You are on page 1of 8

This article was downloaded by: [46.165.210.

17] On: 12 November 2013, At: 16:05 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Traffic Injury Prevention


Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/gcpi20

Effects of Red Light Cameras on Violations and Crashes: A Review of the International Literature
Richard A. Retting , Susan A. Ferguson & A. Shalom Hakkert
a b a a b

Insurance Institute for Highway Safety , Arlington, Virginia, USA

Transportation Research Institute , Technion City, Haifa, Israel Published online: 15 Sep 2010.

To cite this article: Richard A. Retting , Susan A. Ferguson & A. Shalom Hakkert (2003) Effects of Red Light Cameras on Violations and Crashes: A Review of the International Literature, Traffic Injury Prevention, 4:1, 17-23, DOI: 10.1080/15389580309858 To link to this article: http://dx.doi.org/10.1080/15389580309858

PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the Content) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http:// www.tandfonline.com/page/terms-and-conditions

Trafc Injury Prevention, 4:1723, 2003 Copyright C 2003 Taylor & Francis 1538-9588/03 $12.00 + .00 DOI: 10.1080/15389580390120862

Effects of Red Light Cameras on Violations and Crashes: A Review of the International Literature
RICHARD A. RETTING and SUSAN A. FERGUSON
Insurance Institute for Highway Safety, Arlington, Virginia, USA

A. SHALOM HAKKERT
Downloaded by [46.165.210.17] at 16:05 12 November 2013
Transportation Research Institute, Technion City, Haifa, Israel

Red light running is a frequent cause of motor vehicle crashes and injuries. A primary countermeasure for red light running crashes is police trafc enforcement. In recent years, many police agencies have begun using automated red light cameras as a supplement to conventional enforcement methods. The present study reviewed and evaluated available evidence in the international literature regarding the effectiveness of cameras to reduce both red light violations and crashes. Camera enforcement generally reduces violations by an estimated 4050%. In terms of crash effects, most studies contain methodological aws that, to varying degrees, either overestimate (failure to adjust for regression to the mean) or underestimate (comparison with nearby signalized intersections affected by cameras) crash effects. Mindful of these limitations, the research generally indicates that camera enforcement can signicantly reduce injury crashes at signalized intersections, in particular right-angle injury crashes. Most studies reported increases in rear-end crashes following camera installation. Taken together the studies indicate that, overall, injury crashes, including rear-end collisions, were reduced by 2530% as a result of camera enforcement. Keywords Automated Enforcement; Red Light Cameras; Red Light Running Crashes; Right-Angle Crashes; Signalized Intersections

Red light running is a frequent cause of crashes at signalized intersections. During 199298, almost 6,000 people (about 850 each year) died in red light running crashes in the United States, and another 1.4 million were injured in crashes that involved red light running (Insurance Institute for Highway Safety, 2000). Red light running is also a problem in other countries. Greene (2000) studied crashes at signalized intersections in the Australian states of Victoria, Western Australia, and Queensland during 199498, nding that 15% to 21% of the crashes were related to red light running. A primary countermeasure for red light running crashes is police trafc enforcement. In recent years, many U.S. police agencies have begun using automated cameras as a supplement to conventional enforcement methods. Red light cameras are not a new technology. They were used for trafc enforcement in Israel as early as 1969 (Levinson, 1989), in Europe in the
Received 28 August 2002; accepted 4 October 2002. Address correspondence to Richard A. Retting, Insurance Institute for Highway Safety, 1005 North Glebe Road, Arlington, Virginia 22201, USA. E-mail: research@iihs.org

early l970s, and in Australia on a wide scale in the 1980s. In the United States, about 70 communities have implemented red light cameras (Insurance Institute for Highway Safety, 2002), beginning with New York City in the early 1990s. Red light cameras offer the potential for trafc enforcement 24 hours a day and produce a record of evidence that cannot easily be disputed. Most important is the deterrent effect of discouraging drivers from breaking trafc laws in the rst place. The basic technology behind red light cameras was developed in the 1960s. The camera system monitors the status of the trafc signal by an electronic connection to the signal controller, and most systems determine vehicle presence using electromagnetic sensors buried in the pavement near the intersection entry point. Cameras record images of an offending vehicle and the surrounding scene as well as the date and time of offense, vehicle speed, duration of the yellow signal, and how long after the red signal the vehicle began to enter the intersection. Typically, a second photo is taken to verify that the vehicle proceeded through the intersection on the red signal. Technical advances in video processing and digital technology will allow the use of video and digital cameras as alternatives to conventional 17

