Within-groups wage inequality and schooling: further evidence for Portugal

Abstract

This article provides further evidence on the positive impact of schooling on within-groups wage dispersion in Portugal, using data on male workers from the 2001 wave of the European Community Household Panel. The issue of schooling endogeneity is taken into account by using the latest available instrumental-variable technique for quantile regression, i.e. the control-function estimator due to Lee (2007). The findings are compared with earlier results based on different techniques, i.e. the instrumental-variable estimator due to Arias et al. (2001) and the standard exogeneity-based estimator due to Koenker and Bassett (1978).

Acknowledgements

I gratefully acknowledge financial support from the European Commission (EDWIN Project, HPSE-CT-2002-00108). For valuable comments, I would like to thank Santiago Budría, Joop Hartog, Pedro Telhado Pereira, Robert Wright and the participants at the 22nd National Conference of the Italian Association of Labour Economists. The usual disclaimer applies.

Notes

¹ This section partly borrows from Andini (2008).

² Min and Kim (2004) provide a Monte Carlo comparison between the KB parametric estimator and the nonparametric approach due to Yu and Jones (1998). The results indicate that the nonparametric quantile-regression method is more appropriate, particularly when the underlying model is nonlinear or the error term follows a nonnormal distribution.

³ Consider the regression model Y = βX + ϵ. The estimator of KB is consistent at each quantile θ of the conditional distribution of Y if the orthogonality condition holds at each quantile, i.e. ∀θ, while the consistency of the OLS estimator requires that the conditional mean of regression residuals is null, i.e. E(ϵ|X) = 0.

⁴ Jobs in some industries may require more years of schooling than jobs in other industries.

⁵ Pereira and Martins (2004) properly argue that in order ‘to obtain the full impact of education on wages, one should be careful not to include in the wage equation covariates whose value can depend on education. In the extreme case, one should only regress the ln(wage) in education’ (p. 526). See also Andini (2007).

⁶ Oliveira and Santos (2005) find evidence of heterogeneity in efficiency among secondary schools in Portugal. To the best of our knowledge, a more comprehensive study on school quality in Portugal is not available.