Swedish versus USA Traffic Safety:
What Comparing Fatalities Tells Us
Leonard Evans, March 2007
Objective. To show that while the USA and Sweden differ greatly in safety performance, the distribution of traffic fatalities by age and sex in each country is similar.
Methods. 2005 fatality data for the US and Sweden are analyzed.
Results. In 2005 Sweden recorded its lowest number of fatalities since the 1940s, while the US recorded its highest number in 15 years. The Swedish total is 66% below the highest number recorded in Sweden (in 1966), while the US total is only 20% below the highest number recorded in the US (in 1972). If the US total had dropped from its high by the same percent as the Swedish total did, in 2005 the US would have suffered 18,293 traffic deaths, 25,150 fewer than the actual total of 43,443. Despite this enormous difference in safety improvement between the US and Sweden, the distributions of US and Swedish fatalities by age and sex are found to be similar.
Conclusions. There are robust effects in traffic safety, including the overrepresentation of younger drivers, the overrepresentation of males, and the risk of harm when vehicles crash. Harm in traffic can be reduced by large amounts, but only by policies aimed at the attainable goal of reducing the overall risk of crashing, not the unattainable goals of producing vehicles that are safe to crash, getting young people to behave like older people, or getting males to behave like females.
The following are two well established traffic safety findings:
1) Young drivers are overrepresented in crashes
2) Males are overrepresented in traffic deaths
These findings are supported by an enormous body of literature accumulated in over 70 years of scientific study (Evans, 2004). They approach, but do not reach, the level of laws that play so pivotal a role in physical science.
The purpose of the present paper is to show that these findings apply to the mix of fatalities in countries with very different safety performance. Data from the USA (National Highway Traffic Safety Administration [NHTSA], 2007) and Sweden (European Conference of Ministers of Transport (2006), Swedish Road Administration, 2007) are used.
In 2005 the US recorded 43,443 traffic deaths, the highest total in 15 years, including 4,881 pedestrian deaths, the highest total in four years. Sweden recorded 440 total traffic deaths in 2005 – the lowest total since the 1940s. Among US states with smaller human populations than Sweden, 23 recorded more deaths than Sweden. Of these, 11 recorded more than twice as many deaths as Sweden, and one (NC with 1,534) more than three times as many.
The data presented here show that despite such enormous differences in overall safety, the mix by sex and age of who is killed is quite similar for both countries.
Distributions by age and sex
Figure 1 shows normalized distributions according to victim age and sex based on the 43,443 traffic deaths in the US and the 440 in Sweden (the male and female fatalities do not add to exactly these values because a few cases do not have age or sex coded). It is apparent that the two distributions exhibit the same general characteristics.
Figure 2 shows the normalized distributions for driver deaths. These figures show that in Sweden, as in the US, young drivers play a central role.
Figure 1. Normalized distributions of all fatalities in Sweden and the US in 2005. For each country the area under the male plot plus the area under the female plot = 100%.
Figure 2. Normalized distributions of fatalities to drivers of any motorized vehicles in Sweden and the US in 2005.
Male to female ratios
The World Health Organization (2003) estimates that of 1,194,115 people killed in 2001 in traffic worldwide, 848,234 were male and 345,881 female, giving a male-to-female ratio of 2.34 to 1. A large proportion of the world's traffic deaths occur in countries in which females are denied rights available to males, so one might think the large sex discrepancy reflects mainly social custom and law. In an extreme case, societies that do not issue driver licenses to women will expect few female driver deaths.
Figure 3. shows a 2005 male to female ratio for all US traffic deaths of 2.31, essentially the same as the world-wide ratio. For Sweden the ratio is greater, at 2.87.
Figure 3. The number of male fatalities divided by the number of female fatalities for 2005 data for Sweden and the US.
For drivers, male to female ratios are larger still, 5.43 for Sweden and 3.15 for the US. These larger values are due in part because a vehicle driver is more likely to be male in both countries (and more so in Sweden).
