Better to see Chapter 14 How you can reduce your risk  of Traffic Safety  (2004)


Comment:  As of 20 May 2014, I still have never crashed! (after 59 years driving - with no collision insurance)


Words only (no formatting, figures, tables, or photographs) from 1991 book



Paperback copy of complete unchanged book available from , list price $29.95


Chapter 12.  HOW YOU CAN REDUCE YOUR RISK (From 1991 book Traffic Safety and the Driver)


            At a White House news conference in 1960 a reporter asked President Eisenhower, "Sir, do you realize that on your upcoming birthday you will be the oldest President ever to serve?".  Ike smiled and answered, "I believe it's a tradition in baseball that when a pitcher has a no-hitter going for him, nobody reminds him of it" [Humes 1975, p. 155].  Perhaps I would be wiser to take note of that tradition, and not start this chapter by stating that, at time of writing, this researcher of traffic crashes has never actually experienced one in 34 years of driving.  It is, of course, not possible to determine with confidence whether this is merely the result of sheer blind luck, or whether it reflects safer than average driving.  One might, however, note that if a specific individual's crash risk remained constant at the overall average of one per 10 years (Table 13-2), then Eqn 1.1 with x = 3.4 indicates that the probability of having 34 consecutive crash-free years is 3.3% ; an assumed rate of one crash per five years gives a corresponding probability of 0.1%.

            There is no reason to believe that my experience is typical of traffic safety researchers in general, based on the finding in Chapter 6 that knowledge is not the most central factor in crash involvement rates.  An informal (and not too confidential) survey I conducted at an international traffic safety meeting provided clear evidence that some of those studying the subject had involvement and violation rates well above the average.  Although the average rates reported by the researchers are lower than average rates for
the general public, I consider this an unreliable indication because of two large biases -- imperfect memory, and a tendency towards more socially acceptable answers.  Both biases operate in the direction of lowering self-reported rates.  My casual observation is that those of us who attend meetings on traffic safety, including meetings that focus on the role of alcohol, are not dramatically more sober than the general run of humanity.  More objective data on the behavior of safety researchers were obtained by Summala [1987], who measured the speeds of Finnish road-safety researchers as they approached a hotel in which a national road safety meeting was to be arranged.  Of the 13 researchers who could be tracked by radar in a 60 km/h speed-limit zone, nine exceeded 70 km/h, six of these exceeded 80 km/h, and three of these reached 90 km/h!  The researchers' speeds were, on average, higher than those of the general public.

            In the previous chapters many results relating to aggregate effects have been presented, and interpretations offered of factors that affect overall traffic safety.  Here we address the more personal question of what steps an individual driver can take to reduce his or her personal crash risk.  We do not cover the whole range of driving situations featured in many "How to drive safely" books.  Rather, we treat in detail a few specific aspects of driving with a view to gaining insight into general approaches to increasing personal safety.




            A theme we touched on in Chapter 6 is that direct experience continuously reinforces an impression that driving is extremely safe.  Let us invoke the mental construct of a hypothetical "average driver", who has a 0.1 probability of crashing per year.  Such an individual has a better than even chance of
driving six and a half years without crashing (or over 100 000 km, assuming 16 000 km per year); in the same period, this driver also has a 14.0% probability of being involved in two or more crashes, which is how the average comes out to be one per 10 years (computed using Eqn 1.1 with x = 0.65).  The rich repetitive feedback from frequent driving cements the impression that appropriate and safe driving behavior is being used, and fits the "zero-risk" interpretation of Na?"a?"ta?"nen and Summala [1976].  This average driver has no direct way of knowing that a natural and essentially inevitable consequence of such average driving is involvement in one crash per 10 years, or 6 crashes in a 60-year driving career.  

            As the probability that our hypothetical driver experiences at least one crash over a driving career is greater than 99.7%, one might view it as near certain that this driver will be involved in a crash.  A common reaction of drivers involved in crashes is to view them as rare unpredictable events outside reasonable human control, and of such a unique nature that nothing like them will ever recur.  Yet, for our hypothetical driver, rather than being unpredictable, a crash is essentially inevitable.  To realistically expect less than six crashes over a driving career, our hypothetical driver must adopt changed driving behavior so that it no longer matches that same average which copious direct feedback indicated is appropriate.  A crash rate of 0.1 per year would be an unthinkable average for airline pilots, yet flying a plane appears to involve intrinsic risks far in excess of those associated with driving.  In my view, it is within the control of the ground-vehicle driver to approach the same low crash risks which commercial pilots achieve flying (although conventional wisdom is that pilots have above-average crash risks on the road).  Before addressing how a driver might greatly reduce individual crash risk, it is helpful to divide all crash-involved drivers into categories.




            Police procedures generally categorize drivers involved in crashes as being either "at fault" or "not at fault".  Our use of these terms does not deny that each crash has many antecedents, nor that vehicle, roadway or environmental conditions may have played a role.  Nor does it suggest, let alone imply, that countermeasures other than changed driver behavior could not have prevented the crash.  Replacing all rural two-lane roads by divided highways greatly reduces crashes for unchanged driver behavior.  The focus of the present chapter is how an individual driver can reduce personal crash risk within the present infrastructure, which includes as a given the behavior of all the other drivers.  

            Placing a crash-involved driver in the "at fault" category implies some violation of traffic law, while the "not at fault" category implies an absence of evidence of a traffic-law violation.  Although such a categorization is tidy for legal purposes, it tends to convey an erroneous impression that not-at-fault drivers are helpless victims of crashes which occurred entirely out­side their control.  It may be more illuminating to place all crash-involved drivers into three categories, although this is also a simplification:


1. At fault

2. While not legally at fault, the driver could still have avoided involvement

3. Unavoidably involved


            Of these categories, the first and third are the easiest to determine.  Most would agree that the majority of single-vehicle crashes could have been avoided by more prudent driver behavior.  The presence of vehicular or
environmental factors does not necessarily imply no possibility of driver control.  If the vehicle has worn tires or brakes, the roadway is icy, or it is foggy, the driver could still have adjusted behavior accordingly.  Chapter 11 presents considerable evidence that drivers do in fact adjust their behavior for changed conditions.  Much of the content of driver education and public service messages focuses on correcting behavior that tends to involve drivers in at fault crashes.

