U.S. Department of Transportation	http://www.nhtsa.dot.gov

EXECUTIVE SUMMARY

The accompanying NPRM proposes to amend the rear and side impact tests of FMVSS 301, Fuel System Integrity. For the rear impact test, the agency is proposing a more stringent offset test using a lighter deformable barrier, but at a higher test speed of 80 km/h (50 mph). The agency is also proposing to replace the side test for FMVSS 301, with the FMVSS 214 (side impact protection) test. No changes are proposed for the frontal barrier crash test for FMVSS 301.

Costs: The average costs for vehicles which would need corrective action is $5.00 per vehicle. Based on an estimate of 46 percent of the fleet not meeting the standard currently, when spread out over the entire fleet, costs are estimated to be $2.30 per vehicle. Total costs for the fleet are estimated to be $35 million annually.

Target Population: There are 1,200 to 1,300 passenger vehicles with fire annually in FARS (Fatality Analysis Reporting System) in all types of crashes. In addition, there are an estimated 4,000 passenger vehicles involved in injury crashes with fire and 5,000 property damage only crashes with fire (based on NASS-GES). The requirements of this proposal are designed to provide protection to occupants who may be injured by fires that originate in passenger vehicles that are struck in the rear by another passenger vehicle. Based on FARS and NASS-CDS data, there are 57 fatalities and 119 non-fatal injuries annually in the target population. The non-fatal burn injuries were mostly minor and were typically not the maximum injury to the occupant.

Benefits: Vehicles that pass the proposed rear impact requirements will provide protection against crashes in which the impact produces a 33 to 50 percent higher delta V (which corresponds to 110 percent more energy being dissipated in the crash) compared to the current requirements. Benefits are estimated to range from 8 to 21 lives saved annually, once all vehicles on-the-road meet the proposal. While we believe the FMVSS 214 side impact test is somewhat stricter than the current side impact test in FMVSS 301, we could not quantify any benefit since only one out of more than 100 vehicles failed the proposed fuel leakage requirements using the FMVSS 214 proposal.

While it is possible that countermeasures to strengthen the fuel system integrity against rear impact crashes could provide some benefits in other crash modes or in rear impacts involving heavy trucks as striking vehicles, this analysis assumes no benefits in these instances.

Cost effectiveness: $1.67 to $4.38 million per life saved.

TABLE OF CONTENTS

EXECUTIVE SUMMARY

I.   Introduction

This Preliminary Regulatory Evaluation accompanies a Notice of Proposed Rulemaking (NPRM) to upgrade the rear and side impact performance requirements of Federal Motor Vehicle Safety Standard (FMVSS) No. 301, Fuel System Integrity. The purpose of this rulemaking is to reduce fatalities and injuries caused by fires that are the result of rear and side impacts in motor vehicle crashes. Specifically, the agency is proposing to make the current rear impact crash performance requirements more stringent for vehicles with gross vehicle weight rating (GVWR) of 10,000 pounds (4,536 kg) or less, and to replace the present FMVSS No. 301 side impact test requirement with the FMVSS No. 214 side impact test requirements.

This evaluation provides background information on the events leading up to this notice, discusses the agency's proposed rear and side impact fuel system upgrade, and analyzes data on crashes involving fire and the likelihood of fire occurring in rear and side impact crashes. This evaluation also discusses the costs, benefits and other impacts that could result from the proposed upgrade.

II.   Background

Preserving fuel system integrity in a crash to prevent occupant exposure to fire is extremely important. Although vehicle fires are relatively rare events (occurring in only one percent of towed vehicles in crashes), they tend to be severe in terms of casualties. According to an analysis of the agency's Fatality Analysis Reporting System (FARS), from 1991 through 1998, fires occurred in 2.5 to 2.8 percent of light vehicles involved in fatal crashes. The largest changes required by this rulemaking would involve crashes in which passenger vehicles are struck in the rear by other passenger vehicles. There could also be some benefit in vehicle to vehicle side impact crashes, as well as some low speed crashes involving heavier vehicles or stationary objects. Most other crashes would probably not be affected, including fires in other parts of the vehicle aside from the rear fuel tank, fires which started after being struck by a large truck or bus (except at very low speeds), fires which started after striking the front of the vehicle or in non-contact rollovers, and most cases where the rear of the fire-involved vehicle spun into a pole or tree.

To reduce deaths and injuries occurring from fires caused by leaking fuel during and after a crash, Standard No. 301, Fuel System Integrity sets performance requirements for fuel systems in crashes. The Standard limits the amount of fuel spillage from fuel systems of vehicles tested under the procedures in the standard during and after specified front, rear, and lateral barrier impact tests. The standard limits fuel spillage due to these required impact tests to 28.4 grams (1 ounce) by weight during the time from the start of the impact until motion of the vehicle has stopped and to a total of 142 grams (5 ounces) by weight in the 5-minute period after the vehicle stops moving. For the subsequent 25-minute period, fuel spillage during any 1-minute interval is limited to 28.4 grams (1 ounce) by weight. Similar fuel spillage limits are required for the standard's static rollover test procedure. The rollover test is conducted after the front, rear or lateral impact tests.

Previous Post-hoc Studies

There have been five NHTSA studies evaluating the FMVSS 301 standard as it evolved ^{(1), (2), (3), (4), (5)}. All have used data from selected states where fires, and sometimes fuel leaks, could be identified. Cases from national databases were too few for statistical analysis, but were useful in identifying factors that affected fire incidence. The first three focused on fuel leaks and fires rather than injuries and deaths. The fourth and fifth added analyses of injury, death, and overall cost effectiveness.

Study one (Flora) was based on six years of police crash data in Illinois and three years of police crash data in Michigan. These states were chosen because of the quality of their data, especially regarding fuel leakage. Results showed significant reductions in fire rates and leakage rates in passenger cars, and some reduction in leakage rates in light trucks. It could not be proven that the standard was the only cause of these improvements. Data regarding leakage and fires are rare, because these events are quite rare.

Study two (Reinfurt, 1981) looked at fire incidence, comparing passenger cars built before 1968 to passenger cars built after the standard was implemented. Since a fire is more likely in an older car, groups of cars of like ages were compared. The authors used data from the National Crash Severity Study (NCSS) and North Carolina crash data. There were too few data for investigating fuel spillage in cars. The data from the NCSS were too scant for anything but obvious conclusions, i.e., fires were more likely when cars collided with trucks or fixed objects. The data from North Carolina showed a significant increase in the probability of fires in cars built after the standard took effect. This conclusion held even after confounding factors (such as speed, side of the car hit, and age of the car) were statistically controlled through logistic regression.

Study three (Reinfurt, 1982) was similar to the second but assessed the effectiveness of the 1976 version of FMVSS 301. They again used North Carolina data, and replicated the findings from Maryland data. Fire rates in 1969-1975 model-year vehicles were compared to fire rates in 1976-1981 model-year vehicles. This time, there were decreases in the probability of fires. The decreases were statistically significant in North Carolina, but not in Marlyland, which had less than half the cases that North Carolina had.

