PDF A Nationwide Survey of Self-Reported Red Light Running (2001)

Accident Analysis and Prevention 33 (2001) 735– 741 www.elsevier.com/locate/aap A nationwide survey of self-reported red light running: measuri...

Accident Analysis and Prevention 33 (2001) 735– 741 www.elsevier.com/locate/aap A nationwide survey of self-reported red light running: measuring prevalence, predictors, and perceived consequences Bryan E. Porter a,*, Thomas D. Berry b a Department of Psychology, Old Dominion Uni6ersity, Mills Godwin Building, Room 250, Norfolk, VA 23529 -0267, USA b Department of Psychology, Christopher Newport Uni6ersity, 1 Uni6ersity Place, Newport News, VA 23606 -2998, USA Received 11 January 2000; received in revised form 31 July 2000; accepted 12 September 2000 Abstract A United States probability sample of 880 licensed drivers participated in a telephone survey of red light running perceptions and behaviors. Despite most drivers believing red light running was problematic and dangerous, approximately one in five respondents reported running one or more red lights when entering the last ten signalized intersections. Among several demographic and attitude variables, only age group predicted recent red light running. Specifically, younger respondents were more likely to be violators. Drivers also reported being more likely to run red lights when alone, and were typically in a hurry when speeding up to be beat red lights. Contrary to expectations, frustration was not as important for predicting red light running as it was for other driving behaviors, such as speeding, tailgating, weaving, and gesturing angrily at others. Additionally, drivers perceived and received few consequences for running red lights. Less than 6% had received a traffic ticket for red light running and most believed that police would catch less than 20% of violators. Slightly more than one in ten had been involved in a red light running crash. Respondents most commonly suggested legal initiatives to reduce red light running. Accordingly, we recommend traffic safety experts pursue interventions that apply immediate and consistent negative consequences to violators to change the public’s red light running perceptions and behavior. © 2001 Elsevier Science Ltd. All rights reserved. Keywords: Red light running; Traffic safety; Roadway discourtesy; Aggressive driving; Road rage 1. Introduction lights (Deutsch et al., 1980; Retting and Williams, 1996; Porter and England, 2000), where and when red light All drivers are familiar with red light running running occurs (Retting et al., 1995), and what inter- whether we are the guilty party or the near victim. ventions exist to reduce this risky driving behavior However, the reality of red light running may not be so (Retting et al., 1999a,b). These studies suggest red light well known. Retting et al. (1998) found that 3% of all runners do more than run red lights. They are less fatal crashes between 1992 and 1996 involved red light likely to wear safety belts (Deutsch et al., 1980; Porter running. Red light running related fatalities increased and England, 2000) and tend to have more driving approximately 15% during this time period (from 702 in violations on their records (Retting and Williams, 1992 to 809 in 1996). Perhaps not surprisingly, urban 1996). areas are at greater risk for red light running crashes. The present study adds to this growing literature by Retting et al. (1995) found that red light running was surveying self-reported behaviors and perceptions of involved in 22% of urban crashes. red light running in a national sample. The aim of the The dangers of red light running have led to an survey was to address five red light running issues. increasing interest in understanding people who run red First, the prevalence of recent red light running was assessed (i.e. the frequency of running a red light one or * Corresponding author. Tel.: + 1-757-6834439; fax: +1-757- more times when entering the last ten signalized inter- 6835087. sections). Driver characteristics and attitudes were E-mail address: [email protected] (B.E. Porter). tested as predictors of recent red light running. Second, 0001-4575/01/$ - see front matter © 2001 Elsevier Science Ltd. All rights reserved. PII: S 0 0 0 1 - 4 5 7 5 ( 0 0 ) 0 0 0 8 7 - 7 736 B.E. Porter, T.D. Berry / Accident Analysis and Pre6ention 33 (2001) 735–741 the presence of passengers was thought to be impor- 2.2. Inter6iew procedures tant. Red light running was expected to be less likely when there were passengers in the car, particularly