Moral hazard among the employed: Evidence from regression discontinuity

We exploit policy discontinuities in Poland's unemployment insurance to examine the causal effect of changes to both benefit durations and levels. Using a regression discontinuity approach, we uncover three findings: (1) Higher benefit levels distort employment more than benefit extensions. (2) Benefit durations and levels interact: Longer durations substantially increase the distortionary effect of more generous payments. (3) Higher payments increase the transition of employed workers into unemployment. We develop a model of optimal unemployment insurance that accounts for moral hazard among both employed and unemployed workers. Notably, for level increases, distortionary costs are larger among the employed than unemployed.