We identify vulnerable groups through the examination of their employment status in the face of the initial coronavirus disease 2019 (COVID-19) shock through the application of tree-based ensemble machine learning algorithms on a sample of individuals over 50 years old. The present study elaborates on the findings through various interpretable machine learning techniques, namely Shapley values, individual conditional expectations, partial dependences, and variable importance scores. The structure of the data obtained from the Survey of Health, Aging and Retirement in Europe (SHARE) dataset enables us to specifically observe the before versus the after effects of the pandemic shock on individual job status in spatial labor markets. We identify small but distinct subgroups that may require particular policy interventions. We find that the persons in these groups are prone to pandemic-related job loss owing to different sets of individual-level factors such as employment type and sector, age, education, and prepandemic health status in addition to location-specific factors such as drops in mobility and stringency policies affecting particular regions or countries.
|Number of pages||26|
|Journal||Regional Science Policy and Practice|
|Publication status||Published - 28 Apr 2023|
- machine learning
- YOUTH UNEMPLOYMENT
- BIG DATA