EXPLORING THE INFLUENCE OF ENVIRONMENTAL FEATURES ON RESIDENTIAL BURGLARY USING SPATIAL-TEMPORAL PATTERN ANALYSIS By XIAOWEN YANG A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2006
Copyright 2006 by Xiaowen Yang
iii To my parents, my husband and my son
iv ACKNOWLEDGMENTS My special thanks go to my dissertation committee members. I deeply appreciate Dr. Richard Schneider for his valuable advi ce, intellectual stimulation and guidance in the development of this dissertation. I would like to thank Dr. Zwick , Dr. Beljeri and Dr. Wu for their suggestions and comments. I also want to thank my family for their support and encouragement during the dissertation journey.
v TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iv LIST OF TABLES.............................................................................................................ix LIST OF FIGURES..........................................................................................................xii ABSTRACT.....................................................................................................................xi v CHAPTER 1 INTRODUCTION........................................................................................................1 1.1 What is Burglary?...................................................................................................1 1.2 Why Focus on Burglary?........................................................................................2 1.3 Burglary Patterns and Possible C ontributing Environm ental Factors....................3 1.4 Repeat Burglary......................................................................................................4 1.5 Urban Planning and Burglary Prevention...............................................................5 1.6 Summary and Description of Chapters...................................................................6 2 LITERATURE REVIEW.............................................................................................8 2.1 Physical Environment and Crime...........................................................................8 2.1.1 Why Should the Physical Environmen t Be Explored for Crime Prevention Research?...........................................................................................................8 2.1.2 Basic Theories about Built Environment and Crime....................................9 220.127.116.11 Defensible space...............................................................................10 18.104.22.168 Critique of defensible space.............................................................11 22.214.171.124 Crime prevention through environmental design.............................13 126.96.36.199 Critique of CPTED...........................................................................14 188.8.131.52 Situational crime preventi on and rational choice theory..................15 184.108.40.206 Routine activity theory/Life style theory.........................................18 220.127.116.11 Critique of routine activity theory....................................................20 18.104.22.168 Environmental criminology/Crime pattern theory...........................21 22.214.171.124 Displacement and Diffusion of benefits...........................................23 2.2 Repeat Victimization............................................................................................26 2.2.1 What Is Repeat Victimization and Repeat Burglary?.................................26 2.2.2 Why Does Repeat Victimization Matter?...................................................26 2.2.3 Application Projects on Repeat Victimization...........................................30 2.3 Summary...............................................................................................................31
vi 3 RESEARCH DESIGN AND FRAMEWORK...........................................................33 3.1 Purpose of the Study.............................................................................................33 3.2 Geographical Area and Data.................................................................................34 3.2.1 Geographical Area......................................................................................34 3.2.2 Data Sources...............................................................................................37 3.3 Geocoding.............................................................................................................38 3.4 Analysis Scales and Units.....................................................................................40 3.4.1 Modifiable Area Unit Problem (MAUP) and Ecological Fallacy..............40 3.4.2 The Analysis Scale Employed in This Research........................................43 3.5 Environmental Factors and Measurement............................................................44 3.5.1 Permeability................................................................................................46 3.5.2 Measurement of Permeability.....................................................................48 126.96.36.199 Distance to the closest major arteries...............................................51 188.8.131.52 Street layout patterns........................................................................52 184.108.40.206 Street types around parcels...............................................................53 220.127.116.11 Distance to public transportation stop..............................................53 18.104.22.168 Corner location.................................................................................53 22.214.171.124 Block length.....................................................................................54 126.96.36.199 Connectivity index...........................................................................55 3.5.3 Land Use and Adjacency............................................................................56 3.5.4 Environmental Variables Relating to Measurement of Land Use and Adjacency........................................................................................................58 188.8.131.52 Adjacent land use types....................................................................58 184.108.40.206 Degree of land use types mix...........................................................60 220.127.116.11 Residential units mix........................................................................61 18.104.22.168 Vacant buildings and dilapidated houses.........................................64 3.5.5 Density........................................................................................................65 3.5.6 Measurement of Density.............................................................................66 22.214.171.124 Dwelling density..............................................................................67 126.96.36.199 Built area intensity...........................................................................68 3.6 Analysis Framework and Methods.......................................................................69 3.6.1 Analysis Framework...................................................................................69 3.6.2 Analysis Methods Employed......................................................................72 188.8.131.52 Location quotient..............................................................................72 184.108.40.206 Knox test..........................................................................................73 220.127.116.11 K Nearest Neighbor test...................................................................75 18.104.22.168 Mantel test........................................................................................75 22.214.171.124 Match case control methodology.....................................................76 3.7 Summary...............................................................................................................79 4 FINDINGS..................................................................................................................80 4.1 Complete Burglary................................................................................................80 4.1.1 Spatial, Temporal, and Spatial-Temporal Analysis....................................80 126.96.36.199 Temporal analysis--Seasonal vari ation of residential burglary........81
vii 188.8.131.52 Spatial analysis--Clusters of residential burglary.............................83 184.108.40.206 Spatial-temporal analysis--Clust ers of residentia l burglary across time.........................................................................................................85 4.1.2 Environment Variable Analysis.................................................................87 220.127.116.11 Identify important social-economic variables..................................87 18.104.22.168 Environment variable analysis.........................................................89 4.2 Repeat Burglary..................................................................................................103 4.2.1 Spatial and Temporal Analysis.................................................................105 22.214.171.124 Temporal analysisâ€“time c ourse of repeat burglary........................105 126.96.36.199 Spatial analysisâ€“repeat single-family burglary in hot spots...........111 4.2.2 Environmental Variable Analysis.............................................................113 188.8.131.52 Permeability...................................................................................114 184.108.40.206 Land use and adjacency..................................................................118 220.127.116.11 Density...........................................................................................121 4.3. Near Repeat Burglary........................................................................................121 4.3.1 Spatial-temporal Analysis........................................................................123 18.104.22.168 Knox test........................................................................................123 22.214.171.124 K-Nearest neighbor test..................................................................124 126.96.36.199 Mantelâ€™s test...................................................................................125 4.3.2 Environmental Variable Analysis.............................................................125 188.8.131.52 Permeability...................................................................................126 184.108.40.206 Land use and adjacency..................................................................128 220.127.116.11 Density...........................................................................................131 4.4 Summary.............................................................................................................132 5 DISCUSSION AND CONCLUSION......................................................................136 5.1 Discussion...........................................................................................................136 5.1.1 Permeability..............................................................................................138 5.1.2 Land Use and Adjacency..........................................................................139 5.1.3 Density......................................................................................................141 5.1.4 Summary...................................................................................................142 5.2 Crime Prevention Guidance................................................................................142 5.3 Constraints on Generalizations...........................................................................144 5.4 Future Research..................................................................................................145 APPENDIX A LAND USE CODE IN ALACHUA COUNTY.......................................................152 B MONTHLY KERNEL DENSITY MAP AN D LOCATION QUOTIENT MAP FOR RESIDENTIAL BURGLARY IN GAINESVILLE, FL, 2000-2003.......................155 C SPATIAL-TEMPORAL CLUSTER ANALYSIS FOR NEAR REPEAT BURGLARY............................................................................................................160 D CODE FOR ANALYSIS..........................................................................................161
viii LIST OF REFERENCES.................................................................................................190 BIOGRAPHICAL SKETCH...........................................................................................202
ix LIST OF TABLES Table page 2.1 Connections Between Defensible Space and CPTED Strategies.............................14 2.2 Twenty-Five Techniques of Situational Prevention.................................................17 2.3 Routine Activity and Rational C hoice; Comparing and Contrasting the Approaches.........................................................................................................20 3.1 Overview of Burglary in Gainesville.......................................................................37 3.2 Overview of Repeat Burglary in Gainesville...........................................................38 3.3 Indicators of Permeability and Accessibility...........................................................49 3.4 Socioeconomic and Demographic Variables...........................................................78 4.1 Stepwise Regression Model of Soci al-economic Demographic Variables..............88 4.2 Permeability Indicators............................................................................................91 4.3 Distribution of Burglarized Site s and Control Sites among Street Layout Patterns.........................................................................................................92 4.4 Distribution of Burglarized Sites a nd Control Sites among Street Types................93 4.5 Distribution of Burglarized Sites and Control Sites Between Corners or Middle Block Lots................................................................................................................94 4.6 Land Use Type Indicators........................................................................................95 4.7 Land Use Mix Indicators..........................................................................................99 4.8 Residential Units Mix Indicators............................................................................100 4.9 Distribution of Burglarized Sites and Control Sites among Relationships With Substandard Dwelling Units...................................................................................101 4.10 Density Indicators..................................................................................................102 4.1 Repeat Single-family Burglary (N on Single-family Burglary Excluded)..............103
x 4.12 Repeat and Non-repeat Single-family Burglary by Inside/Outside Hot Spot........112 4.13 Permeability Indicators for Repeat Burglary.........................................................114 4.14 Distribution of Single Burg lary Sites and Multiple Burglary Sites among Street Layout Patterns.......................................................................................................116 4.15 Distribution of Single Burg lary Sites and Multiple Burglary Sites among Street Types......................................................................................................................117 4.16 Distribution of Single Burgla ry Sites and Multiple Burg lary Sites between Corners or Middle Block Lots.............................................................................................118 4.17 Land Use Type Indicators for Repeat Burglary.....................................................119 4.18 Land Use Mix Indicators for Repeat Burglary.......................................................120 4.19 Residential Units Mix Indi cators for Repeat Burglary...........................................120 4.20 Distribution of Single Burglary Site s and Multiple Burglary Sites among Relationships With Substandard Dwelling Units...................................................121 4.21 Density Indicators for Repeat Burglary.................................................................121 4.22 Knox Test for near repeat burglary........................................................................124 4.23 Permeability Indicators for Near-repeat Burglary..................................................126 4.24 Distribution of Non Near-repeat Burglary Sites and Near-repeat Burglary Sites among Street Layout Patterns................................................................................127 4.25 Distribution of Non Near-repeat Burglary Sites and Near-repeat Burglary Sites among Street Types................................................................................................128 4.26 Distribution of Non Near-repeat Burglary Sites and Near-repeat Burglary Sites between Corners or Middle Block Lots.................................................................128 4.27 Land Use Type Indicators for Near-repeat Burglary..............................................129 4.28 Land Use Mix Indicators for Near-repeat Burglary...............................................130 4.29. Residential Units Mix Indicat ors for Near-repeat Burglary...................................130 4.30 Distribution of Non Near-repeat Burglary Sites and Near-repeat Burglary Sites among Relationships With Substandard Dwelling Units.......................................131
xi 4.31 Density Indicators for Near-repeat Burglary..........................................................131 4.32 Environmental Features and Complete Burglary, Repeat Burglary and Near Repeat Burglary..................................................................................................................133
xii LIST OF FIGURES Figure page 3.1 Land use of Gainesville............................................................................................36 3.2 Assessed value of re sidential parcels.......................................................................36 3.3 Inconsistence between street geocoding and parcel address....................................40 3.4 Evolution of street patterns......................................................................................52 3.5 Example of street block length methodology...........................................................55 3.6 Example of street conne ctivity index methodology.................................................55 4.1 Seasonal variation of resi dential burglary in general in Gainesville, Florida..........81 4.2 Seasonal variation of residential burgl ary for apartments in Gainesville................82 4.3 Kernel density map and location quoti ent map for residential burglary in Gainesville................................................................................................................84 4.4 Isolated community with high risk...........................................................................85 4.5 Location quotient for residentia l burglaries in October, Gainesville, FL 2000-20003.....................................................................................86 4.6 Location quotient for residential burgla ries in December, Gainesville, FL 200020003........................................................................................................................87 4.7 Distribution of burglarized sites and control sites am ong street layout patterns......92 4.8 Distribution of burglarized sites a nd control sites among street types.....................94 4.9 Distribution of burglariz ed sites and control site s among relationships with substandard dwelling units.....................................................................................102 4.10 Single family repeat bur glary in Gainesville..........................................................104 4.11 Time course for single family rep eat burglary in Gainesville (Month).................107 4.12 Time course for residential burglary in Gainesville (Week)..................................108
xiii 4.14 Time course for residential propertie s, Beenleigh, June 1995 to November 1996 (inclusive ). Source: (T ownsley, Homel et al. 2000).............................................111 4.15 Repeat single-family burglary a nd hot spots of all single-family burglary incidents..................................................................................................112 4.16 Hot spots of single-family burgl ary with and without repeat.................................113 4.17 Distribution of single burglary sites and multiple burglary sites among street layout patterns...................................................................................................................115 4.18 Distribution of single burglary sites and multiple burglary sites among street patterns...................................................................................................................117 4.19 Distribution of non near-rep eat burglary sites and near -repeat burglary sites among street layout patterns...............................................................................................127 5.1 Near repeat residential burglary in Gainesville, May,2003 â€“ June, 2003..............144
xiv Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy PHYSICAL ENVIRONMENT AND RESIDENTIAL BURGLARY By Xiaowen Yang May, 2006 Chair: Richard Schneider Major Department: Urban and Regional Planning With the help of Geographic Informati on Systems and statistical tools, this dissertation intends to (a) ex plore the spatial and temporal patterns of burglary, (b) examine the correlation between burglary and environmental variable s, and (c) identify specific features of the physical environment that contribute to burglary in general and to repeat burglary and â€œnear repe at burglaryâ€ in particular . We hypothesize that some environmental variables such as accessibili ty, house location on the block, and adjacent land uses have strong contributions to bur glary, repeat burglary, and â€œnear repeatâ€ burglary propensity, despite sociodemographi c neighborhood differences. To test this hypothesis, this empirical resear ch uses a case study approach and analyzes data from the Gainesville, Florida, Police Department for residential burglaries from January 2000 to December 2003.
1 CHAPTER 1 INTRODUCTION Although there has been c onsiderable research in recent decades on the identification of features in the physical environment that may contribute to crime and burglary, few studies have considered spatio temporal correlation while using traditional statistical analysis methods. Geographi c Information Systems technology (GIS), a powerful tool in data linking and spatial analysis, provides new opportunities to analyze crime data. With the help of Geographic Info rmation Systems and statistical tools, this dissertation intends to (a) explore the spat ial and temporal patterns of burglary, (b) examine the correlation between burglary and en vironmental variables, and (c) identify specific features of the physical environment that contribute to burgl ary in general and to repeat burglary and â€œnear repeat burglaryâ€ in particular. We hypothesize that some environmental variables such as accessibility, ho use location on the block, and adjacent land uses have strong contributions to bur glary, repeat burglar y, and â€œnear repeatâ€ burglary propensity, despite sociodemographi c neighborhood differen ces. To test this hypothesis, this empirical resear ch uses a case study approach and analyzes data from the Gainesville, Florida, Police Department fo r 3,100 residential burglaries from January 2000 to December 2003. 1.1 What is Burglary? Burglary, or "breaking and entering," is one of the most common crimes worldwide. The definitions of the term burglary are not consistent among countries. In the United States, the term refers to â€œunlawfu l entry of a structure to commit a felony or
2 theftâ€ ( FBIâ€™s Uniform Crime Code , 2000). In Britain, the term burglary refers to incidents in which the offender enters or trie s to enter any building as a trespasser with the intention of committing theft, rape, grievous bodily harm, or unlawful damage. 1 In Canada, the definition of the term burglary 2 includes the elements of breaking and entering with intent, committing an offence, or breaking out. Despite the varying definitions, â€œbreakingâ€ or the â€œintention of breakingâ€ is essential to the definition of burglary. 1.2 Why Focus on Burglary? Compared to violent crime, burglary may a ppear to be a minor crime. However, the consequences of this crime are often serious. It can be very costly. For example, in 1996, the United States National Institute of Justice estimated that burglary cost 9 billion every 1 A person is guilty of burglary if a) he enters any building or part of a building as a trespasser and with intent to commit any such offence as is mentioned in subsection (2) below; or b) having entered any building or part of a building as a trespasser, he steals or attempts to steal anything in the building or that part of it or inflic ts or attempts to inflict on any person therein any grievous bodily harm. Sub section 2 The offences referred to in sub section(1) (a) above are offences of st ealing anything in the building or part of a building in question, of in flicting on any person therein any grievous bodily harm or raping any person therein, and of doing unlawful damage to the building or anything therein. Source: THEFT ACT 1968 SECTION 9(1,2), United Kingdom 2 Everyone who a) breaks and enters a place with intent to commit an indictable offence therein b) breaks and enters a place and commits an indictable offence therein, or c) breaks out of a place after i) committing an indictable offence therein, or ii) entering the place with intent to commit an indictable offence therein, is guilty
3 year, excluding the cost of pr evention and the cost of the criminal justice system (Miller, Cohen, & Wiersema, 1996). According to th e Australian Institute of Criminologyâ€™s estimation in 1992, residential and commercial break-ins cost nearly $900 million per year (Grabosky, 1995). This crime also cost s 4.3 million pounds per year to British society (Brand & Price, 2000). In addition to the socioeconomic cost, burglary, especially residential burglary, engenders significant psyc hological effects in victims. The British Crime Survey found that nearly 4 in 10 victims of "entered" burglaries said they had been very much affected; and 68% of victims of all incidents (of burglary or attempted burglary) responded that they felt angry. Shock, fear, and difficulty in sleeping were also fairly common experiences in burglary vic tims (Budd, 1999). Other studies also showed that the enduring psychological effects of burgl ary on its victims are ju st as severe as the effects related to violent crimes such as assault and robbery (Hough, 1984). 1.3 Burglary Patterns and Possible Co ntributing Environmental Factors To fight this crime, it is very important to understand burglary patterns and possible contributing factors. The spatial distri bution of burglary is not random. Some communities have few burglaries, whereas others are notorious for a high concentration of burglaries. The distribution of crime ove r time is not random, either. For example, analysis of residen tial burglary in the London Borough of Croydon revealed that peaks existed in residential burglary temporal pa tterns (Crime Reduction Center, n.d.). Thus, if burglaries at high-density places and peak times can be prevented, then the total burglary rate can be reduced. With knowledge of where and when burgl aries occur, other factors that are associated with burglary shoul d be explored. Crime, includ ing burglary, is related to a multitude of attributes. Earlier research th at focused on socioeconomic variations in
4 crime has examined the effect on crime of variables such as poverty, ethnicity, age composition, income, education, gender, and residency expressed by percentage of foreign-born residents in a c ity (Byrne & Sampson, 1986). Othe r researchers have argued that the crime rates for communities that share common socioeconomic characteristics can be different because of the differences in environmental features (Newman, 1972). The second argument represents another trend in crime research, which shifts attention away from the motivation of criminals and toward crime events. Without disputing the importance of socioeconomic variables, e nvironmental criminologists argue that the analysis of physical environment variables (land uses, lighting, th e design of landscape and space) â€œis promising because, once understood, available technologies can be used to modify these patterns and abate some crim es without doing significant damage to basic human rightsâ€ (Brantingham & Brantingham, 1981, p. 4). The examination of the association between specific crimes and ch aracteristics of the physical environment would improve the pool of knowledge and tech nology to prevent crime, but should not play the role of supplanting, pr ecluding, or derogating social and psychological research. 1.4 Repeat Burglary Repeat victimization, which means the sa me person or place suffers from more than one crime incident, is an important issue in crime analysis for several reasons First, there is considerable potential for reducing overall crime rates by reducing or eliminating repeat victimization. According to the Br itish Crime Survey (2000), 4% of victims account for between 38% and 44% of all crime reported to the survey (Crime Reduction Center, n.d.). For burglaries, â€œa growing body of research has revealed that in highburglary neighborhoods most resi dences have no burglaries, but a few residences suffer repeated burglariesâ€ (Eck, 1997, p. 1). This finding of repeat burglaries in a few
5 residences has been replicated worldwide, in England, Australia, Canada, and the United States (Farrell & Pease, 1993; Forrester, Chatterton, & P ease, 1988; Forrester, Frenz, O'Connell, & Pease, 1990; Polvi, Looma n, Humphries, & Pease, 1990; Townsley, Homel, & Chaseling, 2000). The data gathered in this research for the City of Gainesville, Florida case study tend to suppor t this conclusion. For example, in Gainesville, repeat burglaries accounted for 41.19% of all residential burglary incidents in the years 2000-2003. If all repeat burglary co uld be eliminated, the overall residential burglary rate in Gainesvill e could be reduced by 28.23%. Th e considerable potential for reduced burglary rates related to repeat burglaries warrants research attention. 1.5 Urban Planning and Burglary Prevention Of all crime types, burglary has arguably the strongest envi ronmental component. Crimes that are attacks against aspects of the physical environment are much more likely to be deterred or encouraged by environmenta l features than are violent crimes against people. As a crime against the physical envi ronment--real property --residential burglary is therefore likely to exhibit a strong rela tionship with physical environment features. The physical environment is a realm th at urban planners and designers can influence because of their expertise a nd responsibilities. Through comprehensive planning, subdivision regulations , landscaping ordinances, as well as building and design regulations and guidelines, pla nners and designers have the potential to mitigate physical variables that contribute to crime and enhance the elements that reduce crime. This leads to safer places for citizens. At the same ti me, â€œsafeguarding the public health, safety, and general welfareâ€ (Schneider & Kitchen, 2002, p. 7) is one of obligati ons of planning and urban design professionals. However, they â€œhav e taken much less of a lead in the area of crime prevention than their public interest missions w ould suggestâ€ (Schneider &
6 Kitchen, 2002, p. 23). This may be caused partly by the lack of str ong empirical evidence that environment variables infl uence crimes, including burglary. Through analyzing burglary data in Gaines ville, Florida, this research aims to explore burglaryâ€™s spatial-tem poral pattern and identify physi cal features that correlate with burglary, while controlling sociodemographic variables. Th e result of this analysis can be meaningful in a variety of ways. In a very practical sense, knowledge about where and when burglary happens could help to tactically deploy lim ited police and crime prevention resources. Furthermore, understa nding how different environmental design features affect the incidence of burglary w ould help building, re gulatory, and planning agencies; law enforcement departments; and ad ministrators take proactive measures that decrease overall criminal activ ities. Moreover, this analysis could provide evidence to support a number of the so-called â€œplace-ba sed crime prevention theories,â€ including defensible space, crime prevention through environmental design (CPTED), environmental criminology, routine activity theory, situational crime prevention, and rational choice theory (Sc hneider & Kitchen, 2002). 1.6 Summary and Description of Chapters With the help of Geographic Information Systems and statistical tools, this dissertation will analyze three residential burgl ary patterns: complete residential burglary, repeat single family residential burglary, a nd near repeat residential burglary. For each burglary pattern, two categories of analysis--s patial-temporal analysis and environment variable analysis--are applied. With this anal ytic focus, this disse rtation intends to (a) explore the spatial and temporal patterns of residential burgl ary, (b) examine the correlation between burglary and environmental variables, and (c) identify specific features of the physical environm ent that contribute to burglary in general and to repeat
7 burglary and â€œnear repeat burgl aryâ€ in particular. The resu lts of these analyses should help us better understand resi dential burglary patterns in the present case study, and may lend credence to basic principles of place-b ased crime prevention theory generally. As we have discussed above, Chapter 1 o ffers an overall descri ption of the scope and purpose of this research. Chapter 2 introduce s relevant theories and research from the literature. Chapter 3 describes the methodol ogy used for this exploratory research. Chapter 4 presents a discussion of the anal ytical findings and their results. Chapter 5 discusses these results and provides conc lusions and recommendations for further research and to planners and law enforcement agencies.
8 CHAPTER 2 LITERATURE REVIEW This chapter provides an overview of th eories about the linkages between the physical environment and crime, specifically burglary and repeat victimization. It discusses why it is important for research on crime and crime prevention to explore these two issues. It also summarizes basic theories and international resear ch on these topics. In the following chapter, discussions of these th eories and related research will make the importance of relevant environm ental variables demonstrable. 2.1 Physical Environment and Crime Causes of crime are complicated, and so are strategies of crime treatment and crime prevention. A range of factor s--including social, economic , demographic, biological, psychological, and physical variable--influence the occurrence of crime. Because features of the physical environment influence human be havior, it should come as no surprise that attributes of the physical environment infl uence criminal behavior. Research that correlates crime with features of the envir onment grew significantly during the past few decades. This area of study has attracted scho lars from disciplines such as criminology, planning, geography, environmental psychology, and architecture. 2.1.1 Why Should the Physical Environment Be Explored for Crime Prevention Research? Traditional criminological research focuse d on the origins of criminal motivation, analyzing the social, economic, biological , and psychological infl uences that were thought to contribute to crime. For exampl e, innate criminal theory (Lombroso 1876)
9 asked what biological features motivated in dividuals to commit crimes, whereas social disorganization theory (Shaw and Mc Kay 1931) focused on the conditions in neighborhoods that motivated individuals to commit crimes. The central feature of these social, economic, and biological theories of crime concerned the origins of offender motivation. However, as the Brantingha ms suggest (Brantingham and Brantingham 1981), a crime is composed of at least five di mensions: a law, an o ffender, a target, and a time and place. The motivational approach ne glects three of the dimensions of the criminal event. Jeffery also suggests (1971; 1977) that the shift in the focus of research-from the motives of offenders to criminal events and their settings--may prove more useful when the objective of analysis is crime rather than â€œcriminalityâ€ and when the policy objective is crime control rather than â€œoffenderâ€ control. 2.1.2 Basic Theories about Built Environment and Crime Over the past 40 years, research has begun to examine the links between the physical characteristics of crime sites and the choices criminals make when selecting such sites. The purpose of this research has b een to suggest measures to be taken so that crime can be reduced. The most prominent theori es that have come out of this research are those concerning defensible space, cr ime prevention through environmental design (CPTED) and environmental criminology (which focus on physical changes most directly), routine activ ity theory, situational crime prevention, and rational choice theory (which focus largely on offender behavior rela tive to crime "settings "). Despite the fact that these theories have ri sen out of different disciplin es, they have become woven together through time.
10 18.104.22.168 Defensible space The major founder of defensible space theory, Oscar Newman observed public housing and crime in New York City, findi ng a relationship between urban design and crime rates. He suggested that features built into the environment th at act as territorial displays and provide natural surveillance oppor tunities increase residentsâ€™ concern for their living space. â€œThe potential criminal perceives such a space as controlled by its residents, leaving an intr uder easily recognized and d ealt withâ€ (Newman 1972, p. 3). Consequently, property in these spaces may be less vulnerable to crime such as residential burglary than it is in more randomly designed spaces. The fundamental principles of this theory include territor iality, natural surveillance, image, and milieu. Territoriality. Territoriality is â€œthe capacity of the physical environment to create perceived zones of territori al influenceâ€ (Newman 1972, p. 51). Physical space should be subdivided and identifiable. Then it can pr ovide both resident and outsider with a perceptible statement of individual and group control over spaces. The notion of territoriality does not require that residents be owners. Th rough territorially sensitive design, tenants can be spurred to protect and defend outdoor spaces adjacent to apartments they live in but do not own. Natural surveillance. This is â€œthe capacity of physical design to provide surveillance opportunities for residents and their agents â€ (Newman 1972, p. 78). In other words, design can facilitate the visual and a uditory monitoring abil ities of residents and other â€œlegitimateâ€ users of interior and exte rior spaces. However, the ability to observe criminal activity will not automatically impe l the observer to respond with assistance to the person or property being victimized. Excuses of avoiding engagement because the observer does not know the victim--or because the event happened on a public street--
11 preclude intervention, in some cases. The e ffectiveness of surveillance depends on â€œwhether the area under surveillance is id entified by the observer as falling under his sphere of influenceâ€ (Newman 1972, p. 79). Image and milieu. Image and milieu can be defined as â€œthe capacity of design to influence the perception of a projectâ€™s uni queness, isolation and stigmaâ€ (Newman 1972, p. 102). For Newman, this involves mitigati ng the stigma of public housing and the juxtaposition of land uses. There are two more concepts that were advanced later and mentioned frequently: access control and boundary definition . Boundary definition m eans clearly dividing and marking spaces to identify the shadings between public and private use. Access control means impeding the movements of poten tial offenders and help ing alert residents to their presence. Access control and boundary definition are closely related to the concept of territoriality and natural surveillanc e, both in theoretical terms and in practical applications. 22.214.171.124 Critique of defensible space In spite of its popularity, defensible sp ace has been criticized concerning its methodology, research techniques, and practical application. On e major criticism is that in analyzing data, Newmanâ€™s theory syst ematically understates and marginalizes socioeconomic and demographic factors. Davidson (1981, p. 84) argues that â€œby prematurely dismissing social factors and by concentrati ng on physical factors,â€ the theory of defensible space â€œmight obscure the importance of other factors which might nullify attempts to make use of the theory to control crime.â€ In defense of Newman, he does recognize the social founda tion of crime, by attempting to locate truly comparable areas where socioeconomic and demographic factors are controlled in his research.
12 However, it is highly complicated to find such â€œtwinâ€ localities in real life. Armitage (1999), in her research, has highlighted the difficulties for this issue. Indeed, it is questionable whether truly comparable areas can be found at all (Coz ens, Hillier et al. 2001). There are also concerns about the applicability of defens ible space principles in practice. MacDonald and Gifford (1989, p. 194) claimed that â€œthere remains a lack of evidence that burglars are deterred by terr itorial displays.'' Territoriality was not supported in the study by Phelan (1977) and those by Bennett and Wright (1984), either. However, Brown and Bentley (1993) found territo riality to be an im portant factor in burglars' decision-making processes. Booth (1981), also, found territoriality to be of limited utility. Furthermore, although Newman (1972) talked about th e interaction of all the principles, there was no data analysis rele vant to interactions between the variables. Another important issue is â€œthe possibili ty that these elements might contain contradictions within them selvesâ€ (Mawby 1977a, p. 176), a poi nt that Newman did not deal with. In this sense, it is possible that each category in his theory may contain some dimension that simultaneously decreases and enhances security. For example, increasing the number of people in an area may increase the number of possi ble witnesses, but it may also increase the number of potential offenders and victims. Despite all these criticisms, Newmanâ€™s theo ry still has had great appeal. Yet, these criticisms may suggest that more thorough a nd rigorous reexaminati on of the concept of â€œdefensible spaceâ€ is necessary. â€œDesigning de fensible space is neither the panacea that some proponents have hoped, nor is it as irrelevant to crime and fear as some detractors have contended'' (Krupat and Kubzansky 1987). Just as all other crime prevention
13 strategies, defensible space can produce variable levels of effectiveness, but cannot cure all problems. 126.96.36.199 Crime prevention through environmental design Crime prevention through environmental de sign (CPTED), as distinct from the architectural context of defensible space, is the brainchild of C. Ray Jeffery, a criminologist. The phrase began to gain acceptance after the publication of his 1971 book, which used the same phrase as its ti tle. In his 1977 revision, Jeffery totally redesigned his theoretical mode l for crime prevention. In this edition Jeffery developed a bioenvironmental model wherein the physical en vironment interacts with the organism by means of the brain. This model was more fully developed in Jefferyâ€™s Criminology: An Interdisciplinary Approach (1990). The basic assumption of Jefferyâ€™s CPTED model is: The response [i.e., behavioral adaptation] of the individual organism to the physical environment is a product of the brain; the brai n in turn is a product of genetics and the environment. The environment never influen ces behavior directly, but only through the brain. Any model of crime prevention must include bo th the brain and the physical environment. (Jeffery and Zahm 1993, p. 330) There are two critical elements in this model. First, the physical environment can be changed or managed to produce be havioral effects that will reduce the incidence and fear of crime. This is the part of the theory that most people th ink about. It is also what we refer to in this paper as CPTED. The second element of CPTED is the individual organism , or human brain, which is the basis of the biomedical approach to crime prevention. However, this biomedical appr oach to crime prevention has not been subjected to widespread empiri cal testing because it is hard to depict brain activity during criminal behavior. Most CPTED research has been conducted with regard to the effects of modifications to the physical envi ronment (Paulsen and Robinson 2004).
