Paper Writing Service http://StudyHub.vip/Age-And-Gender-Differences-In-Perceived 👈

1. Risk Analvsis, zyxwvutsrqponm
Vol. zyxwvutsrqp
16, No. 6, 1996 zyxwvutsrqpo
Age and Gender Differences in Perceived Accident
Likelihood and Driver Competences
A. Ian Glend~n,'.~
Lisa Dorn,Z D. Roy Davies,' Gerald Matthews? and Ray G. Taylor'
Received April 10, 1995; revised May 16, 1996 zyxwvu
Road traffic accident involvement rates show clear age and gender differences which may in part be
accounted for by differences in risk perception and perceptions of driving competence. The present
study extends and replicates that of Matthews and Moran (1986). Young (18-30 years) and older
(4540 years) male and female drivers responded to a questionnaire on perceived accident risk and
driving competence (judgment and skill) with respect to themselves and four target groups, and also
rated a series of videotaped driving sequences with respect to likelihood zyx
of accident occurrence and
perceived driving competence. Results showed that effects of rater characteristics were generally
confined to the questionnaire. Younger males were perceived zyxw
as most likely to experience an accident
and were judged to be lower than other groups in driving competence. Younger groups showed little
bias against older groups and vice versa, but gender-related bias was apparent. The findings of
Matthews and Moran were generally confirmed. The results are discussed with reference to four main
issues: (1) demographic bias effects-which are generally weak; (2) stereotyping on the basis of
gender and/or age of driver; (3) group-specific bias; zyxwv
(4) self-appraisal bias.
KEY WORDS: Perceived accident risk; driving competence; age differences; gender differences.
1. INTRODUCTION
Investigations into how individuals perceive risk
may have important applicationsin respect of their driv-
ing behavior. Thus, speed of hazard perception appears
to predict accident risk.(' zyxwvutsr
) Hugenin(*)
maintainsthat driv-
ing interventions should take account of such driving
characteristics as risk perception, while Matthews and
M~ran'~)
conclude that studies of risk perception have
implications for countermeasures intended to reduce ac-
cident risks.
Links between perceptions and behavior are com-
I Human Factors Research Group, Aston University, Birmingham,
United Kingdom.
2Department of Social Psychology, School of Social Sciences, De
Montfort University, United Kingdom.
Department of Psychology, Dundee University, United Kingdom.
To whom all correspondence should be addressed at School of Ap-
plied Psychology, Griffith University, Gold Coast Campus, PMB 50
Gold Coast Mail Centre, Queensland 4217, Australia.
plex but there is reasonable agreement that perceptions
of risk on the road affect dri~ing.(~.~)
Experimental stud-
ies demonstrate the importance of risk perception, for
examplethat speed choice on blind curves is affected by
information about risk@)and that intention to speed is
predicted by beliefs about accident likelihood.'') Risk
perception bears some relation to objective risk levels-
for example, estimates of accident frequency at particular
locations correlate with accident frequencies at those lo-
cations.@)
More generally,while more frequentlyoccurring
causes of death tend to be underestimated and uncommon
causes tend to be overestimated,numbers of deaths result-
ing from motor vehicle accidents have been found to be
accurately estimated.'y)This may be due to the media and
other publicity given to such statistics.Nevertheless, gen-
erally people are imperfectjudges of risk, so that risk rat-
ings tend to be biased by decision makmg heuristics."0)
Personal biases in risk perception may be associated with
age and gender variables. For example,someobservational
755
0272-4332/Y6/1200-0755$0Y.5uil zyxw
8 1996 Society for Risk Analysis

2. 756 zyxwvutsrqpo
Glendon, Dorn, Davies, Matthews, and Taylor zy
studies(”J*)
suggest that women take fewer risks than men
do when driving. The aim of the present study is to ex-
amine group differences in risk perception in a controlled
laboratoryenvironment.Althoughrealism is sacrificedwith
this method, it allows for control of extraneous and pos-
sibly confounding variables.
