EPA/AA/CTAB/FE-81-6
Technical Support Report for Regulatory Action
Light Duty Vehicle Fuel Economy Labeling
October, 1980
Notice
Technical support reports for regulatory action do not necessarily
represent the final EPA decision on regulatory issues. They are intended
to present a technical analysis of issues and recommendations resulting
from the assumptions and constraints of that analysis. Agency policy
constraints or data received subsequent to the date of release of this
report may alter the recommendations reached. Readers are cautioned to
seek the latest analysis from EPA before using the information contained
herein.
Control Technology Assessment and Characterization Branch
Emission Control Technology Division
Office of Mobile Source Air Pollution Control
Office of Air. Noise and Radiation
U.S. Environmental Protection Agency
2565 Plymouth Road
Ann Arbor, Michigan 48105
-------
ABSTRACT
In response to the Congressional concerns which mandated the EPA "Section
404 Report", and which motivated the "Moffett Hearings", EPA is initiating
rulemaking proceedings to upgrade certain aspects of the EPA/DOE Fuel
Economy Information Program. This report provides technical input to
those rulemaking proposals.
The results of five surveys of consumers' opinions on fuel economy, and
on various vintages of the fuel economy information program, are summarized,
They indicate a general credibility problem with the current EPA fuel
economy figures.
Design parameter sensitivity algorithms are presented, and the need is
established for their use in deriving fuel economy ratings for untested
vehicles.
Road adjustment factors developed from recent in-use fuel economy
survey data are given for a number of different labeling strategies, and
the strategies compared and contrasted.
Results of a detailed analysis of the effects on label "accuracy" of
inclusion/ exclusion of road adjustment factors and varying levels of
vehicle-specificity in labeling are given. From this study, the incre-
mental benefit of progressive levels of departure from the current
(1981) model type labeling system can be estimated.
-------
-2-
I. BACKGROUND
In 1978, Congress recognized that there was a discrepancy between EPA
fuel economy numbers and actual consumer in-use experience with fuel
economy, and mandated a detailed EPA investigation under Section 404 of 1
the National Energy Conservation Policy Act of 1978. EPA's "404 Report"
is the response to that mandate. That report constitutes an initial
technical basis upon which to develop new labeling rulemakings and
Gas Mileage Guide protocols for the 1982 models.
In February 1980, a set of hearings, known as the "Moffett Hearings"
concerning "Automotive Fuel Economy", were conducted by the U.S. House
of Representatives Subcommittee on Environment, Energy and National
Resources. One of the recommendations in the hearings report by the
Subcommittee was:
"EPA devise a new MPG system for labeling new cars and for the
Gas Mileage Guide that provides fuel economy values, or a
range of values, that most drivers can reasonably expect to
experience".
II. PURPOSE
In response to the above-mentioned Congressional concerns, EPA has
initiated rulemaking proceedings to upgrade certain aspects of the Fuel
Economy Information Program. This Technical Report provides additional
technical backup to that process, beyond that of the 404 Report, by
concentrating on specific issues being addressed in the rulemaking.
* Superscripts indicate references, listed at the end of the text.
-------
-3-
III. DISCUSSION
In this, the body of this report, four areas of discussion are presented;
A. General evaluation of the Label/Guide system;
Prior to discussions of specific technical analysis aimed at
modified fuel economy label figures, this section reviews
survey-type information on public awareness and credibility of
the Fuel Economy Information Program in recent years;
B. Design parameter sensitivity;
This summarizes the technical analysis leading to equations
which account for the fuel economy effects of difference in
vehicle weight, axle ratio, and road load horsepower;
C. Road MPGfactors;
This summarizes the technical analysis which defines and
evaluates labeling strategies and develops road adjustment
factors for those strategies;
D. Adjusted labeling;
Using the 1980 fuel economy data base, this example analysis
shows how various options for methods of labeling affect
distributions of label values.
-------
-4-
A. General Evaluation of Label/Guide System
1. Data Sources
The sources of data used in evaluating the Federal Fuel Economy Information
Program, both Fuel Economy Labels and the Federal Gas Mileage Guide,
were as follows:
3
a. AST Telephone Survey - A survey conducted among approximately
600 people nationwide, using a questionnaire-type prompt list to
gauge their opinions on fuel economy and related subjects. The
respondents were all purchasers of 1976 vehicles. The survey was
conducted in February 1976 under FEA (now DOE) sponsorship.
4
b. EPA Focus Groups - To support the 1979 rulemaking, group
discussions were conducted for EPA by Porter and Novelli Associates
in three cities (Philadelphia, Pa., New Orleans, La., and Van Nuys,
Ca.), utilizing two groups per city, 7-12 people per group (50%
women, 50% men). Total number of participants was approximately
60 people. The participants were either purchasers or potential
purchasers of new vehicles. The focus group discussions were conducted
prior to March 1978.
c. J.D. Power Focus Groups - Focus Group discussions in 17
cities were conducted for DOE by the J.D. Power organization.
There were approximately nine people per group, with two (2) groups
for 13 cities and one group for four cities, resulting in a total
of 30 groups and 246 people. The respondents were all purchasers of
new vehicles for model year 1979.
d. J.D. Power Questionaire Survey - For both MY 1978 and MY
1979, questionaire surveys on the Federal Fuel Economy Information
Program were conducted for DOE by J.D. Power. The survey was a
nationwide random sample of approximately 4000 new car car buyers
and 1000 new truck buyers in each of the two model years noted.
The MY 1978 survey was conducted in January 1979. The MY 1979
survey was conducted in March 1979 with a further follow-up in June
1979. This technical support report deals only with the 1978 and
March 1979 surveys.
e. N.Y. Times Auto Industry Outlook - A survey conducted on its
own initiative by the N.Y. Times in April 1980 of both consumers
and automobile dealers in regard to automotive attitudes, behavior,
and buying plans. There were approximately 1000 consumers and 200
dealerships queried in this marketing survey.
-------
-5-
2. Analysis/Summary
A summary of the outputs from each of the previously noted data sources
follows; results related specifically to the Fuel Economy Label Program
and the Gas Mileage Guide are emphasized.
a. ABT Telephone Survey
Importance of Fuel Economy - 50% of new vehicle buyers considered good
fuel economy an extremely important parameter in their vehicle purchase
decision. 16% considered it the single most important factor.
Fuel Economy Label Awareness - 53% of the respondents actually saw the
label on the vehicle they bought.
Guide Awareness - Awareness of the Guide was very low with only 9% of
the buyers indicating awareness of this publication.
Credibility of Gas Mileage Estimates - 64% of "Label aware" buyers did
not believe the estimates.
Actual Gas Mileage - Buyers' actual gas mileage averaged about 1 mpg
less than the EPA combined estimate.
b. EPA Focus Groups
Representation of Fuel Economy Values - Most participants in the focus
groups indicated that single point estimates, regardless of driving mode
represented, were misleading. An approach which would convey fuel
economy as a range, rather than a point, was deemed most suitable.
Label Information - Information such as range of fuel economy for
comparable vehicles, annual fuel costs, and combined fuel economy estimate
were deemed confusing by most participants in the focus group studies.
Ironically, the first two items are mandated by law.
Gas Mileage Guide - Few participants knew how EPA tested cars; few had
ever seen the Guide.
c. J. D. Power Focus Groups
Credibility of EPA Fuel Economy Figures - In general a lack of credibility
was apparent among most participants; however, some "well informed"
buyers were aware of the comparative nature of the figures and of their
application when making a purchase decision.
Fuel Economy Importance - Fuel economy was a very important factor for
the vast majority of new car purchasers surveyed. The dominant theme
throughout the focus group discussions was that fuel economy was a major
factor in vehicle choice.
-------
-6-
Desired Content of Label - Most respondents indicated that "more accurate"
fuel economy numbers sbould be displayed. Specifically, 44% preferred a
two-number label, with the numbers representing city and highway driving
modes; 22% preferred two numbers representing a range within which most
drivers would fit; 22% preferred three numbers. Only 12% of the participants
indicated a preference for a one-number fuel economy information system.
Availability of Guide - Most respondents indicated that the biggest
problem with the Guide was its lack of availability.
d. J. D. Power Questionnaire Survey
Since this survey represents the most extensive data collection performed,
the output is the most detailed. The following discussion has two
parts: one dealing with the entire survey, with both model years pooled,
and one which compares the two model years' results, highlighting possible
differences between 1978's three-number information system and 1979's
one-number system.
(1). Analysis of J.D. Power Survey (Combined MY78 & 79)
Selection Factor In Buying a Vehicle - Although fuel economy ranked
uppermost in people's choices of vehicle selection factors, EPA gas
mileage figures ranked 8th in importance. As to general, rather than
relative, importance, 64% rated "Fuel Economy" important, while 53%
rated "EPA Mileage Figures" important.
Sources of Information on Gas Mileage - The EPA label was rated the most
popular and most important item in terms of sources of information; the
Gas Mileage Guide ranked 8th. J.D. Power suggests that people associate
the label with fuel economy and since it is readily available (prominently
displayed) it comes to mind. Since the Guide is not readly available
and/or prominently displayed, it does not come to mind as readily as the
label.
