Neighborhood Disinvestment, Abandonment, and Crime Dynamics

:This article develops a conceptual framework of neighborhood crime dynamics based on a synthesis of criminology and neighborhood change literatures which suggests that neighborhood decline can produce a nonlinear response in crime rates. The authors probe this relationship using a rich Detroit data set containing detailed, block-level information about housing, land, abandonment, population, schools, liquor outlets, and crime reports of various categories. Negative binomial models reveal that several neighborhood attributes are consistently associated with all types of crime (renter occupancy, population density, establishments with liquor licenses) while other attributes are only associated with particular types of crime. A simulation using estimated parameters suggests that processes of disinvestment and abandonment can generate a nonlinear pattern in the rate of growth in neighborhood crimes that vary in intensity by crime type. The authors explore the implications of their findings for anticrime strategies focusing on demolishing abandoned housing, "right-sizing" urban footprints, and regulating liquor-selling establishments. NotesFor additional references, see Miethe and Meier (Citation).One of the few consensual findings in the criminological literature is the positive correlation between neighborhood poverty rates and crime rates (Crutchfield, Glusker, and Bridges, Citation; Hannon, Citation, Citation; Hipp, 2007; Warner & Pierce, Citation; Warner & Rountree, Citation), although recently Hipp and Yates (Citation) found that crime rates were highest in tracts with 35% poverty rates and lower in tracts with even greater poverty concentrations.The vacant lot designation used for this study includes both "unimproved" (74%) and "improved" (24%) vacant lots (n = 91,488) as recorded by the Detroit Residential Parcel Survey. Both types are defined as parcels without residential structures, and "improved" parcels are those with some visible improvement, such as a paved lot, accessory structure, fence, or park.The Herfindahl index of ethnic heterogeneity is expressed as , where G represents the proportion of the population of race/ethnic group j out of J groups. In this instance, the index includes Hispanic or Latino, non-Hispanic white alone, non-Hispanic black or African American alone, non- Hispanic American Indian and Alaska Native alone, non-Hispanic Asian alone, non-Hispanic Native Hawaiian and Other Pacific Islander alone, non-Hispanic some other race alone, and non-Hispanic two or more races. The index takes on a value of zero with a homogenous composition and a limiting value of one with many groups identically represented.Alternatively, we employed the conventional poverty rate available from ACS but found it less predictive.Our Herfindahl index of earning heterogeneity uses four groups: residents who have not worked in the past 12 months and are assumed to have zero earnings (from ACS), employed residents earning $1,250 per month or less, employed residents earning $1,251 to $3,333 per month, and employed residents earning $3,333 per month or more. The inclusion of ACS data for this measure required aggregation of OTM data to block group–level, and application of the calculated values to blocks comprising each block group.All these lags employ a distance-to-block centroid within cutoff threshold of 2,232 feet. These spatial lags assume that the relationship between crime in the focal block and vacancy/blighting influences in proximate blocks is linear and symmetric regardless of whether nearby blocks are more or less blighted than the focal block.The spatial lag for burglary and drug crime was based on queen's-based contiguity; for vandalism, robbery, and all property crime it used distance within a threshold of 2,232 feet; for all other crimes it used distance within threshold of 4,680 feet.This variable was positively associated with rates of violent crime in all three categories analyzed, but in no case did the relationship approach statistical significance.Of 11,289 blocks in the sample, 4,673 (41.4%) have at least one DFV, and 2,574 (22.8%) have two or more DFVs.For the simulation we are forced to switch to the census tract scale because the variable distributions at the block scale were so positively skewed that distinct quintile breaks often could not be identified. Note all the zero medians for crime in Table and appendix table.For tractability we excluded spatial interactions in this simulation, both endogenously among blocks within our hypothetical tract and from other blocks nearby.The first proportional change was based on the number of crimes at the 10th percentile of the distribution, assuming this would appertain to the neighborhood in Stage 1; thereafter in the simulation the proportional change was based on the prior stage's simulated level of crime. In the simulation we assumed that all blocks within the stylized census tract changed uniformly, since our coefficients are based on underlying block, not tract, data. We also assume that the negative binomial regression parameters estimated cross-sectionally appertain to this implicitly longitudinal process being simulated.Additional informationNotes on contributorsErica RaleighErica Raleigh is the Acting Director of Data Driven Detroit, a nonprofit organization committed to the equitable access of information for data-driven decision making. She holds a Master of Urban Planning from Wayne State University, and a Bachelor of Arts in Hispanic Studies from the University of Michigan. Her research explores neighborhood quality-of-life issues, especially the interplay between neighborhood change and crime dynamics.George GalsterGeorge Galster is the Clarence Hilberry Professor of Urban Affairs at Wayne State University. His research has focused on urban neighborhoods and housing markets, exploring how they change and how they change the people who live within them. This has resulted in over 130 peer-reviewed articles, 30 book chapters, and seven books. His latest book is Driving Detroit: The Quest for Respect in the Motor City (2012). He earned his PhD in Economics from the Massachusetts Institute of Technology.