- Email: [email protected]
The optimism bias and traffic accident risk perception
Accid. Anal. & Prev. Vol. 21. No. 4. pp. 333-340. Printed in Great Bntam.
THE OPTIMISM
Department
19X9 D 1989 Maxwell
BIAS AND TRAFFIC RISK PERCEPTION
0001-4575iR9 $3.(X1 + .lKI Pergamon Macmillan plc
ACCIDENT
DAVID M. DEJOY of Health Promotion and Behavior, University of Georgia, Athens, GA 30602, U.S.A.
(Received 21 June 1988; in revised form 8 February 1989) Abstract-Research suggests that people are excessively and unrealistically optimistic when judging their driving competency and accident risk. In this study, college-age drivers compared their risk of being involved in a variety of described traffic accidents relative to their peers. They also rated each of the accidents along a number of dimensions hypothesized as being related to optimism. In addition, subjects provided global estimates of their driving safety, skill, and accident likelihood. Significant optimism was evident for both the accidents and the global ratings. Optimism increased with driving experience and marginally with age. Those with more driving experience considered human factors to be more important in accident causation; those assigning more importance to human factors also rated themselves as more skillful drivers. For the specific accidents, perceived controllability was a strong predictor of optimism. The findings for controllability are interpreted in terms of other recent data and hypothesized explanations of the optimism bias. In general, it appears that optimism arises because people persistently overestimate the degree of control that they have over events.
INTRODUCTION
There is fairly general agreement today that self-protective behavior is essentially voluntary (e.g. Cohen, Smith, and Anger 1979; Green, Kreuter, Deeds, and Partridge 1980), and that whether an individual will take steps to protect himself or herself from a hazard depends to a considerable extent on how that person appraises the risks involved. Moreover, what appears to be important is the individual’s subjective estimate of risk rather than the objective facts associated with the hazard (Rosenstock 1974; Janz and Becker 1984). As such, considerable interest has been generated in understanding risk perception and in identifying systematic sources of bias in these processes. Research indicates that unwarranted optimism may be a particularly robust source of bias. Optimism refers to the tendency of people to be excessively and unrealistically optimistic and overconfident when judging the degree of personal risk associated with various events or situations. With respect to traffic accidents, evidence suggests that while most drivers possess fairly accurate perceptions of total, societal traffic risks (e.g. Lichtenstein, Slavic, Fischhoff, Layman, and Combs 1978), they tend to believe that these aggregate estimates of risk do not apply to them personally. Most drivers consider themselves to be safer, more skillful, and less likely to be involved in an accident than the average driver (Svenson 1981; Svenson, Fischhoff, and MacGregor 1985). Optimistic judgements of driving risk have also been obtained for specific driving situations and behaviors, using still photographs and videotapes (Finn and Bragg 1986; Matthews and Moran 1986). Optimistically biased judgements of personal risk are not limited to driving behavior; similar biases have also been demonstrated for a variety of health problems and other threats to well-being (Weinstein 1980, 1982, 1984). Although optimism appears to be a rather stable characteristic of risk perception, relatively little is known about the factors that produce or influence this particular bias. Indirect evidence from previous research suggests that optimism may be closely tied to the perceived controllability of the event or hazard in question. For example, Svenson et al. (1985) found that although judgements of driving safety, driving skill, and accident likelihood all showed an optimistic bias, the magnitude of the bias for accident likelihood was considerably less than that displayed for the skill and safety measures. If a high degree of control is a precondition for optimism, then it is reasonable to expect that optimism will be more evident for ratings of safety and skill than accident likelihood. 333
334
D. M. DEJOY
In general, people do have more control over their driving safety and skill than their ultimate accident involvement. Matthews and Moran showed that among young male drivers optimism was most pronounced in situations involving vehicle handling and driving reflexes, that is, situations involving a relatively high degree of driver control. Weinstein had his subjects rate the degree of controllability of various health threats and found that perceived controllability was a major predictor of optimism. Other explanations of the optimism phenomenon have also been offered (Weinstein 1982). One possibility is that optimism is an ego-defensive response directed at reducing the degree of threat associated with adverse events. This would suggest that optimism should increase as outcome severity increases. Another possibility is that optimism is a byproduct of egocentrism. People have difficulty in adopting the perspective of others and tend to lose sight of the fact that the same factors that make them feel relatively free of risk also apply in equal or greater measure to others. Accordingly, factors that influence our own perceptions of risk can be sources of error when we compare our own risk to that of others. Thus, it is reasonable to propose that lack of experience with a particular threat might increase optimism, and that people might also be particularly optimistic in response to highly improbable events. A third alternative is that optimism occurs when people have a stereotypic image of the kind of person who would succumb to the threat, particularly if they perceive themselves as not fitting this image. The major purpose of the present study was to extend the investigation of optimism to specific traffic accidents (as compared to driving situations or behaviors) and assess the relative contribution of various hypothesized factors in producing optimistically biased judgements of traffic accident risk. To accomplish this, subjects were presented with a series of traffic accidents and were asked to compare their risk of being involved with that of other members of their group. The advantage of using accident scenarios is that accidents can be readily configured that vary in terms of controllability, seriousness, and the other dimensions thought to be related to optimism. If optimism systematically increases or decreases as a function of one or more of these dimensions, then progress will have been made toward identifying the determinants of the optimism which characterizes our personal assessments of traffic accident risk. Combining Weinstein’s results with those from studies dealing specifically with traffic risks suggests that the controllability dimension might well be a principal determinant of optimism. Weinstein also found that lack of personal experience and the existence of vivid stereotypes contributed to optimism. As such, these factors may also be related to traffic accident risk perception. Previous research has provided little support for the importance of outcome severity. A secondary purpose of this study was to provide additional confirmatory data on the basic optimism phenomenon. In a fashion similar to Svenson, subjects provided global estimates of their accident likelihood and driving safety and skill. They also rated the overall importance of human, technical, and chance factors in accident causation. The inclusion of these measures provided an opportunity to compare the subjects’ comparative risk judgements for specific accidents with their responses to the global optimism and general accident causation questions. METHOD
