effects (Welch, 1979) and studies on the impact of early career problems on later outcomes.
The general approach taken by these studies is to assess the initial wage or employment penalties from entering the labor market in a bad year and to test whether this initial impact persists over time. Raaum and Røed (2006), e.g., show for Norway that school leavers facing particularly depressed labor market conditions at the start of their career face a higher risk of unemployment both initially and after ten years. Oreopoulos et al. (2008) study careers of Canadian college graduates and find a high initial wage penalty of entering in a recession, but the penalty fades away during the first decade of a worker’s career.
In this paper we study a different aspect of the impact labor market entry conditions can have on career development. We depart from the traditional strategy of modeling wage or employment outcomes at a particular point in time and focus on mobility throughout the complete career path instead. Thereby our aim is twofold. First, we want to identify specific career patterns that characterize the earnings development of individuals after entry in the labor market. The idea is to extend the traditional mover-stayer classification to a wider variety of career types.
Intuitively, some individuals may be in stable employment relationships throughout their working lives, while others are observed in more volatile jobs; still others could be considered as social climbers with a consistent upward mobility, while others could be characterized as losers with a high tendency of downward mobility. Our second goal is to find out whether labor market conditions at the start of one’s career have an impact on the type of career pursued over the lifetime. While entering the labor market in a recession might impose an immediate penalty in the form of lower starting wages, it might also influence the life-time career path; i.e. an individual might be characterized by a different career-type when entering the labor market in a recession as opposed to a boom period.
The statistical problem behind our empirical analysis consists of finding groups of similar time series in a set or panel of time series that are unlabeled a priori. In this paper we introduce new clustering techniques which determine subsets of similar time series within the panel. Compared to cross-sections, distance-based clustering methods are rather difficult to define for time series data. Frühwirth Schnatter and Kaufmann (2008) demonstrated recently that model-based clustering based on finite mixture models (Banfield and Raftery, 1993; Fraley and Raftery, 2002) extends to time series data in quite a natural way. The crucial point in model-based clustering is to select an appropriate clustering kernel in terms of a sampling density which captures salient features of the observed time series. Various such clustering kernels were suggested for panels with real-valued time series observations by Frühwirth-Schnatter and Kaufmann (2008) and Juárez and Steel (2010).
