Participant Exercise #4
The Dramatic Failure of U.S. Safety Policy
Engineering is important, public policy is crucial.
The author is a researcher, writer, and lecturer on traffic safety and a member of the National Academy of Engineering.
The dramatic failure of U.S. traffic safety policy seems to be one of our best kept national secrets. Policies aimed at reducing the toll of 42,000 Americans killed annually rarely make news. When they do, it is usually because the very institutions responsible for the failure announce what impressive progress is being made. To obtain a realistic sense of our own progress we need to compare it to progress in other countries.
My 2004 book Traffic Safety (1) presents such a comparison. The main findings of that comparison and my interpretation of the findings are summarized here. Numerical details of the calculations, identification of data sources, references and detailed documentation are in the book. The material in my book led to comments in a recent TR News article by Brian O’Neill (2). Here I respond to those comments.
Prior to the mid 1960s the U.S. had the safest traffic in the world, whether measured by deaths per registered vehicle, or deaths for the same distance of travel. By 2002, in terms of deaths per registered vehicle, the U.S. had dropped from first place into sixteenth place, behind Australia, Austria, Canada, Denmark, Finland, Germany, Great Britain, Iceland, Japan, Luxembourg, the Netherlands, New Zealand, Norway, Sweden, and Switzerland.
Comparing fatalities in 2002 to fatalities in 1979
The decline in U.S. safety relative to other countries is explored by comparing changes in specific U.S. fatality rates with the changes in the same rates in other countries. Three traffic fatality rates are examined:
1. Fatalities per year (the raw fatality rate).
2. Fatalities per thousand registered vehicles (the vehicle rate).
3. Fatalities per 100 million vehicle miles of travel (the distance rate).
Three countries, Great Britain, Canada, and Australia are selected for the comparisons. These three countries are chosen because they have much in common with the U.S. in terms of language, beliefs, and traditions. Performance is compared over the 23-year period from 1979 to 2002. It was in the late 1970s/early 1980s that the safety policies of the U.S. and other countries began to diverge. The results are not all that different if the initial and final times were a few years earlier or later, or if various different comparison countries were chosen.
Fatalities per year comparisons
Figure 1 shows the change in the simplest measure of safety performance, the raw total traffic deaths per year. While fatalities in the 23-year period declined in the U.S. by 16.2%, declines of 46.0%, 49.9%, and 51.1% occurred in Britain, Canada, and Australia. If U.S. fatalities had declined by the same 46.0% as in Britain, U.S. fatalities would have been 27,598 instead of the 42,815 reported. By matching the British decline, 15,217 fewer Americans would have been killed in 2002. The corresponding fatality reductions for matching Canadian and Australian performance are 17,229 and 17,837.
FIGURE 1 Traffic fatalities per year in the U.S. and in three comparison countries relative their number in 1979.
Because of the disparate numbers of vehicles in different countries, raw fatality rates can be compared effectively only by renormalizing in some way as in Figure 1. However, rates such as fatalities per thousand registered vehicles can be plotted without the need to select a reference year, as shown in Figure 2. Prior to the late 1970s the comparison countries, in common with all countries, had rates higher than the U.S. The U.S. rate shows no indication of a drop in response to the requirement that all 1968 and later models must satisfy a number of Federal Motor Vehicle Safety Standards (FMVSS). The only notable downward spike, in 1974, is unrelated to vehicles, but reflects various behavior changes stimulated by the 1973 oil embargo, especially travel speed reductions in response to changes in speed limits.
FIGURE 2 Traffic fatalities per thousand registered vehicles in the U.S. and in the three comparison countries.
While the U.S. vehicle rate declined by 46.2%, the rates in Britain, Canada, and Australia declined by 67.1%, 63.5%, and 71.9%. If the U.S. rate had declined by these same percents, U.S. fatalities in 2002 would have been lower by 16,606, 13,718 and 20,429.
Fatalities for the same travel distance comparisons
The best estimates of distance of vehicle travel are for Great Britain, based on observations supported by the Department of Transport at fifty sites. Reliable estimates over a long period are not available for most countries, so the comparison is confined to Great Britain. Figure 3 shows that while the distance rate in Britain was (as in all countries) previously higher than in the U.S., in 2002 it was lower. By matching the British decline, 15,670 fewer Americans would have been killed in 2002.
FIGURE 3 Traffic fatalities per 100 million miles of vehicle travel in the United States and in Great Britain.
Additional Americans killed over the 1979-2002 period
The estimates above are all for reductions in the number of Americans who would have been killed in 2002 if the various rates in the U.S. had declined by the same percents as occurred in the comparison countries. By calculating corresponding differences for all the intervening years it is straightforward to estimate the cumulative additional American lives lost. The average for all rates and the three countries is 214,286. So, if U.S. safety progress had kept pace with that in the comparison countries, about 200,000 fewer Americans would have died on our roads in the 23-year period.
