Skills, Tasks, and Occupational Choice∗

This paper seeks to quantify two important sources of workers’ wage growth skill accumulation and occupational human capital by constructing and estimating a structural occupational choice model. Accumulated skills are assumed to be more transferable between jobs that utilize similar tasks. Unlike previous work, I allow task usage to vary not only by occupation, but also by hierarchical level within an occupation. Furthermore, I separately estimate the impacts of occupation-specific and occupationlevel specific human capital, where occupation-specific human capital can be transferred across levels within an occupation. This allows workers to adjust their task usage vectors by adjusting their hierarchical level (i.e. through promotion or demotion), while maintaining their accumulated occupation-specific human capital. Taking labor market histories from the German Socio-Economic Panel, and task usage data from the German Qualification and Career Survey, I estimate my model using Indirect Inference. Results show that occupation-specific human capital is a significant component of wage growth, which confirms that workers can transfer some occupational human capital across levels within an occupation. Counterfactual simulations show that eliminating skill change reduces the overall mean wage level by 6.4%, while eliminating occupational human capital accumulation reduces mean wage level by 39%. These results highlight the importance of occupational human capital, even when accounting for task-specific human capital. ∗I would like to thank Audra Bowlus, Chris Robinson, Jed DeVaro, Todd Stinebrickner and Salvador Navarro for helpful comments and guidance. I would also like to thank participants at the Economics PhD Workshop at the University of Guelph and the 2012 Atlantic Canada Economics Association Conference. I am grateful for the data used in this survey provided by the German Institute for Economics Research and the Federal Institute for Vocational Education and Training. Computational resources from the Shared Hierarchical Academic Research Computing Network (SHARCNET) are also greatly appreciated. All errors are my own. Contact the author: [email protected]