child The Social Science Research Center (SSRC) at Old passengers. Third, we assessed the importance of being Dominion University in Norfolk, VA was contracted to frustrated on the likelihood of running red lights. make the survey calls. SSRC staff trained interviewers Fourth, perceptions of red light running consequences and pilot-tested the survey among their managers prior were assessed. We asked drivers how many red light to beginning official calls. After these tests, authors runners they thought would be caught by police. We edited questions for clarity. Official calls were com- also asked them if they had received a ticket for red pleted between 15:00 and 23:00 h. Eastern Standard light running or been in a crash involving a red light Time on weekdays and between 10:00 and 18:00 h on runner. Fifth, drivers gave their ideas to reduce red Saturdays. The survey required 12–15 min to complete, light running. depending on how detailed respondents were with an- swering the questions. SSRC employees followed standardized procedures 2. Methods for making calls and obtaining participants’ agreement to complete the survey. Callers told respondents who 2.1. Sur6ey instrument they were, where they were calling from, and that they were conducting the survey for traffic-safety educators. The 58-item survey addressed several red light run- Participants were told that their responses would re- ning issues (a copy of the survey is available upon main confidential. Finally, participants were asked if request). Survey questions were designed to be useful to they were at least 18 years old, had a driver’s license, research sponsors of the ‘Stop Red Light Running’ and drove a motor vehicle. If so, they were asked to (SRLR) program who also sponsored this research — complete the survey. If not, they were asked if someone DaimlerChrysler Corporation, the American Trauma else was present who met the criteria. Minors were Society, and the Federal Highway Administration. excluded to avoid the necessity of obtaining parental Sponsor representatives collaborated in designing the permission as would be required per the university’s survey’s content. Institutional Review Board policies. One key item assessed how recently respondents had run red lights. Specifically, respondents were 2.3. The national sample and response rates asked ‘recalling the last ten traffic lights you drove through, how many of them were red when you The national sample used in this study included 880 entered the intersections?’ Respondents were also asked respondents who were randomly selected from a pre-ex- how likely they would run red lights when they had isting database of 5024 USA drivers. The previous passengers or when they were frustrated. Other key database was created by randomly sampling all listed items addressed whether respondents believed red light residential telephone numbers in the United States, running was a problem or dangerous as well as their weighting each state by its 1997 population of residents views on red light running consequences. Finally, they over 15 years of age. Ten states, however, were over- were asked about their ideas to reduce red light run- sampled to provide regional information for the partic- ning. ular research sponsors (DaimlerChrysler Corporation, The survey was assessed for both face and content the American Trauma Society, and the Federal High- validity. We generated a domain of questions tapping way Administration; Porter and Berry, 1999). We re- into both the context and content of red light running. weighted the over-sampled states and included them Questions were created and selected for their relevancy with the others to form a national probability sample and efficiency. The principal criterion for formulating that was weighted appropriately by each states’ popula- questions was to tap the antecedents, behaviors and tion of those over 15. The final n-size of the national consequences that were thought to be important to red sample was limited by the original dataset’s minimum light running. samples of four for Alaska, Delaware, and Wyoming. Reliability was difficult to assess given the different Statistically, the national sample had a 3.3% margin of styles and foci of questions, and different scales for error for a 95% confidence interval. The margin of responses (continuous, categorical, and open-ended). error was useful for interpreting confidence in individ- However, at least one series of questions that should ual items that were answered by the entire sample (e.g. have been