Federal Motor Vehicle Safety Standards; Rollover Prevention

DEPARTMENT OF TRANSPORTATION

National Highway Traffic Safety Administration

49 CFR Part 571

[Docket No. 91-68; Notice 5]

RIN 2127-AC64

Federal Motor Vehicle Safety Standards; Rollover Prevention

AGENCY: National Highway Traffic Safety Administration (NHTSA), DOT.

ACTION: Denial of petitions for reconsideration.

-----------------------------------------------------------------------

SUMMARY: This notice announces the denial of petitions for

reconsideration of the agency's decision to terminate rulemaking to

develop a vehicle rollover stability standard.

FOR FURTHER INFORMATION CONTACT: The following persons at the National

Highway Traffic Safety Administration, 400 Seventh Street, S.W.,

Washington, D.C. 20590:

For non-legal issues: Gayle Dalrymple, Office of Crash Avoidance

Standards, telephone (202) 366-5559, facsimile (202) 366-4329.

For legal issues: Steve Wood, Office of the Chief Counsel, NCC-20,

telephone (202) 366-2992, facsimile (202) 366-3820.

SUPPLEMENTARY INFORMATION:

I. 1994 Notice Terminating Rulemaking on a Vehicle Rollover

Stability Standard

On June 28, 1994, NHTSA published a notice in the Federal Register

announcing two agency actions: (1) the termination of rulemaking to

develop a Federal Motor Vehicle Safety Standard on vehicle rollover

stability; and (2) the proposal of a consumer regulation for labeling

vehicles with rollover stability information. (59 FR 33254)

In the portion of the 1994 notice terminating rulemaking, the

agency examined the suitability of using a variety of vehicle stability

metrics 1 as a basis for a rollover standard. NHTSA concluded that

two such metrics, tilt table angle (TTA) 2 and critical sliding

velocity (CSV),3 can each separately account for approximately

half of the variability in rollover risk in single vehicle accidents

remaining after considering driver, roadway, and environmental factors.

NHTSA stated:

1 A vehicle stability metric is a measured vehicle

characteristic that is analyzed to determine whether it is related

to a vehicle's likelihood of rollover involvement.

2 The tilt table test involves placing the vehicle on a

platform which is then tilted about an axis parallel to the

vehicle's longitudinal axis. TTA is the angle at which the last tire

on the upper side of the platform loses contact with the platform

and the vehicle begins to fall off the platform. This metric is

influenced by changes in a vehicle's mass, center of gravity height,

track width, and suspension movement, all of which are physically

related to rollover stability.

3 Critical sliding velocity includes the roll moment of

inertia as well as the various static factors included in tilt table

angle. CSV is calculated from an equation which can be found in the

June 28, 1994 notice, as corrected on July 26, 1994 (59 FR 38038).

---------------------------------------------------------------------------

The suitability of a vehicle safety standard based on rollover

stability depends on the importance of rollover stability, as

represented by a vehicle metric, relative to other rollover

influences, such as vehicle handling properties, vehicle condition,

the nature of the roadway and shoulder terrain, and driver behavior.

The agency sought to determine whether vehicle stability metrics are

significant variables in a statistical model of the risk of

rollover. If they are, then a standard regulating stability might be

justified, depending on the results of a comparison of benefits and

costs for such a standard.

After analyzing a number of static and dynamic rollover metrics,

the agency concluded that two vehicle metrics, tilt table angle and

critical sliding velocity, can account for about 50 percent of the

variability in rollover risk in single vehicle accidents, after

considering driver, roadway, and environmental factors. (Rollover

risk is the number of single vehicle rollovers involving a

particular make/model divided by the number of single vehicle

crashes of all types involving the same make/model.) This

statistical analysis was conducted on all light duty vehicles

treated as a group. However, analysis of accident data indicated

that certain subgroups of light duty vehicles are more likely to

roll over than other subgroups. For example, sport utility vehicles

and compact pickup trucks tend to be the most likely vehicles to

roll over. Large passenger cars tend to be the least likely to roll

over.

59 FR 33254, at 33258.

While NHTSA concluded that the two vehicle stability metrics were

of some value in estimating the likelihood that a single vehicle

accident involving a particular model of vehicle would result

[[Page 28551]]

in a rollover, the agency emphasized that analyses also ``show that

other factors in addition to those analyzed are affecting rollover

risk.'' (Id., at 33260) As the agency noted, ``[t]he suitability of a

vehicle safety standard based on rollover stability depends on the

importance of rollover stability, as represented by a vehicle metric,

relative to other rollover influences, such as vehicle handling

properties, vehicle condition, the nature of the roadway and shoulder

terrain, and driver behavior.'' (Id., at 33258) In other words, the

issue was not simply whether there is a statistical relationship, but

also whether that relationship is strong enough, considering other

influences, so that improvements in the stability metrics, especially

relatively small improvements, would generate benefits commensurate

with the costs. If the relationship is not sufficiently strong, even

significant changes in the stability metrics may be overwhelmed by the

other influences and thus fail to cause a significant change in

rollover experience.

The agency concluded that while each of the stability metrics has

some causal relationship to the potential for rollover and a

statistical relationship to real-world rollover frequency, a standard

based on either of the metrics would yield measurable benefits only if

it required that the metrics be increased to an extent that would

impose excessive costs and necessitate radically redesigning one or

more types of light trucks.4 The agency reached this conclusion

after examining the merits of establishing a single rollover standard

for all light duty vehicles (i.e., passenger cars and light trucks).

---------------------------------------------------------------------------

\4\ The term ``light trucks'' includes sport utility vehicles,

vans, and pickup trucks with a gross vehicle weight rating of 4,536

kilograms (10,000 pounds) or less.

---------------------------------------------------------------------------

With respect to a single standard, the agency stated:

The agency also determined that, considering the costs and

benefits involved, proposing a safety standard specifying a single

minimum stability value for both cars and light trucks could not be

justified. While light trucks have lower stability measurements than

cars do, the greatest number of rollover-related deaths and injuries

occur in passenger cars because of their larger population size.

Therefore, if the agency wished to set a stability minimum high

enough to realize significant reductions in the number of fatalities

in all light duty vehicles, it would have to set the minimum above

the stability number of most light trucks. The costs of such a

standard, in terms of the cost of vehicle redesign and the loss of

consumer-desired attributes, were determined to be very high, as

entire classes of light trucks would probably need to be

substantially redesigned to meet such a standard. This redesign

could result in the elimination of some vehicle types, e.g., sport

utility vehicles, as they are known today.

Id., at 33258.

To avoid such drastic consequences for light trucks, the agency

considered whether it would be appropriate to set one standard for cars

and separate standards for various classes of light trucks.5 NHTSA

concluded that it was not appropriate. Since its analysis of the

ability of the two vehicle stability metrics to account for the

variability in rollover risk in single vehicle accidents was conducted

on all light duty vehicles as a group, the agency examined the ability

of the metrics to account for variability within individual subgroups

of those vehicles. Regarding the results of that examination, NHTSA

stated:

\5\ Id., at 33258.

---------------------------------------------------------------------------

[I]t was necessary to determine whether either of the stability

metrics exhibited sufficiently high levels of correlation to assure

the agency that a requirement applying to only one class of vehicle

would be expected to reduce the incidence of rollovers for vehicles

in that class. * * * [T]he agency found that the statistical

correlations of the metrics with rollover accident data within a

class of vehicles was not so consistent as for all vehicles grouped

together. This weakening of the predictive ability of the metric is,

to some extent, the result of the smaller range of the metric within

any class of vehicles together with the inherent variability in the

data. Based on this analysis, and the general analysis of costs and

benefits discussed later, the agency determined that proposing a

standard specifying one minimum stability value for cars and others

for various classes of light trucks could not be justified.

Id., at 33528.

II. Petitions for Reconsideration of Decision To Terminate

Rulemaking

In July 1994, the agency received two petitions for reconsideration

of its decision to terminate rulemaking on a rollover stability

standard. One petition was submitted by Advocates for Highway and Auto

Safety and the Insurance Institute for Highway Safety (Advocates/IIHS)

and the other by Randall and Sandy Vance, Doug White, and Robert and

Glenda Cammack (Vance, et al.). Both petitions asked NHTSA to

reconsider its decision to terminate rulemaking to establish a minimum

standard for vehicle rollover stability. The Vance et al. petition

expressed general disagreement with that decision, while the Advocates/

IIHS petition identified detailed points of disagreement. For this

reason, unless otherwise specified, references below to ``the

petition'' or ``the petitioners'' are references to the Advocates/IIHS

petition.

While the petitioners made numerous contentions, they focused on

four general areas: the character of a reasonable rollover standard,

the agency's statistical analysis of how a standard could be selected,

the agency's benefit calculations, and the agency's statements

concerning cost burden to the manufacturing industry. The following is

a summary of the more important contentions addressed in this notice

and the appendix to this notice:

NHTSA should have more thoroughly considered establishing

separate standards for separate classes of vehicles.

