NHTSA Report Number DOT HS 809 662October 2003

Vehicle Weight, Fatality Risk and Crash Compatibility of Model Year 1991-99 Passenger Cars and Light Trucks

Charles J. Kahane, Ph.D.

ABSTRACT

Logistic regressions calibrate crash fatality rates per billion miles for model year 1991-99 passenger cars, pickup trucks, SUVs and vans during calendar years 1995-2000 – by vehicle weight, vehicle type, driver age and gender, urban/rural, and other vehicle, driver and environmental factors – a cross-sectional analysis of the fatality rates of existing vehicles.  These analyses suggest that, after controlling for driver age/gender, urban/rural, annual mileage, and other factors:

Logistic regression analyses of fatalities per billion miles in two-vehicle collisions show that MY 1991-99 LTVs were more aggressive than MY 1991-99 cars when they struck other vehicles.  The analyses show correlations between occupants' fatality risk in the struck car and the frontal height-of-force and rigidity of the striking LTV.

EXECUTIVE SUMMARY

The National Highway Traffic Safety Administration's (NHTSA) 1997 report on vehicle weight and fatality risk estimated the effects of 100-pound reductions in light trucks and vans (LTVs) and in passenger cars.  In the 1997 report, statistical analyses of model year (MY) 1985-93 vehicles in calendar year (CY) 1989-93 crashes found little overall effect for a 100-pound reduction in LTVs, but an increase of about 300 fatalities per year in cars. However, they also produced the doubtful findings that vehicle weight reductions do not increase fatality risk in car-to-car or LTV-to-LTV crashes and even reduce fatality risk in pedestrian crashes. 

NHTSA took a good, hard second look at the subject, identified anomalies in the 1997 report, and applied different analysis techniques to more recent crash data.  This new statistical analysis of MY 1991-99 vehicles in CY 1995-2000 crashes supersedes NHTSA's 1997 report.

The new study expands the analyses by separately estimating the effects of 100-pound reductions in heavy LTVs, light LTVs, heavy cars and light cars.  It compares the fatality rates of LTVs and cars, to quantify differences between vehicle types, given drivers of the same age/gender, etc. In support of NHTSA's research on car-LTV compatibility, it analyzes fatality rates in two-vehicle crashes based on the mass and rigidity of each vehicle and the height mismatch between vehicles.

Effects of 100-pound weight reductions on fatality rates

In MY 1991-99, and earlier, heavy vehicles had lower fatality rates per billion miles of travel than lighter vehicles of the same general type.  When two vehicles collide, the laws of physics favor the occupants of the heavier vehicle (momentum conservation).  Furthermore, heavy vehicles were in most cases longer, wider and less fragile than light vehicles.  In part because of this, they usually had greater crashworthiness, structural integrity and directional stability.  They were less rollover-prone and easier for the average driver to control in a panic situation.  In other words, heavier vehicles tended to be more crashworthy and less crash-prone. Some of the advantages for heavier vehicles are not preordained by the laws of physics, but were nevertheless characteristic of the MY 1991-99 fleet.  Offsetting those advantages, heavier vehicles tended to be more aggressive in crashes, increasing risk to occupants of the vehicles they collided with.

The statistical analysis uses the Fatality Analysis Reporting System (FARS), R.L. Polk registration data, State crash data and the National Automotive Sampling System (NASS).  Logistic regressions calibrate crash fatality rates per billion miles for model year 1991-99 vehicles during calendar years 1995-2000 – by vehicle weight, driver age and gender, urban/rural and other factors discussed and quantified in this report: availability of air bags, ABS, or 4-wheel drive; vehicle age; annual mileage; speed limit; day/night; wet/dry road; high/low State fatality rate; and calendar year. “Crash” fatality rates include fatalities to occupants of the case vehicle, occupants of the other vehicles it collides with, and any pedestrians.  The key is to compare fatality rates of heavy and light vehicles “on a level playing field” by adjusting for differences in the age and gender of the drivers, the types of roads they travel, and the other factors.  In each of six crash modes that, together, account for over 96 percent of the nation's crash fatalities, the analysis calibrates the average increase in the fatality rate for vehicles weighing W-100 pounds relative to vehicles weighing W pounds, after controlling for driver age/gender and the other factors – a cross-sectional analysis of the fatality rates of existing vehicles.  (Throughout this study, a vehicle's “weight” is its “curb weight”: the actual weight of the vehicle with a full tank of fuel and other fluids needed for travel, but no occupants or cargo.)

