Safety-in-numbers: A systematic review and meta-analysis of evidence

Abstract

This paper presents a systematic review and meta-analysis of studies that have estimated the relationship between the number of accidents involving motor vehicles and cyclists or pedestrians and the volume of motor vehicles, cyclists and pedestrians. A key objective of most of these studies has been to determine if there is a safety-in-numbers effect. There is safety-in-numbers if the number of accidents increases less than proportionally to traffic volume (for motor vehicles, pedestrians and cyclists). All studies reviewed in the paper are multivariate accident prediction models, estimating regression coefficients that show how the number of accidents depends on the conflicting flows (pedestrians, cyclists, motor vehicles), as well as (in some of the models) other factors that influence the number of accidents. Metaanalysis of regression coefficients involves methodological problems, which require I:\SM-AVD\3398 Kjerne 21\Artikkelarkiv 2013-\Elvik_10.1016_j.ssci.2015.07.017.docx 2 careful consideration of whether the coefficients are sufficiently comparable to be formally synthesised by means of standard techniques of meta-analysis. The comparability of regression coefficients was assessed. It was concluded that a formal synthesis of regression coefficients in studies of the safety-in-numbers effect is defensible. According to a random-effects inverse-variance meta-analysis, the summary estimates of the regression coefficients for traffic volume are 0.50 for motor vehicle volume, 0.43 for cycle volume and 0.51 for pedestrian volume. Estimates are highly consistent between studies. It is concluded that a safety-innumbers effect exists. It is still not clear whether this effect is causal, nor, if causal, which mechanisms generate the effect.

Safety in numbers for cyclists beyond national-level and city-level data: a study on the non-linearity of risk within the city of Hong Kong

“…31 If only bicycle exposure is modelled (ie, not including the five control variables), the coefficient is equal to 0.37 and 0.33 in the two periods, respectively, which are comparable with other studies regardless of the spatial scale they used. This suggests that controlling other variables may help better evaluate the effect of bicycle exposure.…”

Rethinking safety in numbers: are intersections with more crossing pedestrians really safer?

“…Our estimates were much lower than those reported by Elvik and Bjørnskau,20 suggesting a stronger SIN effect.…”

Safety-in-numbers: Estimates based on a sample of pedestrian crossings in Norway

“…In the third place, nearly all studies of safety-in-numbers rely on cross-sectional data, which make it difficult to establish causal relationships. In a recent review, Elvik and Bjørnskau (2016) concluded that no studies of safety-in-numbers have controlled adequately for all relevant confounding variables and that one cannot conclude that these studies have uncovered a causal relationship.…”