To achieve a better relationship between costs and

benefits, NHTSA should have considered the alternative of setting a

standard for the most rollover-prone vehicles within one or more of the

following groups: sport utility vehicles (SUVs), vans, and pickup

trucks.

Compact SUVs 6 are the most rollover prone group of

light duty vehicles.

---------------------------------------------------------------------------

\6\ The vehicles considered compact SUVs in NHTSA's analysis

were: Ford Explorer, Chevy S10 Blazer, Jeep Cherokee, Jeep Wrangler,

Toyota 4-Runner, Nissan Pathfinder, Geo Tracker, GMC S-15 Jimmy

(essentially a twin of the Blazer), Isuzu Trooper, Isuzu Rodeo,

Suzuki Sidekick (essentially a twin of the Tracker), Mazda Navaho

(essentially a twin of the 4WD Explorer), Mitsubishi Montero, Isuzu

Amigo, and Suzuki Samurai.

---------------------------------------------------------------------------

Minor vehicle changes (e.g., suspension changes) could be

used to achieve stability improvements at reasonable cost.

NHTSA did not provide any factual support for its

assertion that there are serious safety problems associated with

improving vehicle stability metrics through suspension changes.

NHTSA did not explain the nature and extent of the major

design changes that it said were necessary to meet any stability

metric, nor how much such changes would cost.

The level of projected benefits of a rollover standard was

understated by the agency because it:

used average class values in lieu of model specific

rollover accident data for the rollover experience of some vehicle

models;

used inappropriate statistical measures; and

viewed rollover prevention as accident mitigation instead

of accident prevention.

Although Congress did not mandate the issuance of a

rollover stability standard, it expected that such a standard would be

issued.

Contrary to NHTSA's position, the statute governing the

agency's vehicle

[[Page 28552]]

safety rulemaking readily permits the elimination of a class of

vehicles widely accepted in the marketplace.

The agency may not consider the policy concerns underlying

the Regulatory Flexibility Act without preparing a regulatory

flexibility analysis.

III. Response To Petitions for Reconsideration

In response to the petitions, the agency has reconsidered its

decision to terminate rulemaking on a rollover stability standard. As

explained below, the agency is, on reconsideration, reaffirming that

decision.

The petitions raise several points that are not disputed by NHTSA;

however, they do not compel the conclusion that NHTSA should establish

a rollover standard based on vehicle stability metrics. For example,

the agency agrees that single vehicle rollover is a significant safety

problem. NHTSA also agrees that the two vehicle stability metrics are

useful in estimating the likelihood that a single vehicle accident

involving a particular model of vehicle will result in a rollover.

Finally, the agency agrees that it is appropriate in determining

the desirability of a rollover standard to consider a rollover standard

regulating vehicles in the most rollover-prone groups. While the 1994

notice focused primarily on the approach of a single standard for all

light duty vehicles, the agency did analyze separate standards for

separate classes of vehicles. The notice explained that the predictive

ability of the vehicle stability metrics decreased as the vehicle

population was divided into smaller groups. As noted above, the agency

concluded that ``a standard specifying one minimum stability value for

cars and others for various classes of light trucks could not be

justified.'' (Id., at 33257). Since the petitioners suggest issuing a

rollover standard regulating the most rollover-prone vehicles, NHTSA

has focused on such an approach in responding to the petitions for

reconsideration. The agency agrees with the petitioners that, in

theory, the comparatively high rollover rate of compact SUVs makes a

standard regulating that group of vehicles appear more likely to

generate benefits commensurate with its costs than would a standard

regulating any other group of vehicles.

These areas of agreement are insufficient, however, to lead the

agency to the conclusion reached by petitioners. To the contrary, the

agency's detailed analysis below of a rollover stability standard based

on TTA or CSV demonstrates that the costs and other impacts of such a

standard manifestly outweigh the estimated but uncertain benefits.

A general response to the petitioners' arguments appears below.

Certain issues are covered in greater detail in the Appendix to this

notice.

IV. Rationale for Reaffirming Decision To Terminate

A. Summary

Following its examination of the arguments raised by the

petitioners, the agency has revisited and, in some respects expanded,

its rationale for terminating rulemaking on a vehicle stability

standard. The agency again concludes that it is not appropriate to

establish a vehicle rollover stability standard based on a vehicle

stability metric.

If a stability standard were set at a level that would require only

minor vehicle changes in order for the affected models to achieve

compliance, the standard would not produce any safety benefits. Minor

vehicle changes, which consist predominately of suspension changes,

would not produce significant improvements in the vehicle stability

metrics and would not be likely to result in any reductions in

fatalities and injuries.\7\ Moreover, there is reason to conclude that

such suspension changes would, in fact, produce negative safety side

effects.

---------------------------------------------------------------------------

\7\ As noted above, the agency stated in the 1994 notice that a

standard limited in its application to a vehicle subgroup (e.g.,

sport utility vehicles) is particularly unlikely to reduce

fatalities and injuries given the weaker statistical relationships

between the stability metrics and the rollover involvement for

vehicle subgroups. (Id., at 33528)

---------------------------------------------------------------------------

If a stability standard were set high enough to require significant

improvements in the vehicle stability metrics, it would necessitate

full vehicle redesigns and major vehicle changes. However, the safety

benefits of such changes would nevertheless be relatively modest.

Moreover, the overall costs and loss of consumer choice resulting from

full vehicle redesigns involving major vehicle changes would be

substantial and excessive. On balance, the potential for improved

vehicle safety associated with such improvements in the vehicle

stability metrics is not sufficiently large to justify such redesigns.

B. Vehicle Changes To Increase Vehicle Stability Metrics

There are two general categories of vehicle changes that would

increase the vehicle stability metrics (TTA and CSV). One consists of

relatively minor vehicle changes (i.e., suspension changes); the other,

of major vehicle changes (i.e., widening the vehicle track and lowering

the center of gravity) that could only be achieved through full

redesign of the vehicle. The petitioners appear to believe that a

vehicle can be redesigned so it will be significantly less likely to

roll over, that the means for accomplishing this will be ``invisible''

to the consumer, and that the vehicle will look and function as it did

before the redesign. As discussed below, redesigning a vehicle to

significantly reduce its likelihood of rolling over necessarily

involves making fundamental changes in the vehicle's dimensions (making

it wider, longer, lower, heavier) and compromising its utility to

consumers (e.g., by reducing its fuel efficiency, ground clearance,

load-carrying capacity, off-road capability, or driveability on snowy

roads).

1. Minor Vehicle Changes To Increase Vehicle Stability Metrics

Minor vehicle changes have very little effect on the vehicle

stability metrics. Moreover, they do not result in net safety

improvements.

As the petitioners correctly point out, the Preliminary Regulatory

Evaluation (PRE) for the 1992 Advance Notice of Proposed Rulemaking

(ANPRM) suggested that there were grounds for optimism about the

ability of minor vehicle changes, such as suspension tuning, to affect

stability metrics and improve rollover stability. (57 FR 242; January

3, 1992) However, after reviewing the comments on the 1992 ANPRM, the

agency concluded in the 1994 notice that minor vehicle changes could

not, in fact, significantly affect the vehicle stability metrics.

Comments from Advocates itself,\8\ as well as Ford and General Motors,

on the ANPRM indicate that suspension changes result in very little

improvement in rollover stability.

---------------------------------------------------------------------------

\8\ In commenting on the ANPRM, Advocates indicated that it did

not share the agency's optimism at that time about the desirability

of relying on suspension changes to improve rollover stability

metrics. Advocates commented that the selection of TTR as the

parameter to be regulated would ``permit a manufacturer to attempt

manipulation of other stability-related elements of the vehicle's

design, such as its suspension, in order to secure a barely passing

tilt- table score.'' It also expressed concern that the agency ``may

be already tending towards selection of TTRs [see footnote 12] that

will not move the industry towards safer overall vehicle designs,

particularly with regard to wheelbase, width, length, and center of

gravity height, but rather will encourage the perpetuation of the

status quo designs especially with regard to very small cars, small

pickups, and SUVs that will continue to show high rollover

propensities.''

---------------------------------------------------------------------------

Moreover, vehicle rollover stability is not the same as vehicle

handling and control. Some measures that improve

[[Page 28553]]

TTA or CSV do not necessarily result in improved directional control

and stability. Available information suggests that directional control

and stability would be adversely affected as a result of relying upon

suspension changes to make small increases in the vehicle stability

metrics. This information was supplied in comments from Advocates,

Ford, and General Motors on the ANPRM expressing concern with the side

effects of suspension changes to improve TTA.

For example, Ford used a computer simulation of a compact pickup

truck to evaluate the effect of a series of suspension changes on

directional stability and side-to-side load transfer in cornering.