Table 1 shows the average fatality increase per 100-pound reduction in LTVs.  As stated above, the “fatality increase per 100-pound reduction” does not mean the effect of literally removing 100 pounds from a specific LTV.  It is the average increase in the fatality rates of 1991-99 models weighing W-100 pounds relative to other 1991-99 models weighing W pounds, given drivers of the same age/gender and equal values on the other factors.  The analysis comprises pickup trucks, SUVs, minivans and full-sized vans.  The top half of Table 1 shows the effect in light trucks weighing 3,870 pounds or more (this was the median weight of LTVs in MY 1991-99, but the majority of trucks after MY 1995 were heavier).  As curb weight decreased by 100 pounds, fatality rates increased by 2.5 to 3 percent in rollovers and fixed-object collisions.  Fatal crashes with pedestrians and heavy trucks were hardly affected.  However, in collisions of heavy LTVs with cars (where 83 percent of the crash fatalities were occupants of the cars) or with other, usually lighter, LTVs, the 100-pound reduction resulted in a modest net benefit, because it somewhat reduced risk to the occupants of the other vehicles.

In each crash mode, the percentage effects calibrated for MY 1991-99 vehicles were applied to the baseline of all CY 1999 crash fatalities in the United States (all model years) to estimate the annual net fatality change if the mix of LTVs weighing 3,870 pounds or more on the road that year had averaged 100 pounds lighter – i.e., if the public had purchased fewer of the very heavy LTVs and more of the make-models weighing not so much in excess of 3,870 pounds.  The increase in rollovers and fixed-object crashes was partly offset by the reduction in LTV-to-car and LTV-to-LTV fatalities.  The point estimate of the net change for all crash modes was an increase of 71 fatalities, not statistically significant, as evidenced by the interval estimate ranging from –156 to +241.  The interpretation of these interval estimates will be discussed after the presentation of all the results for LTVs and cars.  The point estimate for the percentage change was a nonsignificant increase of 0.48 percent.  The results for the heavier LTVs suggest that there may have been some weight above 3,870 pounds beyond which overall fatality rates tended to increase, rather than decrease, as weight increased.

The lower half of Table 1 shows the effect in LTVs weighing less than 3,870 pounds.  As curb weight decreased by 100 pounds, fatality rates increased in every crash mode – although the observed increases in collisions with pedestrians (1.24 percent) and with cars (1.13 percent) were small and not statistically significant.  In rollovers and collisions with fixed objects, heavy trucks or other (usually heavier) LTVs, fatality rates increased substantially (3.15 to 6.98 percent) as the weight of the “case” LTV decreased.  The point estimate of the net change for all crash modes in baseline CY 1999, per 100-pound reduction among the LTVs weighing less than 3,870 pounds, was an increase of 234 fatalities per year (interval estimate: 59 to 296).  The point estimate for the percentage change was an increase of 2.90 percent.

TABLE 1
FATALITY INCREASE PER 100-POUND WEIGHT REDUCTION, LIGHT TRUCKS

(Baseline = CY 1999 total fatalities, MY 1996-99/CY 1996-2000 fatality distribution)
 
Crash Mode Annual
Baseline
Crash
Fatalities
Effect (%) of
100-Pound Reduction
Annual Net
Fatality Change
Point
Estimate
Interval
Estimate
Point
Estimate
Interval
Estimate
LIGHT TRUCKS WEIGHING 3,870 POUNDS OR MORE
Principal rollover 2,183 2.56 .81 to 3.94 56 18 to 86
Fixed object 2,639 3.06 1.41 to 4.34 81 37 to 115
Ped/bike/motorcycle 2,043 .13 -1.56 to 1.45 3 - 32 to 30
Heavy truck 860 .62 -1.61 to 2.48 5 - 14 to 21
Car 5,186 -.68 -1.79 to .06 - 35 - 93 to 3
Light truck < 3,870 1,010 -1.50 -3.20 to - .17 - 15 - 32 to - 2
Light truck 3,870 +* 784 -3.00 -6.40 to - .34 - 24 - 50 to - 3
OVERALL 14,705 .48 -1.06 to 1.64 71 - 156 to 241
 
