Association for Public Policy Analysis & Management

The job polarization literature classifies occupational tasks very differently and therefore it has been very difficult to compare results across studies. One of the criticisms of this literature points out that the classification of occupations into a specific task group is not always transparent and difficult to replicate. Using US data, leading economists in this field, such as Autor, Levy and Murnane (2003), proposed a task-based categorization into manual, routine and abstract occupations and actively recommend researchers its use as the standard occupational task-based categorization. However, the application of this approach for data of European countries (or other non-American contexts) has been problematic due to differences in the American (SOC) and European (ISCO) occupational classification systems.

This article proposes an objective and reproducible approach towards the task-based categorizion of ISCO occupations by exploiting the recent availability of official crosswalks between the American (US) and European (EU) occupational classification systems.

Applying this method to the Swiss Labour Force Survey (SLFS), I study job polarization in Switzerland and compare the results obtained from two different task-classification procedures: the classification proposed by Goos, Manning and Salomons (2014) and the one using official crosswalks developed here. As both classifications methods assign few occupations to different task categories, the resulting polarization patterns differ. Results illustrate the dependence of task-based polarization from the underlying task classification approach. Also, given the uncertainty around task categorizations, it seems difficult to attribute job polarization to the routinization hypothesis. Heterogeneous occupational groups (clerks and salesmen) are found particularly prone to inconsistent task-based classification.

Finally, this study points out that maps or crosswalks between US and EU occupational classification systems provide an objective and reproducible approach for task classification of EU occupations; however they are available only up to the ISCO 2-digit aggregation level. This study highlights that this level of disaggregation is too coarse for accurate task-based classification, and therefore it is needed to increase the level of granularity of occupational information for occupational classifications.