This study estimates the causal effects of traffic crashes on subsequent driving behaviour, focusing on total annual kilometres driven, the proportion of kilometres driven above the posted speed limit, and the share of driving in urban areas and at night. The analysis relies on weekly data provided by an insurtech company, collected over a two-year period.

We implement a difference-in-differences (DiD) quasi-experimental design that evaluates changes in driving outcomes in the weeks following a crash. The treatment group consists of individuals who experienced an at-fault crash during the overall observation period and for whom complete information is available for the design’s observational window, which spans over 25 weeks. The treatment period corresponds to the week in which the crash occurs (week 13), preceded by a 12-week pre-treatment period and followed by a 12-week post-treatment period. The control group comprises individuals who did not experience a crash during the observation period. The causal effects are estimated using a two-way fixed effects (TWFE) panel model that incorporates both additive and multiplicative control covariates. A semiparametric approach based on inverse probability weighted estimation is employed to assess the robustness of the results.

We conclude that being at fault in a crash has a statistically significant negative effect on the four driving outcomes. The largest negative effects one week after the crash are observed in the share of urban and night-time driving. For all four outcomes, these effects gradually diminish and approach zero after twelve weeks. The decline is faster for the share of urban driving. Different driver and vehicle characteristics are associated with distinct behaviour. The effects for young women and men differ with respect to speed outcomes: while the effect for women is more negative one week after the crash, it decreases more rapidly than the corresponding effect for men.