related were internally consistent. Questions percent of recent red light runners). asking how likely respondents would tailgate, weave, Two response rates were calculated for the pre-exist- gesture angrily, speed, or run red lights when frustrated ing database prior to creating the national probability demonstrated internal consistency (h =0.71 for the five sample. First, the completion rate was 10.5% of all calls items). attempted (total calls= 47 895; 1939 of failed attempts  B.E. Porter, T.D. Berry / Accident Analysis and Pre6ention 33 (2001) 735–741 737 were business or non-residential numbers that were intersections. On the other hand, almost a fifth (19.4%) incorrectly included in the residential sample). The had done so. Logistic regression analysis was used to completion rate was 21.6% if only live contact calls test demographic and attitudinal predictors of recent were considered (live contact calls=23 222). Ten states red light running. Specifically, a forward-stepwise hi- and DC had live response rates less than 10% (DC had erarchical logistic regression model tested whether the a 0% participation rate out of 43 calls). Besides DC, respondent had run one or more recent red lights. Sex four of the states were in the northeastern United States and age group were entered first into the model to — Connecticut, Delaware, New Jersey, and New York. control for over-sampling of women and older respon- The other six low-responding states were California, dents. On the next step, the following variables were Florida, Hawaii, Iowa, Nevada, and Texas. Calls that compared with stepwise techniques to determine if any were not answered, received a busy signal, or were predicted recent red light running above and beyond answered by an answering machine resulted in the sex and age group — educational level, occupation, telephone numbers being returned to the phone bank parental status (yes vs. no), previous involvement in a for possible re-use. red light running crash (yes vs. no), previous receipt of a traffic ticket for red light running (yes vs. no), believ- 2.4. National sample demographics ing red light running to be a problem (yes vs. no) or dangerous (yes vs. no), residential city size (] 100 000 Most of the 880 respondents in the national probabil- population vs.B 100 000), and degree of frustration ity sample had at least some college experience (65.9%). when driving on urban roads (scaled 1– 10, with 1 being Professional occupations were most common (31.0%), ‘not at all frustrated’ and 10 being ‘very frustrated’). A followed by retired respondents (18.4%), blue collar third step tested the sex by age group interaction. workers (11.7%), and homemakers (10.6%). Most re- Only age group significantly predicted recent red spondents lived in cities with fewer than 100 000 resi- light runners. The results did not change when age dents (76.1%; populations based on 1998 census group was entered into the model simultaneously with estimates; US Census Bureau, 2000b). Most partici- other variables. Therefore, the model was re-calculated pants were women (61.9%) and parents (75.3%). Of the for parsimony with only age group. Analyses of age parents, 48.3% had children less than 20 years old. All group vectors revealed that the 36–45, 46– 55, and age groups were represented, however, the sample was 56+ age groups were less likely than 18– 25-year-olds weighted more heavily toward older drivers. Drivers to report recent red light running. As shown in Table 1, aged 46 and over accounted for 51.7% of the sample drivers aged 56 and older were only 0.30 times as likely (46–55=21.1%; 56 and older=30.6%). Other age as the youngest age group to report recent red light groups were represented in the following percentages, running. Drivers 26– 35 years of age were not signifi- drivers 18–25 years made up 11.3%; 26– 35 made up cantly different from the youngest group. See Table 1 15.0%; and 36– 45 made up 22.0%. for odds ratios, Wald  2-statistics, and significance tests US Census data for 1998 showed that the distribu- for each age group comparison. tion of the sexes and age groups generally matched our sample with some exceptions. Women were over-sam- 3.2. Passengers and red light running pled (52.0% of the US population aged 18 and older; US Census Bureau, 2000a), as were the older drivers Passengers reduced drivers’ tendencies to run red (45.8% of the US population were 45 and older). The lights. When alone, 25.6% of drivers reported being at percentage of younger drivers who participated was least somewhat