The largest city in Maryland, Baltimore is infamous for its high crime rates, with levels of property crime and violent crime that by far exceed the national average (Fenton, 2014; FBI UCR Data, 2008-2014). Baltimore city has also been heavily impacted by the foreclosure crisis; it is the second hardest hit jurisdiction in Maryland. In 2009, most of Baltimore city was designated a foreclosure “hot spot” by the Department of Housing and Community Development. Moreover, even as foreclosure rates in most other parts of the country began to subside, foreclosures in Baltimore nearly tripled between 2012 and 2013. This increase represented the biggest annual gain among the twenty largest U.S. metropolitan areas, and Baltimore was one of only five cities that showed an increase in properties receiving default, auction and repossession notices, bucking a decline nationwide. (Perlberg & Dexheimer, 2013)
I examine point-specific foreclosure and victim-based crime data, with observations from 2008 to 2013. For the former, I utilize BNIA-JFI address-specific data for homes that entered foreclosure during the observation period. For the latter, I have compiled a comprehensive list of crimes reported in Baltimore City during that time period, combining information from the Baltimore Police Department’s Victim Based Crime Data (2011-2013) with related extracts from SpotCrime (2008-2010). Each entry is categorized as violent crime or property crime. Violent crimes include homicide, rape, robbery,and assault. Property and public order crimes include burglary, larceny, and motor vehicle theft. Total crime encompasses all reported offenses. Because property crimes occur much more frequently than violent crimes, total crime counts are highly correlated with property crime counts. Additionally, due to data irregularities, not all records could be matched to an address and subsequently geocoded; as result, approximately eight percent of reported crimes were excluded from this study.
I apply a fixed-effect spatial lag model to point-specific panel data on foreclosures, property crime, and violent crime in Baltimore between 2008 and 2013. I utilize a fishnet of uniform 100 x 100 meter grid cells in order to investigate the proximity effects of foreclosure data on criminal activity. The null hypothesis is that foreclosures do not impact the spatial distribution of either property crime or violent crime, while the alternative hypothesis is that higher levels of foreclosures lead to higher levels of crime in the surrounding area. To test the potential distance decay of the effect of foreclosures on crime, I overlay my map of yearly foreclosure and criminal activity in Baltimore with a fishnet grid made up of uniform 100 x 100 meter cells. I compare the number of crimes (in each category--property and violent) that occur within a given grid cell with the number of foreclosures that occur within the same cell, and in surrounding cells (see Appendix A). Within this framework, the annual number of crimes that occur in each grid cell is my dependent variable.
My results indicate a significant and positive spatial relationship between foreclosures and crime. Foreclosure and crime show significant spatial proximity. The magnitude of the relationship is greatest within the 100 x 100 meter grid cell. Specifically, one foreclosure within the 100 x 100 meter grid cell increases the incident rate of violent crimes and property crimes by 1.15 percent and 1.83 percent, respectively. That rate drops to about a 0.57 percentage increase in violent crimes and a 0.66 percentage increase in property crimes for each foreclosure within a kilometer.
The distance measures are jointly significant (χ2at p<0.05) in all models. The distance of foreclosures to property crimes is independently significant (p<0.05) to at least 500 meters, and marginally significant between 500 meters and one kilometer. For violent crimes, however, the distance of foreclosures ceases to be independently significant beyond 200 meters. The weakness of estimates associated with violent crimes may be to some extent a function of the relative infrequency of such crimes. Nevertheless, the empirical pattern of a distance decay effect remains apparent across all models.
Excluding cells that contain only water, on average there are about six crimes per grid cell during the observation period. The number of crimes per grid cell ranges from zero to 278, while foreclosure listings per cell range from zero to sixteen. Both crime and foreclosures tend to cluster, exhibiting a contagion-like pattern across urban space. Estimates suggest that the influence of these clusters is significant in its own right. For grid cells that contain three or more foreclosures in a given year, the associated spike in nearby criminal activity (both property crime and violent crime) is nearly triple that of cells that contain no foreclosures. Moreover, beyond 200 meters, foreclosures have a marginally significant effect on violent crime only after there are at least two foreclosures within a cell.
My results suggest that foreclosures exert a negative externality through increased crime that extends beyond their immediate surroundings; in some cases, the effect is detectable (albeit diminished) up to a kilometer away. Though the general distance effect may vary from place to place, these findings suggest that foreclosures may encourage crime beyond their immediate vicinity.
My findings come with several caveats. First, my model is underequipped to determine whether foreclosures encourage new crimes of opportunity, or instead displace crimes that would have otherwise occurred elsewhere. Thus, my results do not necessarily imply that cities reeling from the foreclosure crisis are consequently likely to experience more crime overall. Instead, my findings suggest that it might behoove police and residents to monitor affected areas more closely.
My analysis also does not tease apart the various stages of the foreclosure process; this makes it difficult to precisely estimate which aspects of foreclosure cause different types of crime. The theoretical literature suggests that vacancies play a particularly significant role in neighborhood crime levels, and it is unclear from my results whether the observed distance-decay effect on crime results from the foreclosure process overall or from vacancies specifically. Finally, my models rely on yearly foreclosure and crime counts. It is possible, however, that shorter time intervals would yield better insight into the relationship between foreclosure and crime and the processes that drive it.
My analysis prioritizes examination of foreclosure effects that decline with distance over rigorous inspection of the cumulative effects of foreclosure within an area. While some effort was made to study whether grid cells with more than one foreclosure impacted their surroundings differently, this paper does not fully and deliberately investigate the extent to which the influence of clustered foreclosures may become disproportionate once their number reaches certain thresholds. Social disorganization, broken windows, and routine activities theories, which focus on reduced informal social control and surveillance and a subsequent increase in criminal opportunity, leave ample room for such a scenario.
Even taking these caveats into consideration, my findings are strongly suggestive of a causal distance-decay effect of foreclosures on property crime and, to a lesser extent, violent crime. My results also intimate that foreclosures not only lead to elevated crime in their immediate vicinity, but also to more modest increases up to a kilometer away. The estimated effects are relatively small, and their magnitude and significance are greater for property crimes than for violent crimes. Nevertheless, these findings provide additional impetus for policy initiatives that aim to reduce foreclosures as a way of curbing the negative externalities associated with them.
** Please check out the full paper for sources, background information, appendices, and more!