The effects of wage compression on general and firm-specific training

Abstract

The issue of who finances on-the-job training has important implications for labour markets. It is persistently difficult to test empirically whether it is the employer or the employee who carries the costs of general training. This article presents a simple method that compares the effect of wage inequality on duration of general training and firm-specific training. The result is consistent with the proposition that it is the worker who bears a greater part of the costs associated with general training than in the case of firm-specific training.

Acknowledgements

I thank Thomas Andrén, Dominique Anxo, Anders Björklund, Lennart Flood, Matthew Lindquist, Erik Mellander and Per Skedinger for helpful comments and suggestions. I have also received comments from seminar participants at the Department of Economics in Uppsala University, Göteborg University, Stockholm University, and at The Research Institute of Industrial Economics in Stockholm. Financial support from The Institute for Labour Market Policy Evaluation (IFAU) is gratefully acknowledged. Data used in this article is available at the IFAU.

Notes

¹ The sample size is around 13 000 individuals between 16 and 64 years of age who are registered to be living in Sweden and participating in the labour force. The response rate for these individuals was about 80%, resulting in a sample of about 10 000 individuals. The number of individuals who had a job at the time of the interview was around 7000.

² I tested if the lagged wage inequality was exogenous to training duration in (1) by assuming that W _it was an endogenous explanatory variable in (1), and used W _{it −1} and W _{it +2} as instrumental variables for W _it in a 2SLS estimation. The Sargan test for overidentifying restrictions could not reject the hypothesis that W _{it −1} was a valid instrumental variable for W _it , provided that W _{t +2} also was an exogenous variable. W _{t +2} is the latest observations on wage inequalities (year 2003). An alternative strategy would be to use W _{t −2} as the additional instrumental variable. However, compatible occupational data were not available for the year 1999.