Awareness of Gas Mileage Guide - Approximately 4 out of 5 respondents
were not even aware that there was a Gas Mileage Guide.
Consulting of EPA Booklet - Only 1/3 of the people aware of the Guide
actually consulted it.
Selection or Elimination of Vehicle Based on Guide - Only 1/3 of the
people who did consult the Guide used the fuel economy information as a
criterion to select or eliminate a vehicle.
Why is Label Not Useful - Most car and truck buyers who felt the label
was not useful felt so because indicated EPA figures were inaccurate,
unrealistic, non-representative.
-------
-7-
Attitude Toward Fuel Shortage - Only half the respondents indicated they
believed there was a fuel shortage. (Note - these responses were obtained
prior to the 1979 fuel shortage.)
55 MPH Speed Limit - 60% of the respondents indicated they believed the
55 MPH limit would save fuel, but only 48% felt it should be vigorously
enforced.
Government Mandate to Build Fuel-Efficient Vehicles - Approximately
3/4 of the respondents indicated the Government should require the
manufacture of fuel-efficient vehicles.
Driving Style Affects Gas Mileage - The respondents agreed overwhelmingly
(90%) that driving style affects fuel economy.
Gas Mileage Record Keeping Practices - Approximately 40% of the respondents
indicated they frequently checked gas mileage; another 50% indicated
they have some feel for their gas mileage.
Usefulness of the Guide Among Those Who Consulted It - Approximately 55%
of the passenger car buyers who were aware of the Guide indicated that
the Guide was of some use to them; for truck buyers, the figure was
lower, 39%.
Availability of Guide in Auto Dealerships - Only half of the respondents
who were aware of the Guide indicated that it was available at the
dealerships they visited.
How Did You Locate the Booklet (Guide) - Less than half of those people
who obtained Guides got their Guides from dealers.
Awareness of Label on Vehicle - Approximately half of the truck buyers
and 1/4 of the car buyers did not recall seeing the fuel economy label
on the vehicle.
Use of Label Information for Vehicle Comparisons - Approximately 1/3 of
car buyers and 1/5 of truck buyers who were aware of the label used the
label information for comparisons.
Usefulness of Label for Purchase Decision - Less than half of label-
aware car and truck buyers felt it was of use in their purchase decision.
Knowledge of EPA City Rating for New Vehicles - More than half of the
car and truck buyers did not know the EPA city rating for the vehicle
they purchased.
Actual Gas Mileage Compared With EPA Rating - Approximately 40% of the
respondents indicated that they equaled or bettered their vehicle's EPA
city rating. Another 40% achieved MPG lower than the city ratings.
With less than half knowing what the EPA ratings were (see previous
item), these percentages represent perceived rather than actual results.
-------
-8-
Why is Your Mileage Less Than EPA Rating Approximately 2/3 of
those who felt they achieved less than the ratings stated it was because
the EPA figures are overly optimistic; about 1/4 indicated their driving
or vehicle/engine may be causes.
EPA Figures Overstate What Owners Get - 60% to 70% of all respondents
agreed with this observation.
Why are EPA Figures Overstated - More than half of the respondents who
felt the EPA figures were too high blame the EPA test procedures,
either as intrinsically unrealistic, or as not being followed.
EPA Figures Serve Useful Purpose for Comparisons - Approximately 2/3 of
car and truck buyers did feel that, even though the EPA figures over-
stated in-use MPG, they were useful for comparison.
(2) Analysis of J.D. Power Survey (1979 versus 1978)
One of the key items which can be inferred from the J.D. Power survey by
comparing the model year 1978 & 1979 responses is relative consumer
preference for either a single number or a multi-number fuel economy
label program. Since no direct questions were included in the survey
regarding the "number of numbers", however, this evaluation of consumer
preference had to be arrived at by inference from some of the existing
questions in the survey. In MY 1978, EPA fuel economy was expressed as a
three number system; City, Combined, and Highway. In MY 1979, EPA fuel
economy was expressed as a single number system, the City number. In
reviewing the Power survey, the following information is pertinent to
the "number of numbers" issue*:
Importance of EPA Gas Mileage Figures in New Car Purchase Decision - In
1978, 86% of the respondents indicated EPA figures were important in
their purchase decision; in 1979 83% felt that way.
Usefulness of Label Gas Mileage Figures - In 1978, 46% of the respondents
indicated EPA figures were useful; in 1979 38% felt that way.
Usefulness of Gas Mileage Guide F.E. Figures - In 1978, 83% of the
respondents indicated EPA Gas Mileage Guide figures were useful; in 1979
80% felt that way.
Too Much Information on Label - In 1978, 59% of the respondents indicated
the label had too much information; in 1979 63% felt that way!
EPA Gas Mileage Figures Useful for Comparison - In 1978, 76% of the
respondents indicated the label figures were useful for comparison; in
1979 72% felt that way.
* Reference 8 contains a more extensive evaluation of this issue,
including in its appendices much more detail on the survey
results than are given here.
-------
EPA Fuel Economy Test Procedure Is Not Representative - In 1978, 17% of
the respondents indicated that the EPA fuel economy test procedures were
not representative; in 1979 20% felt that way.
Label MFC Should Reflect More Representative Driving - In 1978, 28% of
the respondents indicated that the EPA tests should reflect the "real
world" to a greater degree; in 1979 35% felt that way.
Considering the directions in which the response percentages moved from
1978 to 1979, it could be concluded that the 1979 system was consistently less
satisfactory than the 1978 system. Viewing the statistical insignificance
of the generally small numerical differences, however, it can be only
concluded that there certainly was no improvement brought about by the
change from 1978 to 1979 in the Federal Fuel Economy Program (i.e. from
a three number to a one-number system).
According to the ABT survey, in 1976 only 16% of prospective buyers
indicated that fuel economy was their primary concern, yet by MY 1980,
according to the N.Y. Times survey, 61% of the potential buyers indicated
that fuel economy was their primary concern. During this time interval,
according to the J.D. Power survey, there was an apparent slight decrease
in interest in fuel economy information as reported by the Federal
Government. This leads to the conclusion that although people became
more interested in fuel economy, they became less interested in Federal
Fuel Economy data, and suggests that the change from the three number to
the one number system may have had something to do with this loss of
interest.
e. New York Times Auto Industry Outlook Market Survey
Importance of Fuel Economy - Fuel economy is the primary concern of 61%
of potential new car buyers.
Best Way to Cut Fuel Consumption - Reducing unnecessary driving is the
most frequently (27%) mentioned method of reducing fuel consumption by
potential new car purchasers.
Sales versus Size - Over half (59%) of the dealers queried indicated
increased sales of compacts and sub-compacts while larger sized vehicles
showed decreasing sales.
Reason for Depressed Sales of U.S. Autos - Beyond "economic recession"
and "price of gasoline", the most often mentioned teason for depressed
sales of U.S. autos was "poor gas mileage of American cars".
Government Regulation of Minimum MPG - Over half (54%) of the dealers
queried indicated that they opposed government mileage regulations.
This question was not asked in the consumer survey.
Sales Message Most Effective in Moving Cars - Emphasis on economy/gas
mileage is considered the most effective sales message by dealers (97%)
regardless of vehicle type (import or domestic).
-------
-10-
3. Conclusions
It is readily apparent that, upon reviewing the results of the aforementioned
data sources, fuel economy is a dominant consideration for prospective
purchasers of automobiles.
It is interesting to note that in much of the survey and focus group
work done prior to the spring of 1979, fuel economy was mentioned as
important; but subsequent to the spring of 1979 the importance of fuel
economy rose dramatically. It is highly likely that this was due to the
cutoff of oil from Iran and the subsequent 60% increase in the cost of
gasoline. It is important to note that based upon the latest survey
data available to EPA (N.Y. Times), 61% of potential new car purchasers
indicated fuel economy as their primary concern in selecting a vehicle
for purchase.
There were three areas identified by the focus groups and studies which
showed a need for improvements of the Federal Fuel Economy Information
Program. These were as follows:
a. Awareness - Time and time again, a majority of respondents
indicated that they were not aware of the Gas Mileage Guide. A
significant percentage were not even aware of the label, although
all of the people surveyed were new vehicles buyers and had
presumably visited dealer showrooms.
Since awareness plays such an important role in any information
program, it is important that steps be taken to increase the
public's awareness of the Federal Fuel Economy Information
Program. Some of the actions which EPA could consider are
as follows:
(1) Increase the use of media advertising, including television
and radio along with the printed medium (e.g. consumer and
automotive trade publications).
(2) Use strategically placed signs and billboards in high density
locations where a large majority of drivers will be exposed to
the Federal Fuel Economy Information Program during the course
of their travels.
(3) Work closely with the FTC on fuel economy advertising, including
encouragement that the FTC require that all manufacturer ads
(in all media) which mention fuel economy indicate availability
of the Guide.
