Abstract

This paper analyses the personnel records of a Taiwanese auto dealer employing three distinct internal labor markets (ILMs), adding new evidence that builds upon recent empirical and theoretical works on ILMs. We show that the public learning model proposed by Farber and Gibbons (1996) is not supported in general by our data because the behaviors of empirical wage residuals covariance matrix contradict the martingale predictions derived from the their model. However, “public learning” may not be unrealistic once individual specific learning speed is introduced. Furthermore, we find that the positive effects of levels, on both salary and bonus equations, are smaller under a fixed effects model than under an OLS (combined) model. However, part of the wage variations is contributed by individual heterogeneity rather than the hierarchy itself. Evidence also shows that education plays an important role in the determination of levels.