The fourth study (Parsons, of NHTSA, 1983) was an overall evaluation of the FMVSS 301 which was in force in 1975. It was based on state data from Illinois, North Carolina, Maryland, Pennsylvania, and especially Michigan because they had additional data on fuel spillage. It concluded that FMVSS 301 "has significantly reduced post-crash fires in passenger car crashes;.

The reduction in crash fires has resulted annually in: 400 fewer fatalities, 520 fewer serious injuries, 110 fewer moderate injuries, and 6,500 fewer crash fires." (Page i) However, these numbers were derived from one state's data, and were rescinded in a later study (see next).

The fifth study (Parsons, 1990) was a re-examination of FMVSS 301 which concluded there were reduced fires due to the various permutations of FMVSS 301, but there were not enough data to emphatically conclude that lives had been saved. The author suggested that the speed threshold in the present standard may be too low: " . . .[I]t appears that the bulk of the fire hazard for vehicle occupants involved in fire crashes is focused at the upper end of the severity spectrum - i.e., the risk of serious injury or fatality. Since these crashes typically involve high levels of crash or impact severity, it is possible that these levels typically exceed the 20 to 30 mile per hour threshold set by FMVSS 301. Data developed in Chapter 3 indicate that most fatal crashes involving fire occur at speeds higher than these." (Page 4-6).

Overall, these studies are inconclusive. There is general agreement that fires have been reduced, but estimates of death and injury benefits from the original FMVSS 301 are low to none. Generally, they were not able to collect sufficient, useful data from the proper model years to conclude that deaths and injuries due to motor vehicle fires have been reduced significantly due to the standard. In addition, they investigated all post-crash fires, whereas the proposed upgrade is expected to affect mainly fuel-tank fires resulting from rear impacts. The previous studies did not have enough data to significantly differentiate fuel-tank fires from motor-compartment fires, or rear impacts from other impacts.

Studies Leading to the Proposed Upgrade

NHTSA contracted with GESAC, Inc. ⁽⁶⁾ to examine FARS and NASS crash data to determine the types of crashes that were causing fire-related fatalities and injuries and developed a new crash test procedure to simulate the most frequent crash scenario that leads to fire and fire related fatalities and injuries. GESAC selected for detailed analysis 150 NASS cases involving fire and any occupant injury of AIS 2 or greater. One of the objectives of the study was to suggest a crash test to simulate crashes that cause fire. The suggested crash simulation includes impact mode, speed, barrier type, impact location and barrier orientation.

For vehicles receiving rear damage, the report indicated that a moving deformable barrier with a partial overlap (the percentage of the rear width of the vehicle involved in the crash) would simulate the most common crash. Overlap ranged from 30 to 95 percent with an average level of 71 percent. In rear impacts, the delta V ranged from 11 km/h to 73 km/h (7 to 45 mph) with a medium delta V of 42 km/h (26 mph). There were 11 rear impacts with delta V estimates and burn injuries and NHTSA concluded that further study was needed.

A detailed case study of 214 fire related crash cases from 1990 to 1993 FARS data was conducted to help determine the relationship between vehicle crash specifics and fire fatality outcome. ⁽⁷⁾ Crash records were retrieved from seven states which recorded more complete case histories regarding fire crashes. There were 65 burn related (not trauma related) fatalities in 45 of these cases. Thirty of these fatalities occurred in rear impacts. The determination of whether the fatality cause was fire related or trauma related was based upon autopsy reports (5), coroner or death certificate (14), or the author's judgment (11). A thorough review of the crash conditions in these rear impact cases concluded that striking a stationery vehicle at 50-55 mph with a moving deformable barrier (MDB) at a 70 percent overlap would provide a reasonable crash simulation of real world cases involving a rear impact resulting in fatal burns.

This study also estimated that there are 143 burn fatalities annually in rear impact crashes (a confidence interval around that estimate was also provided at 95 to 195 burn fatalities annually in rear impact crashes). However, these estimates appear high. These estimates are based on the16 rear-end crashes that resulted in 30 fatalities with fires. One crash in the sample involved eight fatalities. The number of fatalities per crash in this sample is much higher than typical 30/16 = 1.875 compared to the 1998 totals of 112 fatalities in 103 vehicles (112/103 = 1.087) in rear impacts. In a more narrowly defined group, passenger vehicles struck in the rear by passenger vehicles, the ratio for 1998 was 40 fatalities in 26 crashes (40/26 = 1.538). Thus, the sample may over represent the importance of rear impacts as part of the fire population. In addition, the 143 burn fatalities in rear impacts estimated in the case study is much higher than the 112 fatalities reported in FARS for 1998 in which a passenger car or light truck were struck in the rear and there was a fire, without trying to control for whether the fatality were burn related or trauma related. Thus, this estimate is not reliable and not used in this analysis.

When considering the test set-up for the standard, one must consider the striking vehicle weight. In the 16 cases of the sample, there were 8 passenger cars, 2 light trucks, and 6 heavy trucks. Thus, heavy trucks are over represented in the fatal rear crashes that involve fire. This finding, that heavy trucks are over represented as being the striking vehicle causing fires, appears in both the sample of crashes investigated and in the yearly FARS counts.

The cases in this study included photographs and witness accounts, which greatly enhances estimates of impact speed. However, only eight of these cases had sufficient detail to estimate impact speed, and three of these involved a heavy truck as the striking vehicle, probably a more forceful crash than the proposed test.

The 1995 ANPRM

On April 12, 1995, NHTSA published an Advance Notice of Proposed Rulemaking (ANPRM) (60 FR 18566) on FMVSS No. 301. We announced our plans to consider research and rulemaking activities in three areas:

1)   define performance criteria for fuel system components,

2)   modify the existing FMVSS No. 301 crash test procedures and performance criteria to better simulate the crashes that lead to serious injury and fatalities in fires, and

3)   define the role of environmental and aging factors, such as corrosion, as it affects fuel system integrity

This NPRM only addresses the second of these research agendas.

The Proposal

The current rear impact test of FMVSS 301 requires that the entire rear of a test vehicle is to be impacted by a 1,814 kg (4,000 lbs.) rigid barrier at speeds up to 48 km/h (30 mph). The proposal would require an offset rear crash test procedure specifying that 70 percent of the rear of the vehicle would be impacted by a 1,368 kg (3,015 lbs.) deformable barrier at 80 km/h (50 mph).

The current side (lateral) barrier test of FMVSS 301 requires the 1,814 kg (4,000 lbs.) moving barrier at 32 km/h (20 mph) to strike the side of the vehicle centered on the driver's seating reference point. This test would be replaced by measuring the amount of fuel spillage in the FMVSS 214 test with a 1,368 kg (3,015 lbs.) crabbed deformable barrier striking the vehicle at a specified point (see FMVSS 214, S6.11) in a 53.6 kph (33.5 mph) impact. Since the FMVSS 214 test is run anyway by manufacturers whose vehicles weigh less than 2,722 kg (6,000 lbs. GVWR or less), using this test for the lateral barrier test of FMVSS 301 saves the testing cost and the cost of a vehicle.