-------
-11-
b. Availability - In addition to making the public more aware of
the Program, the information output of the Program should be
readily available to the public. Some of the actions which could
be considered to improve availability are:
(1) Use facilities in high density locations such as service
stations, banks, post offices, etc. to distribute the Guide,
in addition to new car dealers.
(2) Establish a toll free telephone exchange to disseminate and
clarify fuel economy information.
(3) Insert pamphlets concerning fuel economy within popular publications
as a "free item" with the magazine.
c. Credibility - Having people aware of fuel economy and having
access to such information is not very useful if the information
itself is not credible. Therefore to enhance the credibility of
the Federal Fuel Economy Information Program the following actions
might be considered:
(1) Improve vehicle labeling accuracy (such as would be achieved
by going to vehicle-specific fuel economy labels).
(2) Implement a road MPG factor program for fuel economy labels.
Details of the above two programs can be found in subsequent sections
of this report.
Implementation of program improvements such as those above, in cooperation
with DOE, DOT, FTC, IRS, and GPO, could go a long way toward improving
the Fuel Economy Information Program to a point where the motoring
public will have access to, and confidence in, information of importance
to them in making new car purchase decisions.
-------
-12-
B. Design Parameter Sensitivity
1. Introduction
The EPA currently calculates a general "model type" label for vehicle
fuel economy. This general label value is usually based upon the average
of several tested vehicle designs. Thus it occurs that vehicles having
a variety of different design characteristics (e.g. weight, axle ratio,
road load horsepower) all receive the same label fuel economy value.
For those designs which are tested, the label value may not (and frequently
does not) correspond to the test MPG values; for those designs which
are not tested, design differences from the tested cars are not accounted
for. To provide a method for resolving such potential discrepancies,
vehicle design parameter sensitivities have been determined which allow
adjustment of fuel economy test data for specific vehicle design parameter
variations.
To quantify the relations between vehicle fuel economy and specific
vehicle design parameters, the fuel economy impacts of differences in
weight, road load horsepower, and axle ratio have been determined.
(Although more technically appropriate than axle ratio, the parameter
N/V is not addressed in this section because it is not as readily available
for all vehicles for fuel economy adjustment. Sensitivity regression
equations for N/V are given in Appendix F.)
2. Data Sources
Sources for the data for the vehicle design parameter sensitivity analysis
are as follows:
a. Weight and axle ratio sensitivity; Model Year 1975-1980
General Label files;
b. Road load sensitivity; Based on unpublished EPA data, and on
Ref. 9.
3. Analysis
Design adjustment factors were obtained for weight (W), road load horsepower
(RLHP), and axle ratio (AR). The methodology used to obtain the adjustment
factors is described below.
a. Within discrete basic engines, test vehicles identical in
all configuration specifications except the design parameter in
question were paired and their fuel economy and design parameter
values used to determine sensitivity coefficients for city and
highway fuel economy. Approximately 2000 vehicles were involved in
the analysis. A sample calculation follows:
W-L = 3000 Ib.
FEj_ =19.6 MPG
W2 = 3500 Ib.
FE =19.3 MPG
-------
-13-
Sensitivity, S, is calculated for city FE using the following
equation:
AFE/FE
S =
AW/W
(FE1-FE2)/(FE1 + FE2)
(W1-w2)/(w1 + w2)
Using the values given earlier:
(19.6-19.3)7(19.6 + 19.3)
2
(3000-3500)/(3000 + 3500)
2
S- -.1002
Based on this coefficient, a 10% weight increase would result in a
1.002% fuel economy decrease.
b. The W, AR, and RLHP sensitivity factors were then regressed
against W, AR, and RLHP, respectively for city and highway values,
resulting in two regression equations for each of the three design
parameters.. See Figure 1.
4. Results
a. Typical sensitivity factors determined from the above analyses
are as follows:
City Highway
Sensitivity to AWeight -0.37 -0.31
(at 3000 Ib.)
Sensitivity to ARLHP -0.17 -0.35
(at 10 HP)
Sensitivity to AAxle Ratio
without overdrive -0.30 -0.55
(at AR = 2.0)
with overdrive -0.30 -0.55
(at AR = 2.31)
-------
-14-
Figure 1
Design Sensitivity Parameter
Regression Equations
e
CU
•H
o
•H
m
14-1
0)
o
o
MPG Sensitivity to Axle Ratio
(without overdrive)
•H
O
-.70
-.60
-.50
-.40
-.30
-.20
-.10
Std. Error = .483
Corr. = .468
c
0)
•rl
u
•rl
4-1
4-1
01
O
O
•H
•rl
4J
•H
CO
g
4-1
•rl
O
-.70
-.60
-.50
-.40
-.30
-.20
-.10
MPG Sensitivity to Axle Ratio
(with overdrive)
Std. Error = .483
Corr. = .468
2.5 3.0 3.5 4.0
Axle Ratio
2.5 3.0 3.5 4.0
Axle Ratio
c
01
H
a
0)
o
u
-1.0
-.80
•H -.60
co
a
0)
to
cfl -
00
H
SO
.40
.20
Std. Error = .506
Corr. = .402
2.5 3.0 3.5 4.0
Axle Ratio
a
QJ
•H
o
•rl
o
U
4J
•rl
•rl
CO
g
C/3
-.80
-.60
-.40
R)
a -.20
rG
00
•H
Std. Error = .506
Corr. = .402
2.5 3.0 3.5 4.0
Axle Ratio
-------
-15-
c
0)
•H
U
•H
4-1
4-1
CU
O
MPG Sensitivity to Vehicle Weight
Std. Error = .274
Corr. = .268
•H
w
c
0)
CO
•H
U
.40
,35
.30
,25
-.20
2000 2500 3000 3500 4000
Inertia Weight (Ibs)
MPG Sensitivity to Road Load
Horsepower
c
cu
•ri
O
•H
4-4
LW
CU
O
CJ
4-1
•H
•r)
CO
C
0)
4-1
•H
U
.18
.16
.14
.12
Std. Error = .198
Corr. = .063
10 12 14
RLHP
16
C
0)
•H
O
(U
O
O
-.40 -
•H
4->
•H
CO
C
0)
en
cd
S
fi
bO
•H
W
Std. Error = .305
Corr. = .290
2000 2500 3000 3500 4000
Inertia Weight (Ibs)
4-1
CU
•H
O
H
M-l
4-1
01
O
>^
4J
•H
•H
CO
cfl
I
60
•H
.375
,350
,325
.300
.275
Std. Error = .lob
Corr. = .117
8 10 12 14 16
RLHP
-------
-16-
Note - When using the parameter regression equations given below, the
sensitivity coefficients obtained are only valid for values of the
design parameters which yield negative coefficients. A calculated
sensitivity coefficient greater than zero should be set equal to zero.
b. Effect of AWeight
Regression Equation: s|f = -0.657 + 9.542(10~5) W + 3.512(10~10) W2
Examples: SW (W = 2000) = -0.465
SW. (W = 4000) = -0.270
city
Regression Equation: SJJ = -0.626 + 1.024(10~4) W + 8.174(10~10) W2
Examples: sJJ (W = 2000) = - -0.418
s£ (W = 4000) = -0.203
c. Effect of AAxle Ratio
AR __
Regression Equations: S = 1.025 - 0.437 (AR) No Overdrive
AR -
Scity = l'028 ~ °'376 (AR) with ^^drive
AR
Examples: Sci(. No O.D. (AR =3.0) = -.286
AR —
Scity °'D- (A* = 3'0) = ~'100
AR
Scity N° °-D> (A^ = 4'0) = ~'723
AR
Scity °>D- (~^ = 4'0) = ~'476
AR — —
Regression Equations: Sh = 0.578 - 0.380 (AR) No O.D.
AD __
Shwy = °'580 ~ °'327 (AR) with °'D-
AR
Examples: S No O.D. (AR =3.0) = -.562
AR —
Shwy O.D. (AR =3.0) = -.401
SAR
hwy No O.D. (AR = 4.0) = -.942
AT> __
Shwy °'D- (AR = 4'°> = --728
-------
-17-
d. Effect of ARLHP
Regression Equation: S . = -0.247 + 0.756(10 ) RLHP
DT Up
Examples: S^ (RLHP = 8) = -0.186
qRLHP
city (RLHP = 16) = -0.126
DT UP _9 .--r-.--.r-
Regression Equation: Sr~ = -0.483 + 1.325(10 ) RLHP
nwy
•nr up
Examples: S^ (RLHP =8) = -0.377
..RLHP
S* (RLHP = 16) = -0.271
hwy
The mere existence of parameter sensitivity algorithms does not in
itself assure that their use would have any practical benefit. Use of
such algorithms is worth considering only if their interaction with
design parameter differences encountered in actual practice is "strong
enough" to change fuel economy label values bv one whole number or more.