(Docket 91-68-N01-21) Ford evaluated substantial suspension changes,

including a 30 percent increase in spring rates, removal of stabilizer

bars, and a change in the front suspension roll center by 1.5 inches.

It also examined a ride height change that would lower the center of

gravity by 0.5 inch. Ford noted that, in general, tuning a suspension

system such that both the front and rear tires lift from the tilt table

simultaneously would maximize the TTA. However, this optimization

requires either decreasing the front roll stiffness (by removing the

front stabilizer bar), or increasing the rear roll stiffness (by using

a 30 percent greater rear spring rate). The simulation showed that,

among the suspension changes examined by Ford, these two changes made

the greatest improvements in TTA (an increase of 0.62 and 0.55 degrees,

respectively). However, these changes were also shown to alter

directional stability toward oversteer (i.e., these changes tend to

make a vehicle turn more sharply than a driver intends). Ford's

simulation showed that other suspension changes, such as an increase in

front spring rate or a decrease in front roll center height, could

increase TTA (to a lesser degree than those mentioned above), while

altering directional stability toward understeer (i.e., these changes

tend to make a vehicle turn less sharply that a driver intends). The

only minor change mentioned by Ford in its comment which improved TTA

without influencing directional stability was lowering the vehicle c.g.

height by 0.5 inch, resulting in only a 0.17 degree increase in TTA.

Based on its consideration of such comments, the agency concludes

that suspension changes would not produce significant improvements in

rollover stability and would have the potential to cause undesirable

changes in directional stability and handling, which in turn could lead

to an increase in crashes. In view of this conclusion, the agency has

not examined whether those changes could be made at a reasonable cost,

since they are unlikely to yield net safety benefits.

2. Major Vehicle Changes To Increase Vehicle Stability Metrics

Thus, significant improvements to the vehicle stability metrics

could be achieved only through making major changes to the vehicle

chassis and body to increase the track width and/or lower the center of

gravity. These major changes would require full vehicle redesigns that

substantially change the parameters affecting vehicle stability

metrics. The necessary extent of such redesigns is illustrated in the

following example. Given that the center of gravity height for a

typical compact SUV 9 is 27 inches, to raise its TTA (42.9

degrees) to that of the typical full- size SUV (46.4 degrees), it would

be necessary to increase the track width (i.e., the distance between

the left and right tires on an axle) more than 6 inches. Further, such

a track width increase would require a corresponding wheelbase (i.e.,

the distance between the front and rear axles) lengthening of 10 inches

to retain the braking stability of the smaller SUV. As noted later in

the sections regarding cost and impact on consumer choice, such

modifications would eliminate most of the compact SUVs as they

currently exist, converting the typical compact SUV into a full-size

SUV.

---------------------------------------------------------------------------

\9\ A ``typical compact SUV'' and a ``typical full-size SUV''

are hypothetical vehicles with the average TTA and dimensions of all

the vehicles in their class.

---------------------------------------------------------------------------

Citing the example of the GMC Jimmy, which was redesigned for 1995,

petitioners argued that vehicle manufacturers can gradually redesign

their compact SUVs so as to increase their vehicle stability. The

petitioners presented an article from Automotive News stating that the

new Jimmy is longer, lower, and wider than its predecessor.10 The

petitioners further attributed to the new Jimmy ``a chassis

modification that can result in better stability metrics and in lower

rollover crash rates.'' 11

---------------------------------------------------------------------------

\10\ Advocates/IIHS petition, attachment 2.

11 Advocates/IIHS petition, page 18.

---------------------------------------------------------------------------

NHTSA draws a very different lesson than do petitioners from the

example of the Jimmy. In the agency's view, the petitioners

underestimate the extent to which the parameters affecting a vehicle's

stability metrics must be changed to significantly improve those

metrics. As explained below, the overall lessons of the new Jimmy are

that even a significant partial redesign of a vehicle will change its

vehicle stability metrics little in the absence of major changes to the

vehicle's c.g. height and track width, and that even minor changes in

those parameters may come at the cost of adversely affecting other

attributes desired by consumers. For example, the new Jimmy is heavier

and more costly than the prior model.

The agency agrees that the vehicle stability metrics of the new

Jimmy are likely to be somewhat better than those of the old Jimmy.

Although the agency has no TTA or CSV data on the new design, its lower

body height and wider track suggest that it has a slightly better TTA

than its predecessor and its longer, wider, and heavier body suggests

that it may have a greater roll moment of inertia and, therefore, a

slightly greater CSV.

However, the increases in the Jimmy's vehicle stability metrics are

likely to be very small. The reason is that the changes made to the

parameters affecting those metrics were relatively minor. Although the

changes increased the size and weight of the Jimmy, the magnitude of

those changes fell short of the levels needed to make a significant

improvement in its TTA or CSV. The body height of the 2WD model was

reduced by 1.6 inches, but the associated reduction in center of

gravity height is likely to be much less, since the location of the

engine, drive train, suspension, and passenger accommodation component

masses remained unchanged. The height reduction of the 4WD model was

only 0.8 inches. Likewise, the body width was increased by 2.4 inches,

but the front and rear track widths of the 4WD model were increased

less: 1.6 and 1.0 inches, respectively. The 2WD model track width

increases were even less: 0.9 inch at the front and 0.5 inch at the

rear.

Taken together, these changes to the Jimmy's parameters affecting

the rollover stability metrics are very minor compared to the ones

described above as being necessary for a typical compact SUV to achieve

a TTA of 46.4 degrees. Thus, these changes predict at best a very small

improvement in TTA or CSV.

The impact of such small improvements in vehicle stability metrics

on rollover risk is unknown. Since this is a new model for 1995,

neither NHTSA nor the petitioners have data on the rollover experience

of the new Jimmy. There is no way to know at this time if the changes

will actually lead to a reduced risk of rollover.

C. Benefits of Improvements in Vehicle Stability Metrics

NHTSA's 1994 notice estimated that the benefits of a rollover

standard requiring a TTA of 46.4 degrees for all light duty vehicles

included a modest

[[Page 28554]]

amount of benefits for compact SUVs. The agency's estimate that 63

fatalities and 61 serious injuries might be prevented for all light

duty vehicles included the prevention of 31 fatalities and 22 serious

injuries for compact SUVs. The potential compact SUV benefits were

predominately attributable to those particular compact SUV models that

would require significant changes in track width and/or center of

gravity height to achieve the required TTA.

As part of its review of the petitions, the agency recomputed its

estimate of the benefits of making significant design changes in order

to raise the TTR 12 of compact SUVs to 1.05,13 using data

that were not available for some makes and models when the analysis was

done for the 1994 notice.14 The agency estimates that, using the

more current data and certain optimistic assumptions (discussed below),

22 serious injuries and 32 fatalities might be prevented annually if

all new compact SUVs were redesigned to the extent necessary so that

each vehicle in that class had a TTR of 1.05 and if all existing

compact SUVs with a lower TTR were retired from the vehicles-in-use

fleet. The potential benefits for a rollover stability standard are

computed by considering:

---------------------------------------------------------------------------

12 TTR is the tangent of TTA. In its analysis prior to

the 1994 notice, the agency used TTR. Because TTA is an easier

concept to depict on labels for the general public, the agency

proposed the use of TTA rather than TTR for the vehicle label under

a consumer information regulation that was proposed in the 1994

notice. NHTSA used TTA throughout the 1994 notice for that reason.

However, NHTSA has not converted the TTR values to TTA values when

discussing its statistical and benefits analyses in this document.

13 A TTR of 1.05 is the equivalent of a TTA of 46.4

degrees. On page 33261 of the 1994 notice, the agency explained

that, if the agency were to adopt a rollover stability standard

applicable to all vehicles, a TTA of 46.4 degrees was the highest

practicable standard. The agency explained that a TTA of 46.4

degrees is representative of the average full-size SUV. Since the

design changes to increase TTA to that level would cause a compact

SUV to approach the size of full-size SUVs, establishing any higher

standard, whether for all vehicles or for compact SUVs alone, would

lead to the virtual elimination of compact SUVs as that class

currently exists.

14 The recomputation was performed using the same

procedures used for the 1994 estimates and explained in detail in

the document ``Potential Reductions in Fatalities and Injuries in

Single Vehicle Rollover Crashes as a Result of a Minimum Rollover

Stability Standard.'' That document is in Docket 91-68; Notice 3.

However, while the procedures were the same, an expanded set of data

(the number of rollover accidents and single vehicle accidents) were

used in the recomputation to increase its accuracy. The use of the

new data adequately addresses the petitioners' concerns about the

agency's use in the 1994 notice of weighted averages for models for

which there was insufficient data to determine the actual rollover

rate.