A relatively high accident rate amongst young male
dnvers is thought to be partly due to increased risk-tak-
ing in this gr0up.(I3)
Summala(14)
suggeststhat during the
first 50,000 km, novice drivers gradually develop im-
proved dnving practices and hazard control skills. How-
ever, he maintains that overconfidence among younger
drivers arises from early mastery of basic car-control
skills, while their hazard control skills remain defective.
argues that compared with their older counter-
parts, younger drivers may be less able to recognize a
hazardous situationwhen it arises and thereforeare more
likely to take risks on the road. Most drivers consider
themselves to be more competentand safer than the “av-
erage” driver,(16)
this being particularly true for males
shortly after passing their driving test.“’)
Studies which have assessed drivers’ perception of
risk in various driving situationshave reported that driv-
ers underestimate certain traffic hazards and overesti-
mate their own driving abilitie~.(~.’~,’~.’~’
Matthews and
Moranc3)found that perception of risk among males and
confidence in driving abilities is dependent on age. Us-
ing both a questionnaire and a series of videotaped se-
quences of various dnving behaviors, young and older
males gave ratings of vehicle-handling,driving reflexes
and drivingjudgment for themselvesand for other target
groups. Results showed that compared with older driv-
ers, young drivers gave higher estimations of future ac-
cident involvement, but gave lower ratings of accident
risk for driving which required fast driving reflexes or
vehicle-handling skill. They were more confident than
older drivers in their own driving abilities. Further,
younger drivers believed that their peers were signifi-
cantly more at risk than they themselves were and pos-
sessed poorer abilities.
Matthews and Moran’so)questionnairedata showed
no significant correlation between ratings for risk and
ratings for driving ability amongtheir sampleof younger
male drivers. However, data for older male drivers
showed significantpositive correlationsbetween risk rat-
ings and ratings for ability, reflexes, judgment, and ve-
hicle-handlingskill. Risk ratings were strongly inversely
related to ability ratings for the videotape data for both
young and old groups.
This study is a replication and extension of that of
Matthews and Morad3’to include gender as well as age
of rater and of “target groups” of drivers rated. As in
Matthews and M~ran,‘~)
ratings from a general question-
naire and for videotaped sequences of specific maneu-
vers were obtained. Effects on ratings of both the
between-participantsvariables of age and gender of tar-
get groups, or ratees, were analyzed. Four types of bi-
asing effects were postulated.
First, age and gender differences in risk perception
such as those identified by Matthews and MoranI3)are
instances of a general demographic bias in raters’ per-
ceptions which generalize across a variety of traffic sit-
uations. Demographic bias would be demonstrated
empirically by effects on ratings of between-participants
variables of rater age and gender.
Second, in stereotyping particular groups, there
may be a general societal tendency for older people to
be rated as less competent than younger people.(2o)
Hence, older drivers may generally be seen as being
more likely than younger drivers to be involved in an
accident. Operationof stereotypeswould be indicated by
effects of within-participants variables, ratee age, and
gender.
Third are group-specific biases-whereby rater
groups may show particular biases in their perceptions
of target groups. For example, in occupational settings,
younger raters of job aptitude show greater age stereo-
typing than do older raters.cZ0)
Younger raters might as-
sign higher risk ratings to older ratees than older raters
would. Such bias would be manifested by interactions
between rater and ratee variables.
Finally, bias in self-appraisal and the extent to
which it varies with rater characteristics can be evalu-
ated. Matthews and Moran(3)found that younger male
drivers overestimatedtheir own ability relativeto that of
their peers, while older male dnvers did not. A study of
different driving scenarios found that drivers rated their
own driving skillsas superiorto those of other drivers.(2’)
Participants rated other drivers’ competence as average
rather than poor, suggestingthat they adopted a positive
self-bias rather than a negative other-bias.Males showed
greater self-enhancementbiases across all driving skills,
whereas females showed less bias in certain specific
skills. With the experience effect taken into account,
gender differences were not significant.
2. METHOD
Participants were 60 drivers who were Aston Uni-
versity students and members of mainly nonacademic
staff, including cleaners, porters, and clerical grades.
There were four groups of 15: younger and older men
and younger and older women. The younger groups

3. Perceived Accident Likelihood in Driving zyxwvu
757 zy
Table I. Group Means Showing Age, Experience, and Exposure for
Four Target Groups (Young and Older Males, Young and Older
Females)
Group means
Males Females
Young Older Young Older
Age 24.9 51.7 22.0 50.0
Years since license 5.9 29.7 3.7 24.3
Miles driven per year 9650 11,150 7650 8650
were aged between 18 and 25 years and the older be-
tween 45 and 60 years. Demographically, the sample
was comparableto that used by Matthews and M~ran,’~)
except that the older group in their study of undergrad-
uates, university faculty and staff was aged 35-50 years
and they studied only male drivers. Information was col-
lected on driver age, driving experiencedefined as
time since obtaining a full driving licence, and on ex-
posure-defined as estimated number of miles driven
during the previous 12 months (see Table I). The
younger participants in our sample were, on average,
more experienced drivers than those examined by Mat-
thews and Moran.t3)Thus, even the younger samplewere
relatively experienced drivers. Further, males in the cur-
rent sample had driven on average approximatelytwice
as many miles as had female participants, were older by
an average of nearly 3 years, and had held a driving
licence for over 2 years longer. Whether these differ-
ences might account for any observed differences in rat-
ings described is uncertain. However, it is possible that
differences in driving experience between all groups re-
flected population level differences.
