Congestion by accident? A two-way relationship for highways in England

Abstract

This paper aims to estimate the causal effect of accidents on traffic congestion and vice versa. In order to identify both effects of this two-way relationship, I use dynamic panel data techniques and open access ‘big data’ of highway traffic and accidents in England for the period 2012–2014. The research design is based on the daily-and-hourly specific mean reversion pattern of highway traffic, which can be used to define a recurrent congestion benchmark. Using this benchmark, I am able to identify the causal effect of accidents on non-recurrent traffic congestion. A positive relationship between traffic congestion and road accidents would yield multiplicative benefits for policies that aim at reducing either of these issues. Additionally, I explore the duration of the effect of an accident on congestion, the ‘rubbernecking’ effect, as well as heterogeneous effects in the most congested highway segments. Then, I test the use of methods which employ the bulk of information in big data and other methods using a very reduced sample. In my application, both approaches produce similar results. Finally, I find a non-linear negative effect of traffic congestion on the probability of an accident.