LIGHT TRUCKS WEIGHING LESS THAN 3,870 POUNDS
Principal rollover 1,319 3.15 .64 to 4.30 42 8 to 57
Fixed object 1,687 4.02 1.71 to 4.97 68 29 to 84
Ped/bike/motorcycle 1,148 1.24 -1.26 to 2.38 14 - 14 to 27
Heavy truck 584 5.91 3.10 to 7.36 35 18 to 46
Car 2,062 1.13 -.92 to 1.82 23 - 19 to 38
Light truck < 3,870* 247 6.98 1.92 to 9.32 17 5 to 23
Light truck 3,870 + 1,010 3.49 .96 to 4.66 35 10 to 47
OVERALL 8,057 2.90 .73 to 3.67 234 59 to 296
__________________
* Assumes both light trucks in the collision were reduced by 100 pounds.

Table 2 shows the average fatality increase per 100-pound reduction in passenger cars.  The regression analyses are based exclusively on data for 4-door cars, excluding police cars.  During MY 1991-99, only 24 percent of new passenger cars were 2-door models, and fewer than 1 percent of new 4-door cars were police cars.  The upper section of Table 2 shows the effect in cars weighing 2,950 pounds or more (close to the median curb weight of cars throughout MY 1991-99).  As curb weight decreased by 100 pounds, fatality rates increased strongly in rollovers (4.70 percent), decreased non-significantly in pedestrian crashes (0.62 percent reduction), but increased moderately in all other crash modes (1.59 to 3.18 percent). In absolute terms, though, the largest increase was in collisions with LTVs (83 per year). The point estimate of the net change for all crash modes was an increase of 216 fatalities per year (interval estimate: 129 to 303).  The point estimate for the percentage change was an increase of 1.98 percent.  Those estimates were somewhat weaker than the effects in light LTVs but much stronger than the effects in heavy LTVs.

The lower section of Table 2 shows moderate-to-strong effects in every crash mode for cars weighing less than 2,950 pounds.  In rollovers and in collisions with heavy trucks and LTVs, fatality rates were 5 to 6 percent higher as cars got 100 pounds lighter.  Even in pedestrian collisions, fatality rates rose 3.48 percent.  No such increase of pedestrian fatalities was seen in the heavier cars or either group of LTVs.  The point estimate of the net change for all crash modes was an increase of 597 fatalities per year (interval estimate: 226 to 715), well over double the increase in the heavier cars or the lighter LTVs.  The point estimate for the percentage change was an increase of 4.39 percent.

The strong increase in pedestrian fatalities for the lightest cars is surprising.  At least at first glance, the weight of the vehicle shouldn't have had much effect on the fatality risk of pedestrians. Perhaps, heavier vehicles were simply driven better, even after adjusting for the drivers' age/gender, urban/rural and other factors.  For example, safety-conscious drivers might have selected heavier cars because they considered them safer. Heavier cars, more expensive on the average, might also have attracted higher-income owners with a more health-conscious, less risk-prone lifestyle. This study, however, found that light and heavy 4-door cars, pickup trucks and 4-door SUVs of MY 1991-99 all had remarkably similar incidence of high-risk driving behavior: drinking, speeding, previous crashes, license suspensions, etc.  (Two-door cars had substantially higher-than-average incidence of high-risk driving behavior, but they were not included in the data used to calibrate the weight-safety relationships.)  NHTSA research suggests that the geometry of small cars might, in fact, have increased the risk of serious injury to pedestrians (shorter hoods, more head impacts with the windshield frame).  Finally, small cars, because they felt more maneuverable, might even have induced drivers to weave in traffic or take other risks they would ordinarily have avoided in a larger vehicle.

We do not know how much of the observed effect in pedestrian crashes was due to self-selection – better drivers picking bigger cars – but we are confident that much of the effect, quite possibly even all of it was “real.”  Thus, the maximum proportion that was self-selection may have been as low as zero, but it was definitely less than 100 percent.  In the absence of evidence supporting any specific proportion between zero and 100 percent, this report takes the midpoint and assumes at most half the observed effect in pedestrian crashes was due to self-selection.