            In contrast to cases in which the driver has control, there are cases in which drivers are involved in crashes in which their involvement is unavoidable, given that they had decided to drive.  Cases of vehicles driving over bridges which collapse due to structural failure or earthquake, or vehicles stopped in traffic being struck by crashing airliners provide unmistakable, if extreme, examples.  Indeed, such crashes are not even included in the FARS data if judged to have resulted directly from an "act of God" [National Highway Traffic Safety Administration 1989, Chapter 5, page 3].  Drivers involved in such crashes are victims of largely random events over which they have essentially no control.  There are no realistic changes they can make in their driving behavior to reduce such risks.  In my view only a very small fraction of drivers involved in crashes have involvements which can be characterized as entirely outside their control.  Although the fraction is small, one must always remember that, because of the enormous magnitude of the traffic crash problem, the absolute number of drivers fatally injured in such events still exceeds deaths from many other causes which command much more public attention and resources.

            There has been some tendency to view not-at-fault involvement in multiple-vehicle crashes as possessing characteristics in common with the same random model outlined above.  Basically, the driving environment, in the form of the other drivers, is judged to have visited upon the hapless driver some
unfortunate event over which no measure of control is possible.  Many drivers seem to have a view that because they cannot control the behavior of other drivers there is little they can do beyond driving carefully and obeying the law to prevent themselves from being struck.

            Rather than accepting this, I believe that the individual driver can reduce crash risk substantially by taking various steps to avoid not-at-fault involvement.  Indeed, I consider that a large majority of not-at-fault drivers involved in multiple-vehicle crashes fit into the second category, with only a small fraction being in the third category (again, their absolute numbers are large).  Below we discuss some ways in which drivers in the first two categories might reduce crash risk by specific behavior modifications; we discuss rear-end crashes in detail, and leave it to the reader to apply similar notions to other driving situations.




            In most jurisdictions, when a rear-end crash occurs, the following driver is presumed to be at fault, whereas the driver of the lead vehicle is presumed to be not at fault.  Legally, this is how it should be -- it is the responsibility of following drivers to not rear-end vehicles they are following, whereas drivers are entitled to slow down or stop, as the need arises, without incurring legal jeopardy.  Close following, or tailgating, places a minimum of two vehicles at risk of rear-end collision.


What can a following driver do?


            Fig. 12-1 shows the distribution of following headways on a US urban Interstate freeway in Michigan in 1978 [Evans and Wasielewski 1982].  For this
figure, headway is defined as the elapsed time between the front of the lead vehicle passing a point on the roadway and the front of the following vehicle passing the same point; various headway definitions are used in the literature, but the differences are unimportant in the present discussion.  The figure shows data for drivers in two groups, based on their driving records from police files.  One group had one or more violations of any type (most commonly a speed limit violation) in a seven-year period, whereas the other group was violation free; 27.5% of the drivers with violations were observed following at headways less than one second, compared to 21.6% for the violation-free drivers.  Although short headways indicate risk taking, the reverse is not necessarily so; in any stream of traffic there will be many vehicles sufficiently distant from others that it is inappropriate to consider them in a vehicle-following mode [Wasielewski 1979].


Fig. 12-1 about here


            Relative to the reaction times mentioned in Chapter 5, and to the advice in most driving manuals that headways should be at least two seconds, driving with a headway of less than one second must be viewed as behavior that greatly increases rear-end crash risk.  Indeed, most drivers are choosing following headways less than the recommended two seconds.  Wasielewski [1979] in fact finds drivers who are following other vehicles do so with an average headway of 1.32 seconds; that is, the average headway is considerably shorter than the recommended minimum!

            Sivak et al. [1989] investigate differences in driving risk perception in the US, West Germany, Spain, and Brazil by inviting 80 subjects from each of these countries to estimate subjectively the degree of risk implicit in 100 traffic scenes presented in color slides, 50 of which were photographed from
the driver's viewpoint.  The subjects from Brazil judged that close following did not play an important role in increasing risk.  Close following is given the highest risk-increasing rating by the subjects from the US, followed by those from West Germany, Spain, and, as mentioned, Brazil.  These results are consistent with the notion that the danger thought to be associated with close following increases through a collective, rather than individual, learning experience as nations have more historical experience of widespread motorization.

            There are two likely reasons why drivers tend to become comfortable following at headways that unreasonably increase the risk of involvement in rear-end collisions.  First, a dominant cue when following is the relative speed between your vehicle and the one in front.  In normal vehicle following, relative speed is very close to zero.  There is no risk of rear-end collision if both vehicles maintain identical speeds, no matter how high that speed is.  I believe that the largely static visual impression in vehicle-following tends to lower awareness and concern regarding speed.  If the speed of the vehicle in front changes suddenly, then the ensuing dynamical behavior of both vehicles is strongly speed-dependent, with the amount of energy to be dissipated in the event of a crash even more so.  The second reason why drivers have become comfortable following too closely is that they have learned, from repeated experience, that it is safe to do so, in the sense that they have been doing it for years without adverse consequences.  Experience teaches that the vehicle in front does not suddenly slow down.