Subjects
The subjects in this study were 106 college students (52 males; 54 females). Their ages ranged from 18 to 36, with the mean age being 21. All subjects were licensed drivers and the mean length of driving experience was 5.3 years. Materials
Subjects were asked to compare their risk of being involved in each of 10 different accidents relative to other students at the university (7-point scales; - 3 to + 3 with zero midpoint). The accidents ranged from the relatively mundane to the spectacular and differed in terms of outcome severity and controllability. The selected accidents were:
Optimism bias and risk perception
335
(1) scraping the side of your vehicle while driving up to the drive-in window at the bank, (2) being involved in a serious accident as a passenger in someone else’s vehicle, (3) being killed in an accident because you failed to yield the right-of-way, (4) being bumped from behind while waiting at a stoplight, (5) running over a piece of debris dropped off the truck traveling in front of you, (6) being killed in an accident caused by another motorist who was driving under a suspended license, (7) losing control of your vehicle at a high rate of speed and crashing into another vehicle, (8) backing into another vehicle while pulling out of a parking space, (9) having your vehicle struck by a speeding hitand-run driver, and (10) causing a serious accident while driving under the influence of alcohol. Subjects next rated each accident in terms of its probability, seriousness, controllability, imaginability, worry, and previous experience. In rating accident probability, subjects were provided with the approximate enrollment of the university (25,000) and were asked to estimate the number of students that would be likely to experience that particular accident during the next two or three years. Pilot testing showed this method to be effective in yielding probability estimates. Seriousness was rated along a 5-point scale (not at all serious to extremely serious). Subjects also rated the degree of controllability of the accidents along a 5-point scale (people cannot do anything to completely controllable). The ease of imagining the kind of person who would be involved in the accidents was assessed along a 3-point scale (no image, unclear image, or clear image). Judgements about worry were obtained using a 4-point scale (not at all to quite worried). Experience judgments were made using either a 3-point or a 5-point scale, depending on whether the outcome was fatal or nonfatal (no previous experience, has happened to acquaintance, has happened to relative or good friend, has happened to me, or has happened to me more than once). The questions and response formats used for the six accident attributes were adopted from Weinstein (1982). To reduce the time demands on the subjects, half of the subjects rated one set of five accidents along the six dimensions, while the other half of the subjects rated the other five accidents. The two sets of accidents were equated across the controllability and severity dimensions. The order of the six attribute ratings was the same for each accident. Subjects were also asked to make global estimates of their accident likelihood and levels of driving safety and skill relative to the average driver (13-point scales; -6 to +6 with zero midpoint). The global optimism questions were similar to those used by Svenson et al. (1985). Subjects also rated the importance of human (errors made by people), technical (problems with vehicle or road design), and chance (bad luck or fate) factors in accident causation (13-point scales; 1 not important to 13 very important). Finally, data were collected on each subject’s length of driving experience, miles driven in a typical year, age, sex, and license status. Driving experience was assessed using a 7-point scale (no experience to more than seven years). A six-point scale was used to assess miles driven per year (less than 3,000 to over 15,000). Subjects indicated their age in years, their sex, and whether they were a licensed driver. Procedure
Subjects met in groups and completed the forms individually. Instructions were printed on the materials; the experimenter read these and provided an opportunity for questions. RESULTS
Accident situations
The existence of optimistically biased estimates of risk should be reflected by comparative risk judgments of less than zero. As can be seen in Table 1, negative means were obtained for nine of the accidents. Student’s t (one sample test) was used to test whether the mean for each accident was significantly different from zero. The results indicated that subjects considered their chances of being involved as significantly less AAP21:4-c
D. M. DEJOY
336
Table 1. Comparative risk judgements for the 10 accidents Mean risk judgement
Accident situation
-1.85 - 1.58 - 1.57 - 1.24 --0.75 -0.45 -0.27 -0.22 -0.03 0.04
Causing serious accident while intoxicated Killed because you failed to yield right-of-way Scraping side of vehicle at drive-in window Losing control of your vehicle at high speed and hitting another vehicle Bumping another vehicle while pulling from parking space Serious accident as passenger Killed by another motorist who was driving under suspended license Having your vehicle struck by speeding hit-and-run driver Running over piece of debris dropped from truck Bumped from behind at stoplight
(1.51) (1.21) (1.38) (1.38) (1.45) (1.42) (.98) (.94) (1.22) (1.00)
t - 12.60*** - 13.42*** - 11.68*** -9.24**” -5.30*** -3.26** -2.78”* - 2.39* - 0.24 0.39
Note: A comparative risk judgement of less than zero indicates a belief that one’s risk is less than that of other drivers. Standard deviations are in parentheses. *p < .05. **p < .Ol. ***p < ,001.
than others in their group for eight of the accidents. The two accidents that failed to yield optimistic ratings were among those with the most innocuous outcomes (being bumped from behind at a stoplight and running over a piece of road debris). The relationships among optimism and the various accident dimensions were analyzed using accident situation as the unit of analysis (n = 10). The measure of optimism for each accident was the mean comparative risk judgement from Table 1. The measures of the various accident dimensions were the group mean ratings. The correlation matrix for these measures is presented in Table 2. As Table 2 shows, optimism was strongly correlated with controllability and moderately correlated with imaginability. The correlation with worry approached significance (p < .06). Subjects were much more optimistic when they perceived the event as controllable, and optimism was associated with greater ability to mentally picture the type of person who would be involved and with less worry. Optimism was not affected by event seriousness. Not surprisingly, less serious accidents were rated as more probable, and subjects were more likely to have had experience with such events. Controllable accidents engendered less worry and possessed greater imaginability. Multiple regression was used to determine the contribution of the various accident dimensions to the prediction of optimism. Stepwise analysis using the six attributes as predictors showed controllability (F(1,8) = 75.33, p < .OOl) to be the primary predictor of optimism (R2 = .90). No other factors entered the equation. Global ratings
Substantial optimism was also evident for the global measures of driving safety, driving skill, and accident likelihood (see Table 3). Student’s t was again used to test whether each of these means was significantly different from zero. Consistent with previous research, subjects rated themselves as significantly safer, more skillful, and less Table 2. Accident attribute correlation matrix
Optimism Probability Worry Controllability Imaginability Seriousness Experience
*p < .05. **p i .Ol. ***p < ,001.