Different rates – same picture
The estimated number of fatalities prevented by the U.S. performance matching that in the comparison countries is relatively independent of which rate is chosen. Taking Great Britain to illustrate, the number of American lives saved in 2002 if the U.S. had matched the British percent decline in raw fatalities is 15,217 compared to 16.605 from the vehicle rate, or 15,670 from the distance rate.
If the percent growth in vehicles and growth in travel were identical over the 23 year period in both countries, then each calculation would produce an identical estimate. The somewhat larger difference computed using the vehicle rate reflects simply that percent vehicle growth in Britain was somewhat higher than in the U.S., as is expected since Britain started at lower levels of vehicle ownership. The slightly larger difference computed from the distance rate reflects slightly greater increase in total distance traveled in Britain compared to in the U.S..
The crucial point is that we are examining percent changes. Large fairly stable differences between the countries should not much affect percent changes in time, as such factors similarly influence the 1979 and 2002 fatalities. For example, Britain is more urban than the U.S. and has more (as a percent) pedestrian fatalities– but such factors do not differentially change all that much between 1979 and 2000.
Comments on Brian O’Neill’s Improving U.S. Highway Safety
In his recent article in this publication (2), Brian O’Neill correctly points to progress in U.S. safety. Indeed, if the other countries are deleted from Figures 1, 2 and 3 the declines in U.S. rates can be, and often are, interpreted as impressive progress.
O’Neill identifies great variation among U.S. states, and shows that some states had lower rates than those in the comparison countries. The central point made here is that the reductions in the rates in the U.S. are so much less than in other countries. It is not just that the U.S. rates in all cases in Figures 2 and 3 are higher in 2002 than in the comparison countries, important thought this is, but that the U.S. rates were formerly lower.
Certainly, some U.S. states have lower rates than the aggregate national rates of comparison countries. Within the comparison countries, regions, provinces, or states also have rates that are lower than their national average. However, what is central to examining progress is not what the values are in 2002, but how their percent changes since 1979 compare to changes in the comparison countries.
Given that percent changes over time in raw fatalities, in the vehicle rate, and in the distance rate are similar, it is easiest to compare changes in an individual U.S. state’s performance by simply comparing raw fatalities. This fails to capture that population growth was higher in some regions (the South) than in others, which will contribute to increased fatalities, but it avoids the uncertainties from crashes by out-of-state vehicles, unreliable estimates of travel, etc. Comparing 2002 and 1979 fatalities for all 50 states and DC shows that only one state enjoyed a decline of more than 50%; VT fatalities declined from 159 in 1979 to 78 in 2002 – a 50.9% drop. Even though this is in part a statistical outlier attributable to small numbers, the decline is still less than the 51.1% decline in Australia. The state with the second highest decline is MA, with a 49.9% decline. All other states had smaller declines than any of the comparison countries.
States in the East and Midwest tended to have the largest declines, while those in the South the lowest declines (14 had increases). DC had an increase of 6.0%, although the vehicle rate is low for all years because DC is urban. The wide variations among the states is in part, as O’Neill states, due to differing policies in the different states relating to speed, alcohol, and mandatory use of occupant protection devices. This supports my core thesis that it is public policy aimed at driver behavior that really makes a difference. The vehicles in all sates are subject to identical safety standards. Comparing changes over time in the various states shows that the failure of U.S. safety policy largely applied throughout the nation, even if not uniformly.
Search for an explanation
While straightforward analyses of publicly available data show some 200,000 additional U.S. fatalities over a 23-year period, it is not possible to explain why this occurred by a correspondingly simple analysis. One specific factor, safety belt use, does account for a substantial portion of the difference. If the U.S. had introduced belt-wearing laws on the same schedule as Canada, 95,000 fewer Americans would have died between 1979 and 2002. This still begs the question “Why did the U.S. not adopt this effective intervention earlier?” and leaves unanswered the question “What accounts for the other 100,000 plus additional deaths?”
Such large and robust effects likely reflect fundamental differences in philosophy and approaches to safety. In democracies, elected legislators normally base policy on inputs from many sources, including technical experts. In the U.S. this is so for most subjects, but not for traffic safety. Instead, legislators, most of whom are lawyers, receive their inputs on traffic safety almost exclusively from other lawyers.
U.S. traffic safety policy began to diverge sharply from that in the rest of the world when activist lawyer Ralph Nader convinced the U.S. media, government, and public that the problem was unsafe and defective vehicles. Even if claims in his 1965 book Unsafe at Any Speed had merit, their effect was to focus attention on a dozen or so deaths occurring over a number of years, while ignoring the 50,000 deaths occurring annually at that time.