The experiment was in three parts. In Part A, par-
ticipants used 100 mm visual analogue scales (anchored
by zero and 100) to estimate the likelihood of being
involved in a traffic accident (defined as any collision
involving other vehicles or objects) in the year ahead.
They made estimations in respect of themselves, their
peer group, and the three other participant groups.
In Part B, participants rated two aspects of driver
performance: skill and judgment. These ratings were
again made for themselves, for their peer group and for
the three other target groups. Driving skill was defined
as a driver’s skill in vehicle-handling and in controlling
the speed and direction of the car. Drivingjudgment was
defined as a driver’s ability to assess the traffic situation
and to choose a maneuver which would be safe in that
situation.Fivejudges rated these measures and were able
to distinguish between driving skill and driving judg-
ment.
In Part C, 19videotaped driving sequences,ranging
in duration from 11 to 48 seconds, depicted a target ve-
hicle in a driving situation. In each sequence a target
vehicle had to take avoiding action in order to prevent
an accident. The viewer had an overall perspective of
the road and prevailing traffic conditions. One was used
as a practice sequence.Fivejudges graded the remaining
18 sequences for the degree of risk involved using
Transport Research Laboratory Guidelines for traffic
conflict. These guidelines are used as criteria for deter-
mining three categoriesof risk-high, medium, or low-
based on observed sequences. Following the practice
sequence, the order of presentation of the remaining 18
sequences was block randomised according to degree of
risk-low, medium, or high. There were six sequences
at each risk level. Four types of driving situation shown
were: (1) target vehicle has to avoid a vehicle which
emerges from a T-junction into its path; (2) target ve-
hicle has to avoid another vehicle which emerges from
a crossroads into its path; (3) target vehicle is closely
followed by other vehicles when approaching a round-
about (a multi-intersection); (4) target vehicle is shown
driving at speed and taking comers sharply. The four
types of driving situations were devised so as to be as
comparable as possible to those used by Matthews and
Moran”)as well as being intended to elicit ratings based
on a variety of road vehicle encounters. The final type
of driving situationwas the only one to portray the target
vehicle driver as the prime source of danger.
The order in which ratings were requested in parts
A, B, and C was randomly varied so as to minimize the
likelihood that participants would apply “set rules”
when rating accident likelihood and driving competence
for self and target groups. For each video sequence,par-
ticipants were instructed to focus on the target vehicle
and, using the same 100 mm visual analogue scale as in
Parts A and B, to rate: (1) the likelihood of an accident
occurring (on a per driver basis) in the situation for
themselves and for each of the target groups as driver;
(2) their confidence in their own driving skill and driving
judgment in the situation depicted in the sequence; and
(3) to make the same confidenceratings ofjudgment and
skill with respect to each of the target groups. Partici-
pants thus made 18 ratings for each videotaped se-
quence.
The visual analogue rating scale required partici-
pants to give ratings based on these semanticallydiffer-
ent anchor points for skill ratings: very high skill/very
low skill, forjudgment ratings: very goodjudgmentlvery
poor judgment and for likelihood of an accident: not at
all likely/extremely likely.

4. 758 zyxwvutsrqp
Glendon, Dorn, Davies, Matthews, and Taylor
Table 11.zyxwvutsrqp
Questionnaire Ratings (Mean Values) of Driving Skill and Judgment Made by Four
Groups of Drivers (Young and Older Males, Young and Older Females) in Respect of Four
Target Groups (Young and Older Males, Young and Older Females) and Self
Rater
Skill ratings Judgment ratings
Males Females Males Females
Ratee Young Older Young Older Young Older Young Older
Young males 55.67 62.67 62.33 59.00 49.67 47.00 54.33 50.00
Older males 53.67 61.33 67.00 68.33 56.00 64.00 67.33 65.67
Young females 49.00 47.67 55.33 61.33 54.67 51.00 56.33 63.00
Older females 44.67 38.00 58.33 66.67 47.67 43.67 59.67 67.00
Self 63.33 67.33 61.67 64.67 71.67 73.33 66.67 62.67
3. RESULTS AND DISCUSSION
Questionnaire data for target groups were analyzed
in a zyxwvutsrqpo
2 zyxwvutsrqpo
X 2 X 2 X 2 ANOVA with rater gender and age
as between-participants factors; and “ratee” gender and
age as within-participants factors. Data were further an-
alyzed in a 2 X 2 X 2 ANOVA with rater age and
gender as between-participants factors, and group (self
or peer group) as a within-participants factor.