Optimism
Probability
Worry
.15 .61 - .95*** - .64” - .30 .45
- .30 .Ol -.ll - .71* .76**
- .63* -.49 .52 .17
Controllability
.76** .19 - .42
Imaginability
Seriousness
.20 - .60
-.49
Experience
337
Optimism bias and risk perception Table 3. Global perceptions of driving competency and accident risk
Measure
Mean rating
I
Percent optimistic
Driving safety Driving skill Accident likelihood
2.03 (2.34) 3.01 (1.77) - 1.38 (2.18)
8.94* 17.56* - 6.49*
75.5 88.7 54.3
Note: Percent optimistic refers to percentage of subjects who judged their driving safety and skill to be greater than that of other drivers and their accident likelihood to be less. Standard deviations are in parentheses. *p < ,001.
likely to be involved in an accident than the average driver. All three comparisons were significant at the .OOl level. Also, as reported previously (Svenson et al. 1985), optimism was less pronounced for ratings of accident likelihood than for ratings of driving safety and skill. None of the three global optimism measures were correlated with the age or sex of the subject. The only such correlation that even approached significance was between age and safety (r = .16, p < .09). Older subjects rated themselves as somewhat safer drivers than did younger subjects. The absence of more pronounced effects for age may be due, in part, to the restricted age range of the study sample. The ratings of safety and skill were correlated with years of driving experience, with optimism increasing as driving experience increased ( YS= .27 and .26, respectively; ps < .Ol). Perceived accident likelihood showed a similar relationship with experience, but fell short of significance (r = .15, p < .12). The self-ratings of driving skill were also correlated with miles driven (Y = .28, p < .Ol); subjects who drove more miles in a typical year considered themselves to be more skillful. Miles driven was also correlated with sex (r = - .24, p < .Ol), with males driving more miles in a typical year then females. In general, subjects attached considerably greater importance to human factors (M = 11.13) in causing accidents than to either technical (M = 7.79) or chance (M = 5.26) factors. Subjects who assigned greater importance to human factors in accident causation also rated themselves as more skillful (r = .22, p < .Ol). The correlation between human factors and driving safety was not significant. The importance assigned to human factors also increased with driving experience (r = .26, p < .Ol). Finally, females assigned greater importance to technical factors than did males (r = .22, p < .OS). Other analyses
Correlational analyses were also performed to assess the relationship of the global optimism ratings to the comparative risk judgements for the various accident situations. A mean comparative risk score was computed for each subject by averaging the subject’s comparative risk estimates for the 10 accidents. These mean risk scores were correlated with the individual ratings of driving safety, driving skill, and accident likelihood. As might be expected, those who displayed greater optimism with respect to the specific accidents also considered themselves safer (Y = - .45), more skillful (r = - .37), and less likely to have an accident (r = .45). All three correlations were significant at the .OOl level. Correlations were also computed between the mean risk scores and the importance assigned to human, technical, and chance factors in accident causation. The only effect to reach significance involved the rated importance of technical factors. Subjects who were more optimistic tended to assign less importance to technical factors as accident causes (r = .19,p < .05). Correlations approaching significance were obtained between mean risk and age (r = - .18, p < .07), and between mean risk and driving experience (r = - .18, p < .06). Optimism increased somewhat with increased age and driving experience. Nonsignificant correlations were obtained for sex and miles driven per year.
338
D. M. DEJOY
Finally, a stepwise multiple regression was performed using the three global measures of optimism, the three accident causation measures, age, sex, driving experience, and miles driven to predict mean comparative risk. Driving safety (F(1,103) = 11.08, p < .OOl), accident likelihood (F(1,103) = 8.13, p < .Ol), and the importance assigned to technical factors (F(1,103) = 5.69, p < .05),all made significant contributions to the prediction of mean risk (R? = .31). DISCUSSION
The preceding results show that the college student subjects in this study were quite optimistic in evaluating their risk of being involved in a wide variety of accidents, and were particularly optimistic for accidents that they perceived as being controllable. The present findings along with those of Finn and Bragg (1986) and Matthews and Moran (1986) indicate that optimism transcends global perceptions of driving competency and accident likelihood. People tend to be optimistic not only when offering abstract evaluations of their own driving safety and skill, but also when judging the risk associated with a variety of specific driving behaviors, situations, and accidents. The observed importance of controllability in the present study is consistent with Weinstein’s (1980, 1982, 1984) findings with various health threats and life events, and suggests that optimism may arise because people persistently overestimate the degree of control that they have over events. Matthews and Moran (1986) reported that young drivers were most optimistic for driving situations involving vehicle-handling and driving reflexes. They concluded: “that when the driver can view himself as being in control of the situation the rated risk is low; when he sees himself as a statistic, however, the rated risk is higher, possibly as a result of factors not under direct personal control, such as the driving behavior of others” (p. 310). The present findings directly support this analysis in that subjects were optimistic for those accidents that they had judged to be most controllable. Moreover, as shown in Table 1, the accidents associated with the greatest optimism were those involving either driver negligence or poor vehicle-handling. Optimism, while still present, was much less pronounced for situations in which the motorist was seemingly “victimized” by external circumstances beyond his or her control. Optimism does not appear to be an ego-defensive response, in view of the lack of association between optimism and the seriousness of the accident. Weinstein has also consistently failed to find a significant relationship between outcome severity and optimism (see Weinstein 1987 for summary). People do not appear to be optimistic as a way of coping with threatening events. If they did, optimism should be most obvious in response to accidents that are both serious and beyond individual control. This was clearly not the case (see Table 1). Similarly, no support was found for the egocentrism hypothesis; optimism was not related to lack of personal experience or low probability of occurrence. Optimism was associated, however, with the ability to mentally picture the type of person who would be involved in the accident. Possessing a vivid stereotypic image may be associated with unrealistic optimism. Although correlated with optimism, imaginability did not make an independent contribution to predicting optimism in the regression analysis. Considerable optimism was also present on the three global measures. Subjects rated themselves as safer, more skillful, and less likely to have a traffic accident than the average motorist. Consistent with Svenson et al. (1985), optimism was less pronounced for accident likelihood than the other two measures, once again suggesting that individuals are less optimistic when judging outcomes over which they have limited direct control. Global optimism was found to increase as driving experience increased. Experience had a similar effect on the comparative risk judgements for the 10 accidents (mean risk), but this correlation was only marginally significant. Spolander (1982) also reported increased optimism as a function of experience. Spolander compared drivers who had had licenses for one month, one year, or three years. New drivers perceived themselves as poorer than average in most respects, but after only one year considered themselves as better than average on many technical driving skills. At three years, most