Figure 4 shows Nader demonstrating an airbag simulator “safely” deploying into the face of an unbelted three-year-old girl. When the photograph was taken, on 5 July 1977 at a Washington, DC press conference convened to support airbags, the technical community had been long aware that deploying airbags posed risks to occupants, particularly children. In 1972 a U.S. study titled “Airbag effects on the out-of-position child” was published and presented at a Society of Automotive Engineers meeting in Detroit. It used child dummies and baboons of size and weight similar to children to investigate if deploying airbags posed injury threats to children. It found that they did. In 1974 a Swedish study titled “Possible effects of air bag inflation on a standing child” was presented at technical meetings. It used pigs to simulate what would happen to an out-of-position child who leaned against the air bag as it deployed. In eight of the twenty-four trials the airbag deployment killed the pig.
The person critically influencing U.S. safety policy was a lawyer untrained, uninformed, uninvolved, and uninterested in technical matters. Press conferences, not technical meetings, were Nader’s milieu. Those influencing safety policy in other countries were mostly technically trained, and attended technical meetings such as those at which the information on harm from airbags was communicated, as well as information on interventions that really reduced casualties.
The irrelevance of technical knowledge to safety policy was taken to new heights when Nader’s protégé, lawyer Joan Claybrook, became NHTSA Administrator in 1977. A NHTSA official is quoted as saying, “Joan came to NHTSA with a mission and that mission was air bags.” More specifically, the mission was an airbag mandate, a government requirement that airbags be installed in vehicles. Technical information, such as the effectiveness of airbags, never impeded that goal. Indeed, Claybrook’s claims of effectiveness and estimates of lives saved were not simply exaggerated, technically indefensible, wrong, or plucked out of thin air. Relative to what was known, and published in the literature of the time, they were absurd.
In 1970 belt-wearing became required by law in the Australian state of Victoria. Evaluations were so positive that most of the world followed by passing belt-wearing laws. In the U.S., the agency responsible for safety rejected this proven effective countermeasure.
Not only did Claybrook not support mandatory wearing laws, she did not even accept that belts were effective when used. In a November 1983 television interview she says of airbags:
“They’re much better than seat belts, according to the government’s most recent data.”
Large public health problem in which vehicle factors play a minor role
The major harm the airbag mandate caused is not that it has killed over two hundred people in crashes they would otherwise have survived with no injuries or minor injuries, nor its staggering monetary cost. The cost of airbags on U.S. roads, $54 billion dollars, exceeds the gross domestic product of more than half of the countries in the United Nations. Its real cost to the U.S. was that it encouraged a false belief that safety was achievable by gadgets, rather than understanding that it is a an enormous public health problem requiring a balanced mix of interventions grounded in the scientific understanding of the relative effectiveness of various countermeasures. It fostered the belief that the best approach to improving safety was to make crashes more survivable, ignoring that the only way to sharply reduce harm in traffic (once occupant protection devices are worn) is to reduce the number of crashes. The priority of avoiding crashes rather than surviving them is at the core of the extraordinary success of aviation safety, in which the U.S. leads the world. In 2002 there were zero fatalities in U.S. scheduled airline travel.
The modest driver behavior change of traveling 2 mph slower produces a larger safety benefit than airbags. Public policy can achieve such changes. However, scientifically determined relationships between travel speed and safety play a lesser role in our legislative process than in those of the comparison countries. The history of the airbag mandate, in which ideological zeal triumphed over technical knowledge, contains lessons that have the potential to eventually benefit U.S. safety policy. We must not forget George Santayana’s celebrated aphorism: Those who do not remember the past are condemned to relive it.
Sadly, it is not history. The U.S. continues to obsessively focus on vehicle factors. We ignore that Canada cut its traffic deaths in half even though Canadian and U.S. vehicles are not all that different. The architects of policies that have killed over 200,000 Americans are routinely referred to in U.S. media as Safety Advocates. The most important elected official with safety responsibilities, Senator John McCain, is unreceptive to scientific information, but his door is open to Claybrook, as indicated in the newspaper headline “Republican senator, auto-safety advocate form unlikely alliance.” The one cause for optimism is that once the U.S. does recognize it has a problem, it moves with a speed and energy unequalled in other countries. By adopting new thinking it is possible to cut our fatalities in half with policies that the public would welcome (1, p. 412).
FIGURE 4 Ralph Nader demonstrates an air bag simulator on an unbelted 3-year-old girl at a Washington DC press conference supporting airbags in 1977, long after the technical community had documented concerns about risks to children from deploying airbags.
[credit: AP/Wide World Photos – permission will be needed for publication – likely cost $150]]
1. Evans L. Traffic Safety. Bloomfield Hills, MI: Science Serving Society; 2004
2. O’Neill B. Improving U.S. Highway Safety. TR News