            Why do drivers choose to follow so closely?  It seems to me that it becomes largely a driving habit, rather than reasoned conscious behavior.  Drivers appear to do many things more for their own sake than for any utility benefit; Katz [1988] suggests that even some criminal behavior is indulged in, not for the expected gain, but for the enjoyment of the activity.  Given that
there are some workers in every type of job who would work even if they were not paid, why should some burglars not also enjoy their work without regard to the normally modest material gain?  Returning to the driving question, we have all observed one vehicle dangerously tailgating another on a stretch of multi-lane freeway containing no vehicles other than our own and these two.  Such tailgaters could often reduce their risk by passing the followed vehicle, and thereby in fact save time.  Unlike many other forms of increased risk-taking in driving, such as speeding, overtaking, or running red lights, tailgating generally provides the driver very little in the way of time savings.  If you ignore the question of other vehicles cutting into the gap in front of you, then following at a headway of 2.0 seconds instead of 0.5 seconds means that you arrive 1.5 seconds later.  If the 1.5 seconds is critical, it can all be recaptured by, say, closing up on the vehicle in front just prior to exiting from the freeway.  In that way the risk of a closer following gap is incurred for just a few seconds, rather than for the entire freeway trip.  Larger gaps do increase the probability that another vehicle will cut in in front.  With a gap of two seconds, this probability is still low, and only arises at high flow levels.  If you are in a slower lane, drivers in a faster lane would rarely want to come in front of you so that they could also travel slower.  If you are in the faster lane, drivers in the slower lane may be more cautious and thereby less likely to perform the risky maneuver of merging into a two second gap.  Even if a few vehicles do cut into the gap in front (my guess is about one per 10 km of freeway travel), this adds only about 2 seconds per such incident to the overall trip time.

            Drivers probably object to other vehicles cutting in front of them not because it delays them a couple of seconds, but because it is interpreted as some sort of personal affront, an assault on manhood or womanhood.  If detached rationality cannot dispel such feelings, comfort might be sought in
the confident expectation that the offending driver is likely to be experiencing more than the average crash rate of one per 10 years.  Let such drivers have their fun -- they are paying a high price for it; recapture your two seconds by walking faster to your vehicle.

            A case in which changing behavior can increase safety and save time arises in single-lane vehicle following when the lead vehicle signals an intention to turn right.  Many following drivers maintain an almost constant headway in such situations, so that the following vehicle's trajectory largely matches that of the lead vehicle, even though the lead vehicle will normally have to reduce speed substantially to execute a tight right turn; if the lead vehicle slows more than expected, the following driver may have to reduce speed even more than the lead vehicle.  A better following practice, from the perspective of both safety and efficiency, is to aim at reducing speed as little as possible by allowing a wider gap to open as soon as the lead vehicle indicates the intention to turn.  When the lead vehicle completes the turn, your vehicle, although travelling slower than prior to the initiation of the turn signal, is still travelling faster than the turning vehicle, thereby saving you time and fuel; by being further from the lead vehicle during its main deceleration, you have also increased your safety margin.

            An individual driver can dramatically reduce the risk of being involved in an at-fault rear-end crash by developing the habit of adopting more generous following headways than experience teaches.  The experience that the vehicle in front does not suddenly slow down must be replaced by the intellectual appreciation that for it to do so is not an event of miraculous rarity.  As Summala [1985, p. 51] comments, the traffic system is not as deterministic as the driver's internal representation of it.  There is simply nothing to be gained commensurate with the potential cost by driving in such a way that you will certainly crash into the vehicle in front if it brakes unexpectedly for
some rare, but not cosmically rare, event such as a small animal unseen by you running onto the roadway.


Tailgating and platoons of vehicles


            Tailgating can produce particularly catastrophic results when a platoon of many consecutive tailgaters forms.  This is because of intrinsic platoon dynamics which may amplify disturbances as they propagate down a line of vehicles [Herman et al. 1959].  If the lead vehicle of a many vehicle platoon slows down gently and then regains its prior speed, the second vehicle may respond by slowing down more rapidly (depending on parameters such as reaction time and headway).  The third vehicle will then be confronted with a more rapidly decelerating lead vehicle, so that as we progress down the platoon, each driver produces a larger deceleration, until eventually braking capability is exceeded.  For a sufficiently long platoon of vehicles with identical following parameters, a multiple-vehicle pile-up becomes inevitable. 

            Fig. 12-2 shows results from a mathematical model of a situation in which a stream of identical cars follow each other, initially all travelling at the same speed and separated by 40 feet.  The position of each car is shown relative to the position of the first car (labelled 1), assuming that this first car continued at a constant speed.  At time zero the first car reduces speed, but then returns to its initial speed.  The curves show how all the following cars respond.  As we proceed  down the platoon, each car approaches closer to the one in front, until the trajectories for cars 7 and 8 intersect.  In other words, car 8 crashes into car 7.  One dear lady confessed that after being exposed to this figure she always made sure she was never car number 7 in a platoon with more than 7 cars! 


Fig. 12-2 about here


            Naturally, her concerns should not have been so specific, as Fig. 12-2 represents model output based on assumed parameters, and different choices would have predicted crashes involving other vehicles.  However, it does illustrate the intrinsic instability of rows of tailgaters, which manifests itself in the real world in the form of multiple-vehicle pile-ups, sometimes with multiple fatalities.  Crashes involving the largest number of vehicles generally occur in fog where rather different principles (such as the lead vehicle suddenly appearing, rather than being followed) may apply.

            An individual driver in a platoon following at a large headway may damp out the disturbance so that no collisions occur.  If you find yourself following many consecutive tailgaters, and are yourself closely followed, then adopt a headway larger than the normally safe two seconds.  Drivers choosing safer headways for themselves may thus make safety contributions to the system, and thereby prevent harm to drivers who will be entirely unaware that somebody else prevented them from being involved in a crash.

