Essays on the Environmental Determinants of Crime.

Abstract

This dissertation examines the effect of several environmental variables on criminal activity. Chapters 1 and 3 are the first academic works to my knowledge that examine the effect of temperature expectations and errors in those expectations (i.e. “forecast errors”) on crime. Chapter 1 develops a theoretical model in which temperature expectations determine the distribution of people inside and outside of residences, and thus the level of opportunity for crime. I then examine the impact of temperature expectations and forecast errors on crime using a panel of U.S. cities. My results suggest that crime on day t increases with day t expected maximum temperature, and falls when the future is expected to be warmer. Furthermore, violent crime responds strongly and positively to forecast errors, but property theft is only weakly affected by the same variable. Chapter 3 finds many of the same results in the specific context of Chicago, IL, and also examines the impact of temperature expectations and forecast errors on daily public transit ridership. As the model from Chapter 1 predicts, I find that ridership increases with current day expected temperature, and falls when the future is expected to be warmer. Taken together, these two chapters convincingly demonstrate a link between temperature expectations, time use, and criminal activity.

Chapter 2 studies the relationship between the number of people in a physical space and the level of criminal activity. Using ridership data from the city of Chicago to proxy for the number of people inside and outside of residences, I find that higher ridership is associated with more violent crime and property theft outside of residences, though in almost every case a 1% increase in ridership leads to a less than 1% increase in crime. In other words, larger crowds have more crime in an absolute sense, but are safer on a per capita basis. I also find that property theft inside of residences increases with ridership, but violent crime in the same setting falls. In the remainder of the paper, I consider several sub-samples of interest, and consider an instrumental variables approach to estimation.