Objective To assess the effectiveness of speed cameras in reducing the numbers of crashes and people injured on the arterial roads of Barcelona, and to assess their long-term effectiveness on the beltway.
Methods Time series analyses were performed separately for the arterial roads and the beltway. The stretches of arterial roads encompassing 500 m before and after the location of a speed camera were considered the enforced stretches, the remaining stretches of arterial roads being considered the comparison group. The outcome measures were the numbers of crashes and of people injured. Quasi-Poisson regression models were fitted, controlling for time trend, seasonality and implementation of other road safety measures.
Results Both on the enforced and non-enforced arterial road stretches, the risks of crashes and people injured were similar in the two periods. On the beltway, reductions of 30% (95% CI 38% to 20%) and 26% (95% CI 36% to 14%) were observed, respectively.
Conclusions Speed cameras do not reduce the numbers of crashes or people injured on the arterial roads of Barcelona. However, they are effective in the short and in the long-term on the beltway. Speed enforcement through fixed speed cameras is thus effective in medium–high-speed roads, although effectiveness could not be generalised to roads with lower speed limits and traffic lights.
- evaluation studies
- intervention studies
- public health
- wounds and injuries
Statistics from Altmetric.com
- evaluation studies
- intervention studies
- public health
- wounds and injuries
Road traffic injuries cause great mortality and morbidity worldwide. Speed is related both to the risk of being involved in a crash and to its severity.1–4 Speed limits are applied to promote road safety. Enforcement of these speed limits must be intense enough to ensure that drivers perceive the risk of being caught if they exceed the limit. In this context, the use of speed cameras is increasing.5
Several studies have evaluated the impact of speed cameras on road safety. Two systematic reviews have been performed on the issue.5 6 The most recent review reports a 9–35% crash reduction attributable to the installation of speed cameras and a 7–30% injury reduction.5 Published after these reviews, a study by Jones et al7 observed a 19% crash reduction and a 44% reduction in serious and fatal injuries in England.8 However, both reviews emphasise that the level of evidence of speed cameras' effectiveness is relatively poor due to methodological limitations of the studies. More rigorous studies are needed that account for potential confounders, such as regression to the mean, long-term trends or other interventions. Moreover, most of the studies published to date have assessed mobile speed cameras, whereas little research has evaluated fixed speed cameras.
In the city of Barcelona, over 10 000 traffic crashes occur every year, in which more than 13 000 people are injured and approximately 50 die. With the aim of reducing the number of crashes and their consequences, fixed speed cameras were installed on the beltway of Barcelona on 26 March 2003 and on some of the arterial roads on 1 August 2005.
A previous study that assessed the effectiveness of speed cameras on the beltway of Barcelona after 2 years of activity observed an overall reduction of 27% (95% CI 37% to 15%) for road crashes, whereas no reduction was observed in the comparison group.9
Since the installation of the speed cameras, other road safety measures have been implemented in Barcelona, such as the introduction of the penalty points system on 1 July 2006 and a reform of the Penal Code on 1 December 2007, which considers as crimes certain road use behaviours such as exceeding speed limits or blood alcohol concentration limits. In addition, a regulation was approved in October 2004 that allows car drivers with more than 3 years of experience to ride motorcycles up to 150 cc without needing to pass a motorcycle riding examination.10
The objective of the present study is to assess the effectiveness of speed cameras in reducing the number of crashes and the number of people injured on the arterial roads of the city of Barcelona, and to re-assess their effectiveness on the beltway given the longer follow-up period available.
Description of the intervention
The Barcelona city road network includes a total of 1275 km, of which 24.1 km correspond to a beltway that surrounds the city, 43.4 km are streets considered to be arterial roads and the remaining kilometres correspond to secondary roads.
The arterial roads are defined as trunk roads of two to four lanes that cross the city. They absorb 21.4% of the total vehicle-kilometres travelled. The speed limit is 50 km/h (31.1 m/h) and they are crossed by intersections regulated by traffic lights. The mean speed on these roads during the study period was 21 km/h. Seven visible fixed speed cameras were installed on four of the arterial roads on 1 August 2005, accounting for 29% of the vehicles-kilometres of these roads. They were located at points of entry to the city, where the speed limit changes from 80 km/h (49.7 m/h) to 50 km/h.