TABLE 2
FATALITY INCREASE PER 100-POUND WEIGHT REDUCTION, PASSENGER CARS

(Baseline = CY 1999 total fatalities, MY 1996-99/CY 1996-2000 fatality distribution)
 
Crash Mode Annual
Baseline
Crash
Fatalities
Effect (%) of
100-Pound Reduction
Annual Net
Fatality Change
Point
Estimate
Interval
Estimate
Point
Estimate
Interval
Estimate
CARS WEIGHING 2,950 POUNDS OR MORE
Principal rollover 715 4.70 2.40 to 7.00 34 17 to 50
Fixed object 2,822 1.67 0.63 to 2.71 47 18 to 76
Ped/bike/motorcycle 1,349 -.62 -1.83 to .59 -8 -25 to 8
Heavy truck 822 2.06 .67 to 3.45 17 6 to 28
Car < 2,950 1,342 1.59 .70 to 2.48 21 9 to 33
Car 2,950 +* 677 3.18 1.40 to 4.96 22 9 to 34
Light truck 3,157 2.62 1.74 to 3.50 83 55 to 110
OVERALL 10,884 1.98 1.19 to 2.78 216 129 to 303
 
CARS WEIGHING LESS THAN 2,950 POUNDS
Principal rollover 995 5.08 .87 to 7.55 51 9 to 75
Fixed object 3,357 3.22 .25 to 4.45 108 8 to 149
Ped/bike/motorcycle 1,741 3.48 .22 to 5.00 61 4 to 87
Heavy truck 1,148 5.96 2.50 to 7.68 68 29 to 88
Car < 2,950* 934 4.96 -.72 to 7.16 46 -7 to 67
Car 2,950 + 1,342 2.48 -.36 to 3.58 33 -5 to 48
Light truck 4,091 5.63 2.85 to 6.67 230 117 to 273
OVERALL 13,608 4.39 1.66 to 5.25 597 226 to 715
__________________
* Assumes both cars in the collision were reduced by 100 pounds.

If so, self-selection also played a role in the other crash modes, not just pedestrian crashes.  Therefore, the interval estimates of this study include not only sampling error but also an adjustment – up to half of the observed effect in pedestrian crashes – to account for possible effects due to self-selection.

The interval estimates in Tables 1 and 2 (and also Table 4) are defined as follows: the upper bound is the point estimate plus 1.96 standard deviations of sampling error (from various known sources).  The lower bound is the point estimate, minus 1.96 standard deviations of sampling error, minus half the observed pedestrian effect (and, in Table 1, minus an additional allowance for some uncertainty in the model formulation).  The interval estimates are a tool for gauging uncertainty, but they are not rigorous 95 percent confidence bounds.  When the range in the interval estimate includes zero, the point estimate can be called “not statistically significant.”  When the interval is entirely positive, or entirely negative, it provides some evidence that the observed effect is “real” – the tighter the interval, the stronger the evidence – but the intervals are not rigorous confidence bounds, as they would be, for example, in a simple, controlled experiment.

Table 2, showing a strong increase in fatality risk per 100-pound reduction in cars weighing less than 2,950 pounds, is based on an analysis including drivers of all ages.  When the analysis was limited to drivers age 60 or older, all the size-safety effects became even more severe, in some crash modes more than double.  That suggests older drivers had serious problems controlling the lightest cars and/or that the crash environment in light cars in some way amplified older occupants' general vulnerability to injury.

The point estimates in Tables 1 and 2 are approximately linear and additive.  If, in general, vehicles weighing W-100 pounds had on the average 1 percent higher fatality rates than vehicles weighing W, then vehicles weighing W-200 pounds would have had approximately 2 percent higher rates than vehicles weighing W.  The effect of reducing all LTVs by 100 pounds would have been close to the sum of the effects of reducing LTVs over 3,870 pounds and under 3,870 pounds by 100 pounds each: 71 + 234 = 305.

This study estimates a substantially larger fatality increase per 100-pound weight reduction than NHTSA's 1997 report.  A review of the 1997 report reveals flaws in the calibration procedure leading to a systematic underestimate of the size-safety effect in every crash mode, for both LTVs and cars.  This study's results supersede the 1997 report and, in particular, correct its findings on car-to-car crashes.  Table 2 now shows fatality risk in car-to-car crashes increased as car weight decreased, consistent with intuition and most of the literature.  The lighter cars had higher crash involvement rates and higher fatality risk, given a crash, for their own occupants.  That more than offset the reduction in fatality risk of occupants in the “other” car.

In summary, Tables 1 and 2 suggest that the association between curb weight and fatality risk in MY 1991-99 vehicles was weakest – in fact, nonsignificant – in the heavier LTVs.  It was strongest in the lighter cars.

