Intrafirm human capital externalities and selection bias

Abstract

This article investigates the existence of intrafirm human capital externalities in the Spanish economy. In doing so, it focusses on two estimation issues. First, a semiparametric methodology is used to estimate externalities in a more flexible way that allows to affect all parameters in the equation. Second, two approaches to control for selection bias are designed for use in cross-section data. Our results show a higher externality effect on wages when using the semi parametric approach as compared to the standard estimation approach. Additionally, the estimations even when controlling for selection bias give evidence of externalities.

Acknowledgements

We gratefully acknowledge the funding provided by the Spanish Ministry of Education under project SEJ2006-14849. We would like to thank Ezequiel Uriel for useful comments. All remaining errors are ours.

Notes

¹ However other explanations may depend on externalities of a pecuniary nature as in Acemoglu (1996).

² Rauch (1993) is a seminal paper dealing with this approach.

³ See Moretti (2004) for a detailed review.

⁴ Ciccone and Peri (2006) develop what they call ‘constant composition approach’ to control for imperfect substitutability. Evidence of externalities disappears when applying this approach.

⁵ See Abowd et al. (1999).

⁶ National, sector/province or firm level.

⁷ The number in parentheses is the number of years required to obtain the degree.

⁸ See Bayard and Troske (1999), Barth (2002) and Battu et al. (2003). For the Spanish case, Alcalá and Hernández (2006) is a recent reference.

⁹ The Education Act of 1970 introduced a comprehensive system with 8 years of compulsory education. The education of most of the workers in our sample took place under the system established by this act.

¹⁰ Different definitions for the reference group are possible. In any case, all the definitions lead to positive externalities. So, for example, when the reference group is made up of all the establishments whose maximum number of years of schooling of their employees is 8, following a similar methodology, the average human capital externalities are evaluated at 0.24. This is a consequence derived from the fact that the definition of the reference group is more constraining.

¹¹ The results obtained by using this methodology are provided in the Appendix.