Leading Indicators and Spatial Interactions: A Crime‐Forecasting Model for Proactive Police Deployment

Abstract

We develop a leading indicator model for forecasting serious property and violent crimes based on the crime attractor and displacement theories of environmental criminology. The model, intended for support of tactical deployment of police resources, is at the microlevel scale; namely, 1-month-ahead forecasts over a grid system of 141 square grid cells 4000 feet on a side (with approximately 100 blocks per grid cell). The leading indicators are selected lesser crimes and incivilities entering the model in two ways: (1) as time lags within grid cells and (2) time and space lags averaged over grid cells contiguous to observation grid cells. Our validation case study uses 1.3 million police records from Pittsburgh, Pennsylvania, aggregated over the grid system for a 96-month period ending in December 1998. The study uses the rollinghorizon forecast experimental design with forecasts made over the 36-month period ending in December 1998, yielding 5076 forecast errors per model. We estimated the leading indicator model using a robust linear regression model, a neural network, and a proven univariate, extrapolative forecast method for use as a benchmark in Granger causality testing. We find evidence of both the crime attractor and displacement theories. The results of comparative forecast experiments are that the leading indicator models provide acceptable forecasts that are significantly better than the extrapolative method in three out of four cases, and for the fourth there is a tie but poor forecast performance. The leading indicators find 41-53% of large crime volume changes in the three successful cases. The corresponding workload for police is quite acceptable, with on the average 5.2 potential large change cases per month to investigate and with 31% of such cases being positives.

Likelihood Based Inference and Prediction in Spatio-Temporal Panel Count Models for Urban Crimes

“…Explanations for this temporal development of crime hot spots emphasize that public signs of disorder like vandalism, gambling and 'broken windows' foster increases in more serious crime, since they signal a loss in the ability to exercise social control, further attracting and perpetuating crimes (Wilson and Kelling, 1982). This`broken-windows' phenomenon suggests that the intensity of soft crimes observed in a given regional area may serve as a leading indicator for the number of serious crimes in that area (Anselin et al, 2000, p. 225 and Cohen and Gorr, 2005, Cohen et al, 2007. In order to make use of the leading-indicator properties of soft crimes when specifying a predictive panel model for severe crimes we shall include them as lagged explanatory variables.…”

“…See, e.g., Wilson and Kelling (1982), Anselin et al (2000, p. 225), Cohen and Gorr (2005) and Cohen et al (2007). In particular, our model allows us to compute both short-term and long-term elasticities of severe crimes w.r.t.…”

“…A further motivation for our paper originates from three earlier papers who analyzed a variety of forecasting models for severe crime in Pittsburgh Cohen et al, 2007) and in Pittsburgh and Rochester, New York (Cohen and Gorr, 2005). In the absence of potential socio-economic covariates and spatially and temporally lagged dependent variables, they nd that a linear multivariate leading indicator model is typically best for forecasting large decreases in crime volume, while various forms of exponential smoothing provide the best average forecast point accuracy according to mean squared and mean absolute percentage forecast error.…”

Analyzing Local Spatio-Temporal Patterns of Police Calls-for-Service Using Bayesian Integrated Nested Laplace Approximation

“…It would also be advantageous to analyze hourly patterns in context of daily, monthly, and seasonal patterns, which may be suited for a multilevel spatio-temporal modeling approach. Finally, future research may adapt the presented spatio-temporal Bayesian model for probabilistic forecasts [69,70].…”

ROC-based model estimation for forecasting large changes in demand

“…The dependent variable is the count of serious violent crimes (homicide, rape, robbery, and aggravated assault), while the 12 leading indicators are one-, two-, three-, and four-month time lags of illicit drug 911 calls for service, shots-fired 911 calls for service, and offense reports of simple assaults (Cohen, Gorr, & Olligschlaeger, 2007;Gorr, 2009). The data span the period from January 1990 to December 2001 across 175 census tracts, with 24,500 observations available out of 25,200 after dropping the first four months' observations used for time-lagged variables.…”