The causal impact of a business cycle shock on road crashes and its determinants – A synthetic control group analysis

Abstract

Introduction: Research suggests that recessions correlate with reductions in crash counts. However, knowledge is still scarce regarding the causality of this association, and the mechanisms through which economic shocks affect crash numbers are not well understood. We address these research gaps by applying an econometric methodology that has so far not been used for these research questions. Method: We use a quasi-natural experimental approach as our identification strategy. By exploiting the spatial heterogeneity of a shock, we define affected and less affected regions as treatment and control units. A synthetic control approach is applied to identify the causal impact of a shock on crash counts and explore the mechanisms contributing to this effect. As a case study, we use the 2008/09 financial crisis in Germany and exploit its high spatial variation. Results: We find that the crisis caused a significant crash rate reduction of 8% in the treated region. Only 1/4 of this reduction can be attributed to the decline in exposure. The remaining 3/4 are associated with the crisis-induced decrease in crash risk. Decomposing this effect shows that the crash rates in rural areas, of newly registered vehicles, of young adults, related to alcohol and speeding decline more than the overall crash rate. In contrast, crash rates of severe crashes, of heavy-goods vehicles, at night and on weekends are not the driving factors of the decrease in crash rates. Several robustness tests validate the results. Conclusions: Crash counts declined significantly due to the economic crisis. However, the magnitude of the influence is highly dependent on the crash characteristics. Practical applications: Understanding the trajectory of crash counts is crucial for implementing traffic safety measures and working towards vision zero. Our study shows that macroeconomic parameters are important potential confounding factors that should be considered in accident analysis.