Parametric and semiparametric model representations of the conditional mean of wages: do they make a difference against nonparametric ones?

This study is concerned with the validation of parametric functional representations of the conditional mean of wages using Philippine data. This is done in the light of studies confirming relative weaknesses associated with parametric approaches to the modelling of human capital earnings functions (Zheng, 2000; Heckman <italic>et al</italic>., 2003; Lemieux, 2003; Miles and Mora, 2003). Extending the approaches in Lemieux (2003), Zheng (2000) and Miles and Mora (2003), we conduct consistent specification tests not only on parametric functional forms but also on the semiparametric partially linear model to verify the effect of modelling choice on the schooling-earnings as well as the experience-earnings relationships. Test results indicate that parametric models may still be valid representations of the wage function. In some instances, the semiparametric partially linear wage function holds promise as a modelling alternative to parametric models.