            The system-wide effects of safe headways are not necessarily all positive.  If all drivers were so selfish as to reduce their personal risk of rear-end crash involvement by choosing two second headways, the capacity of freeways would decline, and would indeed have a theoretical upper limit flow of 1800 vehicles per lane per hour; incredibly, flows of 2650 vehicles per lane per hour, which corresponds to an AVERAGE headway of 1.36 seconds, have been recorded on a British motorway [Wasielewski 1974].  As this Chapter is aimed at the individual driver, the wise individual decision is to protect yourself by choosing a safe following headway of about two seconds, and let other more altruistic drivers bear the personal risk of keeping congested traffic flow
high by following at headways which place them at personal risk with, as we have discussed, little personal benefit.


What can a lead-vehicle driver do?


            When another driver follows you too closely, you bear the risk of being involved in a rear-end crash without enjoying even the modest time savings of the tailgater.  In many cases you can reduce or avoid this risk by using a variety of techniques to discourage other drivers from following you too closely.  To do so involves frequent use of rear-view mirrors, which is in general a good driving practice.  If a vehicle follows mine too closely on a non-crowded freeway, I simply speed up to get away from it, or slow down to encourage it to pass.  If traffic is congested, increasing your own headway and level of attention is indicated.  Generally, your attention, or mental load assigned to the driving task, is a factor under your control.  Uncrowded freeway driving in which you are neither following nor being followed requires only a fraction of your maximum driving mental load.  It is not necessary, or desirable, to apply this maximum mental load at all times.  Indeed, to attempt to do so may reduce safety by generating fatigue.  Novice drivers tire quickly because the task, in its pre-autonomous state, does consume most mental capacity.  Maximum attention should be invoked when circumstances, as might be identified by glancing in a rear view mirror, merit it.

            One situation in which being tailgated is particularly unacceptable is merging onto a freeway.  Here you may have to change your speed abruptly, and the possibility of having to abort an attempt to merge between two vehicles cannot be excluded.  The tailgater's need to share attention between the merging and following tasks places you at additional risk.  If I am tailgated on a freeway entrance ramp I monitor the tailgater, reduce my speed
substantially well before the freeway, and when a potentially acceptable gap comes along, accelerate rapidly, observing the tailgater recede into the distance in my rear view mirror.  The time lag before the tailgater responds to the acceleration is readily observed, and provides an interesting indication of his or her likely time lag if the lead car had instead braked!

            Another effective way to deter tailgaters is to flash your brake lights; this is most satisfying if it is followed by braking by the following vehicle.  This approach should not be used if the tailgater is also being tailgated; you don't really want to play the role of vehicle 1 in Fig. 12-2!  When in a particularly feisty mood, I have occasionally applied the brakes mildly, followed by acceleration, with most satisfying results.  If the following driver concludes you are crazy, and unsafe to drive close to, then you have achieved your goal of increased safety.  When being tailgated on quiet rural two-lane roads, I have pulled onto the shoulder forcing the tailgater to pass.  It is not uncommon for the tailgater to then proceed at a slower speed than that used previously to tailgate me, providing additional indications that the source of the behavior is habit rather than time savings.  In congested city traffic, slowing down and using a sweeping hand gesture to invite a tailgating driver, who can see you clearly, to back off usually produces the intended result.  You occasionally get satisfying indications that the offending driver receives additional helpful comments from a spouse wise enough to share your views on the subject.

            The risk of being struck in the rear while approaching a red light, or the risk of being struck while stationary waiting at the light, can be influenced by behavior approaching the light.  Rather than proceeding at your prior cruising speed and braking strongly just in front of the stop line, you can reduce crash risk by gently coasting towards the stop line while maintaining only enough pressure on the brake pedal to activate the brake lights; a moving
vehicle is more visible than a stationary one.  Basically, the goal should be to arrive in front of the the stop line just as the light turns green having reduced your speed as little as possible.  This strategy minimizes vehicle wear and saves fuel [Chang et al. 1977; Evans 1979].  Such an approach is not possible if you are already close to the signal when the light turns red.  Then a brisk deceleration followed by a period of stationary waiting is unavoidable.  For the period when your vehicle is the only one waiting, you are at some risk of being struck in the rear.  Although there is not a great deal you can do, it is still worth keeping an eye on the rear view mirror.  If any vehicle approaches in a threatening way, it might be helpful to flash your brake lights off and on because visual sensitivity to dynamic cues is greater than that to static cues, especially in peripheral vision.  Some cases of being struck in the rear fall clearly into the category of events unpreventable by the struck driver.




            We treated rear-end crashes in detail to bring out a number of points which also in other situations.  Experience teaches us to adopt inadequate safety margins.  To avoid crashes over long periods of time requires adopting safety margins that incorporate the possibility of events of much greater rarity than are encountered in everyday driving.  In a driving career, we will encounter a number of extremely rare and unexpected events.  The aim is to behave so that we do not convert such rare events into crashes.  In particular, there are many steps we can adopt to avoid involvement in crashes for which we have no legal culpability.  Below a few comments, based largely on personal experience, are offered on a number of such situations.




            It is not an event of extreme rarity for a vehicle, especially a large truck [Evans and Rothery 1983], to proceed after a traffic light has turned red (to "run the red").  If you are the first vehicle in line, it is prudent to glance left, and then right, before proceeding when the light turns green.  The presence of stationary, or stopping vehicles, in each lane crossing in front of you confirms that it is safe to proceed.  Such increased caution is particularly important if you are able to approach the intersection without stopping just as the light turns green.  In this case, by reaching the center of the roadway in less time than an initially stationary vehicle, you could surprise a driver running the red light.  The increased incidence of side impact crashes to older drivers [Viano et al. 1990] should motivate the older driver to increase vigilance and attention when at intersections, and when turning in the face of oncoming traffic.  It may be that the subjective safety margins learned in youth continue to be applied even as the senses and information processing capacities decline.