18

R. A. RETTING ET AL.
Table I Studies of effects of red light cameras on violations Percent change 42 27 +17 22 +13 78 67 56 44 34 +5 40 50 4 69 38

Downloaded by [46.165.210.17] at 16:05 12 November 2013

wet-lm devices. Most red light camera systems are portable and can be deployed at numerous intersections that are equipped with the necessary sensors and connections to the trafc signal. Camera housings can be placed at multiple intersections without cameras actually being installed. By placing such housings at many different intersections, more areas per camera can be covered and driver deterrence potentially can be increased. Published evaluations of red light camera effectiveness as well as numerous anecdotal reports generally indicate that cameras reduce both red light violations and associated crashes. However, the methods and results of these studies vary considerably, and to date there has been no effort to synthesize changes in crash data resulting from these research efforts. To the extent that camera enforcement reduces red light running, it would be expected to reduce the frequency of right-angle crashes at signalized intersectionsthe principal type of crash associated with red light running. Such enforcement also may result in increases in rear-end crashes. It has been known for some time that trafc signals themselves increase rear-end crashes (Hakkert & Mahalel, 1978). Differences in driver speeds and stopping behavior during the signal change can contribute to rear-end crashes. For example, if drivers stop more often for red lights, drivers not intending to stop may strike them from behind. Red light cameras, because they can increase stopping behavior, may exacerbate this effect. Therefore, it is important to monitor the incidence of rear-end crashes as well as right-angle crashes after camera enforcement. The present study provides a critical review and synthesis of international research regarding the effectiveness of red light camera enforcement. The principal focus of this synthesis is to better understand the effects of camera enforcement on crashes. However, before examining crashes it is important to better understand the effect of camera enforcement on red light violations across the community, as this can affect the interpretation of crash studies. EFFECTS OF RED LIGHT CAMERAS ON VIOLATIONS Although red light cameras have been in use for several decades, controlled evaluations of their effects on signal violations and associated crashes are relatively few. A review of the international literature revealed seven studies (see Table I) that included information on violation rates before and after initiation of camera enforcement. Two of the studies (Oei et al., 1997; Thompson et al., 1989) monitored red light violations only at camera sites. Reductions in violations of 22% to 56% were reported in these studies, although one author (Thompson et al.) failed to nd a reduction at one of the two sites studied. These ndings are supportive of camera enforcement, but the absence of comparison observations at similar intersections without camera enforcement makes it difcult to determine how much of the reduction was caused by the cameras and how much was due to other factors. For example, if problem intersections were selected for camera enforcement, some reduction in violations would be expected from the phenomenon of regression

Study Chin (1989)

Country Singapore

Study sites 23 Camera sites 20 Noncamera sites 14 Control sitesa Camera site 1 Camera site 2 3 Camera sites 3 Noncamera sites 4 Camera sites 5 Camera sites 2 Noncamera sites 2 Control sites 9 Camera sites 3 Noncamera sites 2 Control sites 4 Noncamera sites after 1 mo 4 Noncamera sites after 6 mo

Thompson et al. (1989) Arup (1992) Oei et al. (1997) Retting et al. (1999a)

Great Britain Australia The Netherlands United States

Retting et al. (1999b)

United States

Chen et al. (2001)

Canada

a Control sites in this study were located in the same community as the treatment sites; however, signs indicating the presence of cameras were installed at camera sites as well as noncamera sites, but not at control sites.

