Spatial-Temporal Analysis Of Crime: Gun And Knife Crime In London

This dissertation report is a description and analysis that I have done during the summer. The dissertation responses to the analysis of the crime, in particular, the “Violence crime” which involves the possession of weapons in London. The dissertation deployed a various form of analysis, combining both statistical and spatial analysis which aims to discover the patterns and explain the underlying causes of the occurrences of crime, supplemented to the current study of environmental criminology. The purpose of this dissertation report seeks to build a foundational analysis for further hypothesis testing based on both theory and observation. In particular, this report will be presented with five sections followed by the research questions and further refinement of the hypothesis.

Introduction

The context of this report is based on the previous dissertation proposal “spatial-temporal analysis of crime: gun and knife crime in London”. The methods are inspired by the theoretical framework of crime pattern theories which I further contextualized with spatial analysis. In the first sections, I will elaborate my findings and methods spatial and temporal clustering using the app I have, detecting how different types of crime exhibits the pattern of hot-spots of crime in London which providing a fundamental mechanism to further narrow down to specific locations and features. In the second section, I further my analysis through analyzing the spatial significance of crime, using local moron’s I and G-stat, arguing that the distribution of crime is not random, allowing me to identify the area as similar or different to their nearby pattern. Significance testing validates my analysis with the crime pattern theory, aiming to show spatial repeat and near repeat victimization (NRV) which identify those people, properties and places that are at a disproportionate risk of victimization. The third sections, I would incorporate temporal patterns as factors, based on the theory of offender-as-forager hypothesis, testing whether there is an interaction between spatial location and time, using Knox Index and Mantel Index that show the relationship between 'closeness in time' and 'closeness in the distance’. The fourth sections, I will reference prior studies and theories on crime, in particular, focusing on the underlying causes of violent crimes. Aiming to develop more details hypothesis on social-economics factors, geodemographics, and land use based on the causal analysis. This section is designed to help improve the explanatory content of analytical products such as problem profiles and principal component construction. In the last sections, I will briefly talk about my next stage of analysis based on the work I have done.

Section1: Web App integration of hot spot analysis and methods approach

Crimes are not spread evenly in space. These areas of concentrated crime are often referred to as hot spots, which many researchers and crime analyst would observe these abnormal spatial patterns of crime and try to infer the cause of these crimes from different social theories. For example, Sherman et al (1989) showed that 3% of dress accounted for 50% of total crime calls to the police I Minneapolis, US. The main idea of the first research questions takes a rather automated and descriptive approach to data collection and visualization, which seeks to identify patterns in the phenomenon of interest, using points pattern analysis at different level of studies. The nature of crime is extremely complex, different crime types may exhibits totally opposite distribution and significance in different places. The recent innovation in GIS and Crime mapping is interactive mapping, offering users for initiating the next analytical stages that explain the problem within an automatic and simplifying manner. Therefore, with this in mind, I have developed an interactive web apps which allow user to select temporal period, different boroughs, different crime types, different types of point pattern analysis and parameters which control spatial statistics. For the purpose of this web app, I would use point data for more details spatial analysis.Figure 1.1 Control panels for shiny apps To demonstrate this app, Figure 1.2 shows all the records of crime from 2016 to 2018 with different colors specifying different crime data as point maps. As you can see it might be difficult to identify the location, relative scale, size and shape of hotspots when crime data are presented as points. The large volume of occurrence with the anti-social behavior related crime make it challenging to visualize and interpret specific patterns In the data’s spatial distribution. However, this interactive tool could visualize each crime type by moths which can be more visually interpretable as shown in Figure 1.3. Taking the main research crime type “possession of weapons” with the recently published data as an example, the patterns become slightly more transparent than the previous visualization. Nevertheless, an increasingly popular method for visualizing the distribution of crime and identifying hotspots on a holistic level is to create a smooth continuous surface to represent the density or volume of crimes distributed across the study areas. Specifically, utilizing dual kernel density estimation to explore the relative intensity of crime occurrences for all locations between sample points and areas. KDE was conducted using the bandwidth and cell size parameters chosen by Ratcliffe and Chainey in a case study of “theft of vehicle” incidents for the borough of Camden to ensure that KDE application would be comparable to others. Figure 1.4 shows the dual kernel density smoothing map of possession weapons which indicates the relative proportions of this particular crime type, comparing to all the criminal records in London, the scale of the right showing the intensity of kernel estimation. A natural progression of this methods may lead to further analysis based on place theories, asking such questions as “At what places re Possession of weapon crimes occurring and at what places are they not occurring? From this maps, it revealed several interesting patterns of “hot spots” and “cold spots”, which majority of the hot spot is clustered around the north and east-west of London, while North-west London and South of London illustrates a “cold spot” of possession of weapon crimes. The boroughs which have the higher proportion of possession of weapons are Enfield, Haringey, Hackney, Newham, Barking and Dagenham, Havering, Camden, North Croydon. The following kernel density demonstrates with a focused analysis on a borough level, identifying a crime that is closeted around a particular place of interest. The Figure 1.5(a) shows one the hot spot regions, hackney whereas the figure 1.5(b) is the city of London where the proportion of possession of weapons crime is lower than other regions. As suggested by Wilson and Evertt (2004), case studies in criminal activities can not only to be empirical generalizations, which means that case studies are used to determine issues of causality and process, allowing us to investigate the underlying causal factors and its theory.

