Related Links

Insurance Institute's March 19, 2005 Status Report,  "The Risk of Dying in One Vehicle versus Another" (PDF)
 

Michelle White's article, "The Arms Race on American Roads"(PDF)
 

Safety Researcher Leonard Evans' Science Serving Society Web site 

Other Stories this Issue:
From the Battlefield to the Soccer Field

Scapegoat Utility Vehicle?
 

A Wooly Mammoth for Teenage Eyes
 

The Evolution of the SUV: A Pictorial Timeline
 

List of References
 

Download printer-friendly PDF of Newsletter  (969KB)

Download printer-friendly PDF of this article only

Tracking Fallout from the SUV
"Arms Race"

 A New, Holistic Look at Old Data

 
Whatever gains in safety people might achieve when they "buy up" in size by replacing their passenger cars with SUVs and light trucks—and the gains are not as clear as was once thought—they are likely to come at the expense of the safety of other road users, according to a new analysis by UC San Diego economist Michelle White. In carrying out her examination she brought a new analytical tool to traffic safety data and, in the process, highlighted how more data would help us understand the role that vehicle type plays in traffic safety.  

White presented her findings, which were published in the October 2004 Journal of Law and Economics as "The 'Arms Race' on American Roads: The Effect of SUVs and Pickup Trucks on Traffic Safety," to the UC Berkeley Traffic Safety Seminar on March 14, 2005.   

White found that for each fatal crash that occupants of an SUV or light truck avoid being killed in because of their vehicle's size and weight, occupants of other passenger vehicles suffer at least 4.3 additional fatal crashes. That makes the SUV/light truck class extremely deadly, she said. 

Additionally, White found that for every 1 million light trucks and SUVs that replace cars, between 34 and 93 additional other road users are killed each year, and the value of the lives lost is costing the country between $242 and $652 per year, according to a model White used to analyze data from the National Highway Transportation Safety Administration's General Estimates System (NHTSA GES).  

But White, and traffic safety researchers in the audience, offered caveats as well. Driver behavior such as risk-taking, aggressiveness and inattention is notoriously hard to quantify and difficult to even find as a consistently measured entity in most databases, yet it is critical to an accurate understanding of vehicle safety. On average, drivers of cars drive in a safer manner than drivers of SUVs and light trucks. That is one reason why the range in risks and costs mentioned above is so wide. There is uncertainty over how, or if, drivers would change their behavior if they were to change vehicles.  

In the example involving the 1 million vehicle replacements cited above, if the statistically more dangerous behavior shown by drivers of light trucks and smaller SUVs were to carry over to the same drivers in passenger cars, White is postulating that the higher numbers would pertain. If they were to change their way of driving  to more closely resemble statistically safer passenger car drivers, the lower numbers would apply.  

Other limitations include inaccurate reporting by police officers at the scene, omissions of crashes from the database, and coding errors. Another potentially confounding element is lack of a consistent definition of vehicle make, model and type across data sets, which most seriously affects SUV and light truck studies because they share categories in some data sets with other vehicle types such as minivans, but not in all. Also, SUVs and light trucks in the databases are not always distinguished by their weight and height, so that smaller models are lumped in with heavier and taller ones, all of which can affect their safety records.  

Sometimes access to data itself—vehicle registration information, for example—is not freely available, so studies cannot be duplicated under identical conditions by independent researchers—a key component of analytical research.   

As this article will show, attempts to account for these shortcomings can contribute to the adoption of certain conclusions as traffic safety dogma that may not, as White's analysis suggests, be entirely accurate. In this case,  it was the widely accepted wisdom that linked a vehicle's weight to the amount of protection it affords its passengers, which the auto industry, and much traffic safety theory, had propounded for years.

A Dogma Is Challenged 

One of the leading supporters of that theory was Leonard Evans, an industry authority who has published many papers on the "bigger/heavier-is-safer" subject that have been used by automakers in defending their manufacturing decisions.   

It was a view that was widely held. “It’s been well established for over three decades that when traffic crashes occur, occupants in heavier/larger vehicles are at lower risk than occupants in lighter/smaller vehicles. … Policies aimed at reducing fuel use have led to lighter vehicles, which have increased traffic fatalities,” Evans wrote in a 2004 article that he published through his traffic safety research institute, Science Serving Society, where he explains how the thinking has changed.  

He stated that it is necessary to qualify the "bigger-is-safer" assertion, citing the influence of vehicle design in addition to weight, rather than weight alone, and also warning against the dangers the heavier vehicle pose to others. Evans then re-analyzed FARS data to see if altering weight independent of size affects net risk. His quantitative conclusions point to the need for better design.   

