EPA-AA-ECTD-78-1
Evaluation of the Representativeness
of EPA Fuel Economy Estimates
Thomas Cackette
January 1978
Emission Control Technology Division
Office of Mobile Source Air Pollution Control
Environmental Protection Agency
-------
I INTRODUCTION
Energy conservation has become a major national commitment. An im-
portant element of this commitment is a national strategy to improve
the fuel economy of the new car fleet through a regulatory program in-
volving stringent fuel economy standards that each manufacturer must
meet for its corporate sales-weighted fleet. The basis for this re-
gulatory program is contained in the Energy Policy and Conservation
Act. To insure consumer cognizance of the fuel economy implications
of the consumer's purchase decision, the Act established a fuel
economy labeling program whereby each new vehicle is posted with its
expected fuel economy. The results of the fuel economy labeling
program are also published as an annual Mileage Guide.
The keystone to both of these programs is the Federal Test Procedure
developed by EPA to quantify and regulate emissions reductions for
environmental purposes. These same procedures are also utilized as
the yardstick of fuel economy measurement. The fuel economy of cars
is measured on pre-production prototype vehicles on a test apparatus
that simulates on-road operation over a driving cycle that has been
determined to yield representative emissions for urban driving conditions.
Recent suggestions that EPA fuel economy estimates significantly over-
predict actual vehicle fuel economy prompted EPA to conduct a number of
studies designed to explore the representativeness of the EPA meth-
odology. The objective of this report is to summarize the results
and conclusions of these studies.
The key issues addressed by this report are:
o Representativeness of the Mileage Guide fuel economy
value obtained from EPA certification prototype vehicles
compared to the owner's reported fuel economy for the
comparable vehicle.
o Sensitivity of the EPA measurements to test procedure
parameters, and to limitations imposed by the laboratory
environment.
o Improvements that can be made in the EPA test procedures.
Historical Perspective - The EPA first published mileage information for
model year 1973, at the direction of the President. Beginning with model
year 1974, manufacturers were encouraged by EPA to label new cars with
fuel economy information. The label data were also compiled in a Mileage
Guide published by EPA. The mileage information was intended to provide
potential new car buyers with information on the relative fuel economy
performance of new models so that they could more effectively consider
fuel economy in purchasing new vehicles. EPA cautioned vehicle owners
on the label and in the Mileage Guide against expecting to achieve the
absolute mpg estimate indicated for each vehicle in view of the wide
range of driving conditions in the country that can affect vehicle fuel
economy.
-------
With the passage of the Energy Policy and Conservation Act (EPCA) in
December 1975, fuel economy labeling and publication of the Mileage
Guide became required by law, effective in mid-model year 1976. EPCA
also established average fuel economy standards that must be met by
each manufacturer on a fleetwide, sales weighted basis, and required
EPA to conduct a fuel economy testing program. EPCA requires the
fuel economy testing to be performed, to the maximum extent possible,
in conjunction with EPA's emission certification testing. The fuel
economy test procedure is required by EPCA to be comparable to that
used in the voluntary program in model year 1975. While there is no
express EPCA requirement that fuel economy testing for developing con-
sumer mileage information and for compliance with EPCA standards be con-
ducted in the same way, doing so results in large savings in terms of
costs to the taxpayer, as well as in savings to the industry which are
reflected in the cost of new cars.
The Fuel Economy Test - EPA fuel economy values are obtained by driving
the vehicle, on a chassis dynamometer, over two prescribed driving
cycles. Fuel economy results are calculated from the measured con-
centrations of tailpipe emissions of carbon-bearing exhaust consti-
tuents (hydrocarbons (HC), carbon monoxide (CO), and carbon dioxide (CC^)
and a knowledge of certain characteristics of the fuel (fuel density and
fuel hydrogen to carbon atomic ratio). A direct measurement of the
mass of fuel used is not made in the EPA test.
A chassis dynamometer, a treadmill-like device, simulates the forces the
vehicle would actually experience on the road. Although the vehicle
remains still, the vehicle's engine turns its drive shaft and rear
wheels (front wheels on front wheel drive vehicles), which in turn drive
the dynamometer rolls. Attached to the rolls are steel disks which
simulate the vehicle's inertia (resistance to acceleration) and a pump
which simulates the aerodynamic drag that normally acts on a moving
vehicle. Thus the vehicle's engine must provide power to overcome
driveline and tire frictions (as it would on the road) in addition to
the forces provided by the dynamometer.—' A schematic of the dynamometer
is shown in Figure 1.
I/ An assumption in the design of the test is that the power dissipated by
the two rear tires driven on the dynamometer simulates the power dissi-
pated by all four tires driven on the flat road. EPA is reevaluating
this assumption for various tire types in a current test program.
-------
CHASSIS DYNAMOMETER ELEMENTS
V*"EHICLE DRIVING
WHEEL(S)
X.