to the mean. (This statistical issue is discussed in greater detail in the following review of camera enforcement and crashes, but it is also relevant for violations.) In addition, these studies do not address the issue of general deterrence; that is, whether red light cameras at a few intersections can generate a halo effect of community-wide reductions in signal violations. General deterrence is a goal, and many communities post warning signs about camera enforcement in areas without cameras to enhance this effect. To address these issues, four studies in Table I compared changes in red light violations at signalized intersections with and without camera enforcement (Arup, 1992; Chin, 1989; Retting et al., 1999a,b). All four studies included comparisons between intersections with and without cameras within the same community, which provides a measure of the general deterrence effect. In addition, three of the studies included comparisons with control intersections either in other communities without camera enforcement (Retting et al., 1999a,b) or in the same community at locations where motorists knew camera enforcement was not deployed (Chin, 1989). The control sites were chosen to be similar to the camera sites, and these comparisons were included to assess whether changes observed in red light violations in the camera enforcement communities were, in fact, due to the cameras. All four of these studies reported a large halo effect. Arup (1992) and Retting et al. (1999a) reported reductions in red light violations at noncamera sites in the same community that were nearly as large as the reductions at the camera sites; Retting et al. (1999b) reported reductions at noncamera sites that were larger than at camera sites. At the same time, comparison with similar sites for which no camera enforcement was expected indicated that these reductions were not likely to have been caused by regression to the mean or other statistical artifacts; violation

EFFECTS OF RED LIGHT CAMERAS

19

Downloaded by [46.165.210.17] at 16:05 12 November 2013

rates at these sites showed little change (Retting et al., 1999a,b) or even increased during the periods of the studies (Chin, 1989). One additional study (Chen et al., 2001) investigated general deterrence effects of red light cameras at noncamera sites in two cities that implemented camera enforcement, but did not monitor effects at camera-enforced intersections. The analyses revealed a 69% reduction in the rate of red light violations at the noncamera intersections 1 month after introduction of cameras and a 38% reduction after 6 months. Overall, the results suggest that camera enforcement leads to large reductions in red light running violations, ranging from 22% to 78%. The study that reported the smallest reduction (Thompson et al., 1989) included an intersection with increased red light violations after camera enforcement was introduced. However, it is noteworthy that camera enforcement warnings were not posted in this study, and it may be that the very low response reected a lack of driver awareness of camera enforcement. In addition to the typically large reductions at the camera sites, there was reduced red light running at other intersections around the community, demonstrating a strong halo effect.

EFFECTS OF RED LIGHT CAMERAS ON CRASHES The review of the international literature revealed eight studiesfrom Australia, Singapore, and the United Statesthat provided crash data before and after introduction of camera enforcement and that included data from noncamera comparison sites as well as sites subject to camera enforcement. Three additional studies were identied but not included because they did not contain crash data for comparison sites (County Surveyors Society, 1990; Hooke et al., 1996) or because data were not provided to allow a before-and-after comparison (Ofce of the Auditor General, 1996).
Table II Study Studies of effects of red light cameras on crashes Period Treatment sites

Most of the studies reviewed had at least one of two common methodological aws. Most compared crashes at cameraequipped signalized intersections with other signalized intersections in the same jurisdiction that did not have cameras. Nearby signalized intersections cannot serve as adequate controls because, as indicated by the red light violation studies, there typically is a large halo effect of camera enforcement at noncamera sites. Crash reductions estimated by comparing camera with noncamera sites therefore are likely to underestimate the effect of cameras on crashes. The second methodological problem is associated with the sites chosen for camera enforcement. Most of the camera sites were chosen because they had a high number of crashes. It has been demonstrated that extreme values in a distribution, such as a particularly high number of crashes in an area during one time period, will tend to move toward the average of the entire group of such areas in the succeeding time period, even if nothing is done to affect the crashes (Campbell & Stanley, 1963). This phenomenon, recognized for years by statisticians and researchers in many elds (Galton, 1889), is known as regression to the mean. This can result in an overestimation of the effects of camera enforcement on crashes. Choosing comparison sites with similar characteristics can partially, but not fully, address regression to the mean. Crash rates over time may be more variable at camera sites than at comparison sites because of the selection bias associated with selecting intersections with the highest number of crashes for camera enforcement. Statistical techniques have been developed for estimating and adjusting for the regression-to-the-mean effect individually at each site (see Hauer, 1997), but such techniques were not employed in these studies. The studies were combined into three groups for analysis, according to how they addressed these methodological challenges (Table II). The rst group (Andreassen, 1995; Mann et al., 1994; Ofce of Road Safety, 1991; Queensland Transport, 1995) did

