The worker-job surplus is of great importance in any matching model of the labor market. It is a key determinant of workers' employment transitions and their wages, and it is crucial for measuring the gains from sorting. In this paper, we develop a theory-founded test to determine which of the observable worker and job characteristics enter the worker-job surplus in the data, where we exploit workers' mobility choices. Our test also indicates which worker and job attributes are relevant for sorting on the labor market. We implement the test on US data using the Survey of Income and Program Participation and the O*NET. The results suggest that a relatively sparse model underlies the data. On the job side, a cognitive and an interpersonal skill requirement impact the surplus along with the (dis)amenity of work duration as well as the workplace size. On the worker side, we find that most of the relevant characteristics are symmetric to the selected job requirements. We reject that there exists a single index reprsentation of these relevant multi-dimensional worker and job attributes.Finally, we use our test results to shed light on multi-dimensional sorting along the relevant dimensions in the data, both in the cross-section and over time.
Joint work with Fabien Postel-Vinay.