Optimism bias and risk perception
339
drivers considered themselves to be better than average on a variety of both technical and defensive driving skills. Subjects, in the present study, also assigned greater importance to human factors in accident causation as their driving experience increased. Moreover, those who held human factors to be important also considered themselves to be more skillful drivers. If optimism is related to an exaggerated sense of control, then as optimism develops with driving experience the importance assigned to human factors should also increase. It is also reasonable to expect that those who assign greater importance to human factors will also claim that they take advantage of that control. Optimism also increased somewhat with age. Marginally significant correlations were obtained for both global driving safety and mean risk. These results must be interpreted with considerable caution, both in view of their magnitude and restricted age range of the sample. McCormick and colleagues (McCormick, Walkey, and Green 1986), using a somewhat broader age range sample, failed to detect effects for age. However, other data exist suggesting that optimism, although evident in older subjects, diminishes with age (Svenson et al. 1985). Unfortunately, in Svenson’s study age was confounded with cultural differences in the study sample. Matthews and Moran (1986) also reported that older drivers were less likely than younger drivers to set themselves above their peers in judging their driving abilities. With respect to the perceived risk associated with various driving situations, Finn and Bragg (1986) found that younger drivers tended to perceive less risk in various driving situations than did older drivers. One possible explanation for the inconsistent findings in the literature is that optimism increases during the first several years postlicensure, but decreases thereafter. Assuming such a curvilinear relationship, increased optimism as a function of driving experience and/or age would be expected for a young subject sample (as in the present study), while opposite effects would be expected for older subjects. This reasoning is consistent with Brown’s (1982) analysis of the differential acquisition of vehicle control skills and road craft over time. According to Brown, because young drivers acquire perceptual-motor skills quite easily, they quickly gain a high level of confidence in their skills. With continued experience, however, drivers acquire a greater appreciation of the complexities of the driving task, and a better match exists between their level of confidence and their decision skills. Another interesting finding from the present study is that male and female subjects did not differ substantially either in terms of their global judgements of driving safety, skill, and accident likelihood, or their comparative risk judgements for the 10 accidents. Svenson et al. (1985) and McCormick et al. (1986) also failed to find consistent effects for sex. Other recent studies have tended to use only male subjects (e.g. Finn and Bragg 1986; Matthews and Moran 1986). The well-documented excess involvement of young male drivers in traffic accidents (e.g. National Accident Sampling System 1981; Williams and Karpf 1984), combined with the absence of differences in optimism between males and females in recent research, suggests that the excess involvement of males may not be directly related to sex differences in risk perception per se, but rather to how these perceptions are translated into driving behavior and risk-taking. In practical terms, the existence of such robust optimism with respect to traffic accident risks has important implications. Referring to Fuller’s (1984) threat avoidance model of driving behavior, excessive optimism should decrease the probability of appropriate anticipatory avoidance responses. That optimism appears to be strongly related to an exaggerated sense of control suggests that efforts should be directed at reducing the driver’s sense of perceived control over the driving environment. Driver education programs are needed that provide direct demonstration of the limits of one’s driving skill in accident avoidance. Young drivers are especially in need of objective data demonstrating the mismatch that exists between their perceived and actual levels of driving skill (see Brown 1987). Special attention should be given to providing drivers with more realistic appraisals of perceived vulnerability. Research indicates that people possess fairly veridical perceptions of overall traffic risks (Lichtenstein et al. 1978); younger drivers may even appreciate that their age group is at greater risk than are older drivers (Finn and Bragg
340
D. M. DEJOY
1986). Indeed, in some instances, young drivers may actually consider driving to be a more dangerous activity than do older drivers (Berger and Persinger 1980). However, the problem is that most drivers, and especially young drivers, do not believe that these risks apply to them personally. As such, interventions must seek to counter the perception of exaggerated control by means that personalize the risk. Standard educational efforts, warnings, and statistics are not likely to be effective. REFERENCES Berger, R. J.; Persinger, G. S. 1980 survey of public perception of highway safety. McLean, VA: Automated Services Inc. ; 1980. Brown, I. D. Exposure and experience are a confounded nuisance in research on driver behavior. Accid. Anal. Prev. 14:345-352; 1982. Brown, I. D. Predisposing factors in the alcohol- and drug-impairment of young drivers’ performance. In: Beniamin, T., editor. Young drivers impaired bv alcohol and other drugs. London: Roval Societv of Medicine Series; 1987:165-172. Cohen, A.; Smith, M. J.; Anger. W. K. Self-protective measures against workplace hazards. J. Safety Res. 11:121-131; 1979. Finn, B.; Bragg, B. W. E. Perception of the risk of an accident by younger and older drivers. Accid. Anal. Prev. 18:289-298; 1986. Fuller, R. A conceptualization of driver behavior as threat avoidance. Ergonomics. 27:1139-1155; 1984. Green, L. W.; Kreuter, M. W.; Deeds, S. G.; Partridge, K. B. Health education planning: A diagnostic approach. Palo Alto, CA: Mayfield; 1980. Janz, N. K.; Becker, M. H. The health belief model: A decade later. Health Educ. Q. ll:l-47; 1984. Lichtenstein, S.; Slavic, P.; Fischhoff, B.; Layman, M.; Combs. Judged frequency of lethal events. J. Exper. Psychol. Human Learn. Memory 4:551-578; 1978. Matthews, M. L.; Moran, A. R. Age differences in male drivers’ perception of accident risk: The role of perceived driving ability. Accid. Anal. Prev. 18:299-313; 1986. McCormick, I. A.; Walkey, F. H.; Green, D. E. Comparative perceptions of driver ability-a confirmation and expansion. Accid. Anal. Prev. 18:205-208; 1986. National Accident Sampling System, Abt Associates, Inc. Analysis of 1979 computerized data. Abt; 1981. Rosenstock, I. M. The health belief model and preventive health behavior. Health Educ. Mono. 2:354-386; 1974. Spolander. K. Oerfarna forares beteende, presentations-formage och attityder (Inexperienced driver’s behavior, ability, and attitudes). Linkoping: Swedish National Road Traffic Research Institute; 1982. Svenson. 0. Are we all less risky and more skillful than our fellow drivers? Acta Psychol. 47:143-148; 1981. Svenson, 0.; Fischhoff, B.; MacGregor. D. Perceived driving safety and seatbelt usage. Accid. Anal. Prev. 17:119-133; 1985. Weinstein, N. D. Unrealistic optimism about future life events. J. Personal. Sot. Psychol. 39:806-820; 1980. Weinstein, N. D. Unrealistic optimism about susceptibility to health problems. J. Behav. Med. 5:441-460; 1982. Weinstein, N. D. Why it won’t happen to me: Perceptions of risk factors and illness susceptibility. Health Psychol. 3:431-457; 1984. Weinstein. N. D. Unrealistic optimism about susceptibility to health problems: Conclusions from a communitywide sample. J. Behav. Med. 10:481-500; 1987. Williams. A. F.; Karpf, R. S. Teenaged drivers and fatal crash responsibility. Law Policy Q. 6:101-113; 1984.