Video sequence ratings were analyzed in a 3 (risk
level: high, medium, and low) X 2 (rater age) X 2 (rater
gender) X 2 (ratee age) X 2 (ratee gender) ANOVA.
Rater age and gender were between-participants factors,
and risk level and ratee age and gender were within-
participants factors. Ratings were also obtained for self,
peer group and the video dnver under three levels of
risk. These were analyzed in a 3 (risk level) X 3 (group:
self, peer and video driver) X 2 (rater age) X 2 (rater
gender) ANOVA, with the first two as within-partici-
pants factors and the other two as between-participants
factors.
Results and Discussion are organized by type of
bias to establish whether demographic, stereotyping,
group-specific, or self-biasing effects are found within
questionnaire and video data. Only results which were
significant at the one percent level or better are presented
and discussed, except for results at the 5% level which
are in accord with such findings, which relate to the
general hypotheses or which support Matthews and Mor-
an’s(” findings. The basis for this decision rests upon a
number of factors; in particular our concern to avoid
Type I1 errors with respect to hypothesized effects, the
possibility that with N = 60, relatively large between-
participants effects may exist between 1% and 5% prob-
ability levels and that interaction effects which are
significant at 5% may reflect stronger effects within
some of the cells.
3.1. Demographic Bias
The first hypothesized effect related to main effects
of rater gender and age on accident likelihood and driv-
ing competence ratings. Results revealed main effects of
rater gender for questionnaire ratings of skill and judg-
ment (skill: F[1,56] = 10.26, p < .01; judgment:
fl1,56] = 7.18, p < .01) and, to a lesser extent, for
accident likelihood (F[1,56] = 5.11, p < .05). Female
participants gave significantly higher ratings for both
these aspects of driver competence than did males (see
Table 11) but saw accidents as more likely than males
did.
We did not replicate findings that young drivers
perceive driving as generally more hazardous than do
older or Brown’s“5) suggestion that younger
drivers may be less able to perceive hazardous situa-
tions-perhaps because of the relatively high driving ex-
perience level of the younger sample. The demographic
bias revealed by the questionnaire data may not be sus-
tained when participants are confronted with behavioral
data because the standard stimulus video material an-
chors raters’ perceptions so as to eliminate gender dif-
ferences.
3.2. Stereotyping
The second hypothesized effect was bias in respect
of ratees. From the questionnaire, but not from the video
data, there was a main effect of ratee gender zy
(F[1,561 =

5. Perceived Accident Likelihood in Driving zyxwvu
759
41.66, p zyxwvuts
< .001) on accident likelihood ratings. Male
drivers were rated as more likely than female drivers to
be involved in an accident. For both questionnaire and
video data, there was a main effect of age of ratee for
accident likelihood ratings (questionnaire: 61,561 =
18.59,~
< .001; video: F[1,56] = 20.42,~
< .001) and
a significant interaction of ratee gender and age (ques-
tionnaire: F[1,56] = 35.24, p < .001; video: F[1,56] =
9.42, p < .OOl). The main effect shows that younger
drivers were rated as more likely than older drivers to
have an accident. The interaction indicates that young
male drivers were rated as much more likely to be in-
volved in an accident in the next 12 months than were
other groups.
For questionnaire skill ratings there was a main ef-
fect of ratee gender (F[1,56] = 17.28,p < .001) with
women being rated as lower in driving skill. From the
video data, interaction effects for skill ratings were
found between risk level and ratee gender (F[2,112] =
8.75, p < .001), risk and ratee age (F[2,112] = 13.58,
p < .001) as well as risk and ratee gender and age
(F[2,112] = 9.85,p < zyxwvu
.001). Women were rated as su-
perior to men in skill at each level of risk, as were young
drivers compared with older drivers. The interactionsbe-
tween risk, and ratee gender and age, indicate that
younger women were rated as superiorin skill compared
with other groups.
The questionnaire data revealed main effects of ra-
tee age for judgment ratings (fl1,56] = 7.40,p < .01)
indicating that older drivers were rated more favorably
than young drivers were (young = 53.25; old = 58.88),
an effect which was not found in the video data.
For both questionnaire and video data there were
also interaction effects between ratee gender and ratee
age on driving judgment (questionnaire: F[1,56] =
36.22,~
< .001; video: fl1,56] = 4 . 2 6 , ~
< .05). These
interactions indicate that for the video sequences,young
females were rated as being superiorto all other age and
gender groups, whereas in the questionnaire data, older
males were rated as superior in judgment to all other
gender and age groups.
