Does CAFE Kill? Oral Remarks By Dr. Leonard Evans
“Does CAFE Kill?”
Transcript of Oral Remarks by
Dr. Leonard Evans
A Discussion of the Report on CAFE
of the National Research Council of the National Academies of Sciences
Competitive Enterprise Institute
January 17, 2002
[The Powerpoint presentation that accompanies this discussion can be found here]
I would like to thank CEI for supporting my trip here, which enabled me to participate at yesterday’s meeting of the Transportation Research Board to address this report from a panel of the National Academies. I could mention, tongue in cheek, that three factors seemed to have rendered me ineligible for inclusion on that panel. First, I have published more of the peer-reviewed literature on the effect of vehicle size on safety than anyone else in the world, and probably more than the entirety of all the other researchers in the world. Second, I have a doctorate in physics, and many of the questions seemed to involve the need for somebody with that particular technical background. Third, I myself am a member of the National Academies. (Only two of the panel’s 13 members were also members of the National Academies; both of them supported the panel’s majority report.)
Does CAFE kill? The majority of the NAS panel found that it does indeed do so, through its downsizing effect on cars. Specifically, they concluded that CAFE has contributed to between 1,300 and 2,600 traffic deaths a year. In my view these findings were sensible. But there was a dissent by two members, published in the appendix of the report, which alleged that this finding was premised on two logical fallacies. One of the dissenters was David Green, an environmental zealot. The other was an automotive journalist, Mary Ann Keller.
The dissenters cite two logical fallacies. I’m going to address the second one first. The dissenters correctly state something that is at the very core of traffic safety, but they then proceed to make quite indefensible inferences from it. They say that the majority’s findings involve many confounding factors, most importantly that of driver behavior, and that this is a far more important determinant of crash occurrences than vehicle mass. That is at the core of traffic safety. And it is correct.
The main body of the report contains data showing that during the period when CAFE was in place, our fatality rate continued to decline. As the report correctly points out, this has led some people to erroneously conclude that this continued improvement in the fatality rate implies that CAFE did not increase deaths. The dissent is less clear on whether it too rejects this view of CAFE.
I want to extend this comparison a little bit by contrasting what happened in our nation to what happened in some nations that did not have CAFE. In Sweden, for example, the fatality rate dropped substantially faster than in the United States. In the mid 1960s, the United States had the lowest fatality rate in the world by far. We are now in 13th place and continuing to decline. Similarly, we do poorly in comparison to Canada, Australia, and Great Britain. If our fatalities had declined along the same pattern as occurred in in Sweden, Britain, Canada, and Australia, then over the last two decades 200,000 fewer Americans would have been killed.
The dissenters are correct about driver behavior as a factor that overwhelms other factors in its magnitude. In fact, driver behavior is overwhelmingly the dominant factor when it comes to the total number of deaths in the rolls. If we had matched those other countries, our current rate of traffic deaths would be about 15,000 fewer fatalities per year.
All of the vehicle factors are relatively unimportant compared to the overwhelming influence of driver behavior. And I really do stress relatively, because these other factors still are associated with the deaths of many thousands of people on our roads. Among the vehicle factors, mass is overwhelmingly the largest.
CAFE unquestionably leads to lighter vehicles, because fuel is consumed in ways that are intimately related to mass. The energy required to accelerate a body from rest to 30 mph is directly proportional to the mass of the body. So the heavier the vehicle, the more fuel you must use, other things being equal.
Now, there’s a technical question which leads to some people to think that the size-safety issue is insoluble: can the effects of important smaller factors be reliably determined in the face of this really enormous factor of driver behavior? The answer to this is a clear yes, in many cases. This what we do in science all the time. We try to isolate the influence of one important factor, even though there are other factors that may have an even larger influence.
The technique that the dissenters keep resorting to is regression analysis, as if this is what one rushes to do in the first instance. It is not. Regression analysis is a technique of desperation, a last resort. It is what one uses when all better scientific methods are somehow unavailable and you can’t come up with some proper scientific way to look at it.