A review of fifteen sales-significant 1980 engine families showed
that axle raio difference of 0.5, weight differences of three ETW
classes (375 Ib), and road load horsepower differences of 1.5 Hp are
quite common within the same configuration. Using these typical design
differences the algorithms show that:
an axle ratio difference of 0.5 in the neighborhood of
AR=3.5 will cause MPG differences exceeding 1 MPG for all
label values >_ 24 city and/or >^ 12 highway;
a weight difference of 375 Ibs. in the neighborhood of 3000
Ibs. will cause MPG differences exceeding 1 MPG for all label
values >_ 22 city and/or ^_ 26 highway;
a road load horsepower difference of 1.5 Hp in the neighborhood
of 8 Hp will cause MPG differences exceeding 1 MPG for all
label values >_ 29 city and/or >^ 14 highway.
5. Conclusion
Certain vehicle designs vary sufficiently from those of the nearest
tested vehicle designs to cause errors in fuel economy label values of 1
MPG or greater; application of weight, axle, and road load sensitivity
algorithms should be used to reduce vehicle mislabeling due to unaccounted-
for design variances for untested vehicle subconfigurations. Section D
illustrates the improvements in labeling that would result from the use
of design sensitivity factors. In situations where a manufacturer can
show that the above universal algorithms are not reflective of his own
vehicles' behavior, EPA could consider the use of alternate, manufacturer-
specific sensitivity algorithms.
-------
-18-
C. Road MPG Factors
1. Introduction
A number of "real world" influences which can cause differences
between the EPA MPG figures and in-use experience have been identified
and quantified by the 404 Report. Because of uncertainties as to how
these influences act in combination, however, it is not practicable at
this time to devise EPA-to-road adjustment factors on an analytical
basis. Fortunately, in-use data exists in sufficient quantity to permit
empirical estimates of road adjustment factors.
2. Data Sources
Although numerous bodies of in-use MPG data are available from fleet
operators' records, it appears, as pointed out in the 404 Report, that
such non-consumer data may not be fully appropriate for developing road
adjustment factors for fuel economy information intended primarily for
consumer use. Hence, this report's analysis concentrates on three
groups of available data representing consumer driving.
a. General Motors 1975 Customer Survey; This data base
includes in-use MPG for some 1500 privately-owned new 1975 GM
models, based on postcard logs of odometer mileage and fuel quantity
for three successive fuel purchases, with (at least) the first and
last purchases being fill-ups. This was a nationwide random sample,
and covered 47 nameplates from all five GM divisions, operated
under both summer and winter driving conditions.
b. Ford 1978 and 1979 Lease Car Survey; This is also a
new-car postcard type survey, covering some 20,000 Ford Motor Co.
vehicles split about equally between the two model years covered.
Winter and summer driving conditions are both included. Strictly
speaking, these cars were not privately owned, being leased to Ford
Motor Co. employees; they were, however, used in personal/commuter
transportation rather than fleet service. A valuable parameter
logged in this survey was the estimated fraction of urban driving
for each car during its survey interval.
c. J.D. Power Survey; This DOE-sponsored survey was a questionnaire
survey which gathered in-use MPG estimates (along with a large
amount of other information related to consumer opinions on the
usefulness of the label/guide program and the importance of fuel
economy in vehicle purchase decisions) from about 10,000 purchasers
of new 1978 and 1979 vehicles. The survey was a nationwide random
sample, and covered vehicles from most domestic and import manufacturers,
As did the Ford survey, this survey acquired data on urban driving
fraction. While the J.D. Power data base has the advantage of
multiple manufacturer representation, it has two disadvantages
which may be of some significance:
(1) The in-use fuel economy economy values are single MPG estimates,
rather than figures calculated from odometer readings and fuel
qantities. In many cases, the respondents indicated the figures
-------
-19-
were not based on record-keeping at all, but were literally only
perceptions of their vehicles' in-use MPG performance. It has been
observed that owner-perceived in-use MPG generally is overly
optimistic (possibly due to measuring MPG only When convenient, as
in a long trip), so the influence of this type of data could result
in underestimation of the in-use shortfall.
(2) The questionaires used for this survey did not include blanks
for specifying the displacement of the engine in the respondent's
vehicle; only the number of cylinders. Consequently, the EPA MPG
values assigned to many vehicles were the average of all label/guide
values for all available engines of a given number of cylinders.
While this could introduce dispersion in the data moreso than
affecting mean values, it is possible that some directional error
(bias) could exist in the resulting shortfall values.
3. Analysis
Early in the course of analyzing these available data for road adjustment
factors, it became clear that there is not just one labeling strategy to
be addressed. There are a number of plausible interpretations of the
data, uses of the various EPA values, and options for presentation of
label values, depending upon what information the label figure(s) is/are
intended to convey. Accordingly, this study developed unique road
adjustment factors for each of six labeling options, defined as follows:
Option 1, "Two-mode Median" - two MPG values are given, representing
the median values of expected in-use fuel economy in "urban" driving
and in "non-urban" driving, respectively. The values are derived
from the EPA City and Highway figures.
Option 2, "Range, 90% Inclusion" - two MPG values derived from EPA
City and Highway figures are used, but they do not relate to specific
driving modes; instead, they reflect, respectively, a low MPG
figure which 95% of drivers exceed, and a high MPG figure which
only 5% of drivers exceed. The two figures thus bound an MPG range
within which 90% of all driving experience falls.
Option 2a, "Range, 60% Inclusion" - this is a variant of the 90%
inclusion range approach. Again, two values based on the EPA City
and Highway figures are used, but the inclusion range is 60%; 20%
of drivers achieve MPG below the lower figure, and 20% exceed the
higher figure.
Option 3, "Two-mode, Ten Percentile Estimate" - this is a conservative
version of Option 1's two-specific-mode strategy, except instead of
reflecting average respective City and Highway experience, the
values presented are such that 90% of urban driving MPG will exceed
the lower figure and 90% of non-urban driving will exceed the
higher figure; the two figures, then, are "9-out-of-10 guarantees"
of MPG performance in each of the two driving modes.
-------
-20-
Option 4, "One-mode Median, City Basis" - this is a one-number
system based on the EPA City figure only. In effect, it assumes
that in-use average MPG for all driving is linearly proportional to
the EPA City figure for all vehicles, whether their Highway-to-City
,MPG relationships are equivalent or widely different.
Option 4a, "One-mode Median, 55/45 Basis" - this one-number system
uses the EPA combined City-Highway value as its basis.
The road factors for each of these options were determined separately
for each of the three data sources. The sample sizes in these data
bases were large enough so that the various percentile cutpoints needed
to determine the road factors did not have to be calculated from statistical
formulae — they were read directly from histograms of ratios of road
MPG to EPA MPG. For options 1 and 3, the ratios histographed were Road
Urban/EPA City and Road Non-urban/EPA Highway. "Urban" and "Non-urban"
vehicle subsets of the Ford and J.D. Power data were defined to include
those vehicles with urban fractions of 90% or greater, and 20% or less,
respectively. For illustration, the characteristics of these two subsets
from the Ford data are shown below. This illustration shows that the
shortfall from either of the EPA values depends strongly on the type of
driving involved.
TABLE I
Characteristics of Urban and Non-Urban Driven Cars
— 1978-79 Ford In-Use Data —
Urban Non-Urban
Driven Cars Driven Cars
Definition ^_ 90% Urban <_ 20% Urban
Number of Cars 4376 4868
Average Miles/Day 27.0 91.5
Average Odometer 3766 6562
Average EPA City MPG 16.28 16.21
Average EPA Hwy. MPG 22.83 22.94
Average Road MPG (actual) 14.02 17.62
Average Road MPG (4000 mi) 14.14 17.55
Average Weight, Ib 3864 3908
Average CID 281 287
Reference 10 gives specific values for alternate definitions of
urban and non-urban cars, and indicates no significant sensitivity
between the subset definitions and their shortfall characteristics.
-------
-21-
For the GM data, which did not include urban fraction, the two subsets
had to be defined using average miles per day (AMPD) as a surrogate
determinant of urban fraction. Those low AMPD records which yielded an
aggregate average AMPD of 27.0 were defined as the urban subset, and those
high-AMPD samples which averaged 91.5 were defined as the non-urban subset,
For options 2 and 2a, the ratios histographed were Road/EPA city
and Road/EPA Highway, the Road values in both cases being those of
all of the data, not just modal subsets.
For option 4, the analysis was based on the Road(all)/EPA City
ratio and for option 4a the Road(all)/EPA combined ratio.
The table below gives the road adjustment factors for the three
separate data sources. For this analysis, which is only an illustration
of the use of road adjustment factors, the average values (in the last
column) were used.
Table II
Summary Table of Adjustment Factors
J.D.Power Ford GM Average
Option 1978 - 79* 1978 - 79 1975 All 3
1. Two-mode Median, City .88 .86 .96 .90
Highway .82 .75 .79 .79
2. Range, 90% inclusion, City .69 .70 .73 .71
Highway 1.03 .91 .91 .95
2a. Range, 60% inclusion, City .85 .82 .87 .84
Highway .84 .82 .83 .83
3. Two-mode, 10%-tile estimate, City .65 .69 .77 .70
Highway .65 .62 .66 .64
4. One-mode Median (city basis) 1.00 .94 1.00 .98
4a. One-mode Median (55/45 basis) .88 .81 .89 .86
* The factors from the J.D. Power survey
were derived from consumer perceived
fuel economy, not actual.
