Abstract
People have multi-dimensional abilities, which affect their schooling choices and how they sort into occupations throughout their careers. To capture this empirical reality, this paper presents a dynamic model of schooling, labor supply and occupation choices with two key features. First, abilities are treated as multi-dimensional (cognitive, socio-emotional, manual and routine), and they dynamically generate later life outcomes, including schooling choices, occupation choices and wages. Second, occupations are modeled as bundles of discrete tasks that correspond to the four dimensions of abilities. This approach explicitly accounts for dynamic sorting behavior and affords an interpretation of how ability, schooling and human capital accumulated on the job affect wages based on the task content of occupations. It accommodates the inclusion of 16 occupations, a large improvement over previous attempts to estimate such dynamic labor supply models, while retaining computational feasibility. Moreover, when an individual changes occupation, it allows human capital accumulated on the job to be partially transferable based on the task content of occupations.
The model is estimated with the Maximum Simulated Likelihood method using data from the 1979 National Longitudinal Survey of Youth and it generates three main sets of results. First, wage returns to education differ by both levels (high school vs. college) and task content of occupations. For example, college education yields a large wage premium for cognitive tasks and a small premium for interpersonal tasks but no premium for manual or routine tasks. Second, task content and pairwise task complementarity are important in explaining wage dispersion. In particular, the cognitive-interpersonal tasks pair and the manual-routine tasks pair are complementary, and performing either pair in the same occupation brings a large additional wage premium. Third, aside from a direct effect on wage income, abilities also directly affect current period schooling utility, and hence workers' terminal educational attainment. While higher abilities to perform cognitive and routine tasks increase the utility from schooling, socio-emotional ability has no direct effect on schooling utility, and higher manual ability decreases schooling utility.
Simulation results reveal an important weakness of the current model. The model under-predicts enrollment rates for school-attending ages, and hence the mean wage levels associated with cognitive and interpersonal task intensive occupations and the number of workers who later take these occupations. I propose specific model changes to resolve this issue. For future work, I will estimate the revised model and use updated model estimates to conduct counterfactual simulations to evaluate various human capital development policies proposed to promote welfare (e.g., universal 4-year college subsidies).