Evidence was thus obtained for biasing through
general stereotyping effects. The age by gender inter-
action was significantfor both sets of accident likelihood
andjudgment measures. Our data suggestthat Matthews
and Moran’d3)finding that older people are rated as be-
ing at lower risk of accident involvement,and higher on
most performance ratings, may be limited to male driv-
ers. Our participants appeared to possess fairly clearly
delineated stereotypes of drivers of differing demo-
graphic characteristics.
3.3. Group-Specific Bias
A third hypothesized effect was for rater by ratee
interactions, demonstrating group-specific rater bias
against those rated. For both questionnaire and video
data, significantrater gender by ratee gender interactions
were found with respect to accident likelihood ratings
(questionnaire:F[1,56] = 9 . 7 5 , ~
< .005;video: 61,561
= 5.64, p < .025). Both gender groups rated males as
more likely than females to be involved in an accident.
The interaction indicated that males were rated as more
likely to have an accident by female raters than they
were by male raters. No evidence for age-specificbiases
were found. Group-specific biases on this evidence de-
pend on driver gender but not on age. This finding may
again be accounted for by the relatively high level of
experience of this sample.
A significant interaction between rater gender and
ratee age was also found from the video data (fl1,56]
= 11.39, p < .001) indicating that older drivers were
rated by females as being less likely to have an accident.
In contrast, males rated both young and older drivers as
being about zyxw
of equal risk of an accident. The absenceof
a comparable effect from the questionnairedata suggests
that in making more general judgments about accident
likelihood, participants did not make discriminatory
judgments about age group of driver. However, a weak
comparable biasing effect in questionnaire skill ratings
(F[I,56] = 6.81, p < .05) was found.
3.4. Self-Appraisal Bias
On the fourth hypothesized effect, that raters would
show bias in rating self compared with peers, from both
questionnaire and video data there was a main effect of
group (questionnaire:F[1,56] = 5.69, p < .05; video:
61,561 = 31.94, p < .001) indicating that raters’ peer
groups (and driver in the video sequences) were rated
as more likely to be involved in an accident than were
raters themselves.
From the questionnaire,though not from the video
data, a weak but significant age by group interaction
(61,561 = 6.35, p < .05) was obtained, indicating that
younger drivers rate their peers as being more likely than
themselves to be involved in an accident, providing
some support for Matthews and zyx
M0ran.O)The reason for
the video data not supportingthe questionnaire data may
again lie in the specific driving behaviors observed in
these sequences, from which participantswere unable to
make significantdiscriminating accidentlikelihoodjudg-

6. 760 zyxwvutsrqpo
Glendon, Dorn, Davies, Matthews, and Taylor
Table 111. Videotaped Sequence Ratings (Mean Values) for
Accident Likelihood Made by Four Groups of Drivers (Young and
Older Males, Young and Older Females) in Respect of Self, Peer
and Driver of Target Vehicle, at Three Levels of Risk (Low,
Medium and High)
Rater zyxwvutsrq
~~~ ~~~
Young males Older males
Risk level
Group Low Medium High Low Medium High
Driver zyxwvutsrqpon
44.32 60.23 72.27 47.68 65.23 68.58
Peer 41.71 56.61 64.98 43.07 58.50 64.62
Self 35.66 47.78 58.98 41.46 55.02 59.99
Driver 41.29 60.28 71.54 47.89 67.68 78.31
Peer 40.79 55.16 63.34 41.11 55.01 58.83
Self 38.64 52.72 62.23 38.63 52.16 56.24
Young females Older females
ments. However, there was a rater age by rater gender
by group interaction from the video data (fl1,56] =
5.99, p < zyxwvuts
.05), indicating that older males considered
themselves to be more likely than their peer group to be
involved in an accident, whereas other gender and age
groups consideredthemselvesto be less likely than their
peers to be involved in an accident.
With respect to driver competence ratings, there
were significantmain effects of group on skill andjudg-
ment ratings for both questionnaire (see Table 11) and
video data (skill:questionnaire:F[1,56] = 5 . 3 8 , ~
< .05;
video: fl1,56] = 5.32, p < .05; judgment: question-
naire: q1,56] = 17.79, p < .001; video: fll,56] =
14.36,p < .001) indicating that participants rated them-
selves as higher on these aspects of driving than either
their peers, or the driver depicted in the video sequences.