-------
-22-
Figure 2 illustrates the distribution cuts and average road factors
from table II.
Each of the six labeling strategies has its own combination of
advantages and disadvantages. These are given in the pro/con
analysis following figure 2.
-------
-23-
OPTION 1
Two-mode
Median
Figure 2
Options for Use of
Road Adjustment Factors
50%
0.90 X City
0.79TC Hwy.
OPTION 2
Range 90%
Inclusive
0.71 X City
0.95 X Hwy.
OPTION 2a
Range 60%
Inclusive
0.84 X
60%
.20%
City 0.83 X Hwy.
OPTION 3
Two-mode
10 - Percentile
Estimate
.90% 10%
90%
0.70 X City
0.64 X Hwy.
OPTION 4
One-mode
Median
(City Basis)
50%
50%
0.98 X City
OPTION' 4a
One-mode
Median
(55/45 Basis)
0.86 X Ccmbined (5b/45)
-------
-24-
Pro/con analysis of labeling options
Option 1; Two-Mode Median
Advantages
-Bimodal driving patterns
perceived by the public
-Two numbers more useful for
comparison shopping, depending
upon intended vehicle use.
-Consistent with current
Q 11
advertising practices '
-Half of the population will
get better than the MPG number
presented for each mode
-Consumer studies indicate that
EPA values are perceived as
Disadvantages
-Half the population will get
less than the MPG number
presented for each mode
-The adjusted highway number in
some rare instances may be
to, or even less than the city
number
6811
discrete rather than a range ' '
-Highway shortfall will be reduced
substantially
Option 2: Range, 90% Inclusive
Advantage
-90% of all drivers will get within
specified fuel economy range.
-Range of fuel economy values favored
by some consumers because "it was an
indicator not an absolute prediction"
of fuel economy they should expect
. 4,8
to receive
Disadvantages
-Consumers seem to perceive EPA
values as discrete rather than
6,8,11
a range
-Range may not be as convenient
for comparison shopping for city
or highway specific drivers
-90% range could be perceived as
overinclusive, opening EPA to
more criticism(i.e., "why don't
they use one label saying all
vehicles will get between one and
100 MPG").
-------
-25-
Option 2a; Range, 60% Inclusive
Advantages
-60% of the population will achieve
fuel economy within the range
-The 60% range is not overly inclusive
-Range is a good indication of fuel
economy rather than an "absolute
indication
.,,4,8
Disadvantages
-Consumers seem to perceive EPA
values as discrete rather than
6,8,11
a range
-Range may not be as convenient for
comparison shopping for city or
highway specific drivers
-40% of the population will not
achieve within the specified range
Option 3; Two Mode 10%-tile Estimate
Advantages
-Bimodal driving patterns perceived
by the public
-Two numbers useful for comparison
• 11
shopping
-Consistent with current advertising
8,11
practices
-90% of population will achieve the
MPG number or better for both city
and highway modes
Disadvantages
-Guaranteeing 90% of the population
will achieve better than the MPG
numbers may be perceived as intentional
underestimation of MPG numbers to
minimize complaint letters rather
than convey meaningful information
on achievable fuel economy.
-The two MPG values resulting from
this scheme are numerically very close;
those who insist on perceiving the
numbers as a range will interpret it
as a virtual guarantee of MPG to be
achieved.
-Since MPG distributions cannot be
discerned from one tail, these figures
cannot be used for any meaningful fuel
demand forecasting or annual fuel cost
computation.
-------
-26-
Option 4; One Mode Median (City Basis)
Advantages
-Compatible with current system
and label format.
-Half of population will get better
than MPG number presented
-One number may be less confusing
to consumers
Disadvantages
-Consumers have indicated overwhelming
preference for a multi-number system,
specifically two numbers.
-One number system not consistent
with current advertising practices.
-Less useful for comparison shopping,
especially for specific highway
drivers (see chart below)
-Small adjustment to city numbers may
appear to be only a token effort by
EPA to revise its FE information system
-Half the population will get lower
value than the MPG number presented
Car
Three Cars, Same City MPG
H/C
Ratio Hwy
1.3 29
1.5 33
1.7 37
55/45
25
26
27
This chart highlights a shortcoming of using a city-only esimate.
Three vehicles with the same City number, but with varying Highway-to-City
Ratios have different values for Highway and Combined estimates. All
car purchasers who do any significant non-city driving will be led to
believe that all these cars are equivalent in MPG, which they are not
in terms of those purchasers' needs.
-------
-27-
Option 4a; One Mode Median (55/45 liasis)
Advantages
-Compatible with current label
format used
-Half of population will get better
than MPG number presented
-Compatible with CAFE compliance
and gas guzzler programs:
if the fudge factor is known,
the raw EPA 55/45 number can be
determined for each car.
-One number may be less confusing
to consumers
Disadvantages
-The 55/45 split is representative of
the population average, but only
represents a small fraction of
individuals' driving habits
-Consumers indicate a preference for
a two number system
-One number system not consistent with
current advertising practices
-Least useful for comparison shopping
(see chart below)
-Half the population will get worse
than MPG number presented
Three Cars, Same 55/45 MPG
Car
1.3
1.5
1.7
Hwy
30
33
36
55/45
26
26
26
Note differences in City and Highway MPG ranking for vehicles with varying
H/C ratios when a combined one number system is used to represent estimated
MPG. City drivers are led to believe the cars have equivalent MPG, when car
A9 would be the better choice; Highway drivers are also led to believe
equivalence, whereas car C_ should be better for their driving pattern.
-------
-28-
4. Conclusions
Except for the two-mode 10 percentile option, all of the two-number
labeling schemes are special cases in a continuum of interrelated
two-mode descriptors and range descriptors; they are not independent.
The Two-mode Median option has range inclusion properties, and the
range options have two-mode properties, as shown below:
Inclusion City
Option Range Factor
of City Driving
2
2a
90%
60%
(not
shown) 50%
1 40%
0.71
0.84
0.87
0.90
Below
11%
35%
43%
50%
Above
65%
57%
50%
Highway
Factor
0.95
0.83
0.81
0.79
% of Highway Driving
Below Above
91%
63%
56%
50%
9%
37%
44%
50%
Of the one-number systems evaluated, the one based on the EPA combined
55/45 figure (Option 4a) is preferable in that it provides a link to the
EPA MPG rating used in the CAFE compliance and Gas Guzzler programs;
however, it may be inferior to the City-MPG based system (Option 4)
in that it could mislead both city-driving car buyers and highway-
driving car buyers.
Either of the one-number systems would be somewhat "cleaner" than multi-
number systems in the cosmetic sense, offering less clutter in the
label and Guide; however, reducing clutter may not provide all the
information desired by consumers.
For all road labeling options, the road adjustment factors are developed
from data sources that surveyed mainly gasoline-fueled, rear-wheel driven
passenger cars. The resulting factors may or may nol accurately reflect
road shortfalls for vehicles which are Diesel powered and/or front-wheel
driven and/or trucks.
Three-number or four-number labeling schemes could be> developed from
the six options evaluated, e.g., adding option 4a's combined city-highway
based value to the two figures from Option 1, combining Options 1 and 3
to yield 10 percentile and 50 percentile numbers for city and highway
driving, etc.
The ease of calculation of an annual fuel cost (AFC) figure varies among
the options estimated:
-for both of the one-number systems, AFC calculation is completely
straightforward:
-------
-29-
- for the Two-mode Median option, AFC can be calculated from a
55/45 weighted average of the two figures, since the figures do
represent modal driving;
- for both of the Range options, AFC could be calculated from a
55/45 weighted average of the two figures, although the two
figures are not directly associated with driving modes;
- for the Two-mode Ten Percentile option, there jts_ np_ way _t£
calculate AFC, because the location of the means of the two
modal distributions cannot be discerned from only the 10 percentile
cutpoints.
Policy and regulatory considerations beyond the scope of this Technical
Report must figure in the final selection of options to be proposed, and
the system to be adopted, for the fuel economy labeling regulations.
It is important to note that additional analysis of the data sources
used herein, and other data sources which may become available to EPA,
must and will be conducted; the rulemaking process in fact requires
continued analysis through the duration of the public input/comment
period. Some important issues which will be receiving analytic attention
between publication of this report and the shaping of the final rule
are:
- the form of the road adjustment. Instead of the use of constant
multiplicative factors, it may be more accurate and appropriate
to specify sets of constant MPG offsets (e.g. Road MPG = EPA MPG
minus two MPG) or constant GPM offsets (e.g. road GPM = EPA GPM
plus 0.007 GPM), or MPG-dependent versions of these.
- the applicability of unique road adjustment schemes to specific
classes of vehicles. It may well be that separate road adjustment
systems are warranted for certain generic types of vehicles,
such as trucks, Diesel-powered vehicles, front-wheel drive
vehicles, certain combinations of these, and/or other vehicle
types. At this point there is simply not enough data to develop
unique road adjustment schemes down to the individual model type
level, but there is already some preliminary evidence that some
broad classes of vehicles may be deserving of unique adjustments.