The current frontal barrier test of FMVSS 301 is a 48 km/h (30 mph) impact into a fixed collision barrier. No changes are proposed for this test.

International Harmonization

There are three other standards concerning fuel system integrity in the world:

1)   The Canadian CMVSS No. 301, Fuel System Integrity Standard is identical to the U.S. FMVSS No. 301.

2)   The Economic Commission for Europe (ECE) Regulation No. 34, Uniform Provisions Concerning the Approval of Vehicles with Regard to the Prevention of Fire Risks (01 Series, Amendment 1, January 1, 1979) has been adopted by 13 European countries. This regulation requires a 48 to 53 km/h frontal fixed barrier impact test and a 35 to 38 km/h rear moving flat barrier impact test. The flat barrier weighs 1,100 kg(+ or - 20 kg). A pendulum can be used as the impactor. ECE Reg. No. 34 does not require a rollover test. The regulation does require a hydraulic internal-pressure test for all fuel tanks and special tests (impact resistance, mechanical strength, and fire resistance) for plastic tanks.

3)   The Japanese Standard, Technical Standard for Fuel Leakage in Collision etc. (Amended on August 1, 1989). The Japanese standard requires a 50 km/h (+ or - 2 km/h) frontal fixed barrier impact test and a 35 to 38 km/h rear moving flat barrier impact test. The flat barrier weighs 1,100 kg(+ or - 20 kg). A pendulum can be used as the impactor.

Thus, no other country has a standard with a deformable moving barrier hitting the vehicle in an offset mode with the combined stringency of the test proposed, considering both the weight of the barrier and the speed of the test.

III.   Test Results

REAR IMPACT TESTS

The agency has conducted 20 rear impact crash tests at 80 km/h (50 mph) with the proposed moving deformable barrier (MDB) to determine what percent of the vehicles meet the proposed criteria and to demonstrate that the proposed procedure can be withstood by even the smallest passenger vehicles in the US market today. This was not a representative sample of vehicles that were tested, but a selection biased toward smaller vehicles where it was believed that more problems would be found.

In eleven of the 50 mph MDB rear impact crash tests, the vehicles met all the FMVSS No. 301 fuel leakage criteria requirements. These tests included a 1993 & 1996 Ford Mustang, a 1996 Plymouth Voyager, a 1996 Chevy Blazer, a 1998 Chevy Metro, a 1999 Mazda Miata, a 1998 Nissan Sentra, three 1998 Honda Civics and a Chevy Cavalier. Two other Cavaliers failed the test, thus for cost/ benefit purposes the Cavalier is considered a failure.

In nine 50 mph MDB rear impact crash tests, the vehicles failed at least one of the FMVSS No. 301 fuel leakage criteria requirements. These tests included two 1993 Ford Mustangs, a 1996 Suzuki Sidekick, a 1996 Dodge Neon, a 1996 Geo Prizm, two 1998 Chevy Cavaliers, a 1998 VW Jetta, and a 1998 Ford Escort.

Table III-1 presents the 13 different nameplates tested (only the 1996 Ford Mustang is used since it is the more current model). Seven vehicles passed and six vehicles failed the proposed test. The 13 models ranged in weight from 2,198 pounds (1998 Chevy Metro) to 4,290 pounds (1996 Plymouth Voyager). Test weight is defined as curb weight plus 300 pounds to account for the driver and passengers or luggage. The pattern of failures did not show a direct relationship to weight (see Table III-1, which are in order by vehicle weight). The two lightest and the three heaviest vehicles passed the proposed test and there was no linear pattern among the other eight. However, since larger vehicles have more structure to absorb collision forces, and since the deformable barrier for this test is a fixed weight (3,015 lbs.) regardless of the size of the tested vehicle, there is a logical reason to expect that larger vehicles are more likely to pass. Although the data show no direct correlation between passing and weight for most of the sample, the 3 heaviest vehicles all passed. This indicates that there may be a weight threshold beyond which failure is unlikely, even at 50 miles per hour (80 km/hr). To understand the implications of this for benefits and compliance costs, three compliance scenarios were examined. These are described in the following models:

Model 1: In Model 1 it was assumed that 60% (6 failures /10 vehicles tested that were lighter than the Mustang) of all vehicles below the weight of the Ford Mustang (1,628 kg or 3,582 lbs.) would require a fix. The Ford Mustang was used as a cutoff because it was the lightest of the three heaviest vehicles which passed the test. This model assumes that had NHTSA tested any other vehicles above that weight, they also would have passed.

Model 2: In Model 2, it is assumed that 46% (6 failures /13 vehicle models tested) of all vehicles at or below the weight of the Plymouth Voyager (1,946 kg or 4,281 lbs.) would require a fix. The Voyager was the heaviest vehicle tested.

Model 3: This model assumes that 46% (6/13) of all vehicles, no matter how heavy, would require a fix. Obviously, it is the most conservative and expensive model.

Table III-2 shows how the vehicle sales are distributed between the weight options and the models.
Under Model 1, 22 percent of the fleet would need improvements to pass the proposed test.
Under Model 2, 33 percent of the fleet would need improvements to pass the proposed test.
Under Model 3, 46 percent of the fleet would need improvements to pass the proposed test.

Table III-1
Test Results Used in the Analysis

Table III-2
Estimating the Percent of Sales Needing Improvements
Weight vs. Model Designations

Lower Weight Range = Up to 1,361 kg. or (3,000 lbs.)
Medium Weight Range = From 1,361 to 1,707 kg. or (3,000 to 3,765 lbs.)
High Weight Range = Over 1,707 kg. or (3,765 lbs.)

Rear Impact Test Impact Energy

The proposed 50 mph FMVSS 301 rear impact test is significantly more stringent than the present 30 mph requirement, since it will increase the barrier's impact kinetic energy (KE) by about 110%. The barrier's initial KE = � [the barrier's mass (m_b) x test speed (v_b) x test speed (v_b)]. To perform the calculation, the test speeds of 50 and 30 mph are converted to the equivalent metric units 22.35 and 13.4 m/sec.

Thus, the present barrier impact KE is � [1,814 kg x 13.4 m/sec x 13.4 m/sec] = 162,860 joules, and the proposed barrier impact KE is � [1,368 kg x 22.35 m/sec x 22.35 m/sec] = 341,673 joules.

