Nightless City: Impacts of Politicians' Questions on Overtime Work of Bureaucrats

We quantify the impact of unexpectedly assigned tasks on overtime work in the context of Japanese government officials. Data on overtime work are typically not reliable. We overcome this problem by using mobile phone location data, which enables us to precisely measure the nighttime population in the government office district in Tokyo at an hourly frequency. Exploiting the exogenous nature of task arrivals, we estimate dynamic responses of overtime work. We find that, in response to an unexpected task, overtime work initially decreases and then increases persistently. Institutional changes, such as relaxing time constraints and adopting a paperless system, were intended to mitigate overtime work. However, our findings suggest that these changes had limited success in reducing overtime or improving the quality of work, as evidenced by our survey experiment. We provide a simple model of optimal work allocation and show that distortion in intertemporal task allocation can account for the observed responses.