Employer learning and the returns to schooling

Abstract

Theoretical models of employer learning suggest that an employee's education is an important signal to the employer initially. Over time, however, the returns to schooling should decrease with labor market experience and increase with initially unobserved ability, since the employer gradually obtains better information on the productivity of an employee. Replicating US studies using a large German panel data set (GSOEP), we find no evidence for the employer learning hypothesis. Differentiating blue-collar and white-collar workers and estimating quantile regressions, however, leads to the conclusion that employer learning takes place for blue-collar workers at the lower end of the wage distribution.

Introduction

The dynamic relationship between the returns to schooling and labor market experience has often been used to test the signaling model against the human capital model. This test relies on the hypothesis that if education acts as a signal, the partial effect of schooling on earnings will decline with increasing labor market experience of an individual, since employers gradually obtain better information on the true productivity of a worker. In recent articles, Farber and Gibbons (1996) and Altonji and Pierret, 1998, Altonji and Pierret, 1999 investigate the dynamic relationship between the returns to schooling and labor market experience in a model of employer learning. According to their models, the return to formal education, as measured by years of education, is independent or even decreases with labor market experience, whereas the return to natural ability, as measured by a standardized aptitude test, parents' education or the wages of siblings, increases with labor market experience when estimating a standard Mincer-type wage equation. Formal education provides some information to the employer as to the general productivity of an employee. However, the more an employer learns first-hand about the employee's true productivity on the job, the faster the measured effect of schooling on wages should decline and the measured effect of natural, initially unobserved, ability should increase with labor market experience.

This paper tests the employer learning hypothesis using a large German panel data set, the German Socio-Economic Panel (GSOEP), for the period from 1984 to 1996. Based on a sample of West German prime-aged males, we do not find evidence for employer learning in Germany. In fact, both the relationship between the reward to education and experience and the relationship between natural ability and experience are positive. To investigate the robustness of our results, we perform several types of sensitivity analysis. In a first step, we analyze to what extent our results are influenced by the assumption that information on the productivity of a worker is public. We find that experience effects on the returns to schooling and unobserved ability are mainly driven by tenure at the current job and not through previous labor market experience, indicating that information on a worker's productivity is private, rather than public.

In a second step, we investigate to what extent the degree of employer learning differs between different types of workers and jobs, as suggested by economic theory and existing empirical evidence. For learning to take place, the employer must be able to observe the productivity of workers. The performance of blue-collar workers should be easier to observe compared to white-collar workers, since the former in general perform tasks where they use relatively more physical work to produce tangible goods, whereas the latter perform tasks with “intellectual” inputs and largely intangible outputs. Furthermore, the theory and empirical evidence on firms' hiring strategy suggests that employers invest more in the screening of applicants for positions which require a higher schooling level and more training. Hence, at the time of hiring, employers should have more information and therefore lower learning rates on high-skilled and high-paid workers than on low-skilled and low-paid workers. Differentiating between blue-collar and white-collar workers and estimating quantile regressions, we find that employer learning in Germany takes place only for blue-collar at the lower end of the earnings distribution.

The paper proceeds as follows. In Section 2, we present the basic intuition behind the idea of employer learning and test the hypotheses using German data. Section 3 provides a sensitivity analysis to investigate the robustness of our results. Section 4 concludes.