Comparison sites

Group 1Studies that did not account for regression to the mean and spillover effects Ofce of Road Safety (1991) 198489 15 Camera-equipped intersections in Adelaide, All other signalized intersections in South Australia Metropolitan Adelaide Mann et al. (1994) 198393 8 Camera sites where no other engineering changes were made. 14 High-volume signalized intersections, Study conducted in Adelaide, South Australia selected to be as comparable as possible Queensland Transport (1995), 198699 79 Camera-equipped intersections in Queensland, Australia Other signalized intersections (presumably all with updated data provided other signalized intersections in Queensland) Andreassen (1995) 197989 41 of the 46 Camera-equipped intersections from the All signalized intersections in Metropolitan previous study by South et al. (1988). Study conducted Melbourne in Melbourne, Victoria Australia Group 2Studies that did partially address regression to the mean but not spillover effects South et al. (1988) 198186 46 Camera-equipped intersections in Melbourne, 46 Noncamera sites Victoria, Australia Hillier et al. (1993) 198691 16 Camera intersections in Sydney, New South Wales, Australia 16 Matched noncamera sites Ng et al. (1997) 198593 42 Camera-equipped intersections in Singapore, phased in 42 Matched noncamera sites between 1988 and 1990 Group 3Study that did account for regression to the mean and spillover effects Retting & Kyrychenko 199599 125 Signalized intersections in Oxnard, California, where 11 3 California cities and nonsignalized (2002) intersections were equipped with red light cameras intersections in Oxnard

20

R. A. RETTING ET AL.

Table III

Estimated percent crash effects of red light camera enforcement (margin of error) Total crashes 8a (11) +6a (15) 48 (5) +7a (9) 8a (21) 7 (6) Injury crashes 23 (22) 20a (27) 46 (8) 7a (16) 26a (29) 9a (17) 29 (11) Right-angle crashes Total 38 (14) +8a (38) 13a (15) +29a (156) 32 (25) Injury 54 (21) 26a (53) 32 (30) 10a (21) 68 (10) Rear-end crashes Total +14a (17) +12a (20) +20 (18) +108 (100) +3a (174) Injury +25a (54) 1a (53) 31a (34) +6a (56)

Study Ofce of Road Safety (1991) Mann et al. (1994)b Queensland Transport (1995) Andreassen (1995)b South et al. (1988) Hillier et al. (1993) Ng et al. (1997) Retting & Kyrychenko (in press)
a Not

Note. Margin of error is dened to be one-half the width of the 95% condence interval. statistically signicant; all other percentage changes are statistically signicant. b Crashes in all groups include only right-angle, right-turn, and rear-end crashes.

not account for halo effects or regression to the mean. Not only were the cameras located at high crash locations (leading to overestimation), but the studies used signalized intersections in the same community as comparison sites (leading to underestimation). The second group of studies (Hillier et al., 1993; Ng et al., 1997; South et al., 1988) partially addressed the regression-tothe-mean effect by selecting comparison intersections without cameras that had attributes similar to camera sites, thus reducing the overestimation problem. These studies did not, however, address the halo effect; thus, camera effects probably were underestimated. The third group (Retting & Kyrychenko, 2002) comprised only one study that accounted for both effects. Crash data were not uniform across the eight studies. For example, the studies used different denitions for rear-end crashes. Some included rear-end crashes in which the lead vehicle was turning, some excluded rear-end turning crashes, and others did not specify the types of rear-end crashes studied. Data regarding total crashes were provided for all studies except Ng et al. (1997) and South et al. (1998). Data regarding injury crashes were provided for all studies except Andreassen (1995). Data regarding right-angle and rear-end crashes were not provided for the Queensland Transport study. Table III summarizes the reported crash effects. Results are provided (where available) for total crashes, total injury crashes, and for two types of crashes of particular interest with regard to red light camerasright-angle crashes, which could be expected to decrease following camera enforcement, and rear-end crashes, which might increase if drivers stop more often for red lights following camera enforcement. Total crashes and total injury crashes include data for crash types other than right-angle or rear-end.