14 Although the academic origin of CPTED di ffers from defensible space, some fundamental principles of CPTED are nearly id entical to those of defensible space. They are surveillance, boundary definition, access c ontrol, the relationship between land use and activity locations, and terr itoriality. Subsequent research ers expanded this theory, and some nonspatial elements were incorporated. For example, Crowe (1997) identifies nine major CPTED strategies th at include scheduling and communication. The comparison between defensible space and CPTED strategi es is listed in the following table. Table 2.1 Connections Between Defens ible Space and CPTED Strategies Defensible space principles (Newman) CPTED strategies (Crowe) Territoriality boundary definition Border definition of controlled space Territoriality boundary definition access control Clearly marked transitional zones Surveillance access control Attent ion directed to gathering area Image and milieu: activity generation Place safe activities in unsafe areas Image and milieu: activity generation Place unsafe activities in safe locations Boundary definition access control Reduce use conflicts with natural barriers None Better scheduling of space Surveillance Increase percep tion of natural surveillance in spaces by design None Overcome distance and isolation by communication Source: (Schneider and Kitchen 2002) 188.8.131.52 Critique of CPTED Although the conception of th e concept CPTED is broader than that of defensible space, it attracted less attention in 1970s. Pa rtly, this may have been because American criminology was unreceptive to the genetic explanation of behavior (Clarke 1997). Furthermore, Jeffrey did not provide detailed recipes for how to reduce crime in his 1971 original work, although the government and publ ic were looking for those (Paulsen and Robinson 2004) By contrast, Newmanâ€™s book in 1972 contained specific suggestions for crime prevention in public housing facilities, such as increasing lobby visibility, altering
15 entrance design and lowering building height . Moreover, some projects conducted under CPTED did not prove to be very effective (Clarke 1997). One example of the unsuccessful projects discussed by Clarke is the Westinghouse project, in which Westinghouse Electric Corpor ation designed (a) a school crime prevention project in Broward County, Florida; (b) a commercial cr ime prevention project in Portland, Oregon; and (c) a residential and mixed-use crime pr evention project in Ha rtford, Connecticut. Named CPTED projects, they were based on Newman's ideas rather than Jeffery's (Jeffery 1977, p. 225). This research produced unconvincing results, â€œperhaps because â€˜territorialâ€™ behavior is le ss natural outside of resident ial settingsâ€ (Clarke 1997, p. 8). There are also some empirical projects th at support the conten tion that the physical environment influences crime. Hunter and Jeffery (1991) undertook a statewide survey of 110 convenience stores in Florida and found th at the presence of (a)concealed access, (b)gas pumps, and (c)cash handling show a statistically significant correlation with convenience store robbery. Because of this rese arch, the cities of Ga inesville, Florida, and Kent, Ohio, passed ordinances about c onvenience stores. These ordinances were aimed at reducing robberies. 184.108.40.206 Situational crime prevention and rational choice theory In 1968, Gary Becker, 1992 Nobel Prize Laureate in Economics, published his seminal work on the rational behavior of crim inals, a work in which he claimed that a criminal evaluates costs and benefits not only in choosing his li festyle but also in deciding whether or not to car ry out a particular crime. However, this economic model has been criticized by Clarke as â€œseriously limited in a num ber of respectsâ€ (Clarke and Felson 1993,p. 5), such as ignoring noncash e quivalent rewards. C onsistent with the rational perspective of crime, Clarke modified the model in which â€œrelationships between
16 concepts are expressed not in mathematical te rms but in the form of â€˜decisionâ€™ diagramsâ€ (Clarke and Felson 1993, p. 9). The model s uggests that offendersâ€™ rationales are â€œlimitedâ€ and less perfect than are those of â€œeconomic man.â€ Rational choice theory posits that criminals make decisions to â€œcomm it crime in specific situations based upon expenditure of effort , balanced by risk factors and expected rewards â€ (Schneider and Kitchen 2002, p. 106). Rational choice theory is a vital supporting element of situational crime prevention, a fundamentally â€œtacticalâ€ appr oach. Situational crime preven tion, which is also set forth by Clarke, is aimed primarily at eliminating opportunities for crime. It is composed of four main elements: (a) an articulated th eoretical framework, (b) a standard methodology for tackling specific crime problems, (c) a se t of opportunity-reducing techniques, and (d) a body of evaluated practice (C larke 1997). The theoretical framework is informed by a variety of "opportunity" theo ries, including the rational ch oice perspective, routine activity and environmental criminology. The standard methodology is a version of the action research paradigm, in wh ich researchers and practitioners work together to analyze and define the problem, to identify and implem ent possible solutions, and to evaluate the results (Clarke 1997). The opportunity-reducing techniques range from simple target hardening to more sophisticated methods of deflecting offenders and reducing inducements (Clarke 1997). Some of thes e techniques extend be yond physical features, such as credit card control and drinking-age laws (see Table 2.2). The body of evaluated practice aims to help practitioners find out â€œwhich measures work best, in which combination, deployed against what kinds of crime and under what conditionsâ€ (Clarke
17 1997, p. 28). It is important to realize that situational measures do not always work in intended ways and measures that work in one setting may not do so in another. Table 2.2 Twenty-Five Technique s of Situational Prevention Increase the Effort Increase the Risks Reduce the Rewards Reduce Provocations Remove Excuses 1. Target harden: Steering column locks and immobilisers Antirobbery screens Tamper-proof packaging 6 Extend guardship: Take routine precautions: go our in group at night, leave signs of occupancy, carry phone â€œCocoonâ€ neighborhood watch 11. Conceal targets: Off-street parking Gender-neural phone directories Unmarked bullion trucks 16. Reduce f rustrations and s tress: Efficient queues and polite service Expanded seating Soothing music/muted lights 21 Set Rules: Rental agreements Harassment codes Hotel registration 2. Control access to f acilities: Entry phones Electronic card access Baggage screening 7. Assist natural surveillance: Improved street lighting Defensible space design Support whistleblowers 12 Remove targets: Removable car radio Womenâ€™s refuges Prepaid cards for pay phones 17. Avoid disputes: Separate enclosures for rival soccer fans Reduce crowding in pubs Fixed cab fares 22 Post instructions: â€œNo parkingâ€ â€œPrivate propertyâ€ â€œExtinguish camp fireâ€ 3. Screen exits: Ticket needed for exit Export documents Electronic merchandise tags 8 Reduce anonymity: Taxi driver Ids â€œHowâ€™s your driving?â€ decals School uniforms 13. Identify p roperty: Property marking Vehicle licensing and parts marking Cattle branding 18. Reduce emotional arousal: Controls on violent pornography Enforce good behavior on soccer field Prohibit racial slurs 23 Alert conscience: Roadside speed display signs Signatures for customers declarations â€œShoplifting is stealingâ€ 3. Deflect offenders: Street closures Separate bathroom for woman Disperse pubs 9 Utilize place managers: CCTV for doubledeck buses Two clerks for convenience stores Reward vigilance 14 Disrupt markets: Monitor pawn shops Controls on classified ads License street vendors 19. Neutralize peer p ressure: â€œIdiots drink and driveâ€ â€œItâ€™s OK to say Noâ€ Disperse troublemakers at school 24 Assist compliance: Easy library checkout Public lavatories Litter bins 5. Control tools/weapons: â€œSmartâ€ guns Disabling stolen cell phones Restrict spray paint sales to juveniles 10 Strengthen formal surveillance: Red light camera Burglar alarms Security guards 15 Deny benefits: Ink merchandise tags Graffiti cleaning Speed humps 20 Discourage imitation: Rapid repair of vandalism V-chips in TVs Censor details o f modus operandi 25 Control drugs and alcohol: Breathalyszers in pubs Server intervention Alcohol-free events
18 Most criticism about situati onal crime prevention relates to displacement. This is an issue to which all environment relevant crim e prevention approaches have to respond. It is discussed below. 220.127.116.11 Routine activity theory/Life style theory Cohen and Felson produced their routine activities theory in 1979. The phrase routine activity includes â€œany recurrent and prevalent activities which provide for basic population and individual needsâ€ (Cohen a nd Felson 1979, p. 593). This theory focuses on when and where people are, what their activ ities are, and what happens to them as a result (Clarke and Felson 1993) . This emphasis reflects its in tellectual roots, which are found in the human ecology theory of Amos Hawley (Hawley 1950). It also echoes lifestyle theory (Hindelang, Gottfredson et al. 1977), which recognizes that the differential risks of victimization are partly a function of the victim â€™s lifestyle, including both work and leisure activities, which vary potential victimsâ€™ exposure to potential offenders. The routine activity approach assumes that crime results from a convergence in time and space of three minimal elements: a likely offender , a suitable target , and the absence of a capable guardian against crime (Cohen and Felson 1979). A guardian is a person whose presence or proximity would pr otect a target or discourage crime from happening. Although the category of guardian includes formal au thorities such as police officers or security guards, in most cases the typical guardians are ordinary citizens, such as neighbors, coworkers, friends, relatives, or property owners. Felson and Clark (1998) put forth four el ements, designated by the acronym VIVA, that influence a targetâ€™s risk of being victimized by crime. They are value , inertia , visibility , and access . All four of these dimensions are considered from the offenderâ€™s
19 viewpoint. Value means that the ta rget must be rewarding. Inerti a is the ability of a target to be moved. Visibility refers to the expos ure of targets to offenders. Access involves all those environmental and situational features that may facilitate offenders getting to the target. The authors believe their approach explai ns the rise in burgl ary between 1960 and 1980. In this period, far more homes were left unguarded in daytime, as more women worked full time. The absence of a guardian during the workday le d to the increased probability of criminal activity. At the same time, an increase in the number of portable electronic goods in people's home--such as televisions and video cassette recorders-provided more suitable targets for burglary. Numerous studies have shown relation ships between the daily activities of individuals and their likeli hood of criminal victimization (Riley 1987). For example, Miethe et al. (1987) found that persons with low levels of daytime and nighttime activity outside of the home have the lowest risk of property victimization. So far the theory of routine activity/life style can explain the conditions that must apply if an offence is to be committed. In practice, it can be used as a foundation for crime prevention. Because an offender, a targ et, and the absence of a capable guardian are needed for a crime to happen, as long as one of the preconditions is removed, the the theory argues that the offence can be prev ented from happening. So, removing valuable items from visibility or providing a capable guardian may be useful methods to prevent crime. Furthermore, the theory can also be a pplied as a risk analysis tool in determining who and which locations are more likely to be victimized. Potential offenders, like ordinary people, â€œhave day-to-day schedules (routine activity) â€“ trips to and from work,
20 visiting friends, going shopping â€“ a nd that in the course of su ch travels they search out likely targetsâ€ (Schneider and Kitchen 2002, p. 107). The design and location of road networks and mixtures of residential a nd nonresidential land uses may influence a potential burglarâ€™s daily movement pattern s, change the proximity of targets and offenders, and therefore enhance or reduce the victimization risk. As we have stated previously, despite th e different intellectual roots of place focus crime prevention theories, they weave together through time. Rational choice theory and routine activity theory, â€œt hough differing in scope and purpose, are compatible, and indeed, mutually supportiveâ€ (Clarke and Felson 1993). Table 2.3 Routine Activity and Rational Choice; Comparing and Contrasting the Approaches Routine Activity Rational Choice Organizing perspective Yes Yes Explanatory focus: Criminal event Criminal dispositions Yes No Yes Yes Level of explanation Macro Micro Causal theory Yes No Situational focus Yes Yes Crime specific Yes Yes Rational offender Implicit Explicit Policy orientation Implicit Explicit Disciplinary parentage Geography, demography, Human ecology Psychology, economics, sociology of deviance, environmental criminology Resource: (Clarke and Felson 1993) 18.104.22.168 Critique of routine activity theory Although routine activity and lif estyle theories are very popular, most research of these theories is limited in several respects. One major problem is that very few studies directly measure the key concepts of thes e theories. Cohen and Cantor (1980, p. 145) concluded that â€œmany of the data needed to operationalize and rigorous ly test the routine
21 activities approach to criminal victimization ar e not yet available.â€ These data are still not widely available today. As a re sult, demographic and socioec onomic variables are used as proxy indicators for the key concepts of routin e activity theory. In fa ct, in the original formulation paper of routine ac tivity theory, rather than inde pendent measures of routine activities and lifes tyles (e.g., amount of time spent in or outside the home, frequency of nighttime activity outside the home), C ohen and Felson (1979) used household composition and employment status to explai n the national crime trend. Without separate and independent measures of key concepts of th ese theories, it is ha rd to distinguish the influence of routine activities from other factors such as proximity to a high crime district. Itâ€™s even harder to examine to what extent routine activities influence criminal victimization. 22.214.171.124 Environmental criminology /Crime pattern theory Environmental criminology claims that cr ime happens when five things come together: a law , an offender , a victim or target , at a time and place (Brantingham and Brantingham 1981, pg.7). The physical components of this theory are largely developed out of CPTED and defensible space th eories (Brantingham and Brantingham 1981,pg.18), with more focus on â€œgeographicâ€ elemen ts rather than â€œdesignâ€ elements of places. Crime pattern theory evolves from environment criminology, but incorporates more theories, such as routine activity theo ry, rational choice theor y, strategic analysis, life-style theory, situ ational crime prevention, hot spot analysis, and oppo rtunity theory. It explores patterns of where, when, and how crimes occur. Crime pattern theory has three main concepts: nodes , paths , and edges . Nodes , a term from the field of transportation, refers both to points of depart ure and destinations (e.g., home, school, entertainment area). Such pl aces not only can generate crime within
22 their own locus, but also nearby. The term paths refers the main areas of travel between nodes, such as streets and pub lic transportation systems. As people travel along these paths from activity nodes with some regular ity, the "paths and narrow areas surrounding them become known spaces to the peopl e who travel themâ€ (Brantingham and Brantingham 1998, p. 35). Finally, â€œedgesâ€ are the â€œp laces where there is enough distinctiveness from one part to another that the change is noticeableâ€ (p. 17). Because people from different neighbor hoods do not know each other, e dges are more likely to provide anonymity, and therefore â€œedges cons titute areas that experience high crime ratesâ€ (p. 18). According to Brantingham and Brantingham, four elements are important to understand crime patterns. They are (a) even t process, (b) crime template/activity backcloth, (c) readiness/willingness, and (d) inte raction of all the former elements. In the case of burglary, the event process can be desc ribed, such that a potential burglar is â€œtriggeredâ€ by an event to commit the offense. The trigger event may be that the potential criminal is short of money or being pushe d by friends. This trigger event leads the offender to search for â€œsuitableâ€ or â€œgoodâ€ ta rgets. The search acti on â€œrests on a general backcloth formed by routine activities and on a template that helps identify what a â€˜greatâ€™ chance is or what a â€˜goodâ€™ opportunity woul d be or how to search for chances and opportunitiesâ€ (Brantingham and Brantingha m 1993, pg.268). By engaging in his or her daily non-criminal, routine activities, the triggered potential burglar develops an â€œawareness space.â€ In accordance with his or her idealized crime template, his or her target will be searched in the awareness sp ace. When the potential burglar finds such a target, he will commit the burglary. It is importa nt to realize that th e triggered event, the
23 probable crime template, the activity back cloth, and the criminal readiness are interrelated. The objectiv e of crime pattern theory and environmental criminology is to explore aggregate patterns. 126.96.36.199 Displacement and Diffusion of benefits Place-focused crime prevention efforts may cau se two side effects: displacement of crime and diffusion of prevention benefits. Th e former is discussed widely, as it can seriously undermine crime prevention motivatio ns. However, there is little reason to believe that the effect of displacement is gr eater or even close to the reduction of crime prevention efforts. Furthermore, the second si de effect of crime prevention â€“ diffusion of benefits--can enhance prevention. Displacement is one of the most serious cri ticisms leveled against all crime prevention theories. The displacement argument claims that crime prevention strategies do not solve the problem of crime, but merely displace it to some other targets, times, places, tactics, or even other categories of crime, etc. (Reppetto 1976). In other words, crime prevention strategies would produce no ne t reduction in crime and only be a wasted effort for society as a whole (Clarke and Felson 1993, pg. 4). The antic ipation of â€œtotal displacementâ€ could seriously undermine crim e prevention motivations. Actually, several researchers recorded encounters with this attitude. Town (Town 2002, Displacement and â€˜Common Senseâ€™, para. 4) re ported one officerâ€™s comment in 1996, â€œWhatâ€™s the point? They will get in anyway, and ev en if they donâ€™t, they will just go somewhere else.â€ Barr and Pease (1990) also documented the same issue. The widespread acceptance of â€œtotal displa cementâ€ is based on the assumption that offenders must commit crime as long as thei r motivations havenâ€™t been changed. Because none of the environment-related crime theori es discussed above emphasize manipulating
24 criminal motivations, their effectiveness in reducing crime in total is doubted. If burglaries in one area are redu ced by application of these theories, displacement, the assumption that the burglaries may simply m ove to other areas, is proposed as the first alternative account. However, th ere is no reason to assume mo st offenders are so driven by motivations that they have to maintain a certain level of offendi ng whatever the cost (Clarke 1997). Felson and Clarke (Felson and Clarke 1998) argued that displacement theory gives far too little importance to the causal role of opportunity. When easy opportunities for crime are reduced, most potentia l offenders, who view criminal life as a relatively easy way to make a living, will commit fewer offences and even be encouraged to explore non-criminal alternat ives (Felson and Clarke 1998). In addition to theoretical explorations that find no basis for believing â€œtotal displacement,â€ empirical evidence also finds little support for this assumption. Over the past decades there have been several empi rical reviews of the displacement phenomenon. Studies in the United States, Canada, Great Britain, continental Europe, and Australia have found that there is often no displacement, but that when displacement occurs it is far from complete (Cornish a nd Clarke 1987; Barr and Pease 1990; Eck 1993; Hesseling 1995; Bowers and Johnson 2003). To sum up, there are neither theoretica l reasons nor empirical evidence for believing â€œcomplete displacement.â€ Crime prevention initiatives can produce very substantial net reductions in crime, and comm only with very little or no displacement. As Eck (Eck 1997, p. 7-49) argued, â€œConcern abou t displacement is usually based more on pessimism than empirical fact.â€
25 Diffusion of benefits is the opposite of displaceme nt. Contrary to the results theorized by displacement theory, the risk of victimization for potential â€œuntreatedâ€ targets that are within close proximity to the â€œtreatedâ€ target area--o r for targets that are similar to them in some way--can be redu ced because of crime prevention actions. For example, a burglary reduction project app lied at a target area may also cause the surrounding areaâ€™s burglary level to be reduce d. It may also result in lower crime rates overall in the target area, as th e surveillance leve l is improved. This effect is also called the â€œfree riderâ€ effect (M iethe 1991), the â€œhalo effectâ€ (Scherdin 1986), the â€œbonusâ€ effect (Sherman 1990), the â€œspilloverâ€ effect (Clarke 1989), and the â€œmultiplierâ€ effect (Chaiken and St evenson 1974). Despite the variety of names, they all refer to the same phenomenon Clarke and Weisburd defined as â€œthe spread of the beneficial influence of an intervention be yond the places which are di rectly targeted, the individuals who are the subject of control, the crimes which are the focus of intervention or the time periods in whic h an intervention is brought .â€(Clarke and Weisburd 1994, p. 169). The realization of diffusi on of benefits can bring cons idered added value to crime prevention strategies. As with crime displacement, there are many forms of diffusion of benefits. It is difficult to make definitive measurements of both phenomena. There is less research for diffusion of benefits than for displacemen t; the evidence for the former is weaker. However, exploration of both phenomena pr ovides an opportunity to maximize crime prevention benefits. By understanding the reasoning and processes that lead to both phenomena, crime prevention programs coul d be more clearly designed to enhance
26 diffusion of benefits and reduce crime di splacement and the debate about crime prevention strategies and effect s can be better balanced. We move now from a discussion of placebased (environmental) crime prevention theories involved to a focus on the crimes th emselves, and specifially repeat victimization and repeat burglaries. 2.2 Repeat Victimization 2.2.1 What Is Repeat Victimization and Repeat Burglary? Repeat victimization occurs â€œwhen the same person or place suffers from more than one incident over a specified period of timeâ€ (Bridgeman and Hobbs 1997). A repeat burglary is defined in terms of locus: wh erever there has been a previous burglary reported at the same location. For the operationa l definition used by most police forces in the United Kingdom, the temporal span is li mited to a period of 12 months (Farrell, Edmunds et al. 2000). Definitions in other coun triesâ€™ research, such as Australia and New Zealand, are mostly consistent with this one . However, the 1-year definition is used mostly for reasons of convenience, not princi ple. Pease (1998) argue d that the definition should vary with the crime reduction purpose, ot herwise the strengths of the approach are lost if it is applied mechanically. In our pr esent research, the definition of time span for repeat burglary is consistent with the data c overage, which is four years. With the longer definition of time span, we have more opportunity to find varied patterns. 2.2.2 Why Does Repeat Victimization Matter? The extent and significance of repeat vi ctimization has only recently been widely recognized. This concept has gripped the criminological im agination so much that Skogan observed that â€œprobably the most impor tant criminological insight of the decade has been the discovery in a very systematic fashion of repeat multiple victimization. This
27 has tremendous implications both for crim inological theory a nd. . .practice in the fieldâ€(Brady 1996, p. 3). Repeat victimization has become a significan t issue for several reasons. First, there is considerable potential for reducing overa ll crime rates by reducing or eliminating repeat victimization. According to the Brit ish Crime Survey, conducted in the year 2000, it has been estimated that 4% of victims acc ount for 38-44% of all crime reported to the survey (Kershaw, Budd et al. 2000). Because repeat incidents comprise a large proportion of overall crime, if they can be prevented, th en a significant proporti on of all offenses can be prevented. As we shall see, the data ga thered in this research for the City of Gainesville case study tend to support this conclusion. For ex ample, in Gainesville, repeat burglaries accounted for 41.19% of all re sidential burglary in cidents in the years 2000-2003. If all repeat burglaries could be el iminated, the overall residential burglary rate in Gainesville could be reduced by 28.23%. Second, repeat victimization is preventable. Victims are not equa lly prone to repeat victimizatio n; change in a target after a crime (for example, taking simple defe nsive measures like strengthening doors or window openings) diminishes the chances of its repetition. Targets victimized are not equally likely to suffer again. The likelihood of repeat victimization may be influenced by environmental and individual charact eristics. As Pease (1998) argued, Key reasons for repeats are believed to be the presence of good, and lack of bad, consequences of the first crime for the offender, and the stability of the situation which presents itself to an offender on the first and subsequent visits to the scene of his or her crime (Pease 1998, p. 6). Meanwhile, it is especially important to recognize, as suggested above, that changes in a target after a crime diminish th e chances of its repeti tion. This recognition is
28 consistent with the broken windows theory (K elling and Coles 1996), in which neglect of the first attack on a building or person may imply to an offender that no one cares, and that the attacks can continue with impunity. Third, knowledge of re-victimization patterns can direct limited crime-fighting and crime-pr eventing resources effectively. Research to date has concluded that victim ization is the best single pred ictor of further victimization, and that when victimization recurs it tends to do so quickly. For example, research in Kirkholt (England) showed that â€œonce a house ha d been burglarized it s chance of repeat victimization was four times the rate of houses that ha d not been burgled beforeâ€ (Forrester, Chatterton et al. 1988). The time in terval between the in itial event and revictimization was first explored by Polvi et al . (1990). This research showed that the risk of re-victimization was greates t immediately after the offenc e. The likelihood of a repeat burglary within 1 month is 12 times the expect ed rate, declining to 2 times the expected rate at 6 months. Half of all second victim izations occurring with in 1 month after the initial event occur w ithin 7 days. Another study in Aust ralia found that 65% of repeat burglaries occurred within 1 month of the original event, and 83% within 2 months (Guidi, Townsley et al. 1997). This means attention given to people and places over a relatively short time frame is much more productive in crime reduction terms when it is concentrated on targets already victimized. F ourth, as previously noted, the impact of repeat victimization on victims is profound, both in economic and psychological terms. Research on repeat victimiza tion in Scotland showed that victims do not become inured to crime, and that they suffer many negative emotional effects, even when victimization episodes appear trivial. â€œIt can be compared to a bereavement process where victims go through various stages after each incidentâ€ (Shaw and Pease 2000, p. 51). This research
29 also showed that many repeat victims have lo w expectations of what the police can do to help them, and these low expectations may lead to a failure to report future crimes. In another analysis by Pease (1998), it was estim ated that 40% of cr imes that happen to single-incident victims were reported to poli ce, compared to a 28% reporting rate among those victimized six times or more. Therefore, preventing crime from recurring means fewer victims suffer these compounding a nd disproportionate crime effects. In addition to the general be nefit of investigating repeat victimization, there are two more reasons that suggest repeat burglary analysis is important to explore, especially in relation to environmental variables. First, areas with high levels of crime can benefit from repeat victimization prevention, as there is evidence that high -crime-level areas are also areas with high levels of multiple victimization (Trickett, Osborn et al. 1992). Bennett (1995) showed that just over one third of a ll domestic burglaries o ccurring within a hot spot were part of a series of burglaries. In addition, some research even has claimed that the main reason the incidence of victimization (total crime incidents per capita) is so high in some areas is due largely to large proportions of repeat vic tims, rather than to the sheer increase in the number of different vi ctims. (Hope 1995; Pease 1998). Second, the analysis of repeat burglary targets is useful in iden tifying environmental factors contributing to burglary. Some factors that mark a place as a compelling target cannot be changed quickly and may lead to repeat vi ctimization. These factors would likely be more distinctive in analysis of repeat victimization target gr oups than analysis of a full group with some random chosen targets. In short, research on repeat victimization is promising and important for a number of reasons, including (a) the extraordinar y costs to victims and society, (b) the
30 identification of environmental variables that may be linked to repeat victimization, (c) crime prevention and policing strategies, and (d) fundamental crime pattern analysis. 2.2.3 Application Projects on Repeat Victimization The study of Johnson, Kerper, et al. (Johns on, Kerper et al. 1973) may be the earliest work on repeat victimization, wher eas this phenomenon has exerted extensive research interest from the 1990s. To a large extent, the success of the Kirkholt Burglary Prevention Project in the United Kingdom (Fo rrester, Chatterton et al. 1988; Forrester, Frenz et al. 1990) acted as a cata lyst for much of the recent work in repeat victimization. The Kirkholt project was aimed at reducing the high level of residential burglary on a public housing estate in Rochdale, in the northwest region of E ngland. In the first research phase, the researcher interviewed the burglar, the victim, and the victimâ€™s neighbors. In addition, the re searcher contacted local ag encies and groups within Kirkholt. Combining this information with the analysis of available burglary data, it was found that the â€œchance of a second or subseque nt burglary was over four times as high as the chance of a firstâ€ (Forrester, Chattert on et al. 1988). These data showed also that nearly half of those households burglarized in December 1986 had been burglarized earlier in the year. In light of these find ings, the value of the strategy of prevention relative to residential burglary was apparent to Kirkholt researchers, and has had an impact far beyond the borders of this British es tate: the prevention of repeat victimization resulted in significant impacts on absolute levels of residentia l burglary. Within 5 months, a 60% drop in burglary was observed, with no signs of displacement to other nearby areas or to other forms of crime. Following the apparent success of the Kir kholt project, there were some crime prevention projects thatâ€”through repeat-burgl ary preventionâ€”aimed to reduce burglary
31 re-victimization. However, few of them have achieved results as impressive as those of the Kirkholt project. In 1998, a demonstration project was conduc ted in Queensland, Australia. Like the Kirkholt project, this project involved making an immediate response to burglarized addresses, such as (a) providing burglary prevention booklets and property-marking kits to victims; (b) conducting home security qu ick assessments and identifying risk-reducing measures to victims; (c) providing portable s ecurity alarms, locks, or similar security equipment to victims who had been burglar ized within the past 12 months; and (d) contacting near neighbors and en couraging them to take immediate steps to improve their own home security. The Queensland project used other approaches also, such as increasing police patrols in hot spots and encouraging the establishment of Neighborhood Watch programs. This project produced a mode st decline in the ab solute number of repeat victimizations. It also produced a substantial reduction in the probability of becoming a repeat victim. In addition, ther e occurred a reduction in the number of offences in the targeted hot spots, wit hout any apparent displacement to neighboring areas. However, there is no evidence that th e total number of bur glaries was reduced. Another Australian project was designe d to prevent repeat victimization (Henderson 2002). It used a range of local and community re sources, rather than police responses, and produced results similar to those of the Queensland project. Repeat burglaries were prevented successfully, wher eas the total number of burglaries was not reduced significantly. 2.3 Summary This chapter provides an overview of th eories about the linkages between the physical environment and crime. The most promin ent theories in this area are defensible
32 space, crime prevention through environmen tal design (CPTED), situational crime prevention and rational choice theory, routin e activity theory and life style theory, environmental criminology and crime pattern theo ry. In this chapter, we lay out the main hypotheses and major research findings of each of these schools of place-based crime prevention theory, and then outline major lim itations and criticism of these theories. This chapter also discusses the importance for exploring repeat burglary and repeat victimization and summarizes basic theories a nd international research on these topics. In short, research on repeat victimization is promising and important for a number of reasons, including (a) the extraordinary costs to victims and society, (b) the identification of environmental variables that may be linked to repeat vict imization, (c) crime prevention and policing stra tegies, and (d) fundamental crime pattern analysis.
33 CHAPTER 3 RESEARCH DESIGN AND FRAMEWORK This chapter presents methodological issues relevant to the current study. First, the purpose of the empirical research is de scribed, followed by a description of the geographical area in which the research was conducted and data sources. A following section describes the unit of analysis employed. The third se ction introduces environment factors and measurements explored in this research. Finally, the analysis methods are described. 3.1 Purpose of the Study As discussed above in Chapter 1, this dissertation intends to use Geographic Information Systems and statistical tools to (a) explore the spatial and temporal patterns of burglary, (b) examine the correlation betw een burglary and envir onmental variables, and (c) identify specific featur es of the physical environment that contribute to burglary in general and to repeat burgl ary and â€œnear repeat burglaryâ€ in part icular. We hypothesize that some environmental variables such as accessibility, house lo cation on the block, and adjacent land uses have strong contributions to burglary, repeat burglary, and â€œnear repeatâ€ burglary propensity, despite sociodemographic neig hborhood differences. To test this hypothesis, this empirical research uses a case study approach and analyzes data from the Gainesville, Florida Police Department, fo r 3100 residential burglaries from January 2000 to December 2003. The analysis includes three crime patterns: complete residential burglary, repeat burglary, and near repeat burglary. For each crime pattern, two categories of analysis--
34 spatial-temporal analysis and environment variable analysis--are applied. Complete residential burglary refers to all the resi dential burglary cases from January 2000 to December 2003. Every residential burglary cas e that occurred during this period will be viewed as one point in this pattern anal ysis. Addresses that experienced multiple burglaries will have several overlapping points at the same location. Repeat burglary and near-repeat-burglary incidents are neither distinguished nor excluded from the complete data set. In the current res earch, the term repeat burglary is defined as the same single family residential unit that is subject to more than one burglary over a 4-year period. Because in many cases it is difficult to id entify the precise dwelling unit burgled at multiple-family units, repeat burglary in multiple dwelling units is not explored in this research. Near repeat burglary refers to burglary cases whose occurrences are close in both time and space, such that an area is at higher risk than would be expected by random distribution. Recent research sugge sts the possibility that near re peat burglaries are a real phenomenon (Townsley, Homel et al. 2003; Ya ng 2005). The intent of the near repeat burglary research is to answer the followi ng questions: (a) Do near repeat burglary phenomena exist in Gainesville? If the answer is yes, (b) what are the methodologies to detect this phenomenon, and (c) what kind of environment variables are correlated with near repeat burglary? 3.2 Geographical Area and Data 3.2.1 Geographical Area This research was conducted in Gainesville, Florida. Gainesville is located in the north central region of Florida. It occ upies approximately 17,648 acres, has 37 thousand of households, and a population of 100 thous and (U.S. Census, 2000). As a university city, Gainesville has a populat ion with a median age that is only 26.4, much younger than
35 the median age (35.3) of the general populat ion in the United States (U.S. Census Bureau). In Gainesville, the percentage of people in the total population between 15-24 years of age that has been proved to be significantly correlated with crime is 32.59%, whereas the percentage in th e general U.S. population is 13.9 %. Gainesville is about 12 miles across both north-south and east-west. Th e city has the north -south highway, I-75, running along its western border. University Avenue and Main Street are the east-west and north-south arteries running through the ci ty (Figure 3.1). Resi dential parcels make up 34% of the total land area in the Gainesvi lle city limit. The median assessed estate value for single-family parcel in 2000 was $64,135. Although, within the city limits, the value of over 80% of single family re sidential parcels is below $100,000, there are several areas that incl ude parcels valued at more than $200,000. Most of them are located around the northwest part of Gainesville (Figure 3.2). The University of Florida is the leading employer in Gainesville. The institution has more than 48,000 students and 4,000 faculty members. Most of them live in and around Gainesville. As one of the largest unive rsities in the United St ates, the University of Florida has a 2,000-acre campus, and half of th at is located within the Gainesville city limits. This research does not include burgl ary within the bounda ries of the campus.
36 Figure 3.1 Land use of Gainesville. Figure 3.2 Assessed value of resi dential parcels.
37 3.2.2 Data Sources Data for this research can be classified in three categories: (a) burglary data, (b) socioeconomic and demographic data, and (c) environmental data. Burglary data were gathered from the Gainesville Police Depa rtment. The data record all the reported burglaries in Gainesville, Florida from January 2000 to December 2003. Information about burglary location, type of burglary, the points and met hods of entry, estimated date and time of the offenses, and item stolen are included. Within this time frame, 8,220 cases of burglary in Gainesville were re ported. Among all the burglary cases, 2,221 were repeated burglary. Socioeconomic and demographic data ar e gathered from year 2000 census data. Although this research is concentrated on th e environmental effects on burglary, we are not trying to make an â€œenvironmental de terminismâ€ argument. In this context, the influences of socioeconomic and demographi c attributes are also considered when environmental effects are explored. The use of socioeconomic and demographic factors in this research will be explained below. The main sources of environment data are Florida Geogra phic Data Library (FGDL) and local Property Appraiser parcel da ta. These data provide information about the physical context of offenses, such as surrounding land use types, lot density, connectivity, etc. Table 3.1 Overview of Burglary in Gainesville 2000 2001 2002 2003 Total Residential Burglary 776 791 771 762 3,100 Business Burglary 387 492 390 368 1,637 Conveyance Burglary 1058 906 879 640 3,483 Total 2,221 2,189 2,040 1,770 8,220 (January 2000December 2003, Gainesville, FL)
38 Table 3.2 Overview of Repeat Burglary in Gainesville Type of Burglary Overall Count Repeat Count Percentage (Repeat /Overall) Residential Burglary 3,100 1,277 41.19 Business Burglary 1,637 1,078 65.85 (January 2000December 2003, Gainesville, FL) 3.3 Geocoding The analysis of residential burglaryâ€™s pattern is based on the ability to locate where crime incidents took place. The process of associating a data point with a geographic location based on address is called geocoding or address matching . Automated street geocoding is the most popular method used to assign geographic coordi nates to incidents based on their street addresses. This method calculates the positions of addresses based on a reference street centerline file, which contains lo cation information about street names, address number ranges, and/or zip code s. For example, if an address number is 120, this address will be allocated to a posit ion 20% away from the starting point of the street centerline feature that has the specifi ed street name of the address in address numbers ranging from 100 to 200. In real life, address num bers do not increase in a strictly proportional manner. This automa tic geocoding process introduces positional error in the geocoded point. Fi gure 3.3 demonstrates this type of error. An alternative method of address locating is parcel addre ss matching, which is based on the location address of property parcel data. This met hod provides more precise coordinates of addresses. However, there are two shortcomi ngs for parcel address matching. First, the parcel address matching method produces a lo wer matching rate than does the street geocoding method, because parcelsâ€™ location ad dresses are collected only from 2004 in Gainesville. Further, not all pa rcel addresses have been r ecorded. Second, the quality and format of address recording are somewhat inconsistent between records. For example, for
39 some records, there are two text spaces be tween street directions and street names, whereas for other records there is only one text space. Moreover, in contrast to some records--in which street types such as Street, Road, and Avenue are recorded in full--for many records abbreviations such as St, R d, and Ave are used. For example, NW 10th STâ€ may be documented as NW 10 St reet.â€ Because the mechanism of parcel address matching is address text string co mparison, spelling variations could reduce matching rate significantly. Because spatial analysis of burglary and crime relies not only on accurate locating but also on a high level of matching rate, th is research employed both parcel address matching and automated street geocoding to pr oduce an given incident location point file. The geographic coordinates of burglary incide nts were located first by property parcel data. To overcome the problems of inconsis tent address recording, addresses both for burglary incidents and parcel lo cations were standardized wi th the help of the ArcGIS geocoding engine. The address components were separated using a sing le text space. All street names and directions were edited according to the standardized form. This procedure maximized the matching rate of the parcel address matching method. For crime incidents where the lo cation could not be found by pa rcel address matching, we employed street geocoding to pinpoint their pos itions. GDT street cen tral line data were used as the base map. Employing this strategy, we matched 3,134 r ecords from the to tal of 3,262 records labeled as residential burglary ; the match rate is 96%. A total of 29 records that fall outside the city limit of Gain esville were dropped. There were records that occurred in the same nonapartment household on the sa me day, we believe these are duplicated
40 records for same case. A total of 5 more cases were dropped af ter these duplications were cleaned. With all the data validation, 3,100 records were kept for this analysis. Figure 3.3 Inconsistence between street geocoding and parcel address. 3.4 Analysis Scales and Units In spatial analysis, the unit of analysis can affect observations a nd conclusions. It is such a key issue that it should be considered prior to any research related to spatial values. In the current research, the analysis is applied in various scales and units discussed below. 3.4.1 Modifiable Area Unit Proble m (MAUP) and Ecological Fallacy Modifiable area unit problem (MAUP) and ecological fallacy are common major difficulties for geography data analysis. These two problems are closely related. The modifiable area unit problem is special fo r all spatially aggregated data, whereas ecological fallacy is a more general statistical problem. The ecological fallacy â€œconsists in thinking that relationships observed fo r groups necessarily hold for individualsâ€ (Freedman 1999, p. 1). For instance, if House A is located in a community where the
41 average house price is higher than in House Bâ€™s community, the ecological fallacy would lead us to conclude that Hous e Aâ€™s house value is higher than that of House B. But this is not necessarily the case. The modifiable ar ea problem arises from the imposition of artificial units of spatial reporting on continuous geogra phical phenomenon resulting in the generation of artificial spatial patterns (Heywood 1998). MAUP applies to two interrelated components: the zoning probl em and the scale problem. The phrase zoning problem refers to the effect that conclusions from the same area data set can be different when the area units are combined into z ones that are of the same size but located differently. One example of this problem is the 2000 U.S. presiden tial election. If one northern Florida county had been switched to Georgia or Alabama, Al Gore, would have won more counts in an aggregated (state ) level, and would have won the election (O'Sullivan and Unwin 2003). What is referred to as the scale problem in MAUP is similar to the concept of ecological fallacy , which states that when data are aggregated into different levels, the conclusions and in ferences may be altered. Sometimes, inference from aggregate data may be misleading or obscure some important underlying patterns. For instance, some census tracts may include major traffic arteries. Such roads may attract a large number of burgl aries. As only a small portion of each tract that runs along the major road has a very high bur glary rate, if all other parts of those tracts have a very low burglary rate, the median bur glary rate for those census tr acts may be also very low. In such a situation, statistical analyses from the aggregate data at census tract level may infer that the major road will not influence the burglary rate significantly. However, with data analyses at the block or site level, we may find there is a significant relationship between burglary and major traffic arte ries (Brantingham and Brantingham 1981).