From the questionnaire, though not from the video zyxwv
data, there was a gender by group interaction (fl1,56]
= 8.19, p < .Ol) for judgment ratings. The difference
between self and peer ratings was much greater for male
than for femaleparticipants. zyxwvuts
An age by group interaction
forjudgment ratings (fl1,56] = 9 . 5 4 , ~
< .01) was also
found, indicating that although both older and younger
participants rated themselves as superior in judgment to
their peers, the difference in rated judgment (see Table
11), is much greater for younger than for older partici-
pants. A risk level by group interaction was obtained
from the video ratings (fl4,224] = 9.40, zyxwvuts
p < .OOl).
Rated accident likelihood involvement rose with situa-
tional risk level for all three groups, but much more
steeply for the driver than for self or peer (see Table
111).
For skill ratings there was also a rater gender by
rater age by group interaction (F[2,112] = 9.02, p <
.01+older males rated themselves as only marginally
better in skill than their peers (49.39 vs. 46.17), com-
pared with other groups who rated themselves much
higher than their peers in skill. Raters generally per-
ceived themselves as being of higher competence and at
lower risk than their peers, consistent with previous re-
~earch.(~.’~)
The discrepancy between peer and self rat-
ings of skill increased with risk level.
For the questionnaire measure of accident likeli-
hood, similar results to those of Matthews and Morad3)
for male drivers were obtained. Young males saw them-
selves as much less likely than their peers to have an
accident, but this effect did not extend to older males.
Both young and older females saw themselves as being
at less risk than their peers. As in Matthews and
M~ran,’~’
these effects were less apparent in the video
data. The expected age by group interaction was ob-
tained for ratings of drivingjudgment, with younger par-
ticipants showing a greater overestimate of judgment
relative to their peers compared with older participants.
A separate gender by group interaction suggests that
males may be more prone than females to overestimate
their own drivingjudgment. In the video sequence meas-
ures, older males estimated their driving judgment as
being only marginally better than that of their peers,
whereas the other three groups gave much higher ratings
to themselves than to their respective peers. Young
males showed the greatest mismatch between self and
peer ratings. Males rated themselvesas more skilledthan
their peers, but females did not. Table IV summarises
significant findings for questionnaire and video data.
3.5. Correlations Between Ratings
For both target group and self-rating data, Pearson
correlations were computed for questionnaireand video-
sequence ratings. All pairs of correlations (range 0.50-
0.92) for driver performance ratings (skill andjudgment)
on the questionnaire and video sequences were signifi-
cantlypositively correlated(p < .001 in all cases).How-
ever, the correlation between accident likelihood ratings
for questionnaire and video sequences (0.21) was not
significant. A possible reason for this finding is that
when asked to rate accidentlikelihoodwithin the context
of the questionnaire, participants are making general as-
sessments of accident likelihood for each group. How-
ever, after observing each video sequence, participants
are rating target groups in the specific context of that

7. Perceived Accident Likelihood in Driving zyxwvu
761
sequence.Thus, while a small general stereotypingeffect
may carry over from questionnaire ratings, this is insuf-
ficient to overcome the powerful influence of the (same)
objective behavior observed in the video sequences.
Significant negative correlations were obtained be-
tween ratings of accident likelihood and driver skill (r zyxwv
= -0.47, p < .001) and judgment (r = -0.59, p <
.001) for the video data. However, comparable correla-
tions for questionnaire data and correlations between
questionnaireand video data were nonsignificant.A pos-
sible reason for these findings is that when asked to rate-
specific instances of skill, judgment, and accident
likelihood-as in the case of the video sequences, in
order to achieve consonant rating patterns, participants
produced accident likelihood ratings which were consis-
tent with their skill and judgment ratings. However,
when responding to the questionnaire, participants gave
generalizedratings of these variables.
3.6. Methodological Issues
While no claims for ecological validity are made,
some brief comments on methodology are pertinent. In
laboratory measures of risk perception,particularlyusing
small samples, the extent to which results can be gen-
eralized to driving situations is problematic, although
such measures of risk perception have shown significant
positive correlations with accident inv01vement.(’~~~~*~)
However, as this study only compared zyxwvut
risk perceptions
of different dnver groups, relating risk perception com-
ponents to driving behavior requires further study.
A further issue is that not all significant findings
are replicated between questionnaireand video sequence
data (see Table IV). The questionnaireessentiallymeas-
ures general beliefs, which may not influence (measur-
able) appraisals of specific driving situations. The video
sequences, particularly at the highest risk level, may
have been unrepresentative of ordinary driving, partic-
ularly in the “speeding” clips where the target driver
engages in behavior which participantsmight themselves
rarely or never engage in. This indicates a need to con-
sider age, gender, and risk perception within specific
drivingbehaviors. For example, Bragg and Finn(22)
found
that younger drivers rated speeding and tail-gating as
less risky than did older drivers.