We can calculate the approximate resultant change of velocity of the moving barrier and test vehicle due to the impact if we ignore the energy absorbed by the aluminum honeycomb on the MDB and crush deformation of the test vehicle's rear end, by applying the principle of the conservation of linear momentum. Where:

m_b v_b = (test vehicle mass) m_t x (final velocity of the barrier & test vehicle) v_f + m_b v_f

Which can be simplified to:     v_f = v_b [m_b � (m_t + m_b)]

Figure III-1 shows the resultant delta v for vehicles between 907 and 2,721 kg (2,000 and 6,000 lbs) when impacted with the present barrier test compared to being impacted with the proposed barrier test. With a 2,000 pound vehicle for example, the present barrier test would result in the vehicle experiencing a 20 mph delta V, while the proposed barrier test will cause the delta v to increase to 30 mph, which is a 50% increase in delta v. Similarly, the delta v of a 4,000 pound vehicle is increased from 15 mph to 21.4 mph, which is a 43% increase in delta v, and the delta v of a 6,000 pound vehicle is increased from 12 mph to 16.7 mph, which is a 33% increase in the delta v.

SIDE TESTS

In response to the ANPRM, Chrysler, Ford, GM, Volvo, AAMA, Advocates, and IIHS all supported replacing the current Standard No. 301 side impact test with the current Standard No. 214 test. Most commenters argued that Standard No. 214's test was more stringent and more representative of real world crash conditions than Standard No. 301 lateral impact test.

Only a few vehicles have failed the fuel leakage requirements of Standard No. 301 in side impact tests. Since 1994, one vehicle out of 43 tested in compliance tests to Standard No. 301 failed. In compliance tests for Standard No. 214, the fuel leakage is routinely measured (although not required), and one out of more than 100 vehicles failed the Standard No. 301 leakage requirement. In side impact NCAP tests, run at 61.6 km/h (38.5 mph), three of 103 vehicles leaked fuel in excess of the Standard No. 301 requirements.

NHTSA has compared the crash test results of a Standard No. 301 lateral impact compliance test condition and a Standard No. 214 compliance test for the same make/model. Our analysis indicates that the fuel system components are exposed to more stringent requirements in the Standard 214 test than in the present Standard No. 301 test. The Standard No. 214 test exposes the subject vehicle to higher crash forces, greater changes in velocity, and higher absorbed crush energy than the present Standard No. 301 test. Thus, the agency believes the Standard No. 214 test is stricter, providing a small benefit, and will reduce testing costs at the same time.

Since only one out of more than 100 vehicles failed the fuel leakage requirements in the proposed lateral test using the FMVSS 214 procedure, it is assumed that there would be essentially no quantifiable vehicle costs and no quantifiable injury or fatality benefits for the change in the lateral impact test. Changing to the slightly stricter FMVSS 214 procedure will provide some benefit, but with less than 1 percent of the vehicles failing this test, the benefits are not quantifiable. There are quantifiable benefits in terms of cost reduction, which will be discussed later.

NCAP TESTS

The agency has performed fuel leakage tests in its New Car Assessment Program (NCAP). The NCAP program is run at speeds that are 5 mph higher than the safety standards. Thus, what are called "failures" regarding NCAP tests below are not compliance failures, but a failure to meet the same performance standards in terms of fuel leakage at a higher impact speed.

In frontal impacts at 35 mph into the barrier, there have been 10 failures out of 406 tests, since 1979. On MY 1995 to MY 2000 vehicles, there have been four failures out of 232 vehicles tested (1.7 percent).

In rear impact tests run at 35 mph on MY 1979 to MY 1981 vehicles, there were 14 failures out of 52 vehicles tested (26.9 percent).

In side impacts run at 38.5 mph on MY 1997 to MY 2000 vehicles, there have been three failures out of 103 vehicles tested (2.9 percent).

REAR IMPACTS

Fatal fires are very rare. The percentage of passenger vehicles with fire in the FARS data has ranged from 2.5% to 2.8% between 1991 and 1998. In FARS there is little or no data which would allow us to classify crash energies as 1) below the present standard, 2) between the present and proposed standard, or 3) above the proposed standard. Only those crashes falling into category 2 are in the target population that could show benefits for this upgrade in the rear test.

An Analysis of FARS Data

We searched the Fatality Analysis Reporting System (FARS) for crashes where the change in the rear test for fuel system integrity might reduce fatalities. Table IV-1 shows how we selected the cases.

FARS data do not contain crash estimates of force, aside from "extent of deformation," which in these cases was always at the highest value. Nor do they contain speed estimates, except in rare cases. However, based on data collected in the NASS-CDS, some of the crash speeds were well above the proposed test speed. Therefore, the fatalities in Table IV-1 represent a high estimate of the number of lives in the target population that might be affected by the proposed change in the rear 301 test.

The obvious problem with these data is there is no indication of the forces involved, other than each crash was forceful enough to cause a fire and kill an occupant. There is also no evidence of what killed the occupants, or whether the crash was survivable had there been no fire. Nevertheless, the number of cases where the occupants would have survived without the fire is very probably smaller than the numbers in Table IV-1.

An inquiry was made concerning linkage between FARS fatalities and the Medical Examiner data from the National Center for Health Statistics. In the case of fiery crashes, these linkages have been of low quality and lead to no useful conclusions.

An examination of the cases in 1998 FARS provides more information of how these numbers fit together. In 1998 FARS in any type of impact (front, side, rear, rollover, multi-vehicle or single vehicle ⁽⁸⁾), there were 1,638 vehicles of any type (passenger car, light truck, heavy truck, motorcycle, bus, others) which had a fire in the vehicle, in which there were 1,610 occupant fatalities. In rear impacts (clock positions 5, 6, or 7, using Principal Impact rather than Initial Impact), there were 127 vehicles of any type which had a fire in the vehicle, in which there were 122 occupant fatalities.

When looking at only passenger cars and light trucks, in 1998 FARS in any type of impact, there were 1,303 vehicles which had a fire in the vehicle, in which there were 1,411 occupant fatalities. In rear impacts, there were 103 vehicles which had a fire in the vehicle, in which there were 112 occupant fatalities.

Taking this one step further, when looking at only passenger cars and light trucks being struck in the rear, in 1998 FARS in multi-vehicle rear impacts in which the striking vehicle was a passenger car or light truck, there were 37 vehicles which had a fire in the vehicle, in which there were 41 occupant fatalities. Thus, by taking out those cases in which a striking vehicle was a heavy truck or bus, those cases in which the fire was in a striking passenger car or light truck, and the single vehicle crashes, the number of occupant fatalities in the target population in 1998 dropped from 112 fatalities to 41 fatalities. [Note that the analysis uses the average number of fatalities over the 8 year period (49) and not just for 1998 FARS.] This does not say that there is no chance that improvements made to vehicles to pass the proposed test would be beneficial in the other crash modes. The target population analysis is designed to conservatively estimate how many fatalities there are in the particular crash mode being simulated by the test. The test uses a 3,015 pound MDB, simulated by a multi-vehicular crash with the striking vehicle being a passenger car or light truck. We took out those cases with large vehicles as striking vehicles because of the greater forces involved. We acknowledge that some of these crashes may be below the test speed, but to be conservative we took out those cases with large vehicles as striking vehicles.

The test only considers fuel leakage in the struck vehicle. Remember that this target population still has not taken into account the possibility that the occupant may have been killed by the impact of the crash. The data only indicate that a fire was present in the vehicle in which the occupant died.