42 To deal with the modifiable area unit problem in spatial studies, multiple approaches have been suggested in the geogr aphical literatures such as optimal zoning approach, sensitivity approach and basic en tity approach (Openshaw 1977; Fotheringham 1989). In this research, we applied two a pproaches, basic entity approach and multiple scale approach to minimize and test the influence of MAUP. Basic entity approach for MAUP advocates the identification of individual entities that are ecologically meaningfully and not modi fiable, and to perform analysis directly on them (Fotheringham 1989). For residential burgl ary, the basic entity is the residential parcel. Environment variable analysis for co mplete burglary and near burglary sites are applied on this level. However, not all anal ysis can be applied on the individual parcel. For example, the identification of social ec onomic and demographic variables that have significant influence on the burgl ary rate has to be explored in aggregate level â€“ census block, as it is the smallest aggregate scale at which most data for social-economic and demographic variables are available. We have to presume that the analytic result for burglary rate and social-economic and demograp hic variables, which is applied on census block area level, may yield reliable estimates of underlying individua l-level relationships for burglary sites and social-economic and de mographic variables. Only based on this assumption, we can use match control case methodology to find un-burglarized control sites for burglarized case sites by controlli ng the identified variab les in census block analysis. This assumption may raise the issu e of MAUP or ecological fallacy, however, some researchers who focus closely on the MAUP have â€œconclude that the use of well chosen grouping variables [ may provide] at least a partia l solution to the MAUP with respect to the â€˜ecological fallacyâ€™, the drawi ng of individual-level inferences based on
43 area-level analysesâ€ (Bivand 1998). Since census block is consistent w ith street block in most cases, and the median size of census blocks â€™ area within the Gainesville city limit is only around 4 acres, the intrazonal variation of all these variab les would be very small. We can, therefore, conclude that census bl ock is an appropriate grouping scale. 3.4.2 The Analysis Scale Employed in This Research It is not difficult to define the microanaly sis unit; it is the site where the burglary occurred. For single-family houses, it is the parcel of the burglarized house. For multifamily buildings, it is the house unit that was victimized. However, since some environment features have to be collected at the macrolevel, such as lot density, the decision concerning the appropriate aggrega tion unit is complex. The complexity stems from the issue of the modifiable area un it problem (MAUP) and ecological fallacy, as described above. In Brantinghamâ€™s (1976) resear ch, the burglary patter n was analyzed at a fivelevel cone of resolution: (1) the national pattern, based on state level data; (2) the Florida pattern, based on county level data; (3) the pattern fo r the city of Tallahassee, based on census-tract-level data; (4) a census tract patt ern, based on block group data; and (5) a block groupâ€™s burglary pattern, base d on data for an individual city block. For most aggregated crime pattern analysis, censu s block and/or census tract are the most popular units, because of the ability to li nk the decennial demographic-economic census data. As the smallest geographic area for wh ich the Census Bureau tabulates data from Census 2000, a census blockâ€™s boundaries can be (a) visible features, such as roads, rivers, railroads, and power lin es (which must be in the TIGER data base); (b) political boundaries, such as counties, townships, a nd cities; and (c) sta tistical area boundaries, such as census tracts, census county divisions, and census designated places (U.S. Census Bureau, n.d.). Census tracts are aggregations of census blocks. The boundaries of a
44 census tract not only follows visible featur es, governmental unit boundaries, and in some instances other nonvisible featur es, but also are designed to divide census tracts into relatively homogeneous units w ith respect to popul ation characteristics, economic status, and living conditions. Census tracts usually contain 2,500-8,000 people, and generally average approximately 4,000 inhab itants (U.S. Census Bureau, n.d.). Street blocks can be another potential unit for crime pattern analysis. Street blocks are the two sides of a street between two cross streets (T aylor 1998). As block boundaries are more easily defined and less ambiguous than are some other conceptions such as neighborhood boundaries, the street block is also a popular unit in crime research related to spatial values. In the current research, the first aggrega tion analysis unit will be census blocks, because (a) most of census block boundaries ov erlap with street block boundaries and (b) the use of census blocks is advantageous in linking the demographic-economic census data. However, it would be too limited to onl y evaluate the environment features of the block where studied site lo cates. Because several scholar s have found that potential burglars tend to penetrate two to three bl ocks from roads, entertainment places, and employment centers (Luedtke and Associat es 1970; Buck, Hakim et al. 1993), and the average size of street blocks in Gainesv ille is around 400 feet, we use a 1,200 foot buffer of the studied site as another option of aggr egation level. In add ition, the major street block, which is a unit defined by major streets, is also employed in the current research. The choices of aggregate level for each envir onment variable will be explained below. 3.5 Environmental Factors and Measurement To analyze the relationship between envir onmental factors and residential burglary, it is important to know what kind of factors may influence residential burglary and how
45 to measure them. There are two objectives fo r the following section. The first objective is to establish a conceptual framework of burglar y-relevant environmental factors. For this study, we characterized the physic al environment in terms of four categories of features: (a) permeability, (b) land use and adjacency, and (c) housing density. The choice of these three categories is based on a review of the literature and on the avai lability of suitable data. Each category is described by several different variables. Some categories of environmental features, such as land use ad jacency and density, are important to new urbanists as outcomes of mixed land use a nd sustainable development. Although some urban planners advocate these principles beca use of perceived advantages in creating social justice and â€œhealthyâ€ developments, th ey are believed to be related to decreased safety in several empirical environment cr iminality studies. Every principle can be applied by controlling different aspects of the physical environment. For example, enhancing the land use mix can be achieved by mixing different type s of land use in a district, by mixing diffe rent types of homogenous nei ghborhoods in a district, or by mixing different types of dwe llings in a neighborhood. By exam ining different aspects of these characteristics the researcher may find that some features in these categories are significantly correlated with re sidential burglary, whereas th e contribution of some other features to burglary is not so significant. This knowledge may provide planners with better opportunities to find an appropriate balance among competing goals for urban development, from environmental efficiency and social justice, to public safety. The second objective of the following secti on is to develop objective methods to measure characteristics of the physical envi ronment that should be included in this research. Although there were many studies th at attempt to explore the relationship
46 between features of the physical environmen t and specific crimes, few researchers have measured the physical environment objectiv ely (Perkins, Wandersman et al. 1993). For example, Jeffery (1977) described the relatio nship between urban environment and crime target sites, based on his obs ervations in New York. Many other studies have been based on personal subjective perception surveys. Howe ver, some research argues that crime can be better predicted through obj ective measures of the environment than through studies of social perception (Gifford 1993; Hiller 1996) . Furthermore, the outcomes of subjective measurement are difficult use in zoning guidance. For example, Jeffery (1977) suggested that â€œphysical features whic h divide or isolate areas a nd activities can be crime producingâ€ (Jeffery 1977,p. 194). When used as guide to urban design or crime site prediction, this vague statement leaves a gr eat deal open to inte rpretation. The current research will measure the physical envir onment features objec tively rather than subjectively. 3.5.1 Permeability The term permeability refers to the level of intrus ion difficulty for a district. The relationship between permeability and burglary has been strongly and consistently emphasized in the literature concerning envir onmental factors associated with burglary Areas with higher permeability are more likely to become part of the awareness space of people, including potential burglars. At the same time, high permeability also provides easy avenues of escape. Neighborhoods with high-permeability property are more likely to be evaluated as â€œsafeâ€ and c hosen as targets by offenders. Permeability is influenced by traffic ar ound and in the district. Whiteâ€™s (1990) research found neighborhood permeabilityâ€”i.e., the number of access streets from traffic arteries to the neighborhood--had a significant effect on burglar y rates. It is also noted
47 that paths between high-activity nodes tend to concentrate criminal offenses, as many crimes occur on the main roads that carry si gnificant traffic and have major public transit stops (Brantingham and Brantingham 1993). In Massachusetts, the Cambridge Police Department reported that areas near public transportation have a higher crime rate than inaccessible areas (Cambridge Police De partment, 1998). In addition to auto accessibility, reducing pedestrian movement in certain neighborhoods also suggests possible beneficial effects. A 62% decline in the rate of serious cr ime (burglary, robbery, and assault) is reported by Newman ( 1980; 1996) on the effects of changes to the Clawson Point public housing complex in the Bronx. Restricting the movement of pedestrians was part of the cha nges instituted to reduce crime. Street layout is another factor of permeab ility. Some research suggests that â€œareas with grid networks have hi gher potential crime rates than areas with organic street layoutâ€ (Brantingham and Brantingham 1981). In grid areas, because street layouts are more predictable, a potential offender can es cape more easily and mo re rapidly than in areas with cul-de-sacs, winding roads, or d ead-ends, where he or she can more easily become disoriented and lost. The organic stre et layout also projects a private atmosphere that may cut down on the level of use by strangers and nonresidents, because nonresidents are more easily identifiable. Fo r these reasons, potential offenders may find it more difficult and confusing to penetrate such areas. Bevis and Nutter (1977) revealed in a study in Minneapolis that the safest stre ets were dead-end streets with only one-way in or out. In their study of liquor-related crime incidents, Bloc k and Block (1995) found that many hot-spot areas are located at major intersections, especia lly intersections of grid and diagonal streets.
48 However, not all studies support this argument. Hillier and Shu found cul-de-sacs may be preferred by burglars, because th ey deter passers-by and reduce natural surveillance (1999). Furthermore, new urba nism, the prominent movement in urban planning from last century, disagrees with the concept of cul-de-sac street networks. Interconnected street grid networks, wh ich provides permeability, are one of new urbanismâ€™s prime tenets. New urbanist theory claims that â€œthe design of streets and buildings should reinforce safe environments, but not at the expense of accessibility and opennessâ€ (Gindroz 2000, p. 133). Despite all the disagreement between new urbanism and place-based crime prevention theory, th ey share the same premise: The physical environment can influence human behavior. Further, they share the same objective: providing better life quality to pe ople. With the help of empirical data analysis, we hope to provide evidence empirically grounded b earing on the discussion rather than just dogmatic arguments. 3.5.2 Measurement of Permeability There are many studies regarding the relationship between permeability and burglary. The measuring methodologies applied in these studies vary. Table 3-3 lists the indicators and their measuring methods and summarizes the conclusions. Some of these indicators may overlap each other. Among all of these indicators, there is no one widely accepted standard that can measure all aspects of an areaâ€™s permeabilit y, accessibility, and connectivity. Considering both intuitiveness as a potential guide to pl anners and data ava ilability, the following indicators are selected for this research. Some of thes e indicators are correlated.
49 Table 3.3 Indicators of Perm eability and Accessibility Variables Coding method Conclusion Source Permeability Number of access lanes leading from each artery into the neighborhood Permeability has a significant influence on neighborhood burglary rate when neighborhood economic factors, instability, and structural density are controlled. (White 1990) No boundaries(fences, shrubs, etc.) One or more dimensions greater than 400 feet Three or more people present Signs of use(debris, defacement, wear) Bounded by public facilities Dichotomy Accessibility and opportunities to observe outdoor public areas are not important factors in whether or not households will be the victims of a crime. (Booth 1981) Located on deadend street Dichotomy Houses located on dead end street are less likely to be burglarized. (Hakim, Rengert et al. 2001) Corner location Dichotomy Houses at corner locations are more likely to be burglarized. (Hakim, Rengert et al. 2001) Distance from highway exit Within .25 mile, .25-.5 miles, .5-1 mile, > 1 mile Houses close to a highway exit are more likely to be burglarized. (Hakim, Rengert et al. 2001) Corner location Dichotomy There were higher rates of commercial burglaries in facilities on corner lots. (Nasar 1981) Street type Major or minor There were higher rates of commercial burglaries in facilities along major streets. (Nasar 1981) Distance to major thoroughfare <.5 mile or >.5 mile There were higher rates of commercial burglaries in facilities near major thoroughfares. (Nasar 1981) Distance to downtown < 2 miles or > 2 miles There were no significant effects of commercial burglary related with distance from downtown. (Nasar 1981)
50 Table 3.3 Continued Variables Coding method Conclusion Source Distance from highway exit Within .25 mile, .25.5 mile, .5-1 mile, > 1 mile Houses located within .25 mile of an exit are more likely to be burglarized, those located .5-1 mile away are least likely. (Rengert and Hakim 1998) Street type Cul-de-Sac, quiet residential street, commercial street, busy residential street, back road/local traffic The most victimized locations are houses located on a busy residential street or on a back road with local traffic only. (Rengert and Hakim 1998) Street type Dead-end, cul-de-sac, â€œLâ€ type, â€œTâ€ type, and through-traffic streets A noticeable pattern of lower residential burglary rates in housing located on blocks with lower accessibility. (Bevis and Nutter 1977) Beta (Bevis and Nutter 1977) Turns Directions from a person in a vehicle can enter or exit a street segement Property crimes are more likely to occur on street segments that are highly accessible within the road network. (Beavon, Brantingha m et al. 1994) Flow Feeder, minor artery, major artery, highway Property crimes are most likely to occur on street segments that have high levels of traffic. (Beavon, Brantingha m et al. 1994) Street type Major thoroughfare or small neighborhood street Low crime neighborhoods were more likely to have small one-way and 2-lane neighborhood streets. (Greenberg and Rohe 1984) Alley Dichotomy There was a significant relationship between the presence of bus stop crime and the existence of an alley. (LoukaitouSideris, Liggett et al. 2001) Mid block connection Dichotomy There was no statistically significant relationship between bus stop crime and mid-block connection. (LoukaitouSideris, Liggett et al. 2001)
51 Table 3.3 Continued Variables Coding method Conclusion Source Pedestrian traffic Light, moderate, heavy There was no correlation between pedestrian traffic and bus crime rates. However, with some other measures, such as average bus headway, it is found that â€œeyes on the streetâ€ have a modest positive effect on crime. (LoukaitouSideris, Liggett et al. 2001) Vehicle traffic Actual number Vehicle traffic showed a weak but significant negative correlation with bus stop crime. (LoukaitouSideris, Liggett et al. 2001) Street width Number of traffic lanes There was no correlation between street width and bus crime. (LoukaitouSideris, Liggett et al. 2001) Sidewalk width There was no correlation between sidewalk width and bus crime. (LoukaitouSideris, Liggett et al. 2001) Street type Major thoroughfare or small neighborhood street Low crime neighborhoods tended to have fewer major streets and more small, neighborhood streets than high crime neighborhoods. (Greenberg, Rohe et al. 1982) Street type Through street, cul-desac, integrated (more movement potential) or segregated (less movement potential) Cul-de-sacs may be preferred by burglars, as they deter passers-by and reduce natural surveillance (Hillier and Shu 1999) Street pattern Traditional street pattern (grid) or hierarchical street pattern There are strong correlation between layout type and crime, with traditional street patterns the best and the most â€˜modernâ€™ hierarchical layouts the worst. (Hillier 2004) 188.8.131.52 Distance to the closest major arteries Research has demonstrated that locations n ear major arteries are more vulnerable to burglary than are those far from major t horoughfares, because burglars tend to extend their search area approximately 2-3 blocks from roads, entertainment places, and employment centers (Luedtke and Associates 1970; Buck, Hakim et al. 1993). In this
52 research, the actual number of distance by feet will be used instead of categorized distance like half mile, 1 mile. The dist ance will be geometric line distance. 184.108.40.206 Street layout patterns Some research suggests â€œareas with grid ne tworks have higher potential crime rates than areas with organic street layoutâ€ (Bra ntingham, 1981, p. 51). As we have previously suggested, this is because a grid street patt ern is more predictable and permeable whereas an organic street layout may cut down on the level of use by strangers and nonresidents, as they may more easily get lost. An organic street layout also helps to build a private atmosphere, which may deter potential burglar s. However, this argument is fiercely challenged by smart growth, new urbanism, and traditional neighborhood development and space syntax theories. To test the in fluence of a neighborhood street pattern on burglary, this research codes street patterns from 1 to 5. The Gridir on street pattern is coded as 1, fragmented parallel as 2, warped parallel as 3, loops a nd lollipops as 4, and lollipops-on-a-stick as 5. It is hoped this analysis can fi nd relationships between street layout pattern and security. With this knowledge , urban planners may have better chances to find appropriate street pa tterns that balance efficiency and quality as well as reconciling functionality and safety. Source: (Southworth and Owens 1993) Figure 3.4 Evolution of street patterns
53 220.127.116.11 Street types around parcels In a study in Minneapolis, Bevis and Nutter ( 1977) examined the relative effect of dead-end, cul-de-sac, L-type stre ets, T-type streets, and thro ugh-traffic streets in a study in Minneapolis. They revealed a clear relati onship between the accessibility of street type and burglary. The safest street s were dead-end streets, as they are the least accessible. However, some research conflicts with this conclusion. Buck et al. (1993) found greater burglary levels on cul-de-sacs in three Philadelphia suburbs. As noted previously, Hiller and Shu also found cul-de-sacs may be prefer red by burglars, as they deter passers-by and reduce natural surveillanc e (1999). The relationship between street types and burglary in Gainesville will be examined by data. In the coding schema, street types around a parcel will be coded from 1 to 4, re presenting dead-end, L-t ype streets, T-type streets, and through-traffic st reets. This coding system is very similar to Bevis and Nutterâ€™s study (1977), however, dead-end and cul-de-sac are not distinguished in the current study as there is no clear cut difference. 18.104.22.168 Distance to public transportation stop In the crime reports for some metropolitan areas, it is found that areas near public transportation, particularly subways, have a higher crime rate than inaccessible area (Cambridge Police Department, 1998). However, it is not certain for small cities such as Gainesvilleâ€”where bus routes are the onl y public transportation system--how public transportation influences burglary. In this an alysis, the variable will be coded as the actual distance in feet from the pa rcel to the nearest bus stop. 22.214.171.124 Corner location Several studies indicated that corner lots are at higher risk of residential burglary than those in the middle of a block (Nasar 1981; Hakim, Rengert et al. 2001). It may be
54 because burglars perceive corner lots to have fewer neighbors, and thus there is a reduced chance of being observed (Rengert & Wasilc hick, 1985). Another reason may be because corner houses are â€œnaturalâ€ places for a potentia l burglar to ring a doorbell to check for a residentâ€™s occupancy, by using the excuse of asking directions (Hakim, Rengert et al. 2001). Furthermore, a browsing burglar may more easily determine the vulnerability of a corner house for breaking and entering than th at of a house located in the middle of a block, as more parts of a corner house can be viewed. This is a dic hotomous variable. 0 represents non-corner location, wher eas 1 represents corner location. In addition to the indicators used by e nvironmental criminologists and relevant researchers, civil engineers a nd urban planners have also developed multiple methods to measure street connectivity of street layout such as bl ock length, number of polygons, number of intersections, street densit y, and connectivity index. Among all these measures, two approaches have been used mo st frequently by pla nners to address the issue of connectivity: block le ngth and connectivity index (H andy, Patterson et al. 2003). The following is a simple summarization of these two measures: 126.96.36.199 Block length Block length refers to the average block length in the studied area. â€œThis methodology relies on the premise that the shorter the leng th of a block is, the better the connectivity will beâ€ (Steiner, Bond et al. 2004, p. 5). It is intuitive as a design standard; however, this methodology doesnâ€™t calculate a connectivity index. For example, the following two street networks have the same block lengt h, whereas their connectivity is different.
55 Figure 3.5 Example of street block length methodology. 188.8.131.52 Connectivity index Connectivity index refers to the ratio of the number of roadway segments to the number of intersections or cul-de-sac ends. This methodol ogy â€œrelies on the premise that when more links are connected to more nodes there is greater network connectivityâ€ (Steiner, Bond et al. 2004, p. 6). It is a usef ul measure; however, there is still one flaw that is inherent within it. This measure doesn â€™t consider the density of a street network, the connectivity index for an area with a perfect grid network with a 500-feet average block length is same as an area having a perfect grid network w ith a 1,000-feet average block length, whereas the actual connectivity fo r these two street networks is different. Both block length and the connectivity index should be used together to measure the connectivity of a street network. Connectivity Index = 24/17 = 1.41 Figure 3.6 Example of street connectivity index methodology. Major streets Local streets Link Node
56 Based on literature reviewed and the rele vant studies noted above, the following variables are employed in the current study to measure permeability. They are a) distance to the closet major arteries, b) street layout patterns, c) street types around parcel, d) distance to public transpor tation stop, e) corner loca tion, f) block length and g) connectivity index. 3.5.3 Land Use and Adjacency Type of land use--of the crime target as well as of the target â€™s surrounding areas--is another category of environmental factors that consistently has s hown a relationship to the risk of criminal victimization. The importa nce of this variable can be explained from several perspectives: target attractiveness (when a number of potential victims can be found at the setting), spatial at tractiveness (when the physical features of a setting make criminal acts easier to occur and to go unnoticed) (Rhodes and Conly 1981), the â€œspillover effectâ€(when crime rate of a lo cale is influenced by neighboring localeâ€™s characteristics), and potential offender cl oseness (when the accessibility of a neighborhood to prospective criminals infl uences its crime rate) (Katzman 1981). Some types of land use and residential units are more likely to attract or generate crime than are others. For example, Rhode s and Conly (1981) found that in terms of target attractiveness, â€œtransitional and commercial areas should be placed high on the list. Residential areas rank low, and industrial areas fall somewhere in betweenâ€ (p. 182). Another study found that specif ic commercial types of land use--such as liquor stores, bars, and taverns--are more likely to gene rate crime than are others (Block 1995 138). Although the present res earch focuses on single family residential burglary, where the land use of event sites would typically be fo r human habitation, the target attractiveness for different densities and types of housing units may vary considerably. Sampson (1986)
57 asserted that the percentage of residential units in structures of five or more units was a significant positive predictor of victimization rates, regardless of other explanatory factors such as racial composition and poverty. The types of land use in the area surrounding a potential target is another important issue. The Brantinghams argued that if en tertainment complexes are located near residential areas, associated crime increas es disproportionately (Brantingham and Brantingham 1981). This phenomenon can be explained by spatia l attractiveness. Because certain types of land use attract activities and peopl e, residential locales around certain types of land use are more likely to be included in the cogni tion map of outsiders. Consequently, target vulnerability in such locales is more likely to be apparent to potential offenders, and such areas are more likely to be victimized. However, not all types of land use juxtapositions may impair security. Jane Jacobs (1961) and Newman (1972) argued that properly juxtaposing types of land use might actually enhance surveillance. The positive and negative types of land use juxtaposition will be reviewed thoroughly in this research. Even for districts used mainly for re sidential purposes, the continuity of characteristics across adjacent areas also affects burglary. The Brantinghams (1975) found burglary occurs at high rates in blocks where characteristics of adjacent areas are dissimilar. On the other hand, they found that burglary occurs at much lower rates in blocks that are similar to surrounding areas . In a study of suburban crime, Katzman (1981) also found that for property crime, including burglary, the characteristics of surrounding neighborhoods like median housing valu e are a better predic tor of crime than are the economic characteristics of a nei ghborhoodâ€™s local populat ion. These phenomena
58 can be explained by â€œspillover effectâ€, which cl aims crime rate of a locale is influenced by neighboring localeâ€™s characteristics. Furthermore, some research found that pr oximity to dilapidated houses and wooded areas might reduce the chance of surveillance, thus augmenting susceptibility. All these arguments will be explored in this empirical research. 3.5.4 Environmental Variables Relating to Measurement of Land Use and Adjacency The environmental indicators for land us e and adjacency will include the variables listed below, most of which are aggregate da ta of the broader areas in which the event sites are located. However, â€œvacant buildings and dilapidated housesâ€ is a variable which evaluates the immediate surr oundings of event sites. 184.108.40.206 Adjacent land use types Mixing land uses is one of the principles most advocated by urban planners while some research on environment criminology ha s indicated that mixing diverse types of land use with residential land use may endange r safety. Urban planne rs claim mixing land uses can restore vitality, environmental qualit y, social equity, and e fficiency. Jane Jacobs, in her benchmark book, The Death and Life of Great American Cities (1961), proposes mixing types of land use. Furthermore, the movements known as sustainable development, new urbanism and smart growth â€”all important theories in urban planning in recent decadesâ€”support mixed types of la nd use. Although the principle of mixed land uses is well established in the contemporar y academic planning world, its influence in the practical world still needs to be explored a nd clarified (Grant 2002). The consequences of mixed land uses should be analyzed from multiple perspectives, including economics, environment and safety. Because public security is an important issue for urban planning,
59 the analysis of this category of environmental variables aims to find out which types of land use are compatible with residential land use, in terms of enhancing--or at least not endangering--the safety of urban residences. Much research has explored mixing different types of land use with residential land use, and how this affects crime, especially residential burglary. Gree nberg et al. (1982; 1984) found that homogeneous residential ne ighborhoods had lower rates of crime. Dietrick (1977) found that resi dential burglary occurred more frequently near commercial areas. Moreover, a recent study claims that, â€œregarding burglary, the presence of sc hools is non-significa nt. Presence of businesses increases burglary, though the effect is partially mediated by physical disorder. The effect of businesses is also moderated by residential (in)stability. Presence of playgrounds increases burglary risk re gardless of neighborhood social-structural characteristicsâ€ (Wilcox, Quisenberry et al. 2004). However, Jacobs (1961) and Newman (1972) assert that diverse land use can attract a continual flow of people, provide more chan ces for informal surveillance, and therefore prevent crime. In contrast, domination of a singl e type of land use can result in areas that are deserted for long periods of time and leave the area unguarded. To find the relationship between adjacen t land use and residential burglary in Gainesville, both the type and the extent of adjacent types of land use should be taken into account. It is also im portant to define the terms adjacent or near . Because several scholars have found that potenti al burglars tend to penetrat e two to three blocks from roads, entertainment places, and employmen t centers (Luedtke and Associates 1970; Buck, Hakim et al. 1993), and the average size of street blocks in Gainesville is around
60 400 feet, we use 1,200 feet as th e analysis range. To identify type and extent of land use diversity, we use two approaches: (a) the presence of each specific type of land use in the studied sitesâ€™ 1,200 feet buffer zone, (a dichotomous variable ) and (b) the percentage of lots dedicated to each specific land uses in the studied sitesâ€™ 1,200 feet buffer zone, (the actual number.) In the current research, inst ead of exploring certain preselected land use types, all land use types coded in Alachua county a ppraiser parcel data are screened. The panoramic scan may provide the researcher a chance to have a better understanding of the relationship between adjacent land use types and burglary. 220.127.116.11 Degree of land use types mix In addition to identifying types of land us e that are compatible with residential usage, the degree of land use homogeneity or heterogeneity will also be explored. We apply two measures of land use mix in this research. The first indicator is general land use mix , which is an indicator established by pla nners and civil engineers. This indicator measures land use mix in terms of diversity among spatial units of a sketch area (Planners/Engineers 2002). By laying an imagin ary grid of 0.1-acre ce lls over the top of the explored area, every subject cell will be assigned a value determined by the number of dissimilar land use cells adjacent to it. Th is process is repeated for all cells and summed into a single value for the entire area . Instead of character izing the amount of different uses in an area, it measures the fr equency of encountering different uses when moving across an area. The higher the value, the greater the land use mix. The following is the formula for calculating the general land use mix indicator: Formula: Di/ Ci Di = number of dissimilar cells adjacent to Cell 1 Ci = number of cells adjacent to Cell 1
61 The second indicator is relative land use mix, which measures the ratio of nonresidential types of land use to housing units in the sket ch area. This indicator is derived by dividing the number of acres of nonresidential (commercial, industrial, etc.) types of land use in the ar ea by the number of housing uni ts. The higher the ratio, the greater the relative land use mix. In the current research, both general land use mix and relative land use mix are measured on two levels, census block and major street block. 18.104.22.168 Residential units mix Burglary, however, is influenced by factors more than adjacency of nonresidential land uses. For instance, in districts used ma inly for residential land use, the mix of different housing forms and densities may also affect the frequency of burglary. There are several theories that address this issue; but th eir conclusions are fiercely conflicting. Furthermore, there are very few empirical studi es to support either of the sides. It is hoped that this research can shed so me light on this tangled issue. The theoretical models concerning the mi x of housing units can be summarized from two perspectivesâ€”that of residents a nd that of burglars--a lthough the conclusions from each perspective still c onflict with each other. First, from the perspective of resident s, homogeneous residential communities may enhance social cohesion or rest rict the people in a given area to â€œlegitimateâ€ users. These results are expected, in turn, to lead to an in crease in residentsâ€™ se nse of â€œterritoriality,â€ thus enhancing safety. However, some theori es believe that a socially diverse housing mix is preferable to create a â€œtruerâ€ community (Saville and Cleveland 1998). According to Britishâ€™s Planni ng Policy Guidance,
62 â€œ The Government believes that it is importa nt to help create mixed and inclusive communities, which offer a choice of housing and lifestyle. It does not accept that different types of housing and tenures make bad neighbors. Local planning authorities should encourage the development of mixe d and balanced communities: they should ensure that new housing developments help to secure a better social mix by avoiding the creation of large areas of housing of simila r characteristics.â€ (PPG 3, Housing, paragraph 10) . In addition to the advantages in social mix and sustainable development, it is argued that mixed housing can increase safe ty by encouraging people with different lifestyles to mix. Therefore, the communityâ€™s surveillance can be improved, with people coming and going throughout the day and evenin g, as compared to the single-use zones that become deserted during the working day, making the oppor tunities for crime easier. Second, from the potential burglarsâ€™ perspe ctive, there are some studies that find that burglars â€œpreferâ€ to commit crimes in communities that are similar to their own, so that they wonâ€™t stand out as suspicious stra ngers. The socioeconomic differences between the targetâ€™s community and the offenderâ€™s ma y inhibit the offender from penetrating, as the offenderâ€™s sense of unfamiliarity might be heightened. But there is also a possibility that the physical proximity of socially and economically disparate communities may increase the level of relative deprivation and endanger an areaâ€™s safety. Block concludes (1979, p. 52) that â€œit is clear that neighborhoods in which poor and middle class families live in close proximity are likely to have higher crime rates than other neighborhoods.â€ A related explanation for this phenomena may be that low income neighborhoods tend to have disproportionally higher numbers of motivated offenders living there.
63 With the theoretical models leading to completely opposite conclusions about the mix of different housing forms, there are onl y very few empirical studies to test the influence of housing mix on crime; and th eir conclusions are conflicting, too. The Brantinghams (1975) found that burglary rate s decreased rapidly to wards the core of homogeneous residential areas, but were high in transitional areas. They also found that blocks that had a high percentage of single fa mily housing units were at much lower risk of burglary than blocks that had a high pe rcentage of apartmen ts (Brantingham and Brantingham 1977; Greenberg and Rohe 1984) . In the Netherlands, Saville and Cleveland (1998) suggest that housing image and territoriality can be reinforced by placing a homogenous neighborhood (100-500 dwe llings) in heterogeneous districts (no larger than 3,000 dwellings). This statement not only supports housing mix but also underscores the importance of exploring extent and scale , and their interactions on housing mix. What impact does the intermixtu re of different types of dwelling units have? Do residential units located in blocks that have a higher degree of mix of multiple family dwellings and single-family homes r un greater risks of being burglarized than residential units located in homogeneous bloc ks? What impact does the intermixture of varied housing value single-family houses have ? For blocks with few or no apartments, do single-family dwellings located in bloc ks that have similar housing value have burglary risk similar to blocks that have varied housing value? Are residential units located in homogeneous blocks and heteroge neous districts safer? To answer these questions, we explore two measures â€“ the pe rcentage of residential land devoted to multifamily housing and the coefficient of va riation of housing valu e--in two different scales â€“ census blocks a nd major street blocks.
64 The percentage of residential land devoted to multifamily housing Although there are nine separate residential codes in the la nd-use maps for the city of Gainesville, only two of them appear to be important for our pur poses â€”single family (SF) and multifamily (MF). This variable is derived by dividing the area devoted to multifamily housing by all the area devoted to re sidential use in the studied district. This measure is an indicator of the extent of mix for different types of residential units in an area. The coefficient of varia tion of housing value Housing value is an index of manifest wea lth. Coefficient of va riation is a measure of how much variation exists in relation to the mean of the studied observations. It is computed by dividing the standard deviat ion of housing value by the mean of housing value in the studied district. This measure is an indicator of the extent of mix for different types of single-family dw ellings in an district. 22.214.171.124 Vacant buildings and dilapidated houses Vacant buildings have been found to be desirable targets of crime. Research suggests that substandard housi ng is associated with higher rates of crime on the block (Spelman 1993) and this is consistent w ith the â€œbroken windows theoryâ€ (Wilson and Kelling 1982; Kelling and Coles 1996). In the coding system for this research, event sites that are substandard houses are coded as 0, event sites that ar e adjourn adjacent to substandard houses are coded as 1: all other sites all coded as 2. The identific ation of substandard houses is based on the outcome of a study accomplished by Zwick and Papajorgji (2003).
65 3.5.5 Density The relationship between housing density and crime is not clear. There is a common myth in America that increases in de nsity lead to social problems and crime. This belief is invoked by the crime and disord er issues that are correlated with dense inner cities and reinforced by John R. Calhounâ€™ s experiment on rats, which asserted that crowding causes "social pathol ogiesâ€ (1962). However, this popular conception has been disputed in the last 20 years. Some research did find that higher density is related to higher crime rates. For example, Gillis (1974) found that structural density (the proportion of multiple dwellings) was the best predictor of juvenile delinquency. In Cambridge, Massachusetts, it was also f ound that densely populated neighborhoods experienced higher residential crime rate. At the same time, some studies found that density is not a reliable predictor of crime. Freedman(1975) found no relationship between juvenile delinquency and population density, with income level and ethnicity controlled. â€œQuite startlingly, there is a reve rse relationship between density and juvenile delinquency in low-income areasâ€(Freed man 1975, p. 62). Schichor, Decker, and Oâ€™Brian (1979) found that density was related positively to property crimes but negatively to assault offenses. In the area of urban planning, density is also a dichotomy. Increasing density is supported by many urban planners, from Jane Jacobs, to the followers of new urbanism. Intensive development is advocated because of its presumed advantage in environment sustainability. However, higher density may increase crowding and environmental stress. â€œIn a crowded situation people become soci ally withdrawn and relatively unconcerned with others, and their interac tion with other people becomes more utilitarian, superficial and transientâ€ (Leung 1993). Therefore, pe ople are less likely to develop a sense of
66 personal control over their re sidential environment, and publ ic safety may be endangered. How dense is over-dense is an impor tant question to urban planners. 3.5.6 Measurement of Density Surprisingly, within these intense debates, there is a lack of clarity about what counts when considering density and how to measure it. Three concepts are used to address the issue of density and how density affects peopleâ€™s liv es: perceived density, physical density, measured density (Alexande r 1993). Perceived density is affected by physical density, individual cognitive factors, and social and cultura l factors. Physical density is made up of measured density a nd the qualitative density of the physical environment, such as openness or closure of a site layout, diversity, light levels, and landscaping. Both perceived de nsity and physical density are qualitative measures rather than quantitative. Measured density is the c oncept that when the term â€“ density--is used what most people think about and refer to. In general, housing de nsity is a number of units in a given area (Forsyth 2003). This principle for measuring density is straightforward in theory but in practice it may be very complicated. First, there is a bewildering variety of measures to express measured density. â€œThe numerator may be th e number of persons, families, households, habitable rooms, bedrooms, housing units or dwelling units (Dus)â€ (Alexander 1993, p. 186). All these measures are correlated with each other, but not proportionally related. For example, population density is af fected both by neighborhood density and by household size. To make the issue more co mplicated, there is no widely accepted definition for the common denominator in every measurement--what is included and what is excluded. For example, to derive the structural densit y in one residential neighborhood, one must ask what should be calculated as the base land area?
