Human capital is widely acknowledged as a key factor for economic performance at both the micro and macro level. Despite the fact that a large fraction of human capital accumulation takes place after the entry into the labor market, most of the existing literature that investigates the returns to investment in human capital has focused on education, due to measurement problems and data availability. Relatively little evidence is available, instead, on the accumulation of human capital through the lifelong training of workers and, more specifically, on the effects of training on productivity.
A number of studies have tried to fill this gap by analysing the impact of training on productivity using firm-level data. However, this literature does not provide a consistent picture, as the lack of longitudinal data has generally made it difficult to control for unobserved heterogeneity and endogeneity of training (e.g. Bartel, 1994, Bishop, 1994, Black and Lynch, 1996, Barrett and O’Connell, 2001). Some recent studies have tackled this problem by focusing on panel data at industry-level (e.g. Dearden et al., 2006, Conti, 2005). This approach, however, does not allow to estimate the private returns to training, as analyses based on industry-level data also capture spillover effects between firms. There exists a recent literature that investigates the effects of training on productivity using firm-level panel data, but it is generally hampered either by the specificity of the sample (e.g. Almeida and Carneiro, 2006), or by the limited number of observations in the sectional dimension (Ballot et al., 2006; Zwick, 2005, 2006) or in the time dimension (Black and Lynch, 2001).