A final problem is the possibility of rating scale
artefacts. The accident likelihood rating scale is essen-
tially logarithmic, so that likelihood ratings are analo-
gous with probability judgments (e.g., 1 in 100, 1 in
1000, etc.) whereas the skill and judgment scales are
Table IV. Summary of Significance Levels @ Values) for
Questionnaire and Video Rating Effects’
Ratings of perceived
Variable, effect Accident
bias (medium) likelihood Judgment Skill
______________ _____
Demographic
Stereotype
Rater gender (Q)
Ratee gender (Q)
Ratee gender X ratee age (Q)
Ratee gender X ratee age (V)
Risk X ratee gender
Risk X ratee age
Risk X ratee gender X ratee age
Ratee age (Q)
Ratee age (V)
Risk
Rater gender X ratee gender (Q)
Rater gender X ratee gender (V)
Rater gender X ratee age (Q)
Rater gender X ratee age (V)
Group specific
Self-appraisal
Groupzyxwvu
(Q)
Group (V)
Rater age X group (Q)
Rater gender X group (Q)
Risk X group
Rater age X rater gender X group (V)
,001
,001
.01
-
-
.001
.oo1
,001
.01 zyx
.05
,001
-
.05
.o1
.05
,001
.05
.o1
-
,001 zyx
.05
.oo1
.oo1
.01
.o1
.oI
.001
-
-
.oo1
,001
,001
-
.05
.05
-
-
.01
Q=Questionnaire ratings;V=Video ratings. For the variable “risk,”
only video ratings were taken.
linear. The absence of intermediate scale values might
lead to overestimations of accident likelihoods. Alter-
natively, participants’ views as to what constituted an
“accident” could have varied. The accident likelihood
ratings obtained suggest that participants may have
sought to respond consistentlyacross items. There might
also be group differences in scale usage.
4. CONCLUSIONS: RISK PERCEPTION AND
GROUP DIFFERENCES IN ACCIDENT
INVOLVEMENT
Matthews and Moranc3)sketch out a model describ-
ing the role of risk perception in accidentcausation.Risk
perception is affected by beliefs about the risks of var-
ious driving situations and the driver’s abilityto perform
various dnving actions within these situations, as well
as immediate perceptual feedback. This analysis sug-
gests several possible explanations for group differences
in risk perception, and hence in driver behavior.

8. Glendon, Dorn, Davies, Matthews, and Taylor zy
Overall, reasonable support was obtained for Mat-
thews and Moran’sI3)contention that younger male driv-
ers tend to see themselves as relatively immune to the
hazards threatening their peers, although the results are
not entirely consistent and there are gender effects on
the extent of self-appraisal bias. Unlike Matthews and
Moran,(3)
we found no evidence that younger males ap-
praise driving as more risky than other groups do,
though it remains likely that group differences in risk
estimation exist with respect to specific driving behav-
iors.(22)
Our data confirm that this group may underesti-
mate their own personal risk and overestimate their
competence compared with their peers. However, al-
though both young males and young females gave lower
self than peer accident risk ratings, younger males esti-
mated their peers’ accident risk to exceed their own by
69.7%: the corresponding figure for younger females
was 3 1.3%.This difference suggests that risk perception
contributes to the gender difference in accident risk
among young people. The data also suggest that age dif-
ferences in risk perception described by Matthews and
Moranc3)are confined to males: younger and older fe-
males underestimated their personal accident risk to a
similar degree. Given the different experience levels be-
tween our sample of younger drivers and that used by
Matthews and Moran,”) it is likely that for at least some
of the variables studied, experience is a more important
variable than age-a proposition that requires testing by
controlling for each effect systematically.
ACKNOWLEDGMENTS
We would like to acknowledge the U.K. Economic
and Social Research Council for financial support in car-
rying out the research described in this paper (Award
Reference No. 202252005). We also acknowledge the
U.K. Transport Research Laboratory and the City of Bir-
mingham Department of Transport for use of material
used in the video sequences. Finally, we appreciate the
useful comments of two anonymous referees on an ear-
lier draft of this paper.
REFERENCES zyxwvutsrq
1. A. R. Quimby and G. R. Watts, zyxwvutsrqpo
Human Factors and Driving Per-
.formance (Laboratory Report 1004, TRRL, Crowthome, 1981).
2. R. D. Hugenin, “The Concept of Risk and Behavior Models in
Traffic Psychology. Special Issue: Risky Decision-Making in
Transport Operations,” Ergonomics 31, 557-569 (1988).