GROUP 1 STUDIES The Ofce of Road Safety (1991) analyzed changes in crash frequency at 15 camera intersections in Adelaide, South Australia, before and after camera enforcement (ve cameras rotating among 15 sites). Warning signs were posted at all camera intersections, and a publicity campaign was conducted. All

other signalized intersections in Metropolitan Adelaide served as comparison sites. Sites were selected on the basis of high incidence of right-angle crashes. Following camera enforcement, there was an 8% reduction in total crashes (not signicant) and a 23% reduction in injury crashes. Regarding specic types of crashes, total right-angle crashes and those involving injuries were reduced 38% and 54%, respectively, at camera sites relative to comparison sites. These right-angle crash reductions were offset only partially by increases in rear-end crashes. Mann et al. (1994) analyzed changes in crash frequency at eight camera sites in Adelaide, South Australia (six additional camera sites were studied by Mann et al., but they received trafc engineering improvements in addition to cameras and thus were excluded from this review). Warning signs were posted at all camera intersections, and a general publicity campaign was conducted. For a comparison group, 14 high-volume signalized intersections, similar in terms of conguration and location, were selected. However, the authors reported that some general engineering improvements and an increase in yellow signal timing were implemented at the comparison sites over the years, raising further questions about the comparability of the sites. Following camera enforcement, total crashes tended to increase (+6%) whereas injury crashes decreased (20%). This pattern was true for both right-angle and rear-end crashes, but none of the estimated effects were statistically signicant. In Queensland, Australia, 79 camera intersections were the subject of one of the largest evaluations of red light camera enforcement to date (Queensland Transport, 1995). For comparison sites, the authors used all other signalized intersections in Queensland. Warning signs were posted only at intersections with cameras, but it was not reported whether a publicity campaign accompanied enforcement. The study authors provided updated crash data through 1999 specically for this review. Data for the study period cover 14 years. During the study period, there was a large growth in crashes at the comparison sites, part of which may have occurred because intersections were added to the comparison sites (new trafc signals were installed). This would lead to overestimates of camera effectiveness. Following camera enforcement, there was a 48% decline in total crashes (not signicant) and a 46% decline in injury crashes at camera

Downloaded by [46.165.210.17] at 16:05 12 November 2013

EFFECTS OF RED LIGHT CAMERAS

21

Downloaded by [46.165.210.17] at 16:05 12 November 2013

sites relative to comparison sites. No other study has reported such large effects of red light cameras on total crashes. Information was not available to separately analyze right-angle or rear-end crashes. The Andreassen (1995) study was a follow-up evaluation of 41 of the 46 camera intersections from a previous study by South et al. (1988) in Melbourne, Australia (this study is discussed in group 2). Warning signs were posted at all camera intersections, and an extensive publicity campaign was conducted. Five of the original 46 camera sites were excluded because crash data were unavailable. The authors included as comparisons all signalized intersections in Metropolitan Melbourne (data for the 41 treatment sites apparently also were included). Data for the study period cover 11 years, during which there were many changes in the composition of both camera and comparison sites (as indicated by the author) as well as an extensive area-wide speed camera enforcement program in Melbourne. The author stated that trends in accidents at the camera sites cannot be compared reliably with those produced by the VicRoads database. In other words, data for the comparison group (taken from VicRoads) were not as complete as data for the camera sites. Following camera enforcement, there was a nonsignicant 7% increase in total crashes at camera sites relative to comparison sites, primarily because of a 20% increase in rear-end crashes. Information was not available regarding injury crashes. GROUP 2 STUDIES South et al. (1988) analyzed injury crashes before and after camera enforcement at 46 camera intersections and 46 noncamera comparison sites in Melbourne, Australia. This was a tighter analysis than the follow-up study by Andreassen (1995) because the comparison sites were randomly drawn from the same pool of high-crash locations as the treatment sites (Andreassen compared the camera sites with all signalized intersections in Melbourne). Following camera enforcement, there was an overall 7% decline in injury crashes (not signicant) at camera sites relative to comparison sites. Both right-angle and rear-end injury crashes declined. South et al. did not report on noninjury crashes. Hillier et al. (1993) analyzed changes in crash frequency at 16 camera intersections and 16 matched noncamera sites in Sydney, Australia, before and after introduction of cameras. Warning signs were posted at all camera intersections, and a general publicity campaign was conducted. To reduce site selection bias, camera and noncamera comparison sites were matched on the basis of crash history, trafc volume, and intersection conguration. Following camera enforcement, there were reductions in both total crashes (8%, not signicant) and total injury crashes (26%) at camera sites relative to comparison sites. This reduction occurred despite a 108% increase in rear-end crashes at camera sites relative to comparison sites. Right-angle crashes increased 29% relative to noncamera sites. In one of the few crash-based studies conducted outside of Australia, Ng et al. (1997) analyzed changes in crash frequency

