Blood and Oil: Vehicle Characteristics in Relation to Fatality Risk and Fuel Economy

Abstract

I examined the potential for a lower risk of death compatible with increased fuel economy among 67 models of 1999–2002 model year cars, vans, and sport-utility vehicles (SUVs) during the calendar years 2000 to 2004. The odds of death for drivers and all persons killed in vehicle collisions were related to vehicle weight, size, stability, and crashworthiness.

I calculated that fatality rates would have been 28% lower and fuel use would have been reduced by 16% if vehicle weights had been reduced to the weight of vehicles with the lowest weight per size, where size is measured by the lateral distance needed to perform a 180-degree turn. If, in addition, all vehicles had crashworthiness and stability equal to those of the top-rated vehicles, more than half the deaths involving passenger cars, vans, and SUVs could have been prevented by vehicle modifications.

INCREASED DEMAND FOR OIL, the war in Iraq, and devastating hurricanes elevated gasoline prices during 2005. Methods that accurately predicted the peak of US oil production in the 1970s forecast both a peak in world production in 2005 and a decline after a period at this peak.1 The increased risk of death resulting from a shift to more fuel-efficient small vehicles may again raise the issue of “blood for oil,” a much-debated topic after the gasoline supply shortages in the 1970s.2

The energy of a moving vehicle is its mass multiplied by the square of its velocity. In other words, at any given speed, the higher the weight of a vehicle (weight is indicative of mass) and its contents, the more energy there is to be managed in a crash. Because increased weight decreases fuel economy and increases braking distance, it would appear that weight reduction would decrease both injuries and fuel consumption. Increased weight is an advantage to the occupants of the heavier vehicle in a 2-vehicle collision, and in single-vehicle crashes, increased weight may bend or break relatively fixed objects, such as small trees, that would otherwise compromise the integrity of passenger compartments. But in head-on crash tests of 2 cars of different sizes, the lighter cars do not stop; they are forced backward, resulting in the exertion of far greater forces on their occupants.3

Based on the ratio of driver deaths in lighter vehicles to driver deaths in heavier vehicles in 2-car crashes, Evans and Frick opposed increased fuel economy standards in the 1990s.4 But, as they noted, less than a quarter of all road deaths occur in crashes between passenger cars. Their claim that weight is more important than vehicle size was also based only on 2-car crashes.5 Public policy should not be made on the basis of such a small minority of crashes. From a public health perspective, the issue is whether vehicle characteristics increase or decrease risk to all road users and whether fuel economy can be improved without increasing risk of death.

Vehicle size provides space for air bags and seat belts to restrain occupants who move at precrash speeds, during a crash. If the weight differential in the 2-vehicle crash is large enough, the passenger compartment of the lighter vehicle may be penetrated, neutralizing the advantage of space. In the past, most studies of vehicle weight and size used wheelbase—the distance from the front to rear axles of passenger vehicles—as the indicator of space.6

Vehicle wheelbase and weight are correlated; vehicles with longer wheelbases usually are heavier. Analysts in the 1970s noted that manufacturers could reduce fuel use without compromising safety or size by using materials that reduce vehicle weight.7 Nevertheless, most of the vehicle manufacturers developed heavy sport-utility vehicles (SUVs), promoted them as safer than cars, and sold them at premium prices. Recent studies have noted the severe consequences to occupants of cars8 and other road users9 struck by these vehicles, which are heavier, on average, than cars. Also, the high center of gravity from the ground (H) relative to track width (T) for many of these vehicles increased rollover death rates of occupants. The coefficient T/2H is a commonly used indicator of stability.10 Risk of death to vehicle occupants in crashes is also related to crashworthiness, the extent to which the vehicle absorbs energy outside the passenger compartment and minimizes forces exerted on vehicle occupants, particularly to the face and chest, where most fatal injuries occur.

I attempted to assess the effect on fatalities of each of these factors, controlling statistically for the effects of the others. The subject vehicles were l999–2002 model year passenger vehicles, including vans and SUVs but excluding pickup trucks, during their first year of use through 2004. I performed separate analyses for driver deaths and all road user deaths in which these vehicles were involved. “All road users” include drivers, other occupants of the subject vehicles, and those involved in collisions with the subject vehicles. Pedestrians, bicyclists, and motorcyclists who died from collisions with the subject vehicles are included because their risk of death may be increased by the longer stopping distances associated with increased vehicle weight.

METHODS

The National Highway Traffic Safety Administration’s Fatality Analysis Reporting System collects data throughout the United States on motor vehicle fatalities that occur within 30 days of a crash. Included are data on curb weight and wheelbase of the passenger vehicles as well as environmental conditions and driver data such as age, gender, blood alcohol concentration, and prior crash and moving violation records. Included were 1999–2000 model year passenger vehicles in use during 2000 to 2004, used for more than 200 000 cumulative years during that period, for which data on stability were available.11 Collisions involving more than 2 vehicles were eliminated to minimize double counting. In 2-vehicle collisions, occupant deaths were assigned only to the vehicle in which the fatal injury occurred. Pickup trucks were excluded as subject vehicles because of lack of sales data on the substantial variation in weight and wheelbase within the same make and model. Where the subject vehicles collided with any type of truck and 1 or more occupants of either vehicle died, the cases were included. The data were available for 67 make and model combinations.