Fatal-crash and fatality rates by vehicle type, model years 1996-99

LTVs of the 1990's included some models that had high rollover fatality rates per billion miles.  They also included models that, when they collided with other vehicles, the occupant fatality rate was high in the other vehicle.  These LTV models may be characterized as “rollover-prone and/or aggressive vehicles.”  The fatal-crash involvement rates and occupant fatality rates of different vehicle types were compared on as “level a playing field” as possible, by adjusting for differences in driver age/gender, annual mileage, vehicle occupancy (where appropriate), distribution of the mileage by urban/rural, speed limit, and other vehicle, driver and environmental factors – but not for vehicle weight.

The statistical approach, based on logistic regressions and data similar to the preceding analyses, was to compare fatal-crash rates per billion vehicle miles for ten groups of model year 1996-99 vehicles during calendar years 1996-2000: four size groups of 4-door cars, three size groups of 4‑door SUVs, two sizes of pickup trucks, and minivans.  All vehicles were equipped with air bags.  Heavy-duty (200/300-series) pickup trucks and full-sized vans were not included in this analysis.  A single “prorated fatal-crash rate” per billion vehicle miles, comprising all crash modes, was computed for each vehicle group, after adjustment for driver age/gender, urban/rural, and other factors.  The prorated fatal-crash rates included fatalities to occupants of the case vehicle, occupants of the other vehicles it collided with, and any pedestrians.  Each crash was weighted by the number of fatalities; however, in order to prevent double-counting, the number of fatalities in multivehicle crashes was divided by the number of cars/LTVs involved in the crash (e.g., in a 2‑vehicle crash, each vehicle was assigned half the crash fatalities).

Table 3 compares the average curb weights and the overall fatal-crash rates of the ten groups.  Groups that included numerous rollover-prone and/or aggressive vehicles in MY 1996-99 had greater fatal-crash rates.  For example, mid-size 4-door SUVs of model years 1996-99 had an average fatal-crash rate of 13.68.  Similarly, large SUVs and pickup trucks had higher fatal-crash rates than some groups of cars or minivans. The four vehicle groups with the lowest overall prorated fatal-crash rates in Table 3 were large cars (7.12), minivans (7.97), mid-size cars (9.46) and large (100-series) pickup trucks (9.56).  Very small 4-door cars had the highest rate (15.73).  However, by 1996-99, these cars only accounted for well under 1 percent of vehicle sales.

TABLE 3
ADJUSTED FATAL-CRASH INVOLVEMENT RATES
PER BILLION CASE VEHICLE MILES, BY VEHICLE TYPE

(Case vehicles are MY 1996-99 light trucks and 4-door cars with air bags in CY 1996-2000, adjusted for age/gender, rural/urban, day/night, speed limit, and other factors)
Vehicle Type and Size Average Curb Weight Prorated* Fatal Crash
Involvements Per Billion Miles
Very small 4-door cars 2,105 15.73
Small 4-door cars 2,469 11.37
Mid-size 4-door cars 3,061 9.46
Large 4-door cars 3,596 7.12
Compact pickup trucks 3,339 11.74
Large (100-series) pickup trucks 4,458 9.56
Small 4-door SUVs 3,147 10.47
Mid-size 4-door SUVs 4,022 13.68
Large 4-door SUVs 5,141 10.03
Minivans 3,942 7.97
* Each fatal crash involvement by a case vehicle is weighted by: the number of crash fatalities divided by the number of cars/LTVs involved in the crash.

Furthermore, 1996-99 SUVs had higher fatality risk for their own occupants than large cars or minivans.  Here, for example, are drivers' fatality rates per billion vehicle miles (adjusted for driver age/gender, urban/rural, and other factors):

Driver Fatalities per Billion Vehicle Miles
Very small 4-door cars11.56
Small 4-door cars7.85
Mid-size 4-door cars5.26
Large 4-door cars3.30
Compact pickup trucks6.82
Large (100-series) pickup trucks4.07
Small 4-door SUVs5.68
Mid-size 4-door SUVs6.73
Large 4-door SUVs3.79
Minivans2.76

The four vehicle groups with the lowest fatality rates for their own drivers were minivans (2.76), large cars (3.30), large SUVs (3.79), and large (100-series) pickup trucks (4.07).