There is a possibility that the real target population is larger due to: (1) the fuel leakage or fire could start in the struck vehicle and spread to the striking vehicle, causing a fatality in the striking vehicle due to fire, (2) as discussed earlier, there could be some heavy truck striking vehicle impacts that the countermeasure could be effective in reducing, and (3) there could be some single-vehicle impacts that the countermeasure could be effective in reducing. Thus, the target population chosen is conservative.

The average number of fatalities over these eight years is 49, with a standard deviation of 10. The average number of case vehicles is 34, with a standard deviation of 7. Note that there are no significant annual trends up or down in either vehicles or fatalities.

We queried the Crashworthiness Data System (CDS) from 1991 through 1998 to get an estimate of the percentage of deaths caused by fire. There were 12 cases involving passenger vehicles being struck in the rear by another passenger vehicle resulting in the death of an occupant in the struck vehicle, where the struck vehicle caught on fire. In 10 of those 12 cases, the deaths were determined to be caused by burn injuries (83 percent). On a weighted basis, you get about the same percentage (132 out of 153 weighted deaths were caused by fire or 86 percent). Therefore, the 49 deaths were multiplied by (132/153), which amounts to an annual estimate of 42 fire-caused deaths in these struck vehicles under the test conditions.

Additionally, there are burn fatalities in the striking vehicle in some crashes when the fire starts in the struck vehicle but then spreads to the striking vehicle. In the 1998 Ragland and Hsai study, eight of the thirty fatalities in rear-end crashes studied were in the striking vehicle. All of these fatalities were due to burns. Therefore, we adjusted the 42 deaths upward by a factor of (30/22=1.36), so 1.36 x 42 = 57 as the final estimate of the annual target population of burn related fatalities that could be affected by this proposal.

We could have used the most harmful event code from the FARS file to distinguish cases in which the fires was the cause of death. However, the agency does not believe this coding has the best accuracy. There was a study dealing with the FARS coding of fire or explosion as the most harmful event. ⁽⁹⁾ This study found significant differences between the states for the percent of vehicles for which fire was coded and for which fire or explosion is coded as the most harmful event. The analysis suggests that it is "extremely unlikely" that the states are measuring the same phenomenon. The percent of vehicles experiencing fire ranged from 0.11 percent to 5.30 percent, among the states. The percent of fire or explosion as the most harmful event, given that fire was coded as occurring during the crash, ranged from 0.56 percent to 95.92 percent, among the states. Thus, this was not considered a reliable code without further analysis of the case.

The same study looked at the Multiple Cause of Death file and found inconsistencies between the Multiple Cause of Death file and the FARS Most Harmful Event data. Again, the agency determined that it is best to examine the cases individually, with as much data as are available on each case, to get the most useful and correct information.

Table IV-2 provides a breakdown of 1998 FARS data by single vehicle and multi-vehicle collision involving fire and by type of collision. The top of the table provides the number of vehicles with fire and the bottom of the table provides the number of fatalities in those vehicles with fire.

Table IV-1
Selected data on passenger vehicles with fire from 1991-1998 FARS

Table IV-2
A Breakdown of 1998 FARS Data

Number of Vehicles
Vehicles in FARS with Fire	1,638
Passenger Cars and Light Trucks with Fire	1,303
PC and LTV with no heavy trucks involved	1,053

Non-Fatal Injuries

Based on the 1998 National Automotive Sampling System - General Estimates System (NASS-GES) there were an estimated 4,000 passenger vehicles involved in injury crashes with fire and 5,000 property damage only crashes with fire.

We used the National Automotive Sampling System - Crashworthiness Data System (NASS-CDS) to estimate the number of injuries and the injury levels that presently occur in fire-involved rear passenger vehicle crashes. We examined eight years of data, from 1991 through 1998 and found only one case where a non-fatal MAIS was caused by fire in a passenger vehicle struck in the rear. (The Maximum Abbreviated Injury Scale (MAIS) indicates the most heavily injured area of the victim's body). In this one non-fatal case the burn injury was an AIS 5 injury and there was another non-burn related AIS 5 injury. On an annualized basis, this one case represented 2 occupants nationwide. All the other cases in which a burn was the maximum injury to the occupant were MAIS 5 or MAIS 6, and were eventually fatal.

Therefore, we re-analyzed the data to look for any fire-caused injury, not just the worst injury to the occupant. We ran a NASS-CDS analysis that looked directly at the Abbreviated Injury Scale (AIS rather than the MAIS). As before, we only looked at passenger vehicles with fire, struck in the rear by another passenger vehicle. We could have included motorcycles as a striking vehicle, but there were no such cases where the car or truck caught on fire after the impact. We did investigate where the fire started. It is possible for a vehicle to have a motor compartment fire, pull off the roadway, and be struck in the rear by an inattentive driver, without causing a fuel-tank fire. However, no such cases appeared. All cases that met the criteria above originated in the fuel tank. Only four cases (not counting the one case of AIS 5 discussed above) in eight years were found, with an average annualized sampling weight of 117. Therefore, we estimate that there were 119 non-fatal injuries annually due to fire in the target population. The distribution of those injuries is 108 AIS 1 injuries, 9 AIS 3 injuries, and 2 MAIS 5 injuries.

V. Costs and Leadtime

The agency believes that most of these fuel leak test failures could be eliminated with minor design revisions. The agency believes these minor design changes will not translate into significant consumer cost increases. The changes would include improvements in the fuel filler neck in cases like the Geo Prizm and Ford Escort; an additional weld to the suspension component failing on the Chevy Cavalier; and rerouting of the fuel lines on the Dodge Neon. The failure mode of the tested MY 1996 Suzuki Sidekick was not determined. However, the 1996 Geo Tracker (which is the same design as the Suzuki Sidekick) failed the FMVSS 301 compliance test due to a deformed gusset plate puncturing the fuel tank wall. The MY 1997 vehicles were redesigned and the MY 1997 Suzuki Sidekick passed the FMVSS 301 compliance test.

The improved performance of the redesigned 1996 Ford Mustang vs. the 1993 Ford Mustang is an example of how vehicles that have failed the proposed upgraded requirements can be redesigned to meet the upgraded rear impact fuel system performance. The agency believes that most current vehicles that would fail the proposed requirements would only require minor design changes and given adequate lead time, the costs to comply could be minimized.

