Capturing Crime at the Micro-place: A Spatial Approach to Inform Buffer Size

We define an abandoned property as any physical structure that is permanently unoccupied.

For example, consider micro-place research that uses the buffer zone or distance decay hypotheses to spatially capture the geographic scope of offenders’ criminal behaviors. The buffer zone hypothesis claims that offenders avoid committing crimes close to home. Two primary arguments support this hypothesis. Proponents of the buffer zone hypothesis argue that offenders avoid committing crimes close to home due to lack of suitable targets (Rengert et al. 1999; O’Leary 2011). Offenders may also avoid committing crimes close to home out of fear of being recognized (Brantingham and Brantingham 1981; Cromwell et al. 1991; Wright and Decker 1994). Alternatively, the distance decay hypothesis claims that offenders are more likely to commit crimes in areas close to their homes of which they readily frequent and are familiar. These hypotheses are in opposition with one another, yet draw from the same reservoir of crime and place theories to explain offending behavior. Furthermore, they lead to different scale selections to capture the geographic scope of the phenomenon of interest (see Bernasco and van Dijke 2020).

For example, the multi-scale error analysis method, developed by Malleson et al. (2019), represents a serious advancement in spatial criminology. However, this method is not appropriate when the aim is to understand the global impact of an intervention on crime at treatment sites and surrounding areas.

While not the focus of our discussion, it should also be known that the MAUP can result in variation in analytical results due to the zoning effect. To elaborate, the zoning effect arises when analytical results are affected as a result of alterations to the structure of spatial units, but not their number (Openshaw and Taylor 1979). As it relates to buffers, concerns regarding the zoning effect heighten when buffers vary in shape.

A high degree of similarity between the underlying processes of the spatial features influences the observed spatial patterns towards attraction.

We further omitted crimes that were irrelevant to the study, such as retail fraud for larcenies or telephone harassment for assaults.

Given its public nature, it is possible that the DLBA data could influence the behaviors of individuals or groups that utilize abandoned properties for illicit activities. To this point, a helpful reviewer raised the possibility of an anticipation effect, whereby individuals or groups leverage the DLBA data to determine whether the abandoned properties they utilize should be deserted for alternatives not slated for demolition. We do not believe this scenario to be very likely. Properties that are slated for demolition are often not assigned demolition contracts in a timely manner. Individuals or groups that would utilize the DBLA data in the manner described would quickly become wise to Detroit’s untimely demolition process and would likely lose motivation to find alternative abandoned properties. In the unlikely scenario in which individuals or groups select alternative abandoned properties, we suspect—drawing from well-established research on offender decision-making – that they would target properties in spatially proximal areas (e.g., Rengert and Wasilchick 2000; Bernasco 2010; Brantingham et al. 2017). If alternative abandoned properties are not identified within these areas, individuals or groups will likely forgo their search. In addition, the same helpful reviewer raised the possibility of non-interference, whereby knowledge of Detroit’s demolition program motivates individuals or groups to conceal illicit activities in abandoned properties in an attempt to avoid demolition. While some abandoned properties are slated for demolition as a consequence of known illicit activities that occur within them, the vast majority are slated for demolition based upon their condition alone. Thus, we do not believe non-interference poses a significant issue for our analytical strategy.

Contractors are invited to bid on packages of houses—ranging from 1 house to over 100—released by the DLBA. Bidders are scored on several criteria, including whether the company is based or headquartered in Detroit, the price per bid, and the company’s ability to handle the workload (City of Detroit 2020). Once a contract has been awarded, the contractor has 120 days to complete the demolition. Importantly, not all properties that are slated for demolition are assigned demolition contracts in a timely manner. If a property has not been assigned a contract it may not be demolished for years. The time to demolition largely depends on whether a property is located within a federally designated zone. These zones were selected based upon the marketability potential of Detroit neighborhoods for redevelopment investment and cover approximately 43% of the total area of the city (City of Detroit 2020). Furthermore, demolitions that occur within these zones are supported by the Hardest Hit Fund, a federally-allocated fund for the purpose of spurring economic development and reducing blight. Given access to these funds to finance demolitions, abandoned properties that are located within federally designated zones are demolished more quickly than abandoned properties located elsewhere. That being said, demolitions occurred both within and outside of federally designated zones during the time period of our study.

We considered the entire city of Detroit for analysis, with the exception of the Downtown and Midtown region of the city. These areas represent a significantly gentrified and more urban environment than the outlying neighborhoods. In addition, the number of residential demolitions was too small to reliably estimate an effect. Therefore, this portion of the city was not included in our analysis. We should note, even with this area included the results were unchanged.

A cutoff of 1,000 feet helps lessen the risk that the studied phenomena are obscured by processes associated with competing spatial scales.

For example, other covariates become redundant when all lagged outcomes are included in the ADH method (Kaul et al. 2015).

Considering our analysis, this explanation is less compelling for cases in which crime is significantly elevated in post-demolition time periods because it suggests that the vacant land created by demolition efforts are more problematic than the abandoned properties that previously occupied them.

Separate analyses could be performed utilizing the bivariate K-function for neighborhoods of varying typologies that are created based upon contextual factors of interest.