Hiropoulos, A., & Porter, J. (2016). Visualising property crime in gauteng: Applying GIS to crime pattern theory. South African Crime Quarterly, (47) doi:10.17159/2413-3108/2014/v0i47a802

 

Utilizing the framework of crime pattern theory (CPT), the objective of this research is to visualize neighbourhood-level patterns of motor vehicle crimes in the ‘home of crime’ of South Africa, Gauteng. Although Gauteng is the smallest province in South Africa, it accounts for 50% of the country’s crimes, most of which being property crimes (thefts out of motor vehicles). This inspired the authors to employ GIS to better explain the occurrences of property crimes in the context of the geography these crimes were committed.

Before delving into the analysis, the authors did a thorough literature review on CPT and ecological analysis of crimes to set the stage for their own research. They provided a detailed explanation for their choices of independent variables and statistical models with reference to CPT. They then employed two statistical models, Moran’s I and Local indicators of spatial association (LISA), to analyze crime patterns in their study area.

The main argument presented in this article is that geography matters to crime occurrences in Gauteng. Thefts out of motor vehicles are highly correlated spatially and are concentrated around the two central business districts (CBDs). CBDs are identified as crime generators and attractors, two common types of crime-ridden locations according to CPT. Hence, their analysis confirms that CPT is an effective framework to predict crime occurrences in Guateng.

They came to this conclusion based on the results from 3 analytical approaches. First, they tested the presence of spatial autocorrelation using Moran’s I, which indicated there is a positive spatial autocorrelation across crime event. They furthered this analysis by looking at a local measure of autocorrelation, LISA, to reinforce the result of spatial autocorrelations by placing a spatial limit on areas of highest crime event concentration. They also derived a hot spot map from LISA. Finally, they looked at the environment in which crimes are committed. They did so by overlaying layers of nodes (locations an individual uses regularly) and paths (actual paths an individual takes to and from their personal activities), such as shopping malls and main roads, onto the hot spot map. This allowed them to visually examine where crimes tend to clusters and determine how well CPT explains crime occurrences in Gauteng.

While one of the main arguments of this article is that GIS is a powerful and practical tool in crime data presentation, I found the use of GIS in this study extremely limited. GIS was applied merely to overlay layers of crime attractors on the LISA map for visual identification of hotspots. However, there is so much potential for their analysis to go beyond visual representation and provide more evidence to support their argument statistically. For example, the authors could have used more sophisticated software such as CrimeStat to calculate different indexes of spatial autocorrelations. Even within the scope of ArcMap, they could have utilized spatial analyst tools such as kernel density, hot spot analysis, and grouping analysis to present crime data in a more accurate and statistical way. Furthermore, the theory they used to base their results on did not account for other factors that could affect crime rate, such as socio-economic status, culture, ethnicity, etc. The results were rather intuitive—as the authors mentioned twice in the article, the results are ‘not surprising’. Another problem with this research is related to the normalization of data using population density. Population density is generally lower in CBDs due to the dominance of commercial land uses; however, crime occurrences tends to be high because they are easily accessible and crowed with vulnerable targets. Using population to calibrate the data may, therefore, exaggerate crime problems in CBD.

For these reasons, I would rate this paper a 6.5 out of 10.

---

This week, I found Baris’ review on the article below extremely intriguing.

Caplan, J. M., Kennedy, L. W., & Miller, J. (2011). Risk terrain modeling: Brokering criminological theory and GIS methods for crime forecasting. Justice Quarterly, 28(2), 360-381. doi:10.1080/07418825.2010.486037

Risk terrain modelling (RTM) was used in this article to forecast the crime of shootings. Maps that were produced from RTM use a range of contextual information relevant to the opportunity structure of shootings to estimate risks of future shootings as they are distributed throughout the geography. Baris did a great comparison of the 2 commonly used GIS methods, risk terrain modelling and retrospective hot spot mapping, and demonstrated the benefit of the former in crime forecasting. Risk terrain modelling seems to be an effective alternative to the analytical method used in the paper I reviewed. Baris also explained in detail why he believed in this method; hence I’d be interested to read this paper and learn more about risk terrain modelling.