Per-vehicle costs for those which currently fail the proposed rear-impact test depend on the type of failure. Cost, Weight, and Lead Time Analysis (Contract No.DTNH22-96-12003, Task Order 004) studied three types of failure that appeared in NHTSA's series of crash tests using the proposed moving deformable barrier. The contractors looked at a 1993 Ford Mustang, a 1996 Geo Prizm, and a 1996 Dodge Neon. All three of these cases were fixed by redesigns. Thus, we had noncomplying and complying vehicles with the proposal. The Mustang failed due to a fuel tank rupture, which the researchers decided could be fixed by a fuel tank guard. The 1999 Mustang has such a guard, and was the basis for the first set of cost and lead time estimates. The Geo Prizm fractured at the junction of the filler neck and the fuel tank. The 1999 Geo Prizm has a filler neck which is flexible and compressible, allowing more deformation between the rear fender and the fuel tank before rupturing. It was the basis for the second set of cost and lead time estimates. A flexible filler neck should fix problems on the Geo Prizm and Ford Escort test vehicles. The Neon failed when the sending unit and fuel lines, located on the bottom of the fuel tank, were ruptured. The solution here was to place the sending unit on top of the fuel tank and re-route the fuel lines.

The contractors cost estimates are dependent upon the size of the vehicle. The fix for the first failure type (guard and strap for fuel tank), is estimated to be $3.93 for a lower weight vehicle (e.g. Geo Prizm), $4.86 for a medium weight vehicle (e.g. Ford Mustang), and $5.04 for a heavy weight vehicle( e.g. Mercury Marquis). The fix for the second failure type (flexible filler neck) is estimated to be $3.22 for a lower weight vehicle, $4.64 for a medium weight vehicle, and $6.06 for a heavy weight vehicle. Thus, the per-vehicle consumer cost would range from $3.13 to $6.06, depending on the nature of the required modification. Note also that this is not an average across the entire vehicle fleet. That figure and others are presented in Table V-1, which estimates consumer cost ⁽¹⁰⁾. The fix for the third failure (relocation of sending unit and fuel line) is estimated to add zero retail cost.

Variation of Cost by Make/Model Weight: The only information available on cost by weight was for the three cars described in Cost, Weight, and Lead Time Analysis, namely a Geo Prizm, a Ford Mustang, and a Mercury Marquis. Although the relationship between part cost and vehicle weight was linear for the filler tube, it was not very linear for the tank guard. Therefore, a simpler rule was used to assign cost to each make/model. Two cut-off points (1,361 and 1,707 kg) were used, midway between the weights of the Prizm and the Mustang, and midway between the weights of the Mustang and the Marquis. Below the first cut-off weight, the averaged cost estimated for the Prizm was applied for the lower weight vehicles. Between the two cut-off weights, the averaged cost estimated for the Mustang was applied for the medium weight vehicles. Above the second cut-off weight the averaged cost estimated for the Marquis was applied for the heavy weight vehicles. Light trucks and vans were also classified by their weight.

Costs are estimated using three weight groups. These weight groups are different than those used in Table III-2 for determining the percent of the fleet that needed improvements, since the cost breakdowns were based on the specific models estimated in the cost contract for which specific fixes had been developed. The cost weight breakdowns, which were based on the vehicles identified in the cost contract, are defined as:

Low Weight - up to 1,361 kg (3,000 lbs.)
Medium Weight - 1,361 kg to 1,707 kg.
High Weight - over 1,707 kg (3,765 lbs.)

Cost Averaging: Since there was no way to tell for vehicles not tested, which vehicle make/model would require which fix (or for that matter, some other fix), the two fixes (tank guard and filler neck) which had per-vehicle costs were summed and divided by two for each of the three weight categories. This resulted in consumer cost estimates of $3.58 for the low weight vehicles, $4.75 for medium weight vehicles, and $5.55 for heavy weight vehicles. This assumption ignores a no-cost fix (e.g., the Dodge Neon), but is also ignores potentially more expensive fixes that might be needed for some other vehicles.

Once the cost was applied appropriately for each make/model, it was multiplied by the number of vehicles sold. Sales figures for the 1998 calendar year (15,179,501 passenger cars and light trucks) were used (see Automotive Weekly, January 11,1999). Weights were from Ward's Automotive Yearbook 1999, with a few weights coming from the 1997 edition for models which were sold in 1999 but not in Ward's 1999 table. Some models with few sales and no longer manufactured were omitted. Total cost estimates are summarized in Table V-1.

Cost estimates:

Model 1: In model one it was assumed that 22 % of all vehicles (which includes 60% of all vehicles lighter than the Mustang) would require a fix. This model resulted in a total cost to consumers of $14 million or an average or $4.28 per vehicle affected, or $0.93 over all vehicles.

Model 2: This model is similar, except that it assumes that 33% of all vehicles (which includes 46% of all vehicles at or below the weight of the Plymouth Voyager) would require a fix. This model resulted in a total cost to consumers of $24 million or an average or $4.78 per vehicle affected, or $1.58 over all vehicles.

Model 3: This model assumes that 46% of all vehicles, no matter how heavy, would require a fix. Obviously, it is the most conservative and expensive model. This model resulted in a total cost to consumers of $35 million or an average or $5.00 per vehicle affected, or $2.31 over all vehicles.

Total costs are estimated to be $14 million to $35 million depending on whether the changes are only required for the smallest vehicles or are applied to all vehicles, including pick-ups, vans, and sport-utility vehicles.

At this time, the agency believes that the test failures are more the result of differences in design than they are related to the weight of the struck vehicle. Thus, for the cost effectiveness analysis, the agency believes the estimate of 46 percent of the fleet needing improvements at $5.00 per vehicle is more likely than the other cost models and estimates.

Table V-1
Cost estimates by model

Items not included in the cost estimates:

Fuel costs for added weight were considered inconsequential, given that the heaviest fix, the fuel-tank guard, was less than seven pounds, the added weight for the flexible filler neck was less that 4 ounces, and the relocation of the sending unit and fuel lines added no weight.

Testing Cost Savings

Replacing the side test for FMVSS 301, with the FMVSS 214 (side impact protection) test, would eliminate the cost to conduct the FMVSS 301 test, as well as the cost of a test vehicle. Based on contractor testing costs for NHTSA, the average lateral test for FMVSS 301 costs $4,064 in the year 2000, not counting the costs of the vehicle. An average test vehicle costs about $20,000. Total savings would be about $24,000 per vehicle model (roughly $4,000 to conduct the test and $20,000 for an average vehicle).

The agency believes the cost for the proposed procedure will be essentially the same as the current rear impact test, with one exception. The proposed test includes a deformable barrier. The deformable face, which costs $1,025 each, is destroyed with each test. Current rear impact tests for FMVSS 301 cost $6,300, but include instrumenting the driver dummy for research purposes at a cost of about $1,300 per test. Under the proposal, the tests will cost about $7,325 ($6,300 + $1,025).

Lead Time

Factors that affect the amount of leadtime necessary for the rear impact test proposal include:

1) All vehicles must be tested with the new test, which is at a higher speed than currently used by most manufacturers in their testing. Thus, essentially all make/models must be tested to determine current compliance. This will take about 5 months.

2) For all non-complying vehicles (6 of 13 vehicles tested did not comply) a remedy must be determined, a prototype solution fabricated and installed in the vehicle and the vehicle retested. With potential iterations, this process can take ten months. The design changes we believe are necessary for the vehicles we have tested have been moderate, not requiring retooling.