at 42 camera intersections in Singapore, phased in between 1988 and 1990, and 42 noncamera sites. Warning signs were posted only at camera intersections, but it was not reported whether a publicity campaign was conducted. Camera and noncamera comparison sites were matched on the basis of crash history and intersection conguration. Following camera enforcement, total injury crashes declined 9%, right-angle injury crashes declined 10%, and rear-end injury crashes increased by 6%. None of these differences were statistically signicant. Information was not provided for noninjury crashes.

GROUP 3 STUDY Retting and Kyrychenko (2002) evaluated the crash effects of red light camera enforcement in Oxnard, California, where 11 intersections were equipped with red light cameras. Cameras were publicized by various means, including posted signs throughout the city and postcards mailed to city residents. On the assumption that cameras should affect behavior and crashes at all signalized intersections in Oxnard, the authors analyzed crash data for all signalized intersections, specically not the 11 camera intersections. This procedure removes both the halo effect and regression-to-the-mean difculties. However, to be sure that any changes in crashes observed at signalized intersections in Oxnard did not result from general trends in California, the changes were compared with data from signalized intersections in three other California cities. To ensure that these changes were independent of general trafc changes in Oxnard, the changes were compared with crash experience at nonsignalized intersections. Following camera enforcement, there was a 7% reduction in total crashes, a 29% reduction in injury crashes, a 32% reduction in right-angle crashes, and a 68% reduction in right-angle injury crashes at Oxnards signalized intersections relative to comparison sites. There also was an increase in rear-end crashes, but it was small and not signicant. Although not part of the principal analysis, a comparison of injury crashes at the 11 camera intersections and other signalized intersections in Oxnard showed similar reductions. Injury crashes decreased 25% (from 52 to 39 crashes) at camera sites and 19% (from 247 to 200) at signalized intersections without cameras. Right-angle crashes and right-angle injury crashes also decreased at both camera and noncamera signalized intersections. These ndings conrm the expected halo effect.

COMBINED RESULTS To better understand how the estimated effects of red light cameras are inuenced by the methodological problems of regression to the mean and uncontrolled generalizations of camera enforcement effects to neighboring intersections, weighted odds ratios were constructed that combined the effects of the different studies within each group. These ratios considered only total

22
Table IV

R. A. RETTING ET AL.
Estimated combined effect of red light camera enforcement on injury crashes Treatment Study Beforea Aftera Comparison Beforea Aftera Odds ratio Weight 95% Condence interval Weighted mean effect

Group 1Studies that did not account for regression to the mean and spillover effects Ofce of Road Safety (1991) 367 63 4,884 1,086 0.772 50.70 Mann et al. (1994)b 147 86 220 160 0.804 34.22 Queensland Transport (1995) 430 445 3,803 7,256 0.542 201.06 Average 0.605 Group 2Studies that did partially address regression to the mean but not spillover effect South et al. (1988) 596 450 625 544 136.27 (4.06)c (3.49)c (4.41)c (4.06)c 0.934 Hillier et al. (1993) 127 82 123 108 0.735 26.70 Ng et al. (1997) 520 386 510 415 0.912 112.56 Average 0.904 Group 3Study that did account for regression to the mean and spillover effects Retting & Kyrychenko (2002) 299 239 173 194 0.713