Years of use were calculated as the sale of a given make and model in a given month12 multiplied by the number of months remaining in 2004, with the total divided by 12. A total of 14438 deaths occurred to people as occupants or other road users in collisions of these vehicles during 104 970 000 years of use. There were 7263 driver deaths in the vehicles, 50% of total deaths overall and 66% of total deaths to occupants of the subject vehicles.

Lateral distance needed to perform a 180-degree turn (turn distance) was an indicator of vehicle size and was less correlated with weight than with wheelbase.13 An index of crash-worthiness was created using vehicle ratings obtained from the Insurance Institute for Highway Safety. Ratings were given to vehicles on the basis of the institute’s frontal offset crash tests of those vehicles. These tests were done at 40 mph with 40% of the total width of the vehicle striking a fixed barrier on the driver side.14 The tests better simulate a common type of severe crash than the National Highway Traffic Safety Administration’s full-front barrier crashes at 35 mph.15 The Insurance Institute for Highway Safety rates vehicles on a 4-point scale on several factors. In this study I employed the scores on life-threatening factors: structural integrity; forces on the head and, separately, the chest of a test dummy; and performance of the restraint systems (seat belts and air bags) in restricting movement of the dummy. I used a summary measure to average the ratings (good = 1, acceptable = 2, marginal = 3, and poor = 4) for the 4 factors on each vehicle.

I used logistic regression to estimate the effects of the selected predictor variables on the odds of mortality in the years of use per make and model. I used least squares regression to estimate the potential for confounded effects. For example, if younger drivers more often drive smaller vehicles, some or all of the correlation of vehicle size and odds of death should be attributed to the factors that produce higher risk in driving by younger drivers. Because there are no data on the use of specific makes and models of vehicles in high- or low-risk environments or by high- or low-risk drivers, it is necessary to assess the potential for confounding indirectly from the crash data.16 If there is confounding of the effect of vehicle weight by age of driver, the ratio of older to younger drivers among the makes and models must be negatively correlated with the vehicles’ weights. Therefore, the ratios of low to high risk of 10 major environmental factors and 10 major driver factors per vehicle were correlated to parameters of the vehicles to rule in or out the potential for confounding.

RESULTS

A multiple regression analysis that included the predictor variables and fuel economy indicated that weight was the only variable with a significant effect on fuel economy. Turn distance was strongly correlated to wheelbase, but weight and turn distance were less strongly correlated than weight and wheelbase. Therefore, I eliminated wheelbase from the analysis to minimize the effects of colinearity. None of the other correlations showed colinearity.

The coefficients relating the predictors to log odds of driver deaths and all deaths in which given makes and models were involved are shown in Table 1 ▶. Because vans and SUVs are more often used as family vehicles and less often used for higher-risk activities such as drag racing, they each have their own coefficient in the analysis. Coefficients on passenger cars exclusive of sports cars (cars classified as sports cars by the auto industry, the insurance industry, or the government), vans, and SUVs are shown separately to illustrate that the effects are not the result of special characteristics of vans and SUVs, such as larger engines or 4-wheel drive.

TABLE 1—

Coefficients (SEs) on Log Odds of Driver Deaths and All Deaths by Vehicle Parameter: 2000–2004

	Cars, Vans, SUVs		Passenger Cars Excluding Sports Cars
	Driver Deaths	All Deaths	Driver Deaths	All Deaths
Intercept	−5.5029 (0.9244)	−7.7635 (0.6148)	−9.5931 (0.2427)	−9.6126 (0.1775)
Curb weight	−0.0133 (0.0039)	0.0124 (0.0028)	0.0015 (0.0045)	0.0229 (0.0033)
Turn distance	−0.0143 (0.0077)	−0.0110 (0.0055)	−0.0219 (0.0086)	−0.0196 (0.0064)
Crash test average	0.4525 (0.0213)	0.3801 (0.0155)	0.5236 (0.0250)	0.4751 (0.0186)
T/2H < 1.2, else 1.2a	−3.1742 (0.7614)	−1.4101 (0.5039)	. . .	. . .
Vanb	−0.8842 (0.0671)	−0.6440 (0.0404)	. . .	. . .
SUVb	−0.4552 (0.0615)	−0.3652 (0.0409)	. . .	. . .

Note. SUV = sport-utility vehicle.

^a There are no passenger cars with T/2H < 1.2.

^b There are no vans and SUVs under the heading “Passenger Cars Excluding Sports Cars.”

Driver death rates decreased as weight increased, but death rates of all road users increased in correlation with increased weight. The weight coefficient for drivers of passenger cars was not significant. Higher stability ratios, longer turn distances, and “good” ratings on all 4 elements of crashworthiness lowered the risk of death. Vans and SUVs were less involved in fatalities when the other factors were controlled statistically. The size, stability, crashworthiness, and vehicle type coefficients were greater in magnitude for drivers’ deaths than for all deaths, indicative of occupant protection, whereas the risk of death to other road users increased as a function of vehicle weight.