3) Finally, the changes must be implemented on the production line. This is about a twelve month process.

Thus, in total we estimate the necessary lead time to be about 27 months. If retooling is necessary for some vehicles (we only tested 13 make/models), at least an additional 6 months must be added to the process making the total 33 months.

The agency is proposing about a three year leadtime after issuance of the final rule for the new rear impact test procedure. Since this is a different test than the current Standard No. 301 test, it is not known how many manufacturers have experience with this test procedure.

Since almost all vehicles pass the Standard No. 214 test without fuel leakage and all manufacturers have done these tests on their passenger cars and light trucks and vans up to 6,000 pounds GVWR, the agency is proposing a one year leadtime after the final rule for implementing the Standard No. 214 test requirement for the lateral test.

VI. Benefits Analysis

Rear Impacts

Estimating the effectiveness of countermeasures in reducing the fire-caused fatality and injury problem is very difficult. First, while the rear impact crash scenario represented by the test is typical of fire-causing crashes, one test cannot represent the continuum of crashes, angles and speeds which the rear of light passenger vehicles are exposed to. Thus, while the countermeasure chosen could solve the problem shown in the compliance test, it may not expose other potential problems that could occur in the variety of real world crashes. On the other hand, there might be cases where improvements made to comply with the proposed offset rear impact test could provide benefits in other crash modes (frontal, side, or rollover). Second, there are not many cases available with fire, injury, and known delta V to distribute the cases between those represented by the test procedure and those cases not represented.

The agency examined fire data on a make/model basis and compared it to the test results for the 13 vehicles it tested against the proposed standards. This was an attempt to determine whether there was a correlation between test performance and the probability of fire and burn injuries. Unfortunately, there were not enough data to draw any conclusions in this regard.

In this analysis, we have severely limited the target population to just those rear impact fatalities in which the striking vehicle is representative of the barrier being used in the test (striking vehicles under 10,000 pounds GVWR) and in which the fire was the cause of the fatality. In addition, the proposed test is a relatively severe test, imparting over twice as much energy at impact than the current test. Given these factors, the agency believes that there should be a relatively high effectiveness against fuel leakage and fires for vehicles designed to meet the proposed requirements.

The agency examined the 10 cases in the special study of FARS cases ⁽¹¹⁾, in which a fatality was believed to be caused by fire, of light vehicles being struck in the rear by light vehicles to determine how many of these cases were similar to the test procedure in the impact mode, the percentage overlap, and in the speed of impact. Of these 10 cases, four cases were very similar to the test setup. All four were estimated to be 50 to 55 mph impact speeds, with three of the four being 70 to 80 percent offset crashes; the fourth was 50 percent offset. Two of the 10 cases are not really represented by the test; one because the impact mode wasn't right (more of a rear/side slap than an offset test), and one because the tongue of a trailer was pushed into the fuel tank. One case had an impact speed somewhat higher than the test speed (estimated at 60 mph), and the countermeasure might have been effective given that the struck vehicle was one that we tested and failed at the test speed. The remaining three cases had unknown delta V and unknown percent offset.

The agency also examined the 1981-1993 NASS and FARS cases investigated in the GESAC, Inc. study ⁽¹²⁾. In this study there were 21 rear impact cases where the fire started in the struck vehicle. Eight of these cases resulted in occupant deaths. One case was the same as one of the 10 cases studied above. In two of these cases, death occurred from the trauma of the impact, not from the fire. In the remaining five cases, death occurred from burn injuries. This study used delta V rather than impact speed as the parameter to discuss the force of the impact. As shown in Figure III-1, the test represents a delta V of about 20-30 mph, depending upon the mass of the struck vehicle. Of those five fatal cases, one had a very high delta V's of 42 mph. One case had a delta V slightly higher than the test speed (estimated at 32.1 mph), and the countermeasure might have been effective. Two cases had delta V's similar to the proposed test condition of 21.2 mph and 24.3 mph. One case had a lower delta V of 15 mph, but had the left wheel come off the vehicle allowing it to skid across the road tearing the fuel line. This case is not well represented by our test.

We also examined 1993-2000 NASS-CDS cases and found five rear impact cases in which a light vehicle was struck by a light vehicle with a fire and a fatality. None of these cases matched the previous cases examined. In all five cases, the occupant died from burns. In two of the cases, the delta V was much higher than the test speed (36 mph into a Crown Victoria, and 74 mph into a Nissan). In three cases, the delta V was unknown.

Thus, we have retrieved as much information as possible on a total of 20 fatal cases in our target population. Of those 20, six are represented very well by our test, two have slightly higher speeds, 6 have unknown speeds, 3 have much higher speeds than our test and 3 have different crash circumstances that are not represented by out test. Our best estimate is that 8 to 14 (40 to 70 percent) of the 20 cases we have detailed information on will be well represented by our proposed test. Assuming these 20 cases are representative of the 57 fatalities that occur per year in which a light vehicle is struck in the rear by another light vehicle, resulting in a fire which causes the death, then 40 to 70 percent of the fatality cases could be like the test setup.

We also examined the non-fatal fire cases in the GESAC and later NASS files to determine how many of them were like the test setup. Of the 18 non-fatal fire cases, 7 cases resulted in non-fatal burn injuries, 9 cases had no burn injuries, and 2 cases involved parked cars with no occupants. The burn injuries were one MAIS 5 injury, one MAIS 2 injury ( a case not represented by our test because a trailer hitch ruptured the fuel tank), one AIS 3 injury where the occupant had another AIS 3 head injury, three AIS 1 injuries where the occupant had other AIS 1 injuries, and one AIS 1 where the occupant had another AIS 2 head injury. Of these 18 cases, 8 had delta V much higher than the test condition (33.4 mph, 34.8 mph, 41 mph, 41 mph, 42 mph, 45.3 mph, 53 mph and 54 mph). The delta V of the other 10 cases varied from 10 mph to 25.8 mph, with two unusual cases occurring at low speeds when an electrical fire started in the rear lights and a trailer hitch ruptured a fuel tank. Thus, for these non-fatal fire cases, our test procedure would represent 8 of 18 cases or 44.4 percent.

The next question is: What percent of the fatalities are represented by the vehicles that do not currently pass the proposed test and would make improvements. Some of the 57 fatalities in the target population undoubtedly occur in vehicles that already pass the proposed test. A look at the FARS fatalities found a few in vehicles that had passed the test and a few in vehicles that did not pass the test. Out of the 13 make/models tested, 6 failed the test and would need modifications to certify compliance (46 percent).

