shane johnson
DESCRIPTION
Primera presentación de Shane Johnson en el marco de la Primera Cumbre Internacional de Análisis Criminal Científico. 21 de abril de 2014TRANSCRIPT
Crime pattern theory and the influence of street networks on burglary risk
Professor Shane D Johnson (Toby Davies, Peng Chen)
UCL Department of Security and Crime Science [email protected]
Street network layout and crime
• Background
• Existing research
• Methods to examine the issue
• Findings
Burglary concentration on street segments
“A major research project should be undertaken to determine the ways in which urban design contributes to crime, and criminologists should work with urban designers in the planning of cities, in the same way in which medical experts, transportation experts, lighting and heating experts, and pollution experts are involved in city planning.”
C. Ray Jeffery (1971) “Crime Prevention through Environmental Design” p.217.
Jacobs - ‘Eyes on the street’
• She advocated streets of mixed use, lined with stoops and small businesses used by day, evening and night (stores, bars and restaurants), with buildings oriented to the street, not turning their backs or blank sides on it
• The physical layout and ‘vibrancy’ of the street encourages residents to watch each other and “self police”
• Mixed usage encourages vibrancy, as well as provides guardianship from shopkeepers and other business owners
Newman - Defensible Space
• Territoriality - creating spaces that residents feel they can control and identify strangers
• Natural surveillance - the ability of residents to keep a watchful eye on potential criminal activity
– Design of spaces can increase or decrease cohesion and anonymity – Juxtaposition of public spaces and entry points to known safe areas
Defensible Space = territorial definition reinforced with surveillance opportunities – creating capable guardians
Theories of Risk & Street Configuration
• Jacobs (1961) – “eyes on the street” – Natural policing, mix land use – New urbanism
- sustainability agenda - increase permeability, avoid cul-de-sacs
“..we believe that the physical structure of our environment can be managed and that controlling it is the key to solving numerous problems confronting government today – traffic congestion, pollution, financial depletion, social isolation, and yes, even crime”
(Plater-Zyberk, 1993: p. 12)
Lines of sight and Cul-de-sacs
“If cul-de-sacs are simple and linear, have a visual connection to a through street (part of the classic urban formula), and also have the local conditions right, then they can also be very safe places. However, when they are combined together to form hierarchical systems of interconnected cul-de-sacs, they can become extremely vulnerable.....”
(Hillier, 2004: p. 39)
How might streets shape risk?
• Cohen & Felson (1979) – Routine Activity Theory – Ecological explanation for crime patterns – Motivated offender, capable guardian, suitable target
• Cornish & Clarke (1986) – Rational Choice Perspective – Risk, reward, effort
• Brangtingam & Brantingham (1982) – Crime Pattern Theory – Explicit spatial dimension – Offender awareness spaces shaped by legitimate activity
Crime Pattern Theory
• Brantingham and Brantingham (1981) - Crime pattern theory – Simple permeable street topology will enhance offender awareness of
more opportunities – Limit opportunities for the development of offender awareness space
“... The relative magnitude of an opportunity is proportional to its relative degree of accessibility which will partially determine its probability of being exploited”
(Rengert, 1998: p. 21)
Summary of how might streets shape risk
• Eyes on the street (guardian focus) – Low risk expected at connected places
• Defensible space (guardian focus)
– Low risk expected at less connected locations
• Crime Pattern Theory (Offender/ecological focus) – Streets shape awareness and hence more crime anticipated at
highly connected locations
Overview of previous work – Robust?
Aggregation and Risk of Errors of Inference
• Descriptive analysis (3 of the 9 reviewed) – – aggregating counts across street segments of a particular type across
different types of areas with different characteristics….
• Correlation (2 of the 9 reviewed) – – Ignores the role of other factors
• Ordinary regression (4 of the nine reviewed) - – Ignores the spatial structure to the data
• Homes are located on segments, segments in neighbourhoods etc – Ignores unmeasured factors
• SOLUTION – Hierachical Linear Models for count data
Study 1- Merseyside UK
H1 – the risk of burglary will be greater on Major roads and those intended to be most frequently used (estimated using road category)
H2 – on streets that are the most (least) connected (particularly to major roads), the risk of burglary will be higher (lower) than for those that are not
H3 - the risk of burglary will be lower in cul-de-sacs and, in particular, in those that
are non-linear
Johnson, S.D., and Bowers, K.J. (2010). Permeability and Crime Risk: Are Cul-de-sacs Safer? Journal of Quantitative Criminology, 26(1), 89-111.