Here are data from a paper of mine that appeared in the July issue of the American Journal of Public Health. It shows the risk of death in a smaller car in a two-car collision, compared to the risk of death in the larger car, related to the mass of the larger car divided by the mass of the smaller car. The rates at which drivers crash are strongly influenced by driver behavior, but the relative risk to each driver when a crash occurs is not affected in any obvious way by driver behavior. This relative risk is enormously influenced by the relative masses of the involved cars as shown in the plotted relationship. It says: if two cars crash into each other, and one of them is twice as heavy as the other, then the driver in the lighter car is about twelve times as likely to be killed, as the driver in the heavier car.
If all drivers drove in a crazier way, we’d get roughly twice as much data, but the relationship would look the same. If drivers were more careful, we’d have fewer data, but the relationship would still look the same. So this relationship is essentially independent of driver behavior. This relationship is so strong, and so clear, that you can see the effect if you take a collision between two otherwise identical cars, and then add the mass of a passenger to one of those cars. Let me make it personal: I have a passenger in my car, you have an identical car but without an added passenger, and we crash head on. The result--I am 14.5 percent less likely to be killed than you are, just because my car is heavier due to the weight of that added passenger.
The presence is this added passenger protects me, but it actually increases the risk to the person in the car that I collide with. The graph you see does not disentangle these two effects. It simply shows that when we divide one by the other, the difference is 14.5 percent. I am 14.5 percent safer than you are, but we don’t know just how this is split between your increased risk and my decreased risk.
In order to determine that, we need another relationship, and the other relationship is one that we will address in terms of the other alleged fallacy that this dissent claimed. They contend that it is a fallacy that reducing the mass of all vehicles will increase risks in collisions between vehicles. They are wrong; the data shows very clearly that this is no fallacy.
Here we see results from five different data sets. Look what happens when two cars of the same mass crash into each other. What we see is that if two very light cars crash into each other, the risk to each driver is approximately twice what it is when two heavy cars crash into each other. It is plotted against weight, which is not necessarily identical to mass but which, practically speaking, is usually representative of it. But what really causes this difference in risk is not the fact that the cars are heavier, but that heavy cars are also larger. This is intrinsic. A big car is heavier, and heavy cars are bigger. The extra size provides additional crush material in front of the driver, and additional occupant space. This provides more time for the driver to come to rest, and that reduces risk.
So if a completely homogeneous fleet of very heavy cars replaced by a completely homogeneous fleet of very light cars, this would roughly double the number of driver injuries and deaths. There is a very fine safety researcher in Europe who addressed question theoretically by considering the way in which the components of vehicles crush in a crash. He came up with this relationship by plugging in the different size components that go into the different-sized vehicles.
So you see there’s a great coherence here. Five data sets, one theoretical relationship – all say that big, heavier cars crashing into big, heavier cars are safer than small, lighter, more fuel-efficient cars crashing into other small, light, fuel-efficient cars.
We are now in a position to pull apart this 14.5 percent number when I, with my extra passenger, crash into you. We both have identical cars, but I have my passenger and you don’t. My risk is down by 7.5 percent. Your risk goes up by 8.1 percent. And so in this case there is a net social loss in that crash. The extra passenger adds to my protection but decreases yours. On balance, social risk has gone up a little bit.
But that is a pure mass effect; that is, the effect of just adding mass to one of the cars. Suppose that instead of carrying a passenger, I swapped my present car for another car that is heavier by that passenger’s weight. Suppose too that my new car is also larger by the amount that you’d expect of a car that is 165 pounds heavier. When this happens, the relationship changes. Now my risk is lower by 11.5 percent and your risk is higher by 6.9 percent. Overall, society doesn’t lose; it gains. The reason that your risk is doesn’t increase as much as it did before is that you’re now crashing into a slightly larger vehicle with slightly more crushable structure. So you also do better hitting a slightly larger car of the same mass than a smaller one of the same mass. On balance, there is a societal gain.
So when I purchase this larger car, society is on average better off. Now there are many complicated questions about equity, but those issues are in a different arena. But in terms of just the total number of lives, when I purchase a larger car, there is a reduction of risk. I’m safer, and so is society overall.
The dissent contended that the relationships between vehicle weight and safety are complicated and not measurable with any reasonable degree of certainty at present. Nothing could possibly be further from the truth in the area of traffic safety. Vehicle mass and size have a better established, and better quantified, influence on safety than any other factor that has ever been studied, including speed limits, safety belts, belt-wearing laws. I am unaware of any safety relationship that has not attracted constituencies of enthusiastic deniers.