            In city streets do not place your faith in other drivers obeying stop signs, or adhering to right-of-way rules.  Many drivers seem to attack stop signs at high speed, and brake at the last moment, even when they can clearly see traffic on the major road.  If the pavement turns out to be unexpectedly slippery, such behavior can be disastrous.  This seems to be another driving behavior rooted in habit, rather than the pursuit of a goal of minimizing trip time.  When I, and many other drivers, travelling on a major road see such a driver heading for our path, we slow down to see what develops.  After the driver on the minor road makes the required legal stop, we regain our prior speed, and proceed with caution.  Thus aggressive drivers delay themselves (as well as the prudent drivers on the major road).  It is another case in which
more dangerous, aggressive driving is rewarded by increased crash risk, increased vehicle wear, increased fuel use, and increased delay!  Pedestrians often similarly increase delays to themselves and others by standing so near the curbside that prudent approaching motorists slow down.




            Many drivers tailgate a vehicle they desire to overtake.  Assuming that such behavior does not influence the lead vehicle, it will generally increase overtaking risk.  The maximum overtaking risk occurs when the overtaken and overtaking vehicle are adjacent, so that the overtaking vehicle cannot quickly return to its original lane.  The time the vehicles are adjacent is reduced if the relative speed between the vehicles when they are level is increased.  If the following vehicle starts very close to the lead vehicle, then, the initial relative speed at the commencement of the overtaking maneuver is zero.  If the following vehicle is further back, its speed can exceed substantially that of the overtaken vehicle by the time the vehicles draw level.  Just prior to drawing level, the following vehicle can abort if there is any reason to do so.  This method may lead to missing some overtaking opportunities, but making others available.

            On relatively deserted freeways I often observe vehicles driving alongside each other on adjacent lanes.  Such behavior increases crash risk for no apparent reason.  If you find yourself alongside another vehicle, especially a long truck, then speed up or slow down.  Be particularly wary of drivers who locate themselves behind your vehicle in positions in which they cannot be seen in your rear-view mirrors.  In general, keep as much space around you as possible; major driving errors, skids, tire blow-outs, and such incidents are far less likely to lead to crashes if your vehicle is not close to any other
vehicles or objects.  If a crash does occur, lots of object-free space surrounding the vehicle is the safest occupant protection environment.  In moments of lax concentration, drivers do drift out of lanes, change lanes without sufficient care, etc.  Such threats cannot always be avoided in high flow traffic, but there is no point incurring them unnecessarily.




            The above comments on methods for reducing risk all involved minimal, or no, delay.  As speed involves a clear trade-off between safety and mobility, it is something on which different rational drivers might make different decisions on different occasions.  Increased average speed can be obtained with the least increase in risk by focusing the speed increases preferentially on the least risky portions of the trip.  Roadway portions with wide shoulders pose less risk than those with close guardrails, and guardrails pose less risk than solid structures such as walls, utility poles or trees.  However, the basics should be kept firmly in mind.  Increased speed increases crash risk, and given that a crash occurs, injury severity increases steeply with speed (Eqns 6-1 to 6-3).




            The simplicity of the earlier claim that average driving leads to average crash risk is complicated by drivers' misconceptions about their own driving ability.


Most drivers think they are better than other drivers


            Svenson [1981] had groups of subjects from the US State of Oregon and from Stockholm, Sweden, rank their own safety and driving skill relative to that of other members of the group, each of whom could be seen performing the same ranking task concurrently in the same room.  The results for both groups of subjects combined are shown in Fig. 12-3.  The interpretation is that the highest bar in the top distribution, for example, indicates that 28% (21 out of 75 subjects) claimed that they drove more safely than 80% of drivers, but less safely than the safest 10% of drivers; one subject (the lowest bar at the extreme left) claimed to be less safe than 90% of the other members of the group.  The bottom distribution, for driving skill, is for 86 subjects (all different from those in the safety study).  Both distributions would be horizontal lines at 10%, as shown by the dashed line, for any objectively measured characteristic; by definition, 10% of drivers are more skilled than (say) the least skilled 30%, but less skilled than the 60% most skilled, and so on for all values.  The dominant feature of both distributions is the subjects' greater likelihood of ranking their safety and skill to be well above the 50th percentile, or median, value; 76% of the drivers considered themselves safer than the driver with median safety, and 65% of drivers considered themselves more skillful than the driver with median skill.


Fig. 12-3 about here


            In an study in which Australian drivers were interviewed at home, Job [1990] finds drivers systematically self-rate their abilities as higher than average.  The overestimation is greater for older than younger drivers, and is greater for males than for females of the same age, the sex differences being
large and systematic over all the seven age categories investigated.  The subjects similarly overrated their abilities, compared to the average, to drive safely after consuming alcohol.   A study in New Zealand [McCormick, Walkey, and Green 1986] finds that up to 80% of drivers rate themselves above average on a number of important characteristics, but also tended to rate themselves below "a very good driver".  Of US students participating in the study of DeJoy [1989], 75% considered themselves to be safer than others, 89% to be more skillful, but only 54% to have lower crash likelihood.  Matthews and Moran [1986] find that young and old Canadian drivers systematically rate their overall driving ability, and their abilities at specific driving tasks, as above average, but, unlike the New Zealand study, slightly greater overconfidence is exhibited by the younger drivers.

            These systematic misjudgments are not necessarily corrected by personal experience.  Preston and Harris [1965] compared 50 drivers whose driving involved them in crashes serious enough to require hospitalization with 50 crash-free drivers matched in relevant variables.  When asked about how skillful they were, the two groups gave similar results, indicating that most drivers, irrespective of crash records, considered themselves to be more skillful than most other drivers.  On the other hand, Matthews and Moran [1986] report that estimates of personal crash risk increase and estimates of driving ability decline with crash involvement, and DeJoy [1989] reports that optimism regarding crash non-involvement increases with experience.  It is possible that driver attitudes and knowledge have changed in the time that has elapsed since the Preston and Harris [1965] study.

