Exploring the effect of motor traffic on street crime

Jose Pina-Sánchez & Toby Davies

Picture: Mike Malone

Thanks for the invite. Toby used to work here, in the Crime Science department.

Introduction

• A huge literature documenting how traffic harms quality of life

  – accidents, pollution, noise, loss of exercise, social capital, …

• An even bigger literature exploring predictors of street crime

  – +5K studies identified in the Handbook of Crime Correlates

• Yet, only a couple of studies have specifically looked into the effect of motor traffic on street crime

Picture: Mike Malone

Why? The reason for that might be that the connection is not that apparent at first sight.

Why?

Picture: Paul Farmer

Multiple small indirect effects?

• Actually, many different crime theories could be used to predict an effect of traffic on crime

Picture: Paul Farmer

Can you think of specific mechanisms through which motor traffic affects street crime. Let’s think for example, violent crime, like common assault - strain theory. Let’s think burglary - eyes on the street. Let’s think antisocial behaviour - broken windows theory. Many different famous crime theories could be used to predict or expect that motor traffic leads to crime, even though none of them deal with traffic as a criminogenic phenomenon. So in this paper we try to estimate the effect of motor traffic on street crime.

Analytical strategy

Picture: Sandy B

Longitudinal data

• Three waves from Understanding Society (2012, 2015 and 2018)

  – necessary to observe changes in neighbourhoods across time

  – remove those who moved to a different address

  – lots of attrition (about 60% missing in 2015 and 90% in 2018)

  – adjusted using multiple imputation

• Subjective measures of motor traffic and perceptions of crime

  – derived from the interviewer and the interviewee

  – which eliminates methods effects

Picture: Sandy B

– whether the interviewer found the participant’s neighbourhood to be affected by heavy traffic

– interviewee’s perceptions of vandalism, theft and violence in their neighbourhood

2-Way fixed effects models

• We model the within-person change across time

  – \((Y_{it} - \bar{Y}_i) = \beta_0 + \beta_1 (X_{1it} - \bar{X}_{1i}) + \sum_{t=2}^{7} \beta_k X_{kit} + \epsilon_{it}\)

  – controlling for average change in perceptions of crime across time and for time of the day when the interview took place

• We focus on the total effect of traffic on crime

  – and indirect effects through social capital and disorder

Picture: Sandy B

We already know that in areas with more traffic there is more crime, as they are likely central areas with more amenities, nightlife, etc. We want to estimate the causal effect of traffic on crime.

Findings

Picture: Jack Fifield

Total effects

• When a neighbourhood goes from “non-heavy traffic” to “heavy traffic”…

  – we estimate an 8.6%, 6.4% and 6.9% increase in perceptions of vandalism, theft, and violence

Picture: Jack Fifield

Mediating effects?

• Traffic is +ly associated with perceptions of disorder (graffity, litter, and boarded houses)

• And -ly associated with social capital (whether neighbours are perceived to help each other)

• These in turn are +ly associated with perceptions of crime

Picture: Jack Fifield

Robustness Checks

Picture: Allen Watkin

Missing data not at random

• High levels of attrition

  – 60% in wave-2 and 90% in wave-3

• Replicated our analysis just for the first two waves

  – found very similar results

Picture: Allen Watkin

Confounding bias

• We cannot anticipate an obvious time-changing confounder

  – we cannot rule it out either

• A “small” confounder could render our findings non-significant

  – our robustness values range from 2.8% to 4%

Picture: Allen Watkin

We cannot think of any confounders

Measurement error

• Interviewee’s records of ‘heavy traffic’ are noisy

  – can see this as classical measurement error

• Attenuation bias (roughly) proportional to the errors

  – e.g. a reliability of 0.8, attenuates our estimates by 1.2

Picture: Allen Watkin

Pun intended.

Discussion

Picture: Colin Babb

Theoretical implications

• Motor traffic appears to cause crime

  – worth investigating the specific mechanisms behind that

  – silly to keep proposing monocausal crime theories

• Cars & Crime as a nascent subdiscipline

  – Why don’t we use driving bans more often?

  – Why is fear of cars not seen as a form of fear of crime?

  – Why are collisions against pedestrians not part of crime prevention strategies?

Picture: Colin Babb

– we cannot rule out confounding bias entirely, but cannot think of a clear one either

– we are possibly underestimating the causal effect because of how motor traffic was measured

Policy implications

• Reducing car dependency is even more beneficial than we thought

  – beyond all its known benefits, it also seems to reduce crime

• Crime prevention policy needs to recognise the criminogenic effect of motor traffic

  – Secured by Design HOMES should be reconsidered

  – e.g. culs-de-sac might be less prone to burglary but they force higher crime elsewhere by encouraging driving

Picture: Colin Babb

Let’s build cities where we don’t need cars

Picture: Colin Babb