3.
4.
5.
6.
7.
8.
9. zyxwvu
10.
I I .
12
13
14
M. L. Matthews, and A. R. Moran, “Age Differences in Male
Drivers’ Perception of Accident Risk: The Role of Perceived Driv-
ing Ability,” Accident Analysis and Prevention 18, 299-313
(1986).
K. Rumar, “The Role of Perceptual and Cognitive Filters in Ob-
served Behavior,” In L. Evans and R. C. Schwing (eds.),Human
Behavior and Traffic Safety (Plenum, New York, 1985).
J. A. Groeger, and J. D. Brown, “Assessing One’s Own and Oth-
ers’ Driving Ability,” Accident Analysis and Prevention 21, 155-
168 (1989).
L. Hendrickx and C. Vlek, “Effects of Risk Information on Speed
Choice in Blind Curves,” In G. B. Grayson and J. F. Lester (eds.),
Behavioral Research in Road Safety (PA2038191, TRRL, Crow-
thome, 1991),pp. 139-147.
D. Parker, “Intentions to Violate,” In G. B. Grayson and J. F.
Lester (eds.), Behavioral Research in Road Safety (PA2038191,
TRRL, Crowthome, 1991), pp. 118-130.
B. Brehmer, “The Psychology of Risk,” In W. T. Singleton and
J. Hovden (eds.), Risk and Decisions (Wiley, Chichester. 1987).
pp. 25-39.
S. Lichtenstein, P. Slovic, B. Fischhoff, M. Layman, and B.
Combs, “Judged Frequency of Lethal Events,” Journal of Ex-
perimental Psychology: Human Learning and Memonj 4.55 1-578
(1978).
J. A. Groeger and P. R. Chapman, “Errors and Bias in Assess-
ments of Danger and Frequency of Traffic Situations.” Ergon-
omics 33, 1349-1 363 (1990).
E. B. Ebbesen and M. Haney, “Flirting with Death: Variables
Affecting Risk-Taking at Intersections,” Journal zyx
qf Applied Psy-
A. Katz, D. Zaidel, and A. Elgrishi, “An Experimental Study of
Driver and Pedestrian Interaction During the Crossing Conflict,”
Human Factors 17, 514-527 (1975).
A. Jonah, “Accident Risk and Risk-Taking Behavior Among
Young Drivers,’’ Accident Analysis and Prevention 16, 255-271
(1986).
H. Summala, “Young Driver Accidents: Risk Taking or Failure
chology 3, 303-324 (1973).
of Skills?” Alcohol, Drugs and Driving 3(4), 79-91 (1987).
15. 1. D. Brown, “Exposure and Experience Are a Confounded Nui-
sance in Research on Driver Behavior,” Accident Analysis and
Prevention 14, 345-352 (1982).
16. 0. Svensson, “Are We Less Risky and More Skillful than Our
Fellow Drivers?” Acta Psychologica 47, 143-148 (1981).
17. K. Spolander, Biwo ’ ‘rares uppfattning om egen zyx
k
o zyx
“rfo ’ ’ maga
(in Swedish). (VTI-rapport No 252, Linko” ping, 1983).
18. I. A. McCormick, F. H. Walkey, and D. E. Green, “Comparative
Perceptions of Driver Ability-A Confirmation and Expansion,”
Accident Analysis and Prevention 18, 205-208 (1986).
19. zyxwvu
P.Finn, and B. W. E. Bragg, “Perception of Risk of an Accident
by Young and Older Drivers,” Accident Anal.vsis and Prevention
18, 289-298 (1986).
20. B. J. Avolio and G. V. Barrett, “Effects of Interviewing in a
Simulated Interview,” Psychology and Ageing 2, 5643 (1987).
21. F. P. McKenna, R. Stanier, and C. Lewis, “Factors Underlying
Illusory Self-Assessment of Driving Skill in Males and Females,”
Accident Analysis and Prevention 23, 45-52 (1991).
22. B. W. E. Bragg and P. Finn, Younger Driver Risk Taking Re-
search: Technical Report of Experimental Studv. Report prepared
for National Highway Traffic Safety Administration, US.DOT
23. N. Yamashita, “The Cognitionof Risk and the Evaluation of Traf-
fic Desirability as the Attitudinal Predicting Factor of Behaviors.”
Japanese Journal o
f Trafic Psychology 2, 3342 (1986). (in Jap-
anese with English summary).
24. N. Fukazawa, Risk Perception and Driver Improvement. Paper
presented to the 22nd lntemational Congress of Applied Psychol-
ogy, Kyoto, Japan (1990).
HS-806-375 July 1982.