            It should be kept in mind is that over 50% of all drivers do in fact have crash rates lower than the average.  While this might appear paradoxical, it is a necessary consequence of the skewed nature of the distribution of crashes discussed in Chapter 1; some drivers have crash rates many times the average
value, whereas no driver can have a rate below zero. (Similarly, more than half of salaried workers earn less than the average salary).  By definition, half of all drivers have crash rates lower, and half higher, than the median crash rate, which is lower than the average crash rate.  The focus of most of the studies reviewed is on medians; the public tends to think of average as meaning median, so the distinction is of little practical importance in those few studies which asked subjects to compare themselves to average drivers.


Why do most drivers think they are better than average?


            Groeger and Brown [1989] extended the investigation of Svenson [1981] by asking subjects to rank themselves not only at safe driving and skillful driving, but also for eight additional abilities or dispositions (gardening, clumsiness, house painting, intelligence, happiness, cooking, competitiveness, and musical ability).  Their finding of a similar general tendency for subjects to consider themselves better than most for many attributes led them to consider the result for the case of driving to be expected on more general principles, and therefore of reduced importance.  Weinstein [1989] reports on a body of research showing that people have a systematic bias in favor of thinking that their personal risk is lower than that of others for a wide range of hazards; he further cites research emphasizing benefits of illusions, such as reduced depression.  Counterbalancing such benefits can be substantial costs in specific instances.  People betting on horse races often think that they know better than the odds-makers.  As in the case of drivers, experience does not often change this view.  The tendency of drivers to think that they are safer than other drivers, whatever its source, tends to move driver behavior in the direction of less prudent safety margins.

            Job [1990] mentions some mechanisms by which experience actually teaches us that we are better than most drivers.  Reports of fatalities rather than inducing fear tend to confirm our perceptions of driving superiority, in that other people are being killed and we have not been.  Most drivers have not been injured, so reports of serious injuries similarly confirm that we drive better than others.  While fear is present in learning to drive, the goal is to eliminate it; good driving is relaxed and confident.  The longer one drives, the greater is the accumulation of evidence that all the really bad things happen to others.  Fuller [1988] writes:

                        From behavioural theory we can predict that every time a driver takes risks, either knowingly or unknowingly, and `gets away with it' without any undesirable consequence, then that behaviour will be reinforced; that is, made more probable in similar circumstances in the future.

            We receive day-to-day feedback that we drive better than others.  We notice when other drivers maintain poor lane position, or turn corners with inappropriate trajectories, or without signalling.  We are generally unaware when others are making similar judgments about us.  As Robert Burns laments, "O wad some Pow'r the giftie gie us -- To see oursels as others see us!"  The mechanism that systematically biases our perceptions about our own driving is somewhat akin to people's perception that they find more coins than they loose, because they are aware of finding but generally unaware of losing.  As Brehmer [1980] points out, experience can be a false teacher.


Implications for the individual driver


            The above findings suggest two points for the individual driver.  First, the majority of drivers you encounter in traffic consider themselves to be
better than most drivers, and thus pose a greater threat to you than they think they do.  Second, it is possible that your self-evaluation of driving skills and safety may be biased in the same direction as that of most drivers.




            Prior to need for continuous decision making in traffic, the driver makes a more long-term decision regarding choice of vehicle; this decision influences safety to some degree in all traffic circumstances.  Let us first consider motorcycles.  Since FARS data have been collected, annual US motorcycle rider fatalities have varied from a high of 5144 in 1980 to a low of 3661 in 1988.  In 1988 there were 4.58 million motorcycles driven an average of 2188 miles per year [Federal Highway Administration 1989, p. 172], for a fatality rate of 36.5 deaths per hundred million miles of motor-cycle travel.  The corresponding data for cars are 26 069 occupant deaths in a fleet of 141.25 million cars driven an average of 10 118 miles per year, giving a fatality rate of rate of 1.82 deaths per hundred million miles.  That is, the motorcycle fatality rate is 20 times that for cars.  While the types of drivers who choose motorcycles are likely to have higher crash risks even when driving cars than the average driver, the major portion of the fatality risk difference is due to the limited occupant protection a motorcycle provides in a crash.  A car is involved in a crash every 10 years or so; most crashes involve only minor, or no, injury.  If one makes the assumption that a motorcycle is about as likely to be involved, per year, in a crash as is a car, then we would expect a motorcyclist to be involved in one crash per 10 years.  However, the outcome to motorcyclists is likely much more severe than to car occupants.  Perhaps any young person contemplating motorcycle travel
should find out how many of his friends have been involved in car crashes, and become familiar with the generally injury-free, or minor injury, outcomes of most such crashes. If the same crashes had occurred to motorcyclists, what would the outcomes have been?  If the decision to choose a motorcycle is made despite the high risk, then certainly a helmet should always be worn.  These reduce fatality risk by 28% (Chapter 9), which, using the use rates in Chapter 10, enables us to calculate that the fatality risk per unit distance of travel for the helmeted motorcyclist is about 17 times what it is for car occupants (typically unbelted), compared to 24 times for the motorcyclist without a helmet.  The intrinsic higher risk to motorcycle riders, with or without helmets, inspired one wit to grossly overstate the situation by quipping, "Buy your son a motorcycle for his last birthday."