---------------------------------------------------------------------------

(1) The reduction of rollovers per single vehicle accident (RO/SVA)

predicted for increases in TTR;

(2) The number of single vehicle accidents experienced by vehicles

that would need to be altered in order to comply with the standard; and

(3) The degree of harm mitigation (in the number of fatalities and

serious injuries) as a result of rollover prevention given that a

single vehicle accident has occurred.

The following table, corresponding to Table 1 of the document

``Potential Reductions in Fatalities and Injuries in Single Vehicle

Rollover Crashes as a Result of a Minimum Rollover Stability

Standard,'' contains the results of this latest computation. For an

explanation of the headings and entries in the table, see that

document.

------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Est % Projected AIS3 + Fatality

MY 1991 1986-88 5 1986-90 of Est AIS3 Est RO/SVA @ reduction reduction

Compact SUV make model Drive production TTR state SVA/RV Michigan compact + fatalities min TTR @ min TTR @ min TTR

RO/SVA SUV ROs injuries 1.05 1.05 1.05

------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Vehicle A.................................... 2 WD*\15\ 65,515 0.88 **\16\0.0068 **0.359 6 39 28 0.270 9.6 6.9

4 WD 184,554 0.88 **0.0068 **0.359 16 110 79 0.270 27.2 19.5

Vehicle B.................................... 2 WD 29,480 0.95 0.0103 0.342 4 25 18 0.280 4.6 3.3

4 WD 93,866 0.99 **0.0102 0.27 9 63 45 0.244 6.1 4.4

Vehicle C.................................... 2 WD* 19,920 1.08 0.0091 0.317 2 14 10 ........... ......... .........

4 WD 101,541 1.08 0.0091 0.317 11 71 51 ........... ......... .........

Vehicle D.................................... 2 WD 0 ........... ............ ......... ....... ........ .......... ........... ......... .........

4 WD 46,478 1.03 0.0163 0.273 8 50 36 0.263 1.8 1.3

Vehicle E.................................... 2 WD* 4,892 1.01 0.0211 0.362 1 9 7 0.338 0.6 0.4

4 WD 39,989 1.01 0.0211 0.362 11 74 53 0.338 4.9 3.5

Vehicle F.................................... 2 WD* 3,555 0.93 **0.0215 **0.315 1 6 4 0.258 1.1 0.8

4 WD 35,945 0.93 **0.0215 **0.315 9 59 42 0.258 10.8 7.7

Vehicle G.................................... 2 WD 0 ........... ............ ......... ....... ........ .......... ........... ......... .........

4 WD 30,702 0.978 ............ 0.394 5 31 22 0.348 3.6 2.6

Vehicle H.................................... 2 WD 6,479 0.95 0.0114 0.259 1 5 3 0.219 0.7 0.5

4 WD 23,515 0.99 **0.0123 0.252 3 18 13 0.228 1.7 1.2

Vehicle I.................................... 2 WD 0 ........... ............ ......... ....... ........ .......... ........... ......... .........

4 WD 26,776 0.98 ............ **0.481 5 33 24 0.427 3.7 2.7

Vehicle J.................................... 2 WD* 740 0.947 ............ ......... 0 1 0 0.281 0.1 0.1

4 WD 23,870 0.947 ............ ......... 3 20 15 0.281 3.3 2.1

Vehicle K.................................... 2 WD* 1,257 0.978 ............ 0.407 0 2 1 0.360 0.2 0.1

4 WD 10,492 0.978 ............ 0.407 2 13 10 0.360 1.6 1.1

Vehicle L.................................... 2 WD 0 ........... ............ ......... ....... ........ .......... ........... ......... .........

4 WD 11,404 0.88 **0.0068 **0.359 1 7 5 0.270 1.7 1.2

Vehicle M.................................... 2 WD 0 ........... ............ ......... ....... ........ .......... ........... ......... .........

4 WD 10,616 0.93 ............ ......... 1 9 7 0.274 1.7 1.2

Vehicle N.................................... 2 WD* 5,011 1.016 ............ ......... 1 4 3 0.315 0.2 0.2

4 WD 2,818 1.016 ............ ......... 0 2 2 0.315 0.1 0.1

Vehicle O.................................... 2 WD* 832 1.04 ............ ......... 0 1 1 0.329 0.0 0.0

4 WD 3,546 1.04 ............ ......... 0 3 2 0.329 0.1 0.0

[[Page 28555]]

Weighted Averge.............................. ............................. .......... ........... 0.01049 0.335 ....... ........ .......... ........... ......... .........

------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Total \17\............................. ............................. 783,783 ........... ............ ......... 100 669 480 ........... 85.0 61.0

Total \18\............................. ............................. .......... ........... ............ ......... ....... ........ .......... ........... 22.0 32

------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

\15\ An ``*'' in this column indicates that the agency lacked sufficient data for the 2WD version of the model. For these models, the agency assumed that the 2WD version had the same TTR and

the same rollover rate as the 4WD version.

\16\ An ``**'' in this column and in the next one indicates that 1988-91 Michigan accident data were used instead of the data indicated by the column heading.

\17\ The serious injury and fatality reduction figures in this row are the benefits that might result if the standard prevented not only a rollover, but also an accident of any type.

\18\ The serious injury and fatality reduction figures in this row are the benefits that might result if the standard prevented a rollover, but still allowed an injury-causing accident of some

type to occur after the vehicle left the road. The injury and fatality figures in this row were derived by multiplying the figures in the row immediately above by a mitigation factor of 26

percent for injuries and 52 percent for fatalities. For further details on these factors, see section 3 of the Appendix to this notice.

There are two optimistic assumptions incorporated in the

computation process for both the original and new estimates: \19\

---------------------------------------------------------------------------

\19\ The agency made these assumptions because limitations in

available data made it impossible to use more precise values. When

making these assumptions, the agency took an optimistic approach so

as to present the prospects of a vehicle stability standard in the

best possible light.

---------------------------------------------------------------------------

The number of rollover injuries and fatalities prevented

will be proportional to the number of rollovers prevented; and

The fatality and injury rates of the late 1980's will be

representative of future rates.

The effect of these optimistic assumptions is that these new

estimates, like the 1994 estimates based on the same assumptions, may

in fact overstate the actual benefits, i.e., the number of fatalities

and injuries likely to be prevented by improving the TTR of compact

SUVs to 1.05.

The first assumption assumes that the rollover accidents that would

be prevented as a result of requiring an increase in TTR would have the

same fatality and injury rates as rollover accidents in general. There

is reason to believe that this would not be the case. The likelihood of

fatalities and serious injuries in rollover accidents is heavily skewed

toward crashes involving more than one quarter turn. Data show that

light truck rollover crashes involving only a single quarter turn have

about one-third the fatality rate of the average rollover. This

difference in likelihood of harm is significant if moderately improving

TTR would not be equally likely to prevent a multiple quarter-turn

rollover as a single quarter-turn rollover. NHTSA believes that it is

more likely that the prevented rollovers would tend to be the lowest

energy rollovers, i.e., the single quarter-turn rollovers. At best,

improving TTR would only slightly mitigate the more severe rollovers.

Thus, by assuming that rollovers prevented by an improvement in TTR

would be average rollovers instead of the least severe rollovers, the

agency is overstating the benefits obtainable from such an improvement.

Had the agency based its benefit estimates on the fatality rate of

rollovers involving a single quarter turn, the estimated number of

prevented fatalities would have been about 11 instead of 32.

The second assumption, that the fatality and injury rate in

rollovers will remain constant, is likely to overstate the benefits of

a vehicle stability standard since, if recent trends continue, future

increases in safety belt use, as a result of Federal, state, and local

efforts, can reasonably be expected to reduce the overall harm from

rollover accidents. As belt use increases, rollover casualties

decrease, even if the number of rollover crashes remains constant.

Consequently, even with liberal assumptions and using the most

current and complete database available, NHTSA estimates that a

rollover stability standard requiring compact SUVs to achieve the same

TTR (1.05) as the typical full-size SUV would prevent 22 serious

injuries and 32 fatalities annually. While precise quantification is

impossible, the agency believes, for the reasons stated above, that the

actual level of safety benefits would be significantly lower.

D. Costs of Improvements in Vehicle Stability Metrics

The substantial vehicle redesigns necessary to enable many existing

compact SUVs to achieve a TTR of 1.05 and produce the estimated

reductions in fatalities and injuries discussed above would have

substantial negative impacts, both in terms of reduced consumer choice

and unmet preferences and in terms of increases in manufacturer and

consumer costs.\20\ As noted above, the only way to achieve significant

increases in TTR is to increase the track width and/or lower the center

of gravity. Increasing track width or lowering the center of gravity,

using conventional, commonly used designs and production methods, would

necessarily, and significantly, increase vehicle size and weight. For

NHTSA, in effect, to require compact SUVs to approach the size and

weight of full-size SUVs would run counter to consumer preferences that

have led to the existing fleet of compact SUVs. The strength of those

preferences is demonstrated by the fact that compact SUVs outsold full-

size SUVs by a margin of six to one in 1994, the latest year for which

the agency has sales data.\21\ The Ford Explorer, the compact SUV model

with the lowest TTR and therefore the compact SUV which would be most

affected by any minimum standard, is the best-selling SUV and is the

ninth most popular make/model of all car and truck models combined.

