Are older drivers more likely to cause accidents?

ScienceDaily has an article about some new RAND research that shows the usual RAND quality in their papers - i.e. next to none.

Drivers 65 and older are just one-third as likely as drivers 15 to 24 to cause auto accidents, and not much more likely than drivers 25 to 64 to cause accidents, according to a new RAND Corporation study.

Does that RAND study show that? Well, RAND and the authors of the studies says so, but when reading the RAND paper about the study, it becomes clear that this is a highly doubtful claim.

The paper can be found Regulating Older Drivers - Are New Policies Needed? by David S. Loughran, Seth A. Seabury, and Laura Zakaras (.pdf)

In the introduction, the authors write

Although it has been scientifically established that physical and cognitive degeneration at older ages compromises driving ability, it is not clear just how much riskier older drivers are than other drivers. Most published research shows that accidents per mile driven increase when drivers are in their fifties and, by the time they reach their eighties, accidents per mile driven are almost as high as they are for the youngest drivers (see, for example, Li, Braver, and Chen, 2003). As we describe later, however, this measure of risk can be misleading.

Our research departs from previous studies that rely on this measure. Instead, we use an innovative statistical method to estimate the likelihood that older drivers will cause an accident relative to the likelihood that other drivers will cause an accident, controlling for vehicle miles driven. We will refer to this statistic throughout this paper as the relative riskiness of older drivers. Levitt and Porter (2001) first devised and employed our statistical method in their study of the relative riskiness of drunk drivers. Based on our findings, we make a number of policy recommendations aimed at stemming the rise in traffic-related injuries and deaths that are expected as the average age of the driving population increases.

I am probably not alone in thinking that a study using "an innovative statistical method", that reaches a different conclusion from all studies, raises some major warning signs.

The traditional method of looking at accidents per mile driven takes into account that older people drive less than other people, and looks at the relative risk created by the older drivers.

The authors of the paper argue that older drivers tend to be more risk adverse in their driving (avoiding high speed zones, driving in the dark etc.), so they feel that a look of the dangers imposed by giving older driver easy access to renew their driver's license should reflect this.

This sounds somewhat reasonable, until one realizes that even with the risk adverse behaviour, people aged above 75 is more dangerous when they are on the road than any other group of people, except people aged below 25 (see figure 2.1 in the paper), who certainly are not risk adverse. In other words, the old people who renew their diver's license impose a disproportional number of the accidents. Making it harder for older people to renew their driver's license would have a disproportional positive influence on the number of road accidents (as would making it harder for people under 25).

There are also some other problems with the study, since it only looks at the age of drivers involved in accidents resulting in fatalities. This is certainly one parameter of risk, but there are also the accidents resulting in non-fatal injuries (something the authors acknowledge), accidents with only material damage, and accidents caused by other peoples' risky behaviour.

Once the authors have looked at the age of people involved in fatal accidents, they find that more accidents are caused by younger people than older people (two groups that are not precisely defined in the study).

In summary, we find that older drivers are only slightly likelier than other drivers to cause an accident but are considerably likelier to be killed in one. Younger drivers, on the other hand, are considerably likelier than other drivers to cause a crash, drive much more frequently than older drivers, and are less susceptible to fatal injuries than older drivers are. These findings do not mean that driving skills do not, in fact, deteriorate with age as a result of worsening mental
and physical impairments. Instead, our evidence suggests that older drivers adjust their behavior in light of these worsening impairments. Many older drivers cease to drive altogether; many others reduce the miles they drive and avoid the most dangerous driving conditions. Because they are aware of their own limitations and adjust their driving patterns in response, older drivers pose only a slightly increased risk to other drivers. The main danger they pose on the road is not to others but to themselves.

Again this sounds reasonable until you think a little harder about it. Yes, we know that younger people causes more accidents than older people (here I define younger people as below 25), but that's because of very different reasons. Younger people are usually involved in accidents because of inexperience and/or risky behaviour (like drunk driving), something that cannot be tested for when renewing a driver's license. Old people on the other hand, are involved in more accidents due to having their driving ability impaired. Something that could easily be tested for when renewing a driver's license.
In other words, just because a different group of drivers are more risky, doesn't mean that the problem of old people driving past their ability shouldn't be addressed. And the fact that they are more at risk to themselves doesn't mean that they are not unqualified for renewing their driver's licenses. The driver's license is a license to drive a potentially deadly vehicle, and should only be given to people who are actually able to drive it without posing a danger to themselves or others.

I agree with the study that the relative risk caused by younger people should certainly be addressed, but this is not a zero-sum game. Making it harder to get a driver's license for younger people (or easier to loose it) doesn't mean that the states can't at the same time reduce the risks caused by older people driving. Both things could (and should) be done to reduce the number of accidents.

RAND probably publishes some well-researched papers some times, but I have yet to come across any. I am quite disappointed at ScienceDaily for just passing on such bad stuff.