Displaced workers who are forced to switch occupations are found to experience larger earning losses than those who don't (Kambourov and Manovskii (2009)). Occupational heterogeneity may imply that these losses are not necessarily the same across occupations. In addition, if human capital is more important for workers who are better suited for a particular occupation, then any misallocation of workers across occupations will be more pronounced in occupations where human capital is particularly important. For instance, if human capital interacts with how well a worker matches in each occupation and human capital is more important for cooks than truck drivers, then having the wrong worker work as a cook is more costly than having the wrong worker work as a truck driver. <P> The goal of this project is to investigate the importance of occupation-specific human capital across different occupations. A recent literature has argued that occupation-specific human capital is a key component in wage determination. That same literature however typically assumes that the rate of human capital accumulation does not differ across occupations. <P> We introduce a model where workers, while employed in an occupation accumulate specific human capital and also learn about the underlying quality of their occupational match. Our setup allows for occupational heterogeneity: the rates of human capital accumulation, as well as the speed of learning are allowed to differ across occupations. Human capital can also interact with match quality. In addition, workers accumulate general human capital. When making an occupational choice, the worker takes into account his current wage, the occupation-specific human capital that will be accumulated, as well as the information obtained regarding the underlying quality of the occupational match. His state space includes his beliefs about his match quality in each occupation, as well as his human capital level at each occupation. Therefore for more than a handful occupations, the problem quickly becomes very hard to solve computationally. Through the use of Gittins indices however, the problem becomes tractable, even in the case of many occupations. <P> As in the data, our setup predicts that wages are an increasing function of tenure. In particular, wages increase for two reasons: first, workers accumulate human capital and second, selection implies that only the workers who are best matched remain in an occupation. Occupational switching however, which is driven by workers learning that they're not well-matched in an occupation, is a decreasing function of tenure. The rate of decline of occupational switching with tenure identifies the speed at which workers learn about their match quality. Given that estimate, we can separately identify the rate of human capital accumulation in each occupation, by looking at the part the wage increase that is not explain by selection. In addition, we also look at the increase with tenure of the within-occupation cross-sectional wage variance. <P> Our setup follows a large empirical literature that attempts to distinguish between human capital and selection, starting with Abraham and Farber (1987), Altonji and Shakotko (1987), Topel (1990) and later Altonji and Williams (2005) who examine the importance of firm tenure on wages. More recently Kambourov and Manovskii (2009), Gathmann and Schönberg (2010) and Pavan (2011) consider the importance of occupation/industry choice in wage formation. Finally, understanding the importance of human capital and learning in each occupation also contributes to the displacement loss literature (Jacobson et al. (1993)).