---------------------------------------------------------------------------

\20\ As explained in the Appendix, NHTSA made two cost

estimates. The first was based on the assumption that compact SUVs

needing a TTR increase of more than 0.06 would require a full

vehicle redesign. The second was based on the assumption that only

compact SUVs needing a TTR increase of more than 0.04 would require

such a redesign.

\21\ Model year 1994 sales data from Automotive News 1995 Market

Data Book, Crain Communications, Detroit, Michigan, May 24, 1995.

All light trucks--6,097,787 vehicles.

Compact SUVs--21.9% of light trucks, or 1,335,415 vehicles.

Full-size SUVs--3.6% of light trucks, or 219,520 vehicles.

---------------------------------------------------------------------------

Upsizing compact SUVs so as to eliminate much of the size and

weight difference between those vehicles and full-size SUVs also might

have a significant adverse affect on the

[[Page 28556]]

production and sales of SUVs. The body of the average full-size SUV is

currently about 10 inches wider than that of the average compact SUV,

and the track width is about 9 inches greater. The 6-inch increase in

track width necessary to bring the TTR of compact SUVs up to that of

full-size SUVs (assuming no increase in c.g. height) would remove much

of those differences between compact SUVs and full-size SUVs. Given the

admonition in the legislative history of the National Traffic and Motor

Vehicle Safety Act against eliminating vehicle types (see the

discussion in section D of the Appendix to this notice), such a

dramatic potential impact on the design of compact SUVs and on the

market for those vehicles must be carefully weighed.

In addition to impacts on consumer choice and sales, there are

substantial monetary costs associated with redesigning those compact

SUVs that would need significant increases in TTR to meet a standard of

1.05.\22\ The agency estimated those costs using confidential cost data

submitted by domestic automobile manufacturers during the course of

several agency rulemaking proceedings to establish light truck

Corporate Average Fuel Economy (CAFE) standards. The estimated consumer

cost of bringing all such new compact SUVs into compliance with such a

standard is between $310 million and $335 million, depending on which

of two assumptions is made about the vehicles that would require a full

vehicle redesign. A detailed discussion of the method used to estimate

these costs is included in the Appendix to this notice.

---------------------------------------------------------------------------

\22\ As demonstrated by Table 1, the vast majority of the

measurable benefits from such a standard would come from

improvements to these fully redesigned vehicles, instead of those

vehicles that would need only lesser changes to comply with the

standard.

---------------------------------------------------------------------------

The agency believes that the foregoing estimate of the costs of a

rollover standard requiring compact SUVs to achieve a minimum TTR of

1.05 is understated. Those estimates do not include the incremental

costs of material and labor involved in the manufacture of a larger

vehicle. In addition, the estimates do not include any costs for

vehicles that would only need minor changes, instead of a full vehicle

redesign, to comply with the standard. NHTSA has not attempted to

calculate those costs because the benefits of the standard are already

outweighed by the initial cost estimate.

The agency recognizes that providing a lengthy leadtime period

would reduce the costs of compliance to the extent that manufacturers

were able to make their compliance efforts coincide with their normal

model changeover timetable. However, providing additional leadtime

would do nothing to reduce the adverse impacts on consumer preferences.

Further, an extended lead time would not affect the costs of additional

labor or materials.

VI. Conclusion

The discussion above and in the Appendix demonstrates that even a

standard applicable only to compact SUVs, the vehicle type that the

petitioners characterize as one of the two ``most rollover-prone

vehicle types,'' 23 would generate substantial adverse impacts on

manufacturers and consumers, both in terms of monetary costs and in

loss of consumer choice, that would outweigh the benefits of such a

standard. There is no reason to believe that a standard that would

mandate significant increases in TTR/TTA or CSV for any other vehicle

type or group of vehicle types would be any more cost beneficial.

---------------------------------------------------------------------------

23 Advocates/IIHS petition, page 12.

---------------------------------------------------------------------------

Accordingly, NHTSA reaffirms its decision to terminate this

rulemaking without proposing a rollover stability performance standard.

Issued on May 31, 1996.

Barry Felrice,

Associate Administrator for Safety Performance Standards.

Appendix

The Advocates/IIHS petition contained many detailed technical

arguments. Responses to the more significant ones are provided in

this appendix.

A. The Benefits Estimate

1. Replacing Weighted Averages With Actual Rollover Data Now Available

Makes No Appreciable Change in the Estimate

The petitioners criticized the benefit estimates made by the

agency in connection with the 1994 notice because, for those vehicle

models for which the agency lacked sufficient rollover accident

data, it used the average of the rollover per single vehicle

accident rate (RO/SVA) of the class of vehicles to which that make

and model belonged, weighted by the 1991 production of each make and

model for which the agency had RO/SVA. The benefits were calculated

using the TTRs of 1991 makes and models and the accident records of

1991 makes and models (and identical vehicles from prior model

years) to represent a hypothetical future fleet.1

---------------------------------------------------------------------------

1 A detailed discussion of the method can be found in

``Potential Reductions in Fatalities and Injuries in Single Vehicle

Rollover Crashes as a Result of a Minimum Rollover Stability

Standard'' in Docket 91-68, Notice 3.

---------------------------------------------------------------------------

The petitioners pointed out that the average TTR for vehicles

for which the agency did not have adequate RO/SVA data was lower

than the average TTR of vehicles for which it had RO/SVA data, and

therefore claimed that use of weighted averages was inappropriate.

The petitioners' criticism concerning the use of weighted averages

as substitutes for missing data was focused particularly on the use

of those averages for the large number of vehicles in the

hypothetical future fleet that were represented by the Ford

Explorer. The agency had no RO/SVA or single vehicle accident

involvement rate (SVA/RV) data for the Ford Explorer and certain

other vehicles at the time of the notice because they were either

recently introduced or comparatively low production volume models.

The petitioners argued that a higher rollover rate should have been

used for vehicles like the Explorer which have a lower TTA than the

vehicles from which the weighted average was derived.

It is not appropriate to assume that a higher than average

rollover rate is appropriate for the Explorer or the other vehicles

simply based on their having a lower than average TTA. The data

demonstrate that the order of vehicle models ranked according to TTA

is not the same as the order of models ranked according to rollover

rate. See Table 1 in the accompanying notice of denial of petitions

for reconsideration. Thus, although two different vehicle models may

have the same TTA, they may not necessarily have the same rollover

rate. Likewise, a vehicle model with a TTA lower than that of

another model may nevertheless have a lower rollover rate, and vice

versa.

Accordingly, the agency has not assumed a higher rollover rate

for those models for which sufficient rollover data are lacking.

However, the agency has responded to the petitioners' concern about

the use of weighted averages in connection with the 1994 notice by

replacing those averages, where possible, with rates based on actual

rollover accident data that became available after that notice was

prepared.

Where sufficient, the 1988--1991 Michigan accident data were

used to calculate the rollover rate figures for models for which

data were previously missing. Following the practice of previous

analyses, the agency used the accident data to calculate rollover

rates only for makes and models which had at least 25 single vehicle

accidents. Actual rollover rates (RO/SVA) from Michigan were added

for the 4WD Ford Explorer, Nissan Pathfinder, and Isuzu Trooper, and

actual single vehicle accident rates (SVA/RV) were added for the 4WD

Ford Explorer, the 4WD S10 Blazer, the Nissan Pathfinder, and the

4WD GMC S15 Jimmy. The 4WD Explorer data were used for the nearly

identical, but low production volume, Mazda Navajo.

There were still some models for which the agency lacked

sufficient actual make and model accident data. For most of these

models, while the agency lacked sufficient data for the 2WD versions

of those models, it had sufficient data for the more numerous 4WD

versions. In these instances, the agency assumed that the rates for

the 2WD versions were identical to the rates for the 4WD versions of

the same make and model, instead of calculating rates based on

weighted averages. New weighted averages

[[Page 28557]]

were computed on the basis of the expanded data and were used only

where sufficient specific data remained unavailable for a particular

model. The instances in which the agency computed new weighted

averages were limited. Weighted averages of RO/SVA and SVA/RV were

used for less than 10 percent and 19 percent, respectively, of the

example population of compact SUVs. See Table 1.

