Abstract:
Dingel and Neiman (2020) report that 37% of jobs in the US can be performed at home. Their classification algorithm applies a threshold to 15 survey questions on a 1-5 scale. These thresholds are not justified. As a robustness check, I create more lenient (liberal) and more restrictive (conservative) versions of the algorithm by adding or subtracting 0.5 from the threshold. The liberal version counts more occupations as remote, while the conservative version counts fewer. The original estimated share of occupations that can be remote is 37%. The liberal estimate is 59% and the conservative estimate is 25%. The classification is sensitive to small changes in the thresholds.