Pedestrian deaths in Sweden in 2005 were 31 males and 19 females, for a male to female ratio of 1.63. Pedestrian deaths in the US 2005 in 2005 were 3,432 males and 1,437 females, for a male to female ratio of 2.39. There are insufficient Swedish pedestrian fatalities to perform additional analyses (the more successful the safety policy, the more difficult it is to analyze). However, a detailed analysis of US data shows more male than female pedestrian fatalities per capita at every age (in one year increments) for all ages for which sufficient data are available (birth through the late 80s) (Evans, 2004, p. 228-230). The greater male pedestrian risk is clear even in the first year of life – baby boys are 40% more likely than baby girls to become pedestrian fatalities.
For passengers male risk exceeds female risk by much more than the barely greater than unity ratios for both countries shown in Figure 3. This is because older passengers are more likely to be female (Evans, 2007), and there are more females than males in the population, especially at older ages. Older people, and females, are more likely to die in the same severity crash. (Evans 2004; p. 120-146). US data for 2005 show that for the teens through the twenties (aged 13 through 29) there were 2672 male passenger deaths compared to 1675 female passenger deaths, for a male to female ratio of 1.60.
The similarity of the age and sex relationships for Swedish and US fatality data provide additional evidence that intrinsic human characteristics are central to traffic safety. The relationships show similar sex and age dependence to that found for the amount of the hormone testosterone in the body (Evans, 2004, p. 230-232).
While these seemingly immutable factors are at the center of traffic safety in both countries, safety progress is quite different. The mix of fatalities and the total number of fatalities should be viewed as conceptually distinct quantities.
The 2005 Swedish fatality count of 440 is 66% below the Swedish all-time high of 1,313 (recorded in 1966). In sharp contrast, the 2005 US fatality count of 43,443 is only 20% below its all time high of 54,589 (recorded in 1972). If the US total had dropped from its high by the same percent as the Swedish total dropped, in 2005 the US would have suffered 18,293 traffic deaths, 25,150 fewer than the actual total of 43,443.
The 66% reduction in Sweden compared to 20% reduction in the US cannot be due to stable differences between Sweden and the US, because they focus on changes. They can be due only to some major difference, or change in direction, that occurred in the two countries starting around the late 1960s or early 1970s. Either Swedish safety policy became substantially more effective, or US policy became substantially less ineffective.
While the reductions in Sweden are in sharp contrast to those in the US, they are not all that different from reductions achieved in many other countries, including Canada, Britain, and Australia. The departure of safety progress in the US from that in the most successful countries is explained and discussed in detail in the chapter titled The Dramatic Failure of US Safety Policy in Evans (2004, p. 381-411). In this chapter a number of rates (fatalities per year, fatalities per registered vehicle, and fatalities for the same distance of travel) are examined. As the findings are similar for all rates, we examine only fatalities per year in this paper. (Fatalities per registered vehicle for Sweden and the US is shown plotted versus time elsewhere (Evans 2004, p. 41)).
The core of the dramatic US failure is an obsessive focus on vehicle factors that research shows have only modest effects on safety. The data presented above show that the most fundamental characteristics of humans that are central to traffic safety. The problem is primarily one of the behavior of road users, not engineering. Countermeasures that focus on such behaviors as belt-wearing, drunk driving, running red lights, and speeding have been shown to generate huge reductions in total casualties, while vehicle factors produce at most modest reductions to small subsets of the total road-user population, the effects diminished yet further by a long phase-in process.
Advocates of the US approach have even offered explanations of the enormous Sweden compared to US difference in terms of vehicle factors, such as the increase in SUVs. NHTSA concludes (Kahane 2003, p. 200) that SUV’s have lower total fatality rates than small cars, but higher rates than large cars -- so whether more SUVs reduce or increase deaths is far from obvious. What is clear is that greater growth of SUVs in the US than in Sweden, or other changes in vehicle factors, cannot come even close to generating the differences in safety progress in the two countries. Indeed, conventional wisdom in the early 1970s was that vehicles in Sweden were safer than those in the US, providing the US with the opportunity for more potential gains through vehicle factors.
