Exploring the effect of motor traffic on street crime

Jose Pina-Sánchez & Toby Davies

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

Why?

Multiple small indirect effects?

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

Analytical strategy

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

2-Way fixed effects models

  • We model the within-person change across time

    – control for average change in perceptions of crime across time

  • We focus on the total effect of traffic on crime

    – and indirect effects through social capital and disorder

Findings

  • 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

  • Mediating effects

    – traffic leads to perceptions of disorder (graffity, litter, and boarded houses)

    – and undermines social capital (whether neighbours are perceived to help each other)

    – these in turn increase perceptions of crime

Robustness Checks

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%

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

Discussion

Theory and policy implications

  • Motor traffic appears to have a causal effect on street crime

    – to some extent mediated through disorder and social capital

  • Reducing car dependency is even more beneficial than we thought

  • 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