On the beltway the speed limit is mostly 80 km/h. There are no traffic lights or intersections. The mean speed on these roads during the study period was 58 km/h. Eight fixed speed cameras were installed on 26 March 2003. They operate from 22 different locations, from which they are randomly moved, in such a way that motorists have the perception that speed is being enforced throughout the whole beltway.9
For both the arterial roads and the beltway, warning signs inform about the speed enforcement through radars. Before their installation, and during the following few months, they were announced by a publicity campaign in all news media. The speed camera takes a photo of the vehicle when it exceeds the speed limit allowed by 5% or more, and a fine is mailed to the registered owner within 2–3 weeks, which can be up to €300 if the limit was exceeded by 50% or more.
Assessment of the effectiveness of speed enforcement through speed cameras was performed separately for the arterial roads and for the beltway in two substudies. In both cases an evaluation study was carried out using an interrupted time-series design.
A quasi-experimental design was used on the arterial roads: the stretches of arterial roads encompassing 500 m before and after the location of the speed cameras were considered the enforced stretches of arterial roads, the remaining stretches of arterial roads being considered non-enforced arterial roads, acting as the comparison group.
No comparison group was available in this case on the beltway, because the comparison group used previously in the 2-year follow-up study was precisely the same arterial roads where the speed cameras now being evaluated are located.
The study population consisted of injury crashes and people injured in road traffic crashes in the city of Barcelona between January 2002 and December 2007 for the arterial roads and between January 2001 and December 2007 for the beltway.
Sources of information
The source of information was the local police accident database, which collects information on road traffic collisions with people injured.
The number of injury crashes and the number of people injured (fatal and non-fatal) were the outcome measures (dependent variables). The number of deaths was not analysed separately because frequency was very low: between eight and 15 per year on the arterial roads and between one and six per year on the beltway.
The explanatory variable (intervention) consisted of a dummy variable that compared the post-intervention period with the pre-intervention period. On the arterial roads, the pre-intervention period covered from January 2002 to July 2005, and the post-intervention period from August 2005 to December 2007. On the beltway, the pre-intervention period covered from January 2001 to March 2003, and the post-intervention period from April 2003 to December 2007. Similarly, two additional dummy variables were created to account for the new motorcycle regulation implemented on October 2004 and the introduction of the penalty points system on 1 July 2006.
A time series regression analysis was performed, with month as the unit of analysis, using Poisson regression models adjusted for overdispersion (quasi-Poisson).11 Potential confounding by time trend and seasonal patterns were controlled for using a linear trend and sine and cosine functions.12 The model for each outcome can thus be summarised as follows:where t is the time period (t=1 for the first month of the series, t=2 for the second month, etc), k takes values between 1 and 6 (for example k=1 for annual seasonality; k=2 for 6-monthly seasonality), T is the number of periods described by each sinusoidal function (for example T=12 months), Xt identifies the pre and post-intervention periods, Zjt other co-variables introduced (ie, the new motorcycle regulation and the penalty points system), j the number of co-variables introduced, and ε the error term. This model estimates the change in the mean number of collisions (and people injured) between the post and the pre-intervention periods. The change in time trend between the two periods was also estimated including a term in the model for interaction between the intervention and the time period (β6Xtt).13 Only the statistically significant terms were included in the final model. Residual autocorrelation was checked for using simple and partial autocorrelation function plots.
Relative risks (RR) and their 95% CI were derived from the adjusted models. The numbers of collisions and people injured prevented by the speed cameras were calculated as the difference between the observed and expected numbers in the post-intervention period. The expected numbers were predicted with the statistical models. Figures are only presented for the series of collisions, because the series for injured people follow the same pattern.
Statistical analyses were carried out using Stata statistical software, release 9.14
Comparison of the pre and post-intervention periods revealed a small decrease in the monthly median of people injured on the enforced stretches of arterial roads—five and four in the pre and post-intervention periods, respectively. Also, a small decrease on the non-enforced stretches was observed—262 and 252 in the pre and post-intervention periods, respectively. The corresponding figures for the monthly median of collisions were three and three, and 188 and 181, respectively (table 1).
