ccr whitworth inequality & crime
Post on 26-Jun-2015
193 Views
Preview:
DESCRIPTION
TRANSCRIPT
Inequality, crime and the role of geographical scale
Dr Adam Whitworth, Dept of Geography
Background• Oxford
– Crime domain of Indices of Multiple Deprivation
– NDC crime displacement & costs of crime analyses
• Sheffield Jan 2010-– Inequality & crime
• A multiple fraud: neither a criminologist nor a geographer!
Inequality & Crime: Two strands
• Inequalities of crime– Spatial distributions, concentrations, persistence
• Inequality and crime– Focus on issue of scale, sub-national relevance &
variation in inequality-crime links
Inequality and Crime• Much recent attention on inequality as
a driver of social ills
• 3 main theories linking inequality and crime– Economic theory (Becker, 1968)– Strain theory (Merton, 1938)– Social disorganisation theory (Shaw & McKay, 1942)
Inequality and Crime
• Much empirical work• Tends to find positive links• Range of data & methods• Range of scales, though tend to be large & US
focussed• But no sub-national England analyses
Step 1: Inequality & crime across England’s local authorities
• Outcomes: crime rates (per ‘at risk’ population)– Burglary & criminal damage rate (households)– Vehicle crime rate (vehicle owning h’holds)– Violence and robbery rate (resident population)
• Hierarchical data: multilevel models– 7 Years of data: 2002/03-2008/09 (Level 1)– in 352 local authorities (Level 2)– in 39 Police Force Areas (Level 3)
• All variables log transformed– % change in outcome variable for each one % change in the explanatory
Explanatory variables
Total population
Youth population rate
Population density
Population turnover
Non-white percentage
Unemployment rate
% Not achieving 5 GCSE A*-C
Teenage conception rate (per1000)Mean house price
PFA Officers
PFA detection rate
Inequality (weighted Gini of MSOA income 2004/05)
Population controls
Social Disorganisationthesis (Shaw & McKay)
Strain theory (Merton)
Economic Theory (Becker)
Main interest
Burglary Robbery Veh Cri Violence Crim DInequality 0.20* 0.28* 0.27* 0.10* 0.07+Pop Density 0.11* 0.37* 0.22* 0.10* 0.08*Turnover 0.23* 0.35* 0.05 -0.03 -0.37*Total Pop 0.07+ 0.22* 0.10+ -0.06+ -0.04Youth Pop -0.19+ -0.29 -0.18 0.42* 0.36*% Unem 0.06+ 0.35* 0.10* 0.12* 0Av house price -0.41* -0.49* -0.21* -0.18* -0.18*% 5 GCSEs A*-C 0.03 0.08 -0.06 -0.15* -0.18*% Non-white 0.06
0.33* 0.06 0.01 0.10*
Youth conceptions 0 0.08 0.05 0.17* 0.12*PFA Detection rate 0.02 -0.27* -0.11 -0.10* -0.05PFA Officers 0 0.04 -0.02 0 -0.12*2003/04 -0.02 0 -0.06* 0.19* 0.11*2004/05 -0.12* -0.09* -0.20* 0.32* 0.12*2005/06 -0.16* -0.09* -0.24* 0.31* 0.09*2006/07 -0.19* -0.09* -0.29* 0.31* 0.10*2007/08 -0.25* -0.18* -0.41* 0.25* -0.012008/09 -0.26* -0.27* -0.53* 0.15* -0.15*Constant 7.84* 3.28* 6.74* 4.95* 7.52*Between-PFA variance (L3)
0.04 0.06 0.03 0.01 0.01(% explained) 56.20% 89.30% 76.80% 50.90% 62.80%Within-PFA btw-CDRP (L2)
0.04 0.09 0.06 0.03 0.02(% explained) 51.00% 84.60% 65.50% 78.40% 75.80%Within-CDRP btw years (L1)
0.02 0.07 0.02 0.02 0.01(% explained) 62.80% 17.20% 71.30% 27.20% 53.60%
Same models, alternative inequality measures
Implications
• Consistent relationship between inequality & crime
• Greater support for sociological theories
• Risk of individualised focus in policy: need greater recognition of the structural socio-economic inequalities which provide context & drivers for crime
• Step 2: experimenting with varying the understanding of ‘local’ in the local inequality measure
‘Local’ inequality & crime: the conceptual starting point
• Given findings at LA level, why explore relevance of more local inequality?
(Wilkinson & Pickett, 2006)
Step 2: Exploring ‘local’ inequality and crime
• Alternative plausible explanation for Wilkinson and Pickett findings– Local inequality matters but only in certain contexts: local
findings less consistent– National results may be the weighted aggregation of local
findings: more consistent– But processes are locally (not nationally) driven
• Plausibility in relation to crime specifically– Theoretical: esp Becker’s economic theory– Intuitive: wouldn’t we expect potential offenders to notice local
inequality as well as/rather than inequality at larger scales– Some empirical evidence to support its relevance
London & South Yorkshire IMD2010 quartiles
• Rationale for Met & South Yorks case study areas
• Data: 2 years per force
Varying the scale of the ‘local’ inequality measure
Gini (left) & Gini-Robbery correlation (right) across the 10 layers
0 2 4 6 8 10 12Gini
SouthYorks
Met
10
9
8
7
6
5
4
3
2
1
10
9
8
7
6
5
4
3
2
1
-.3 -.2 -.1 0 .1 .2 .3 .4Ineq-Crime Correlation
10
9
8
7
6
5
4
3
2
1
10
9
8
7
6
5
4
3
2
1
Explanatory variables in the modelling
Youth population rate
Population density
Population turnover
Non-white percentage
Unemployment rate
% Adults only basic education
Teenage conception rate (per1000)Mean house price
Richest area dummy
Year dummyCDRP dummies
Spatially lagged crime rateInequality (weighted Gini of MSOA income 2007/08)
Population controlsSocial Disorganisationthesis (Shaw & McKay)
Strain theory (Merton)
Economic Theory (Becker)
Spatial autocorrelationMain interest
Modelling approach• Outcomes: Crime counts at MSOA level
– Burglary, vehicle crime, robbery, violence
• Moran’s I shows spatial autocorrelation in crimes and so spatially lagged outcome added to RHS
• Overdispersed Poisson distribution: negative binomial models
• Total population added to RHS as exposure variable: interpretation shifts from counts to rates per capita– Centres on one: ie 1.04 for a continuous X relates to a 4%
point increase in outcome for 1% increase in X
Modelled inequality coefficients across all 10 contiguous layers
.95 1 1.05 1.1Inequality coefficient
10
9
8
7
6
5
4
3
2
1
Met
Burglary Vehicle Crime
Robbery Violence
.6 .7 .8 .9 1 1.1Inequality coefficient
10
9
8
7
6
5
4
3
2
1
South Yorks
Burglary Vehicle Crime
Robbery Violence
Implications
• Local inequality can have relevance, but context looks to matter
• Findings line up with prior hypotheses across the two case study areas– London: coherent socio-spatial system– South Yorks: tendency for lower crime levels and fractured
socio-spatial system across the contiguity layers
• Processes linking inequality & crime : is this a substantive finding (ie local inequality really does matter but only in certain contexts) or an artefact of the numbers/method?
Next steps
• More systematic analyses around– Whether consistent story builds up around local inequality &
crime– Whether any such variation in finds can be lik to profiles of local
contexts
• How?– Geographically weighted regression models– And map findings onto geodemographic profiles of local
contexts– Quali work with offenders? How does inequality fit in as a
factor, if at all? And does this vary by local context?
top related