If fatalities were evenly distributed over both passing and failing vehicles (46 percent of all vehicles tested failed) and if the effectiveness were 100 percent in those cases with similar crash conditions to the test setup in which the striking vehicle were less than 10,000 pounds GVWR, the number of fatalities reduced by the test would be 10 to 15 (57 fatalities x 0.375 to 0.563 like the test setup x .46 vehicles modified). However, there are two factors that would raise or lower these estimates. Given the narrow target population utilized in this analysis, and the conservative assumptions taken, the agency believes it is reasonable to expect a very high overall effectiveness for these cases, although not 100 percent. For this analysis, we assume an effectiveness of 50 to 75 percent. The test should do a better job of determining those vehicles that are more likely to be involved in fires than an even distribution. In other words, we would expect vehicles that failed the performance test would more likely be over reported in the real world fire population. In addition, once you have a test procedure with engineers trying to determine how to assure compliance, they are likely to find many small changes that can reduce the risk of fires, even in vehicles that originally pass the test. Thus, we would expect that if 46 percent of the vehicles did not pass the proposal, that more than 46 percent of the fires would be in these vehicles. A reasonable expectation would be that the fire risk of the worst performers might be 50 percent higher than the average vehicle or 0.46 x 1.5 = 0.69. Putting together our engineering judgment and assumptions, the range of benefits from meeting the proposed tests are 8 to 21 lives saved (57 fatalities x 0.40 to 0.70 like the test setup x 0.50 to 0.75 effectiveness x 0.69 distribution of fires for failing vehicles).

Side impacts

NHTSA has compared the crash test results of a Standard No. 301 lateral impact compliance test condition and a Standard No. 214 compliance test for the same make/model. Our analysis indicates that the fuel system components are exposed to more stringent requirements in the Standard 214 test than in the present Standard No. 301 test. Thus, the agency believes the Standard No. 214 test is stricter and will provide a small benefit.

Since only one out of more than 100 vehicles failed the fuel leakage requirements in the proposed lateral test using the FMVSS 214 procedure, it is assumed that there would be essentially no quantifiable injury or fatality benefits for the change in the lateral impact test. Changing to the slightly stricter FMVSS 214 procedure will provide some benefit, but with less than 1 percent of the vehicles failing this test and with less than 100 fatalities resulting from fire in multi-vehicle side impacts (see Table IV-2), the benefits are not quantifiable.

Injuries

Although there are non-fatal burn injuries that result from rear impact fires, our data sources (1991 to 1998 NASS-CDS) estimated an average of only 2 cases per year resulted in the burn injury being the most serious injury or tied for the most serious injury suffered by the injured person. This was based on weighting one case and dividing by 8 years. This case also had another non-burn AIS 5 injury at the same injury level. Eliminating the burn injury, although desirable, in itself would not eliminate the most serious injury experienced by the crash victim in any of the cases we examined. For the cost effectiveness analysis, we have chosen to ignore the impact of reduced non-fatal burn injuries.

VII. Cost Effectiveness

To be conservative, this analysis will be based on the highest cost estimate from Table V-1, which is about $35 million. We have also not included a value for the difference in property damage that would result between a crash with a fire and a crash without a fire. The estimated fatality benefits for this rulemaking are 8 to 21 lives saved per year once all vehicles in the on-the-road fleet meet the proposal.

The cost per life saved is estimated to be $1.67 million to $4.38 million ($35 million/21 lives to $35 million/8 lives).

VIII. Small Business Impacts

Regulatory Flexibility Act

The Regulatory Flexibility Act of 1980 (5 U.S.C. �601 et seq.) requires agencies to evaluate the potential effects of their proposed and final rules on small businesses, small organizations and small governmental jurisdictions.

There are two types of businesses impacted by this proposal, first-stage manufacturers and second-stage manufacturers. The types of failures found in testing relate to the surrounding environment of fuel tanks, how the fuel filler necks are attached to the body and how fuel tanks and lines are installed in the vehicle.

Currently, there are about 4 small first-stage motor vehicle manufacturers in the United States. This is not a substantial number. It is unknown how many of their vehicle models currently meet the proposed requirements. Comments are requested on the impact of this proposal on small vehicle manufacturers.

There are a large number of second-stage manufacturers that could be affected by this proposal. Second-stage manufacturers buy a chassis from a first-stage manufacturer and finish it to the consumer's specifications. Many of these manufacturers that put a work-related body on a pickup truck chassis (like a small tow truck) get involved with the fuel system, both in the structure around the fuel tank and where the fuel filler neck attaches to the body. Other second-stage manufacturers use a van chassis or an incomplete vehicle for ambulances, small mobile homes, small school buses, etc. Typically, the first-stage manufacturer provides the second-stage manufacturer with a body builder's guide, which tells the second-stage manufacturer what they can do and still pass along the original equipment manufacturer's certification for compliance with FMVSS 301. In other cases, compliance with FMVSS 301 is not completely a pass-through certification situation, especially for incomplete vehicles. Even when there is a potential pass-through certification, to the extent that a second-stage manufacturer deviates from the guide, they have to certify compliance on their own. The agency does not know how often this occurs. This proposal would make a stricter test for those certifying compliance on their own. However, the agency tentatively concludes that this will not result in a significant economic impact on these companies. Most of the problems these companies will have are in the filler neck area, for which there are known countermeasures being currently used. Comments are requested on this tentative conclusion.

The agency has a negotiated rulemaking underway considering certification problems for first and second-stage manufacturers.

The agency tentatively concludes that this proposal would not have a significant economic impact on a substantial number of small entities.

IX. Cumulative Impacts

Section 1(b) II of Executive Order 12866 Regulatory Planning and Review requires the agencies to take into account to the extent practicable "the costs of cumulative regulations". To adhere to this requirement, the agency has decided to examine both the costs and benefits by vehicle type of all substantial final rules with a cost or benefit impact effective from MY 1990 or later. In addition, proposed rules should also be identified and preliminary cost and benefit estimates provided. A this time, there are no major outstanding proposals that have quantified costs and benefits.

Costs include primary cost, secondary weight costs and the lifetime discounted fuel costs for both primary and secondary weight. Costs will be presented in two ways, the cost per affected vehicle and the average cost over all vehicles. The cost per affected vehicle includes the range of costs that any vehicle might incur. For example, if two different vehicles need different countermeasures to meet the standard, a range will show the cost for both vehicles. The average cost over all vehicles takes into account voluntary compliance before the rule was promulgated or planned voluntary compliance before the rule was effective and the percent of the fleet for which the rule is applicable. Costs are provided in 1997 dollars, using the implicit GNP deflator to inflate previous estimates to 1997 dollars.

Benefits are provided on an annual basis for the fleet once all vehicles in the fleet meet the rule. Benefit and cost per average vehicle estimates take into account voluntary compliance.

Table IX-1
COSTS OF RECENT PASSENGER CAR RULEMAKINGS
(Includes Secondary Weight and Fuel Impacts)
(1997 Dollars)

Table IX-2
BENEFITS OF RECENT PASSENGER CAR RULEMAKINGS
(Annual benefits when all vehicles meet the standard)

Table IX-3
COSTS OF RECENT LIGHT TRUCK RULEMAKINGS
(Includes Secondary Weight and Fuel Impacts)
(1997 Dollars)

Table IX-4
BENEFITS OF RECENT LIGHT TRUCK RULEMAKINGS
(Annual benefits when all vehicles meet the standard)