Table 3 shows the fatal-crash rate was lower for small 4-door SUVs (10.47) than for mid-size 4‑door SUVs (13.68) in MY 1996-99. The drivers' fatality rate per billion vehicle miles was likewise lower in small SUVs (5.68) than mid-size SUVs (6.73).  This was the only exception to the customary trend, where larger size groups of the same vehicle type had lower fatal-crash rates and occupant fatality rates.

A more detailed comparison of the fatality rates of small SUVs, mid-size SUVs and mid-size cars of MY 1996-99 shows that rollovers and occupants of the “other” vehicle in 2-vehicle crashes accounted for the higher risk of the SUVs.  The small SUVs had much lower rollover fatality rates than the mid-size SUVs, although still high compared to the cars.  Similarly, the fatality rate for occupants of other vehicles, per billion case-vehicle miles, was substantially lower for the small-SUV case vehicles than for the mid-size SUVs, but still high compared to the cars.  By contrast, the fatality rates for the vehicles' own occupants in non-rollover crashes, per billion occupant miles, were fairly similar for the three types of vehicles, and actually lowest for the mid-size SUVs:

Fatalities per Billion Miles (Not Prorated)
Small
4-Door
SUVs
Mid-Size
4-Door
SUVs
Mid-Size
4-Door
Cars
Rollovers1.062.71.50
Occupants of other vehicles3.444.462.55
Occupants of case vehicle, in non-rollovers4.383.954.63

As stated above, LTVs of the 1990's included numerous rollover-prone and/or aggressive vehicles.  However, by 1996-99, several new models of small 4-door SUVs with improved rollover stability had been introduced.  For example, one model was measured by NHTSA and rated substantially more stable than most mid-size or large SUVs of the mid-1990's. The above statistics suggest that small 4-door SUVs of 1996-99 may have been the beginning of a new generation of more stable, less aggressive vehicles with lower fatal-crash rates.  This trend appears to have continued and expanded since 1999, comprising entirely new designs such as car-based “crossover” SUVs and less sweeping redesigns of existing LTVs.  Indeed, rollover-resistance ratings published by NHTSA in 2001 show new models of SUVs in all three size groups with greater stability than the models they superseded.  Also, new technologies such as “blocker bars” have been introduced on some LTVs to make them less aggressive in collisions with other vehicles.

Table 3's adjusted fatal-crash rates for the ten groups of MY 1996-99 vehicles can be applied to the baseline of all CY 1999 crash fatalities in the United States (all model years) to estimate the annual change in fatalities if the mix of vehicle types on the road in 1999 had changed – i.e., if the public had purchased more vehicles of one type and fewer of another.  Table 4 estimates the reduction in fatalities given nine hypothetical scenarios in which the MY 1996-99 vehicle mix changed to more of one type of car or minivan and fewer of one type of SUV or pickup truck.  For comparison purposes, it also considers one more scenario: a change from very small 4-door cars to small 4-door cars.   It estimates what might have been the annual effect of a “one percentage point change” in the vehicle mix.  For example, during MY 1996-99, mid-size 4-door SUVs accounted for 8 percent of new-vehicle sales, and large 4-door cars, 12 percent.  Table 4 assumes MY 1996-99 vehicles constituted the entire on-road fleet and estimates the effect on fatalities in baseline CY 1999 if the vehicle mix had instead consisted of 7 rather than 8 percent mid-size SUVs and 13 rather than 12 percent large cars. 

The first nine scenarios in Table 4 all combine a likely reduction in fatalities with a reduction in vehicle weight.  The point estimates of the fatality reductions in Table 4 range from 29 to 200 per year, per percentage point change in the vehicle mix.  The reductions for the scenarios involving changes from mid-size or full-size SUVs to cars or minivans have wholly positive interval estimates.  By comparison, the change from very small cars to small cars is estimated to reduce fatalities by 156 (with a weight increase).  Of course, all these estimates are specifically for MY 1996-99, a time when numerous pickup trucks and SUVs were rollover-prone or aggressive.