- the data sources, particularly those of older model years, must
be further evaluated with respect to their representativeness
and statistical significance. To be useful/representative,
in-use data bases should include, as much as practicable, measured
fuel economy, seasonal representation (summer and winter),
reasonable geographic coverage, some identification of type
of driving involved, e.g. % urban, reasonable coverage of the
product line, and enough individual vehicle data (transmission
type, engine CID, etc.) to facilitate accurate assignment of
EPA MPG values to the vehicle records.
-------
-30-
D. Adjusted Labeling
1. Introduction
Label values are currently calculated for new cars by figuring a sales-
weighted harmonic average of all the tested vehicles in a base level
(unique combination of basic engine, inertia weight and transmission
class) which are then combined for each model type. Since in a base
level many of the configurations receive at least one fuel economy test,
it is possible in some cases to apply a label value that represents the
test results for a given vehicle more specifically than the average base
level value.
Vehicles are tested on a subconfiguration level, which may differ by
test weight or road load horsepower within a configuration, and con-
figurations can vary by axle ratio or engine code within a base level.
If all the subconfigurations were tested, it would be easy to recommend
the use of those specific test results for the labels on those sub-
configurations. Because less than 20% of the technically unique sub-
configurations are tested, some alternative must be sought.
An untested vehicle which differs only slightly from a tested vehicle in
the design parameters of axle ratio, road load horsepower, or test
weight can be labeled with a value derived from the label on the tested
vehicle. A set of fuel economy sensitivity algorithms was created to
account for differences in these design parameters (see section B).
With these algorithms, specific labels can be prepared for 60% of the
new car fleet.
The remaining vehicles differ by at least engine code or transmission
configuration from a tested vehicle. It might be possible, for some of
these, to develop a transmission adjustment or to ignore engine code
differences within an engine family.
2. Analysis
The 1980 light-duty certification fuel economy data base was analyzed,
concentrating on the 60% which is either directly tested or can be
simply adjusted for slight differences in design parameters. A road
factor was also applied to both the City and Highway MFC's, not to
obviate the issues brought up in section C of this report, but to show,
for illustration purposes, a typical road adjustment.
Data were prepared in two steps. The first step was to create a file
containing one entry for each subconfiguration in the certification data
base for 1980. This involved reading 9467 vehicle records ("card 5's"),
collapsing them to 5541 technically unique subconfigurations, and consulting
the results of 2125 manufacturer's tests and 997 EPA tests. The collapsing
was done for records which were identical in all technical specifications,
differing only in car line. Where several tests were applicable to a
given vehicle the results were harmonically averaged, sample weighted.
-------
-31-
The second step was to develop a uniform basis for comparing several
different labeling schemes. These 3 schemes are:
a. Base level labeling - similar to the current system, this
system uses sales-weighted average fuel economy (FE) values
for all the members of a base level.
b. Tested subconfiguration specific labeling- if a subconfig-
uration was tested, the test value is used; if not, the base
level value is used.
c. Sensitivity adj usted labeling - if a subconfiguration was
tested, the test value is used; if not, the FE's of the
tested vehicles with the same engine code, in the same base
level, are adjusted for the differences in design parameters,
and the result is harmonically averaged for this vehicle.
The sensitivity adjustments were made using a FORTRAN subroutine (see
appendix A) based on the following algorithm:
The sensitivity S of FE to perturbations in
design parameter X,
9FE
s ~ "ax"
has been separately determined for test weight, axle ratio,
and road load horsepower. For calculation purposes,
AFE/FE
S =
AX/X
or FE - FEj^ = FE
Where FE is the unknown value and FE- is the tested value.
If there is more than one design parameter adjustment, the AFE's are
added arithmetically. As discussed in the 404 Report, aritmetic addition
of AFE's provides a better estimate of combine MPG effects then does a
product of MPG factors.
The three versions of City FE were multiplied by a road factor of 0.86,
and the Highway FE by 0.75, to form six more FE values, for a total of
12 for each record. Of the 5541 subconfigurations read, 1227 were
rejected because they were untested and projected to have zero sales, and
1781 were rejected because they belonged to untested configurations.
938 base levels were realized for 1980. This analysis addresses only
tested or adjustable subconfigurations and leaves open the question of
how to label untested engine codes.
These procedures were first run on a small data base of 15 high-selling
engine families (524 subconfigurations), and the results agreed with
manual analysis of the 15-family sample.
-------
-32-
3. Results
The twelve MPG values in each subconfiguration record are actually six
pairs of City and Highway values, as described above:
EB = base level label
ET = tested subconfiguration specific label
EA = sensitivity adjusted value
RB = road adjusted base level
RT = road adjusted tested subconfiguration specific label
RA = road adjusted sensitivity adjusted value
The following six histograms show the distributions of differences
between options EB, RB, RT, and RA, using RA as the reference. Because
road adjustment causes a much larger shift than specific test labeling
or sensitivity adjustment, ET and EA were not charted. For Figure 3,
differences were tabulated to the nearest integer, and subconfigurations
belonging to untested configurations were omitted. For tabulations of
actual values, see Appendix B.
Histograms of similar differences using EB and RB as references are
contained in Appendix C, and illustrate how the alternate schemes change
label values relative to these references.
4. Conclusions
The charts in Figure 3 were prepared to contrast four alternative labeling
schemes: base level labeling, road-adjusted base level labeling, road-
adjusted specific-tested subconfiguration labeling, and road-adjusted
specific-tested sensitivity-adjusted subconfiguration labeling. Assuming
the last alternative gives the most accurate label, the charts show
several things:
a. Road adjustment shifts the data by about 2.7 MPG city and 6.5
MPG highway, without affecting its spread very much. Since the FE
vector is multiplied by a road factor, and the results are subtracted,
the resulting spread is due partly to the analysis and not solely
to the data. For this reason the other comparisons are made against
road adjusted labels.
b. Without road adjustment the other improvements have a negligible
effect on reducing shortfall.
c. Road adjustment raises the fraction labeled the same from
about zero to about 75%. Specific test labeling increases this
fraction by 10-15%, and sensitivity adjusting picks up another 10-
15%. No conclusions can be reached about the label accuracy of
subconfigurations whose engine codes were not tested.
d. Without both specific testing and sensitivity adjusting there
are vehicles mislabeled by as much as 4
-------
f-lon/*> ^ .D/fre/iErtfleS DeTu>e&i /jakc/iciJ Onfio/ij
hiju/e ^ J/ime J r
100
w
SO
25
||
.ill,
|00
75
£0
»
... — .llllll..
too
»
so
2ST
-15 -10 -5 O tS f/0 tla -|5 -10 -5 O *5 r|Q t|J
RA-EB cvk KA-E^ Hg^y
^isan'^.?" O.Z^same "*nean~^>-^ O^sawe.
100
J 15"
"1
s 50
•> .»>
J 25
M-
1
1
(00
7S
SO
2J
..1
1.
too
75
SO
25"
0 -15 -10 -5 0 +a t-/0 t|5 -|5 -10 -S O *5 r|0 +|5
4-
J «\>» X"\*J
?{ KA— RD dry KA—1\.ft Hifltiway
^ -/> rtr "7fi C^ «v»
Tnaan f.wi TOO /a same. >iean "0.02. 67% same.
ICO
15
50
25
too
75T
SO
*
too
«
50
2o
-15 -10 -5 0
r/0 +|5
-|5 -10 -S O t5 t-|0 +|5
Ci
ify
same.
same.
-------
-34-
iy. SUMMARY CONCL,USIQNS
1. The Fuel Economy Information Program has a public awareness problem.
Recent surveys show that only slightly more than half of new-car
buyers were aware of the label, and less than 20% were aware that
a Guide existed.
2. There is also an availability problem with the Guide; only half of
the people who were aware of it were able to obtain Guides from
auto dealers.
3. Combining Conclusions 1 and 2, if the objective of the Guide program
is to reach all new-car buyers via auto dealer distribution, the
program's success rate is less than 10%.
4. It is not as though less-than-perfeet awareness of Federal Fuel
economy Information is simply a reflection of a relative unimportance
of fuel economy for consumers. Survey data indicate that vehicle
fuel economy ^s_ an important criterion for vehicle purchase decisions.
It is the EPA fuel economy figures which rank relatively low in the
opinion sense, and which are largely ignored in terms of actual use
of the figures.
5. Compared to other sources of fuel economy information, however, the
EPA label does rate as most popular and important. The Guide ranks
8th in importance as a viable source of fuel economy information.
6. Some of the surveys suggest, and others show with specificity, that
there is a serious credibility problem with the EPA numbers, and
that the 1979 interim revision of the information system had no
effect at all in shrinking this credibility gap. Terms like
"unrealistic", "inaccurate", and "nonrepresentative" continue to be
used in describing the reasons for this credibility problem.
7. For some cars, part of the EPA-to-road fuel economy difference can
be traced to administrative causes, wherein sales-weighted averaging
has resulted in certain vehicle configurations receiving label
values that are several MPG different from what they achieved in
tests.
