https://orcid.org/0000-0003-4672-2079
![Sample our Law journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5942/TOC-Subject-Sample-2021-Law.jpg"})
Abstract
Interpretations of two bodies of crime-place research conflict. Land use and crime studies claim particular facilities increase crime. Risky facilities studies show most places of a single type have little or no crime, but a few of that type have a great deal of crime. How can a facility be generally criminogenic and mostly safe? To resolve this conflict, we make use of the fact that a single owner can own multiple facilities and each owner may have consistent management practices in their facilities. We first replicate findings of earlier land use studies with crime and land parcel data from Cincinnati. Second, we cluster land parcels by property owners and re-estimate the land use and crime relationship. The links between land use and crime in the replication decline or disappear after clustering. Findings suggest owners’ place management influences crime at their places, thus resolving the conflict between the two bodies of research.
Disclosure Statement
No potential conflict of interest was reported by the authors.
Notes
1 A negative association between land use type and crime may decline once we account for management, if much of the crime suppressing effects of the land use were due to the management, rather than the land use per se. For a land use that is inherently crime suppressing, it would take considerable bad management to create crime.
2 Other public structures and parcels include city structures (0.4%), federally managed parcels (0.07%), utility parcels (0.02%), and 0.003% of common areas and greenbelts.
3 In Appendix C, we provide the full distribution of crime for all 19 land uses in J-curve.
5 To calculate the unemployment rate of each block group, we use the population of males over 16-year old rather than both male and female populations as the number of female-headed households is already included in the concentrated disadvantage index.
6 In Appendix D, we provide a detailed illustration and explanation of this procedure.
7 Some parcels are tax exempt – churches, schools, and some other types.
8 In Appendix F, we provide a detailed statistical and theoretical rationale for using fixed effects transformation over random effects.
9 What usually happens to a bar (or other land uses) when it is sold but stays a bar? The ownership changes from the previous owner (a) to the new owner (b). The values of Bars variable would remain 1. This particular bar parcel has two different net effects on crime over the study period (we call it “net bar effect” since it is net of each owner’s place management effect). Specifically, there is a net bar effect (Effecta) during the owner (a) period and another net bar effect (Effectb) during the owner (b) period. Since this bar was owned by two different owners, the “average net bar effect” on crime of this bar is (Effecta+ Effectb)/(the number of owners, here 2). Note that there are more bars (bar#2, bar#3, …) to consider in our study. Calculating the “average net bar effect” of each bar (across multiple owners over time) the same way, the statistical package would give the mean of “average net bar effects” from all bars (from bar#1, bar#2, bar#3, …). This would be the inherent pure effect of bars on crime controlling for (net of) all property owners’ place management effect.
10 A Hierarchical Linear Model (HLM), with parcels at level one and owners at level two, might seem preferable. In this study, it is not. First, HLM demands we assume that land uses are randomly allocated to property owners. This is very implausible: owners tend to specialize in the land uses they control (e.g., school owners do not own bars, bar owners are more likely to own other bars than convenience stores). Without this assumption we cannot make causal inferences about the effect of property owners on land uses when using HLM. Consequently, did not use HLM.
11 Since we employ a Poisson model to analyze the number of crimes at each parcel, our model assumes that crimes occurred at each parcel are independent with crimes at different parcels. This allows us to analyze the link between land use and crime at each parcel without the spatial lag effect among adjacent parcels. Further, since our geographic units of analysis are individually separated parcels, assuming an automated spill-over effect across nearby parcels would suck up too much variation in the outcome variable, and the model begins to describe the random error in the data by itself rather than the relationships between variables are explaining. Doing this may lead our results to suffer from an “overfitting problem.” We tested the possibility of spatial autocorrelation using Moran’s-I statistics (with z-score = 0.007038) and found that the distribution of crimes across the parcels does not appear to be significantly different than random.
12 If a variable’s IRR is greater than 1 (say, IRR is 2) then the land use is 2 times riskier than our reference variable, thus the variable increases the likelihood of crime. If the variable’s IRR is between 0 and 1, then the variable decreases the chance of crime.
13 To test whether the three variables (tax paid, area size, and market total value) are appropriate proxies to crime opportunity, we run the same model with and without these variables. We found that eliminating these variables from the model significantly changes some of our findings between land use and crime. For example, full restaurant became non-significant. Further, though the model still remained significant, the model statistics (here, Wald Chi-Square) decreased about 39% when we removed the three opportunity variables.
14 The Intraclass Correlation Coefficient (ICC) indicates that about one-third (or 36%) of the variance is due to property owners, and the remaining two-thirds of the variance in crime is due to land use, control variables, and their interactions. For a detailed explanation, see Appendix G.
15 There are 22 to 23 strictly liquor serving bars (with no foods) for each year, which sums up to 111 bar parcels observed during the five-year study period. The top 10% of the worst bar owners (two of them manages 3 bars for each) experienced about 47% of crimes. According to the J-curve of bars in Appendix C, the top 10% of the worst bars accounted for 51% of crimes. So we have a concentration of crime at the top worst bars, but also at the top worst bar-owners as well.
16 We consider the possibility of suppression effect by checking the variance inflation factor (see Appendix J) of each independent variable. However, all variance inflation factors show that the suppression effect is very unlikely.
17 In Appendix K, we also provide how different land uses experience relative risks (in x-times) compared to public places.
18 In Appendix L, we also provide a full list of estimated coefficients of 18 land uses on crime using single-family homes as an alternative reference variable.
19 In Appendix M, we provide summary of IRR-changes from unclustered to clustered model in percentage of 18 land uses and their number.
Additional information
Notes on contributors
YongJei Lee
YongJei Lee is an assistant professor in the School of Public Affairs at University of Colorado Colorado Springs. His current research examines police effectiveness, spatio-temporal patterns of crime hot spots, crime hot spot forecasting algorithm, concentration of crime at places, offenders, and victims, and measures of crime concentration phenomenon.
SooHyun O
SooHyun O is an assistant professor in the School of Criminology, Criminal Justice and Strategic Studies at Tarleton State University. Her current research examines juvenile delinquency, juvenile victimization at school, crime hot spot forecasting, and the relationship between land use and crime.
John E. Eck
John E. Eck is a professor in the School of Criminal Justice at the University of Cincinnati. He specializes in the study of high crime places and in police effectiveness.