Evans found that a reduction of 32 kilograms, along with a .2 meter increase in length, generated an 8.5 percent reduction in the net fatality risk in crashes between large and small cars. That is substantial when compared to the 12 percent net risk reduction achieved by $50 billion worth of airbags that carmakers have installed.   

“In short: if a car is heavier, it reduces risk to its driver, but increases risk of other drivers. If a car is larger (without also being heavier) it reduces risk to its driver and also reduces risks to other drivers,” Evans concluded. Evans' re-analysis now appears to favor roomy sedans with big crumple zones as the type of vehicle offering the greatest degree of safety for the widest range of road users.

New Ways to Look at the Data 

White's experience of using econometric analysis to reach a similar conclusion illustrates the promise that cross-discipline study holds in helping augment safety researchers' analytical tools in order to extract more knowledge out of the data that is available. Based on her background as an economist, White brought the fundamental economic concept of external costs to vehicle design and safety.  

Her prior work regarding SUV safety was targeted at the efficiency and equity of no-fault insurance laws in different states. They were designed to lower insurance rates and speed up claims processing by having  insurance carriers pay for their own customers' accident-related medical bills, regardless of who was at fault. No-fault agreements meant that injured parties didn't have to sue the people who collided with them to cover basic medical expenses.

A Question of Fairness 

However, if there is a higher likelihood of being injured in a collision with an SUV or light truck than in one with a car, that could affect the equity of the no-fault system, especially with the 20 percent increase in SUVs and light trucks on the road in the past decade or so.   

"In states with no-fault, the liability system does not penalize drivers of large or heavy vehicles at all for causing higher damage. In fact, drivers of cars are at a disadvantage under no-fault, since they suffer more damage in crashes and bear these costs themselves," White says.  

White became interested in uncovering more of the external costs of these vehicles and acquired the GES data set from NHTSA. NHTSA's GES is a nationally representative sample of police-reported motor vehicle crashes of all types, from minor to fatal. Begun in 1988, the system is used to estimate how many motor vehicle crashes of different kinds take place, and what happens when they do. These collision reports are chosen from 400 police jurisdictions in 60 areas that together create a composite picture of the geography, roadway mileage, population and traffic density of the U.S. The jurisdictions provide a random sample of police collision reports from a collection of 50,000 a year. Each report is coded for roughly 90 data elements. As a measure of the complexity of the database, the user manual is 216 pages long.  

White downloaded the NHTSA GES file in 2001 and had the data  recoded and fed into a model to determine measures for light truck/SUV safety compared to cars. The model analyzed roughly 192,000 crashes from 1995 to 2001, and White began to extract conclusions. 

Among them:   

    * The probability of fatalities for car occupants falls by 38 percent when the other vehicle is a car rather than a light truck. For serious injuries the reduction is 19 percent.

    *  The probability of fatalities among light truck occupants falls 55 percent, and the probability of serious injuries among light truck occupants falls 22 percent if the light tuck hits a car instead of another light truck.

    *  The probability of fatalities rises by 45 percent, and serious injuries by 11 percent if a light truck hits a pedestrian or bicyclist rather than a car.

    *  Motorcyclists have a 56 percent higher probability of death if hit by a light truck instead of car and a 26 percent higher probability of injury.

    *  Light truck drivers have a 14 percent higher probability of dying in a single-vehicle crash than if they were in a car, and a 14 percent higher chance of injury in such a crash.

    *  For each 1 million cars replaced by light trucks, between 34 and 93 additional car occupants, pedestrians, bicyclists, or motorcyclists are killed each year in traffic crashes, the lower figure if it is assumed that the former truck drivers will drive more cautiously when driving cars, the higher figure if it is assumed that they won't. 
 

The value of lives lost alone is between $242 million and $652 million per year for each light truck that replaces a car, not counting the costs of injury treatment and productivity lost because of disability (again, depending on the assumption about driver behavior noted in the item above).  

However, White said that there are opportunities to develop deeper and more detailed databases. Other researchers in the audience at Etcheverry Hall expressed similar sentiments, adding that major proprietary data sets like vehicle registration records can be hard to obtain. Also, the data from police reports can be incomplete or inaccurate.

How More Data Would Help 

Even NHTSA's GES has inherent limitations that researchers must work with. It is a sample of a collection of records estimated at 50,000, NHTSA reports, but White said some states understate or overstate the frequency or severity of different types accidents in their jurisdictions. Also, although there are 90 data characteristics, certain correlating data that could sharpen results are missing. The reports do not include things like crash victims' socioeconomic status, number of crashes they had been in during the previous year and other variables shown to affect traffic safety results, White said. 