67 The measured density can be classified as three categories: population density, dwelling density, and built-ar ea intensity measures. Popul ation density is mainly concerned with â€œcrowdingâ€ (which is also a ps ychological construct). It is possible to live in very high density district in a spacious apartment with no crowding, and conversely it is also possible to liv e in a detached farm house that is crowded in terms of having many people per room (Forsyth 2003). Population dens ity can be further classified into two categories: internal population density that can be measured by people per room or per bedroom, and external population density th at can be measured by per square mile. However, we do not have data for internal popu lation density such as people per room. In terms of external population de nsity such as people per sq uare mile, it is difficult to regulate this density directl y, whereas urban planners can influence it by regulating dwelling density and built-area intensity in zoni ng. In this analysis, we will explore only dwelling density and built-ar ea intensity measurements. 126.96.36.199 Dwelling density The most popular measuring method for dwe lling density is dividing the number of dwelling units by the total land area. When the denominator refers only to parcels allocated for residences--excl uding roads, parks, and othe r public lands (Alterman and Churchman 1998), the measurement may be termed parcel density or lot density . When the denominator refers to all the land area devo ted to residential facilities, such as access roads within the site, privat e garden space, car parking areas, incidental open space and landscaping, and childrenâ€™s pl ay areas where these are to be provided, the dwelling density can be termed net neighborhood resident ial dwelling density (NNRDD). However, at the neighborhood level, dwe lling-building forms--such as multifamily dwelling or single family dwelling--have a cl ear association with residential densities
68 (Alexander 1993). Using average value may ma sk the variability caused by building types. To resolve this problem, the concept of net neighborhood resi dential building type density (NNRBTD) (Alexander 1993) is introduced. The calcula tion of this measurement is similar to net neighborhood residential dwelling density, but the numerator counts only the dwelling units of one type in a neighborhood. In the current research, it is difficult to get the denominator of NNRDD and/or NNRBTD for all census blocks in Gainesvill e. A modified conception--residential building type density--is employed. The numerat or of this measurement is similar to NNRBTD, which is the dwelling units of one type in the st udied area. The denominator refers only to parcels which are allocated fo r the specified type of dwelling units. Based upon the most popular dwelling types in Gainesv ille, two basic types are included in this research: single family housing and multifam ily apartments. Multi-family residential building type density is very close to â€œstr uctural density,â€ which was defined by Sampson as percentage of units in structures of fi ve or more units (1983), but may be more applicable to be used for zoning codes. 188.8.131.52 Built area intensity For Built area intensity measures, two popul ar concepts are employed: floor area ratio and coverage. Floor area ratio is the ratio of built floor area to the parcel area. Coverage is the ration of the area covered by build ings to the parcel ar ea. In this research, as it is difficult to get data for covera ge, only floor area ra tion is explored. Some evidence suggests that crime rates may be differently related to different types of measured density. For example, Gall e et al. (1972) found th at internal density (number of persons per room) was a stronger correlate of delinquency than was external density (number of persons per square mile ). Booth, Welch, et al. found that â€œareal
69 density (dwelling units per square mile) is rela ted to a greater extent with property crimes than with personal ones, but household crow ding (proportion of all households containing more than one person per room) is little re lated to either in th e large citiesâ€ (1976, p. 303). As different types of density may ha ve different effects on people and their behavior (Gillis 1974), it is important to explore the relationship between residential burglary and different density m easures and find out the best indicator(s) of density and burglary. Again, all density measurements ar e measured on two spatial scales, census block and major street block. 3.6 Analysis Framework and Methods In the previous section, all the environment feat ures to be tested and measurement methods are presented. This section introduces the analysis framework and the analysis methods employed in th e present research. 3.6.1 Analysis Framework This research explores three crime pattern s: complete residen tial burglary, repeat burglary, and near repeat burglary. For each crime pattern, two categories of analysis-spatial-temporal analysis and environment variable analysis--are applied (Figure 3.7). Complete residential burglary refers to all the residential burglary cases from January 2000 to December 2003. The objective of spatio-temporal pattern analysis for complete residential burglary is to ascertai n where and when burglar ies are most likely to happen and to determine the association of sp atial and temporal patterns of residential burglary. Findings of this analysis would be helpful in allocating crime prevention resources strategically. To obtain these spat io-temporal findings, we expect to (a) produce temporal frequency pl ots of burglary cases, (b) id entify clusters of burglary
70 events (hotspots), and (c) compare locations of hotspots across time. Location quotient, a relatively new technique for crime anal ysis, is employed in this stage. Figure 3.7 Analysis framework Produce temporal frequency plots of burglary cases Identify clusters of burglary events, with location quotient and kernel density method Compare locations of hotspots across time, with location quotient and kerneldensitymethod Complete Burglary Identify significant social economic and demographic variables, with stepwise regression method Identify environmental factors which may deter or encourage burglary, with match cases control method Spatial-temporal Analysis Environment Variable Analysis Explore the time course of repeat single family burglary Compare spatial pattern of repeat burglary and hot spots of complete burglary Repeat Single Family Burglary Find out if repeatedly victimized households are different than houses burglarized once, in terms of physical environment features, with two sample t-test and Chi-square test Spatial-temporal Analysis Environment Variable Analysis Explore the existence of near repeat burglary phenomena, with Knox test , Mantel test , and K nearest neighbor test Near Repeat Burglary Identify environmental factors that significantly correlates with near repeat burglary, with two sample t-test and Chisquare test Spatial-temporal Analysis Environment Variable Analysis
71 In addition to the understanding of where a nd when burglaries are most likely to happen, it is also important to know why they occur. Although this research emphasizes the influence of environmental variable s on residential burglary, the impacts of socioeconomic and demographic factors upon crime are emphasized. Match case control methodology is applied in this stage. For the current research, repeat burglary is defined as the same single housing unit that is subject to more than one burglary over a 4-year period. B ecause in many cases it is difficult to identify the precise dwelling unit burgled at multiple-family units, repeat burglary in multiple dwelling units is not e xplored. Analysis for repeat residential burglary will only concentrate on repeat bur glary in single-family, detached housing units. Spatial-temporal pattern analys is of repeat burglaries incl udes two parts, the first is the time course exploration of repeat bur glaries; the second is the exploration of relationship between repeat burglary a nd hot spots of complete burglary. The environment analysis of repeat burglary ai ms to find out if repeatedly victimized households are different than houses burglari zed once, in terms of physical environment features. The two sample t-test and Chi-square test are em ployed to compare variables. Near repeat burglary refers to burglary cases whose occurrences are close in both time and space, such that an area is at highe r risk than would be expected by random distribution. This is distinguished from repeat burglaries, which are defined as the multiple victimization of the same household. The intent of spatia l-temporal pattern analysis for near repeat burglary research is to explore the existence of near repeat burglary phenomena in Gainesville. Three me thods from epidemiology--the Knox test,
72 the Mantel test, and the K nearest neighbor test--are em ployed in this stage. The environmental analysis of near repeat burglary aims to find out what kind of environmental variables are corre lated with near repeat burglary. The two sample t-test and Chi-square test are empl oyed to compare variables. 3.6.2 Analysis Methods Employed To perform all the analyses mentioned above, multiple analysis methods are employed. Some of them are conventional, lik e two sample t-test and Chi-square test. Some of them are relatively new to crime an alysis or urban planning, such as location quotient, the Knox test, the Mantel test, a nd the K nearest neighbor test for spatialtemporal analysis and match case contro l methodology for environment analysis. The following part introduces th ese â€œnewâ€ approaches. 184.108.40.206 Location quotient Location quotient ( LQ ) is a measure used to compare an area's share of a particular type of activities with the refe rence area's share of the same type of activities. Compared to other traditional crime analysis methods, such as hot spot (kernel density) analysis, it has the advantage of being able to incor porate the discontinuous backgrounds. This analysis method is employed to analyze spatia l and spatial-temporal pattern of complete residential burglary in the current research. The terminology of location quotient ( LQ ) is borrowed from the economic literature. It is originally defined as follows: LQi = ( Eij/Ei )/( Eij/ Ei) Where: Eij = economic activity in subarea i department j Ei = total economic activity in subarea i Eij = economic activity of department j in the whole area Ei = total economic activity in the whole area
73 LQs are frequently calculated on the basi s of employment. The interpretation of location quotients is very simple. When a regi onâ€™s LQ for industry j is larger than 1, it can be concluded that local employment is gr eater than expected. These extra jobs then must export their goods and services to non-lo cal areas. Therefore, the region is defined as specialized in industry j. With careful and case-based design, this tec hnique can be a particularly useful tool in crime analysis. For example, it can be used to assess changes of local crime structures over time, or to compare crime structures across localities. Although burglary cases concentrate around certain areas, there can be certain months that these areas suffer burglary victimization disproportionally compar ed to the standard temporal distribution of burglary. 220.127.116.11 Knox test Because near repeat burglary is defined as burglary cases that are proximate both spatially and temporally, the detection of near repeat burglary is the detection of spatiotemporal clusters of burglary cases. Th ree most widely used statistical techniques-the Knox test, the Mantel test, and the K near est neighbor test--are employed as formal procedures for testing space-time interaction. Knox test is a simple and straightforward method for testing space-time interaction which was proposed by Knox (1964). The Knox method quantifies space-time interaction based on critical space and time distances. W ith the Knox method, each case is paired with every other case, so that N cases woul d produce N(N-1)/2 distinct pairs. Pairs of cases separated by less than the critical space di stance are considered to be near in space. Likewise, pairs of cases separate d by less than the critical time distance are deemed to be near in time. This resulted in a 2*2 contingency table cross-ca tegorizing pairs as close or
74 far in space and time. The test statistic, X , is the count of those pairs of cases that are separated by less than the critic al space and time distances. If X is excessive compared to the number expected by chance, this should be an indication that spatial-temporal interaction is present. Despite the popularity of this simple test statistics, there ar e two potential major problems with the Knox method--population shift bias and arbitrary choice of critical space-time distance. Population shift bias refers to the phenomena that the Knox test is biased if the rate of population growth is not constant for all geographic sub areas(Mantel 1967). As population in Gainesville city limit is relatively stable in the studied period, this study does not suffer from population shif t bias. The issue of arbitrary choice of critical space-time distance can be solved by linking these two values to prior empirical findings. As several scholars have found that potential burglars tend to penetrate two to three blocks from crime generators (Luedt ke and Associates 1970; Buck, Hakim et al. 1993), and the average size of street blocks in Gainesville is around 400 feet, we use 800 feet and 1200 feet as the critical distance. In terms of critical time distance, we have some knowledge from repeat burglary study. Polviâ€™ s (1990) research found that the likelihood of a repeat burglary within one month is 12 tim es the expected rate, declining to twice at six months. Another study in Australia found that 65% of repeat burglaries occurred within 1 month of the original event, and 83% within 2 months (Guidi, Townsley et al. 1997). The temporal pattern exploration of rep eat burglary in Gainesville consistent with these findings. To explore the near repeat burglary, we use 1 month (30 days), 2 month (60 days) as the critical time distance.
75 18.104.22.168 K Nearest Neighbor test The K Nearest Neighbor Test is another popu lar method to test space-time clusters of point data. The test statisti c is the count of the number of case pairs that are K nearest neighbors in both space and time. When spatia l-temporal interaction exists, the number will be large since nearest neighbors in space will also tend to be nearest neighbors in time. Although Knox and Mantel tests are al so frequently used, they have some disadvantages. The choice of cr itical distances in the Knox test can be subjective, and Mantel's test is insensitive to nonlinea r associations between the space and time distances. 22.214.171.124 Mantel test Mantel's test (1967) is a widely used method for assessing the relationships between two distance or dissimilarity matrices . When both spatial and temporal distance matrices are provided, this method can dete rmine whether there exists an excess of burglary incidence--a cluster --above what might be expected by chance alone. Mantel's test statistic, Z , is the sum, across all case pairs, of the time distance multiplied by the spatial distance. Z is also called the Mantel product. The advantage of this method is that it circumvents problems associated with selecting the critical spatial and temporal distances for the Knox test by first calculating space and time distance matrices. However, this method is insensitive to nonlinear associations between the space and time distances. For patter n like contagious disease, Mantel recommended transforming the spacetime distances, such as the reciprocal distance (d(ij) + C) ^-1, to reduce the impact of large dist ances. For our near repeat burglary hypothesis, we expect the small sp ace and time distances to be correlated, but
76 not the large distances, like infectious dis ease. Based on this e xpectation, we choose reciprocal transformation method to test the space and time clus ter of burglary incidents. 126.96.36.199 Match case control methodology Burglary cases are more likely to cluster around downtown, low-income communities. Without consideri ng the correlation between bu rglary and social-economic status, we are more likely to identify environm ental variables that significantly correlates with low income community rather than bur glary. To reduce the compounding effect of social-economic and demographic variables, we use match case control methodology. For every burglarized case site, we will select c ontrol site in Gainesville that has similar socioeconomic and demographic characteri stics, but which was not victimized by burglaries. As a result of controlling the so cioeconomic and demographic characteristics, the findings relative to the occurrence of burgl aries will more likely be correlated with environmental attributes, as distinct from the socioeconomic and demographic variables. To use the match case control methodology, the premise is to know what kind of socioeconomic and demographic characterist ics should be controlle d. As most of the socioeconomic and demographic data are av ailable in the census block level, the identification of social economic and dem ographic variables that have significant influence on the burglary rate is applied to th e aggregate data in the census block level. For this analysis, stepwise regression method is employed. The dependent variable is the l o garithm value of burglary rate, which is calc ulated by dividing the burglary counts reported in each block by the number of households in that block. The explanatory variables used in this model were derived from a list of cr ime rate affecting factors, which were advanced by the Federal Bureau of Investigation (FBI, 1999). The complete list is shown below.
77 Population density and degree of urbanization Variations in composition of the populati on, particularly youth concentration Stability of population with respect to residents' mobility, commuting patterns, and transient factors Economic conditions, including median income, poverty level, and job availability Cultural factors and educati onal, recreational, and re ligious characteristics Family conditions with respect to divorce and family cohesiveness Modes of transportation and highway system Climate Effective strength of law enforcement agencies Administrative and investigativ e emphases of law enforcement Policies of other components of the crimin al justice system (i.e., prosecutorial, judicial, correctional, and probational) Citizens' attitudes toward crime Crime reporting practices of the citizenry (Crime in the United States 2002, n.d.) Not all categories of factors in the above list are included in this research. Data are hard to obtain for factors of some categorie s, such as crime reporting practices of the citizenry and citizensâ€™ attitudes toward crime. The factors of some other categories are similar for the whole research area, such as climate. Based on data availability and informative value, the following categories of factors are employed in this research. Population Density Age Composition Gender Composition Ethnicity Composition Stability of Population Economic Conditions Family Conditions
78 These factors can be represented by variab les as the following table (3.4).With the identification of socioeconomic and demogr aphic variables that have significant influence on the residential burg lary rate, control sites that were not burglarized but have similar socioeconomic and demographic charact eristics to burglarized case sites can be identified. Having completed the foregoing steps, identification of control sites, which is the ground of the environmental variable analysis, will be ascertained. Table 3.4 Socioeconomic and Demographic Variables Category of Factor Variable NameDescription Population Density POPDENS population density PCT_17 percentage of 7-17 years in total population PCT_21 percentage of 18-21 years in total population Age Composition PCT_29 percentage of 2129 years in total population Gender Composition PCT_Male percenta ge of male in total population PCTMNRTY percentage of minority Ethnicity Composition HETERO ethnicity heterogeneity, the likelihood that two randomly selected members of a neighborhood are of same origin Stability of Population PCTR ENTER percentage of housing units renter occupied Economic Conditions AVG_ASSD_V average va lue of single familyâ€™s assessed value PCTHLD1M percentage of 1-person household male householder PCTHLD1F percentage of 1-person household female householder PCTMARCH percentage of family households married-couple family w/ own children under 18 yrs PCTMARNOC percentage of family households married-couple family no own children under 18 yrs PCTMHHC percentage of family households other family male householder no wife w/own children under 18 yrs PCTMHHC percentage of family households other family female householder no wife w/own children under 18 yrs Family Conditions AVE_FAM_SZaverage family size
79 3.7 Summary In this chapter, we have built the framewor k for the current research. This research explores three crime patterns: complete re sidential burglary, repe at burglary, and near repeat burglary. For each crime pattern, tw o categories of analysis--spatial-temporal analysis and environment variable analysis --are applied. The anal ysis is applied on multiple scales and units. The microanalysis unit is the site, which is the parcel for singlefamily houses and the house unit for multifamily buildings. There are three scales for environment features that have to be collect ed in aggregate level such as density. They are census block, 1,200 foot buffer zone of the studied site and street blocks defined by major street. we have also built a framework of environmental features that are included in this research and their corresponding indicators a nd measurement methods based on a review of the literature and on the avai lability of suitable data. Feat ures to be explored include permeability , land use and adjacency , and density . As noted above, permeability will be measured by seven variables: 1) traffic capac ity/volume of the closest major arteries, 2) distance to the closet major arteries, 3) the openness to major arteries, 4) street layout patterns, 5) street types ar ound parcel, 6) distance to public transportation stop, and 7) corner location. Land use and ad jacency will be measured by f our variables: 1) adjacent land use types, 2) degree of land use mix, 3) residential units mix, a nd 4) vacant buildings and dilapidated houses. Density will be measur ed by two variables: 1) dwelling density and 2) built area intensity. Territoriality (Newman 1972) and incivilit ies (Wilson and Kelling 1982), which are widely believed to be relevant to burglary and crime, are not in the list. But some
80 environmental features related to these f actors, such as residential unit mix and substandard housing, are incl uded in this research.
80 CHAPTER 4 FINDINGS This research aims to (a) explore the spat ial and temporal patterns of burglary, (b) examine the correlation between burglary and en vironmental variables, and (c) identify specific features of the physical environment that contribute to burgl ary in general and to repeat burglary and â€œnear repeat burglaryâ€ in particular. Findi ngs for three crime patterns: complete residential burglary, repeat burglar y, and near repeat burglary are presented in this chapter. For each crime pattern, two cate gories of analysis--spatia l-temporal analysis and environment variable analysis--are applied. 4.1 Complete Burglary Complete residential burglary refers to all the residential burglary cases from January 2000 to December 2003. Every residential burglary case that occurred during this period will be viewed as one poi nt in this pattern analysis . Addresses that experienced multiple burglaries will have several overl apping points at the same location. Repeat burglary and near-repeat-burglar y incidents are neither dis tinguished nor excluded from the complete data set. 4.1.1 Spatial, Temporal, and Spatial-Temporal Analysis The objective of spatio-temporal pattern an alysis for complete residential burglary is to ascertain where and when burglaries ar e most likely to happen and to determine the association of spatial and tem poral patterns of residential burglary. Findings of this analysis would be helpful in allocating crime prevention resources strategically. To obtain these spatio-temporal findi ngs, we expect to (a) produ ce temporal frequency plots
81 of burglary cases, (b) identify clusters of burglary events (hotspots), and (c) compare locations of hotspots across time. 188.8.131.52 Temporal analysis--Seasonal var iation of residential burglary Reliable estimates of crime seasonality are valuable for law enforcement and crime prevention. Seasonality affects many decisi ons, from long-term reallocation of police officers across precincts to short-term targeting of patrols for hot spots. Researchers have studied the seasonality of crime for more than 100 years, with sometimes contradictory results (Block 1984; Baumer and Wright 1996). Despite variations in the findings of this literature, researchers often point out two conclusions: (a) that property crimes peak in the fall and winter and (b) that violent crimes peak in the summer months (Baumer and Wright 1996). This trend can also be found in Gainesville (Figure 4.1), with two exceptions : one peak in May and one dip in October for the studied period. 0 50 100 150 200 250 300 350 123456789101112 MonthCount Figure 4.1 Seasonal variation of residential burgl ary in general in Gainesville, Florida.
82 One possible reason for the phenomena is th at Gainesville is a college town and residentsâ€™ activities are heavily influenced by the universityâ€™s calendar. May is the month that the spring semester ends and students ar e packing to leave. Students are more likely to leave their apartments doors and windows unl ocked at this period. Burglars have more opportunities to sneak in easil y. Indeed, Gainesville detec tives report and our data confirms that a primary method of entry for burglars is through front doors left unlocked by residents(Schneider and Yang 2003). The bur glary rate drop in October is a common phenomenon for college town although the reas on is unclear. When the temporal pattern of all residential burglary incidents (Figure 4.2) occurred in apartments (where most students are living) were plot ted, it was found that burglary incidents peak at May and December, which coincides with the e nd of Spring and Autumn semesters. Seasonal Variation of Residential Burglary in Gainesville (Apartment)0 10 20 30 40 50 60 70 80 123456789101112 MonthCount Figure 4.2 Seasonal variation of residential burglary for apartments in Gainesville.
83 184.108.40.206 Spatial analysis--Clusters of residential burglary To explore the spatial pattern of residen tial burglary in Gainesville, two analysis methods are applied in this research. The first is the c onventional kernel density, the second is location quotien t. Location quotient ( LQ ) is an indicator that compares an area's share of a particular activity with the reference area's share of the same type of activity. This terminology is borrowed from the economic literature. It is or iginally defined as follows: LQi = ( Eij / Ei )/( Eij / Ei ) Where: Eij = economic activity in subarea i department j Ei = total economic activity in subarea i Eij = economic activity of department j in the whole area Ei = total economic activity in the whole area Compared to kernel density, location quotie nt has the advantage of incorporating the discontinuous background. For example, when estimating the hot spots of robbery by incidents, results of kernel density usually show concentr ation in a city center where people either living or working are concentrated. Results of LQ , which uses population as a denominator, may find that the risk for a lo w-population-density area is higher. Another character of LQ is it has to be applied in aggregate data, wher eas kernel density is applied in individual point data, in most cases. For th e current research, the LQ method is applied in the census block level, with house hold unit as the area denominator.
84 Figure 4.3 Kernel density map and location quotient map for residential burglary in Gainesville. With the kernel density method, two major burglary â€œhot spotsâ€ are identified: (a) one beside the east side of the University of Florida, along University Avenue and 13th Street and (b) the other between Main Street and Waldo Road (Figure 4.3). The first one is where apartments and rental houses ar e concentrated. The second is where most residential real estate assessed values are below $50,000. The LQ map (Figure 4.3) for residential burglary rate, which indicate s burglary counts versus household counts, reveals a pattern a little bit different than that using kernel density. The burglary rate for apartments beside the university are not abnormally higher compared to the overall burglary rate in Gainesville. It is the high dwelling unitsâ€™ de nsity that marks the area as a hot spot. Furthermore, some districts, like the block that is highlighted by the blue circle in the map, are defined as â€œmoderateâ€ in kern el density map but are defined as high risk in LQ map. A thorough exploration (Figure 4.4) of this block finds out that this block includes an isolated community that suffered a high rate of burglary victimization. The relative risk for households in this block is much higher than the standard risk of
85 households in Gainesville, however, as there are not many households in this area, the density of crime incidents is not high. Figure 4.4 Isolated community with high risk 220.127.116.11 Spatial-temporal analysis--Clusters of residential burglary across time Crimes vary enormously over the year. Reliable estimates of crime seasonality are valuable for tactically deploying limited police and crime prevention resources. Crime seasonality has been explored for more th an 100 years (Baumer and Wright 1996). Most of these studies focus on la rge spatial units of aggregation such as city, regional, and national levels (Farrell and Pease 1994). B ecause of the large scale focus, this methodology may mask variations calibrated in smaller units. Some neighborhoods or areas may have large seasonal fluctuations and suffer victimizati on inappropriately in certain months. Intervening in these areas at their relative peak time can be very effective in reducing crime. With the help of LQ technique, seasonality for small areas across Gainesville is mapped. The LQ value is derived by comparing the share of burglary
86 counts of the subarea to those within the city limit, for a specific month. As many census blocks only have one or two burglary incidents during the st udy periods, the analysis is applied on a larger aggregate level--census tract. Kernel density maps for every month are also provided for comparison purposes. The maps from the LQ model clearly reveal that burglary seasonality varies considerably across the space of a city. For example, in October, not only do general burglary rates for Gainesville drop, the LQ values for neighborhoods around stadium are also lower than normal. In December, the mont h that fall semester ends and students are more likely to leave their apartments empty, LQ values for neighborhoods around campus where student apartments concentrated are higher than average. The monthly LQ map can be very helpful for planning and ev aluating monthly police interventions. Figure 4.5 Location quotient for residential burglar ies in October, Gainesville, FL 200020003 Stadium
87 Figure 4.6 Location quotient for residential bur glaries in December, Gainesville, FL 2000-20003 4.1.2 Environment Variable Analysis 18.104.22.168 Identify important soci al-economic variables Although this research emphasizes the in fluence of environmental variables on residential burglary, the impacts of soci oeconomic and demographic factors upon crime are taken into consideration. As noted in Chapter 3, match case control methodology is applied in this research. W ith the help of geographic information system technology, we selected control sites in Gainesville that have similar socioeconomic and demographic characteristics, but were not victimized by burglaries. By controlling for the socioeconomic and demographic characterist ics in the sample selection stage, the potential confounding effects of these va riables can be greatly reduced. To use the match case control methodology, the premise is to know what kind of socioeconomic and demographic characteristic s should be controlled. As noted in chapter Campus
88 3, we employed stepwise regression to fi nd socioeconomic and demographic variables that correlate significantly with burglary rates. Before actually running a regression model to find the important variables, the re lationships between bur glary rate and every individual variable are explor ed. It is found that percentage of male in total population and percentage of housing units occupied by renters have a quadratic relationship with burglary rate. Therefore more variables, square of PCT_Male and square of PCTRENTER, are added into the model. With a stepwise regression, we found the following social variables play an important ro le in census blockâ€™s bur glary rate. They are AVG_ASSD_V, HETERO, PCTHLD1F, PCTHLD1M, PCTMARCH, PCTMNRTY, PCT_Male, PCTRENTER, PCT_21 and POPDEN S. Table below shows the SAS output of stepwise regression that li sts all variables that are significant for the outcome of burglary rate. Table 4.1 Stepwise Regression Model of Social-economic Demographic Variables Variable Parameter Estimate Standard Error Type II SS F Value Pr > F Intercept 0.69061 0.33215 2.83679 4.32 0.0379 AVE_FAM_SZ -0.06684 0.04580 1.39741 2.13 0.1449 AVG_ASSD_V -0.00000423 9.700835E-7 12.49768 19.05 <.0001 Hetero 0.26998 0.13447 2.64521 4.03 0.0450 PCTHLD1F -0.00557 0.00231 3.79701 5.79 0.0164 PCTHLD1M -0.00617 0.00230 4.72581 7.20 0.0074 PCTMARCH -0.00840 0.00283 5.79122 8.83 0.0031 PCTMNRTY 0.00327 0.00102 6.79543 10.36 0.0013 PCTMale -9.60896 1.03613 56.43542 86.01 <.0001 SQPCTMale 9.73373 0.97128 65.90141 100.43 <.0001 PCTRenter 0.00975 0.00370 4.56748 6.96 0.0085 SQPCTRenter -0.00011593 0.00003851 5.94557 9.06 0.0027 PCT_21 -0.00501 0.00251 2.61145 3.98 0.0464 POPDENS -0.01753 0.00398 12.71066 19.37 <.0001
89 22.214.171.124 Environment variable analysis With the identification of socioeconomi c and demographic variables that have significant influences on the residential bu rglary rate, control sites that were not burglarized but have simila r socioeconomic and demographic characteristics to burglarized case sites can be id entified. All residential parcel s would inherit most of the socioeconomic and demographic factors value from the census blocks that the parcels belong to. However, because the average as sessed value that repr esents the economic condition is found to be significant to the bur glary rate and should be controlled, the control procedure will use the parcelâ€™s ow n assessed value rather than the average assessed value of all single-family residentia l parcels in the census block. To find the control site for every burglarized case site, three conditions were applied. First, the site should not have been burglarized from January 2000 to December 2003. Second, the siteâ€™s controlled socioeconomic and demographi c factorsâ€™ value is within in a certain range of the case site. Th ird, the site is not within the sa me census block as the case site. From all the residential parcels that can fulfill these three conditions within the Gainesville city limit, one of them would be randomly selected as the control site. With all the three conditions app lied, 2,867 burglarized sites among 3,100 cases were found to match sites, whereas 233 cases, which consti tute 7.5% of all burgl ary cases, cannot be matched. The environment analysis includes analysis of variables from three categories. Most of these variables can be calculated by computer automatica lly, e.g., distance to public transportation. However, some of them have to be defined case by case, e.g., street type around parcel. For those variables that can not be defined automatically, values were assigned manually to 310 cases that are identified by randomly sampling from the 3,100
90 reported residential burglary crime sites in Gainesville. Paired t analyses are rendered on these 310 cases and their control sites. Because only 286 of 310 cases have match sites, the analyses were applied on th ese matched pairs of cases. With the matched case control methodol ogy, the potential confounding effects of socioeconomic and demographic characteris tics are greatly reduced. As a result, the findings relative to the occurren ce of burglaries are more likely to be correlated with environmental attributes, as distinct from the socioeconomic and demographic variables. The environmental features to be explored include permeability , land use and adjacency , and density . The analysis output for these three cat egories of variables are demonstrated below. 1) Permeability Permeability is measured by seven variables: (a) distance to the closet major arteries, (b) street layout patterns, (c) st reet types around parcel, (d) distance to public transportation stop, (e) corner location, (f) block length, and (g) connectivity index. Among the seven variables, four are continuous variables, whereas the other three--street layout patterns, street types, around parcel and corner location --are categorical variables. For continuous variables, paired t tests were used to examine whether the differences between burglarized and unburgl arized dwelling-unit pairs ar e statistically significant. Categorical variables were det ected by chi-square tests. Table 4.2 presents the paired t test results for the four continuous variables. The mean value is the mean for all paired differe nces, which is the resu lt of the burglarized case sitesâ€™ value minus the unburglarized control s itesâ€™ value. This tabl e illustrates that all of the permeability continuous variables are statistically different between matched pair
91 sites. The table suggests that burglarized control s ites tend to be located far away from major arteries and public transportation stops. Meanwhile, the average block length around burglarized dwellings tends to be shorte r than that of unburglarized control sites. The connectivity indexes around bur glarized houses are larger th an those of control sites. Both shorter block length and larger conn ectivity index suggest better connectivity. Table 4.2 Permeability Indicators Mean of difference Standard deviation T Value Pr > |t| Distance to the closet major arteries -233.27 801.97 -15.57 < 0.0001* Distance to public transportation -218.52 1858.1 -6.30 < 0.0001* Block length -129.09 3179.5 -2.17 0.0298* Connectivity 0.054 0.3327 8.68 < 0.0001* Table 4.3 lists the counts of residential units by street layout pattern and whether or not those units were burglarized. Pearsonâ€™s chi-square test was used to compare the distribution of burglarized cas e sites and control sites among different layout patterns. The p value of < 0.0001 indicates a significant relationship betw een street layout pattern and burglary. Standardized Pearson residuals we re also calculated and given in the table. Those with values exceeding 2 to 3 in absolute values are more likely to be significantly different from the expected values under th e condition that the stre et layout pattern and burglary are independent. From Table 4.3 we can see that th e only street type pattern associated with burglarized resi dential parcels is the gridir on street pattern. Fragmented parallel, loops and lollipops, and lollipops on a stick tend to have a protective effect. There is no significant difference between matc h pair sites for warp ed parallel. Figure 4.7 shows the distribution of burglarized and control sites among the five street layout patterns.
92 Table 4.3 Distribution of Burglarized Site s and Control Sites among Street Layout Patterns Observed frequency Expected frequency Standardized Pearson residuals Burglarized sites Control sites 1120 695 941 874 Gridiron 11.8 -8.8 1292 1383 1387 1288 Fragmented parallel -4.6 5.3 227 187 214 199 Warped parallel 1.3 -1.2 253 311 292 271 Loops and lollipops -3.2 3.7 175 272 232 215 Lollipops on a stick -4.9 6.5 *Pearsonâ€™s chi-square = 125.56, df = 4, p < .0001 0 200 400 600 800 1000 1200 1400 1600 GridironFragmented Parallel Warped Parallel Loops and Lollipops Lollipops on a StickFrequencies Burglarized Sites Control Sites Figure 4.7 Distribution of burglar ized sites and control sites among street layout patterns.