likely to run a red light. This percentage slightly less than estimated to exist (12.7% of the 1998 drops sharply when passengers are in the vehicle. Only population was 18– 24 years of age). The potential biases resulting from over-sampling women and older Table 1 drivers were addressed in the first analysis of red light Logistic regression statistics for predicting recent red light runners running behavior and were considered possible limita- (N =880)a tions to the study’s generalizability to all of the nation’s drivers. Variable OR 95% CI Wald  2 Age group 29.41** 26–35 vs. 18–25 0.88 0.50–1.54 0.20 3. Results 36–45 vs. 18–25 0.54 0.31–0.94 4.78* 46–55 vs. 18–25 0.34 0.19–0.61 12.79** 56+ vs. 18–25 0.30 0.17–0.53 17.86** 3.1. Recent red light runners a OR, odds ratio; ORs predicted recent red light running. The Most drivers (80.6%) reported that they had not run constant term was included in the model, but was not reported to a red light when entering the previous ten signalized save space. *, PB0.05; **, PB0.001. 738 B.E. Porter, T.D. Berry / Accident Analysis and Pre6ention 33 (2001) 735–741 Table 2 speed 20 mile/h (32.2 km/h) over the limit, (b) tailgate, Perceived time constraints and frustration effects on red light (c) weave in and out of traffic, and (d) gesture angrily runninga at other drivers or pedestrians when frustrated. All Comparisons N Percent likelihoods were scored 1–10, with 1 equal to ‘not at all likely’ and 10 equal to ‘very likely’. As one would Late and approaching intersection that is about to ha6e a red light expect, each behavior’s likelihood was correlated with — what do you do? (n = 877) the likelihood of performing other behaviors (see Table Slow down and prepare to stop 625 71.3 Speed up to beat light 252 28.7 3). Fig. 1 displays each behavior’s likelihood plotted against the overall degree of urban-driving frustration, If slowed down and prepared to stop — reason? ( for n = 625) Safe thing to do 362 57.9 scored as 1 for ‘not at all frustrated’ to 10 for ‘very Following the law 96 15.4 frustrated.’ For example, respondents who were ‘not at Afraid of getting hurt in crash 71 11.4 all frustrated’ on urban roads had an average red light My responsibility to stop 49 7.8 running score of 1.18, or a low likelihood of performing Other 47 7.5 the act. As urban-driving frustration increased toward a If speed up to beat the light — reason? ( for n = 252) score of 10, the likelihood of each behavior generally In a rush 88 34.9 increased. Red light running, however, was an excep- To save time 86 34.1 Frustrated with having to stop again 29 11.5 tion in that its slope was closest to horizontal. Enjoy the thrill of beating the light 7 2.8 Respondents typically reported being more likely to Other 42 16.7 do other behaviors than run red lights regardless of What makes you frustrated on urban roads? ( for those who were urban-driving frustration. A multivariate analysis of frustrated; n = 708) variance (MANOVA) was used to test the significance Discourteous drivers 308 43.5 of this finding. The likelihoods of doing each of the Congestion 147 20.8 behaviors when frustrated were treated as dependent Drivers not following the law 90 12.7 measures and the degree of overall urban-driving frus- Too many stop lights 28 4.0 Long commute 7 1.0 tration was treated as the independent variable. Overall Other 128 18.1 urban-driving frustration predicted the likelihood of doing at least one of the behaviors, F(5,874)=22.12, a Three respondents were incorrectly coded, and were omitted from PB0.001. A planned univariate contrast found that as these analyses. urban-driving frustration increased, respondents were on average more likely to speed, weave, tailgate, and 15.8% of respondents reported being at least somewhat gesture than run red lights, F(1,878)= 36.49, PB0.001. likely to run red lights when one adult passenger was Unlike the other behaviors, red light running may not present and only 4.8% admitted being likely when there be as much a function of frustration. However, it is were children in the car. important to note that overall likelihoods of perform- ing any of the behaviors when frustrated were clustered 3.3. Time pressures and frustrations at the lower end of the 1–10 likelihood scale. Respondents were asked to voice concerns about 3.4. Percei6ed and real consequences of red light time and frustration on urban roads. These results are running summarized in Table 2, demonstrating that a large number of drivers, though not a majority, were willing Respondents were doubtful that police would catch to speed up to beat an oncoming red light. When asked most red light runners. On average, they believed fewer why they would speed up, drivers reported being in a than two out of ten violators (M= 1.92; S.D.