Study Area and Data
OS street network and Address Point data: - 118,161 homes - 10,760 street segments
(9,936 with houses on them) 13,287 burglaries (4 years data)
Census geography
- OA (Sample = 815) and, - Ward (Sample = 25)
Road classification
• OS classification – Major – Minor – Local – Private
• Manual classification (~11k street segments) – Linear – Non-linear cul-de-sacs (Sinuous)
Cul-de-Sacs
Mostly Linear Mostly Sinuous
First-order Connectivity
2 links 3 links
4 links 6 links
Other variables
Plus – estimates of unmeasured factors at neighbourhood and wider area level
Aggregate Results by Segment Type
Multi-level Poisson models
Random intercepts Segment IVs Output area IVs
- Area level intercepts explain a significant amount of the variance
- After accounting for these….
HLM results (segment level)
Higher risk
Lower risk
Lower risk
Study 2 – Birmingham (UK)
• More formal analysis using betweeness centrality
• Closer alignment between estimates of potential awareness and crime risk
• Large N study – Birmingham city – ~0.5M homes – ~27,000 burglaries (4 years)
Davies, T., & Johnson, S.D. (2014). Understanding the effect of urban form on burglary via a quantitative analysis.
Nodes, edges, betweeness
Calculating Betweeness
# shortest paths that pass through edge e
r radius r (varied - topological and metric distance)
Betweeness versus Road categories
Poisson HLM results
Burglary risk greater on street segments with highest estimated usage
Summary ALL OTHER THINGS BEING EQUAL, and accounting for
factors that operate at the street segment and area level Risk is higher on –
– segments with higher intended usage – Segments with more connections to major roads
Risk is lowest on- – Roads connected to private roads – Sinuous and linear cul-de-sacs (in that order) – Effect of cul-de-sacs is conservative as some will be ”leaky”
Policy implications?
• Connectivity would appear to criminogenic (crime risk varies within areas)
• Prioritise (pro-active) crime prevention on highly connected street segments (situational prevention as well as patrols)
• Build less connected street networks, more cul-de-sacs
• BUT, this is based on correlations
An Agent Based Modeling Study of Street Network Topology and Burglary Patterns
Peng Chen, Shane D Johnson, and Hongyong Yuan
Agent Based Model (ABM) study
• Research to date is largely correlational not experimental – Mechanism for pattern unclear – CPT or variation in defensive
action?
• ABM is a bottom-up or generative approach – Heterogeneous agents & virtual environment – Simple (condition action) rules, bounded rationality – Stochastic - RNGs
• Emergent phenomena – macro level patterns (across many runs)
Current Model
• Simple ecological model of burglary influenced by RAT, RCT and CPT
• Generative sufficiency – Is the model sufficient to generate patterns discussed?
• Experiments – How do patterns vary for different network configurations? – What about police?
“Typical” street network construction
• Nodes and Edges
• Streets have common properties: – Scale free distribution (Power law distribution)
• Vehicular flows (Stefan et, et al., 2006) • Mass transit networks (Lu and Shi, 2007) • Barabasi & Albert (2001) algorithm
– Small world • Street network topology (Porta, et al., 2006) • Watts & Strogatz (1998) algorithm
Simulated Street Network Connectivity
! !Model 1 – Small world Models 2-4 – coupled networks (0.01, 0.1, 1.0)
Description of Model Parameter! Rationale!Street%network!N=10,000!!!!!
Agent&rules&!poffender="0.01!!!!
pjump=1.0!!!!
pselect=(0�1)!!!
pback=(0,$0.1$or$0.5)!!!
Tmax=360!!
!Number! of! nodes! used% to% construct! the! networks.! The$ city$ of$Liverpool) (UK)) has) around) 9,000) (see) Johnson) &! Bowers,( 2010).((For$reasons$of$computational$resources,$10!000!nodes!are!used%in!the$simulation.!!
!The$proportion$of$potential$offenders$in$the$population.$Wikstrom)(2003)&reports&that&the!fraction)of)offenders) in)a)sample)of)14815#years&olds&was&1.3%.!!
The$probability$people$agent$starts$ to$move!at!beginning!of!each!day.!We! assume!all! people!perform! their! routine! activities! every!day.!!
The$ probability$ the$ agents$ travel$ in$ small8world& network& or& in&scale8free$network$(sensitivity$analysis$–!0"to"1,"0.2"increments).!!!
The$ probability$ that! agents! return! to! the! original! locations! each!time!step%(sensitivity(analysis).!!
The!number&of&time&steps&per!day!!
!
Description of Model
Agent rules
All Agents – Simple activity movement from home node
Offenders agents
– Planned, or – Opportunistic offending decision models:
• Assess guardianship • Evaluation of condition action rule (f(current resources
+guardians)) – Offend or not (allocated resource)
Basic patterns in footfall
Model 2 Model 4
Mean crime per segment
Model 2 Model 4
Overall crime for each model
- (standard deviations in parentheses)
- Interaction with activity patterns
Adding cops (model 2)
Summary
• Simulated crime levels associated with connectivity – But non-linear effects (tipping points?) – Pattern theory sufficient to generate expected patterns – Interaction with routine activity patterns
• Assumptions of model, parameters?
• Policy simulations can provide insight about complex phenomena and test the plausibility of prevention scenarios