8. The administrative remedy for the deficiency in conclusion 7 may
not be necessarily burdensome in terms of test load; in fact, the
judicious use of design parameter adjustment could reduce the test
burden in some cases.
-------
-35-
9. Data show that drivers, on a "snapshot" basis, drive in one of two
general modes: either predominantly low average speed stop-and-go
travel or predominantly higher average speed quasi-cruise travel;
moreover, surprisingly few individuals average a 55/45 split between
the two modes on an elapsed time basis. See Appendix E.
10. When asked directly what kind of fuel economy information system
(number of numbers) they prefer, 88% of focus group participants
indicate they favor a multi-number system. However, note that this
sample size was only about 250, and it is not certain how leading
the questions put to the focus groups were.
11. In spite of Conclusions 9 and 10, the close agreement between the
EPA City figure (which simulates idealized stop-and-go urban
driving only) and average in-use MPG (which comes from the pooled
results of all types and conditions of driving) is a numerical
coincidence whose espousal on purely pragmatic grounds has been
once, and could be again, an irresistible temptation.
12. The most beneficial single change that can be made to the labeling
system is the incorporation of road adjustment factors, which (for
the Two-mode Median option) raises the fraction of vehicles labeled
"correctly" from less than 1% to about 75%.
13. Once road factors have been applied, specific labeling of tested
subconfigurations increases the percent labeled accurately by an
additional 10-15%; the incremental benefit of going further and
using design parameter sensitivity adjustments for untested sub-
configurations is another 10-15%.
14. The use of specific labeling without application of road factors is
of no significant benefit to label accuracy.
15. A significant fraction (40% in 1980) of vehicle subconfigurations
have labels of undeterminable accuracy, since their engine codes
are untested. Keeping Conclusion 8 in mind, there is the possibility
of improving test coverage over the engine code spectrum without
significantly increasing overall test burden, by using sensitivity
adjustment to reduce test coverage over the design paramenter
spectrum.
-------
-36-
REFERENCES
(1) U.S. Environmental Protection Agency, "Passenger Car Fuel Economy:
EPA and Road," Report EPA 460/3-80-010, September 1980. ("404
Report").
(2) House Committee on Government Operations, "Automobile Fuel Economy:
EPA's Performance," Report 96-948, May 13, 1980 ("Moffett Report").
(3) Scardino, Birch, and Vitale, "Impact of the FEA/EPA Fuel Economy
Information Program", June, 1976 ("ABT Study").
(4) Porter and Novelli, "Report on Focus Groups, Gas Mileage Information
Program," February 1978 ("EPA Focus Group").
(5) M. Hunt, et_ ad, "Preliminary Report on Study and Evaluation of DOE
Vehicle Energy Efficiency Programs", April 1980 ("J.D. Power Report"),
(6) EPA Analysis of J.D. Power data systematized by Energy and Environ-
mental Analysis, Inc. (unpublished), in addition to the Power
report in Ref. 5.
(7) The New York Times, "Auto Industry Outlook—A National Study of
Consumer and Dealer Views," September 1980.
(8) K. Hellman, "Review of the Available Information on the Subject of
how many Different Fuel Economy Numbers to Present", Memo to C.
Gray, ECTD, August 26, 1980.
(.9) Data from S. Martin and K. Springer, "Influence on Fuel Economy and
Exhaust Emissions of Inertia, Road Load, Driving Cycles, and N/V
Ratio for Three Gasoline Automobiles" Southwest Research Institute
Final Report under Task No. 9, EPA Contract No. 68-03-2196, June
1977.
(10) K. Hellman, "Extremes Analysis of Ford In-Use Data Base", Memo to
R. Maxwell, CD, September 3, 1980.
(11) K. Tuckey, "One-Number Versus Two-Number Issue Presentation,"
Draft Issue Paper, August 15, 1980.
-------
-37-
APPENDICES
-------
-38-
APPENDIX A
FORTRAN Subroutine for Sensitivity Adjustment
SUBROUTINE SF(CARA,CARB,OD)
C TAKES UNTESTED CAR A AND DEVELOPS NEW FE VALUES FROM TESTED CAR B
C CARA AND CARB ARE 5-VECTORS. HERE ARE THEIR INDICES:
C I TEST WEIGHT 2 AXLE RATIO 3 ROADLOAD HP 4 CITY FE 5 HWY FE
C FE'S ENTER AND LEAVE CARA INVERTED(GPM) AND CARB UPRIGHT(MPG).
C THE FE'S IN CARA ARE BEING HARMONICALLY ACCUMULATED IN THE PARENT ROUTINE
C OD TKUE INDICATES THE PRESENCE OF OVERDRIVE IN THE TRANSMISSION
REAL CARA(5),CARB(5)
LOGICAL*! OD
RM(A,B)=0.5*(A+B)
TH(X1,X2,D,F2)=D-(F2*2*S*(X2-X1))/(X1+X2+S*(X2-X1))
C TH IS THE SOLUTION TO THE SENSITIVITY EQUATION FOR DELTA FE
DC=0.0
DH=0.0
IF(CARA(J).Eg.CARB(J)) GO TO 20
C TEST WEIGHT FUDGE
WB=RM(CARA(1),CARB(1))
S=-0.657+9.542E-5*WB+3.512E-10*WB*WB
DC=TH(CARA(1),CARB(1),DC,CARB(4))
S=-0.626+1.024E-4*WB+8.174E-10*WB*WB
DH=TH(CARA(1),CARB(1),DH,CARB(5))
20 IF(CARA(2).EQ.CARB(2)) GO TO 30
C AXLE FUDGE
ARB=RM(CARA(2),CARB(2))
IF(OD) GO TO 25
S=l.025-0.437*ARB
DC=TH(CARA(2),CARB(2),DC,CARB(4))
S=0.578-0.380*ARB
DH=TH(CARA(2),CARB(2),DH,CARB(5))
GO TO 30
25 S=J.028-0.37b*ARB
DC=TH(CARA(2),CARB(2),DC,CARB(4))
S=0.580-0.327*ARB
DH=TH(CARA(2),CARB(2),DH,CARB(5))
30 IF (CARA(3).LQ.CARB(3)) GO TO 40
C ROAD LOAD FUDGE
RLB=RM(CARA(3),CARB(3))
S=-0.247+.00756*RLB
DC=TH(CARA(3),CARB(3),DC,CARB(4))
S=-0.483+0.01325*RLB
DH=TH(CARA(3),CARB(3),DH,CARB(5))
40 CARA(4)=CARA(4)+).0/(CARB(4)+DC)
CAkA(5)=CAKA(5)+1.0/(CARB(5)+DH)
RETURN
END
-------
ERRATUM
EPA/AA/CTAB/FE-81-6
Page 38, Line 11 should be:
TH(X1,X2,D,F2)=D+(P2*2*S*(X1-X2))/(X1+X2+S*(X1-X2))
or its equivalent
TH(X1,X2,D,F2)=D-(F2*2*S*(X2-X1))/(X1+X2-S*(X2-XU)
-------
-39-
APPENDIX B
Data for Histograms la Figure 3 and Appendix C
Percent of subconflguratIons
A MFC
-15
-14
-13
-12
-11
-10
-9
-8
-7
-6
-5
-4
-3
-2
-1
0
41
+2
+3
AMPC
0
41
+2
+3
+4
+5
+6
+7
+8
+9
+10
+11
+12
+13
+14
+15
AMPG
-15
-14
-13
-12
-11
-10
-9
-8
-7*
-6
-5
-4
-3
-2
-1
0
+1
+2
+3
+4
RA-EB
City
0.8
3.3
16.6
35.0
36.0
8.2
0.2
Hwy
0.1
0.0
0.6
1.7
1.6
5.0
8.6
12.1
16.0
19.3
17.8
13.0
3.3
0.9
0.1
EB-RB
City
6.1
41.5
34.0
16.4
1.8
0.2
Uwy
1.8
15.7
19.4
lb.9
16.3
13.3
7.5
6.6
1.2
0.9
0.3
0.0
0.1
RB-EB
City
0.2
1.8
16.4
34.0
41.5
6.1
Uwy
0.1
0.0
0.3
0.9
1.2
6.6
7.5
13.3
16.3
16.9
19.4
15.7
1.8
RA-RB
City
0.1
1.2
11.8
78.5
7.9
0.6
Uwy
0.1
0.5
3.4
13.5
67.1
11.3
3.7
0.5
EB-RT
City
0.2
7.4
38.3
34.3
16.7
2.4
0.7
Uwy
0.1
0.5
3.0
12.8
19.1
18.1
16.5
12.8
7.9
5.8
1.4
1.4
0.4
0.0
0.1
RB-RT
City
0.2
3.9
89.3
5.6
0.8
0.1
Uwy
2.1
5.8
83.3
5.9
2.5
0.3
0.1
RA-RT
City
0.1
0.4
6.2
89.2
3.9
0.3
Uwy
0.5
0.9
7.6
83.7
5.4
1.6
0.5
EB-RA
City
0.2
8.2
36.0
35.0
16.6
3.3
0.8
Hwy
0.1
0.9
3.3
13.0
17.8
19.3
16.0
12.1
8.6
5.0
1.6
1.7
0.6
0.0
0.1
RB-RA
City
0.6
7.9
7«.5
11.8
1.2
0.1
Hwy
0.5
3.7
11.3
67.1
13.5
3.4
0.5
0.1
-------
100
7$
SO
25
1
-15 -10 -5 0
EErRB C.fy
»ean *2.1
100
44
t «"
0
»1 50
*
j 25
»
1
-15 -10 -5 0
* pO — CT Q'i
1 ^ ^
ICO
w
50
25
1
,1
-15 -10 -5 0
EB-*A c;u
'
">nean +2.? (•
•
1
ll.