CHASSIS ROLLS
INHRTIAL DISCS
(simulate vehicle mass)
POWER ABSORPTION UNIT (pump)
(simulates aerodynamic drag)
Figure 1 - Chassis Dynamometer
The city driving cycle consists of a 7.5 mile trip with an average speed
of 20 miles per hour, 4.3 minutes (18%) of idling, 18 stops, and a short
freeway section with a top speed of 56 mph. The test vehicle, after
sitting at an ambient temperature of 68 to 86° Fahrenheit for at least 12
hours, is driven over the city cycle on the dynamometer. The engine is
then stopped for ten minutes and the 7.5 mile test is repeated.— The
results of each part of the test, referred to respectively as the cold
and hot start tests, are weighted by the national average of hot and
cold start trips (53 percent cold start and 47 percent hot start) to
yield an overall city fuel economy estimate.
After completion of the city test, the vehicle, in a warmed-up condition,
is driven on the dynamometer over the highway cycle. This 10.2 mile
trip has an average speed of 48 mph, with no stops between the beginning
and end of the test. Speed variations during the trip are small, with
most of the trip occurring between 40 and 60 mph. The speed versus time
history of each trip is shown in Figure 2.
2J In practice, the second portion of the hot start test following the
first 505 second point is not performed. Data indicate that this portion
of the cold start test, which is measured separately, also represents
warmed-up vehicle operation, and repeating this portion of the test would
yield the same results.
-------
City Driving Cycle (LA4)
1300.00 IWO.OO
Highway Driving
Cycle
100.00 200.00
500.00 600.00 700.00 800.00
Time (seconds)
Figure 2 - Dynamometer Driving Schedules
In the Mileage Guide city, highway, and combined fuel economy estimates
are presented for each model type. The combined value provides an
estimate of the vehicle's overall .average fuel economy. It is calculated
by harmonically weighting the city value by 0.55 and the highway value
by 0.45, the two factors being the nationwide fraction of city and
highway trips.—'
Annually EPA tests pre-production vehicles to confirm compliance with
emission standards. The test performed on these cars is identical to
the procedure used to measure city fuel economy. Thus a fuel economy
result is obtained as a by-product of the emission test for these cars.
To fully represent all model lines, additional vehicles (about 30 percent
of the total in 1977) are tested specifically for fuel economy. All
vehicles are pre-production prototypes which have accumulated 4000 miles
prior to testing. In 1977, approximately 750 vehicles were tested for
fuel economy at EPA's laboratory.
II ANALYSIS
At issue is how well does the EPA Mileage Guide fuel economy compare
to on-road fuel economy reported by owners. The data available allow
this issue to be addressed in two parts. By comparing the Mileage
Guide fuel economy to dynamometer fuel economy measurements on in-use
cars, the effect of differing test conditions is removed. Any fuel
economy differences thus identified will be the result of prototype
to production vehicle differences, or of differences in maintenance,
3/ More recent data indicate 58% city and 42% highway weighting factors
would be more appropriate.
-------
method of mileage accumulation or state-of-repair of the vehicle. These
same dynamometer tests can also be compared to owner estimates of on-
road fuel economy for the same vehicle. Such a comparison, to the
extent that owner estimates can be relied on (few owners keep precise
records), should indicate the degree to which EPA test procedures are
representing the typical on-road driving and operating conditions.
Together these results can identify overall differences between Mileage
Guide and on-road fuel economy.
A. Comparision of Mileage Guide Fuel Economy to Dynamometer Tests of
In-Use Production Cars
FY75 Emission Factor Testing - Each year EPA contracts to test approxi-
mately 2000 in-use production vehicles for emissions. The data are used
for a variety of purposes, including generation of emission factors for
use in air quality planning. This testing program is labelled Emission
Factors (EF) and is designated by the fiscal year (FY) funding used
(e.g., FY75 EF Program). Although a wide range of vehicle model years
and types are tested, the emphasis is placed on the newer models. The
vehicles are randomly borrowed from private owners and are tested as
received, without tune-up or servicing, using the same procedures and
equipment types used in EPA certification and compliance testing. Among
the tests performed on the vehicles are the city and highway fuel economy
tests.
The fuel economy data from this test program and the results of EPA
certification and compliance tests provide a basis for comparison of
prototype and production vehicles tested using the same procedures. Any
discrepancy in fuel economy will be the result of inherent differences
in the way the vehicles are produced or equipped (Are the prototypes
typical of production vehicles?) or the condition of the vehicle (Does
wear, method of mileage accumulation or improper maintenance as exper-
ienced in the real world result in lower in-use fuel economy?).
Fuel economy results from the FY75 emission factor program have been
compared to the data in the Mileage Guide. Of the 1975 and 1976 model
year vehicles tested, 812 matches with the city Mileage Guide fuel
economy were made. Because the highway test was not performed on the
1975 model year emission factor vehicles, only 230 highway fuel economy
comparisons, all on 1976 model year cars, were made.