            We first discussed motorcycles because they constitute the low end of the distribution of vehicles by size and mass.  Many goals and constraints other than safety enter into the choice of a vehicle.  However, it should be kept in mind that vehicles obey the laws of physics.  Other factors being equal, the greater a vehicle's mass, the lower is the expected injury severity in a crash.  Other factors being equal, the lower a vehicle's center of gravity, the greater is its resistance to rollover.  There is little evidence of important differences in safety between vehicles of the same model year, type, and mass.  Differences in overall injury rates (injuries per registered vehicle) for different models relate mainly to use factors.  This is particularly apparent in the stable observation that station wagons have lower rates than the corresponding four-door version of the same vehicle, which in turn has a lower rate than the corresponding two-door version.  Clearly, the message here is that different types of users choose different types of vehicles, and not that adding a couple of doors increases crashworthiness.

            The more careful the driver is, the greater is the relative importance of vehicle mass.  This is because the proportion of all crashes that involve more than one vehicle increases with declining driver crash risk, and car mass has a larger influence on fatality and injury risk in multiple-vehicle crashes than in single-vehicle crashes.  In the limit, if the risk of single-vehicle crash is reduced to zero so that the only risk is being struck by another vehicle, then the results in Chapter 4 suggest that the occupant in a 1800 kg car might have a fatality risk as much as 75% lower than an occupant in a 900 kg car.

            For occupants who always fasten manual safety belts, motorized or other types of passive belts offer no discernible advantages.  If a motorized two-point belt is used, then the manual lap belt should always be fastened.  Any decision regarding whether to purchase the additional occupant protection provided by airbags is subject to the same economic and other considerations that apply to vehicle choice.  It is worth noting that, in terms of reducing fatality risk, the additional protection the airbag provides to an occupant wearing a three-point belt system is provided equally by choosing a vehicle with mass 80 kg greater (Chapter 9).




            The notion of incentives to decrease the likelihood of involvement in traffic crashes may seem absurd.  After all, involvement in even the most minor crash is an extremely unpleasant experience, involving a ruined day, bureaucratic entanglements, and the loss of hundreds of dollars.  A major crash may involve the loss of tens of thousands of dollars, the loss of transportation, criminal prosecution, jail, a long stay in hospital, permanent
injury, or death.  What penalties beyond these could possibly further motivate drivers to avoid crashes?  Such an analysis fails to include that behavior still tends to be influenced by changes in the perceived cost (monetary and other) of outcomes.  All drivers I have quizzed admit that they would drive more carefully if their vehicles contained high explosives set to detonate on impact; dramatically increasing the harm from a minor crash can clearly reduce the probability of a minor crash.  I suspect that the potential embarrassment of losing my own crash-free record, which I so foolhardily announced at the beginning of this chapter, has further increased my own driving caution.


Does collision insurance increase collisions?


            If increasing the cost of crash involvement tends to reduce crash involvement, then the following corollary seems inescapable; reducing the cost of crash involvement increases crash involvement.  Insurance sharply reduces the cost of involvement in a specific crash by transferring most of the monetary cost away from those directly involved.  Based on the discussion above, it seems almost certain that insurance must increase the number of crashes, although there is no direct evidence of this, let alone any quantification of the presumed effect.  I would certainly feel safer driving amongst drivers required to pay the full property damage cost of any crash in which they were involved rather than the actual situation in which much of the cost is borne by those uninvolved.  

            The non-purchase of collision insurance is a safety measure which does slightly reduce mobility by encouraging more careful driving; the trade-offs here are a reduction in crash risk, a small reduction in mobility, but a large saving in money.  If you judge that your risk is average, and possess
resources that would allow you to pay the vehicle-repair or replacement costs of a crash without unbearable pain, then there is no investment that I can conceive of which has an expected return even approaching the investment decision of changing from purchasing to not purchasing collision insurance. (Perhaps the decision to change from betting on horses to not betting on horses has a comparable pay off). If your risk is below average, then the return is all the greater.  My casual observation is that I and other motorists who carry little insurance beyond that required by law have crash involvement rates well below average, while motorists with abnormally high crash rates tend to shy away from even driving around the block without collision coverage, their premium for which reflects their driving record.  Such patterns are not reflecting the influence of insurance on driver behavior, but of self-knowledge about driver behavior on insurance purchase decisions; those who know they are safer than average tend to prefer to spend their money themselves rather than give it to the insurance company, while those who have been "unlucky" in the past fear more bad luck. 

            The remarks above have focused exclusively on the monetary and safety benefits of not purchasing insurance to cover the repair or replacement of your own vehicle, provided you can absorb such losses yourself.  Discharging obligations to others is a quite different matter, for which the law rightly requires every driver to carry insurance.  Those driving illegally without insurance tend to be high risk drivers, often at the fringes of society.




            While poor assessment of risk can lead to poor driving decisions, it cannot be denied that riskier driving is often indulged in because it is fun; driving has motivations in addition to transportation (Chapter 6).  Driving is
only one of many activities in which pleasure and safety are in conflict.  People in droves have been denying themselves the pleasure of fried bacon or an ice-cream sundae, not to mention a double martini, in the uncertain expectation that they will stay healthier and live longer.  While pursuing the food analogy, it is worth noting that the restaurant industry did not go broke because of changes towards reduced consumption and simpler food; quite the reverse happened.  When people become more concerned about something, they are creative enough to find ways to spend more money on it.  Many people are willing to endure what, to many of us, seems the agony of jogging.  Part of their motivation is to live longer, a motivation which never seems to stigmatize them as being more cowardly than others less afraid of dying.  It seems to me that a convincing case can be made that taking a little more care in traffic is a substantially less unpleasant and and less time consuming way to increase longevity.  Basically, the expected safety costs of risky driving seem out of all proportion to the benefits, especially when considered in the context of the sacrifices so many people are willing to make for more modest life expectancy gains.