Another example of the US focus on the near irrelevant is vehicle recalls. In 2004 more than 30 million vehicle recall notices were issued - about 1.7 recalls for every new vehicle sold (McDonald, 2006). NHTSA, which is responsible for the massive expenditures and activity associated with these recalls, has no idea how many fatalities there would have been if there had been zero recalls. However, my judgmental estimate suggests that the number would likely have been in the range of three fewer to one more.
The possible increase in deaths due to recalls arises because the recall notices cause additional travel. It is certain that additional travel generates additional risk. However, there is no corresponding confidence that the fatality risk associated with the vehicle is lowered by the fix specified in the recall.
About a quarter of recall noticess are ignored. Despite the prodigious sample sizes generated by decades of recalls, there is no evidence that the vehicles not fixed pose higher risks than those fixed.
To suggest that these recalls, and the massive publicity they generate, represent merely economic waste without safety consequences misses a crucial point. What is believed by those whose lives are at risk in traffic is crucial to their safety. The belief that unimportant factors are important displaces knowledge about what is really important.
There can be little dispute that, other things being equal (which they rarely are), safety is enhanced by engineering that reduces harm when crashes occur. Likewise, measures aimed specifically at the behavior of young, or male, or older drivers can reduce casualties. Such measures can lead to important casualty reductions. Also, the reductions can be increased further by constantly striving to refine these measures. All countries that have made safety progress have addressed vehicle factors, safety recalls, younger drivers, older drivers, etc. It is not the items on the list of countermeasures that is crucial, but the relative emphasis placed on each of them. The largest reductions have been achieved by public policies aimed at reducing crashes (increasing use rates for active occupant protection devices, a behavioral factor, is also important). The goal must be the attainable one of developing public policies that lead to safer traffic in which crash risk is lower, not the unattainable goals of producing vehicles that are safe to crash, getting young people to behave like older people, or getting males to behave like females.
This work rests on data from the Swedish Road Administration, Borlänge, Sweden, provided in a personal communication from Thomas Lekander. Not only did he provide the data, but additionally kindly translated the many column and row headings into English.
Evans L. (2004) Traffic Safety. Bloomfield Hills, MI: Science Serving Society.
Evans L. (2007) Drivers involved in crashes killing older road users. Society of Automotive Engineers, SAE paper 2007-01-1165. Warrendale, PA.
Kahane CJ. (2003) Vehicle weight, fatality risk and crash compatibility of model year 1991-99 passenger cars and light trucks. Report DOT HS 809 662. Washington, DC: US Department of Transportation, National Highway Traffic Safety Administration, Washington, DC.
McDonald KM. (2006). Shifting Out of Park: Moving Auto Safety from Recalls to Reason. Tucson, AZ: Lawyers & Judges Publishing Company.
National Highway Traffic Safety Administration [NHTSA]. (2007) Fatality Analysis Reporting System (FARS) Web-Based Encyclopedia. Data files and procedures to analyze them. Available at http://www fars.nhtsa.dot.gov.
Swedish Road Administration. (2007) Data disaggregated by various factors provided in personal communication from Thomas Lekander. (Total fatalities available at European Conference of Ministers of Transport (2006). Preliminary data on Road Safety in Europe: Improvement continues in the West - Decrease of fatalities in Central and Eastern European Countries - Positive signs confirmed in the CIS. Available at http://www.cemt.org/events/PressReleases/06acc2005.pdf.)
World Health Organization. (2003) Burden of disease project. Global burden of disease estimates for 2001. Available at http://www.who.int/healthinfo/statistics/gbdwhoregionmortality2001.xls