I picked these three for a reason; they’re all situations in which I’ve been involved in some fairly major controversies. For example, some have alleged that speed limits have no effect on safety. There is an academic at the University of Manitoba in Canada who made an enormous splash claiming that if you’re in a crash, wearing a safety belt increases your risk of being killed. This Canadian professor and I were even on a talk radio show with G. Gordon Liddy here in D.C. Similarly, Professor John Adams at University College in London has made a worldwide splash claiming that belt-wearing laws actually increase casualties.
In all of these situations, the people making these claims had far stronger cases than the dissenters to the TRB CAFE report. The evidence to show that these dissenters are wrong is vastly stronger than in the above controversies. Vehicle size effects have been consistently observed using many data sets, in many countries, with many analyses and many methods. This is how you build up a really solid technical understanding in difficult arenas. When you get the same answer no matter how you look at it, you get a more and more confident understanding.
Of course, just as with all aspects of traffic safety, much more is not known than is known. But you can’t let the existence of these unknowns refute everything that you don’t like, which is really what this dissent is doing.
We know beyond a reasonable doubt that, in two-vehicle crashes, if a car is heavier, then it provides more protection to its occupant, and more risk to the people in vehicles it hits. Whether the net effect increases or decreases risks depends on the specific vehicles. I’ve given examples where they’re equal. It gets complicated when they are not, and for the vehicle fleet the net effect is difficult to estimate. But what we can say with very good confidence is that making the entire fleet heavier will, at the very worst, generate a small increase in risk. The far more likely result is a decrease in risk, but there is some uncertainty here and at the very most, it could be a very small increase in risk.
So for two-vehicle crashes we see a clear protective aspect for mass in collision between similar vehicles. When the vehicles are different, the outcome is somewhat complicated. But when we go to single-vehicle crashes, in which roughly half of all occupant fatalities occur, this complexity disappears. In such crashes there is no second vehicle to be concerned about. Greater size gives the occupant more protective space. Greater weight increasingly deforms the obstacle that is hit, thereby reducing the crash forces that act on the car’s occupant. The result here is overwhelmingly clear--greater vehicle mass reduces deaths and injury severity, period. Greater size reduces severity because of the greater crush distance. Larger sizes are associated with larger track width, that is, with larger spaces between the left and right wheels. This reduces rollover risk. So for the single-vehicle crash, all the factors point unambiguously in the same direction and produce substantial effects.
Now the vital thing to remember in this entire debate is that these clear-cut single-vehicle effects overwhelm any possible uncertainties in multiple vehicle cases. We can conclude, beyond any reasonable doubt, that when weight is reduced, as it must be under CAFE, we will increase casualties. Even the proverbial crash into a brick wall has weight as a very important factor. The brick wall is supposedly nondeformable, and so car weight, independent of car mass, is said to be unimportant. But look at this photo of a car that actually crashed part way through a brick wall; its driver is standing there on the side, uninjured, chatting to the policeman. Imagine if this vehicle had been a small, light car, instead of the large vehicle that it is. In that case the wall would have been just be fine, undamaged, but not so for the driver, who might have ended up in the hospital or in the morgue. He would certainly not be having a congenial discussion with the state trooper.
The conclusion is that CAFE has caused, and is causing, increased deaths. Higher CAFE standards will generate additional deaths. Now, of course, this does not necessarily mean that we should not have higher CAFE. We all support innumerable policies that result in deaths. As a citizen, I certainly support the policy of not having a hospital at the end of every street, even though I’ve reached that stage in life where one of the lives lost because of that policy might be my own. We know that not having a hospital at the end of every street will kill people, but we still approve of this policy because we believe the funds could be spent better elsewhere.
The role of the technical community ought to be to tell the political process that not having a hospital does kill people. It should also be to tell us that having policies that lead to lighter vehicles, as CAFE certainly does, will increase casualties. It’s up to the political process, then, to handle it as is politically appropriate. But from a technical point of view, there is no fuzziness or ambiguity of any sort whatsoever regarding CAFE. CAFE kills, and higher CAFE standards will kill even more.