Essays on the economics and econometrics of urban crime and house price prediction

Criminal activity is inherently a local phenomenon. Anyone who has lived in a city knows that crime rates vary considerably across neighborhoods---occasionally even from block to block. City-wide crime rates do not adequately measure this variation and, consequently, cannot be used to test many important theories in the crime literature. This research represents an attempt to understand some of the causes and consequences of criminal activity using neighborhoods (broadly defined) as the geographic unit of aggregation. Our decision to use neighborhood geography will hopefully shed light on several key details of urban crime that broad-based macro studies have missed or ignored altogether.; Many studies have noted serial correlation in housing prices and have built prediction models for national and MSA real estate prices. However, no research has been conducted on predicting neighborhood house prices, despite the fact that most residential moves take place within a metropolitian area. In Chapter 1 we explore the possibility of predicting house prices for city neighborhoods in New York and Chicago, which should be of great importance to intra-MSA movers. In particular, neighborhood crime rates have predictive power for intra-MSA prices, due to positive autocorrelation in crime rates and the strong negative correlation between crime and house prices. Crime rates provide a barometer for gentrification (and urban decay) in city neighborhoods, and neighborhood house prices track crime trends over suitably long horizons. We successfully predict prices out-of-sample for one year forecast horizons in both cities, with RMSE up to seven percent smaller than naive AR models.; High frequncy data is an effective tool for solving the simultaneity problem between crime and various deterrence measures. In Chapter 2 we use quarterly crime and arrest panel data from Los Angeles to test the "Broken Windows hypothesis" in a manner similar to Corman and Mocan (2005). Misdemeanor arrests, which serve as a proxy for broken windows policing, are shown to have significant negative impacts on robbery, assault, burglary, and auto theft for the city of Los Angeles during the period 1983 to 2005. Own arrest rates, unemployment rates, and police manpower are also shown to have significant impacts on felony crime. These findings are broadly consistent with Corman and Mocan (2005) and the literature on crime detterance. However, first-differenced crime and arrest variables display quadratic time trends which inflate the magnitude of misdemeanor arrest coefficient estimates. When time trends are removed from our data the support for broken windows policing diminishes. Some of the evidence for the Broken Windows hypothesis in the literature appears to rest on city-wide trends in crime and misdemeanor arrests, and not on relationships between stationary time series.; Social Disorganzation theory predicts high crime rates in neighborhoods with high levels of renters, single parent families, and ethnic heterogeneity. Routine Activities theory predicts high crime rates in neighborhoods with motivated criminals, desirable victims, and few detterence measures. In Chapter 3, we build a spatial model of crime rates using over 750 New York City neighborhoods and find a number of empirical regularities that are consistent with these theories. 15 different felony and misdemeanor crimes are analyzed, including rarely examined crimes such as purse snatching and illegal weapons possession. Crime rates in New York neighborhoods are significantly associated with standard demographic variables, such as race and income, along with physical infrastructure, such as subway stops, boarding houses, and drug rehabilitation clinics. Surprisingly, crime rates are negatively associated with population density, and almost entirely unaffected by the presence of public housing. These empirical regularities are robust to several estimation methods and robustness checks.; Ideally this study w...