93 Table 4.4 lists the distribution of residentia l units by street type and whether or not they have been burglarized. Observed value, expected value, and standardized Pearson residual are listed. With the Pearsonâ€™s chi-square test, the p value is < .0001, which indicates a significant relations hip between street type and burglary. Table 4.4 also shows that burglarized house units are more likely to be located beside through-traffic streets and less likely to be located beside dead-e nd streets and T-type streets. The opposite relationship for control sites also exists. This implies th at through-traffic streets may contribute to burglary whereas T-type streets and dead-end streets can be protective. However, there is no clear relationship betw een L-type streets and burglary. This is inconsistent with the study applied by Bevi s and Nutter (1977), which claims there is a clear relationship between street type and burglary by the degree of permeability, which is increased with the order of dead-end, cul-de -sac, L-type streets, T-type streets, and through-traffic streets. Table 4.4 Distribution of Burglarized Site s and Control Sites among Street Types Observed frequency Expected frequency Standardized Pearson residuals Burglarized sites Control sites 24 58 42 39 Dead-end streets -3.5353 6.1533 64 72 71 65 L-type streets -1.2361 1.4056 72 104 92 84 T-type streets -3.0327 4.1680 150 52 105 96 Through-traffic streets 12.0261 -5.8551 *Pearsonâ€™s chi-square = 67.07, df = 3, p < .0001
94 0 20 40 60 80 100 120 140 160 Dead endL-type streetsT-type streets Through-traffic streets Frequencies Burglarized Sites Control Sites Figure 4.8 Distribution of burglarized site s and control sites among street types. To understand if corner lots are at higher ri sk of residential burglary than those in the middle of a block, chi-square tests are also applied. The result (Tab le 4.5.) shows that there is a strong correlation betw een burglarized residential un its and a corner location (p = 0.0021). Thus, the evidence suggest s that burglarized lots are more likely to be located at corners. And this finding is compa tible with the Brantingham's research. Table 4.5 Distribution of Burglarized Sites a nd Control Sites Between Corners or Middle Block Lots Observed frequency Expected frequency Standardized Pearson residuals Burglarized sites Control sites 113 71 96 88 Corner 3.57 -2.71 197 215 214 198 Middle of block -2.71 3.57 Pearsonâ€™s chi-square = 9.42, df = 1, p = 0.0021
95 2) Land use and adjacency Land use and adjacency were measured by four variables: (a) adjacent land use types, (b) degree of land use mix, (c) resident ial units mix, and (d) vacant buildings and dilapidated houses. a) Adjacent land use types Adjacent land use types were measured by two approaches, the presence of each specific type of land use and the percentage of lots dedicated to each specific land use in the studied sitesâ€™ 1,200-foot buffer zone. These two variables measure the existence and extent of each type of land use. All 100 t ypes of land use coded in the Alachua county appraiser officeâ€™s parcel data were screened. Table 4.6 show s all types of adjacent land use that were detected to be significantly different between burglar ized sites and match sites by one or both measurement approaches. Table 4.6 Land Use Type Indicators Code L and Use P r > |t| by p ercentage T Value by p ercentage P r > |t| by p resence T Value by p resence 1 Single Family 0.4722 0.72 0.0007 -3.4 2 Mobile Homes 0.9811 0.02 <0.0001 -4.68 3 Multi Family <.0001 6.62 <0.0001 7.05 8 Multi Family less than 10 <.0001 7.36 <0.0001 5.04 9 Undefined reserved for DOR 0.0002 -3.78 <0.0001 -11.83 10 Vacant Commercial 0.0148 2.44 <0.0001 8.82 11 Stores One-Story <0.0001 4.06 <0.0001 9.72 12 Mixed Use, i.e., Store and Office<0.0001 4.46 <0.0001 5.06 13 Department Stores (2 sites) 0.3058 1.02 0.0004 3.58 14 Supermarket (7 sites) <0.0001 4.37 <0.0001 4.6 16 Community Shopping Center 0.1849 -1.33 0.0165 2.4 17 One-story Non-Professional Offices <0.0001 6.36 <0.0001 7.14 18 Multi-story Non-Professional Offices 0.1226 1.54 <0.0001 5.34 19 Professional Service Buildings 0.0156 2.42 0.0037 2.91 21 Restaurants, Cafeterias <0.0001 4.53 <0.0001 6.46 22 Drive-in Restaurants <0.0001 5.76 <0.0001 6.79
96 Table 4.6 Continued Code L and Use P r > |t| by p ercentage T Value by p ercentage P r > |t| by p resence T Value by p resence 23 Financial Institutions 0.0077 2.67 0.0003 3.6 24 Insurance Company Offices <0.0001 4.15 <0.0001 4.77 25 Repair Service Shops 0.4103 0.82 <0.0001 4.77 26 Service Stations <0.0001 6.85 <0.0001 7.24 27 Automotive Repair, Service and Sales 0.0013 3.23 <0.0001 6.21 28 Parking Lots, Mobile Home Sales <0.0001 5.13 <0.0001 8.27 29 Wholesale, Manufacturing, and Produce Outlets 0.0136 2.47 0.2159 1.24 30 Florist, Green House 0.0054 2.78 0.0204 2.32 32 Enclosed Theaters, Auditoriums (1 site) 0.2496 -1.15 0.0321 -2.14 33 Night Clubs, Bars, and Cocktail Lounges 0.0102 2.57 <0.0001 4.51 35 Tourist Attractions (1 site) 0.0083 -2.64 0.0881 -1.71 36 Camps (1 site) <0.0001 -6.84 <0.0001 -6.85 39 Hotels, Motels 0.3663 0.9 <0.0001 6.68 41 Light Manufacturing 0.1124 1.59 0.0011 3.27 45 Canneries, Distilleries, and Wineries (1 site) 0.0001 3.88 <0.0001 3.9 48 Warehouses, and Distribution Centers 0.2221 1.22 <0.0001 5.95 70 Vacant Institutional (1 site) 0.0151 -2.43 1 0 71 Churches 0.7245 -0.35 0.016 2.41 73 Private Hospitals (1 site) 0.0455 2 0.0455 2 74 Homes for Aged 0.1484 1.45 0.0315 2.15 75 Orphanages 0.8622 -0.17 0.0004 3.53 76 Mortuaries, Cemeteries 0.01 2.58 0.0011 3.27 77 Clubs, Lodges, and Union Halls 0.0243 2.25 <0.0001 5.12 82 Forest, Park, and Recreational Areas (6 sites) 0.0303 -2.17 0.1228 1.54 83 Public Schools <0.0001 5.13 0.0002 3.77 84 Colleges (1 site) 0.7222 0.36 <0.0001 5.07 85 Public Hospitals (5 sites) 0.0027 3 0.0009 3.33 86 Other Counties 0.1663 -1.38 0.001 -3.29 87 Other State <0.0001 4.13 0.0009 3.32 88 Other Federal 0.0464 1.99 0.0315 2.15 89 Other Municipal <0.0001 4.93 0.0001 3.8 91 Utilities 0.0209 2.31 0.4868 -0.7
97 According to Table 4.6, at the 0.05 level, 30 types of land use are identified as strongly correlated with burgl ary incidents by the percentage value, whereas 43 types of land use are identified by the existence Bool ean value. Among these land uses, 11 types of land uses have less than 10 sites. Th e test result can be biased easily by other environmental factors correlated with these few sites. We cannot make conclusions for these types of land uses, although the statistica l test showed they have strong correlation with burglarized dwelling units. Excluding thes e land uses with few sites, 23 types of land use are identified as contributing to burglary incidents by both tests. They are offices, service stations, automotive repair, se rvice and sales, parking lots, mobile home sales, florist, green house, night clubs, bars , and cocktail lounges, canneries, mortuaries, cemeteries, clubs, lodges, and union halls, public schools, other state, other federal, and other municipal. Most commercial land uses, except regional shopping malls, are on the list of sites contributing to burglary incident s by either both or one of the extent and presence tests. There are only 7 lots coded as supermarket, whereas Gainesville has many more supermarkets than this count. In a further da ta review, it was found th at all the 7 lots are located at middle to east side of the city, which is generally less safe relative to crime incidence than the west side. The coding i ssue may bias the result. Furthermore, we found the other supermarkets are coded as comm unity shopping center. In the extent test, it is found that a community shopping center can be protective, although not significantly. Although there is an argument that mixing residen ces and business will enhance surveillance, there is also a statement that juxtaposing commercial and residential land
98 uses may expose the residential locales to outsiders, which may raise crime risks. According to the table, there is a trend that la nd uses that attract a small to medium traffic flow, such as cafeteria, services, and parking lo ts may be "unbalanced" as to the stream of "ousiders" vis a vis the necessary local survei llance to deter or prevent burglaries. At the same time, land uses that attract large amount s of activities, such as local shopping malls, large scale supermarkets and movie theate rs, may provide surveillance and enhance safety. The spatial-temporal analysis that found the burglary rate drops during the football season and drops even further for the district around the stadium during the same time period also supports this conclusion. Multifamily sites, which include two types of land use code--Multi-family and Multifamily less than 10 units--are found to highly as sociated with burglarized residential units by both extent and presence test (p < 0.0001). Th is finding is consistent with Sampsonâ€™s (1986) study, which found the percentage of multi residential unit was a significant positive predictor of victimization rates, regardless of other explanatory factors such as racial composition and poverty. The presence of public schools (26 sites) al so tends to increase the risk of burglary, in Gainesville, whereas the presence of pr ivate schools (40 sites, most of them are kindergarten or daycare) is not statistically significant. Th e reason is obvious. Residential units close to the University of Florida (the only college site with in the city limit) are more likely to be burglarized. However, this district also carries other environment features that are correlated with burglary, su ch as higher percentage of multifamily units and higher accessibility. As we have stated before, test resu lt can be biased easily because
99 of other factors correlated with the only campus site. The relationship between college and residential burglary is unclear. b) Degree of land use mix The degree of land use mix consists of two measurements: general land use mix and relative land use mix. The two measurements are measured in two scales: census block and major street block. Table 4.7 shows the result of a paired t test. Again, the mean value is the mean for all paired differe nces, which is the output of the burglarized case sitesâ€™ value minus the unburgl arized control sitesâ€™ value. According to Table 4.7, at both the census block level and the major-street block level, burglarized case sites tend to have a higher general land use mix value than do paired control sites. Relative land use mix, which measures the ratio of nonresidential types of land use to housing units in the sket ch area, is significantly correlated with burglary cases in major street level. Table 4.7 Land Use Mix Indicators Mean of difference Standard deviation Pr > |t| t value General land use mix in census block level 0.0090.23510.03192.15 Relative land use mix in census block level -0.11345.50540.2702-1.1 General land use mix in major street block level 0.0110.1499<. 00013.98 Relative land use mix in major street block level 0.6315.0678<. 00016.66 c) Residential units mix Residential units mix is measured by two measures: the percentage of residential land devoted to multifamily housing and the coefficient variation of housing value, in two different scales--census blocks and major street blocks. Table 4.8 lists the result of paired
100 t test. It is interesting to fi nd out that the coefficient of variation of the housing value is significant at the both the censu s block level and the major-street block level, however in opposite directions. Table 4.8 shows that burgla ry cases are more likely to occur in the major-street block level with a higher degree of intermixture of varied-housing value and in a census block with a lower degree of in termixture. With a further review, we found out that there were 583 burglarized cases occurring in census blocks, with only one residential parcel. The coefficient value of the variation of the housing value for these census blocks is 0, whereas only 235 contro l cases occurred in such census blocks. Because residential uni ts with no neighbor in the same block may imply secluded houses, the analysis results can be biased. All pairs whose coefficient variation value for either case site or control site is 0 are excluded fo r a further test. This new test contains 2346 pairs, the p value is 0.0028 and t value is positive, which implies that when secluded houses are excluded, burglary cases are more likel y to occur in census blocks with higher degree of intermixture of varied-housing value. Table 4.8 Residential Units Mix Indicators Mean of difference Standard deviation Pr > |t| t value Percentage of residential land devoted to multifamily housing in census block level 0.0460.2601<0.00019.46 Coefficient of variation of housing value in census block level -0.01650.24280.00003-3.63 Percentage of residential land devoted to multifamily housing in major street block level -0.00190.20370.6253-0.49 Coefficient of variation of housing value in major street block level 0.0310.2146<0.00017.63 Coefficient of variation of housing value in census block level (excluded cases with value 0) 0.0130.21540.00282.99
101 Table 4.8 also shows that the percentage of residential land de voted to multifamily housing correlates with burglary cases in th e census block level, but not in the majorstreet level. This implies place single-fam ily and multifamily housing in the same block may increase burglary risks. However, the da ta suggest that mixi ng different types of homogeneous blocks at a larger level, e.g., ma jor street block, will not negatively affect burglary risks, but may enhance safety. d) Substandard houses Table 4.9 lists the distributi on and test results of resi dential units by relationship with substandard houses and whether or not th ese units have been burglarized. According to Table 4.9, about 11.40% of burglary occu rred within substa ndard dwelling units, whereas only 4.67% of matched control sites are located on substandard dwelling units. The difference is highly signifi cant. The evidence suggests that burglary incidents are more likely to occur in dwelling units that ar e substandard or adjacent to such dwelling units than unburglari zed control sites. Table 4.9 Distribution of Burglarized Sites and Control Sites among Relationships With Substandard Dwelling Units Observed frequency Expected frequency Standardized Pearson residuals Burglarized sites Control sites 285 110 203 192 Is substandard dwelling unit 11.53 -7.11 146 88 120 114 Adjacent to substandard dwelling unit 3.91 -3.09 2069 2159 2176 2052 Away from substandard dwelling unit -7.75 11.68 a) Pearsonâ€™s chi-square = 89.69, df = 2, p < .0001
102 0 500 1000 1500 2000 2500 Is SubStandAdjacent to SubstandOtherFrequency Burglarized Sites Control Sites Figure 4.9 Distribution of burglarized sites and control sites among relationships with substandard dwelling units. 3) Density Density was measured by two variables, pa rcel density and floor area ratio, at both the census block level and the major-street block level. Table 4.10 shows the paired t test results for the four continuous va riables. Contrary to the widely held belief that increases in density lead to crime, evidence suggests th at there is a signifi cant correlation with burglary cases and low density. Table 4.10 shows that low lot density in both the census block and major street block level are asso ciated with burglary cases. The relationship between burglary and floor area ratio at censu s block level and majo r street block level are opposite. The data shows that burglary case s are more likely to be located in census blocks with higher floor area ra tio and major street blocks with lower floor area ratio. Table 4.10 Density Indicators Mean of difference Standard deviation Pr > |t| t Value Lot density in census block level -1.2267 10.1 < .0001 -6.50 Floor area ratio in census block level 1647 18033 0.0034 2.93 Lot density in major street block level -0.0368 1.8088 0.0023 -3.05 Floor area ratio in major st reet block level-100.34 3414.9 0.0004 -3.53
103 4.2 Repeat Burglary In the current research, the term repeat burglary is defined as the same single residential unit that is subject to more than one burglary over a 4-year period. Because in many cases it is difficult to identify the precise dwelling unit burgled at multiple-family units, repeat burglary in multiple dwelling units is not explored in this research. Analysis for repeat residential burglary concentrates only on repeats in single-family, detached housing units. To identify single-family repeat burglary, only record s that are geocoded by the parcel address matching method--in othe r words, records that can be linked to parcels--are considered. Among the parcel-l inked records, burglary incidents that occurred in parcels with a land use code fr om appraiser data of 01, which represents a single family, were identified as single-family burglaries. Once the data set was cleaned as much as possible, repeat single-family burglaries were identifie d by counting incidents that occurred in each distinct victimized address. This was performed with the Group function in Access. Burglary records that c ould not be linked to parcels or with missing land use code were not considered in this procedure. Before disc ussing the results for repeat burglary, it is worth no ting that the estimated calculati ons are conservative. Figure 4.10 shows the distribution of si ngle family repeat burglary and single family parcel. Table 4.11 Repeat Single-family Burglary (Non Single-family Burglary Excluded) Times burglarized Number of single family addresses Percent of single family addresses Number of burglary incidents Percent of burglary incidents 0 20,54093.530 0 1 1,2505.691,250 76.27 2 1320.60264 16.11 3 310.1493 5.67 4 or more 70.0332 1.95 Total 21,9601001,639 100
104 Figure 4.10 Single family repeat burglary in Gainesville As can be seen from Table 4.11, the tota l amount of single-family burglary was experienced by 6.47% (1,420) of all 21,960 addr esses and 23.73% (389) incidents of burglary were experienced by 11.97% (170) vict im addresses (two or more for each household). Furthermore, 13.36% (219) incident s occurred in addresses burglarized at least once. This means if all repeat burglary can be eliminated, bur glary rate for singlefamily houses can reduce 13.36%. According to the formula described by Trickett et al. (1992), the expected rate of revictimization is 3.36%. The observed rate is 3.7 times the expected rate. This figure is consistent w ith research conducted in Britain (Forrester, Chatterton et al. 1988; Johnson, Bowers et al. 1997) and Canada (Polvi, Looman et al. 1990) which reported that observe d rates of repeat victimizat ion are generally less than 4
105 times the expected rate. Townsleyâ€™s (Towns ley, Homel et al. 2000) re search in Australia claimed that the chance of re victimization is just twice th e initial chance of becoming a victim, lower than most other studies. Howe ver, the authors are comparing two observed values rather than comparing the observed va lue to expected value, as are other studies. According to the data provided by Townsley et al., the observed rate is also 3.7 times the expected rate. Because research applied in different time and geographical locations found consistent results, the finding of the elevated risk of repeat burglary may be a reliable and generalizable finding. 4.2.1 Spatial and Temporal Analysis 126.96.36.199 Temporal analysisâ€“time course of repeat burglary Temporal analyses of revictimization ar e concerned primarily with the amount of time that elapsed between repeat events. Before we run into the actual analysis, there are two issues that need to be addressed. First, it is unlikely to know the exact time of a burglary offense. In the police crime-recordi ng system, two dates: FromDate and ToDate, are recorded for each burglary incident. The duration can vary from several hours to a number of weeks or longer, if the victims have been on vacation. This creates problems when one wishes to calculate the duration between two incidents. To solve this issue, the middle point of the FromDate and ToDate is used in the calculation. The second issue is the time window effect, which refers to th e underestimation of repeat victimization and the short periods bias of the time course distribution, due to using a set time period. For example, Farrell et al. (2002) found that â€œa one year time window captures 42% more repeats than a six-month window [while] a th ree year window captures 57% more repeats than a one year time windowâ€ (2002, p. 19). Furthermore, the short-period repeat burglaries are more likely to be detected than are large-duration repe ats. For example, in
106 the 48-month data set, a repeat occurring with an interval of 1 month can be identified if the initial incident occurs in any time fr om the first month to the 47th month of the sample. However, repeats with a 24-month pe riod can be identified only if the initial event occurs during the first 2 years of the sample. The l onger the time between burglary incidents, the less likely it is that repeats would be identifi ed. There are two methods to overcome the short period bias. The first method is adjusting the time course frequencies by a correction factor. The correction factor can be derived according to the following formula: C = ( T + R ) / ( T R ) Where C is the correction factor applied to time course counts T is the total number of time units of observation R is the number of time units in which repeats are counted. The correction factor increases as the durat ion between repeat in cidents elapses. By multiplying the raw counts for the time c ourse with the correction factor, the undercounted frequencies of lo ng duration can be compensated. The second method to compensate short-period bias is providing data additional to the time period studied. For example, if the maximum repeat interval studied is 6 months, then an additional 6 months of data, which record incident s that occurred before or af ter the period studied, would be provided so that all repeats can be identified reliably. In the current research, the correction factor method is applied to solve the time window issue.
107 0 5 10 15 20 25 30 351 4 7 10 13 16 19 22 25 28 31 34 37 40Months Between IncidentsFrequencie s Actual Adjusted Figure 4.11 Time course for single family repeat burglary in Gainesville (Month). Figure 4.11 shows the revictim ization time courses, both actual and adjusted, for single-family residential burglar y in Gainesville. Most featur es displayed by the graph are consistent with all time course distribu tion research ever published, whereas some differences also exist. These features are: 1. Although repeat victimization tends to occu r soon after the first incident, repeat residential burglary for single-family housing tends to have a short break rather than occurring immediately afte r the prior burglary. As in most other relevant research, this temporal pattern exploration for repeat burglary also found most repeat burglaries occur within 1 to 2 months of the first victimization and the risk of repeat burglary decreases quickly after the first 2 months. However, unlike the risk found in other studi es, the risk of being burgled again was highest in the 2nd month after the initial burglary rath er than the 1st month. To have a better understanding about the heightened risk of revictimization in the very short time
108 after the initial offense, a higher degree of resolution, the week, is used to investigate time course of repeat burglary (Figure 4.12). 0 5 10 15 20 25 30 35 40 450 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 51Weeks between IncidentsFrequencie s All Residential Burgalry Single Family Residential Burgalry Figure 4.12 Time course for residentia l burglary in Gainesville (Week). Inconsistent with almost all of the pion eer research that found â€œthe risk of revictimization is greatest in the period im mediately after the initial offenceâ€(Polvi, Looman et al. 1990; Robinson 1998), data from Gainesville demonstrat e that it is the 3rd week after the initial offense that the risk of revictimization is highest. This finding conflicts with Polvi et al., who conclude that â€œaround 28 percent of repeat burglaries occur on the same day or adjacent days [and] half occur within seven days and 74% within sixteen daysâ€ (Polvi, Looman et al . 1990,p. 11). It also differs from a study in Tallahassee, which states that percent of the burglary re-vic timizations occurred within one week of the initial offen ce, while 51 percent occurred within one monthâ€(Robinson 1998). It is not clear what cau ses the difference. One possible reason is that the previous research did not distinguish single-fam ily housing burglary from other
109 residential burglary such as burglary in apartments and multifamily residential buildings. Furthermore, for these previous studies, burgl aries that occurred in the same building but not the same units are also considered as repeat. When all reside ntial burglaries are considered, including burglaries in the multiun it residential buildings such as apartments that are considered as repeats, the time fr ame of repeat burglary in Gainesville is consistent with other analyses. 0 5 10 15 20 25 1357911131517192123 Months between IncidentsFrequency Observed Expotentia for Observedl 0 5 10 15 20 25 30 35 1357911131517192123 Months between IncidentsFrequency Adjusted Expotential for Adjusted Figure 4.13 Time course for residentia l burglary in Gainesville (Month). 2. The time course of repeat single-family residential burglary vi ctimization conforms to an exponential model, consistent with other studies.
110 To explore the pattern of repeat victimi zation frequencies curve, both the adjusted and actual frequencies are used. Moreover, beca use counts for repeat burglary more than 2 years in the past are continuously less than 5, and the adjusting factor can be as large as 11 times, the estimation may be biased by the adjusted factor. Thus, only frequencies for the time interval up to 2 years are used. The observed repeat victimization curve is clearly almost exponential, a trend also observed in other research (Spelman 1995; Johnson, Bowers et al. 1997; Townsley, Homel et al. 2000). The expone ntial regression can explai n over 65 percent of the variance (R-squared = 0.6517, p < .0001). However, the exponential regression can only explain over 20 percent of the variance for the curve of adjusted frequencies. Figure 4.13 shows the relationship between observed, adju sted, and predicted counts for repeat burglary incidents. 3. Two â€œhumpsâ€ are present for single-family re peat residential burgl ary in Gainesville, whereas most previous research found only one â€œhumpâ€ at around 4 to 5 months. Two â€œhumps,â€ increasing in frequency afte r risk of revictimizations diminishes markedly over time, have been observed in Gainesville. One is around 9 to 10 months, another is around 15 months. Although most re search about the time course of repeat burglary have also observed the â€œhump,â€ th ese researches found only one â€œhump,â€ at around 4 to 5 months (Polvi, Looman et al . 1990; Polvi, Looman et al. 1991); Robinson, 1998; (Townsley, Homel et al. 2000). It may be b ecause little previous research used data across more than 2 years. Actually, In Town sleyâ€™s observation (T ownsley, Homel et al. 2000), the â€œhumpâ€ at 16 months can barely be observed (Figure 4.14).
111 The reason for â€œhumpâ€ is not clear; it has b een speculated that the period of increased risk may reflect an â€œinsurance effectâ€: if householders are insured and replace their goods, after certain months the burglar can be confident that new goods are available. Figure 4.14 Time course for residential prope rties, Beenleigh, June 1995 to November 1996 (inclusive ). Source: (Townsley, Homel et al. 2000) 188.8.131.52 Spatial analysisâ€“repeat single -family burglary in hot spots Having established the temporal pattern of repeat burglary in Gainesville, the spatial pattern was also investigated. Hot s pots of single-family re sidential burglary are detected by the density function in ArcGIS . Figure 4.15 shows the relationship between single-family burglary and hot spots. Table 4.12 lists the numbe r of repeat single-family burglary and whether or not th e address is in the hot spot s. According to Chi-squared statistics, the expected valu e under the independence hypothesi s for repeat single-family burglary inside a hot spot is 112, Chi-squared value is 9.71, P value is less than 0.01, indicating a significant differen ce in rates of repeat burgl ary inside and outside of hotspots. This conclusion is consistent with the previous study: Hot spots have a higher proportion of repeat burglary in cidents than do non-hot-spot ar eas, and the relationship is statistically significant (Trick ett, Osborn et al. 1992).
112 Figure 4.15 Repeat single-family burglary and hot spots of all single-family burglary incidents. Table 4.12 Repeat and Non-repeat Single-family Burglary by Inside/Outside Hot Spot Repeat single-family burglary Nonrepeat single-family burglary Inside hot spot 145327 Outside hot spot 244923 Figure 4.16 shows the hot spots of single-fa mily residential burglary when all subsequent cases of repeat burglary incidents are eliminat ed. It is observed that eliminating repeat victimization may diminish a crime hotspot. This map implies that it is more heavily victimized addre sses rather than more addre sses victimized that accounts for the main reason that the burglary rates are so high in some areas.
113 Figure 4.16 Hot spots of single-family burglary with and without repeat. 4.2.2 Environmental Variable Analysis The environment analysis of repeat burg lary aims to find out if repeatedly victimized households are different than hous es burglarized once, in terms of physical environment features. Osbornâ€™s research found â€œl ittle evidence that repeat victimization have distinctive characteristics compared with single victimsâ€ (Osborn, Ellingworth et al. 1996). However, Osbornâ€™s research concentrat ed more on socioeconomic characteristics such as length of residence, ethnicity, a nd age of household than on environment factors, which are the central interest of our analysis. We did not employ case control methodology for repeat single family and environment variables analysis as Osbornâ€™s study already pr oved there is no significant difference between multiple and single burglary victims. The two sample t-test rather
114 than the paired t-test is used to compare co ntinuous variables. Ca tegorical variables are still compared by Chi-square test. 184.108.40.206 Permeability Table 4.13 Permeability Indicators for Repeat Burglary Mean value of single burglary site Mean value of multiple burglary site Difference between mean values T value by pooled method Pr>| t| by pooled method T value by Scatterthwaite method Pr>| t| by Scatterthwaite method Distance to the closest arteries 757.88 664.24 93.642 2.99 0.0028* 3.21 0.0014* Distance to public transportation 897.89 716.78 181.1 3.90 0.0001* 4.41 <0.0001* Block length 1364.8 1229.9 134.86 1.42 0.1568 1.38 0.1685 Connectivity 1.4972 1.551 -0.054 -5.11 <0.0001* -5.47 <0.0001* Table 4.13 illustrates the two sample t-test results for the c ontinuous permeability indicators. Mean values for single burglary group and multiple burglary group, difference for mean value of the two groups and t va lue, p value for both the pooled method and Scatterthwaite methods are list ed. According to the table, mu ltiple burglary sites tend to locate closer to arteries, clos er to public transportation and have larger connectivity figure. However, the differentiation of bloc k length between the two groups is not significant. Table below (4.14) lists the counts of single burglary and repeat burglary by street layout and the output of Pears onâ€™s chi-squared test. The p value of 0.0018 indicates that there is strong interact ion between street layout pattern and repeat burglary. Single-family households locate in areas with grid networks are more likely to be victims of multiple burglary than households in areas with other t ypes street layout. In the paired t-test for burglarized sites and matched unburglarized sites, fragmented parallel, loops and
115 lollipops, and lollipops on a stick were found to have a protective effect. In the current comparison, the protective effect cannot be observed again. This finding is similar to Osbornâ€™s research which stated households w ith protective characte ristics have this protection reduced for a subsequent ev ent(Osborn, Ellingworth et al. 1996). 0 0.1 0.2 0.3 0.4 0.5 0.6 GridironFragmented Parallel Wraped ParallelLoops and Lollipops Lollipops on a StickPercentage Single Burglary Multiple Burglary Figure 4.17 Distribution of singl e burglary sites and multiple burglary sites among street layout patterns
116 Table 4.14 Distribution of Single Burglary Sites and Multiple Burglary Sites among Street Layout Patterns Observed frequency Expected frequency Standardized Pearson residuals Single Burglary Sites Multiple Burglary Sites 391 157 418 130 Gridiron -2.97 3.62 702 206 692 216 Fragmented parallel 1.14 -1.07 61 11 55 17 Warped parallel 2.15 -1.62 36 6 32 10 Loops and lollipops 1.88 -1.36 60 9 53 16 Lollipops on a stick 2.89 -1.97 *Pearsonâ€™s chi-square = 17.1373, df = 4, p=0.0018 Table 4.15 is the cross tabulation of street type and multiple or single burglary. P value for Pearsonâ€™s Chi-test is 0.24, which suggests that there is no statistically significant relationship between multiple burgl ary and street type around the parcels. Because the observed frequency for repeat bu rglary around dead end streets is 0, the chisquare test cannot find the e xpected frequency for this condition and we are unable to make conclusions. Furthermore, the protec tive effect of T-type street, which was observed in the complete burglary analysis , cannot be found here. Again, this is consistent with Osbornâ€™s research that found protective effects reduced for repeat victimization.
117 Table 4.15 Distribution of Single Burglary Sites and Multiple Burglary Sites among Street Types Observed frequency Expected frequency Standardized Pearson residuals Single Burglary Sites Multiple Burglary Sites 14 0 11 0 Dead end 30 9 31 8 L-type streets -0.43 0.47 41 13 43 11 T-type streets -0.72 0.84 59 15 59 15 Through-traffic streets 0.05 -0.05 *Pearsonâ€™s chi-square = 4.20, df = 3, p=0.24 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 Dead EndL-type StreetsT-type StreetsThrough-traffic StreetsPercentage Single Burglary Multiple Burglary Figure 4.18 Distribution of singl e burglary sites and multiple burglary sites among street patterns Table 4.16 lists the distribution of single-f amily burglary sites between corner lots or middle of block lots by single or mu ltiple burglary. The p value is 0.3426 which indicates that there is no significant correlation. This finding conflicts with the
118 environmental features analysis for complete burglary, which found that corner lots are more likely to be burglarized. In another word, a house located at corner is more likely to be chosen as the target for the first bu rglary. However, when a house is already burglarized, the corner location will not influence the risk fo r subsequent victimization. Table 4.16 Distribution of Single Burglary S ites and Multiple Burg lary Sites between Corners or Middle Block Lots Observed Frequency Expected Frequency Standardized Pearson Residuals Single Burglary Sites Multiple Burglary Sites 59 12 96 88 Corner 3.57 -2.71 85 12 214 198 Middle of Block -2.71 3.57 *Pearsonâ€™s Chi-square = 0.9005, df = 1, p=0.3426 220.127.116.11 Land use and adjacency Land use and adjacency includes four meas urements: (a) adjacent land use types, (b) degree of land use mix, (c) residentia l units mix, and (d) vacant buildings and dilapidated houses. Table 4.17 lists all types of adjacent land use that were detected to be significantly different between single burglary group and repeat burglary group by the extension approach--the percen tage of lots dedicated to each specific land use in the studied sitesâ€™ 1,200-f oot buffer zone. According to the table (4.17), multi-fam ily, vacant commercial, shopping malls, restaurants, cafeterias, automotive repair, se rvice and sales, parking lots, mobile home sales, florist, green house, hotels, motels , private hospitals, mo rtuaries, cemeteries, colleges, and public hospitals are more clustered by repeat burglary rather than single burglary sites. Most of th ese land use types, except shopping malls, were found to
119 contribute to burglary incidents in the complete burglary anal ysis. Closeness to One-story Non-Professional Offices, enclosed theaters, auditoriums, churches, and private schools seem to have protective effects fr om subsequent burglary offenses. Table 4.17 Land Use Type Indicators for Repeat Burglary Code Land Use T value by pooled method Pr>| t| by pooled method T value by Scatterthwaite method Pr>| t| by Scatterthwaite method 1 Single Family 13.33 <.0001 13.88 <.0001 3 Multi Family -9.34 <.0001 -8.55 <.0001 9 Undefined reserved for DOR 2.88 0.0040 3.10 0.0020 10 Vacant Commercial -3.83 0.0001 -3.73 0.0002 15 Regional Shopping Malls (1 site) -2.34 0.0196 -1.97 0.0491 17 One-story Non-Professional Offices 2.77 0.0056 2.83 0.0047 21 Restaurants, Cafeterias -3.57 0.0004 -3.39 0.0007 27 Automotive Repair, Service and Sales -2.68 0.0074 -2.46 0.0141 28 Parking Lots, Mobile Home Sales -2.60 0.0094 -2.42 0.0157 30 Florist, Green House -6.07 <.0001 -5.37 <.0001 32 Enclosed Theaters, Auditoriums (1 site) 2.26 0.0237 2.57 0.0102 39 Hotels, Motels -3.11 0.0019 -2.86 0.0043 71 Churches 5.81 <.0001 6.23 <.0001 72 Private Schools 2.27 0.0234 2.42 0.0156 73 Private Hospitals (1 site) -2.39 0.0168 -2.00 0.0455 76 Mortuaries, Cemeteries -4.78 <.0001 -4.25 <.0001 84 Colleges (1 site) -4.10 <.0001 -3.93 <.0001 85 Public Hospitals (5 sites) -3.69 0.0002 -3.47 0.0005 86 Other Counties 2.18 0.0293 2.38 0.0172 89 Other Municipal 2.12 0.0339 2.14 0.0323 According to Table 4.18, repeat burglary site s are more likely to occur in areas with a high degree of general land use mix, both in census block level and major street block level. At the same time, thereâ€™s no correla tion between with relative land use mix and repeat burglary, at both levels.
120 Table 4.18 Land Use Mix Indicat ors for Repeat Burglary Mean value of single burglary group Mean value of multiple burglary group Mean value of difference T value by pooled method Pr>| t| by pooled method T value by Scatterth waite method Pr>| t| by Scatterth waite method General land use mix in census block level 0.1813 0.2528 -0.072 -6.05 <0.0001 -5.49 <0.0001 Relative land use mix in census block level 0.2078 0.1969 0.0109 0.16 0.8733 0.22 0.8277 General land use mix in major street block level Length 0.208 0.2311 -0.023 -3.51 0.0005 -3.43 0.0006 Relative land use mix in major street block level 0.642 0.8448 -0.203 -1.56 0.12 -1.51 0.1328 Table 4.19 shows that there is no relationshi p between coefficient of variation of housing value and repeat burglary, at both the census block level and the major street block level. This finding means house value around the victimized house will not affect repeat burglary for the same house. Table 4.19 Residential Units Mix Indicators for Repeat Burglary Mean value of single burglary group Mean value of multiple burglary group Mean value of difference T value by pooled method Pr>| t| by pooled method T value by Scatterth waite method Pr>| t| by Scatterth waite method Coefficient of variation of housing value in census block level 0.2628 0.2664 -0.004 -0.40 0.6877 -0.39 0.6979 Coefficient of variation of housing value in major street block level 0.4055 0.4177 -0.012 -1.27 0.2036 -1.24 0.2168 Table 4.20 lists the distribution of single-f amily residential bu rglary incidents by the relationship with substandard houses and whether or not there were multiple victimizations. According to th e table, there is no strong co rrelation because p is larger
121 than 0.05 .The protective effect of not bei ng adjacent to ("away from") substandard dwelling units diminishes. Table 4.20 Distribution of Single Burglary Sites and Multiple Burglary Sites among Relationships With Substandard Dwelling Units Observed Frequency Single Burglary Sites Multiple Burglary Sites Is substandard dwelling unit 87 27 Adjacent to substandard dwelling unit 65 24 Away from substandard dwelling unit 1098 338 *Pearsonâ€™s Chi-square = 0.54, df = 2, p=0.76 18.104.22.168 Density Table 4.21 shows repeat burglary sites are more likely to be located in areas with high density. At the census block level, repeat burglary is correlated with both lot density and floor area ratio. At the major street level, only lot density is associated with multiple burglary. Table 4.21 Density Indicators for Repeat Burglary Mean value of single burglary group Mean value of multiple burglary group Mean value of difference T value by pooled method Pr>| t| by pooled method T value by Scatterthw aite method Pr>| t| by Scatterthw aite method Lot density in census block level 3.862 4.1415 -0.28 -2.56 0.0104 -2.59 0.0099 Floor area ratio in census block level 6571.2 7042.4 -471.2 -2.36 0.0184 -2.03 0.0430 Lot density in major street block level 2.9499 3.0997 -0.15 -2.01 0.0445 -1.97 0.0488 Floor area ratio in major street block level 5946.5 6141.2 -194.7 -1.31 0.1902 -1.18 0.2393 4.3. Near Repeat Burglary Near repeat burglary refers to burglary cases whose occurrences are close in both time and space, such that an area is at highe r risk than would be expected by random distribution. This is different from repeat burglaries, which are defined as the multiple
122 victimization of the same household. Recent re search suggests the possibility that near repeat burglaries are a real phenomenon (Y ang, 2004, Townsley, Homel et al. 2003). If they do exist, crime prevention strategies s hould be more broadly calibrated than on the basis of repeat burglary alone. As we have noted above, there is a gr eat deal of evidence suggesting that burglarized households are more likely to be victimized again th en are unburglarized households and that revictimization is most lik ely to occur in a very short time after the initial offence (Forrester, Chatterton et al . 1988; Polvi, Looman et al. 1990; Polvi, Looman et al. 1991; Guidi, Townsley et al. 1997). The literature also suggests that burglaries influence each other, like a contagi on, such that if one house is burglarized, the neighbors are at higher risk of also being burglarized (Townsley, Homel et al. 2003). In this context, houses proximate to victimized targets are more likely to share similar features with prior targets and fall within th e cognition map of burglars. We suspect then that even in high crime areas, burglaries te nd to cluster in time and space based on predictable patterns. Although near repeat burglary can be an important theore tical issue, as in the concept of repeat burglary, ve ry little research has been co nducted on this topic. In a systematic literature review, the present rese arch is only one of three studies that have been done on this topic. There are only two other empirical studies using both â€œdistanceâ€ and â€œtimeâ€ variables to identify â€œnearness, â€ so that two burglar y incidences can be identified as near repeat burglaries (Tow nsley, Homel et al. 2003; Johnson and Bowers 2004). The lack of knowledge of near repeat bu rglary is partly due to crime analystsâ€™
123 unfamiliarity with using spatial-temporal anal ysis methodologies to detect near repeat burglary. 4.3.1 Spatial-temporal Analysis The objective of spatial-temporal analysis is to find out if th e phenomena of near repeat burglary exists. As near repeat burglary refers to bur glary cases whose occurrences are close in both time and space, the detection of near repeat burglary is the detection of spatial-temporal clusters of burglary cases. Three most widely used statistical techniques for testing space-time interaction--the Knox test, the Mantel test, and the K nearest neighbor test--are employed as fo rmal procedures to detect th e existence of near repeat burglary. 22.214.171.124 Knox test The Knox test is a simple and straightfo rward method to test spatial-temporal cluster. This method has been employed in the previously menti oned two studies about near repeat burglary. To use this method, critical space and time distance for defining â€œnearâ€ first have to be quantified. In the present study, we choose 800 feet and 1200 feet as a spatial critical distance, and 1 month (30 days), 2 month (60 days) as the critical time distance. Listed below (Table 4.22) are the su mmery of four Knox test outputs. The detail reports can be found in appendix (C). P valu es with two methods--Chi-squared statistic and Monte Carlo simulations, are provided. For all the Knox tests of one month a nd two months, 800 feet and 1200 feet, P values are small and significant. We can conclude that space-time interaction for residential burglary exists in Gainesvill e, in other word, near repeat burglary phenomenon exists in Gainesvill e. Since the critical spatia l distance are set as 800 feet and 1200 feet and time distance are set as one and two months, we can conclude that the
124 risk of victimization for dwelling units within a range of 800 and 1200 feet of a burglarized house for a period of 1-2 months elevates. This finding is consistent with similar research (Johnson and Bowers 2004; Bowers and Johnson 2005). Table 4.22 Knox Test for near repeat burglary Spatial Critical Distance Temporal Critical Distance Test Statistics Expect Statistics P-value from ChiSquare P-value from Monte Carlo simulations 800 30 1545 297.43 0 0.001 800 60 2840 533.63 0 0.001 1200 30 2742 380.19 0 0.001 1200 60 4979 648.95 0 0.001 126.96.36.199 K-Nearest neighbor test The K Nearest Neighbor Test is another popu lar method to test space-time clusters of point data. The test statisti c is the count of the number of case pairs that are K nearest neighbors in both space and time. When spatia l-temporal interaction exists, the number will be large since nearest neighbors in space will also tend to be nearest neighbors in time. Although Knox and Mantel tests are al so frequently used, they have some disadvantages. The choice of cr itical distances in the Knox test can be subjective, and Mantel's test is insensitive to nonlinea r associations between the space and time distances. The k nearest neighbor approach avoids these problems. The result of KNearest Neighbor test for all residential burglary cases in Gainesville can be found in Appendix (C). As both the combined P-value and Monte Carlo P-value are smaller than 0.05, we can conclude that there is significant spatialtemporal interaction such that pairs of cases that are nearest neighbors in space tend also to be nearest neighbors in time. With the inspection of the Jk, the P-values are very small for k =1 through k =6 (0.001), and does
125 not show significance for k equal to or larger than k =7. This demonstrates that the interaction occurs at a space-time scale affec ting the first through sixth nearest neighbors. 188.8.131.52 Mantelâ€™s test Mantel's test (1967) is a widely used method for assessing the relationships between two distance or dissim ilarity matrices. For our near repeat burglary hypothesis, we expect the small space and time distances to be correlated, but not the large distances, like infectious disease. Base d on this expectation, we choos e reciprocal transformation method to test the space and time cluster of bur glary incidents. The calculation result for all residential burglary cases in Gainesville by Mantelâ€™s te st can be found in Appendix (C). Since p value for the mantelâ€™s test is highly small and significant, we can conclude that there is significant spatial-temporal inter action for residential burglary in Gainesville. For all of the three methods testing spacetime interaction for burglary incidents, P values are small and significant. Therefore, we can confirm that burglaries do cluster in space and time. Furthermore, as the Knox test found that burglary cases clustered within 800 feet, 1200 feet and 1-2 months, we can conc lude that burglary vi ctimization is the most reliable predictor of future burglary victimization for not only the same household (Pease 1998), but also for nearby houses in th e near future. Crime prevention strategies based on this near repeat burglary findings should be more broadly calibrated than on the basis of repeat burglary alone. 4.3.2 Environmental Variable Analysis The environmental analysis of near repeat burglary aims to find out what kind of environmental variables are correlated with near repeat burglary. Based on findings from spatial-temporal analysis of near repeat burglary, all burgla ry incidents within 1200 feet and 1 month of another burglary incidents are defined as â€œnear-repeatâ€ burglary cases.