Using actual rollover data wherever available, the agency

recomputed the benefit estimates for compact SUVs. Substitution of

the new rollover rates produced very little change in the estimate

of the numbers of fatalities and serious injuries that might be

prevented if a rollover stability standard were adopted for compact

SUVs. Replacing the weighted averages used in the 1994 notice with

rates based on accident data for particular makes and models changed

the result of the analysis very little, i.e., by less than four

percent. This may be seen by comparing the estimates of the benefits

that would be obtained if preventing a rollover meant preventing an

accident altogether. Those benefits were estimated to be 83 serious

injuries and 59 fatalities in the 1994 notice. They have been

recomputed to be 85 serious injuries and 61 fatalities, based on the

new accident data and less reliance on weighted averages. See Table

1.

2. Accident Mitigation, Not Accident Prevention, Is the Proper Measure

of Benefits

Since an accident would still occur in the vast majority of

instances in which a rollover is prevented, the agency reduced those

figures accordingly using an accident mitigation factor. The

resulting new benefit estimate is 22 serious injuries and 32

fatalities.

The petitioners criticized the agency for making the same

adjustment to the benefits in the 1994 notice. Then, as now, NHTSA

assumed that the benefits would come from accident mitigation

instead of accident prevention. It was appropriate for the agency to

assume that the benefits would be in terms of accident mitigation

since over 90 percent of all single vehicle rollovers are off-road,

tripped rollovers, i.e., rollovers that occur when a vehicle leaves

the roadway sideways, encounters a tripping mechanism, and rolls.

Since a vehicle is running off the road in a tripped rollover

situation, such a vehicle will still likely crash into some off-road

object even if the vehicle is prevented from rolling over after it

leaves the road. If a rollover can be prevented in that situation,

then the resulting accident will most likely be one of lower

severity than if a rollover had occurred because rollovers tend to

be more severe than non-rollover accidents. The primary benefits

from a rollover stability standard would result from preventing the

more severe form of off-road accident.

3. A Single Accident Mitigation Factor, Not Separate Factors for

Individual Vehicle Types, Is the Proper Basis for Measuring Benefits

The petitioners also criticized the agency for using a single

accident mitigation factor (52 percent) for fatalities across the

board instead of computing separate factors for different types of

vehicles. In support of their argument for the use of different

factors, they noted that rollover accidents account for 80 percent

of the fatalities of the occupants of compact SUVs in single vehicle

accidents. Based on this, the petitioners concluded that rollovers

in compact SUVs are four times as deadly as non-rollover accidents,

and therefore the agency should have used a mitigation factor of 75

percent for compact SUVs.

The agency rejects the petitioners' argument. A mitigation

factor based on ratios of absolute numbers of fatalities, instead of

on fatality rates, is incorrect unless the same number of occupants

were exposed to rollover accidents and non-rollover accidents. If

the exposure is not the same, then it is impossible to determine the

extent to which the ratio reflects the difference in accident

exposure versus a difference in accident severity. Further, the

issue of a difference in accident severity is not just a matter of

the difference in severity of a rollover accident and a non-rollover

accident at the same speed. It is also a matter of possible

differences in speed. For example, it is necessary to determine

whether the consequences of 60 mph rollovers are being compared to

those of 30 mph non-rollover accidents. Finally, it is also

necessary to examine whether apparent differences between vehicle

groups are a result of differences in crashworthiness, or just a

consequence of smaller sample sizes.

The agency's use of a single mitigation factor for fatalities

takes these considerations into account. NHTSA considered the number

of occupants exposed to rollover and non-rollover single vehicle

accidents as well as the number of fatalities for each accident

type. It also considered the speed limit of the road as a rough

indication of the severity of the accident.

As a first step in determining the mitigation factor, NHTSA

compared the overall fatality rate of rollover accidents to the

overall fatality rate of non-rollover accidents, based on single

vehicle accidents of all cars and light trucks without consideration

of accident severity. The fatality rate of rollover accidents was

slightly more than twice that of non-rollovers, suggesting a 52

percent mitigation factor.

Next, the agency computed a series of relative fatality rates

(with and without rollover), comparing only accidents occurring on

roads with the same range of posted speed limits (25 mph or less;

30-35 mph; 40-50 mph; 55-65 mph). While the accident data do not

indicate the actual accident speed, grouping by speed limit acts as

a rough control on accident severity, because it restricts accident

groups to the same kinds of roads, even though the actual range of

crash speeds may significantly exceed the range of posted speed

limits for a particular group of accidents. The relative fatality

rate for each road speed limit group were added and then averaged.

The result was the same 52 percent mitigation factor for fatalities.

Using the same process led to a mitigation factor of 26 percent for

serious injuries.

In addition, even if the agency were to use different mitigation

factors for different vehicle types, their use would not result in

dramatic changes in benefit estimates. For compact SUVs, the 75

percent mitigation factor suggested by the petitioners would result

in a fatality reduction of 46 rather than the 32 calculated by the

agency. This difference is 0.15 percent of the 9,000 annual rollover

fatalities. Using the estimates prepared for the 1994 notice, for

the entire light duty vehicle fleet, the use of different mitigation

factors resulted in predicting 71, instead of the agency's 63, lives

saved from requiring a TTR of 1.05. This is a difference of 0.089

percent.

B. The Cost Estimate

The petitioners criticized the agency for failing to provide any

costs for the vehicle changes that would be necessary to meet a

minimum rollover stability standard. The agency concluded in the

1994 notice that a large number of vehicles would require

fundamental full redesigns to meet a minimum stability standard.

Because the agency was aware of the magnitude of costs involved in

vehicle redesigns, it was apparent that the costs and other impacts

would substantially exceed the benefits. NHTSA did not, however,

provide a quantification of those costs and other impacts.

To demonstrate the validity of its conclusion about the costs

and other impacts, the agency has conducted a rough cost analysis

for this notice as set forth below.2 To estimate the compliance

costs for those vehicles which would have to be fully redesigned to

make the substantial changes necessary to comply with a minimum

stability standard, the agency used confidential cost data submitted

by domestic automobile manufacturers during the course of several

agency rulemaking proceedings for light truck Corporate Average Fuel

Economy (CAFE) standards. These data are manufacturer estimates of

the costs of full redesigns of compact SUVs that would have been

necessary if the CAFE standards had been set at certain levels.

These submissions include estimates of investment costs for a

redesigned vehicle model, but do not include material and labor

costs for the manufacture of the vehicle. NHTSA believes a full

vehicle redesign for rollover stability purposes would necessitate

similar investment costs. Accordingly, it is appropriate to use the

investment cost figures from the CAFE program to estimate the

investment costs for vehicles which would require a full redesign to

comply with a rollover stability standard.

---------------------------------------------------------------------------

\2\ As explained below, the agency did not calculate all costs

of a standard because it determined that one category of those

costs, the investment costs for vehicles requiring major changes,

would by itself exceed the benefits of a standard.

---------------------------------------------------------------------------

The CAFE submissions include investment cost data for five

models of compact SUVs. Since the specific raw data are

confidential, they cannot be set forth here or otherwise publicly

released. To convert those data into a form in which the original

data can not be determined, the agency divided the per model data by

the applicable manufacturer's estimated average annual production

capacity and then divided by the number of

[[Page 28558]]

years of the vehicle's design cycle life.3 The per-vehicle cost

estimates for these five vehicle design cycle lives were then

averaged to arrive at the estimate used in this analysis. The

individual per-vehicle cost estimates range from $317.37 to $532.37

and the average is $416.77. Since these costs are costs to the

manufacturer, they were adjusted to represent costs to the consumer

by dividing them by 0.75, a standard factor used by the agency in

its vehicle rulemaking in estimating consumer costs from

manufacturer's wholesale costs. The resulting estimated average

consumer cost per vehicle resulting from the redesign of a compact

SUV is $555.69.

---------------------------------------------------------------------------

3 Since the submissions were made in 1986, 1989, 1993,

and 1994, submissions for years prior to 1994 were adjusted to 1994

dollars using the implicit gross domestic product deflator as

calculated by the Bureau of Economic Analysis.

---------------------------------------------------------------------------

The agency then determined the number of vehicles that would

need vehicle redesigns to comply with a vehicle rollover stability

standard requiring a minimum TTR of 1.05. Based on available data,

the agency believes at least some models would have to be fully

redesigned to achieve TTR increases of more than 0.04, and that

almost all models would have to be fully redesigned to achieve at

TTR increase of 0.06. The agency determined next that 558,756

vehicles would need to be fully redesigned if the threshold for

having to make a full redesign were 0.06 and 603,637 vehicles would

need to be fully redesigned if the threshold were 0.04. Multiplying

these numbers of vehicles by the $555.69 per vehicle investment cost

estimate, the agency estimated that the total investment costs of a

standard requiring a TTR of 1.05 would be $310,495,121 to

$335,435,044.