=1.83) rush or wanting to save time. Drivers who slowed down did so for safety reasons. Table 3 Results involving the causes of urban driving frustra- Pearson correlations for the likelihood of drivers performing five tions were particularly interesting. The majority of driv- behaviors when frustrated (N= 880)a ers were more frustrated with discourtesy (as they Number Behavior 1 2 3 4 defined it) than they were with any other driving prob- lem, including congestion. The frequencies of choosing 1 Running red lights each frustration did not significantly differ between 2 Weaving 0.36 recent red light runners and others. 3 Speeding 0.30 0.38 4 Tailgating 0.28 0.51 0.34 Respondents were also asked how likely they would 5 Gesturing angrily 0.28 0.33 0.22 0.32 run red lights when frustrated. As a comparison, re- spondents were also asked how likely they would (a) a All correlations were significant at PB0.001.  B.E. Porter, T.D. Berry / Accident Analysis and Pre6ention 33 (2001) 735–741 739 Fig. 1. Likelihood of engaging in five behaviors when frustrated as a function of overall urban-driving frustration. would be caught. Interestingly, recent red light runners respondents suggesting either more education (15.2%) believed more violators would be caught (M = 2.36) or required attendance at driver improvement clinics than did other participants (M =1.82), t(227.62) = (1.5%). Respondents were not asked to elaborate on 3.11, PB 0.01 (df adjusted for unequal variances giving examples of more public education. Changing signal a more conservative test). Only 5.8% of the respondents timings was offered by only 2.8% of the participants. reported receiving one or more red light running traffic Discouragingly, more than one in five (23.1%) had no tickets. Recent red light runners (6.4%) did not differ ideas to prevent red light running. There were no significantly from others (5.6%) in ticket receipts. statistically significant differences between recent red The majority of respondents believed that red light light runners and others in the percent offering each of running was a problem and dangerous. Overall, 79.8% these ideas. of the sample believed red light running was a problem, and almost all (98.8%) believed it to be dangerous. Further, more than one in ten respondents (10.9%) had 4. Discussion been involved in a red light running crash. Among these variables, recent red light runners (73.7%) differed Limitations of the study should be considered before from others (81.2%) only in believing red light running discussing the results and offering conclusions. First, was less of a problem,  2(1) =4.88, P B0.05. men and younger drivers, who are most likely to be involved in many traffic crashes and dangerous driving 3.5. Ideas for changing red light running beha6ior behavior (Evans, 1991), were not represented as well as women and older drivers. Sex and age group were Finally, respondents were asked to suggest ideas for controlled in analyses predicting recent red light run- changing red light running behavior, particularly the ning, but the generalizability of these and other findings behavior of drivers who may not change easily. We did would have been increased with a more representative not ask directly what would change their behaviors sample. In addition, the reliance on a telephone survey because we felt respondents would be more open and as opposed to behavioral observations was limiting, honest to ideas if they did not fear those ideas would be because as is well known self-report data do not neces- used to change them. Respondents were allowed to sarily reflect an individual’s actual behaviors. However, offer more than one suggestion, but for the purposes of the survey methodology served its purpose as an effi- this article only the first ideas offered were considered. cient means to add more understanding to the red light The most common ideas involved legal initiatives. running literature. Combining police enforcement (14.2%), increased fines Given the above limitations, the data offered insights (13.1%), and photo enforcement strategies (11.5%), into the nature of red light running. Specifically, one of 38.8% suggested greater legal consequences for red light the more interesting results involved the role of frustra- running. Education was second, with 16.7% of the tion and