126 These cases were identified by an applic ation written in VBA. The 3100 burglary incidents are separated into two groups, â€œnear-repeatâ€ group with 1617 cases and non near-repeat group with 1483 cases . The two sample t-test is used to compare continuous variables. Categorical variables are still compared by Chi-square test. 184.108.40.206 Permeability Table 4.23 illustrates the two sample t-test results for the c ontinuous permeability indicators. Mean values for non near-repeat burglary group and near-repeat burglary group, difference for mean value of the two gr oups and t value, p value for both pooled method and Scatterthwaite method are listed. Acco rding to the table, near-repeat burglary sites tend to locate closer to arteries, have shorter block le ngth and larger connectivity figure. However, the differentiation of distan ce to public transportation between the near repeat and non near repeat group is not significant. Table 4.23 Permeability Indicators for Near-repeat Burglary Mean value of non nearrepeat burglary site Mean value of nearrepeat burglary site Differen ce between mean values T value by pooled method Pr>| t| by pooled method T value by Scatterthwaite method Pr>| t| by Scatterthwaite method Distance to the closest arteries 684.1 612.61 71.489 3.98 <.0001 3.94 <.0001 Distance to public transportation 930.16 858.47 71.698 1.46 0.1453 1.47 0.1413 Block length 1865.4 1658.1 207.31 2.58 0.0098 2.58 0.01 Connectivity 1.552 1.5259 -0.071 -9.72 <.0001 -9.66 <.0001 Table below (4.24) lists the counts of non near-repeat burglary and near-repeat burglary by street layout and output of Pearsonâ€™s chi-squared test . The p value is less than 0.0001, which indicates strong inte raction between street layo ut pattern and near-repeat burglary. Consistent with analyses discusse d above relative to complete burglary and
127 repeat burglary, areas with grid networks ar e more likely to have near-repeat burglary. Fragmented parallel and loll ipops on a stick tend to have protective effects against contagious burglary. 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 GridironFragmented Parallel Wraped ParallelLoops and Lollipops Lollipops on a StickPercentage Non Near-repeat Burglary Near-repeat Burglary Figure 4.19 Distribution of non n ear-repeat burglary sites and near-repeat burglary sites among street layout patterns Table 4.24 Distribution of Non Near-repeat Bu rglary Sites and Near-repeat Burglary Sites among Street Layout Patterns Observed frequency Expected frequency Standardized Pearson residuals Non near-repeat burglary Sites Near-repeat burglary Sites 426 694 534 586 Gridiron -7.07 9.63 676 616 616 675 Fragmented parallel 4.75 -4.05 102 125 108 118 Warped parallel -0.84 0.89 131 122 121 132 Loops and lollipops 1.42 -1.30 128 47 83 92 Lollipops on a stick 9.84 -5.52 *Pearsonâ€™s chi-square = 100.7877, df = 4, p<.0001
128 Table 4.25 is the cross tabulation of street type and non near-repeat or near-repeat burglary. P value for Pearsonâ€™s Chi-test is 0.09, which suggests that there is no strong relationship between near-repeat burglary a nd street type around the dwelling units. Table 4.25 Distribution of Non Near-repeat Bu rglary Sites and Near-repeat Burglary Sites among Street Types Observed frequency Non near-repeat burglary Sites Near-repeat burglary Sites Dead end 17 7 L-type streets 32 32 T-type streets 37 35 Through-traffic streets 66 84 *Pearsonâ€™s chi-square = 6.27, df = 3, p=0.09 Table 4.26 lists the distribution of burglary incidents between corner lots or middle of block lots by non near-repeat or near-r epeat burglary. The p value is 0.3914, which indicates that there is no significant correlation. Table 4.26 Distribution of Non Near-repeat Bu rglary Sites and Near-repeat Burglary Sites between Corners or Middle Block Lots Observed Frequency Non near-repeat burglary Sites Near-repeat burglary Sites Corner 64 59 Middle of Block 88 99 *Pearsonâ€™s Chi-square = 0.73, df = 1, p=0.3914 220.127.116.11 Land use and adjacency Land use and adjacency includes four meas urements: (a) adjacent land use types, (b) degree of land use mix, (c) residentia l units mix, and (d) vacant buildings and dilapidated houses. Table 4.27 lists all types of adjacent land use that were detected to be significantly different between non near repeat burglary group and near repeat burglary group by the extension measurement--the percenta ge of lots dedicated to each specific land use in the studied sitesâ€™ 1,200-foot buffer zone.
129 Table 4.27 Land Use Type Indicato rs for Near-repeat Burglary Code Land Use T value by pooled method Pr>| t| by pooled method T value by Scatterthwaite method Pr>| t| by Scatterthwaite method 2 Mobile Homes 2.74 0.0061 2.66 0.0079 3 Multi Family -4.73 <.0001 -4.77 <.0001 7 Boarding Homes (Institutional) -3.97 <.0001 -3.97 <.0001 8 Multi Family less than 10 -11.10 <.0001 -11.25 <.0001 9 Undefined reserved for DOR 2.83 0.0047 2.76 0.0058 10 Vacant Commercial -2.72 0.0065 -2.72 0.0065 11 Stores One-Story -4.45 <.0001 -4.45 <.0001 12 Mixed Use, i.e., Store and Office -3.24 0.0012 -3.22 0.0013 19 Professional Service Buildings -3.76 0.0002 -3.82 0.0001 22 Drive-in Restaurants -8.61 <.0001 -8.77 <.0001 23 Financial Institutions -2.60 0.0094 -2.61 0.0091 25 Repair Service Shops -5.16 <.0001 -5.19 <.0001 26 Service Stations -5.70 <.0001 -5.75 <.0001 28 Parking Lots, Mobile Home Sales -4.13 <.0001 -4.19 <.0001 30 Florist, Green House -2.42 0.0155 -2.45 0.0143 39 Hotels, Motels -33.93 <.0001 -3.90 <.0001 40 Vacant Industrial 2.99 0.0029 2.88 0.0041 45 Canneries, Distilleries, and Wineries (1 site) -5.03 <.0001 -5.15 <.0001 73 Private Hospitals (1 site) 2.09 0.0366 2.00 0.0455 77 Mortuaries, Cemeteries -2.01 0.0443 -2.01 0.0442 78 Colleges (1 site) 2.20 0.0281 2.11 0.0354 82 Forest, Park, and Recreational Areas (6 sites) 2.50 0.0126 2.39 0.0169 85 Public Hospitals (5 sites) -5.11 <.0001 -5.22 <.0001 86 Other Counties 2.57 0.0102 2.52 0.0119 87 Other State -4.01 <.0001 -4.06 <.0001 According to table (4.27) above, near repeat burglary is positively correlated with the following land use types, multifamily, boarding homes (Institutional), multi family less than 10, vacant commercial, stores one-story, mixed use, i.e., store and office, professional service buildings, drive-in restaurants, financial institutions, repair service shops, service stations, parki ng lots, mobile home sales, fl orist, green house, hotels, motels, canneries, distilleries, and wineries (1 site), mortuaries, cemeteries, and public Hospitals (5 sites). Most of these land use t ypes have also been iden tified as contributing to burglary in complete burglary analysis. Mobile homes, undefined reserved for DOR,
130 vacant industrial, private hospitals (1 site), colleges (University of Florida), forest, park, and recreational Areas (6 sites) are found to be clustered more by â€œremoteâ€--non near repeat burglary--incidents than by near repeat burglary. Table 4.28 Land Use Mix Indicators for Near-repeat Burglary Mean value of non nearrepeat burglary site Mean value of nearrepeat burglary site Mean value of difference T value by pooled method Pr>| t| by pooled method T value by Scatterth waite method Pr>| t| by Scatterth waite method General land use mix in census block level 0.2237 0.265 -0.041 -5.36 <.0001 -5.38 <.0001 Relative land use mix in census block level 0.3263 0.2869 0.0394 0.83 0.4063 0.82 0.4102 General land use mix in major street block level Length 0.2099 0.2396 -0.03 -6.97 <.0001 -7.02 <.0001 Relative land use mix in major street block level 0.9806 1.706 -0.725 -4.06 <.0001 -4.13 <.0001 According to Table 4.28, near repeat burglar y incidents are more likely to occur in areas with high degree of general land use mix, both in census block level and major street block level. Similar to the analysis re sult of complete burglary, near repeat burglary is significantly correlated with relative land use mix in major street block level. Table 4.29 Residential Units Mix I ndicators for Near-repeat Burglary Mean value of non nearrepeat burglary site Mean value of nearrepeat burglary site Mean value of difference T value by pooled method Pr>| t| by pooled method T value by Scatterth waite method Pr>| t| by Scatterth waite method Coefficient of variation of housing value in census block level 0.28 0.2656 0.0144 2.18 0.029 2.18 0.0291 Coefficient of variation of housing value in major street block level 0.4125 0.4419 -0.029 -4.94 <.0001 -4.95 <.0001
131 According to table 4.29, the coefficient of variation of the housing value is significant correlated with near repeat burglar y at the both the census block level and the major-street block level, however in opposite di rections. This implies that when houses in the same census block are more similar in terms of price, which may imply more possibility in sharing similar features, near repeat burglary are more likely to occur. However, in major street block level, it is the dissimilarity rath er than similarity contributes to burglary. Table 4.30 Distribution of Non Near-repeat Bu rglary Sites and Near-repeat Burglary Sites among Relationships With Substandard Dwelling Units Observed Frequency Non near-repeat burglary Sites Near-repeat burglary Sites Is substandard dwelling unit 128 157 Adjacent to substandard dwelling unit 59 87 Away from substandard dwelling unit 1296 1373 *Pearsonâ€™s Chi-square = 4.76, df = 2, p = 0.0926 Table 4.30 shows there is no correlation be tween near repeat burglary and the relationships with substandard dwelling units. This output contrasts with earlier findings in complete burglary analysis but consistent with findings for repeat burglary analysis. 18.104.22.168 Density Table 4.31 Density Indicators for Near-repeat Burglary Mean value of non nearrepeat burglary site Mean value of nearrepeat burglary site Mean value of difference T value by pooled method Pr>| t| by pooled method T value by Scatterth waite method Pr>| t| by Scatterth waite method Lot density in census block level 3.219 3.5195 -0.301 -2.98 0.0029 -3.02 0.0026 Floor area ratio in census block level 8355.4 10571 -2215 -3.61 0.0003 -3.66 0.0003 Lot density in major street block level 2.6495 2.8655 -0.216 -4.23 <.0001 -4.26 <.0001 Floor area ratio in major street block level 5946.2 6455 -508.8 -4.38 <.0001 -4.42 <.0001
132 Table 4.31 shows repeat burglar y sites are more likely to locate in areas with high density. Both lot density and floor area ratio are positively correlated with near repeat cases, at both census block level a nd major street block level. 4.4 Summary Findings for three crime patterns: complete residential burglary, repeat burglary, and near repeat burglary are presented in this chapter. For each crime pattern, two categories of analysis--spatial-temporal anal ysis and environment variable analysis--are applied. The spatio-temporal pattern analysis for complete residential burglary suggests that: There are two major burglary â€œhot spotsâ€ in Gainesville: (a) one beside the east side of the University of Florida, along Univer sity Avenue and 13th Street and (b) the other between Main Street and Waldo Road. The burglary rate drops duri ng October (the football seas on) and drops even further for the district around the stadiu m during the same time period Burglary incidents occurred in apartm ents peak at May and December, which coincides with the end of Spring and Autumn semesters. The spatio-temporal pattern analysis for re peat residential burg lary suggests that: Although repeat victimization tends to occu r soon after the first incident, repeat residential burglary for single-family hous ing tends to have a short break rather than occurring immediately af ter the prior burglary, which is inconsistent with almost all of the pioneer research. The time course of repeat single-family residential burglary vi ctimization conforms to an exponential model, consistent with other studies Two â€œhumpsâ€ are present for single-fam ily repeat reside ntial burglary in Gainesville, whereas most previous res earch found only one â€œhumpâ€ at around 4 to 5 months. Hot spots have a higher proportion of rep eat burglary incidents than do non-hotspot areas, and the relationship is statistically significant.
133 Eliminating repeat victimizati on may diminish a crime hotspot. Finally, the spatio-temporal pa ttern analysis for repeat residential burglary suggests that: Near repeat burglary is a real phenomenon in Gainesville. Crime prevention strategies based on near repeat burglary should be more broadly calibrated than on the basis of repeat burglary alone. The environment variable analysis for three crime patterns: complete residential burglary, repeat burglary, and near repeat burglary explore multiple physical environment features which belongs to three category. Table 4.32 list s the findings. â€œNâ€ represents for negative influence, â€œPâ€ represents for protective in fluence and â€œUâ€ represents for uncorrelated. Table 4.32 Environmental Featur es and Complete Burglary, Repeat Burglary and Near Repeat Burglary Complete Burglary Repeat Burglary Near Repeat Burglary Distance to the closet major arteries N N N Distance to public transportation N N U Block length N U N Connectivity N N N Gridiron N N N Fragmented parallel P U P Warped parallel U U U Loops and lollipops P U U Lollipops on a stick P U P Dead-end streets P U U L-type streets U U U T-type streets P U U Through-traffic streets N U U Corner N U U Middle of Block P U U General land use mix in census block level N N N Relative land use mix in census block level U U U General land use mix in major street block level N N N Relative land use mix in major street block level N U N
134 Table 4.32 Continued Complete Burglary Repeat Burglary Near Repeat Burglary Percentage of residential land devoted to multifamily housing in census block level N Coefficient of variation of housing value in census block level U U P Percentage of residential land devoted to multifamily housing in major street block level U Coefficient of variation of housing value in major street block level N U N Coefficient of variation of housing value in census block level (excluded cases with value 0) N Is substandard dwelling unit N U U Adjacent to substandard dwelling unit N U U Away from substandard dwelling unit P U U Lot density in census block level P N N Floor area ratio in cen sus block level N N N Lot density in major street block level P N N Floor area ratio in major street block level P U N
136 CHAPTER 5 DISCUSSION AND CONCLUSION This chapter addresses the results presente d in chapter 4. Implic ations, constraints on generalization and future re search are also discussed. 5.1 Discussion This research aims (a) to explore the spat ial and temporal patte rns of burglary, (b) to examine the correlation between burglary and environmental variables, and (c) to identify specific features of the physical e nvironment that contribute to burglary in general and to repeat burglary and â€œnear-rep eat burglaryâ€ in par ticular. We hypothesized that some environmental variables such as accessibility, house location on the block, and adjacent land uses have strong contributi ons to burglary, repeat burglary, and â€œnear repeatâ€ burglary propensity, despite sociodem ographic neighborhood differences. To test that hypothesis, we explored these burglary types and applied two categories of spatialtemporal analysis and environment variable an alysis. Our research generally confirms the hypothesis, subject to the follo wing caveats, discussed below. For complete burglary, this research finds the burglary patterns in Gainesville are influenced by the campus calendar. For exampl e, the temporal pattern exploration of all residential burglary suggests that burglaries in Gainesville peak in the fall and winter, which is consistent with similar studies. Ho wever, there are two exceptions: one peak occurs in May and one dip in October. May is the month that the spring semester ends. The burglary rate drop in October is a comm on phenomenon for college town. For repeat single-family burglary, this re search finds that although repe at victimization tends to
137 occur soon after the first incident, repeat residential burglary for single-family housing tends to have a short break rather than o ccurring immediately after the prior burglary. Moreover, two â€œhumpsâ€ are present for singl e-family repeat residential burglary in Gainesville, whereas most previous research found only one â€œhumpâ€ that is around 4 to 5 months. Meanwhile, data in Gainesville also demonstrates several conclusions consistent with other studies, such as hot spot areas having a higher proportion of repeat burglary incidents than non-hot-spot areas. This relations hip is statistically si gnificant and the time course of repeat single-family reside ntial burglary victimization conforms to an exponential model, consistent with other studie s. For near repeat burglary, this research confirms by three test methods that â€œnear-rep eat burglaryâ€ is a real phenomena. The three test methods are the Knox test, the Mantel te st, and the K Near Neighbor test. For all the three methods, the spatial-temporal inter action is highly signif icant (p < .0001). In the environmental analysis, complete burglary incidents are sensitive to more environment features than are repeat burgl ary and near-repeat burglary. Once a singlefamily house is burglarized, the preventive effects of some environmental features diminish for subsequent victimization, which is consistent with Osbornâ€™s research about socioeconomic characteristics of repeatedly victimized hous eholds (Osborn, Ellingworth, et al., 1996). In fact, in the present research, ten types of environment characteristics are found to be protective in complete burglary analysis, wherea s no environment variable can protect a burglarized house from repeat burglary. Near-repeat burglaries are more sensitive to macrolevel environment features than microlevel featur es. Actually, for all environment variables that evaluate the im mediate surroundings of event sites, only â€œcloseness to major streetsâ€ contributes to near-repeat burglary. The other microlevel
138 features--â€œdistance to public transportation,â€ â€œstreet t ypes around parcels,â€ â€œcorner location,â€ and â€œrelationship with substandard dwelling unitâ€--have no association with repeat burglary. 5.1.1 Permeability This research finds out that permeability is significantly correlated with residential burglary. All permeability measures that indicate high permeability and openness-- â€œcloseness to major arteries,â€ â€œshort dist ance to public transportation,â€ â€œshort block length,â€ â€œhigh connectivity index value,â€ â€œgri diron street pattern,â€ â€œlocation on throughtraffic streets,â€ and â€œcorner locationâ€--contribut e to complete bu rglary. Among these permeability factors, â€œdistance to major street ,â€ â€œconnectivity index value,â€ and â€œgridiron street patternâ€ are associated with all thr ee types of burglary. Furt hermore, three street patterns (fragmented para llel, loops and lollipop s, and lollipops on a st ick) and two street types (dead-end streets and t-type streets) are found to have protective effects in complete burglary analysis. Two street patterns--fragmented parallel and lollipops on a stickâ€“-may protect dwelling units from near-repeat burglary. These findings do not support the tenet of accessibility and openness suggested by new urbanism. New urbanist th eory claims that â€œthe design of streets and buildings should reinforce safe environments, but not at the expense of accessibility and opennessâ€ (Gindroz, 2000, p. 133). However, safety is pe rhaps the most fundamental issue for urban design and planning, because no one wants to live in a community unless it is safe (Gindroz, 2000). It is a reasonable suggesti on that communities s hould be developed for residents rather than passers-by. In this contex t and based on our analysis in Gainesville, the pure gridiron street pattern tends not to be appropriate for residential communities. The protective effect of the fragmented parallel pattern suggests the potential of
139 enhancing safety by redesigning community circ ulation patterns from grid to fragmented parallel. Longitudinal research in Hartford (Fowler, McCa lla, et al., 1979; Fowler & Mangione, 1986) and Akron (Donnelly & Majka, 1998) also indicate that redesigned community street layout pattern can reduce crime. The findings for permeability also provide evidences to support some principles for the so-called â€œplace-based crime prevention th eories.â€ â€œAccess contro lâ€ is an important component of defensible space and CPTED. Th e finding that more permeability links to more burglary is consistent with these theories. Moreover, the correlation between closeness to major streets and burglary inci dents support the â€œpathâ€ concept in crime pattern theory. 5.1.2 Land Use and Adjacency Land use mix is another important tenet for sustainable development, new urbanism and smart growthâ€”all important theories in urban planning in recent decades. In our exploration of land use mix and safe ty, the answer is complicated. According to Jacob (1961) and Newman (1972), diverse land uses can attract a continual flow of people and provide more chances for informal surveillance, and ther efore enhance safety. However, our Gainesville data and analyses sh ow that businesses that attract a small to medium traffic flow, such as cafeterias, se rvices, and parking lots may be "unbalanced" as to the stream of "ousiders" vis-a-vis the necessary local surveillance to deter or prevent burglaries. At the same time, land uses that attract large amounts of activities, such as local shopping malls, large scale supermar kets and movie theaters, may provide surveillance and enhance safety. In another word s, these uses tend to be compatible with residential land use, in this research. As th ese â€œprotectiveâ€ land us es occupy only 0.8% of nonresidential land use, it is not surprising that we fi nd that other land use mix
140 measuresâ€”general land use mix and relative land use mix in census blocks and major street levelsâ€”either contribute to or are not associated with burglary. No protective effects are seen to be associated with these measures. The answer to the question of mixed housing types is also complicated. The current research shows that combining different type s of residential dwelli ng units (e.g., single family and multifamily) in the same census block is associated with higher crime risks. However, it will not affect safety in large scale in the same manner as a block defined by major streets. At the same time, this res earch also finds mixing houses with different values in small scale (census block) is acceptable whereas the mixture in large scale (major street block) may contribute to burglary. The opposite outputs for mixing house types and house values are very interesting. It is not cert ain whether the phenomenon is caused by the fact that a large percentage of apartments in Gainesville are rented by college students. Studentsâ€™ low incomes can be viewed as â€œtemporary,â€ therefore, the closeness of apartment community and â€œwea lthyâ€ (single-family) community may not create the feeling of â€˜â€˜relative deprivationâ€ that are associated with crime, as it might when adjacent residents are forever locked in to low income and class strata. However, mixing these two types of housing in the sa me census block may â€œimpede environmental social controlâ€ (Sampson, 1983, p. 279), reduce surveillance, and increase opportunities for burglary. On the other hand, single-fam ily house residents ar e less likely to be students. Larger variation in small scale, such as census block, implies houses close-by may share less similar features. Commonly, th e differences may not be large enough to impede social interaction. As a result, risk for near-repeat burglary may be reduced, and hence the overall risk for complete burglary is also reduced. However, in a large spatial
141 scale, such as blocks defined by major str eets, close proximity of different economic status neighborhoods may improve risks for burglary. This research also support the â€œbroken windowâ€ theory. Dwelling units that are substandard or adjacent to such dwelling units are more likely to be burglarized than are houses that are distanced from substandard dwelling units. 5.1.3 Density Contrary to the common beli ef that increasing density leads to crime, the present research finds out when socioeconomic and demographic variables effects are controlled, density can be protective. However, when a community or a house is intruded upon, the protective effects vanish. In the complete bu rglary analysis, density measures are found to be protective. However, in repeat burglar y and near repeat burglary analysis, density variables are found to be either contribute to or not associ ated with burglary. It is probably because the community is already fallen in within the burglarâ€™s awareness space. This output partly supports the concep tion of intensive development that is advocated by many urban planners, from Jane Jacobs, to the followers of new urbanism, smart growth, and sustainable development. It may be because higher density can provide more surveillance to prevent an initial burgl ary. When burglaries do occur, which implies that a community is already intruded upon, few environment features can be protective. In this situation, crime prev ention methods other than environmental design, such as intensive police patrol, tempor ary alert facilities, and window locks should be applied to reduce the risk for subsequent re peat and near-repeat burglary.
142 5.1.4 Summary In summary, this research finds that some environmental variables such as accessibility, gridiron street pattern, and land use mix have strong contributions to burglary, repeat burglary, and near-repeat burglary propensity, desp ite sociodemographic neighborhood differences. Initial burglary inci dents are more likely to be prevented by environment features than are rep eat burglary and near-repeat burglary. Findings of this research partly suppor t the conception of more intensive development. Land use mix and housing mix can be either protective or adverse to residential community safety. Permeability cont ributes to residentia l burglary incidents. Although the application of gridiron street layo uts is strongly discou raged in residential communities, other street la yout patterns are found to be either protective or not associated with burglary incide nts. Given this range, urban designers and planners have significant flexibility to fi nd an appropriate balance among competing goals for urban development, from environment efficiency and social justice, to public safety. 5.2 Crime Prevention Guidance Given this knowledge about the relationshi p between environmental features and burglary, building regulatory and planning agenci es have more chances to take proactive measures that can enhance community safety. However, as we stated earlier, when a house or a community is burglarized once, fe w environment features can be protective. Other crime prevention strategies should be applied to reduce risks for subsequent incidents. Knowledge about the pattern of re peat and near-repeat burglary could help to tactically deploy limited police a nd crime prevention resources. As noted in chapter 4, repeat single-family burglaries are more likely to occur in the first 2 months after the initial incident. Immediate response to a burglary incident has
143 considerable potential for reducing overall burglary rates by reducing or eliminating repeat victimization. In the Kirkholt Burglary Prevention Project (F orrester, Chatterton, et al., 1988; Forrester, Frenz, et al., 1990), found that the burgla ry rate fell to 40% of its previous level by successfully preventing repeat victimiza tion. If we can eliminate all repeat burglaries occurred in 2 months of the previous incide nt, we can eliminate 72 cases, which is 4.4 % of single family residential burglar ies in Gainesville. In addition to the strategy of â€œimmediate response,â€ b ecause weâ€™ve found there are two â€œhumpsâ€ for repeat burglary, informing policing officers and victims about the humps so that they can be alert should also be helpful in reduc ing repeat burglary, hence reducing overall burglary rates. Crime prevention strategies based on near -repeat burglary could be even more broadly calibrated than on the ba sis of repeat burglary alone. If we can eliminate all nearrepeat burglaries that occurred within an 800 ft buffer zone and one month of the initial incident, we can reduce 24.9 % of residential burglaries in Gainesville. Compared to 4.4 %, as noted above for repeat burglary, th is is a significant improvement. Figure 5.1 demonstrates all burglary incidents and the corresponding 800 ft buffer ring for the month of May, 2003. If the polic e officers could have such a map on June 1, 2003 and patrolled intensively among these areas, they ha d the potential to catc h or prevent 38% of burglary cases in June 2003.
144 Figure 5.1 Near repeat residential burglary in Gainesville, May,2003 â€“ June, 2003 5.3 Constraints on Generalizations There are certain constraints in generalizi ng this study. As this research is applied to the data of Gainesville, a college cit y, caution should be taken when generalizing the findings from this research to other kinds of cities. For exam ple, the temporal pattern of burglary in multiple dwelling units has a clear association with the universityâ€™s calendar. The same pattern may not be observed in othe r cities. Some other atypical features of Gainesville include: (a) most of the apartm ents are rented by students, and (b) the percentage of people in the total population betw een 15-24 years of age that has been proved to be significantly corr elated with crime is 32.59%, wh ereas the percentage in the general U.S. population is 13.9%. However, a large portion is made up of college students.
145 5.4 Future Research This research explores the relationship of environmental features and burglary by controlling socioeconomic and demographic variables. The compounding effect of socioeconomic and demographic variables was reduced by case-control methodology. However, the interaction of environment and socioeconomic variables is not explored in the current research. Do environmental el ements play the same role in burglary prevention or attraction in communities with different socioeconomic profiles, such as middle class communities and working class communities? The interaction among environment variables is another avenue for research. For example, for communities with different street layout patterns, is through-tra ffic streets associated with burglary whereas dead-end streets are protective? In the current analysis, all 100 types of land use coded in the Alachua County Appraiser Officeâ€™s parcel data were screened. The output is diluted. Are there any better methods to explore the relationship betw een adjacent land use types and burglary? Furthermore, we stated there is a trend that la nd uses that attract a small to medium traffic flow, such as cafeteria, services, and parking lo ts may be "unbalanced" as to the stream of "ousiders" vis-a-vis the necessary local su rveillance to deter or prevent burglaries. However, how many â€œoutsidersâ€ are enough? Are there methods to evaluate this issue? In the spatial-temporal analysis, major area s for further analysis are the microlevel patterns. For example, is the risk of repeat burglary and near-repeat burglary similar for different communities? If not, what kind of variables may contribute to the differences? We have explored the time course of repeat burglary based on all repeat single-family burglary in Gainesville. However, one study in Perth, Western Au stralia (Morgan, 2001) found that time courses for repeat burglaries in distinct but adjacent districts are different.
146 Can we observe the same phenomena in Gaines ville? If we can, wh at factors make the difference? Another question to answer is if repeat burglaries and near-repeat burglaries are perpetrated by the same indivi dual? What do the points of en try and method of entry tell us? Knowing that information, can we improve the accuracy of crime prediction with the findings from this research and direct po lice force allocation acco rdingly? We would hope that to be one outcome of this work.
