The issue of regional wage differentials is relevant both for policy proposes and general public discussion. A sound knowledge of the distribution of wage inequalities and their causes is essential for defining policy measures for reducing spatial income inequalities. A range of empirical studies have analysed regional wage differentials for a number of countries (Blackaby and Manning, 1990; Blackaby and Murphy, 1995; Duranton and Monastiriotis, 2002; García and Molina, 2002). Typically, these studies are based on OLS estimates and the decomposition method devised by Blinder (1973) and Oaxaca (1973), which focuses on the analysis of wages differences at the mean of the conditional earnings distribution. This approach provides a reasonable description of wage distributions when they are unimodal, symmetric and have similar variances (Butcher and Dinardo, 2002). However, in general, these conditions may be not fulfilled. Therefore, wage differentials should be analysed along the entire wage distribution. This paper seeks to build on previous research in a number of different ways. Firstly, unlike most previous studies, we estimate regional wage equations by quantile regression in order to analyse the effect of covariates at several points on the wage distribution. Secondly, we apply the quantile-based decomposition method suggested by Machado and Mata (2005) and Melly (2005a, 2006) to decompose regional wage differentials at several points of the wage distribution. This method is of a semi-parametric nature, which allows for the estimation of significance tests and confidence intervals of wage decomposition effects (characteristics and returns). This marks a clear difference in relation to the non-parametric method suggested by Dinardo et al. (1996) and Butcher and Dinardo (2002) and applied by Motellón et al. (2011), which does not allow such significance tests to be performed. To the best of our knowledge, this is the first application of the methods proposed by Machado and Mata (2005) and Melly (2005a, 2006) in the context of regional wage differentials. We consider the case of Portugal, a small country with significant and quite stable regional wage differentials (Vieira et al., 2006; Pereira and Galego, 2011). Our findings reveal that coefficients estimates along the wage distribution for each region and between the various regions are not stable. Moreover, these findings confirm previous evidence as to the existence of significant regional wage differences between the Lisboa region and the other regions, and also reveal increasing differentials across the wage distribution. Finally, with regard to the regional wage decomposition, we conclude that both the part relating to differences in characteristics and the part relating to differences in returns to these characteristics are in general statistically significant and increase across the entire wage distribution.
