Does Violent Crime Deter Physical Activity?

Crime has been argued to have important externalities.  One concern is the impact of crime on neighbourhood quality and community. Established negative externalities on neighbourhood include flight to the suburbs (Cullen and Levitt 1999), declining property values (Gibbons 2004), a reduction in the creation of new retail and personal service businesses (Greenbaum and Tita 2004) and geographical sorting of local businesses (Rosenthal and Ross 2010). In this paper we examine whether violent crime (which includes murder, manslaughter, knife attacks and aggravated assault) impacts on individuals’ daily lives and their participation in the most common form of physical activity undertaken in their local area, walking. We also investigate the effect on other forms of physical activity.

This is a substantive issue. First, violent crime is not uncommon. In our country of study, England, there are on average 730 police recorded offences of violent crime with injury per 100,000 local inhabitants each year during our sample period and violent crime accounts for approximately 10% of all recorded crime. Moreover, individuals tend to overestimate their chances of being a victim of violent crime. Individuals get information about crime from the media (Duffy et al. 2008) and media reporting has been shown to be highly selective, focusing on the most serious examples of crime and victimisation (Greer 2007). Second, walking is an important form of exercise. The importance of walking and other physical activity as a determinant of good health has been well established World Health Organization 2002). Despite this, walking has been decreasing in many industrialised countries and this fall has been linked to the rise in obesity and obesity related diseases. Walking also has important externalities: individuals who walk in their local neighbourhood contribute to a sense of community (du Toit et al. 2007). But while concerns about personal safety are commonly cited in research as a barrier to local walking, the literature to date has not found clear results for the relationship between crime and walking or physical engagement in the community. A detailed review concluded that this was partly as a result of non-causal design and small data sets (Foster and Giles-Corti 2008).

We use a sample of nearly 1 million people residing in over 320 small areas in England between 2005 and 2011. We show that concerns about personal safety co-move with police recorded violent crime. To identify the causal effect of recorded violent crime on walking and other physical activity we control for individual-level characteristics, non-time varying local authority effects, national time effects and local authority-specific trends. In addition, we exploit a natural experiment that caused a sudden increase in crime – the 2011 England riots – to identify the causal impact of a large exogenous crime shock on physical activity in a triple difference framework. Our results show a substantive deterrent effect of local area violent crime on walking, pointing to important effects of violent crime on non-victims. The adverse effect of an increase in local area violent crime from the 25^th to the 75^th percentile on walking is equivalent in size to a 6 degree C fall in average minimum temperature.