The Value of Forecast Improvements: Evidence from Advisory Lead Times and Vehicle Crashes

Scientific and technological advances are resulting in improved forecasts of risk, but do better forecasts result in better risk management? I investigate to what extent the improvements in lead time of winter weather advisories affect the frequency of motor vehicle crashes. I construct a data set of winter weather advisories, weather monitor readings, and vehicle crashes at the county-date level in 11 states in the US during 2006-2018. Using within county variation in lead time, I show that receiving winter advisories earlier reduces crash risk significantly. I also examine two potential mechanisms that might lead to these effects. First, using the mobile phone location data from SafeGraph, I show that longer lead times result in fewer visits by people to places outside their homes. Second, using snow plow truck location data, I show that road crews perform a greater level of winter maintenance activities when advisories arrive with longer lead time. Overall, this study provides evidence that improvements in forecast lead times result in meaningful economic benefits to society, and these benefits come from both the individual and institutional response to longer lead times.