+5 r/O t|5
too
75
SO
25
Illli,,
-15 -10 -S 0 t5 fiO +15
too
*
so
25
t* t^ •* i^J5 H "?H w^y
V '
0 % same. _ inean "^6. ST wX sawe.
i
li.
+5 1-/0 t|5
100
«
50
25
I
.lllll..
-15 -10 -5 O *5 r/O +|5
too
75
SO
2a
ait— ^"-11^ O7.«me
|
ll.
+5 **/0 4"|ff
|00
75
50
25
.lllllll
-15 -10 -5 0 t5 r/o t|5
too
75
s-o
25
E^j^FSft |-| (ontu^W
vj
). 1 % Same. ">nean "f"£>.5" 0 */J same
-------
ICO
JS
SO
25
1
.ll
1
too
IS
so
2S
•
...Hill.
|OCJ
IS"
so
2a
-IS -10 -5 0 *5 fl
t^B-EB Clfy
>iean "2.7 0*2 same.
-IS -10
•)near)~6.5~ O*Z Same
ICO
o
J
«)
|00
1 1
-IS
-5 0 +5
-IS -10 -S O
>iean 0.03 89%
"Tneano.oZ QS'Xsame
IS
so
ICO
25
lOO
£•0
Zff
-IS -10
1.
">nean
0 *5 Tto
sawe
-IS -10 -S O
>iean 0,02.
same
|00
-------
-42-
APPENDIX D
Equivalence of 55/45 Road Factors
Given: Empirical road ratios, Road Urban/EPA City = .90 B Rj,/Er
Road Non-Urban/EPA Hwy. « .79
Road Overall/EPA Comb. = .86 «
Now,
0' Cb
£„
.55 + .45 .55 + .45
EC V, , VEC
and, R'0 =
.55 + .45 .55 + .45
•90E0 EL,
^FX E
EC C
EC
.55 + .45
.90 .79 EH/EC
so,
90 .79
.55 Ejj/E
C
.611 Eg/Eg f .570
Hence, the ratio of a constructed R' to E , depends on E /E :
VE_ = 1.0 1.2 1.4 1.6 1.8 2.0
K. ,£ ^
0 cb' .847 .852 .856 .859 .862 .865
Therefore, R'^/Ep, ^ RQ^Cb' in worc*s« f°ad combined 55/45 MPG values
constructed from Two-mode Median Road Urban and Road Non-Urban figures
are essentially equal to road combined MPG values constructed directly
by applying the One-mode Median (55/45) road factor to the EPA combined
figure.
-------
-43-
APPENDIX E
Bimodal Nature of Vehicle Travel
Traffic surveys have shown that vehicle travel is not a smooth continuum
of speed. When plotted against average vehicle speed, miles traveled
are distributed in two partially overlapping humps. Figure X shows this
behavior for a nationally-representative VMT mix (60/40) of travel on
"Non-highway" and "Highway" roads. Figure Y illustrates similar patterns
for travel in metropolitan areas. Even in such urban driving, the
existence of some high-average speed travel (on urban freeways?) is
obvious. The data in figures X and Y are essentially instantaneous, or
"snapshot" distributions.
Another bimodal aspect of vehicle travel shows up on a more "elapsed
time" basis (rather than snapshot), as shown in Figure Z. In this
questionnaire survey, 39% of respondents indicated their travel averaged
more than 60% urban, and 55% had travel which was less than 50% urban.
Only 5% of these 10,000 people approximated a 55/45 urban/non-urban
travel split, and the overall average split is 48/52. As a facetious
analogy, the U.S. population may indeed average about 50% female/50%
male, but very few individuals exhibit the characteristics of the pop-
ulation average.
-------
-44-
Figure X
Table 2 • Road Type Statistics
14
12
| 10
I 8
| 6
4
2
0
75 175 275 37 S 475 575 67 5 77 6
Speed MPH
FIGURE 39
PERCENT OF MILES SPENT IN SPEED BANDS
Road Type
(Highway)
Expressway
Expressway-Business Route
Rural Highway
(Non-Highway)
Suburban Artery
Un paved Rural
Urban Artery
Strip Commercialism
Suburban - No Curb
Suburban - Curb
Unpaved - Suburban
CBD • No Parking
Urban
CBD -Parking
Shopping Center-Parking Lot
Avg. Speed
mph
53.1
467 .
447
286
245
230
222
21.5
190
177
172
17.2
12.4
115
Stops Per Mile
.03
07
10
77
55
1 29
1 41
.80
1 16
37
255
1 77
380
4.35
% Usage
19
14
8
16
•
25
7
1
-
1
4
3
The predominant influencing factor m determining the dis-
tribution of speeds was the fifth category - Road Type This
point is obvious, but what was not so obvious was the over-
whelming influence that road type had Road type so predom-
inantly influenced driving patterns, that other factors such as
vehicle type, traffic density, population density, and even indi-
vidual differences between people, were minor in comparison.
(These other factors, however, can potentially influence the
road type selected by the driver) The road type categories
can be broken down into two modes of driving as is illustrat-
ed by the dashed lines in Figure 39 These two modes of driv-
ing can be separated by lumping the road type categories into
two groups Highway and Non-Highway
Table 2 illustrates the differences between the modes in
terms of both average speed and stops-per-mile Figures 40
and 41 illustrate the speed distributions of the separated
modes Table 3 lists some driving parameters of the two sepa-
rated modes, as well as driving characteristics of the overall
survey It can be seen, for instance, that the majority of idle
time (time spent at speeds less than 2 5 mph) falls into the
non-highway mode, as would be expected Figures 42 and 43
illustrate the distributions of idle times and distance between
stops for the two modes.
Source: T. Johnson, £t_ al, "Measurement of Motor Vehicle Operation
Pertinent to Fuel Economy", SAE Paper 750003, February 1975.
-------
-45-
Figure Y
Figure 1 shows the dis-
tributions of distances covered with the indi-
cated microtrip speed for the Scott and GMPG
data sets. There are no substantial differences
between the two data sets. Both are character-
ized by a prominent peak at about AO km/h, pre-
sumably corresponding to driving on local
streets, and a smaller peak at about 75 km/h,
presumably reflecting freeway and highway dri-
ving. The valley between the two peaks occurs at
about the same speed as the upper limit for the
applicability of the linear fuel versus trip time
relation, found previously to be about 55 km/h
[8], Urban traffic is thus rather naturally
classified into a low-speed regime where fuel
consumption increases linearly with trip time
per unit distance and a high speed regime where
fuel consumption increases with speed due mostly
to increasing aerodynamic drag. The large frac-
tion of urban travel falling within this low
speed regime, 68% for the Scott data, brings out
the potential fuel savings that might be achieved
through improvements in urban traffic flow.
15-1
10-
PERCENT OF
DISTANCE
IN 5 km/h
SPEED
INTERVALS
5-
G H PROVING GROUND
SCOTT RESEARCH
40 60
SPEED, km/h
80
100
Fig. 1 - Percent of distance driven at indicated
microtrip average speed for the two data sets
for all cities studied
Microtrip speed distributions are shown in
Figure 2 for New York and Los Angeles, the cities
with the lowest and highest average speeds, re-
spectively. Both distributions have a qualita-
tively similar bimodal character, but the two
peaks in New York are shifted towards lower
speeds compared with Los Angeles. This suggests
that the lower average speeds in New York are
not due to a different mix of highway and local
driving but rather to slower speeds in each of
these two categories individually.
is-,
10-
PERCENr OF
DISTANCE
IN 5 km/h
SPEED
INTERVALS
5-
NEH »ORK
LOS ANGELES
20
40
60
SPEED, km/h
~I
100
Fig. 2 - Percent of distance driven at indicated
microtrip average speed from the Scott Research
Data for New York and Los Angeles
Source:
P. Wasielewski, et_ al, "Automobile Braking Energy, Acceleration
and Speed in City Traffic", SAE Paper 800795, June 1980.
-------
-46-
Figure Z
Percent of Drivers who
Do the Indicated %
Urban Driving
20% r-
-------
-47-
APPEHDIX F
N/V Sensitivity Regression Analysis
N/V
Regression Equation : S £'. v = 0.803 - 0.0274 N/V
Examples : S (N/V = 30) = -0.019
S ®*.1 (N/V = 60) = -0.841
Regression Equation : S = 0.385 - 0.0238 N/V
Examples : S (N/V = 30) = -0.329
nwy
(N/V = 60) = -1.043
nwy
------- |