TABLE 4
CHANGE IN FATALITIES PER YEAR
GIVEN A ONE-PERCENTAGE POINT CHANGE IN THE ON-ROAD FLEET
FROM MY 1996-99 SUVs AND PICKUPS TO CARS OR MINIVANS

(Baseline = CY 1999 total fatalities, MY 1996-99/CY 1996-2000 fatality distribution)
Fatality Reduction Per Year Versus Weight Reduction Per Vehicle Point Estimate Interval Estimate
Small 4-dr SUVs Mid-size 4-dr cars 86 29 –72 to 64
Mid-size 4-dr SUVs Mid-size 4-dr cars 961 129 79 to 152
Large 4-door cars 426 200 140 to 220
Minivans 80 174 130 to 201
Large 4-dr SUVs Large 4-door cars 1,545 101 27 to 133
Minivans 1,199 72 17 to 111
Compact pickups Mid-size 4-dr cars 278 103 –57 to 163
Large pickups* Large 4-dr cars 862 127 –65 to 194
Minivans 516 82 –80 to 161
…………………………………………………………………………………………………….
Very small 4-dr cars Small 4-dr cars –364 156 –40 to 231
*Large, standard-duty (100 series) trucks.  Excludes heavy-duty 200/300 series pickup trucks.

Car-light truck compatibility

NHTSA has been researching car-light truck compatibility since 1993.  In collisions between LTVs and cars, approximately 80 percent of the fatalities are occupants of the cars.  The objective is to reduce fatality risk in the car, without increasing risk in the LTV.  That may require increasing crashworthiness of the car, but it might be easier to accomplish by reducing the aggressiveness of the LTV, or by a judicious combination of both. Of course, MY 1991-99 LTVs usually outweighed cars but, in addition, there were two sources of mismatch between LTVs and cars that made the LTVs extra “aggressive” when they hit the cars:

·       Structural incompatibility: the LTV's front was more rigid than any part of the car

·       Geometric incompatibility: the LTV's front applied its force at a height above the car's structures designed to withstand force

The databases and logistic-regression analysis methods used to study vehicle weight and fatality risk were also suitable for investigating car-LTV compatibility.  Fatality rates in 2-vehicle collisions, per billion miles of each vehicle, were calibrated as a function of the body type and curb weight of each vehicle (MY 1991-99 in CY 1995-2000), the age/gender of each driver, urban/rural location, speed limit, and other vehicle, driver and environmental factors.  Once again, the objective was to compare the fatality rates in car-to-car and LTV-to-car collisions on as “level a playing field” as possible.  The first goal was to quantify the extra aggressiveness of MY 1991-99 LTVs relative to MY 1991-99 cars of the same weight.  The analysis focused on collisions where the struck vehicle was a car, and the striking vehicle was a car, pickup truck, SUV or minivan.  Table 5 shows how much the fatality risk of the driver of the struck car increased when the striking vehicle was an LTV. 

The first row of Table 5 evaluates left-side impacts to the struck car by the front of the striking vehicle.  Left-side impacts are the most dangerous for drivers, because they sit on the left.  When the striking vehicle was a passenger car of weight W, let us say the driver of the struck car had fatality risk index 100.  When the striking vehicle was a pickup truck of weight W, the fatality risk of the driver of the struck car increased to 177. In other words, it was almost twice as dangerous, on a per-mile basis, to be hit on the left side by a pickup truck as by a car of the same weight as that pickup truck.  When the striking vehicle was an SUV, the risk index was 235. Even when the striking vehicle was a minivan, the risk index was 130, higher than when it was a car.  The risk indices for MY 1991-99 pickup trucks, SUVs and minivans were all significantly higher than 100 in front-to-left impacts.

The second row of Table 5 considers head-on (front-to-front) collisions.  Here, LTVs were much less aggressive.  The risk index for the driver of the struck car was significantly higher than 100 only when the striking vehicle was an SUV (132).  Impacts by pickup trucks and minivans had risk indices just slightly, and not significantly, above 100. When they hit cars on the right side or rear, the aggressiveness of LTVs was higher than in head-on collisions, but not as high as when they hit the car on the left side.  The third row of Table 5 shows that risk indices for pickup trucks and SUVs were both significantly above 100.

TABLE 5
AGGRESSIVENESS OF MY 1991-99 LTVs IN IMPACTS WITH MY 1991-99 CARS
AFTER ADJUSTMENT FOR THE STRIKING VEHICLE'S WEIGHT**

(Fatality risk index of the driver of the struck car, by striking vehicle type;
MY 1991-99 vehicles in CY 1995-2000 crashes)
Striking Vehicle's
Front Impacted the
Struck Car on the
Driver Fatality Risk Index
in the Struck Car by
Striking Vehicle Type
Car Pickup SUV Minivan
Left side 100 177* 235* 130*
Front (head-on collision) 100 114 132* 104
Right side or rear 100 139* 162* 125
Anywhere 100 139* 171* 116*
*Significantly greater than 100.
**For example, in a front-to-left impact, if the risk for the driver of the struck car was 100 when the striking vehicle was a 3,500 pound car, the risk increased to 177 when the striking vehicle was a 3,500 pound pickup truck.

