Objectives: Violent behavior is influenced by individual and societal characteristics, but the role of environmental factors is less understood. Our aims were to use national-level data to identify the association between criminal behavior and short-term temperature conditions, including the departure of daily temperatures from normal conditions. Methods: We conducted a multi-stage hierarchical time-series model across 436 U.S. counties and 14-years representing 100.4 million people to investigate the association between daily mean temperature and daily mean temperatures departing from normal conditions with violent and non-violent crime counts. First-stage comparisons were made within counties to control for population and geographic heterogeneities, while a second stage combined estimates. We evaluated differences in risk based on county sociodemographic characteristics and estimated non-linear exposure-response relationships. Results: We observed a total of 9.0 million violent crimes and 20.9 million non-violent property crimes between 2000 through 2013. We estimated that each 10 °C increase in daily temperature or daily departure from long-term normal temperatures were associated with 11.92% (95% PI: 11.57, 12.27) and 10.37% (95% PI: 10.05, 10.69) increase in the risk of violent crime, respectively. Similar, but lower in magnitude trends, were observed for property crime risks. We found that crime risk plateaus and decreases at high daily temperatures, but for temperatures departing from normal, the association with crime increased linearly. Seasonal variations showed that anomalously warm temperatures days during cool months had the greatest risk. Conclusions: Our study revealed an association between higher temperatures and high departure from normal temperatures with both violent and non-violent crime risk, regardless of community-type. However, our findings on seasonal and daily trends suggest that daily mean temperature may impact crime by affecting routine activities and behavior, as opposed to a temperature-aggression relationship. These results may advance public response and planning to prevent violent behavior.
Bibliographical noteFunding Information:
We thank Ander Wilson (Colorado State University) for his suggestions and advice on the statistical model.
- Temperature departures