Worker specialisation and the consequences of occupational decline

According to a standard Roy model, when the demand for labour in an occupation declines, the utility loss of an incumbent worker is determined by the initial difference between the utility associated with his current occupation and his best outside option in the set of occupations unaffected by the shock. We refer to this difference as his degree of occupational specialisation. Generalist workers with good outside options will leave quickly and be better off, while highly specialised workers willingly remain and tolerate the full effect of the demand shift through lower wages.

This project shows that under certain distributional assumptions on occupation utility, the expected value of the above utility difference can be inferred from the ex ante probability of a worker being observed in the set of unaffected outside occupations, given his traits. We name this function the occupation specialisation index (OSI). To construct it empirically, we train an artificial neural network to predict occupational choice probabilities using data from Sweden on detailed worker characteristics, including multidimensional abilities, age, education, region of residence, and industry-specific work experience.

The index is then used to shed some light on the historical decline in employment in routine-intensive occupations. I compare the difference in long-run occupation switching behaviour and career outcomes between workers initially employed in routine and non-routine occupations across the specialisation distribution by means of a difference-in-differences-styled specification.

A better understanding of the historical shift in the demand for routine work and its consequences for incumbent workers is important in its own right. But the results can also substantiate the general usefulness of the specialisation index in determining which workers are susceptible to negative demand shifts. The index is solely based on current information. Therefore, it could be used to characterise workers employed in occupations today that are believed to decline or even disappear in the future by susceptibility to such shifts.