During the first ten years of labor market experience workers transition from high job turnover into stable employment and have rapid wage growth. About two-thirds of life-time job turnover and wage growth occurs during these early years (see Topel and Ward (1992)). Initial high turnover manifests itself in both high job finding and separation rates for young workers.
The mechanism where young workers transition from rapid turnover to stable employment is not well understood. Given the importance of early work experience, a theory is needed to explain labor market outcomes of young workers. A full understanding of these labor market transitions must account for the decline in both job finding and separation rates over the life cycle.
I propose a model where labor market experience allows workers to learn about the quality of future matches. Past work experience allows a jobless worker to differentiate between good and bad matches before accepting a new job. This simple mechanism generates the decline in unemployment, job separation and job finding rates with age, the decline in separations with job tenure, the rise in wages with labor market experience, and fat right tails in the wage distribution.
This paper uses a discrete time version of Jovanovic’s (1979) search model modified to include accumulated labor market experience. The novel feature of the model is that unemployed workers receive a signal about the quality of a new match. The precision of the signal depends on the worker’s experience. Like in Jovanovic (1979) workers are uncertain about the quality of their match and learn about the quality over time by observing realizations of output.
In this model, workers learn rapidly about the quality of their current job by observing output and this experience is useful in helping them determine the quality of future opportunities when seeking a new job. This result contrasts with classic models of learning that have been used to understand wage growth over workers’ careers. These models seek to explain how transferable information learned at a given job is to future employment.
On one extreme is the Jovanovic (1979) model where workers only learn about their current job and all human capital investments are firm specific. Here information is lost once the worker decides to change jobs. On the other extreme are models where workers learn about their entire vector of abilities when employed as in Farber and Gibbons (1996) and Gibbons et al. (2005). In these models, workers learn about their ability to perform at all jobs at a constant rate. This paper provides a mechanism that is between the two extremes. Being able to parameterize how transferable the information gained from experience is when searching for new jobs is a useful tool in analyzing labor market outcomes.
My model has implications for both job finding and separation rates as workers gain experience. Job finding rates are determined in the model by two factors. First, workers receive job offers at an exogenous rate. Second, workers can choose to accept or reject job offers that they receive. The model with experience predicts a decline in job finding rates as experience allows individuals to differentiate between good and bad jobs. For inexperienced workers jobs are experience goods; they only learn about the quality of the match by trying it out. However, as workers gain experience jobs become inspection goods. Market experience influences decisions by unemployed workers about which jobs to accept. This contrasts to the standard Jovanovic (1979) model where workers accept all or a constant fraction of job offers since their information when unemployed does not change with experience.
Moreover, the model with experience is able to account for the full decline in the job separation rate. In Jovanovic’s (1979) model, job durations are identically and independently distributed random variables and hence the turnover generated by the model is a renewal process. Each time the worker becomes unemployed she is in an identical position; job finding rates are constant. In the standard Jovanovic (1979) model only statistical sorting generates a decline in job separation rates.
In his model, older workers are more likely to have been in their job longer. While this sorting is able to qualitatively match the decrease in separations with age, it does not quantitatively account for the magnitude of the decline found in the data. Adding experience to the model generates a second force that causes separations to decline. Older workers are selective so their new jobs are more likely to be good. This additional feature predicts both a decline in job finding rates and quantitatively captures the full decline in job separation rates.