152 APPENDIX A LAND USE CODE IN ALACHUA COUNTY Value Description 1 Vacant Residential 2 Single Family 3 Mobile Homes 4 Condominia 5 Cooperatives 6 Retirement Homes 7 Boarding Homes (Institutional) 8 Multi-family less than 10 units 9 Undefined reserved for DOR 10 Vacant Commercial 11 Stores One-story 12 Mixed Use, i,e, Store and Office 13 Department Stores 14 Supermarket 15 Regional Shopping Malls 16 Community Shopping Center 17 One-story Non-professional Offices 18 Multi-story Non-professional Offices 19 Professional Service Buildings 20 Airports, Marinas, Bus Terminals, and Piers 21 Restaurants, Cafeterias 22 Drive-in Restaurants 23 Financial Institutions 24 Insurance Company Offices 25 Repair Service Shops 26 Service Stations 27 Automotive Repair, Service, and Sales 28 Parking Lots, Mobile Home Sales 29 Wholesale, Manufactur ing, and Produce Outlets 30 Florist, Greenhouses 31 Drive-in Theaters, Open Stadium 32 Enclosed Theaters, Auditoriums 33 Night Clubs, Bars, and Cocktail Lounges 34 Bowling Alleys, Skating Rings, Enclosed Arenas 35 Tourist Attractions
153 Value Description 36 Camps 37 Race Horse, Auto, and Dog Tracks 38 Golf Courses 39 Hotels, Motels 40 Vacant Industrial 41 Light Manufacturing 42 Heavy Manufacturing 43 Lumber Yards, Sawmills, Planning Mills 44 Fruit, Vegetables, and Meat Packing 45 Canneries, Distilleries, and Wineries 46 Other Good Processing 47 Mineral Processing 48 Warehouses, and Distribution Centers 49 Industrial Storage(Fuel , Equip, and Material) 50 Improved Agriculture 51 Cropland Soil Class 1 52 Cropland Soil Class 2 53 Cropland Soil Class 3 54 Timberland 55 Timberland 56 Timberland 57 Timberland 58 Timberland 59 Timberland 60 Grazing Land Soil Class 1 61 Grazing Land Soil Class 2 62 Grazing Land Soil Class 3 63 Grazing Land Soil Class 4 64 Grazing Land Soil Class 5 65 Grazing Land Soil Class 6 66 Orchard, Groves, Citrus 67 Poultry, Bees, Tropical Fish, Rabbits, etc. 68 Dairies, Feed Lots 69 Ornamentals, Misc, Agriculture 70 Vacant Institutional 71 Churches 72 Private Schools 73 Private Hospitals 74 Homes for Aged 75 Orphanages 76 Mortuaries, Cemeteries 77 Clubs, Lodges, and Union Halls
154 Value Description 78 Sanitariums, Convalescent, and Best Homes 79 Cultural Organizations 80 Undefined 81 Military 82 Forest, Park, and Recreational Areas 83 Public Schools 84 Colleges 85 Public Hospitals 86 Other Counties 87 Other State 88 Other Federal 89 Other Municipal 90 Gov. Owned Leased by Non-Gov. Lessee 91 Utilities 92 Mining, Petroleum, and Gas Lands 93 Subsurface Rights 94 Rights-of-Way Streets, Roads and Canals 95 Rivers, Lakes and Submerged Lands 96 Sewage Disposal, Borrow Pits, and Wetlands 97 Outdoor Recreational 98 Centrally Assessed 99 Acreage not Zoned for Agricultural
155 APPENDIX B MONTHLY KERNEL DENSITY MAP AN D LOCATION QUOTIENT MAP FOR RESIDENTIAL BURGLARY IN GAINESVILLE, FL, 2000-2003
160 APPENDIX C SPATIAL-TEMPORAL CLUSTER ANALYSIS FOR NEAR REPEAT BURGLARY k-Nearest Neighbor Method ************************* Upper-tail Upper-tail k J(k) P(k) DJ(k) DP(k) =============================================================== 1 9 0.00100 9 0.00100 2 44 0.00100 35 0.00100 3 76 0.00100 32 0.00100 4 99 0.00100 23 0.00100 5 124 0.00100 25 0.00100 6 152 0.00100 28 0.00100 7 171 0.00100 19 0.06600 8 194 0.00100 23 0.02200 9 226 0.00100 32 0.00100 10 256 0.00100 30 0.01000 Combined P-value: Statistical Distance Test statistic = 56.219664 Number of Monte Carlo simulations = 999 P-value from Monte Carlo simulations = 0.00100 Bonferroni P-value (J) = 0.01000 Simes P-value (J) = 0.00100 Bonferroni P-value (DJ) = 0.01000 Simes P-value (DJ) = 0.00200 Mantel Method ************* Matrix 1 Distance Transformation: (D+50)^-1 Matrix 2 Distance Transformation: (T+1)^-1 Mantel's r = 0.0128442 Monte Carlo simulation method: Test statistic = 0.0128442 Number of Monte Carlo simulations = 999 P-value from Monte Carlo simulations = 0.001000
161 APPENDIX D CODE FOR ANALYSIS Public Function GeneralLandUseMi x(OriRasterFileName As String, DestinRasterFileName As String) 'get the original rasterdataset Dim pOriginRasterDS As IRasterDataset Set pOriginRasterDS = GetRaste rDatasetFromDisk(OriRasterFileName) Dim pOriBlock As IPixelBlock Set pOriBlock = GetP ixelBlock(pOriginRasterDS) ' Create a default raster a nd QI raster properties interface Dim pDesRaster As IRaster Set pDesRaster = pOri ginRasterDS.CreateDefaultRaster Dim pDesRasProps As IRasterProps Set pDesRasProps = pDesRaster ' Get RasterBand from the raster Dim pDesBand As IRasterBand Dim pDesBandCol As IRasterBandCollection Set pDesBandCol = pDesRaster Set pDesBand = pDesBandCol.Item(0) ' QI RawPixel interface Dim pDesRawPixel As IRawPixels Set pDesRawPixel = pDesBand ' Create a DblPnt to hold the PixelBlock size Dim pDesSize As IPnt Set pDesSize = New DblPnt pDesSize.SetCoords pDesRasProps.Width, pDesRasProps.Height ' Create PixelBlock with defined size Dim pDesBlock As IPixelBlock Set pDesBlock = pDesRawP ixel.CreatePixelBlock(pDesSize) ' Loop through the Array and set value to each pixel Dim I, j As Integer Dim X, Y, Count, VariantTotal As Integer Dim OriVal, NewVal As Variant Dim LandUseMixVal As Double
162 For I = 0 To pDesSize.X 1 For j = 0 To pDesSize.Y 1 'loop through the origin raster to get value LandUseMixVal = 0 Count = 0 VariantTotal = 0 OriVal = pDesBlock.GetVal(0, I, j) ' If OriVal <> 0 Then ' MsgBox OriVal ' End If For X = I 1 To I + 1 For Y = j 1 To j + 1 NewVal = pOriBlock.GetVal(0, X, Y) If NewVal <> "NoData" Then Count = Count + 1 If NewVal <> OriVal Then VariantTotal = VariantTotal + 1 End If End If Next Y Next X pDesSafeArray(I, j) = VariantTotal / Count Next j Next I ' Create a DblPnt to hold the top left corner Dim pPnt As IPnt Set pPnt = New DblPnt pPnt.SetCoords 0, 0 ' Write the PixelBlock to raster band pDesRawPixel.Write pPnt, pDesBlock MsgBox "done" End Function Public Sub ManualInputValue(Burgalr yOrMatchFC As IFeatureClass, _ FldName As String, zoomFactor As D ouble, Optional SqlWhere As String) Dim indexFld As Integer indexFld = Burgalry OrMatchFC.FindField(FldName) Dim pQueryFilter As IQueryFilter Set pQueryFilter = New QueryFilter pQueryFilter.WhereClause = SqlWhere
163 Dim pCursor As IFeatureCursor Set pCursor = BurgalryOrMatchFC.Update(pQueryFilter, False) Dim pFeature As IFeature Set pFeature = pCursor.NextFeature Do Until pFeature Is Nothing zoomToFeature zoomFactor, pFeature HighlightFeature pFeature pFeature.Value(indexFld) = In putBox("Please input the value", FldName, pFeature.Value(indexFld)) pCursor.UpdateFeature pFeature Set pFeature = pCursor.NextFeature Loop MsgBox "done" End Sub Public Sub zoomToFeature(zoomFactor As Double, pFeature As IFeature) Dim pMxDoc As IMxDocument Dim pEnv As IEnvelope Set pMxDoc = ThisDocument Set pEnv = pFeature.Shape.Envelope If Not pFeature Is Nothing Then If pFeature.Shape.GeometryType = 1 Then 'point pEnv.XMax = pEnv.XMax + 5 pEnv.XMin = pEnv.XMin 5 pEnv.YMax = pEnv.YMax + 5 pEnv.YMin = pEnv.YMin 5 End If pEnv.Expand zoomFactor, zoomFactor, True pMxDoc.ActivatedView.Extent = pEnv pMxDoc.ActiveView.Refresh End If End Sub Public Sub HighlightFeatur e(pFeature As IFeature) Dim pMxDoc As IMxDocument Dim pActiveView As IActiveView Dim pGContainer As IGraphicsContainer Dim pDocDefaultSymbols As IDocumentDefaultSymbols Set pMxDoc = ThisDocument Set pDocDefaultSymbols = pMxDoc Set pActiveView = pMxDoc.ActivatedView Set pGContainer = pActiveView.GraphicsContainer pGContainer.DeleteAllElements
164 If Not pFeature Is Nothing Then Dim PElement As IElement Dim PGeometry As IGeometry Set PGeometry = pFeature.Shape If PGeometry.GeometryType = 1 Then Set PElement = New MarkerElement PElement.Geometry = PGeometry pGContainer.AddElement PElement, 0 ElseIf PGeometry.GeometryType = 3 Then Set PElement = New LineElement PElement.Geometry = PGeometry pGContainer.AddElement PElement, 0 ElseIf PGeometry.GeometryType = 4 Then Set PElement = New PolygonElement PElement.Geometry = PGeometry Dim PColor As IColor Set PColor = New RgbColor PColor.RGB = RGB(0, 255, 0) Dim pOutline As ILineSymbol Set pOutline = New SimpleLineSymbol 'Set the line symbol properties pOutline.Width = 4 pOutline.Color = PColor Dim pSFSym As ISimpleFillSymbol Set pSFSym = New SimpleFillSymbol pSFSym.Style = esriSFSHollow pSFSym.Outline = pOutline Dim pFillShapeEle As IFillShapeElement Set pFillShapeEle = PElement pFillShapeEle.Symbol = pSFSym pGContainer.AddElement pFillShapeEle, 0 End If pActiveView.PartialRefre sh esriViewGraphics, Nothing, Nothing End If End Sub Public Function GetLandUseAnalysisValu e(pointFName As String, parcelFName As String, _ BufferDistance As Double) 'Get the point feature class and feature Dim pFeaClass As IFeatureClass Dim pFeaCur As IFeatureCursor
165 Dim pFeature As IFeature Set pFeaClass = getFeat ureClassFromDisk(pointFName) Set pFeaCur = pFeaClass.Update(Nothing, False) Set pFeature = pFeaCur.NextFeature Dim pParcelFClass As IFeatureClass Set pParcelFClass = getF eatureClassFromDisk(parcelFName) Dim LandUseCodeColl As New Collection Dim LandUseFieldColl As New Collection Dim Code As String Dim K As Integer Dim PField As IField Dim pFieldEdit As IFieldEdit For K = 1 To 99 If K < 10 Then Code = "0" & CStr(K) Else Code = CStr(K) End If LandUseCodeColl.Add Code LandUseFieldColl.Add Code Set PField = New Field Set pFieldEdit = PField With pFieldEdit .Name = Code .Editable = True .Required = False .IsNullable = True .Type = esriFieldTypeDouble .Precision = 18 .Scale = 12 End With pFeaClass.AddField PField Next K Dim PGeometry As IGeometry Dim AnArea As IArea Dim pSelectionSet As ISelectionSet 'define the coll ection of land use types Dim pOverallCur As ICursor Dim SqlStr As String Dim pSubQFilter As IQueryFilter Dim pFilter As IQueryFilter
166 Dim pSubCur As ICursor Dim ppCursor As ICursor Set ppCursor = pParcelFClass.Search(Nothing, False) Set pSubQFilter = New QueryFilter Set pFilter = New QueryFilter 'define the varible for data statistics Dim pDataStatistics As IDataStatistics Dim pStatResults As IStatisticsResults Dim OverallAreaSum As Double Dim SubAreaSum As Double Dim ratio As Double Dim j As Integer 'loop through the point features Do While Not pFeature Is Nothing 'Get the buffer area Set PGeometry = pFeature.Shape 'MsgBox pFeature.OID Set AnArea = Buffer Object(PGeometry, BufferDistance) Set pSelectionSet = Ge tFeaturesWithin(pParcelFClass, AnArea) pSelectionSet.Search pFilter, False, pOverallCur Set pDataStatistics = New DataStatistics Set pDataSta tistics.Cursor = pOverallCur pDataStatistics.Field = "Acres" Set pStatResults = pDataStatistics.Statistics OverallAr eaSum = pStatResults.Sum '**** get the analysis value 'loop through the land use types For I = 1 To LandUseCodeColl.Count Code = LandUseCodeColl.Item(I) SqlStr = "USE_CODE = '" & CStr(LandUse CodeColl.Item(I)) & "'" pSubQFilter.WhereClause = SqlStr pSelectionSet.Search pSubQFilter, False, pSubCur 'get the statistics result Set pDataStatistics = New DataStatistics Set pDataStatistics.Cursor = pSubCur pDataStatistics.Field = "Acres" Set pStatRes ults = pDataStatistics.Statistics SubA reaSum = pStatResults.Sum 'get the ration ratio = SubAreaSum / OverallAreaSum 'fill in the table pFeature.Valu e(pFeaClass.FindField(Code)) = ratio pFeaCur.UpdateFeature pFeature Next
167 Set pFeature = pFeaCur.NextFeature Loop MsgBox "Done" End Function Public Function GetMatchPoint(BurgalryPT FName As String, parcelFName As String) 'get the social variables list used to match Dim AnalysisVarNameColl As New Collection Dim AnalysisVarRangeColl As New Collection 'AnalysisVarNameColl.Add "ASSD_VAL" 'AnalysisVarRangeColl.Add (30000) AnalysisVarNameColl.Add "Ave_Fam_Sz" AnalysisVarRangeColl.Add (1) AnalysisVarNameColl.Add "AVG_ASSD_V" AnalysisVarRangeColl.Add (40000) AnalysisVarNameColl.Add "Hetero" AnalysisVarRangeColl.Add (0.25) AnalysisVarNameColl.Add "PCTHLD1F" AnalysisVarRangeColl.Add (15) AnalysisVarNameColl.Add "PCTHLD1M" AnalysisVarRangeColl.Add (15) AnalysisVarNameColl.Add "PCTMNRTY" AnalysisVarRangeColl.Add (33) AnalysisVarNameColl.Add "PCTMale" AnalysisVarRangeColl.Add (0.15) AnalysisVarNameColl.Add "PCTRenter" AnalysisVarRangeColl.Add (30) AnalysisVarNameColl.Add "PCT_21" AnalysisVarRangeColl.Add (20) AnalysisVarNameColl.Add "POPDENS" AnalysisVarRangeColl.Add (12) 'Loop through all burglarized point Dim pBurblaryFC As IFeatureClass Dim pBurblaryFCur As IFeatureCursor Set pBurblaryFC = getFeatur eClassFromDisk(BurgalryPTFName) Set pBurblaryFCur = pBur blaryFC.Update(Nothing, False) Dim pBurglaryFeature As IFeature Set pBurglaryFeature = pBurblaryFCur.NextFeature Dim indexMatchCount As Integer, indexJoinIDInBurglaryFC As Integer Dim indexJoinParcel As Integer, i ndexID As Integer, indexCase As Integer
168 indexMatchCount = pBur blaryFC.FindField("MatchCo") indexJoinIDInBurglaryFC = pBurblaryFC.FindField("Join_FID") indexJoinParcel = pBur blaryFC.FindField("Join_Par") indexID = pBurblaryFC.FindField("ID") 'census ID indexCase = pBurblar yFC.FindField("Casenumb_1") Dim pMatchSiteFC As IFeatureClass Dim pMatchSiteQueryFilter As IQueryFilter Set pMatchSiteFC = getF eatureClassFromDisk(parcelFName) Set pMatchSiteQueryFilter = New QueryFilter Dim pMatchSiteCursor As IFeatureCursor Dim pMatchFeature As IFeature Dim MatchNumber As Integer Dim RndIndex As Integer Dim indexIdInMatchFC As Integer, inde xMatch As Integer, indexParcel As Integer, indexJoinCase As Integer indexIdInMatchFC = pM atchSiteFC.FindField("FID_1") indexMatch = pMatchS iteFC.FindField("Sel_Match") indexParcel = pMatchSiteFC.FindField("TAX_PARCEL") indexJoinCase = pMatchSiteFC.FindField("Join_Case") Dim JoinParcel As String Dim JoinFID As Long Dim CaseNum As String Dim SqlStr As String Dim AnalysisVarName Dim analysisRange Dim OriBurSiteVal As Double Dim I As Integer Do Until pBurglaryFeature Is Nothing 'find parcels which match all conditions SqlStr = "BurglaryCo = 0 and Sel_Match <> 'Y' " For I = 1 To AnalysisVarNameColl.Count AnalysisVarName = AnalysisVarNameColl.Item(I) analysisRange = AnalysisVarRangeColl.Item(I) OriBurSiteVal = pBurglaryFeature.Value(pBurglaryFeatur e.Fields.FindField(AnalysisVarName)) SqlStr = SqlStr & " a nd " & AnalysisVarName & ">" & (OriBurSiteVal analysisRange) & _ " and " & AnalysisVa rName & "<" & (OriBurSiteVal + analysisRange) Next I 'add the condition for census block
169 SqlStr = SqlStr & " and ID <> " & pBurglaryFeature.Value(indexID) 'MsgBox SqlStr pMatchSiteQuer yFilter.WhereClause = SqlStr Set pMatchSiteCursor = pMat chSiteFC.Search(pMatchSiteQueryFilter, False) MatchNumber = GetCountFromCursor(pMatchSiteCursor) 'write the count number to burglary data pBurglaryFeature.Value(indexMatchCount) = MatchNumber 'produce a random number in range of candicates count RndIndex = Int(MatchNumber * Rnd) + 1 'produce a number from 1 to count 'Set pMatchFeatur e = pMatchSiteCursor.NextFeature 'pMatchSiteCursor.Flush Set pMatchSiteCursor = pMat chSiteFC.Update(pMatchSiteQueryFilter, False) For I = 1 To RndIndex Set pMatchFeature = pMatchSiteCursor.NextFeature Next I If MatchNumber > 0 And Not (pMatchFeature Is Nothing) Then JoinFID = pMatchFeature.Value(indexIdInMatchFC) JoinParcel = pMatchFeature.Value(indexParcel) CaseNum = pBurglaryFeature.Value(indexCase) pBurglaryFeature.V alue(indexJoinIDInBurglaryFC) = JoinFID pBurglaryFeature .Value(indexJoinParcel) = JoinParcel pBurblaryFCur.UpdateFeature pBurglaryFeature 'pBurglaryFeature.Store pMatchF eature.Value(indexMatch) = "Y" pMatchFeature.Value(indexJoinCase) = CaseNum pMatchSiteCursor.UpdateFeature pMatchFeature 'pMatchFeature.Store End If Set pBurglaryFeature = pBurblaryFCur.NextFeature Loop MsgBox "done" End Function Public Function MultipleLocationQuote(Quot eRequestTable As String, QuoteFldName As String) Dim casePointFileName As String, censusPolyFileName As String, residenceParcelFileName As String casePointFileName = "C:\Xiaowen\New GPD Data\Geocoding_NewGainesvill e\ResidenceBurglaryInGaine _ClearSameDayReplica.shp " censusPolyFileName = "C:\Xiaowen \New GPD Data\LocationQuoients\POE Analysis\cenblkClear_Gainesville.shp" 'residenceParcelFileName = "C:\Xiaowen\downloaded from H\LocationQuients\singlefamilyWithStanAdd_InGain.shp"
170 'get the table for multiple quotes Dim QuoteTable As ITable Set QuoteTable = getITa bleFromDisk(QuoteRequestTable) 'get the field Dim PFieldIndex As Integer PFieldIndex = QuoteT able.FindField(QuoteFldName) 'loop through the table to run analysis Dim QuoteCursor As ICursor Dim pRow As IRow Set QuoteCursor = Quot eTable.Search(Nothing, False) Set pRow = QuoteCursor.NextRow Dim QuoteName As String Dim QuoteCount As Integer Dim ConditionA As String QuoteCount = 1 Do While Not pRow Is Nothing QuoteName = pRow.Value(PFieldIndex) If QuoteName <> " " Then ConditionA = "STATSDATPO = '" & QuoteName & "'" ' or POE2 = '" & QuoteName & "'" RUNLQ ConditionA, casePointF ileName, censusPolyFileName, "HOUSEHOLDS", QuoteName, QuoteCount, False QuoteCount = QuoteCount + 1 End If Set pRow = QuoteCursor.NextRow Loop End Function Public Sub RUNLQ(ConditionA As St ring, casePointFileName As String, censusPolyFileName As String, _ HouseHoldFldInCensus As String, QuoteName As String, QuoteCount As Integer, ResidenceParcelOrCasePoint As Boolean) 'get the input file from hard disk Dim PointFC As IFeatureClass Set PointFC = getFeature ClassFromDisk(casePointFileName) Dim censusFC As IFeatureClass Set censusFC = getFeatureC lassFromDisk(censusPolyFileName) Dim IndHousehold As Integer IndHousehold = censusFC.FindField(HouseHoldFldInCensus) ' Dim parcelFC As IFeatureClass ' Set parcelFC = getFeatureClassFromDisk(residenceParcelFileName) 'insert fields Dim PField As IField Dim pFieldEdit As IFieldEdit
171 If QuoteCount = 1 Then Dim IndLocaTo As Integer IndLocaTo = censusFC.FindField("LocaTo") If IndLocaTo <> -1 Then censusFC.Delete Field censusFC.Fields.Field(IndLocaTo) End If Set PField = New Field Set pFieldEdit = PField With pFieldEdit .AliasName = "LocaTo" .Name = "LocaTo" .Type = esriFieldTypeInteger .DefaultValue = 0 End With censusFC.AddField PField End If Dim IndLQco As Integer IndLQco = censusFC.FindField(QuoteName) If IndLQco <> -1 Then censusFC.DeleteField censusFC.Fields.Field(IndLQco) End If Set PField = New Field Set pFieldEdit = PField With pFieldEdit .AliasName = QuoteName .Name = QuoteName .Type = esriFieldTypeInteger .DefaultValue = 0 End With censusFC.AddField PField Dim IndLQDividor As Integer IndLQDividor = censusFC.FindField(QuoteName & "Val") If IndLQDividor <> -1 Then censusFC.DeleteField censusFC.Fields.Field(IndLQDividor) End If Set PField = New Field Set pFieldEdit = PField With pFieldEdit .AliasName = QuoteName & "Val" .Name = QuoteName & "Val" .Type = esriFieldTypeDouble
172 .Precision = 18 .Scale = 12 .DefaultValue = 0 End With censusFC.AddField PField Dim IndLQVal As Integer IndLQVal = censusFC.FindField(QuoteName & "LQ") If IndLQVal <> -1 Then censusFC.DeleteField censusFC.Fields.Field(IndLQVal) End If Set PField = New Field Set pFieldEdit = PField With pFieldEdit .AliasName = QuoteName & "LQ" .Name = QuoteName & "LQ" .Type = esriFieldTypeDouble .Precision = 18 .Scale = 12 .DefaultValue = 0 End With censusFC.AddField PField IndLQco = censusFC.FindField(QuoteName) IndLQDividor = censusFC.FindField(QuoteName & "Val") IndLocaTo = censusFC.FindField("LocaTo") IndLQVal = censusFC.FindField(QuoteName & "LQ") 'get the LQ value for whole area Dim pScratchWorkspace As IWorkspace Dim pScratchWorkspaceFactory As IScratchWorkspaceFactory Set pScratchWorkspaceFactor y = New ScratchWorkspaceFactory Set pScratchWorkspace = pScrat chWorkspaceFactory.DefaultScratchWorkspace Dim pWholeAreaSelectionset As ISelectionSet Dim pQF As IQueryFilter Set pQF = New QueryFilter pQF.WhereClause = ConditionA Set pWholeAreaSelectionset = Poin tFC.Select(pQF, esriSelectionTypeIDSet, esriSelectionOptionNorma l, pScratchWorkspace) Dim WholeAreaLQCount As Double Dim WholeAreaLQValue As Double WholeAreaLQCount = pWholeAreaSelectionset.Count If WholeAreaLQCount = 0 Then Exit Sub
173 End If If ResidenceParcelOrCasePoint Then WholeAreaLQValue = WholeAr eaLQCount / 50791 'use total of household unit Else WholeAreaLQValue = WholeAreaL QCount / 3100 'use total burglary cases count End If 'loop through censusfc Dim censusFea As IFeature Dim censusCursor As IFeatureCursor Set censusCursor = cens usFC.Update(Nothing, False) Set censusFea = censusCursor.NextFeature Dim PAllselectionset As ISelectionSet Dim PFindalSelectionset As ISelectionSet Dim pResidenceSelectionset As ISelectionSet Dim pselectionNumber As Double Dim pLocaTo As Double Dim HouseholdTotal As Integer Do While Not censusFea Is Nothing 'find all residence poly inside census block 'Set pResidenceSelectionset = FindFeaturesWithin(parcelFC, censusFea.Shape) HouseholdTotal = censusFea.Value(IndHousehold) 'find all points inside census block Set PAllselectionset = FindF eaturesWithin(PointFC, censusFea.Shape) If ResidenceParcelOrCasePoint = True Then 'use household count as local Total pLocaTo = HouseholdTotal 'pLocaTo = pResidenceSelectionset.Count Else pLocaTo = PAllselections et.Count 'use burglary case count as total End If If pLocaTo > 0 Then 'get the case number Set PFindalSelectionset = GetSubSelection(PAlls electionset, ConditionA) pselectionNumber = PFindalSelectionset.Count 'record all information If QuoteCount = 1 Then 't his is the first time fo r this function to run censusFea.Value(IndLocaTo) = pLocaTo censusCursor.UpdateFeature censusFea
174 End If censusFea.Value(IndLQco) = pselectionNumber censusCursor.UpdateFeature censusFea censusFea.Value(IndLQDividor) = pselectionNumber / pLocaTo censusCursor.UpdateFeature censusFea censusFea.Value(IndL QVal) = (pselectionN umber / pLocaTo) / WholeAreaLQValue censusCursor.UpdateFeature censusFea End If Set censusFea = censusCursor.NextFeature Loop End Sub Public Function runTemporalLQ() Dim casePointFileName As String, censusPolyFileName As String, HouseHoldFldInCensus As String casePointFileName = "C:\Xiaowen\New GPD Data\Geocoding_NewGainesvill e\TemporalAnalysis_Complet e\ResidenceBurglaryInGai ne_ClearSameDayReplica_Clear.shp" censusPolyFileName = "C:\Xiaowen\N ew GPD Data\LocationQuoients\Temporal Analysis\CensusTract_Gaine.shp" HouseHoldFldInCensus = "SUM_HOUSEH" Dim QuoteCount As Integer Dim QuoteName As String, ConditionA As String QuoteCount = 1 QuoteName = CStr(" Month" & CStr(QuoteCount)) Do While QuoteCount <= 12 ConditionA = "Month = " & QuoteCount RUNLQ ConditionA, casePointFileName, censusPolyFileName, HouseHoldFldInCensus, QuoteName, QuoteCount, False QuoteCount = QuoteCount + 1 QuoteName = CSt r("Month" & CStr(QuoteCount)) Loop End Function ublic Sub nearBurAna(burPointFN As String, DistrictPolyFN As String, DisIDFldName As String, _ Optional QueryString As String) 'if want the analysis to be done for the whole point dataset, just i nput DistrictPolyFN as "" 'loop through the district poly features to get polygon
175 'get all the points inters ect with the looped polygon 'create analysis table Dim CreateNearCa seFIDDBF As Boolean If Int(m_LongestDis / m_Spatialunit) = 1 Then CreateNearCaseFIDDBF = True End If Dim DisId As String Dim MatrixFN As String Dim fs As Object Set fs = CreateObject(" Scripting.FileSystemObject") Dim fStringPath As String fStringPath = fs.getparentfoldername(burPointFN) If DistrictPolyFN <> "" Then Dim pDisFC As IFeatureClass Set pDisFC = getFeatureClassFromDisk(DistrictPolyFN) Dim PDisFCur As IFeatureCursor Dim pDisQueryFilter As IQueryFilter Set pDisQueryFilter = New QueryFilter pDisQueryFilte r.WhereClause = QueryString Set PDisFCur = pDisFC.Search(pDisQueryFilter, False) Dim PDisFeat As IFeature Set PDisFeat = PDisFCur.NextFeature Dim pPointFC As IFeatureClass Set pPointFC = getFeatureClassFromDisk(burPointFN) Dim pPointSelectionSet As ISelectionSet Dim IndexDisIDFld As Integer IndexDisIDFld = pDisFC.FindField(DisIDFldName) If IndexDisIDFld = -1 Then IndexDisIDFld = pDisFC.FindField(pDisFC.OIDFieldName) End If Do While Not PDisFeat Is Nothing DisId = PDisFeat.Value(IndexDisIDFld) MatrixFN = fStringPath & "\Dis" & DisId & ".dbf" '& "_D" & CStr(m_Spatialunit) & "_T" & CStr(temporalUnit) & ".dbf" Set pPointSelectionSet = FindFeaturesWithin(pPointFC, PDisFeat.Shape) If pPoi ntSelectionSet.Count > 1 Then nearBurglary burPointFN, MatrixFN, CreateNearCaseFIDDBF, pPointSelectionSet ResidualAnalysis MatrixFN End If Set PDisFeat = PDisFCur.NextFeature
176 Loop Else MatrixFN = fStringPath & "\Comp" & ".dbf" '& DisId & "_D" & CStr(spatialUnit) & "_T" & CStr(temporalUnit) & ".dbf" nearBurglary burPointFN, MatrixFN, CreateNearCaseFIDDBF ResidualAnalysis MatrixFN End If End Sub Public Sub nearBurglary(burPointF N As String, MatrixFN As String, CreateNearCaseFIDDBF As Boolean, _ Optional pPointSelect ionSet As ISelectionSet) 'get the featureclass 'get the table 'loop through the features to get the value '*********************get the featureclass********************** Dim pPointFC As IFeatureClass Set pPointFC = getFeatureClassFromDisk(burPointFN) Dim pFirstPointFCursor As IFeatureCursor Dim pSecondPointFCursor As IFeatureCursor Dim pSecondQF As IQueryFilter If pPointSelectionSet Is Nothing Then Set pFirstPointFCurs or = pPointFC.Search(Nothing, False) Else pPointSelectionSet.Sear ch Nothing, False, pFirstPointFCursor End If Dim pFirstPointFeat As IFeature Set pFirstPointFeat = pFirstPointFCursor.NextFeature '********************get the table for pairs count matrix***************************** Dim pMatrixTable As ITable If theFileExists(MatrixFN) Then Set pMatrixTable = getITableFromDisk(MatrixFN) Else Set pMatrixTable = CreateNearBurglaryMatrix(MatrixFN) End If '********************get the table for NearCaseNum***************************** If CreateNearCaseFIDDBF Then Dim fs As Object Set fs = CreateObj ect("Scripting.FileSystemObject") Dim NearCaseFIDDBFFN As String NearCaseFIDDBFFN = fs.getparentfoldername(MatrixFN) & "\" & fs.GetBaseName(MatrixFN) & "_Case.dbf" If theFileExists(NearCaseFIDDBFFN) Then
177 DeleteFiles NearCaseFIDDBFFN End If Dim pNearCaseTable As ITable Set pNearCaseTable = CreateNearCaseFID(NearCaseFIDDBFFN) Set fs = Nothing Dim pRowBuf As IRowBuffer Dim pInsertCursor As ICursor Set pInsertCursor = pNearCaseTable.Insert(True) Set pRowBuf = pNearCaseTable.CreateRowBuffer End If '********************loop through the features****************** Dim FirstOID As Long Dim pSecondPointFeat As IFeature Dim SpatialDis As Long Dim temporalDis As Long Dim FieldIndex As Long Dim pRow As IRow Dim pMatrixTableCursor As ICursor Dim pQF As IQueryFilter Dim TmpValue As Long Dim FeatureCount As Long FeatureCount = pPoi ntFC.FeatureCount(Nothing) Do While Not pFirstPointFeat Is Nothing FirstOID = pFirstPointFeat.OID Set pSecondQF = New QueryFilter pSecondQF.WhereClause = "FID > " & FirstOID If pPointSelectionSet Is Nothing Then Set pSecondPointFCur sor = pPointFC.Search(pSecondQF, False) Else pPointSelectionSet.S earch pSecondQF, False, pSecondPointFCursor End If Set pSecondPointFeat = pSecondPointFCursor.NextFeature Do While Not pSecondPointFeat Is Nothing SpatialDis = FindSp atialDis(pFirstPointFeat, pSecondPointFeat) temporalDis = FindTem poralDis(pFirstPointFeat, pSecondPointFeat) '************** populate the matrix table If SpatialDis >= 0 And temporalDis >= 0 Then FieldIndex = pMatrixTable.FindField(CStr(SpatialDis)) 'fieldIndex = pMatrixTable.FindField Set pQF = New QueryFilter pQF.Wh ereClause = "days = " & temporalDis Set pMatrixTab leCursor = pMatrixTable.Search(pQF, False) Set pRow = pMatrixTableCursor.NextRow
178 Tm pValue = pRow.Value(FieldIndex) pRow.Value(FieldIndex) = TmpValue + 1 pRow.Store 'pMatrixTableCursor.UpdateRow pRow '************** End If If CreateNearCaseFIDDBF Then If SpatialDis = 0 And temporalDis = 0 Then pRowBuf.Value(1) = pFirstPointFeat.OID pRowBuf.Value(2) = pSecondPointFeat.OID pInsertCursor.InsertRow pRowBuf End If End If Set pSecondPointFeat = pSecondPointFCursor.NextFeature Loop Set pFirstPointFeat = pFirstPointFCursor.NextFeature Loop Set pMatrixTable = Nothing MsgBox "Congratualtions, job done" End Sub Public Function CreateNearBurglaryMatr ix(MatrixFN As String) As ITable 'create table in hard disk 'fields list Dim PField As IField Dim pFieldEdit As IFieldEdit Dim pFields As IFields Dim pFieldsEdit As IFieldsEdit Set pFields = New Fields Set pFieldsEdit = pFields pFieldsEdit.FieldCount = Int( m_LongestDis / m_Spatialunit) + 3 'define the OID field Set PField = New Field Set pFieldEdit = PField With pFieldEdit .AliasName = "ObjectID" .Name = "ObjectID" .Type = esriFieldTypeOID End With Set pFieldsEdit.Field(0) = PField 'define the temporal field Set PField = New Field Set pFieldEdit = PField With pFieldEdit .AliasName = "days"
179 .Name = "days" .Type = esriFieldTypeInteger End With Set pFieldsEdit.Field(1) = PField 'define the matrix field Dim spatialDistance As Long Dim j As Long Dim pFieldName As String j = 2 spatialDistance = m_Spatialunit Dim tmpDistance As Double tmpDistance = 0 Do While tmpDistance < (m_LongestDis + m_Spatialunit) pFieldName = Str(tmpDistance) Set PField = New Field Set pFieldEdit = PField With pFieldEdit .AliasName = pFieldName .Name = pFieldName .Type = esriFieldTypeDouble .Precision = 18 .Scale = 12 .DefaultValue = 0 End With Set pFieldsEdit.Field(j) = PField tmpDistance = tmpDistance + m_Spatialunit j = j + 1 Loop 'create the table Dim fs As Object Set fs = CreateObject(" Scripting.FileSystemObject") Dim fStringName As String, fStringPath As String fStringPath = fs.getparentfoldername(MatrixFN) fStringName = fs.GetBaseName(MatrixFN) Dim pWorkspaceFactory As IWorkspaceFactory Dim pWorkspace As IWorkspace Dim pFeatureWorkspace As IFeatureWorkspace Dim pFeatureClass As IFeatureClass Dim pMatrixTable As ITable Set pWorkspaceFactory = New ShapefileWorkspaceFactory Set pWorkspace = pWorkspaceFactory.OpenFromFile(fStringPath, 0) Set pFeatureWorkspace = pWorkspace
180 Set pMatrixTable = pFeatureWorkspace. CreateTable(fStringName, pFields, Nothing, Nothing, "") 'populate the temporal field Dim temporalDis As Long Dim pRowBuf As IRowBuffer Dim pInsertCursor As ICursor Set pInsertCursor = pMatrixTable.Insert(True) Set pRowBuf = pMatrixTable.CreateRowBuffer temporalDis = 0 Do While temporalDis < (m _LongestTime + m_TemporalUnit) pRowBuf.Value(1) = temporalDis pInsertCursor.InsertRow pRowBuf temporalDis = temporalDis + m_TemporalUnit Loop Set CreateNearBurglaryMatrix = pMatrixTable Set fs = Nothing End Function Public Function CreateNearCaseFID(Near CaseFIDDBFFN As String) As ITable Dim PField As IField Dim pFieldEdit As IFieldEdit Dim pFields As IFields Dim pFieldsEdit As IFieldsEdit Set pFields = New Fields Set pFieldsEdit = pFields pFieldsEdit.FieldCount = 3 'define the OID field Set PField = New Field Set pFieldEdit = PField With pFieldEdit .AliasName = "ObjectID" .Name = "ObjectID" .Type = esriFieldTypeOID End With Set pFieldsEdit.Field(0) = PField 'define the firstFID field Set PField = New Field Set pFieldEdit = PField With pFieldEdit .AliasName = "FirFID" .Name = "FirFID" .Type = esriFieldTypeInteger End With
181 Set pFieldsEdit.Field(1) = PField 'define the Second FID field Set PField = New Field Set pFieldEdit = PField With pFieldEdit .AliasName = "SecFID" .Name = "SecFID" .Type = esriFieldTypeInteger End With Set pFieldsEdit.Field(2) = PField 'create the table Dim fs As Object Set fs = CreateObject(" Scripting.FileSystemObject") Dim fStringName As String, fStringPath As String fStringPath = fs.getparen tfoldername(NearCaseFIDDBFFN) fStringName = fs.Get BaseName(NearCaseFIDDBFFN) Dim pWorkspaceFactory As IWorkspaceFactory Dim pWorkspace As IWorkspace Dim pFeatureWorkspace As IFeatureWorkspace Dim pFeatureClass As IFeatureClass Dim pNearCaseTable As ITable Set pWorkspaceFactory = New ShapefileWorkspaceFactory Set pWorkspace = pWorkspaceFactory.OpenFromFile(fStringPath, 0) Set pFeatureWorkspace = pWorkspace Set pNearCaseTable = pFeatureWorks pace.CreateTable(fStringName, pFields, Nothing, Nothing, "") 'return itable Set CreateNearCaseFID = pNearCaseTable Set fs = Nothing End Function Public Function FindSpatialDis (pFirstPointFeat As IFeature, pSecondPointFeat As IFeature) As Long Dim pFirstX As Double Dim pFirstY As Double Dim pSecondX As Double Dim pSecondY As Double Dim XfieldIndex As Long Dim YfieldIndex As Long Dim distance As Long XfieldIndex = pFirstPoin tFeat.Fields.Fi ndField(m_CoordX) YfieldIndex = pFirstPoin tFeat.Fields.Fi ndField(m_CoordX)
