Recent research in the ‘‘crime at places’’ literature is concerned with smaller units of analysis than conventional spatial criminology. An important issue is whether the spatial patterns observed in conventional spatial criminology focused on neighborhoods remain when the analysis shifts to street segments. In this article, the authors use a new spatial point pattern test that identifies the similarity in spatial point patterns. This test is local in nature such that the output can be mapped showing where differences are present. Using this test, the authors investigate the stability of crime patterns moving from census tracts to dissemination areas to street segments. The authors find that general crime patterns are somewhat similar at all spatial scales, but finer scales of analysis reveal significant variations within larger units. This result demonstrates the importance of analyzing crime patterns at small scales and has important implications for further theoretical development and policy implementation.

Crime rate is a statistic used to summarize the risk of criminal events. However, research has shown that choosing the appropriate denominator is non-trivial. Different crime types exhibit different spatial opportunities and so does the population at risk. The residential population is the most commonly used population at risk, but is unlikely to be suitable for crimes that involve mobile populations. In this article, we use “crowd-sourced” data in Leeds, England, to measure the population at risk, considering violent crime. These new data sources have the potential to represent mobile populations at higher spatial and temporal resolutions than other available data. Through the use of two local spatial statistics (Getis-Ord GI* and the Geographical Analysis Machine) and visualization, we show that when the volume of social media messages, as opposed to the residential population, is used as a proxy for the population at risk, criminal event hot spots shift spatially. Specifically, the results indicate a significant shift in the city center, eliminating its hot spot. Consequently, if crime reduction/prevention efforts are based on resident population based crime rates, such efforts may not only be ineffective in reducing criminal event risk, but be a waste of public resources.

PurposeCrime analysts need accurate population-at-risk measures to quantify crime rates. This research evaluates five measures to find the most suitable ambient population-at-risk estimate for ‘theft from the person’ crimes.Method1.Collect ‘ambient’ datasets: the 2011 Census, aggregate mobile telephone locations, and social media.2.Correlate the population measures against crime volumes to identify the strongest predictor.3.Use the Gi* statistic to identify statistically significant clusters of crime under alternative denominators.4.Explore the locations of clusters, comparing those that are significant under ambient and residential population estimates.Results and DiscussionThe research identifies the Census workday population as the most appropriate population-at-risk measure. It also highlights areas that exhibit statistically significant rates using both the ambient and residential denominators. This hints at an environmental backcloth that is indicative of both crime generators and attractors – i.e. places that attract large numbers of people for non-crime purposes (generators) as well as places that are used specifically for criminal activity (attractors). Regions that are largely residential and yet only exhibit hotspots under the ambient population might be places with a higher proportion of crime attractors to stimulate crime, but fewer generators to attract volumes of people

It is well known that, due to that inherent differences in their underlying causal mechanisms, different types of crime will have variable impacts on different groups of people. Furthermore, the locations of vulnerable groups of people are highly temporally dynamic. Hence an accurate estimate of the true population at risk in a given place and time is vital for reliable crime rate calculation and hotspot generation. However, the choice of denominator is fraught with difficulty because data describing popular movements, rather than simply residential location, are limited. This research will make use of new ‘crowd-sourced’ data in an attempt to create more accurate estimates of the population at risk for mobile crimes such as street robbery. Importantly, these data are both spatially and temporally referenced and can therefore be used to estimate crime rate significance in both space and time. Spatio-temporal cluster hunting techniques will be used to identify crime hotspots that are significant given the size of the ambient population in the area at the time.