            Average driving is perceived by drivers to be safe, a view repeatedly confirmed by copious direct feedback.  Experience teaches drivers to adopt safety margins which, on average, generate about one crash involvement per 10 years.  When crashes do occur there is a tendency to view them as rare unpredictable events outside reasonable human control, and of such a unique nature that nothing like them will ever recur.  They are not normally interpreted as the natural and largely inevitable consequence of average
driving behavior. Indeed, whether they have crashes or not, most drivers consider themselves to be safer than the majority of drivers on the road.  There is no intrinsic reason why a driver has to behave in such a way as to generate one crash per 10 years, or about six over a 60-year driving career; certainly, such a rate could not be contemplated for airline pilots.  The airline pilot does not learn mainly by personal experience, but by a process involving a greater degree of collective learning and intellectualization of the requirements of safe driving.  For road vehicles, we discussed in some detail the example of rear-end crashes.  The driver of a following vehicle can increase headway to values greater than experience has taught to be safe, and do so without incurring appreciable additional delay.  Even though drivers struck in the rear are judged to be not-at-fault, drivers who are followed can and should still exercise control over the behavior of following drivers, and basically refuse to be placed at risk because of the driving habits of others.  To avoid crashes over long periods of time requires adopting safety margins that acknowledge the possibility of events of much greater rarity than are encountered in everyday driving.  In a driving career, we will encounter a number of extremely rare and unexpected events.  The aim is to behave so that we do not convert such rare events into crashes.  

            Increasing the cost of crash involvement tends to reduce the likelihood of crashing; purchasing collision insurance very likely increases crash risk because it insulates drivers from some of the immediate monetary costs of crash involvement.





Brehmer, B.  In one word: not from experience. Acta Psychologica 45:223-241; 1980.

Chang, M-F.; Evans, L.; Herman, R.; Wasielewski, P.  Gasoline consumption in urban traffic. Transportation Research Record 559:25-30; 1977.

DeJoy, D.M.  The optimistic bias and traffic accident risk perception. Accident Analysis and Prevention 21:333-340; 1989.

Evans, L.  Driver behavior effects on fuel consumption in urban driving. Human Factors 21:389-398; 1979.

Evans, L.; Rothery, R.  Influence of vehicle size and performance on intersection capacity. In: Hurdle, V.F.; Hauer, E.; Steuart, G.N., editors. Proceedings of the Eighth International Symposium on Transportation and Traffic Theory. Toronto, Ontario: University of Toronto Press; 1983.

Evans, L.; Wasielewski, P.  Do accident involved drivers exhibit riskier everyday driving behavior?  Accident Analysis & Prevention l4:57-64; 1982.

Federal Highway Administration.  Highway statistics 1988. Washington, DC: US Department of Transportation, document FHWA-PL-89-003; October 1989.

Fuller, R.  Psychological aspects of learning to driver. In: Rothengatter, J.A.; de Bruin, R.A., editors. Road users and traffic safety. Assen/Maastricht, Netherlands: Van Gorcum, p. 527-537; 1988.

Groeger, J.A.; Brown, I.D.  Assessing one's own and others' driving ability: influence of sex, age, and experience. Accident Analysis and Prevention 21:155-168; 1989.

Herman, R.; Montroll, E.W.; Potts, R.B.; Rothery, R.W.  Traffic dynamics: analysis of stability in car following. Operations Research 7:86-106; 1959.

Humes, J.C.  Podium humor: a raconteur's treasury of witty and humorous stories. New York, NY: Harper and Row; 1975.

Job, R.F.S.  The application of learning theory to driving confidence: the effect of age and the impact of random breath testing. Accident Analysis and Prevention 22:97-107; 1990.

Katz, J.  Seductions of crime. New York, NY: Basic Books; 1988.

Matthews, M.L.; Moran, A.R.  Age differences in male drivers' perception of accident risk: the role of perceived driving ability. Accident Analysis and Prevention 18:299-313; 1986.

McCormick, I.A.; Walkey, F.H.; Green, D.E.  Comparative perceptions of driver ability -- a confirmation and expansion. Accident Analysis and Prevention 18:205-208; 1986.

Na?"a?"ta?"nen, R.; Summala, H.  Road-user behavior and traffic accidents. Amsterdam: North Holland; 1976.

National Highway Traffic Safety Administration.  Fatal Accident Reporting System 1988. Washington, DC: US Department of Transportation, document DOT HS 807 507; December 1989.

Preston, C.E.; Harris, S.  Psychology of drivers in traffic accidents. Journal of Applied Psychology 49:284-288; 1965.

Sivak, M.; Soler, J.; Tra?"nkle, U.; Spagnhol, J.M.  Cross-cultural differences in driver risk-perception. Accident Analysis and Prevention 21:355-362; 1989.

Summala, H.  Modeling driver behavior: a pessimistic prediction?  In: Evans, L.; Schwing, R.C., editors. Human behavior and traffic safety. New York, NY: Plenum Press, p. 43-61; 1985.

Summala, H.  Young driver accidents: risk taking or failure of skills? Alcohol Drugs and Driving 3(3-4):79-91; 1987.

Svenson, O.  Are we less risky and more skillful than our fellow drivers? Acta Psychologica 47:143-148; 1981.

Viano, D.C.; Culver, C.C.; Evans, L.; Frick, M.C; Scott, R.  Involvement of older drivers in multivehicle side-impact crashes. Accident Analysis and Prevention 22:177-199; 1990.

Wasielewski, P.  An integral equation for the semi-Poisson headway distribution model. Transportation Science  8:237-247; 1974.

Wasielewski, P.  Car following headways on freeways interpreted by the semi-Poisson headway distribution model. Transportation Science 13:36-55; 1979.

Weinstein, N.D.  Optimistic biases about personal risks. Science 246:1232-1233; 1989.