Speed and safety effect of photo radar enforcement on a highway corridor in British Columbia

Abstract

This study evaluates the effect of the photo radar program on traffic speed and collisions at photo radar (PRP) influence locations (PRP location) and interleaving non-PRP locations on the Vancouver Island portion of Highway 17 (Pat Bay Highway) in British Columbia (BC). Simple before–after comparison was used to summarize the speed effect while observational before–after method was employed to estimate the safety effect. To control for regression to the mean and time effect, Empirical Bayes (EB) method with comparison groups was employed in collision analysis. The study found a 2.8-km/h reduction in mean speed and a 0.5-km/h reduction in speed standard deviation at a monitoring site 2 km south of the treatment area. Corresponding to speed reduction, the study revealed a 14%±11% reduction in expected collisions at the PRP locations, a 19%±10% reduction at the non-PRP locations, and a 16%±7% reduction along the study corridor as a whole. No evidence was found for a localized effect in a 2-km range of the photo radar direct influence area, over and above those at the interleaving non-PRP locations. The results support the hypothesis of a distance spillover effect — that the program not only improved safety at the PRP locations, but along the entire enforcement corridor as well. It suggests that the unpredictable nature of the deployments lead drivers to modify their behavior along the length of the corridor because they could not discern ‘safe’ from ‘unsafe’ segments.

Introduction

Unsafe speed is a major contributing factor to traffic collisions in British Columbia (BC). During 1995 ‘unsafe speed’, as judged by police attending collisions, was involved in 37% of all fatal collisions, 15% of all personal injury collisions, and 9% of all property damage only collisions. More than 8000 people were injured and 184 people killed in the 10 564 unsafe speed related collisions in 1995, resulting in severe social and economic cost to British Columbians. It is generally acknowledged that speeding may also play a role in other collisions, not specifically identified by the police as involving ‘unsafe’ speed.

In response to this problem, the BC provincial government and the Insurance Corporation of British Columbia (ICBC) introduced the Photo Radar Program (PRP) in 1996. A previous study described the implementation and assessed the one-year effect of the program in the province, as a whole (Chen et al., 2000). The present study is intended to estimate the speed and safety effect of the program on a selected highway corridor 2 years after the commencement of the program.

A limited number of studies have been found to address the corridor-specific effects of photo radar programs. Europe and Australia pioneered the technology and provided the earlier studies. Photo Radar was introduced in Norway in 1988. Elvik (1997) conducted a before–after study of the effects of the program on collisions, controlling for general trend and regression to the mean. Empirical data from 64 road sections were collected and Bayes method was used in model construction and analysis. The study found a statistically significant 20% reduction in injury collisions associated with photo radar. Further analysis revealed that the effect varied with prior frequency of collisions at different sites. The higher the number of collisions before the program, the greater the effect. As insightfully pointed out by the author, the study did not investigate the change in speed as an intervening effect of the program. Nor did it address the hypothesis that drivers slow down at the photo radar site, only to speed up after passing the site, causing migration of collisions to the next section of highway.

British authorities have made extensive use of photo radar to control traffic speed in the 1990s. London Accident Analysis Unit (1997) conducted a before–after study on the West London Speed Camera Demonstration project. The study examined the collision data on trunk roads 3 years before and 3 years after the introduction of the project. The remaining roads in the area were used as the control sites. The studies revealed an 8.9% reduction in total collisions and a 12.1% reduction in fatal and serious collisions attributable to the photo radar program. The study is limited by its design. No control for the regression of means was incorporated. Limited control for passage of time effects was introduced by the use of the other roads in the area.

The largest operation of a photo radar program was found in Victoria, Australia. In September 1989, Victoria introduced its program with extended deployment of 60 speed cameras. Rogerson et al. (1994), (phase 3) assessed the localized influence of the program. To alleviate the potential contamination of overlapping alcohol-related interventions, the study used low-alcohol-time collisions as the outcome measure. The study did not find clear evidence of a localized safety impact of the photo radar program. The only significant effect was found at high alcohol hours, when the program is confounded with enhanced drinking-driving enforcement. There is no reduction in the number of collisions on the actual day when the speed camera was used or on the following 6 days within a 1-km radius of the deployment sites.