Section 2: Spatial autocorrelation and statistical testings

Further extending the previous section which illustrates both “hot” and “cold” spot in London, exploring the occurrence of crime rates. Nevertheless, section 1 acts as a descriptive framework, which in this session, it illustrates the use of Global and Local Moran’s I to test the spatial process of routine activities theory and crime pattern theory. At the broader level, regional crime distribution can be explained by many factors can be influenced by, or at least related to, the nearby locations. Crime pattern theory as the examines the “relationship of the offence to the offender’s habitual use of space”, which suggests that offenders are influenced by the daily activities and routines of their familiar space. To put it another way, from a offenders’ perspectives, their day-to-day activities will be actively seeking for suitable target with low-level of guardians or place managers whereas from a spatial perspectives, places with low social order and poor economic infrastructures, there are tendencies for offenders to react similarly to the same opportunities. Therefore, following this theory, detecting spatial autocorrelation can not only suggest clustering of like values but also the possibility of a dispersion of values, so that high-value areas are surrounded by low values areas with testable statistical significance.

Figure 2.1 is a Moran Scatterplot which illustrates the relationship between the values of the chosen attribute at each location and the average value of the same attribute at nearby locations, which I use “possession of weapons” crime as the example. From the figure 2.1, it suggests that large percentages of crime occurrence (points on the right-hand side of the plot) tend to be associated with high local average values of percentage of crime (points toward the top of the plot, which means that crime is spatially clustered in space. Furthermore, this significance is confirmed by a global Moran’s I value of 1.760162e-01 (0.18) with a p-value smaller than 0.05 (figure 2.2), suggesting a positive spatial autocorrelation which supports the prior visual evidence of a North-South divide in possession of weapons. The Local Moran’s I statistic can go a stage further and identify which borough are in statistically significant clusters. The result of this analysis refining the hotspot analysis with significance testing. In central and north London, some boroughs show a high-value clustering with a statistical significance (dark red) whereas there are some areas around the north-west and south London with light blue show represent the low-value clustering with a statistical significance. A natural progression from this methods may lead to further analysis based on comparative case studies, observing whether clustering of gun and knife crime differ from other types of crimes, and how does it change temporally. The third sections, I would incorporate temporal patterns as factors, based on the theory of offender-as-forager hypothesis, testing whether there is an interaction between spatial location and time, using Knox Index and Mantel Index that show the relationship between 'closeness in time' and 'closeness in the distance’.

