The paper investigates the determinants of the Ordinary Least Squares (OLS) bias of the wage return to graduate education for high-school workers in Portugal.
The study uses matched employer-employee data for Portugal, over the 2002–2012 period, to estimate a wage-schooling model that controls not only for individual observed characteristics, firm observed characteristics and year fixed effects, but also for three high-dimensional vectors of fixed effects – one for employees, one for employers and one for job titles.
The main results are the following. First, disregarding individual fixed effects is highly problematic, accounting for 48.5% of the OLS bias. Second, disregarding firm fixed effects is also problematic, accounting for 12.3% of the OLS bias.
The implication for the studies in the labor-supply literature that estimate, by means of instrumental variables, the wage returns to in-school work or to on-the-job schooling is that an instrument dealing with employee’s unobserved ability only may fail to meet the exclusion restriction.
Take the typical instrument based on a policy reform that changes the compulsory schooling level in the population. This instrument may well be argued to be correlated with the education of the employee and uncorrelated with the unobserved ability of the employee, but unfortunately it cannot be seen as orthogonal to the unobserved ability of the employer because of its correlation with the (unobserved) education of the manager. This is a simple corollary of the fact that the employee and the manager belong, in general, to the same population.
Individuals invest a considerable amount of resources in education, which is seen to have positive effects on several dimensions of individual life. Yet, the estimation of these effects is still surrounded by technical difficulties.
To the best of the authors’ knowledge, this is the first study that uses the Gelbach decomposition to investigate the determinants of the OLS bias of the wage return to graduate education for high-school workers.