On the enforced stretches of arterial roads, the monthly number of collisions depicted in figure 1 did not show a clear pattern when comparing the pre and post-intervention periods. The risk of road crashes was similar in both periods (RR 0.98; 95% CI 0.75 to 1.27), and a non-significant increase was observed in the risk of people being injured (RR 1.37; 95% CI 0.79 to 2.38) (table 1). The time trends were similar in the two periods (data not shown).
On the non-enforced stretches of arterial roads, as on the enforced stretches, the monthly number of collisions did not show a clear pattern (figure 1). The risks in the pre-intervention period were similar to those in the post-intervention period, both for collisions (RR 0.99; 95% CI 0.93 to 1.05) and for people being injured (RR 1.00; 95% CI 0.91 to 1.10) (table 1). Again, time trends were similar in the two periods (data not shown).
Analyses were stratified by traffic density at the moment of the collision (fluent or congested), according to police criteria. Similar results were obtained (data not shown). An attempt to stratify analysis regarding one-vehicle and multiple-vehicle collisions was made. However, there were only 29 one-vehicle collisions on the enforced stretches throughout the study period.
A decrease in the monthly median of people injured, comparing the pre with the post-intervention period, was observed on the beltway. The monthly median of people injured during the pre and post-intervention periods was 79 and 62, and the monthly median numbers of collisions were 48 and 38, respectively (table 1).
On the beltway, the RR obtained showed a statistically significant risk reduction in the post-intervention period for both collisions (RR 0.70; 95% CI 0.62 to 0.80) and people injured (RR 0.74; 95% CI 0.64 to 0.86). The time trend was similar in both periods (data not shown). Therefore, 4 years and 9 months after the installation of the speed cameras, 913 collisions had been prevented and 1219 fewer people had been injured attributable to the installation of the speed cameras (table 1). Figure 2 graphically displays the monthly numbers of observed and expected crashes had the speed cameras not been installed. The observed crashes are consistently lower than the expected crashes.
Analyses stratified by number of vehicles involved showed a risk reduction in one-vehicle collisions (RR 0.73; 95% CI 0.55 to 0.98) and in multiple-vehicle collisions (RR 0.92; 95% CI 0.83 to 1.03).
The present study did not show any effect of fixed speed cameras on the arterial roads of Barcelona. Nonetheless, it confirms the long-term effectiveness of fixed speed cameras installed on the beltway of Barcelona, whose short-term effectiveness had previously been shown.9
The results observed on the arterial roads were unexpected. The characteristics of the arterial roads, with lower speed limits and the presence of traffic lights, compared with those on the beltway, could be influencing the results. Also, the publicity campaign might not have had the same media coverage for the arterial roads as for the beltway. Moreover, given the new motorcycle regulation approved in October 2004, there might be more motorcycles, implying a greater exposure to crashes. A stratified analysis on the arterial roads for collisions with and without the involvement of motorcycles was not feasible due to the small number of collisions on the enforced stretches. An additional explanation could be the high traffic density on these roads. However, limiting the analysis to collisions that occurred in conditions of low traffic density did not support this hypothesis. Finally, the small numbers available on the enforced stretches might have hindered the observation of speed cameras' effectiveness.