0.5390.680

39%

0.8031.017 0.6090.834

10% 29%

Downloaded by [46.165.210.17] at 16:05 12 November 2013

of injury crashes. crashes in all groups include only right-angle, right-turn, and rear-end crashes. c Numbers in parentheses are crash rates per intersection per year and are corrected for different before and after periods in both treatment and control sites.
b Injury

a Number

injury crashes because many studies did not provide information separately for right-angle and rear-end crashes; in addition, the study by Andreassen (1995) was excluded because it provided no information on injuries. Table IV summarizes these data. The odds ratios for each study were constructed as the ratio of the number of crashes after camera enforcement to the number before. The ratio for each study was then weighted according to the number of crashes it contributed to its group total and averaged with the ratio of other studies in the group; the weighted average method is described in detail by Elvik (2001). Because all injury crashes are the unit of analysis, the weighted mean effects by group take into account any trade-off associated with an increase in rear-end crashes. Results for the rst group of studies showed a statistically signicant 39% reduction in injury crashes compared with control sites. Because these studies corrected for neither regression to the mean nor the halo effect, there was the potential for both overestimation and underestimation of the crash effects. Results for the second group of studies showed a nonsignicant 10% reduction in injury crashes compared with control sites. These studies partially considered the regression-to-the-mean effect by selecting comparable sites for the control group. However, they failed to account for the halo effect, thus increasing the likelihood that crash reductions were underestimated. Results for the third study, which controlled for both regression-to-the-mean and halo effects, showed a statistically signicant 29% reduction in injury crashes compared with control sites; this estimate lies between those of the group 1 and group 2 studies, as expected by the pattern of methodological problems.

indicates that changes in behavior resulting from conventional enforcement methods generally are limited in time and distance and do not produce lasting changes in driver behavior (i.e., Elvik, 2001; Heidstra et al., 2001; Makinen & Oei, 1992). A considerable increase in the risk of apprehension is required in order to achieve high levels of road user compliance with trafc safety laws. In reviewing the international literature, it is clear that red light camera enforcement is highly effective in reducing red light violations and right-angle injury crashes associated with red light running. Although results vary considerably due in part to the methodological weaknesses of the studies, the results all indicate that red light camera enforcement reduces injury crashes; the best estimate is about 2530%. The Retting and Kyrychenko (2002) study, which attempted to address both regression-to-the-mean and halo effects, indicates red light cameras reduce injury crashes by 29%. Rear-end crashes increased in many studies, but rear-end injury crashes increased less and were more than offset by the reductions in right-angle injury crashes. ACKNOWLEDGMENT
This work was supported by the Insurance Institute for Highway Safety, Arlington, Virginia, USA.

REFERENCES
Andreassen, D. (1995) A Long Term Study of Red Light Camera and Accidents, ARRB Report ARR 261, Australian Road Research Board, Victoria, Australia. Arup. (1992) Red Light Camera Evaluation StudyImplementation in Brisbane, Report No. 6221, Arup Transportation Planning for Queensland Transport, Melbourne, Australia. Campbell, D. T., Stanley, J. C. (1963) Experimental and Quasi-Experimental Designs for Research, Houghton Mifin Co., Boston, MA.

DISCUSSION The goal of police trafc enforcement is to achieve high levels of road user compliance with trafc safety laws. Prior research