NHTSA's GES has other shortcomings. An estimated 50,000 vehicle collisions per year in the 60 sampling areas are not accounted for due to the fact that they are not reported to the police, White states, most likely because they are single-vehicle collisions or cause only minor injury and damage, which is not reported.  

Furthermore, changing definitions of vehicles and vehicle designs can be difficult to account for. For example, in 1998, SUVs as a group were shifted from the car to the light-truck category in GES data, despite the fact that many newer models, like those considered "crossover utility vehicles," incorporate many of the safety elements of cars and have designs that are much less aggressive in collisions with cars. The inaccuracies that can result from such broad classifications are illustrated by the following example. In its March 19, 2005 issue of its magazine, Status Report, the Insurance Institute for Highway Safety found that driving a Chevrolet Blazer 2-door is roughly 25 times riskier to its driver than driving a Toyota 4Runner SUV. The 4Runner had 12 deaths per million registered vehicles, versus 308 for the Blazer, yet both are coded as light trucks in GES.  

Wenzel had similar definition issues with NHTSA’s FARS database in his 2005 study, “Are SUVs Really Safer Than Cars?” (See "Scapegoat Utility Vehicle" in this issue). As Wenzel was forced to do, White divides vehicles into categories using a degree of subjectivity. White uses three—cars, light trucks, and heavy vehicles—but "cars" includes everything from a Ford Mustang to a Honda Accord, while the light-truck category ranges from 5,000-pound, mid-size SUVs to differently shaped minivans which weigh thousands of pounds less. Other categories get similarly combined or even dropped. Skateboarders and scooter riders are classified as pedestrians in White’s study, while collisions with horse carriages (common in a few cultural enclaves in the east) are dropped.   

In another example of missing data, NHTSA’s FARS cannot account for the actual curb weights of vehicles other than cars—a problem if a researcher is seeking to correlate weight with danger.

Statistically applied under-weights and over-weights have a powerful ability to compensate for unknown factors in databases, and they are deployed routinely in data sets to account for expected irregularities. For example, GES data are weighted to compensate for perceived police under- or over-reporting. In another example, because her data set runs from 1995 to 2001 and cannot account directly for the increase in airbags and other safety features, White must weight aspects of the sample to make it representative of all crashes. White said the science of weighting is well known, but it’s also an art, requiring subjective decisions that will influence results. 

If anything, SUVs have slowed the rate of the rise in safety of America's roads.

Interestingly, the advent of these safety features is used by automakers to claim that SUVs have increased safety overall. Aggregate data appear to show that their entry into the fleet in significant numbers correlates with an increase in road safety, but what's not being counted are the savings achieved during the same period as a result of increased seat belt usage, airbags, less drunken driving and other factors over the last 20 years, White said. These are positive developments that would have improved safety no matter what. If anything, SUVs have slowed the rate of the rise in the safety of America's roads, she argues.  

The limitations of GES and other popular data sets like NHTSA’s FARS and the IIHS’s yearly vehicle risk study point to an opportunity for a supplemental national database to enhance their capabilities, or new methods of extracting data from the existing sets. If new information were collected and added to the GES database, it would be possible to correct for safety features of cars and also for vehicle weight, White noted.

The Unanswered Question: Behavior 

Looming over all three  data sets is the influence of the "X" factor of driver behavior and habits. For example, in the issue of the Insurance Institute's Status Report comparing driver death rates in different makes and models, the Institute further weighted FARS data based on observations that women were much safer drivers than any other category of driver. Crash numbers for vehicle makes and models that women were prone to buy had to be weighted to account for the inherently safer demographic. 

A household survey of thousands of homes could correlate crash information with demographic information on age, drinking patterns, seat belt usage, annual vehicle mileage, occupation, and reasons people drive, White said. 

Wenzel echoed such sentiments in his talk one month earlier. “It’s extremely difficult to determine inherent safety of a vehicle type or model because of the difficulty in separating the contribution of driver characteristics and behavior from the contribution of vehicle design,” he stated. “For example, some car models may attract relatively aggressive drivers, who increase the fatalities in the model, independent of its design.” 

While FARS, GES, and IIHS data sets contribute much to understanding crash characteristics, White and Wenzel see a strong opening for a comprehensive database drawn from the entire population and supplemented with follow-up surveys, which could tease out the major factors responsible for a vehicle’s safety. 

Automakers' market acumen, combined with the latest in academic research, could develop powerful new tools to forecast future problems like fleet compatibility before it becomes too late to do much about them. The long view of traffic safety could save thousands of lives.