THE OPTIMISM
Department
19X9 D 1989 Maxwell
BIAS AND TRAFFIC RISK PERCEPTION
0001-4575iR9 $3.(X1 + .lKI Pergamon Macmillan plc
ACCIDENT
DAVID M. DEJOY of Health Promotion and Behavior, University of Georgia, Athens, GA 30602, U.S.A.
(Received 21 June 1988; in revised form 8 February 1989) Abstract-Research suggests that people are excessively and unrealistically optimistic when judging their driving competency and accident risk. In this study, college-age drivers compared their risk of being involved in a variety of described traffic accidents relative to their peers. They also rated each of the accidents along a number of dimensions hypothesized as being related to optimism. In addition, subjects provided global estimates of their driving safety, skill, and accident likelihood. Significant optimism was evident for both the accidents and the global ratings. Optimism increased with driving experience and marginally with age. Those with more driving experience considered human factors to be more important in accident causation; those assigning more importance to human factors also rated themselves as more skillful drivers. For the specific accidents, perceived controllability was a strong predictor of optimism. The findings for controllability are interpreted in terms of other recent data and hypothesized explanations of the optimism bias. In general, it appears that optimism arises because people persistently overestimate the degree of control that they have over events.
INTRODUCTION
There is fairly general agreement today that self-protective behavior is essentially voluntary (e.g. Cohen, Smith, and Anger 1979; Green, Kreuter, Deeds, and Partridge 1980), and that whether an individual will take steps to protect himself or herself from a hazard depends to a considerable extent on how that person appraises the risks involved. Moreover, what appears to be important is the individual’s subjective estimate of risk rather than the objective facts associated with the hazard (Rosenstock 1974; Janz and Becker 1984). As such, considerable interest has been generated in understanding risk perception and in identifying systematic sources of bias in these processes. Research indicates that unwarranted optimism may be a particularly robust source of bias. Optimism refers to the tendency of people to be excessively and unrealistically optimistic and overconfident when judging the degree of personal risk associated with various events or situations. With respect to traffic accidents, evidence suggests that while most drivers possess fairly accurate perceptions of total, societal traffic risks (e.g. Lichtenstein, Slavic, Fischhoff, Layman, and Combs 1978), they tend to believe that these aggregate estimates of risk do not apply to them personally. Most drivers consider themselves to be safer, more skillful, and less likely to be involved in an accident than the average driver (Svenson 1981; Svenson, Fischhoff, and MacGregor 1985). Optimistic judgements of driving risk have also been obtained for specific driving situations and behaviors, using still photographs and videotapes (Finn and Bragg 1986; Matthews and Moran 1986). Optimistically biased judgements of personal risk are not limited to driving behavior; similar biases have also been demonstrated for a variety of health problems and other threats to well-being (Weinstein 1980, 1982, 1984). Although optimism appears to be a rather stable characteristic of risk perception, relatively little is known about the factors that produce or influence this particular bias. Indirect evidence from previous research suggests that optimism may be closely tied to the perceived controllability of the event or hazard in question. For example, Svenson et al. (1985) found that although judgements of driving safety, driving skill, and accident likelihood all showed an optimistic bias, the magnitude of the bias for accident likelihood was considerably less than that displayed for the skill and safety measures. If a high degree of control is a precondition for optimism, then it is reasonable to expect that optimism will be more evident for ratings of safety and skill than accident likelihood. 333
334
D. M. DEJOY
In general, people do have more control over their driving safety and skill than their ultimate accident involvement. Matthews and Moran showed that among young male drivers optimism was most pronounced in situations involving vehicle handling and driving reflexes, that is, situations involving a relatively high degree of driver control. Weinstein had his subjects rate the degree of controllability of various health threats and found that perceived controllability was a major predictor of optimism. Other explanations of the optimism phenomenon have also been offered (Weinstein 1982). One possibility is that optimism is an ego-defensive response directed at reducing the degree of threat associated with adverse events. This would suggest that optimism should increase as outcome severity increases. Another possibility is that optimism is a byproduct of egocentrism. People have difficulty in adopting the perspective of others and tend to lose sight of the fact that the same factors that make them feel relatively free of risk also apply in equal or greater measure to others. Accordingly, factors that influence our own perceptions of risk can be sources of error when we compare our own risk to that of others. Thus, it is reasonable to propose that lack of experience with a particular threat might increase optimism, and that people might also be particularly optimistic in response to highly improbable events. A third alternative is that optimism occurs when people have a stereotypic image of the kind of person who would succumb to the threat, particularly if they perceive themselves as not fitting this image. The major purpose of the present study was to extend the investigation of optimism to specific traffic accidents (as compared to driving situations or behaviors) and assess the relative contribution of various hypothesized factors in producing optimistically biased judgements of traffic accident risk. To accomplish this, subjects were presented with a series of traffic accidents and were asked to compare their risk of being involved with that of other members of their group. The advantage of using accident scenarios is that accidents can be readily configured that vary in terms of controllability, seriousness, and the other dimensions thought to be related to optimism. If optimism systematically increases or decreases as a function of one or more of these dimensions, then progress will have been made toward identifying the determinants of the optimism which characterizes our personal assessments of traffic accident risk. Combining Weinstein’s results with those from studies dealing specifically with traffic risks suggests that the controllability dimension might well be a principal determinant of optimism. Weinstein also found that lack of personal experience and the existence of vivid stereotypes contributed to optimism. As such, these factors may also be related to traffic accident risk perception. Previous research has provided little support for the importance of outcome severity. A secondary purpose of this study was to provide additional confirmatory data on the basic optimism phenomenon. In a fashion similar to Svenson, subjects provided global estimates of their accident likelihood and driving safety and skill. They also rated the overall importance of human, technical, and chance factors in accident causation. The inclusion of these measures provided an opportunity to compare the subjects’ comparative risk judgements for specific accidents with their responses to the global optimism and general accident causation questions. METHOD
Subjects