The study design and the statistical analysis used allowed controlling for the main confounding factors that usually affect road safety intervention evaluation studies. In addition, other confounding factors (changes in traffic, weather, economic conditions and other safety measures, among others)15 were accounted for by using a comparison group for the arterial roads. Although obtaining an adequate comparison group is not a straightforward task, the authors were able to use the non-enforced stretches. A comparison group was not available for the beltway. Nonetheless, although it may add evidence to the results, it is not compulsory when using time series analysis, as percentage change is only compared along time trends in the same series (ie, estimates are obtained comparing the numbers in the series of the intervention group, which are compared with those obtained in the comparison group, when available). In addition, the local police accident database is an exhaustive database, because the local police has a specific unit in charge of collecting information on road traffic collisions in the city of Barcelona. Migration of crashes upstream or downstream of the camera16 is not likely to be a bias in the present study because the crash count considered an area of 1 km around the speed camera on the arterial roads and the whole length of the beltway, and not only the immediate vicinity of the speed camera. Data on people injured was checked for multiple-vehicle collisions or collisions involving high occupancy vehicles. Only one collision with 19 people injured was observed on the beltway throughout the whole period of study—in the post-intervention period, which is not expected to have affected the results. Some authors have reported the empirical Bayes approach with a comparison group and flow correction to be the best.15 However, the estimates obtained depend on the quality of the prediction models used, which, moreover, may become outdated if, for example, trends in crash risk decline.16 Other authors suggest using ARIMA models.17 Nonetheless, previous studies have observed that Poisson regression models yield similar estimates with a similar goodness of fit of the models. Moreover, their coefficients can be interpreted in terms of RR, which provide a straightforward interpretation of the effectiveness of an intervention.18–20
Among the limitations of the study, one is the small number of collisions observed on the enforced stretches of arterial roads. Nonetheless, the long pre and post-intervention periods available provide stability to the analysis. Although there is no information about the effect in travelling speed, the numbers of collisions and of people injured, the ultimate objective of speed cameras, were analysed. Assessment of the second and fourth power effects of speed on serious injury and fatal crashes was not possible due to the small numbers of traffic fatalities and because injury severity was not properly collected before 2005. Nevertheless, serious injuries (ie, people injured admitted to hospital for more than 24 h) only represent 3.5% of all injuries on the arterial roads and 3.0% on the beltway.
Comparison with previous studies is difficult due to the diversity in the type of interventions (mobile vs fixed speed cameras; urban vs rural location of the cameras) and in the type of analysis performed (simple before–after studies vs ARIMA or Poisson regression). Most studies assessed the effectiveness of mobile speed cameras on several types of intercity roads. A 21% reduction was observed in the number of injury collisions and serious casualties in The Netherlands.21 Christie et al22 observed a 51% reduction in the number of injurious crashes at a distance of up to 500 m from the camera, whereas no reduction was observed for larger distances. A 16% reduction in the number of collisions was observed on a highway in British Columbia, along the whole corridor.23 Also in British Columbia, speed cameras reduced collisions by 25%, serious injuries by 11% and fatalities by 17%.8 In England, a 19% reduction was observed in collisions and a 44% reduction in serious and fatal injuries.7 Hidden speed cameras were estimated to reduce collisions by 11% and people injured by 8%.24 25 The only published study assessing the effectiveness of fixed speed cameras—apart from that by Pérez et al.6—observed a 22% reduction of injurious crashes, 17% attributable to speed reductions.26 In that study the speed cameras were located in rural and urban 30 m/h roads, although results were not stratified according to the type of road. The latest published systematic review reported total crash reductions of 9–35%, 7–30% for injurious collisions and 13–58% for serious injuries.5 In the present study similar results were obtained only on the beltway—a 30% reduction for injurious crashes and a 26% reduction for people injured. The effectiveness of speed cameras might be subject to road characteristics, such as the speed limits, or the presence of traffic lights. On the arterial roads, most probably, the effect of the speed cameras is offset by the effect of the traffic lights.
In conclusion, speed cameras do not reduce road traffic crashes and road traffic injuries on the arterial roads of Barcelona. However, they have been proved effective in the short and in the long-term on the beltway of Barcelona. Speed enforcement through fixed speed cameras has thus been proved to be effective for medium–high-speed roads, although effectiveness could not be generalised to roads with lower speed limits and traffic lights.
What is already known on this subject
Speed is related both to the risk of being involved in a crash and to its severity.
Systematic reviews conclude that speed cameras reduce the number of road traffic crashes by 9–35% and road traffic injuries by 7–30%, although the level of evidence is relatively poor due to methodological limitations of the studies.
Most of the studies published to date have assessed the effectiveness of mobile speed cameras, but little research has evaluated fixed speed cameras.
What this study adds
Speed enforcement through fixed speed cameras reduces road traffic crashes and people injured in medium to high-speed roads. However, they are not effective in roads with lower speed limits and traffic lights.
The authors would like to thank Mercè Navarro, Ángel López, Joan Mañosa, Manuel del Haro and Álex Culubret (Àrea de Prevenció, Seguretat i Mobilitat, Ajuntament de Barcelona). This paper will be included in the thesis of one of the authors (AMN), performed at the Pompeu Fabra University (UPF).
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.