The agency believes that this range of estimated costs of a

rollover standard requiring compact SUVs to achieve a minimum TTR of

1.05 is understated. As noted earlier, these cost estimates do not

include estimates of the incremental costs of material and labor

involved in the manufacture of the vehicle. Since vehicles would

need significant increases in track width, and attendant increases

in wheelbase, they would be generally larger and heavier. As a

result, the agency concludes that there would be significant

increases in the costs of material and labor involved in the

manufacture of such vehicles. In addition, these cost estimates do

not include any costs for vehicles which could comply with the

standard by changes that are less than a full vehicle redesign.

C. Objections to the Statistical Tools Used by the Agency in

Reaching Its Decision

The petitioners asserted that the agency did not use the

``typical'' statistical measure, the deviance statistic, to judge

the adequacy of the logistic regression models used by the agency in

its analyses of the relationship of TTA to RO/SVA, and the

importance of the vehicle stability metrics. The petitioners also

objected to the agency's use of two statistical measures, R2

and the C-statistic. Finally, the petitioners questioned the

agency's reliance on data from the State of Michigan.

Although the agency did not use the deviance statistic to judge

the adequacy of the logistic regression models, the agency did use a

mathematically equivalent measure, the likelihood statistic (-

2*ln(likelihood)). Using that measure permitted the agency to

compare the effect of adding variables (specifically the vehicle

stability metrics) to the hypothesized models. Detailed discussions

of the agency's analyses are found in the Technical Assessment Paper

(TAP) (Docket 91-68-N01-03) and the Addendum to the Technical

Assessment Paper (Docket 91-68-N03-02) which were placed in NHTSA's

docket. The TAP and the Addendum present analyses using five

measures: the C-statistic, R2, the percentage change in

R2, the likelihood statistic, and the variables' chi-square. It

is true that the deviance statistic was not reported because the

computer software the agency used to conduct this analysis, SAS

Institute's PROC LOGIST, does not include the deviance statistic as

one of the model diagnostics. However, the agency does not believe

that this affects the general conclusions regarding the importance

of the vehicle stability metrics.

NHTSA believes that it may help to explain this issue in non-

statistical terms. The petitioners' argument amounts to a complaint

that the agency described various glasses of water in terms of how

much water is in the glass, instead of in terms of how much water

could be added to the glass. In either case, the capacity of the

glass is the same. If the capacity is known, and if either the

amount of water or the amount of unused capacity is known, the other

amount can be derived.

Similarly, the deviance statistic preferred by the petitioners

describes how much of the variability 4 in the regression model

is left to be explained. The likelihood statistic, which the agency

used, describes how much of the variability in the model is

explained. In either case, the total variability to be explained is

the same. If, as the agency's analysis showed, the addition of TTR

to the model decreased the value of the likelihood statistic, the

deviance statistic would have increased by the same amount. Using

either measure would lead to the same conclusion about the value of

TTR.

---------------------------------------------------------------------------

4 Variability is the difference between what the

statistical model predicts and actual accident records.

---------------------------------------------------------------------------

The petitioners also assert that the use of R2 was

inappropriate because it is not weighted, i.e., it does not reflect

the number of single vehicle accidents for each vehicle make and

model. The petitioners also state that R2 is sensitive to

extreme values. The agency's use of R2 was described fully on

page 5-66 of the TAP. The agency agrees, as explained in the TAP,

that there are limitations to the use of R2. As also explained

in the TAP, R2 was used as an approach to providing the types

of descriptive statistics of model fit with which more people are

familiar, and not to provide a mathematically rigorous assessment of

model fit. The agency's use of R2 was an attempt to make the

explanation of the analysis understandable to a wider audience, and

was not the sole basis of the agency's decision.

The petitioners' assertions of problems with the use of the C-

statistic are not applicable to the C-statistic as used by the

agency. In an attempt to support their assertions, the petitioners

pointed to an example of how the C-statistic can ``misbehave''

presented on page 146 of Hosmer and Lemeshow.5 The agency's use

of the C-statistic is not the same as that in Hosmer and Lemeshow's

example. That example simply uses a classification table with an

arbitrary cut point to determine, e.g., whether an actual rollover

was predicted to be a rollover. The C-statistic employed by NHTSA

measured the concordance between all possible pairs of observations,

taking one from the actual rollover population and one from the

actual non-rollover population. The C-statistic represents the

percentage of those pairs (which number literally in the millions)

for which the actual rollover had a higher predicted probability

than the actual non-rollover's predicted probability (of rolling

over), minus one-half the number of ties. There is no arbitrary

cutoff point. In addition, the agency's decision was not based on a

single statistical measure. The agency analyzed the data with a

number of statistical measures, all of which pointed to the same

conclusions. Accordingly, the agency remains confident in its

results.

---------------------------------------------------------------------------

5 D.W. Hosmer and S. Lemeshow, Applied Logistic

Regression, Wiley Interscience, New York, 1989.

---------------------------------------------------------------------------

Finally, the petitioners' objection to the agency's reliance on

Michigan data for performing the statistical regressions instead of

using the data from the other four states was based on their concern

that the agency did not examine the extent to which the state is

anomalous because of its generally flat topography. The petitioners

stated that this could lead to a lower proportion of rollovers per

single vehicle accident than the other states in the data base. The

agency relied on Michigan data because they included a large number

of available observations, and were based on a low reporting

threshold and more refined accident reporting variables. The agency

did examine whether the rollover rate in this state was anomalous,

and as stated on page 13 of the Addendum, discovered that ``(t)he

rollover rate in Michigan is near the midpoint of the range for all

five states studied.'' The examination of the relative rollover

rates of the five different states was fully explained in the TAP on

pages 59-65.

D. Legal Arguments

The petitioners also addressed the implications of the

Intermodal Surface Transportation Efficiency Act of 1991 (ISTEA)

(P.L. 102-240), the National Traffic and Motor Vehicle Safety Act of

1966 (the Safety Act) (P.L. 89-563),6 and the Regulatory

Flexibility Act (P.L. 96-354) for rulemaking concerning a vehicle

stability standard. The petitioners also argue that the decision not

to issue a rollover standard is judicially reviewable.7

---------------------------------------------------------------------------

6 After the publication of the termination notice, the Safety

Act was codified in volume 49 of the United States Code. Any cites

to provisions of the Safety Act have been updated to reflect the

codification.

7 The agency agrees with the petitioners that this

termination is ``final agency action'' for the purposes of judicial

review.

---------------------------------------------------------------------------

[[Page 28559]]

The petitioners begin by citing the provision in ISTEA that

required NHTSA to initiate rulemaking concerning a rollover

standard. The petitioners acknowledge that Congress did not mandate

the issuance of a final rule in this area.

Although the petitioners make this concession, it bears

emphasizing how clearly ISTEA and its legislative history

demonstrate that in each instance in which Congress mandated that

the agency initiate vehicle safety rulemaking, it clearly specified

whether the agency had the discretion to decide not to issue a final

rule. In sections 2502-3 of ISTEA, Congress specified that the

agency was to initiate rulemaking regarding five different areas of

vehicle safety performance. With respect to one area, upper interior

head impact protection, Congress specified that rulemaking would be

considered completed only when a final rule was issued. However,

with respect to the other four areas, including rollover, Congress

did not mandate the issuance of a final rule. It expressly provided

that rulemaking on rollover and the other three areas would be

considered completed either when the agency issued a final rule or

when the agency decided, after considering public comments, not to

issue a final rule. The Conference Report on ISTEA emphasized the

discretion which it had reserved to the agency. The conferees said,

with reference to the mandated rulemaking on rollover, ``the

conferees do not predetermine the outcome of [this rulemaking]. The

[NHTSA] is free to conclude the rulemaking in any manner consistent

with the APA and the 1966 Act'' (H. Conf. Rep. 404, 102d Cong., 1st

Sess., at 397 (1991)). Thus, Congress made no judgment in ISTEA

about the ultimate merits of issuing a final rule on rollover.

Instead, Congress provided NHTSA with the latitude to decide that a

rollover standard should not be issued if, in the agency's judgment,

the facts did not warrant such issuance. The agency's conclusion

that such a regulation would not have sufficient benefits to justify

its cost is an ample and proper basis for a decision not to issue a

final rule.

Although the petitioners concede that Congress did not require

the agency to issue a final rule on rollover, they assert that

Congress ``expected the agency to set some form of stability-

enhancing regulation.'' 8 As authority for that assertion, they

cite the legislative history of the Department of Transportation and

Related Agencies Appropriations Act of 1995. (P.L. 104-59) The

Senate committee report on that Act contended that NHTSA had

``effectively abandoned efforts at developing a performance standard

for improved rollover protection.''

---------------------------------------------------------------------------

\8\ Advocates/IIHS petition, page 41.