The subjects in this study were 106 college students (52 males; 54 females). Their ages ranged from 18 to 36, with the mean age being 21. All subjects were licensed drivers and the mean length of driving experience was 5.3 years. Materials
Subjects were asked to compare their risk of being involved in each of 10 different accidents relative to other students at the university (7-point scales; - 3 to + 3 with zero midpoint). The accidents ranged from the relatively mundane to the spectacular and differed in terms of outcome severity and controllability. The selected accidents were:
Optimism bias and risk perception
335
(1) scraping the side of your vehicle while driving up to the drive-in window at the bank, (2) being involved in a serious accident as a passenger in someone else’s vehicle, (3) being killed in an accident because you failed to yield the right-of-way, (4) being bumped from behind while waiting at a stoplight, (5) running over a piece of debris dropped off the truck traveling in front of you, (6) being killed in an accident caused by another motorist who was driving under a suspended license, (7) losing control of your vehicle at a high rate of speed and crashing into another vehicle, (8) backing into another vehicle while pulling out of a parking space, (9) having your vehicle struck by a speeding hitand-run driver, and (10) causing a serious accident while driving under the influence of alcohol. Subjects next rated each accident in terms of its probability, seriousness, controllability, imaginability, worry, and previous experience. In rating accident probability, subjects were provided with the approximate enrollment of the university (25,000) and were asked to estimate the number of students that would be likely to experience that particular accident during the next two or three years. Pilot testing showed this method to be effective in yielding probability estimates. Seriousness was rated along a 5-point scale (not at all serious to extremely serious). Subjects also rated the degree of controllability of the accidents along a 5-point scale (people cannot do anything to completely controllable). The ease of imagining the kind of person who would be involved in the accidents was assessed along a 3-point scale (no image, unclear image, or clear image). Judgements about worry were obtained using a 4-point scale (not at all to quite worried). Experience judgments were made using either a 3-point or a 5-point scale, depending on whether the outcome was fatal or nonfatal (no previous experience, has happened to acquaintance, has happened to relative or good friend, has happened to me, or has happened to me more than once). The questions and response formats used for the six accident attributes were adopted from Weinstein (1982). To reduce the time demands on the subjects, half of the subjects rated one set of five accidents along the six dimensions, while the other half of the subjects rated the other five accidents. The two sets of accidents were equated across the controllability and severity dimensions. The order of the six attribute ratings was the same for each accident. Subjects were also asked to make global estimates of their accident likelihood and levels of driving safety and skill relative to the average driver (13-point scales; -6 to +6 with zero midpoint). The global optimism questions were similar to those used by Svenson et al. (1985). Subjects also rated the importance of human (errors made by people), technical (problems with vehicle or road design), and chance (bad luck or fate) factors in accident causation (13-point scales; 1 not important to 13 very important). Finally, data were collected on each subject’s length of driving experience, miles driven in a typical year, age, sex, and license status. Driving experience was assessed using a 7-point scale (no experience to more than seven years). A six-point scale was used to assess miles driven per year (less than 3,000 to over 15,000). Subjects indicated their age in years, their sex, and whether they were a licensed driver. Procedure
Subjects met in groups and completed the forms individually. Instructions were printed on the materials; the experimenter read these and provided an opportunity for questions. RESULTS
Accident situations
The existence of optimistically biased estimates of risk should be reflected by comparative risk judgments of less than zero. As can be seen in Table 1, negative means were obtained for nine of the accidents. Student’s t (one sample test) was used to test whether the mean for each accident was significantly different from zero. The results indicated that subjects considered their chances of being involved as significantly less AAP21:4-c
D. M. DEJOY
336
Table 1. Comparative risk judgements for the 10 accidents Mean risk judgement
Accident situation
-1.85 - 1.58 - 1.57 - 1.24 --0.75 -0.45 -0.27 -0.22 -0.03 0.04
Causing serious accident while intoxicated Killed because you failed to yield right-of-way Scraping side of vehicle at drive-in window Losing control of your vehicle at high speed and hitting another vehicle Bumping another vehicle while pulling from parking space Serious accident as passenger Killed by another motorist who was driving under suspended license Having your vehicle struck by speeding hit-and-run driver Running over piece of debris dropped from truck Bumped from behind at stoplight
(1.51) (1.21) (1.38) (1.38) (1.45) (1.42) (.98) (.94) (1.22) (1.00)
t - 12.60*** - 13.42*** - 11.68*** -9.24**” -5.30*** -3.26** -2.78”* - 2.39* - 0.24 0.39
Note: A comparative risk judgement of less than zero indicates a belief that one’s risk is less than that of other drivers. Standard deviations are in parentheses. *p < .05. **p < .Ol. ***p < ,001.
than others in their group for eight of the accidents. The two accidents that failed to yield optimistic ratings were among those with the most innocuous outcomes (being bumped from behind at a stoplight and running over a piece of road debris). The relationships among optimism and the various accident dimensions were analyzed using accident situation as the unit of analysis (n = 10). The measure of optimism for each accident was the mean comparative risk judgement from Table 1. The measures of the various accident dimensions were the group mean ratings. The correlation matrix for these measures is presented in Table 2. As Table 2 shows, optimism was strongly correlated with controllability and moderately correlated with imaginability. The correlation with worry approached significance (p < .06). Subjects were much more optimistic when they perceived the event as controllable, and optimism was associated with greater ability to mentally picture the type of person who would be involved and with less worry. Optimism was not affected by event seriousness. Not surprisingly, less serious accidents were rated as more probable, and subjects were more likely to have had experience with such events. Controllable accidents engendered less worry and possessed greater imaginability. Multiple regression was used to determine the contribution of the various accident dimensions to the prediction of optimism. Stepwise analysis using the six attributes as predictors showed controllability (F(1,8) = 75.33, p < .OOl) to be the primary predictor of optimism (R2 = .90). No other factors entered the equation. Global ratings
Substantial optimism was also evident for the global measures of driving safety, driving skill, and accident likelihood (see Table 3). Student’s t was again used to test whether each of these means was significantly different from zero. Consistent with previous research, subjects rated themselves as significantly safer, more skillful, and less Table 2. Accident attribute correlation matrix
Optimism Probability Worry Controllability Imaginability Seriousness Experience
*p < .05. **p i .Ol. ***p < ,001.