Combining all of the preceding crash modes, the last row of Table 5 shows that, overall, every type of MY 1991-99 LTV was significantly more aggressive than a passenger car.  All of these indices apply specifically to MY 1991-99 vehicles and could change for more recent LTVs as new technologies or designs are introduced to reduce aggressiveness in collisions.

The second analysis goal was to test for association between the aggressiveness of model year 1991-99 LTVs in crashes and physical parameters describing the structural rigidity and geometry of the trucks.  Two parameters were readily available, because NHTSA measures them during its frontal crash tests in the New Car Assessment Program (NCAP).  They are:

Frontal rigidity:     The average slope of the force-deflection profile maintained for at least 150 millimeters during the vehicle's initial crush in an NCAP frontal impact with the barrier.

Height-of-force:    The average height-of-force measured by load cells set at various height levels in the NCAP barrier.  It is the weighted average of the effective height of the applied force on the barrier face over the duration of the impact.  

Association was tested by limiting the preceding logistic-regression analyses to crashes where the striking vehicle was an LTV (and the struck vehicle was a car), and adding the two parameters to the regression.  In front-to-left impacts, there was a statistically significant association between the driver's fatality risk in the struck car and the difference in the heights-of-force of the striking and struck vehicles: the greater the height mismatch between the LTV and the car, the greater the fatality risk of the driver of the car.  In head-on collisions, the LTV's frontal rigidity was significantly associated with the car driver's fatality risk: the more rigid the LTV, the greater was the fatality risk of the car driver.

The analyses accept as a given that model year 1991-99 LTVs were, on the average, more aggressive than cars.  These somewhat exploratory findings suggest that the LTVs with the tallest and most rigid frontal structures were even more aggressive than the other LTVs.

These analyses of car-LTV compatibility are intended to supplement and corroborate, not supersede NHTSA's previous work on that subject.  This study's approach, based on fatality rates per billion miles, controlling for each vehicle's weight, each driver's age and gender, urban/rural, and other factors, helps compare fatality rates in car-car and car-LTV collisions “on a level playing field.”  On the other hand, the per-mile approach does not necessarily separate crash-proneness from crashworthiness effects (a disadvantage here, although it was a plus in the size-safety analyses).  It is best to look at these results in combination with NHTSA's previous findings on car-LTV compatibility.  In addition, the statistical findings that show an association of these two parameters with extra aggressiveness of LTVs do not, by themselves, guarantee that these two parameters “caused” the aggressiveness, or that they are the parameters that best explain or measure aggressiveness.  Crash testing with existing and, eventually, modified vehicles is another essential step in learning what makes LTVs aggressive.

Limitations of the analyses

This study is a cross-sectional analysis of the crash fatality rates per billion miles of real MY 1991-99 vehicles in CY 1995-2000: light, mid-size and heavy passenger cars, pickup trucks, SUVs and vans. Statistical tools calibrated the relationships between vehicle weight and fatality rates – the average increase for vehicles weighing W-100 pounds relative to vehicles weighing W pounds – and the differences between cars and LTVs, after controlling for driver age/gender, urban/rural, and other vehicle, driver and environmental factors.  The results specifically describe the performance of MY 1991-99 vehicles; the impact of new designs or technologies in more recent vehicles will be revealed as they now accumulate on-the-road experience.

The analysis is not a “controlled experiment.”  People are largely free to pick whatever car or LTV they wish.  Owner characteristics and vehicle use patterns can and do vary by vehicle weight and type.  This study adjusts for differences in age/gender, urban/rural driving, and other factors, and tries to gauge uncertainty due to less tangible variations in “how well people drive.”  But, ultimately, we can never be sure that a 30-year-old male operating a large LTV on an urban road at 2:00 p.m. in a Western State drives the same way as a 30-year-old male operating a smaller LTV/light car/heavy car at a similar roadway, time and location.  The interval estimates in this study try to depict likely ranges of uncertainty in the principal findings, but rigorous “95 percent confidence bounds” do not apply here, as they would, for example, in a simple, controlled experiment.

pdf graphicThe complete report is available here in pdf format.

Return to Regulatory Evaluation Page