In most instances, there was no significant correlation between the ratio of low to high risk of environmental and behavioral factors in relation to vehicle characteristics. Most of the few modest correlations of any significance were in the opposite direction from any indication of higher risk among the vehicles with higher risk characteristics. The ratio of male drivers to female drivers in fatal crashes was higher as vehicle weight increased, but the correlation was weak. In the aggregate, the low correlations—and the reverse direction of most of the significant ones—suggest no confounding factor that would negate the findings that weight, size, stability, and crashworthiness are primary factors in vehicle mortality rates.

I examined the potential effect of lower vehicle weight with no corresponding change in vehicle size on mortality rates and fuel use using data on the correlation of these 2 characteristics (Figure 1 ▶). Although there is a 43% overlap in the variances of the 2 vehicle characteristics, Figure 1 ▶ clearly shows that many of the vehicles examined weigh far more than the minimum weight achieved in other vehicles with similar turn distances. SUVs and vans tend to be heavier, but several makes and models are within the same weight range as passenger cars.

FIGURE 1—

Vehicle weight and turn distance.

The coefficients for all deaths in Table 1 ▶ indicate the approximate reductions in mortality that would accompany a reduction in weight to the minimum—acheived in currently manufactured vehicles of a given turn distance—assuming that the same number of vehicles would have sold with the reduced weight. The weights of vehicles in Figure 1 ▶ increased by approximately 156 lb per 1-ft increase in turn distance. A line drawn through the points along the bottom edge of the data in Figure 1 ▶ shows the minimum weight of those vehicles at a given turn distance, calculated using the equation 156 × turn distance in feet −3000. Using the equation predicting all deaths from Table 1 ▶, I subtracted the expected lives lost had weight been minimized from the actual lives lost for each make and model combination, given the extant curb weight for each vehicle. The sum of the differences across vehicles is 4032 deaths, about 28% of the number of persons killed in collisions involving these vehicles. The same method for vehicle crashworthiness and stability indicated that 3604 (25%) deaths could have been prevented if all of the vehicles had received the top crashworthiness rating and 415 (3%) could have been prevented if T/2H was 1.2 or greater on vehicles that had less stability.

Regression of EPA-rated highway fuel economy by curb weight of the vehicles indicated about 7 fewer miles per gallon per 1000 lb of added weight. Although EPA-rated mileage overestimates actual mileage, it accurately reflects differences among vehicles under controlled conditions.17 The weight-influenced relative fuel use is therefore considered accurate. Weight accounts for approximately 68% of the variation in fuel economy, a good fit of the data to a linear model. I applied the regression to the difference between actual and minimum weight achieved at a given turn distance, assuming 12000 miles per vehicle per year, which resulted in an estimated 16% reduction in fuel use had the manufacturers minimized weight per vehicle turn distance.

DISCUSSION

The data indicate that fuel economy is not incompatible with societal risk if reductions in vehicle weight are accomplished without reducing turn distance (vehicle size). As predicted by physics, increased weight increases risk of death and fuel consumption. Increased vehicle size reduces risk of death. The increased risk to vehicle drivers of reduced vehicle weight is more than offset by the reduced risk to other road users. Had the 1999–2002 models been manufactured with the minimum weight-to-turn-distance ratio of lighter vehicles, death rates would have been reduced by approximately 28%; gasoline use, by 16%. Additionally, had all vehicles received the highest crash-worthiness rating and had a T/2H of 1.2 or better, more than half of the deaths could have been prevented.

Although it is doubtful that many vehicle buyers have a precise knowledge of the effect of weight and size on personal and societal risk, most have probably heard that bigger is safer. The advantage to drivers and disadvantage to other road users of increased weight presents a dilemma for the knowledgeable buyer. If informed that greater weight reduces user risk somewhat but increases risk to other road users more, how many would opt for the heavier vehicle? The government has the authority to resolve the issue by regulating vehicle crashworthiness and fuel economy but has made only minor adjustments in the past 20 years. In an attempt to increase fuel economy, Congress in 1975 required each manufacturer to achieve an average of 27.5 miles per gallon for its fleet of passenger cars and light trucks.18 The manufacturer that wishes to market heavier vehicles can maintain the average by also marketing much lighter vehicles. Whereas fuel economy is improved when the miles-per-gallon average is increased, mortality risk is increased by the extent to which variance in average vehicle weight is increased. More sensible fuel economy regulation that would not be adverse to safety could be achieved by setting a standard for minimum fuel economy dependent on vehicle size. Manufacturers would have an incentive to minimize weight in vehicles in a given size category and to use more fuel-efficient engines in larger vehicles.

The correlation of increased mortality risk with less than “good” ratings on crash tests suggests that vehicle buyers would reduce their risk of motor vehicle fatalities by avoiding vehicles with one or more ratings less than “good.”