Labor Turnover Bias in Estimating Wages

Cross-sectional earnings function analyses typically do not recognise the potential for misrepresentation arising from a relationship between unobserved ability and the probability of labor turnover. Our point is that if individuals staying with the firm for relatively long periods of time are more, or less, able on average than recent hires, the coefficient estimated on tenure is necessarily biased. This is an important issue for wage determination modelling given the relevance of tenure to wages. In this note we examine the theoretical basis of our claim, and propose and implement a solution to the problem with a novel use of instrumental variables on a large sample of workers employed in the Australian government.