An exciting development in the environmental criminology emphasize the detection of hot-spots is its persistence and coincidence over time as shown by Anselin (2000), the hot-spots reflect high levels of crime initially moderate, but over time, usually, this crime will change to more violent types of crimes. Therefore, research should be considered and controlled in time to prevent more severe incidents to people and property by the hot-spot.

In the fourth sections, I will highlight the prior research which investigates the underlying factors for crime occurrence, establishing a more details hypothesis and refinement of variables that need to be investigated. This sections would explore from three general: social-economics factors, geodemographics and lifestyle and land use from theoretical perspectives. Crimes are not spread evenly in space. The fundamental assumption behind spatial crime pattern analysis is that crimes would correlate with environment settings and this assumption is supported by related theories, such as environmental criminology and broken windows theory. The critical concept of environmental criminology suggests that environmental factors would influence criminal activity while the broken windows theory suggests that areas with persistent crime issues would further deprive the neighborhoods. Offenders’ choices of locations, victims, environments and neighborhoods are governed partially by understanding that their chances of successfully committing a crime are higher in some of these places than in others.

As Nettler (1978) points out, the criminological theory has been dominated by two units of analysis: individuals and communities. From individuals or a micro level perspectives, it is interesting to see how social-economics factors might act as a rational choice for offenders to achieve their criminal goals. Notably, it is argued that criminal decision-making is based on the built environment which might be appealing for certain criminal activities. For example, Spano et al. (2208) specify the occurrence of gun carrying in New York based on the vulnerability hypothesis where the areas with the lowest level of surveillance and related activities such as extreme poverty increase the chance of violent victimization. Applying this framework to this current study in London, I would set consider two aspects. Firstly, the related crimes that increase the use of offensive weapons. Specifically, I will investigate whether the increase of offensive weapons is attributable to, for example, an increase in drug dealing and antisocial behaviors in London. I would conduct further Principle component analysis for the selections of crime types. Secondly, the social-economic factors that would foster the environment with a low level of guardians and surveillance, for example, poverty and parental education as the predictors. Poverty and inequality have become entrenched in localities where, in the absence of employment or significant material assistance, involvement in various forms of crime may be one of the few ways to make a living. The lack of economic opportunities for young people has stimulated the growth in both the US and the UK of illegal economies around drugs, stolen goods and protection. From communities or a neighborhoods perspectives, deviant lifestyle and demographics argue that the risk of victimization is the result of routine activities in a neighborhood. There are some literature explores Gang activity as a dangerous lifestyle that translates into higher rates of both violent offending and victimization. Battin et al. (1998) argued that gang membership was related to delinquency and substance abuse as well as rival gang fights in Denver. Victimization may occur during the commission of criminal behavior, which from this perspectives, is it interesting to identify the geodemographics groups that might increase victimization. Furthermore, recent research by Centre for Crime and Justice Studies (CCJS) at King's College explore the link between youth culture and gun violence in disadvantaged areas in London, which give me the foundation to enhance my research, particularly in the demographics profiling. Nevertheless, the lack of studies and data in London, I would be exploring is to generate cluster using K-means with identifying variables to map the location of gang activity. Furthermore, I would use geodemographics indicators provided by Longley (2015) as a way to investigate whether the deviant lifestyles have the positive correlation with offensive weapons, in particular, the labels classified as struggling suburbs.

Lastly, from a micro level of analysis, the dissertation would employ points of interest data as a new form of analysis. A study by Wang et al. (2016) used POI as a new form of data have shown that using such categorical information of POIs are useful to profile neighborhood functions. Such neighborhood functions could further help us predict the crime rate. This demonstrates that POI data provide additional information about the communities that are not covered by the demographics. For example, the increase in weapon offences is attributable to an increased opportunity for offenders to offend in places where there are fewer police stations or public monitoring.