Optimism
Probability
Worry
.15 .61 - .95*** - .64” - .30 .45
- .30 .Ol -.ll - .71* .76**
- .63* -.49 .52 .17
Controllability
.76** .19 - .42
Imaginability
Seriousness
.20 - .60
-.49
Experience
337
Optimism bias and risk perception Table 3. Global perceptions of driving competency and accident risk
Measure
Mean rating
I
Percent optimistic
Driving safety Driving skill Accident likelihood
2.03 (2.34) 3.01 (1.77) - 1.38 (2.18)
8.94* 17.56* - 6.49*
75.5 88.7 54.3
Note: Percent optimistic refers to percentage of subjects who judged their driving safety and skill to be greater than that of other drivers and their accident likelihood to be less. Standard deviations are in parentheses. *p < ,001.
likely to be involved in an accident than the average driver. All three comparisons were significant at the .OOl level. Also, as reported previously (Svenson et al. 1985), optimism was less pronounced for ratings of accident likelihood than for ratings of driving safety and skill. None of the three global optimism measures were correlated with the age or sex of the subject. The only such correlation that even approached significance was between age and safety (r = .16, p < .09). Older subjects rated themselves as somewhat safer drivers than did younger subjects. The absence of more pronounced effects for age may be due, in part, to the restricted age range of the study sample. The ratings of safety and skill were correlated with years of driving experience, with optimism increasing as driving experience increased ( YS= .27 and .26, respectively; ps < .Ol). Perceived accident likelihood showed a similar relationship with experience, but fell short of significance (r = .15, p < .12). The self-ratings of driving skill were also correlated with miles driven (Y = .28, p < .Ol); subjects who drove more miles in a typical year considered themselves to be more skillful. Miles driven was also correlated with sex (r = - .24, p < .Ol), with males driving more miles in a typical year then females. In general, subjects attached considerably greater importance to human factors (M = 11.13) in causing accidents than to either technical (M = 7.79) or chance (M = 5.26) factors. Subjects who assigned greater importance to human factors in accident causation also rated themselves as more skillful (r = .22, p < .Ol). The correlation between human factors and driving safety was not significant. The importance assigned to human factors also increased with driving experience (r = .26, p < .Ol). Finally, females assigned greater importance to technical factors than did males (r = .22, p < .OS). Other analyses
Correlational analyses were also performed to assess the relationship of the global optimism ratings to the comparative risk judgements for the various accident situations. A mean comparative risk score was computed for each subject by averaging the subject’s comparative risk estimates for the 10 accidents. These mean risk scores were correlated with the individual ratings of driving safety, driving skill, and accident likelihood. As might be expected, those who displayed greater optimism with respect to the specific accidents also considered themselves safer (Y = - .45), more skillful (r = - .37), and less likely to have an accident (r = .45). All three correlations were significant at the .OOl level. Correlations were also computed between the mean risk scores and the importance assigned to human, technical, and chance factors in accident causation. The only effect to reach significance involved the rated importance of technical factors. Subjects who were more optimistic tended to assign less importance to technical factors as accident causes (r = .19,p < .05). Correlations approaching significance were obtained between mean risk and age (r = - .18, p < .07), and between mean risk and driving experience (r = - .18, p < .06). Optimism increased somewhat with increased age and driving experience. Nonsignificant correlations were obtained for sex and miles driven per year.
338
D. M. DEJOY
Finally, a stepwise multiple regression was performed using the three global measures of optimism, the three accident causation measures, age, sex, driving experience, and miles driven to predict mean comparative risk. Driving safety (F(1,103) = 11.08, p < .OOl), accident likelihood (F(1,103) = 8.13, p < .Ol), and the importance assigned to technical factors (F(1,103) = 5.69, p < .05),all made significant contributions to the prediction of mean risk (R? = .31). DISCUSSION
The preceding results show that the college student subjects in this study were quite optimistic in evaluating their risk of being involved in a wide variety of accidents, and were particularly optimistic for accidents that they perceived as being controllable. The present findings along with those of Finn and Bragg (1986) and Matthews and Moran (1986) indicate that optimism transcends global perceptions of driving competency and accident likelihood. People tend to be optimistic not only when offering abstract evaluations of their own driving safety and skill, but also when judging the risk associated with a variety of specific driving behaviors, situations, and accidents. The observed importance of controllability in the present study is consistent with Weinstein’s (1980, 1982, 1984) findings with various health threats and life events, and suggests that optimism may arise because people persistently overestimate the degree of control that they have over events. Matthews and Moran (1986) reported that young drivers were most optimistic for driving situations involving vehicle-handling and driving reflexes. They concluded: “that when the driver can view himself as being in control of the situation the rated risk is low; when he sees himself as a statistic, however, the rated risk is higher, possibly as a result of factors not under direct personal control, such as the driving behavior of others” (p. 310). The present findings directly support this analysis in that subjects were optimistic for those accidents that they had judged to be most controllable. Moreover, as shown in Table 1, the accidents associated with the greatest optimism were those involving either driver negligence or poor vehicle-handling. Optimism, while still present, was much less pronounced for situations in which the motorist was seemingly “victimized” by external circumstances beyond his or her control. Optimism does not appear to be an ego-defensive response, in view of the lack of association between optimism and the seriousness of the accident. Weinstein has also consistently failed to find a significant relationship between outcome severity and optimism (see Weinstein 1987 for summary). People do not appear to be optimistic as a way of coping with threatening events. If they did, optimism should be most obvious in response to accidents that are both serious and beyond individual control. This was clearly not the case (see Table 1). Similarly, no support was found for the egocentrism hypothesis; optimism was not related to lack of personal experience or low probability of occurrence. Optimism was associated, however, with the ability to mentally picture the type of person who would be involved in the accident. Possessing a vivid stereotypic image may be associated with unrealistic optimism. Although correlated with optimism, imaginability did not make an independent contribution to predicting optimism in the regression analysis. Considerable optimism was also present on the three global measures. Subjects rated themselves as safer, more skillful, and less likely to have a traffic accident than the average motorist. Consistent with Svenson et al. (1985), optimism was less pronounced for accident likelihood than the other two measures, once again suggesting that individuals are less optimistic when judging outcomes over which they have limited direct control. Global optimism was found to increase as driving experience increased. Experience had a similar effect on the comparative risk judgements for the 10 accidents (mean risk), but this correlation was only marginally significant. Spolander (1982) also reported increased optimism as a function of experience. Spolander compared drivers who had had licenses for one month, one year, or three years. New drivers perceived themselves as poorer than average in most respects, but after only one year considered themselves as better than average on many technical driving skills. At three years, most