different reckless driving behaviors. A driver’s 740 B.E. Porter, T.D. Berry / Accident Analysis and Pre6ention 33 (2001) 735–741 likelihood of running red lights may not depend as nalized intersections; Roundabouts, 2000), increased much upon frustration as expected (see Fig. 1). When police enforcement or installation of photo enforcement frustrated, drivers reported being more likely to speed, cameras seem best for applying immediate negative tailgate, weave, and gesture angrily at other drivers and consequences to violators. Cameras, in particular, are pedestrians than run red lights. If this finding is repli- effective because every violator is punished at any time cated, an interesting debate arises. Red light running of day. Police after all are often shifted to other loca- has been described as a deliberate aggressive act ‘that tions for revolving duties. More importantly, cameras allows the frustrated driver to move ahead at the cost have contributed to the largest reductions in red light of infringing on other road users’ rights’ (Shinar, 1998, running behavior that have been documented thus far p. 137). Red light running, then, may be uniquely (Retting et al., 1999a,b) different from aggressive driving as described by Shinar Respondents agreed that increased law enforcement (1998) or as conceptualized by a typical sample of was important for reducing red light running, although traffic-safety educators, engineers, and law enforcement only 11.5% offered cameras as their first choice. While officials (Porter and Berry, 1998). it was not surprising that drivers offered other options A second issue with frustration, related to Shinar’s besides enforcement cameras, it is important to note (1998) concern for drivers’ rights, was discourtesy. In that locations with active enforcement cameras for red our sample, discourtesy was chosen as the main reason light running have received significant public support for urban-driving frustration, which seemed inconsis- for their use (Retting et al., 1999a,b). Respondents also tent with drivers’ reported red light running. Most suggested giving the public more education about red researchers including Shinar (1998) would agree that light running. However, most respondents already be- red light running is a discourteous act. The survey did lieved red light running was a problem and dangerous, not include follow-up questions about discourtesy so and a significant number of them run red lights. The recent red light runners could not comment on their value of additional education to the public as a sole own contradiction. However, one hypothesis for the intervention will likely be ineffective in reducing red inconsistency is that discourtesy is not as important to light running. Exceptions may be educating new drivers drivers as is the perception that red light running has who have little experience driving at intersections or few negative consequences. implementing mass media programs to change red light Drivers have a reason to be cynical about red light running norms (see Evans, 1991 for a discussion on running consequences. Although red light running con- changing driving norms). tributes to 22% of urban crashes (Retting et al., 1995) The study of red light running continues to need and 3% of all fatal crashes (Retting et al., 1998), drivers theoretical and conceptual effort. It is a behavior that see the numerous violators each day who never get occurs in a very complex environment of the intersec- caught or cause a crash. Drivers may witness as many tion. Many factors play a role in drivers’ decisions to as ten light cycles per hour with at least one red light run red or stop at a light, complicating efforts to runner (Porter and England, 2000). For the most part, increase intersection safety. Even with the need for respondents believed red light running goes unpun- additional research, it seems clear from this study and ished. Only a minority had experienced a crash result- others that red light running interventions must in- ing from a red light runner. In addition, respondents crease consequences for violations. Whether it is with believed police would catch fewer than 20% of the red photo enforcement cameras or consistent police en- light runners (and this is probably optimistic). Interest- forcement, programs must find ways to teach drivers ingly, recent red light runners thought slightly more that red light running will not be accepted. Perhaps violators would be caught than other respondents. The such programs would also reduce the level