---------------------------------------------------------------------------

The 1995 Appropriation Act legislative history is inapposite

here and lacks any possible binding effect. Since that history

pertains to a different statute, it carries no weight in the

interpretation of NHTSA's duties under ISTEA. NHTSA notes further

that the language cited by the petitioners is part of a discussion

expressing concern about the agency's delay in publishing some of

the ISTEA rulemakings. The discussion does not express any

expectation about the substantive outcome of agency rulemaking on

rollover, but does express an expectation that NHTSA will complete

the remaining ISTEA rulemakings expeditiously. Finally, even if the

Senate committee had specifically expressed an expectation

concerning the outcome of the rollover rulemaking, that expectation

would not impose a binding obligation on NHTSA unless Congress

coupled that expectation with a mandate to issue a final rule on

rollover and enacted that mandate into law. See Center for Auto

Safety v. Peck, 751 F.2d 1336, at 1351 (D.C. Cir. 1985). Congress

did not do so. Instead, it expressly decided not to mandate the

issuance of a final rule on that subject.

The petitioners argued that neither the Safety Act nor the

Regulatory Flexibility Act provide any legal grounds for terminating

rulemaking on a vehicle stability standard. The petitioners quoted

statements in the 1994 notice that 49 U.S.C. Sec. 30111(b)(3) would

preclude NHTSA from mandating any stability requirement that is

``incompatible with certain types of vehicles,'' and that a

stability requirement ``could raise concerns'' under the Regulatory

Flexibility Act. (59 FR 33254, 33263) They interpreted these

statements as implying that the agency believed it was prohibited

from issuing any standard that might require ``the radical redesign

of the characteristics [of] many, and in some cases all, vehicles of

certain classes * * * and possibly even the elimination of certain

classes of vehicles as they are known today.'' The petitioners

countered with alternative propositions, arguing that NHTSA has

authority to eliminate whole classes of vehicles, and that, even if

NHTSA does not have such authority, it failed to consider a less

demanding regulatory approach such as setting different standards

for separate vehicle types which would not require all vehicles in a

class to be altered. The petitioners argued also that NHTSA cannot

rely on the Regulatory Flexibility Act when the agency did not

prepare any analysis of the impacts of a standard on small entities.

The primary bases for the agency's decision to terminate

rulemaking on a vehicle stability standard are the limited safety

benefits, and the excessive costs and market disruption of such a

standard, regardless of whether that standard applies to all light

duty vehicles or to particular class such as compact SUVs. The 1994

notice discussed the high costs of a standard that specifies a

single performance level which was applicable to all light duty

vehicles and was high enough to require the full redesign of at

least some passenger cars. As explained previously, the agency

concluded that such a standard would have costs and other impacts

which outweighed its benefits. NHTSA similarly concludes that the

costs and other impacts of a standard applicable to compact SUVs

would far outweigh its benefits. Logically, if a standard for the

most rollover-prone light duty vehicles would fail this basis test,

it follows that a standard for other groups of light duty vehicles

would not be justified.

It should be noted that neither 49 U.S.C. 30111(b)(3) nor the

Regulatory Flexibility Act impose an absolute legal bar to a minimum

stability standard. The agency is not foreclosing any possibility of

further rulemaking. As stated above, NHTSA might reinitiate

rulemaking in this area if information becomes available

demonstrating the cost effectiveness of a minimum stability

standard.

However, the Safety Act does place limits on the agency's

rulemaking authority. The agency lacks authority to eliminate entire

classes of vehicles. This interpretation reflects the language of 49

U.S.C. 30111(b)(3) and its legislative history. 49 U.S.C.

30111(b)(3) states:

When prescribing a motor vehicle safety standard under this

chapter, the Secretary shall * * * consider whether a proposed

standard is reasonable, practicable, and appropriate for the

particular type of motor vehicle or item of motor vehicle equipment

for which it is prescribed.

The Senate Report accompanying the 1966 Safety Act explained

this provision as follows:

In determining whether any proposed standard is ``appropriate''

for the particular type of motor vehicle or item of motor vehicle

equipment for which it is prescribed, the committee intends that the

[NHTSA] will consider the desirability of affording consumers

continued wide range of choices in the selection of motor vehicles.

Thus it is not intended that standards will be set which will

eliminate or necessarily be the same for small cars or such widely

accepted models as convertibles and sports cars, so long as all

motor vehicles meet basic minimum standards.

(S. Rep. 1301, 89th Cong., 2d Sess., at 6 (1966))

Given this legislative history, NHTSA cannot mandate a stability

requirement so incompatible with the most fundamental

characteristics which define a class of vehicles that implementing

the requirement would cause the elimination of that class. As an

example, the agency noted in the 1994 notice that sport utility

vehicles have features (high ground clearance and narrow track

width) to facilitate off-road use and use on snowy roads. The agency

would not have the authority to set a performance level so stringent

that no vehicles could have these features. This is neither a

radical, nor a new interpretation of the agency's authority. NHTSA

is not suggesting, as the petitioners suggest, that the agency lacks

any authority to issue a standard that requires significant change

to all vehicles in a class. In fact, there are many examples of the

agency using its authority to require changes to all vehicles in a

particular class. Those changes did not, however, eliminate as a

practical matter any recognized classes of vehicles.

Petitioners incorrectly suggested that the agency had a duty

under the Regulatory Flexibility Act to prepare a regulatory

flexibility analysis in connection with either the 1994 notice or

the ANPRM which preceded it. NHTSA did not ``fail'' to prepare any

required report. That Act mandates the preparation of analyses in

connection with notices of proposed rulemaking and final rules only.

NHTSA believes that the Regulatory Flexibility Act was a

relevant concern in

[[Page 28560]]

considering the possibility of proposing a stability standard

applicable to all light duty vehicles because multistage

manufacturers, especially van converters, which are often small

business entities, could be affected by such a standard. NHTSA is

not suggesting that that Act would prevent the issuance of such a

standard or that the concerns about impacts on small manufacturers

were insurmountable regardless of what approach is taken by the

agency in setting the standard. In fact, a standard limited to

compact SUVs would essentially eliminate those impacts because few,

if any, of those vehicles are produced by multistage or other small

manufacturers.

F. NHTSA's Alleged Lack of a Comprehensive Rollover Program

The petitioners characterized NHTSA's identification of seven

separate measures as part of a comprehensive agency plan to address

rollovers as simply ``a chronicle of ongoing or prospective crash

reduction programs that are not aimed uniquely at mitigating

rollover losses.'' The petition went on to complain that some of the

measures ``may never come to fruition,'' and that others have not

been specifically tailored by the agency to address the rollover

problem. The petitioners concluded by stating their belief that

NHTSA's comprehensive program for rollover is really an attempt to

try to persuade the public that the agency is taking action on

rollover safety, notwithstanding the termination of the vehicle

stability rulemaking.

The agency believes that the question of whether the activities

comprising its comprehensive rollover program uniquely address

rollover safety is irrelevant if those activities effectively

address that issue. If NHTSA can take actions, such as issuing a

standard, that significantly reduces the deaths and injuries that

occur in rollover crashes, it should make no difference whether that

reduction is achieved by means that also reduce deaths and injuries

in other types of crashes. The agency agrees that there is a

possibility that some of the regulatory initiatives announced by the

agency as part of its rollover program involve proposals that may

never become final rules. However, this possibility exists with any

regulatory initiative. The agency cannot foretell the nature of the

public comments that it will receive or prejudge the outcome of its

analyses of comments and other information obtained during the

rulemaking process. NHTSA included those initiatives in its rollover

program because preliminary evaluations of those initiatives

indicate that they are promising avenues for addressing rollovers.

The agency will pursue these initiatives expeditiously and

conscientiously. For example, since the 1994 notice was published,

NHTSA has published a final rule to extend the current requirements

for side door latches to rear door latches. (60 FR 50124) This rule

is an attempt to reduce the number of ejections from the rear door

of vehicles, thus reducing injuries and fatalities. Based on data

for years 1988-1992, NHTSA estimates that 147 occupants were fatally

ejected from the rear door of vehicles. Forty two percent of those

fatalities occurred in rollover accidents.

One of the specific initiatives singled out for criticism by

petitioners was the upgrade of Standard 201 to reduce head impact

injuries. The petitioners objected to its inclusion in NHTSA's

comprehensive rollover plan because the proposed compliance impact

speeds ``are often less than those [speeds] responsible for the very

high rate of severe head trauma that is suffered by occupants in

rollover crashes.'' The final rule upgrading Standard No. 201 was

published on August 16, 1995. (60 FR 43031) Even if the petitioners

were correct, the essential fact remains that the final rule will

make substantial reductions in rollover fatalities and injuries. The

agency estimated that 244-334 fatalities and 189-273 serious

injuries would be averted in rollovers as a result of that rule.

Authority: 49 U.S.C. 322, 30111, 30115, 30117 and 30166;

delegation of authority at 49 CFR 1.50 and 501.8.

[FR Doc. 96-14145 Filed 5-31-96; 4:38 pm]

BILLING CODE 4910-59-P