Optimism bias and risk perception
339
drivers considered themselves to be better than average on a variety of both technical and defensive driving skills. Subjects, in the present study, also assigned greater importance to human factors in accident causation as their driving experience increased. Moreover, those who held human factors to be important also considered themselves to be more skillful drivers. If optimism is related to an exaggerated sense of control, then as optimism develops with driving experience the importance assigned to human factors should also increase. It is also reasonable to expect that those who assign greater importance to human factors will also claim that they take advantage of that control. Optimism also increased somewhat with age. Marginally significant correlations were obtained for both global driving safety and mean risk. These results must be interpreted with considerable caution, both in view of their magnitude and restricted age range of the sample. McCormick and colleagues (McCormick, Walkey, and Green 1986), using a somewhat broader age range sample, failed to detect effects for age. However, other data exist suggesting that optimism, although evident in older subjects, diminishes with age (Svenson et al. 1985). Unfortunately, in Svenson’s study age was confounded with cultural differences in the study sample. Matthews and Moran (1986) also reported that older drivers were less likely than younger drivers to set themselves above their peers in judging their driving abilities. With respect to the perceived risk associated with various driving situations, Finn and Bragg (1986) found that younger drivers tended to perceive less risk in various driving situations than did older drivers. One possible explanation for the inconsistent findings in the literature is that optimism increases during the first several years postlicensure, but decreases thereafter. Assuming such a curvilinear relationship, increased optimism as a function of driving experience and/or age would be expected for a young subject sample (as in the present study), while opposite effects would be expected for older subjects. This reasoning is consistent with Brown’s (1982) analysis of the differential acquisition of vehicle control skills and road craft over time. According to Brown, because young drivers acquire perceptual-motor skills quite easily, they quickly gain a high level of confidence in their skills. With continued experience, however, drivers acquire a greater appreciation of the complexities of the driving task, and a better match exists between their level of confidence and their decision skills. Another interesting finding from the present study is that male and female subjects did not differ substantially either in terms of their global judgements of driving safety, skill, and accident likelihood, or their comparative risk judgements for the 10 accidents. Svenson et al. (1985) and McCormick et al. (1986) also failed to find consistent effects for sex. Other recent studies have tended to use only male subjects (e.g. Finn and Bragg 1986; Matthews and Moran 1986). The well-documented excess involvement of young male drivers in traffic accidents (e.g. National Accident Sampling System 1981; Williams and Karpf 1984), combined with the absence of differences in optimism between males and females in recent research, suggests that the excess involvement of males may not be directly related to sex differences in risk perception per se, but rather to how these perceptions are translated into driving behavior and risk-taking. In practical terms, the existence of such robust optimism with respect to traffic accident risks has important implications. Referring to Fuller’s (1984) threat avoidance model of driving behavior, excessive optimism should decrease the probability of appropriate anticipatory avoidance responses. That optimism appears to be strongly related to an exaggerated sense of control suggests that efforts should be directed at reducing the driver’s sense of perceived control over the driving environment. Driver education programs are needed that provide direct demonstration of the limits of one’s driving skill in accident avoidance. Young drivers are especially in need of objective data demonstrating the mismatch that exists between their perceived and actual levels of driving skill (see Brown 1987). Special attention should be given to providing drivers with more realistic appraisals of perceived vulnerability. Research indicates that people possess fairly veridical perceptions of overall traffic risks (Lichtenstein et al. 1978); younger drivers may even appreciate that their age group is at greater risk than are older drivers (Finn and Bragg
340
D. M. DEJOY
1986). Indeed, in some instances, young drivers may actually consider driving to be a more dangerous activity than do older drivers (Berger and Persinger 1980). However, the problem is that most drivers, and especially young drivers, do not believe that these risks apply to them personally. As such, interventions must seek to counter the perception of exaggerated control by means that personalize the risk. Standard educational efforts, warnings, and statistics are not likely to be effective. REFERENCES Berger, R. J.; Persinger, G. S. 1980 survey of public perception of highway safety. McLean, VA: Automated Services Inc. ; 1980. Brown, I. D. Exposure and experience are a confounded nuisance in research on driver behavior. Accid. Anal. Prev. 14:345-352; 1982. Brown, I. D. Predisposing factors in the alcohol- and drug-impairment of young drivers’ performance. In: Beniamin, T., editor. Young drivers impaired bv alcohol and other drugs. London: Roval Societv of Medicine Series; 1987:165-172. Cohen, A.; Smith, M. J.; Anger. W. K. Self-protective measures against workplace hazards. J. Safety Res. 11:121-131; 1979. Finn, B.; Bragg, B. W. E. Perception of the risk of an accident by younger and older drivers. Accid. Anal. Prev. 18:289-298; 1986. Fuller, R. A conceptualization of driver behavior as threat avoidance. Ergonomics. 27:1139-1155; 1984. Green, L. W.; Kreuter, M. W.; Deeds, S. G.; Partridge, K. B. Health education planning: A diagnostic approach. Palo Alto, CA: Mayfield; 1980. Janz, N. K.; Becker, M. H. The health belief model: A decade later. Health Educ. Q. ll:l-47; 1984. Lichtenstein, S.; Slavic, P.; Fischhoff, B.; Layman, M.; Combs. Judged frequency of lethal events. J. Exper. Psychol. Human Learn. Memory 4:551-578; 1978. Matthews, M. L.; Moran, A. R. Age differences in male drivers’ perception of accident risk: The role of perceived driving ability. Accid. Anal. Prev. 18:299-313; 1986. McCormick, I. A.; Walkey, F. H.; Green, D. E. Comparative perceptions of driver ability-a confirmation and expansion. Accid. Anal. Prev. 18:205-208; 1986. National Accident Sampling System, Abt Associates, Inc. Analysis of 1979 computerized data. Abt; 1981. Rosenstock, I. M. The health belief model and preventive health behavior. Health Educ. Mono. 2:354-386; 1974. Spolander. K. Oerfarna forares beteende, presentations-formage och attityder (Inexperienced driver’s behavior, ability, and attitudes). Linkoping: Swedish National Road Traffic Research Institute; 1982. Svenson. 0. Are we all less risky and more skillful than our fellow drivers? Acta Psychol. 47:143-148; 1981. Svenson, 0.; Fischhoff, B.; MacGregor. D. Perceived driving safety and seatbelt usage. Accid. Anal. Prev. 17:119-133; 1985. Weinstein, N. D. Unrealistic optimism about future life events. J. Personal. Sot. Psychol. 39:806-820; 1980. Weinstein, N. D. Unrealistic optimism about susceptibility to health problems. J. Behav. Med. 5:441-460; 1982. Weinstein, N. D. Why it won’t happen to me: Perceptions of risk factors and illness susceptibility. Health Psychol. 3:431-457; 1984. Weinstein. N. D. Unrealistic optimism about susceptibility to health problems: Conclusions from a communitywide sample. J. Behav. Med. 10:481-500; 1987. Williams. A. F.; Karpf, R. S. Teenaged drivers and fatal crash responsibility. Law Policy Q. 6:101-113; 1984.