of urban- ‘gambler’s fallacy’ (Stanovich, 1998) may explain this driving frustration. Without increasing negative conse- perception. Gambler’s fallacy is the mistaken belief that quences, there is no reason to expect red light running past and future events are correlated. In terms of red rates will decrease at most intersections in the near light running, drivers with an increased tendency of future. running red lights may feel more sensitive to the per- ceived odds of being caught (i.e. ‘my time is coming’). The general perception of getting caught for running Acknowledgements red lights will likely increase only with immediate and consistent negative consequences to violators. Behav- The study was funded by a grant from Daimler- ioral psychologists have long recognized the importance Chrysler Corporation to support the ‘Stop Red Light of immediate consequences for shaping behavior (see Running’ (SRLR) program. SRLR was also co-spon- Houston, 1991) and shaping traffic-safety behaviors is sored by the American Trauma Society and the Federal no exception. In the absence of structural changes to Highway Administration. Portions of this manuscript intersections (i.e. building roundabouts instead of sig- were released in a technical report written shortly after  B.E. Porter, T.D. Berry / Accident Analysis and Pre6ention 33 (2001) 735–741 741 the project’s completion. The conclusions are the re- Porter, B.E., Berry, T.D., 1999. A Nationwide Survey of Red Light sponsibility of the authors alone, and are not necessar- Running: Measuring Driver Behaviors for the ‘Stop Red Light Running Program’. Old Dominion University, Norfolk, VA. ily those of SRLR partners. However, SRLR personnel Porter, B.E., England, K.J., 2000. Predicting red light running behav- as well as personnel at Golin/Harris International, a ior: a traffic safety study in three urban settings. Journal of Safety public-relations firm contracted by DaimlerChrysler, Research 31, 1 – 8. provided valuable suggestions and assistance during all Retting, R.A., Williams, A.F., 1996. Characteristics of red light phases of the project. We gratefully acknowledge the violators: results of a field investigation. Journal of Safety Re- search 27, 9 – 15. assistance of Dr Jeffrey Harlow, Tancy Vandecar, and Retting, R.A., Williams, A.F., Preusser, D.F., Weinstein, H.B., 1995. their students at the Social Science Research Center Classifying urban crashes for countermeasure development. Acci- (SSRC) at Old Dominion University. The SSRC as- dent Analysis and Prevention 27, 283 – 294. sisted in survey development and testing, and com- Retting, R.A., Ulmer, R.G., Williams, A.F., 1998. Prevalence and pleted the telephone calls. Finally, we gratefully thank Characteristics of Red light Running Crashes in the United States. Insurance Institute for Highway Safety, Arlington, VA. peer reviewers and our students for helpful comments Retting, R.A., Williams, A.F., Farmer, C.M., Feldman, A., 1999a. on earlier versions of this manuscript. Evaluation of red light camera enforcement in Fairfax, VA, USA. ITE Journal 69 (8), 30 – 34. Retting, R.A., Williams, A.F., Farmer, C.M., Feldman, A., 1999b. Evaluation of red light camera enforcement in Oxnard, Califor- References nia. Accident Analysis and Prevention 31, 169 – 174. Roundabouts, 2000. Insurance Institute for Highway Safety: Status Deutsch, D., Sameth, S., Akinyemi, J., 1980. Seat belt usage and Report. May 13, 2000. vol. 35, pp. 1 – 3, 6. risk-taking behavior at two major traffic intersections. In: Pro- Shinar, D., 1998. Aggressive driving: the contribution of the drivers ceedings of the 24th Conference of the American Association for and the situation. Transportation Research Part F 1, 137 –160. Automotive Medicine, October 1980. Stanovich, K.E., 1998. How to Think Straight about Psychology, Evans, L., 1991. Traffic Safety and The Driver. Van Nostrand fifth ed. Longman, New York. Reinhold, New York. US Census Bureau, 2000a. Annual population estimates by age group Houston, J.P., 1991. Fundamentals of Learning and Memory, fourth and sex, selected years from 1990 to 2000 [On-line]. Available: ed. Harcourt Brace College Publishers, Fort Worth, TX. http://www.census.gov/population/www/estimates/nation2.html. Porter, B.E., Berry, T.D., 1998. An Action Report for Understanding US Census Bureau, 2000b. Place and county subdivision population and Reducing Aggressive Driving and Boating. DRIVE SMART estimates [On-line]. Available: http://www.census.gov/population/ Hampton Roads, Norfolk, VA. www/estimates/citypop.html.