EFFECTS OF RED LIGHT CAMERAS

23

Downloaded by [46.165.210.17] at 16:05 12 November 2013

Chen, G., Wilson, J., Meckle, W., Casey, R. (2001) General Deterrence Effects of Red Light Camera and Warning Signs in Trafc Signal Compliance in British Columbia, J. Trafc Med., Vol. 29, pp. 4653. Chin, H. C. (1989) Effect of Automatic Red-Light Cameras on Red-Running, Trafc Eng. Control, Vol. 30, pp. 175179. County Surveyors Society. (1990) Trafc Signal Cameras, Report No. 1/12, Special Activity Group on Accident Reduction, County Surveyors Society, Derbyshire, England. Elvik, R. (2001) Area-Wide Urban Trafc Calming Schemes: A Meta-Analysis of Safety Effects, Accid. Anal. Prev., Vol. 33, pp. 327336. Galton, F. (1889) Natural Inheritance, Macmillan, London. Greene, F. (2000) Red Light Running, Proc. Conf. Road Safety Research, Policing, and Education. Hakkert, A. S., Mahalel, D. (1978) The Effect of Signalization of Intersections on Road Safety with Special Reference to the Introduction of a Blinking Green Phase, Trafc Eng. Control, Vol. 19, pp. 212215. Hauer, E. (1997) Observational before-after Studies in Road Safety, Elsevier Science Ltd., Oxford, England. Heidstra, J., Goldenbeld, C., Nilsson, G., Makinen, T., Sagberg, F. (2001) New Concepts in Automatic Enforcement, EU Project ESCAPE, Deliverable 6, Technical Research Centre of Finland, Espoo, Finland. Hillier, W., Ronczka, J., Schnerring, F. (1993) An Evaluation of Red Light Cameras in Sydney, Research Note RN 1/93, Road Safety Bureau, Roads and Trafc Authority, New South Wales, Australia. Hooke, A., Knox, J., Portas, D. (1996) Cost Benet Analysis of Trafc Light and Speed Cameras, Police Research Series Paper 20, Police Research Group, Home Ofce, London. Insurance Institute for Highway Safety. (2000) Red Light Running Factors into More than 800 Deaths Annually; More than Half of Those Who Die Are Hit by Red Light Violators. Available from http://www.hwysafety.org/ news releases/2000/pr071300.htm. Insurance Institute for Highway Safety. (2002) Red Light Cameras in Action. Available from http://www.highwaysafety.org/safety facts/rlc cities.htm. Levinson, J. (1989) Gotcha! How Israeli Police Catch Red-Light Jumpers on Film, Trafc Safety, Vol. 6, pp. 1315.

Makinen, T., Oei, H. L. (1992) Automatic Enforcement of Speed and Red Light Violations, Research Report, Institute for Road Safety Research, Leidschendam, the Netherlands. Mann, T., Brown, S., Coxon, C. (1994) Evaluation of the Effects of Installing Red Light Cameras at Selected Adelaide Intersections, Report Series 7/94, Ofce of Road Safety, South Australian Department of Transport, Walkerville, South Australia. Ng, C. H., Wong, Y. D., Lum, K. M. (1997) The Impact of Red-Light Surveillance Cameras on Road Safety in Singapore, J. Road Transport Res., Vol. 6, p. 2. Oei, H. L., Catshoek, J. W. D., Bos, J. M. J., Varkevisser, G. A. (1997) Project Red-Light and Speed (PROROS), SWOV Report R-97-35, Institute for Road Safety Research, Leidschendam, the Netherlands. Ofce of the Auditor General. (1996) Improving Road Safety-Speed and RedLight CamerasThe Road Trauma Trust Fund, Report 1, Ofce of the Auditor General, Western Australia, Australia. Ofce of Road Safety. (1991) Report on Red Light Camera Program: Operation from July 1988 to December 1989, Ofce of Road Safety, Department of Road Transport, Adelaide, South Australia. Queensland Transport. (1995) Queenslands Road TollQueensland Transport Submission to the Parliamentary Travelsafe Committee, Queensland Transport, Queensland, Australia. Retting, R. A., Kyrychenko, S. Y. (2002) Crash Reductions Associated with Red Light Camera Enforcement in Oxnard, California, Am. J. Public Health, Vol. 92, pp. 18221825. Retting, R. A., Williams, A. F., Farmer, C. M., Feldman, A. F. (1999a) Evaluation of Red Light Camera Enforcement in Fairfax, Virginia, ITE J., Vol. 69, pp. 3034. Retting, R. A., Williams, A. F., Farmer, C. M., Feldman, A. F. (1999b) Evaluation of Red Light Camera Enforcement in Oxnard, California, Accid. Anal. Prev., Vol. 31, pp. 169174. South, D., Harrison, W., Portans, I., King, M. (1988) Evaluation of the Red Light Camera Program and the Owner Onus Legislation, Report SR/88/1, Road Trafc Authority, Victoria, Australia. Thompson, S. J., Steel, J. D., Gallear, D. (1989) Putting Red-Light Violations in the Picture, Trafc Eng. Control, Vol. 30, pp. 122125.

You might also like