182 If pFirstPointFeat.Value(XfieldIndex) = 0 Or pSecondPointFeat .Value(YfieldIndex) = 0 Then FindSpatialDis = -1 Else distance = (Sqr(((p FirstPointFeat.Value(XfieldIndex) pSecondPointFeat.Value(XfieldIndex)) * (p FirstPointFeat.Value(XfieldIndex) pSecondPointFeat.Value(XfieldIndex))) + ((p FirstPointFeat.Value(YfieldIndex) pSecondPointFeat.Value(YfieldIndex)) * (p FirstPointFeat.Value(YfieldIndex) pSecondPointFeat.Value(Y fieldIndex))))) * 3.28084 FindSpatialDis = (Int(distance / m_Spatialunit)) * m_Spatialunit ' If Int(distance / m_Spatialunit) <> (distance / m_Spatialunit) Then ' FindSpatialDis = (Int(di stance / m_Spatialunit) + 1) * m_Spatialunit ' Else ' FindSpatialDis = (Int( distance / m_Spatialunit)) * m_Spatialunit ' End If End If If FindSpatialDis > m_LongestDis Then FindSpatialDis = m_LongestDis End If End Function Public Function FindTemporalDis(pFirstPoin tFeat As IFeature, pSecondPointFeat As IFeature) Dim pFirstTime As Date Dim pSecondTime As Date Dim FieldIndex As Long Dim days As Long FieldIndex = pFirstPointFea t.Fields.FindField(m_TimeFieldName) pFirstTime = pFirstPointFeat.Value(FieldIndex) pSecondTime = pSecondPointFeat.Value(FieldIndex) days = DateDiff("d", pFirstTime, pSecondTime) FindTemporalDis = (Int(days / m_TemporalUnit)) * m_TemporalUnit ' If Int(days / m_TemporalUnit) <> (days / m_TemporalUnit) Then ' FindTemporalDis = (Int(days / m_TemporalUnit) + 1) * m_TemporalUnit ' Else ' FindTemporalDis = (Int(da ys / m_TemporalUnit)) * m_TemporalUnit ' End If If FindTemporalDis > m_LongestTime Then FindTemporalDis = m_LongestTime End If End Function Public Function ResidualAnalysis(MatrixFN As String) 'populate the sum row and field
183 Dim pMatrixTable As ITable Set pMatrixTable = GetSum(MatrixFN) 'creat expected value table for analysis Dim pExpectTableFN As String Dim fs As Object Set fs = CreateObject(" Scripting.FileSystemObject") Dim fStringPath As String, fStringBase As String fStringPath = fs.getparentfoldername(MatrixFN) fStringBase = fs.GetBaseName(MatrixFN) pExpectTableFN = fStringPat h & "\" & fStringB ase & "Expect.dbf" Dim pExpectedTable As ITable If theFileExists(pExpectTableFN) Then DeleteFiles pExpectTableFN End If Set pExpectedTable = CreateN earBurglaryMatrix(pExpectTableFN) 'creat residual ration table for analysis Dim pResidualTableFN As String pResidualTableFN = fStringPat h & "\" & fStringBase & "Residual.dbf" '"C:\Xiaowen\nearBurglary\D istrict5\D5R esidual.dbf" Dim pResidualTable As ITable If theFileExists(pResidualTableFN) Then DeleteFiles pResidualTableFN End If Set pResidualTable = CreateN earBurglaryMatrix(pResidualTableFN) 'fill in value for both tables Dim pQF As IQueryFilter Set pQF = New QueryFilter pQF.WhereClause = "OID > 0" Dim pExpectCursor As ICursor Set pExpectCursor = pExpectedTable.Update(Nothing, False) Dim pExpectRow As IRow Set pExpectRow = pExpectCursor.NextRow Dim pResidualCursor As ICursor Set pResidualCursor = pRes idualTable.Update(Nothing, False) Dim pResidualRow As IRow Set pResidualRow = pResidualCursor.NextRow Dim pMatrixCursor As ICursor Set pMatrixCursor = pMatrixTable.Search(Nothing, False) Dim pMatrixRow As IRow
184 Set pMatrixRow = pMatrixCursor.NextRow Dim OIDValue As Long Dim SumFieldIndex As Long SumFieldIndex = pMatrixTable.Fields.FieldCount 1 Dim RowSum As Double Dim FieldSum As Double Dim FieldIndex As Long Dim matrixVal As Double Dim expectVal As Double Dim residualVal As Double Dim ALLSum As Double Dim PTotalValueRow As IRow Set PTotalValueRow = pMatrixTable.G etRow(pMatrixTable.Row Count(Nothing) 1) ALLSum = PTotalValu eRow.Value(SumFieldIndex) If ALLSum = 0 Then MsgBox "something wrong with the matrix, please check it!" Exit Function End If Do While Not pExpectRow Is Nothing OIDValue = pExpectRow.OID 'get the row sum RowSum = pMatrixRow.Value(SumFieldIndex) For FieldIndex = 2 To pMatrixTable.Fields.FieldCount 2 'get the field sum FieldSum = PTotalValueRow.Value(FieldIndex) 'get the expected value expectVal = RowSum * FieldSum / ALLSum pExpectRow.Value(FieldIndex) = expectVal 'get the residualratio matrixVal = pMatrixRow.Value(FieldIndex) If expectVal <> 0 Then residualVal = matrixVal / expectVal Else residualVal = 0 End If pResidualRow .Value(FieldIndex) = residualVal Next pExpectCursor.UpdateRow pExpectRow pResidualCursor.UpdateRow pResidualRow Set pExpectRow = pExpectCursor.NextRow Set pResidualRow = pResidualCursor.NextRow Set pMatrixRow = pMatrixCursor.NextRow Loop
185 End Function Public Function GetSum(MatrixFN As String) As ITable Dim pMatrixTable As ITable Set pMatrixTable = getITableFromDisk(MatrixFN) 'add sum field to table Dim pTotalValueField As IField Dim pFieldEdit As IFieldEdit Set pTotalValueField = New Field Set pFieldEdit = pTotalValueField With pFieldEdit .AliasName = "Sum" .Name = "Sum" .Type = esriFieldTypeInteger End With pMatrixTable.AddField pTotalValueField 'MsgBox PmatrixTable.Fields.FieldCount 'add count row to table Dim pRow As IRow 'get the total valu e of every field and row Set pRow = pMatrixTable.CreateRow Dim PField As IField Dim FieldIndex As Long Dim OIDValue As Long Dim sumValue As Long sumValue = 0 FieldIndex = 2 OIDValue = 1 Dim totalRowCount As Long totalRowCount = pMatrixTable.RowCount(Nothing) Dim totalFieldCount As Long totalFieldCount = pMatrixTable.Fields.FieldCount Dim pTotalRow As IRow Set pTotalRow = pMatrixTable.GetRow(totalRowCount 1) Dim pCursor As ICursor Dim pQF As IQueryFilter Set pQF = New QueryFilter pQF.WhereClause = "OID > 0" 'and OID < totalRowCount -1" Set pCursor = pMatrixTable.Update(Nothing, False) 'loop through rows to get sum value for every row 'Dim pRow As IRow
186 Set pRow = pCursor.NextRow Do While Not pRow Is Nothing sumValue = 0 For FieldIndex = 2 To totalFieldCount 2 sumValue = sumValue + pRow.Value(FieldIndex) Next pRow.Value(tot alFieldCount 1) = sumValue pCursor.UpdateRow pRow Set pRow = pCursor.NextRow Loop 'loop through fields to get sum value Dim pData As IDataStatistics Dim pStatResult As IStatisticsResults For FieldIndex = 2 To totalFieldCount 1 sumValue = 0 Set pCursor = pMat rixTable.Search(Nothing, False) Set pData = New DataStatistics Set pData.Cursor = pCursor pData.Field = pMatri xTable.Fields.Field(FieldIndex).Name Set pStatResult = pData.Statistics 'MsgBox pStatResult.Count sumValue = pStatResult.Sum pTotalRow.Value(FieldIndex) = sumValue pTotalRow.Store Set pData = Nothing Next Set GetSum = pMatrixTable End Function Public Function CalRepeatTimeWindow(RepeatBurglaryPointFName As String, RepeatIDFldName As String, DateFldN ame As String, TimeUnit As String) 'get the featureclass Dim RepeatBurPointFC As IFeatureClass Set RepeatBurPointFC = getFeatureClassFromDisk(RepeatBurglaryPointFName) 'get the repeatIDField id Dim indexRepeatID As Integer Dim indexDate As Integer Dim indexDateFirst As Integer Dim indexDatePrevious As Integer Dim indexFirstBoo As Integer indexRepeatID = RepeatBurP ointFC.FindField(RepeatIDFldName) indexDate = RepeatBurP ointFC.FindField(DateFldName) indexDateFirst = RepeatBurPointFC .FindField("DateFir" & UCase(TimeUnit))
187 indexDatePrevious = RepeatBurPoint FC.FindField("DatePre" & UCase(TimeUnit)) indexFirstBoo = RepeatBurPointFC.FindField("First") 'delete the old time window fields if they exist If indexDateFirst > 0 Then RepeatBurPointFC.DeleteField (R epeatBurPointFC.Fields.Field(indexDateFirst)) End If If indexDatePrevious > 0 Then RepeatBurPointFC.DeleteField (RepeatBurPointFC.Fields.Field(indexDatePrevious)) End If If indexFirstBoo > 0 Then RepeatBurPointFC.DeleteField (RepeatBurPointFC.Fields.Field(indexFirstBoo)) End If 'add the time window fields Dim pAddField As IField Dim pFieldEdit As IFieldEdit Set pAddField = New Field Set pFieldEdit = pAddField With pFieldEdit .Name = "DateFir" & UCase(TimeUnit) .Editable = True .Required = False .IsNullable = True '.DefaultValue = defaultVal .Type = esriFieldTypeInteger End With RepeatBurPointFC.AddField pAddField Set pAddField = New Field Set pFieldEdit = pAddField With pFieldEdit .Name = "D atePre" & UCase(TimeUnit) .Editable = True .Required = False .IsNullable = True '.DefaultValue = defaultVal .Type = esriFieldTypeInteger End With RepeatBurPointFC.AddField pAddField Set pAddField = New Field Set pFieldEdit = pAddField With pFieldEdit
188 .Name = "First" .Editable = True .Required = False .IsNullable = True '.DefaultValue = defaultVal .Type = esriFieldTypeString End With RepeatBurPointFC.AddField pAddField indexDateFirst = RepeatBurPointFC .FindField("DateFir" & UCase(TimeUnit)) indexDatePrevious = RepeatBurPoint FC.FindField("DatePre" & UCase(TimeUnit)) indexFirstBoo = RepeatBurPointFC.FindField("First") 'sort the table Dim pTable As ITable Set pTable = RepeatBurPointFC Dim pTableSort As ITableSort Set pTableSort = New esriGeoDatabase.TableSort Dim pQueryFilter As IQueryFilter Set pQueryFilter = New QueryFilter With pTableSort .Fields = RepeatIDFldName & "," & DateFldName .Ascending(RepeatIDFldName) = True .Ascending(DateFldName) = True Set .QueryFilter = pQueryFilter Set .Table = pTable End With pTableSort.Sort Nothing 'find the time window Dim StartRepeatID As Integer, CurrentRepeatID As Integer Dim startDate As Date, CurrentDa te As Date, PreviousDate As Date Dim TimeWindowFirst As Intege r, TimeWindowPrevious As Integer Dim pCursor As ICursor Set pCursor = pTableSort.Rows Dim pRow As IRow Set pRow = pCursor.NextRow StartRepeatID = pRow.Value(indexRepeatID) startDate = pR ow.Value(indexDate) PreviousDate = startDate Do While Not pRow Is Nothing CurrentRepeatID = pRow.Value(indexRepeatID) CurrentDate = pRow.Value(indexDate) If CurrentRep eatID <> StartRepeatID Then
189 StartRepeatID = CurrentRepeatID startDate = CurrentDate PreviousDate = CurrentDate pRow.V alue(indexFirstBoo) = "First" End If TimeWindowFirst = DateDi ff(TimeUnit, startDate, CurrentDate) TimeWindowPrevious = DateDi ff(TimeUnit, PreviousDate, CurrentDate) pRow.Value(index DateFirst) = TimeWindowFirst pRow.Value(indexDa tePrevious) = TimeWindowPrevious pRow.Store PreviousDate = CurrentDate Set pRow = pCursor.NextRow Loop MsgBox "done" End Function
190 LIST OF REFERENCES Alexander, E. R. (1993). Density Measures: A Review and Analysis. Journal of Architectural and Planning Research, 10 (3), 181-202. Alterman, R., & Churchman, A. (1998). Housing Density: A Guide to Increasing the Efficiency of Urban Land Use . Haifa: The Center for Urban and Regional Studies, Technion. (Hebrew). Armitage, R. (1999). An Evaluation of Secured By Design Housing Schemes Throughout the West Yorkshire Area : University of Huddersfield, The Applied Criminology Group. Barr, R., & Pease, K. (1990). Crime Placemen t, Displacement, and Deflection. In M. Tonry & N. Morris (Eds.), Crime and Justice: A Review of Research (Vol. 12, pp. 277-318). Chicago, IL: University of Chicago Press. Baumer, E., & Wright, R. (1996). Crime Seasonality and Serious Scholarship: A Comment on Farrell and Pease. British Journal of Criminology, 36 , 579-581. Beavon, D. J. K., Brantingham, P. L., & Bran tingham, P. J. (1994). The Influence of Street Networks on the Patterning of Prope rty Offenses. In R. V. Clarke (Ed.), Crime Prevention Studies, Vol II. (pp. 149-163). New York: Willow Tree Press. Bennett, T. (1995). Identifying, Explaining and Targeting Burglary Hot Spots. European Journal of Criminal Policy and Research, 3 , 113-123. Bennett, T., & Wright, R. (1984). Burglary on Burglars . Gower: Aldershot. Bevis, C., & Nutter, J. (1977). Changing Street Layouts to Re duce Residential Burglary. Paper presented at the American Society of Criminology, Atlanta,GA. Bivand, R. (1998). A Review of Spatial Statistical Techniques for Location Studies . Retrieved March,20, 2005, from http://www.nhh.no/geo/gib/gib1998/gib983/lund.pdf Block, C. R. (1984). Is Crime Seasonal? Chicago: Illinois Criminal Justice Information Authority. Block, R. (1979). Community, Environment, and Violent Crime. Criminology, 17 , 46-57.
191 Block, R. L., & Block, C. R. (1995). Space, Place, and Crime: Hot Spot Areas and Hot Places of Liquor-Related Crime. In J. E. A. D. Weisburd (Ed.), Crime and Place (pp. 145-184). Monsey, NY: Willow Tree Press. Booth, A. (1981). The Built Environment as a Crime Deterrent. Criminology, 18 (4), 557570. Booth, A., Welch, S., & Johnson, D. R. ( 1976). Crowding and Urban Crime Rates. Urban Affairs Quarterly, 11 (3), 291-308. Bowers, K. J., & Johnson, S. D. (2003). Measuring the Geographical Displacement and Diffusion of Benefit Effects of Crime Prevention Activity. Journal of Quantitative Criminology, 19 (3). Bowers, K. J., & Johnson, S. D. (2005). Do mestic Burglary Repeats and Space-Time Clusters: The Dimensions of Risk. European Journal of Criminology, 2 (1), 67-92. Brady, T. V. (1996). Measuring What Matters: Part One: Measures of Crime, Fear and Disorder . Washington, DC: National Institute of Justice. Brand, S., & Price, R. (2000). The Economic and Social Cost of Crime (No. Home Office Research Study 217). London: Home Office. Brantingham, P. J., & Brantingham, P. L. (1 977). Housing Patterns and Burglary in a Medium-Sized American City. In J. E. Scott & S. Dinitz. (Eds.), Criminal Justice Planning (pp. 63-74). New York: Praeger. Brantingham, P. J., & Brantingham, P. L. (1981). Introduction: The Dimensions of Crime. In P. J. Brantingham & P. L. Brantingham (Eds.), Environmental Criminology (pp. 7-26). Beverly Hills,C A: Sage Publications. Brantingham, P. J., & Brantingham, P. L. (1998). Environmental Criminology: From Theory to Urban Planning Practice. Studies on Crime and Crime Prevention, 7 (1), 31-60. Brantingham, P. J., Dyreson, D. A., & Bran tingham, P. L. (1976). Crime Seen Through a Cone of Resolution. American Behavioral Scientist, 20 (2), 261-273. Brantingham, P. L., & Brantingham, P. J. ( 1975). Residential Burglary and Urban Form. Urban Studies, 12 , 273-284. Brantingham, P. L., & Brantingham, P. J. (1981). Mobility, Notoriety, and Crime: A Study in the Crime Patterns of Urban Nodal Points. Journal of Environmental Systems, 11 (1), 89-99.
192 Brantingham, P. L., & Brantingham, P. J. (1993). Environment, Routine, and Situation: Toward a Pattern Theory of Crime. In R. V. Clarke & M. Felson (Eds.), Routine Activity and Rational Choice: Adva nces in Criminological Theory (Vol. 5, pp. 259-294). New Brunswick and London: Transaction Publishers. Bridgeman, C., & Hobbs, L. (1997). Preventing Repeat Victimisation: the Police Officerâ€™s Guide . London: Home Office. Brown, B. B., & Bentley, D. L. (1993). Resi dential Burglars Judge Risk: the Role of Territoriality. Journal of Environmental Psychology, 13 , 51-61. Buck, A. J., Hakim, S., & Rengert, G. F. (1993). Burglar Alarms and the Choice Behavior of Burglars: A Suburban Phenomenon. Journal of Criminal Justice, 21 , 497-507. Budd, T. (1999). Burglary of Domestic Dwellings: Findings from the British Crime Survey . London: Home Office. Byrne, J. M., & Sampson, R. J. (Eds.). (1986). The Social Ecology of Crime . New York: Springer Verlag. Calhoun, J. R. (1962). Population Density and Social Pathology. Scientific American, 206 , 139-148. Chaiken, J., M. Lawless , & Stevenson, K. (1974). The Impact of Police Activity on Crime: Robberies on the Ne w York City Subway System (No. No. R-1424-N.Y.C). New York: The Rand Corporation. Clarke, R. V. (1989). Theoretical Background to Crim e Prevention through Environmental Design (CPTED) and Situational Prevention. Paper presented at the Designing Out Crime: The Conference Papers, Canberra. Clarke, R. V. (1997). Situational Crime Prevention Successful Case Studies (Second Edition ed.). Guilderland, New York: Harrow and Heston. Clarke, R. V., & Felson, M. (1993). Routine Activity and Rational Choice (Vol. 5). New Brunswick and London: Transaction Publishers. Clarke, R. V., & Weisburd, D. (1994). Diffusion of Crime Control Benefits: Observations on the Reverse of Displacement. In R. V. Clarke (Ed.), Crime Prevention Studies (Vol. 2, pp. 165). Monsey, N.Y: Criminal Justice Press. Cohen, L., & Felson, M. (1979). Social Cha nge and Crime Rate Trends: A Rountine Activity Approach. American Socioloigical Review, 44 , 588-605.
193 Cohen, L. E., & Cantor, D. (1980). Determ inants of Larceny An Empirical and Theoretical Study. Journal of Research in Crime and Delinquency, 17 (2), 140159. Cornish, D., & Clarke, R. V. (1987). Unders tanding Crime Displacement: An application of Rational Choice Theory. Criminology, 25 , 933-947. Cozens, P., Hillier, D., & Prescott, G. ( 2001). Crime and the Design of Residential Property--Exploring the Theo retical Background Part 2. Property Management, 19 (4). Crowe, T. D. (1997). Crime Prevention Th rough Environmental Design Strategies and Applications. In L. J. Fennelly (Ed.), Effective Physical Security (2nd ed., pp. 3588). Boston, MA: Butterworth-Heinemann. Davidson, R. N. (1981). Crime and Environment . New York: St. Martin's Press. Dietrick, B. (1977). The Environment and Burglary Vi ctimization in a Metropolitan Suburb. Paper presented at the Annual Me eting of American Society of Criminology, Atlanta, Georgia. Donnelly, P. G., & Majka, T. J. (1998). Resi dentsâ€™ Efforts at Nei ghborhood Stabilization: Facing The Challenges of Inner-city Neighborhoods. Sociological Forum, 13 (2), 189. Eck, J. E. (1993). The Threat of Crime Displacement. Criminal Justice Abstracts, 25 (3), 527-546. Eck, J. E. (1997). Preventing Crime at Places. In L. W. Sherman, D. Gottfredson, D. MacKenzie, J. Eck, P. Reuter & S. Bushway (Eds.), Preventing Crime: What Works, What Doesn't, What's Promising . Washington, D.C: National Institute of Justice. Farrell, G., Edmunds, A., Hobbs, L., & Laycock, G. (2000). RV Snapshot: UK Policing and Repeat Victimisation (No. Crime Reduction Series Paper 5). London: Home Office. Farrell, G., & Pease, K. (1993). Once Bitten, Twice Bitten: Repeat Victimisation and Its Implications for Crime Prevention (No. Crime Prevention Unit Paper 46). London: Home Office. Farrell, G., & Pease, K. (1994). Crime Seasona lity: Domestic Disputes and Residential Burglary in Merseyside 1988-1990. British Journal of Criminology, 34 , 487-498. Farrell, G., Sousa, W. H., & Weisel, D. L. (2002). The Time Window Effect in The Measurement of Repeat Victimization: A Methodology for Its Examination, and
194 An Empirical Study. In N. Tilly (Ed.), Analysis for Crime Prevention (Vol. 13). Monsey, New York: Criminal Justice Press. Felson, M., & Clarke, R. V. (1998). Opportunity Makes the Thief (No. Police Research Series,Paper 98). London: Police Research Group. Forrester, D., Chatterton, M., & Pease, K. (1988). The Kirkholt Burglary Prevention Project, Rochdale (No. Crime Prevention Unit, Pa per 13). London: Home Office. Forrester, D., Frenz, S., O'Connell, M., & Pease, K. (1990). The Kirkholt Burglary Prevention Project: Phase II (No. Crime Prevention Unit, Paper 23). London: Home Office. Forsyth, A. (2003). Measuring Density: Worki ng Definitions for Residential Density and Building Intensity. Design Briet, 9 . Fotheringham, A. S. (1989). Scale-independent Spatial Analysis. In M. G. a. S. Gopal (Ed.), Accuracy of Spatial Databases (pp. 221-228). London and New York: Taylor and Francis. Fowler, F. J., & Mangione, T. W. (1986). A Three-pronged Effort to Reduce Crime and Fear of Crime: The Hartford Experiment. In D. P. Rosenbaum (Ed.), Community Crime Prevention:Does it work? (pp. 87-108). Newbury Park, CA: Sage. Fowler, F. J., McCalla, M. E., & Mangione, T. W. (1979). Reducing Residential Crime and Fear: The Hartford Neighborhood Crime Prevention Program . Washington, DC: U.S Department of Justice. Freedman, D. A. (1999). Ecological Inference and the Ecological Fallacy . Retrieved Aug/20, 2004, from http://www.stanford.e du/class/ed260/freedman549.pdf Freedman, J. L. (1975). Crowding and Behavior . New York: Viking Press. Galle, O. R., Gove, W. R., & McPhers on, J. M. (1972). Population Density and Pathology: What Are th e Relations for Man? Science, 176 (30), 23-30. Gifford, R. (1993). Crime and Context: a Complex, Crucial Conundrum. Journal of Environmental Psychology, 13 , 1-2. Gillis, A. R. (1974). Population Density and Social Pathology: The case of Building Type, Social Allowance and Juvenile Delinquency. Social Forces, 53 , 306-314. Gillis, A. R. (1974). Population Density and Social Pathology: The Case of Building Type, Social Allowance and Juvenile Delinquency. Social Forces, 53 (2), 306314.
195 Gindroz, R. (2000). Twenty One. In M. Leccese & K. McCormick (Eds.), Charter of the New Urbanism (pp. 133-140). New York: McGraw Hill. Grabosky, P. N. (1995). Burglary Prevention . Canberra: Australian Institute of Criminology. Grant, J. (2002). Mixed Use in Theory and Practice. Journal of American Planning Association, 68 (1), 71-84. Greenberg, S. W., & Rohe, W. M. (1984). Neighborhood Desi gn and Crime: A Test of Two Perspectives. Journal of American Planning Association, 50 (1), 48-61. Greenberg, S. W., Rohe, W. M., & Willia ms, J. R. (1982). Safety in Urban Neighborhoods: A Comparison of Physic al Characteristics and Informal Territorial Control in Hi gh and Low Crime Neighborhoods. Population and Environment, 5 (3), 141-165. Guidi, S., Townsley, M., & Homel., R. (1997). Repeat Break and Enter Crimes: An Analysis of Police Calls for Se rvice Data in a Brisbane Region. Paper presented at the The 2nd National Outlook Symposium: Violent Crime, Property Crime and Public Policy, Canberra. Hakim, S., Rengert, G. F., & Shachmurove, Y. (2001). Target Search of Burglars: A Revised Economic Model. Regional Science, 80 , 121-137. Handy, S., Patterson, R. G., & Butler, K. (2003). Planning for Street Connectivity. American Planning Association Plan ning Advisory Service Report 515 . Hawley, A. (1950). Human Ecology: A Theory of Community Structure . New York: Ronald. Henderson, M. (2002). Preventing Repeat Residential Burgalry:A Meta-evaluation of Two Australian Demonstration Projects . Canberra,Australia: Commonwealth Attorney-Generalâ€™s Department. Hesseling, R. B. P. (Ed.). (1995). Displacement: A Review of the Empirical Literature (Vol. 3). Monsey, NY: Criminal Justice Press. Heywood, I., S. Cornelius, and S. Carver. (1998). Introduction to Geographical Information Systems . New York: Addison Wesley Longman. Hiller, B. (1996). Space Is The Machine . Cambridge: Cambridge University Press. Hillier, B. (2004). Can Streets Be Made Safe? URBAN DESIGN International, 9 , 31-45.
196 Hillier, B., & Shu, S. (1999). Do Burglars Understand Defensible Space . Retrieved 02/05, 2003, from http://www.spacesyntax.com/housing/B illCrimePaper/BillCrimePaper.html Hindelang, M. J., Gottfredson, M. R., & Garofalo, J. (1977). Victims of Personal Crime: An Empirical Foundation for a Theo ry of Personal Victimization . Hope, T. (1995). Symposium on Repeat Vict imization: The Flux of Victimization. British Journal of Criminology, 35 (3), 327-342. Hough, M. (1984). Residential Burglary: A Profile for the British Crime Survey. In R. Clarke & T. Hope (Eds.), Coping With Burglary: Research Perspectives on Policy (pp. 49-59). Boston: Kluwer Academic Publishers. Hunter, R. D., & Jeffery, C. R. (1991). Enviro nmental Crime Prevention: An Analysis of Convenience Store Robberies. Security Journal, 2 (Apri), 78-83. Jeffery, C. R. (1971). Crime Prevention Through Environmental Design . Beverly Hills: Sage Publications. Jeffery, C. R. (1977). Crime Prevention Through Environmental Design (Second ed.). Beverly Hills: Sage Publications. Jeffery, C. R. (1990). Criminology: An Interdisciplinary Approach . Englewood Cliffs: Prentice-Hall. Jeffery, C. R., & Zahm, D. L. (1993). Cr ime Precention through Environment Design, Opportunity Theory, and Rational Choice M odels. In R. V. Clarke & M. Felson (Eds.), Routine Activity and Rational Choice (Vol. 5). New Brunswick: Transaction Publishers. Johnson, J. H., Kerper, H. B., Hayes, D. D., & Killinger, G. G. (1973). The Recidivist Victim: A Descriptive Study . Huntsville: Sam Houston St ate University: Institute of Contemporary Corrections and Behavioral Sciences. Johnson, S. D., Bowers, K., & Hirschfield, A. (1997). New Insights into the Spatial and Temporal Distribution of Repeat Victimization. British Journal of Criminology, 37 (2), 224-241. Johnson, S. D., & Bowers, K. J. (2004). Th e Burglary as Clue to the Future: The Beginnings of Prospective Hot-Spotting. European Journal of Criminology, 1 , 237. Katzman, M. (1981). The Supply of Criminal s: A Geo Economic Examination. In S. Hakim & G. F. Rengert (Eds.), Crime Spillover (pp. 119). Beverly Hills and London: Sage.
197 Kelling, G. L., & Coles, C. (1996). Fixing Broken Windows . New York: Free Press. Kershaw, C., Budd, T., Kinshott, G., Joanna Mattinson, Mayhew, P., & Myhill, A. (2000). The 2000 British Crime Survey for England & Wales (No. Statistical Bulletin 18/00). London: Home Office. Knox, G. (1964). The Detection of Space-time Interactions. Applied Statistics, 13 , 25-29. Krupat, E., & Kubzansky, P. E. (1987). Designing to Deter Crime. Psychology Today, 21, 58-61. Leung, H.-L. (1993). How Crowded Is Overcrowded. Journal of the Canadian Institute of Planners. Lombroso, C. (1876). Criminal Man (1972 ed.). Cambridger, MA: Schenkman Publishing. Loukaitou-Sideris, A., Liggett, R., Iseki, H., & Thurlow, W. (2001). Measuring The Effects of Built Environment on Bus Stop Crime. Environment and Planning B: Planning and Design, 28, 255 280. Luedtke, G. D., & Associates. (1970). Crime and the Physical City: Neighborhood Design Techniques for Crime Reduction . Washington, DC: U.S. Department of Justice. MacDonald, J. E., & Gifford, R. (1989). Terr itorial Cues and Defensible Space Theory: the Burglar's Point of View. Journal of Environmental Psychology, 1989 (9), 193205. Mantel, N. (1967). The Detection of Disease Clustering and A Generalized Regression Approach. Cancer Research, 27, 209-220. Mawby, R. I. (1977a). Defensible Space: A Theoretical and Empirical Appraisal. Urban Studies, 14, 169-179. Miethe, T. D. (1991). Citizen Based Crime Cont rol Activity and Victimization Risks: An Examination of Displacement and Free-Rider Effects. Criminology, 29 , 419-440. Miethe, T. D., Stafford, M. C., & Long, J. S. (1987). Social Differe ntiation in Criminal Victimization: A Test of Routine Activities/Lifestyle Theories. American Sociological Review, 52 (2), 184-194. Miller, T. R., Cohen, M. A., & Wiersema, B. (1996). Victim Costs and Consequences: A New Look : National Institute of Justice Repor t, NCJ155282, U.S. Department of Justice.
198 Morgan, F. (2001). Repeat Burglary in a Pe rth Suburb: Indicator of Short-term or Longterm Risk? In G. Farrell & K. Pease (Eds.), Repeat Victimization (pp. 83-118). Monsey: Criminal Justice Press. Nasar, J. L. (1981). Environmental Factors and Commercial Burglary. Environmental Systems, 11 (1), 49-56. Newman, O. (1972). Defensible Space. In. New York: Macmillan. Newman, O. (1980). Community of Interest . Garden City, NY: Anchor Press. Newman, O. (1996). Creating Defensible Space . Washington, DC: U.S. Department of Housing and Urban Development. Openshaw, S. (1977). Optimal Zoning Syst ems for Spatial Interaction Models. Environment and Planning A, 9 , 169-184. Osborn, D. R., Ellingworth, D., Hope, T., & Trickett, A. (1996). Are Repeatedly Victimized Households Different? Journal of Quantitative Criminology, 12 (2), 223-245. O'Sullivan, D., & Unwin, D. (2003). Geographic Information Analysis . New York, NY: John Wiley & Sons, Inc. Paulsen, D. J., & Robinson, M. B. (2004). Spatial Aspect of Crime: Theory and Practice . Boston,. MA: Allyn & Bacon. Pease, K. (1998). Repeat Victimisation: Taking Stock (No. Crime Prevention and Detection Paper 90). London: Home Office. Perkins, D. D., Wandersman, A., Rich, R. C., & Taylor, R. B. (1993). The Physical Environment of Street Crime: Defensible Space, Territorialit y and Incivilities. Journal of Environmental Psychology, 13 , 29-49. Phelan, G. F. (1977). Testing Academic Notions of Arch itectural Design for Burglary Prevention, How Burglars Perceive Cues in Suburban Apartment Complexes. Paper presented at the annual meeting of the American Society of Criminology, Atlanta, GA. Planners/Engineers, C. (2002). Smart Growth INDEX Indicator Dictionary , August/2004, from http://www.epa.gov/smartgrowth/ pdf/4_Indicator_Dictionary_026.pdf Polvi, N., Looman, T., Humphries, C., & Pease, K. (1990). Repeat Break-and-Enter Victimization: Time Course a nd Crime Prevention Opportunity. Journal of Police Science and Administration, 17 (1), 8-11.
199 Polvi, N., Looman, T., Humphries, C., & Peas e, K. (1991). The Time Course of Repeat Burglary Victimization. British Journal of Criminology, 31 (4), 411-414. Rengert, G., & Hakim, S. (1998). Burglary in Affluent Communities: A Planning Perspective. In M. Felson & R. B. Peiser (Eds.), Reducing Crime Through Real Estate Development and Management (pp. 39-52): Urban Land Institute. Reppetto, T. A. (1976). Crime Preventi on and the Displacement Phenomenon. Crime and Delinquence, 22 , 166-177. Rhodes, W. M., & Conly, C. (1981). Crime a nd Mobility: An Empirical Study. In P. J. Brantingham & P. L. Brantingham (Eds.), Environmental Criminology (pp. 167188). Beverly Hills,CA: Sage. Riley, D. (1987). Time and Crime: The Link between Teenager Lifestyle and Delinquency. Journal of Quantitative Criminology, 3 (4), 339. Robinson, M. B. (1998). Burglary Revictimizat ion: The Time Period of Heightened Risk. British Journal of Criminology, 38 (1), 78-87. Sampson, R. J. (1983). Structural De nsity and Criminal Victimization. Criminology, 21 , 276-279. Sampson, R. J. (1986). The Effects of Ur banization and Neighbor hood Characteristics on Criminal Victimization. In R. M.Figli o, S. Hakim & G. F. Rengert (Eds.), Metropolitan Crime Patterns (pp. 3-26). Monsey, New York: Criminal Justice Press. Saville, G., & Cleveland, G. (1998). 2ND GENERATION CPTED: An Antidote to the Social Y2K Virus of Urban Design. Paper presented at the 3rd Annual International CPTED Conf erence, Washington, DC. Scherdin, M. J. (1986). The Halo Effect: Psyc hological Deterrence of Electronic Security Systems. Information Technology and Libraries, September , 232-235. Schichor, D., D. Decker, and R. O'Bria n. (1979). The Relationship of Criminal Victimization, Police per Capita and Popul ation Density in Twenty-Six Cities. Journal of Crime & Justice, 8 , 308-316. Schneider, R., & Yang, X. (2003). Residential Burglary in Gainesville . Gainsville: University of Florida,Geoplan. Schneider, R. H., & Kitchen, T. (2002). Planning for Crime Prevention . London and New York: Taylor & Francis Group.
200 Shaw, C. R., & McKay, H. D. (1931). Social Factors in Juvenile Delinquency . Washington D.C: Govern ment Printing Office. Shaw, M., & Pease, K. (2000). Research on Repeat Victimis ation in Scotland: Final Report . Edinburgh: The Scottish Executive Central Research Unit. Sherman, L. W. (1990). Police Crackdowns: Init ial and Residual Deterrence. In M. Tonry & N. Morris (Eds.), Crime and Justice:A Review of Research (Vol. 12, pp. 1-48). Chicago: University of Chicago Press. Southworth, M., & Owens, P. M. (1993) . The Evolving Metropolis: Studies of Community, Neighborhood, and Street Form at the Urban Edge. Journal of the American Planning Association, 59 (3), 271-287. Spelman, W. (1993). Abandoned Bu ildings: Magnets for Crime? Journal of Criminal Justice, 21 , 481-495. Spelman, W. (1995). Once Bitten, Then What? British Journal of Criminology, 35 (3), 366-383. Steiner, R. L., Bond, A., M iller, D., & Shad, P. (2004). Future Directions for Multimodal Areawide Level of Service Handbook Research and Development . Tallahassee,FL: The Florida Department of Transportation. Taylor, R. B. (1998). Crime and Small-Scale Places: Wh at We Know, What We Can Prevent, and What Else We Need to Know. Paper presented at the Crime and Place: Plenary Papers of the 1997 Confer ence on Criminal Justice Research and Evaluation, Washington, D.C. Town, S. (2002). Crime displacement: The perception, problems, evidence and supporting theory , February 2002 Townsley, M., Homel, R., & Chaseling, J. (200 0). Repeat Burglary Victimisation: Spatial and Temporal Patterns. The Australian and New Zealand Journal of Criminology, 33 (1), 37-63. Townsley, M., Homel, R., & Chaseling, J. (200 3). Infectious Burglaries: A Test of the Near Repeat Burglary. British Journal of Criminology, 43 , 615-633. Trickett, A., Osborn, D. R., Seymour, J., & P ease, K. (1992). What Is Different about High Crime Areas? British Journal of Criminology, 32 (1), 81-89. White, G. F. (1990). Neighborhood Permeability and Burglary Rates. Justice Quarterly, 7 (1), 57-67.
201 Wilcox, P., Quisenberry, N., Cabrera, D. T., & Jones, S. (2004). Busy Places and Broken Windows? Toward Defining the Role of Physical Structure and Process in Community Crime Models. The Sociological Quarterly, 45 (2), 185-207. Wilson, J. Q., & Kelling, G. L. (1982). The Police and Neighbourhood Safety: Broken Windows. The Atlantic Monthly, 3 , 29-38. Yang, X. (2005). Near Repeat Burglary. Paper presented at the The Eighth Annual Crime Mapping Research Conference, Savannah,Georgia. Zwick, P. D., & Papajorgji, J. (2003). A GIS Decision Model fo r Detecting Substandard Housing in Alachua County, Florida . Gainesville: GeoPlan Center, Department of Urban and Regional Planning,University of Florida.
202 BIOGRAPHICAL SKETCH Xiaowen Yang is a PhD student in the Ur ban and Regional Planning Department at the University of Florida. She is majoring in Geographic Information Systems research. Her work seeks to explore the relationship between environment variables and burglary with the help of GIS and corresponding spatia l-temporal statistics technology. She is a native of China and holds a bachelorâ€™s de gree of engineering in urban planning and design and a masterâ€™s degree of engineering in urban planning and design. Both degrees are from Tongji University, Shanghai.