Recently, Oei (1998) reviewed studies of speed enforcement (conventional and electronic), its effects on speed behavior and traffic safety and the potential halo effects based on studies in Europe, Australia and North America. At the location or route level, Oei concludes that traffic speed is reduced substantially as a result of speed enforcement. However, the evidence for a safety effect is sparse and unreliable, due to the lack of control for regression to the mean, the time effect, and the large random variation inherent in collision counts. It appears that further research on speed, safety and the halo effect of photo radar programs, using stringent methods and sufficient data, is needed.

The photo radar program started operation in British Columbia in March 1996. A general description of the program can be found in Chen et al. (2000). The photo radar program started operation on Pat Bay Highway, the study corridor, in April 1996. From April to July, the owners of speeding vehicles were issued warning letters for speeding offenses. Starting from August 2, 1996 violation tickets were issued to vehicle owners. The tickets carry a fine of $100.00 to $150.00, depending on the level of speeding. In general, photo radar is deployed at locations of high collision history or where the local community has expressed concerns about speeding. It was operated between 06:00 and 23:00 h on the study corridor, mostly during daytime hours. The camera takes photographs when the offending vehicle exceeds the speed limit by at least 11 km/h.

In addition to general warning signs at the entry of major highway corridors, the British Columbia Government and the Insurance Corporation of British Columbia (ICBC) conducted an intensive media campaign before, during, and after the initial implementation of the program. A pre-implementation survey of randomly selected BC residents showed that about 95% of people in the province were aware of the photo radar program before its introduction (Viewpoints Research, 1995).

BC photo radar program is based on the general deterrence theory and the theories relating speeds and speed variance to collisions. General deterrence is described by Ross (1982) as: ‘the effect of threatened punishment upon the population in general, influencing potential violators to refrain from a prohibited act through a desire to avoid the legal consequences’ (p. 8). Given the likelihood of being detected and punished in the presence of photo radar enforcement, it was hypothesized that the program would lead to a reduction in the mean and the variance of traffic speed at the deployment sites.

A number of theories postulate relationships between speed and collisions (Shinar, 1997). The simplest and relatively robust ones are based on physics, stipulating that the stopping distance increases exponentially with speed and that the energy dissipated upon collision is proportional to the power of two of speed. The higher the speed of a vehicle, the more likely the vehicle is to be involved in a collision and the more severe the collision. The predictions have been supported (or at least not rejected) by most empirical studies in traffic safety (Nilsson, 1981, McKnight and Klein, 1990, Fildes et al., 1991, Rock, 1995). As summarized by Finch et al. (1994), in general, for every 1-km/h increase in mean traffic speed, collisions rise by about 3%.

The variance in speed affects collisions through its impact on potential inter-vehicle conflicts. It seems reasonable that a larger dispersion in traffic speeds would result in a greater number of overtaking maneuvers, which in turn increase the likelihood of conflicts and collisions, as was demonstrated by Hauer (1971) and has been supported by many empirical studies. The results were epitomized by the much-quoted U shaped curve between collision involvement and deviation from mean speed (Solomon, 1964, West and Dunn, 1971, Shinar, 1997). The risk of collision increases with the speed differential of a vehicle from the median speed of the traffic. In sum, the above arguments lead to the prediction that the photo radar program will reduce the mean and variance of speed, resulting in a reduction in traffic collisions at the enforcement locations.

Competing theories exist with regard to the extended effect of photo radar enforcement on the highway segments adjacent to treatment segments. The BC photo radar program is based on the assumption that there will be a positive spillover effect in speed and collisions primarily due to the mobility of the photo radar units and the resulting unpredictable nature of enforcement. However, a reasonable competing proposition is that drivers will slow down at the PRP locations only to increase speed at non-PRP sections of the highway to compensate for time lost. This compensatory behaviour could result in the unintended collision migration to untreated sites. A study to empirically test the validity of these positions is of practical and theoretical significance.

Consequently, this study set outs to evaluate the corridor-specific effect of the photo radar program by addressing the following questions:

• To what extent is traffic speed and speed variance reduced at photo radar deployment locations and at a speed monitoring location on the highway, 2-km from the nearest deployment location, as a result of the photo radar program?

• To what extent are traffic collisions reduced at PRP deployment locations and at non-PRP interleaving locations along the study corridor as a result of the photo radar program?

• Does the evidence support the traffic collision migration hypothesis? Or, conversely, does the evidence support the spillover effect of the program?