This comparison shows that 74 percent of the vehicles tested demon-
strated city fuel economy within 10 percent of the city Mileage Guide
value. This translates to within two miles per gallon for a twenty mpg
car. Eighteen percent of the cars demonstrated city fuel economy that
-------
was more than 10 percent low, and 8 percent were more than 10 percent
better than the Mileage Guide values.—'
The agreement was poorer for the highway fuel economy tests. Sixty-
seven percent of the tests resulted in highway fuel economy within 10
percent of. the Mileage Guide values. Twenty-nine percent exhibited
highway fuel economy more than 10 percent poorer than the guide, and
only 4 percent had fuel economy 10 percent .or more above the guide. For
the highway tests, the results are heavily skewed towards lower than the
Mileage Guide in-use vehicle fuel economies; this is not as apparent for
the city results. The above information is summarized in Table 1.
The vehicle owner is more likely to compare an absolute mile per gallon
difference than a percentage. For the city results the average in-use
vehicle fuel economy was 0.5 mpg lower than the Mileage Guide. For the
highway test the difference grows to 1.7 mpg lower.
Table 1 also presents the same statistics by model year. Because no
highway tests were run on 1975 model year (MY) vehicles in this EF
program, a comparison of city differences to highway differences can be
made only for 1976 models. For 1976 model year cars, on the average,in-
use vehicles were 5 percent lower than the Mileage Guide on the city
test and 6 percent lower on the highway test, i.e:
EF mpg = 0.95 city cycle, 1976 models
Mileage Guide mpg
= 0.94 highway cycle, 1976 models
A criticism of using emission factor fuel economy data to compare with
Mileage Guide numbers is that EF cars are tested in an as-received con-
dition and many of the cars do not meet emission standards. However, as
Table 1 shows, the results for those vehicles meeting all emission
standards (47 percent) are virtually identical with the group of all
vehicles. For these relatively new vehicles, the state-of-tune can not
account for the lower fuel economy of the in-use vehicles. For that
reason, the inclusion of vehicles not meeting emission standards in
4/ The Mileage Guide values are a projected average of the numerous con-
figurations in which a specific model may be built. In the 1975 Mileage
Guide, multiple fuel economy results for a specific model (e.g., Pinto)
were presented only if the engine size or fuel system (e.g., 2 barrel vs.
4 barrel carburetor) differed. Separate values for manual and automatic
transmissions were not presented. Thus, an emission factor vehicle may
have a different transmission type, axle ratio, tire type, test weight,
body style, or optional equipment from the average vehicle represented
by the fuel economy value in the guide. However, since the sample size
in the emission factor programs is large and multiple vehicles of each
model are tested, on the whole the comparison of emission factor data to
the Mileage Guide values is valid. In the 1976 Mileage Guide the data
are presented separately for automatic and manual transmission cars.
-------
Table 1
Comparison of Production and Prototype Vehicle Fuel Economy
(As Measured by the Federal Test Procedure)
Absolute Difference (mpg)
-2 or poorer
-2 to +2
+2 or better
Percentage Difference (%)
90 or poorer
90 to 110
110 or better
All Vehicles
18%
74
8
Highway
28%
68
4
29%
67
4
Vehicles Which
Meet Emission Standards
Highway
31%
67
2
19%
72
9
30%
67
3
City
Sample
c/
All Vehicles
1975 models
1976 models
Subcompacts
Other cars '
Size Percent^1 Diff (mpg)-'
812 98 -0.5
Highway
Sample
Size Percent
273
539
196
614
103,,
95d/
94
98
40.2
-0.9
-1.4
-0.3
231
230
49
180
94
94
88
97
Diff (mpg)
-1.7
-1.7
-4.1
-0.7
a/ read as 10 percent of all EF vehicles have city fuel economy more than 2 mpg lower
than shown in the guide
W Percent defined as EF FE/Guide FE times 100
_c/ Difference defined as EF FE minus Guide FE
d_/ It may be tenuous to conclude from these data that the production to prototype fuel
economy difference is getting worse each model year. A trend of improving fuel
economy with accumulated vehicle mileage has been identified. The 1975 and 1976
model year data agree better after correcting to a common 4,000 mile basis.
Additional data, shown in Figure 3 and discussed in later sections, indicate this
mileage factor accounts for most of the variation in city results from the numerous
data sources.
-------
9
this fuel economy comparison is deemed valid.
The FY75 EF data suggest that the agreement between in-use vehicle and
prototype fuel economy is much worse for the high fuel economy of subcompacts
than for larger cars. The data suggest (Table 1) a 12 percent discrepancy
in highway values for subcompacts compared to a 3 percent discrepancy
for the larger cars. The same comparison for city fuel economy shows a
shortfall of 6 percent for subcompacts versus 2 percent for larger cars.
In terms of absolute highway fuel economy, the subcompacts on the average
are 4 mpg lower than the guide (1976 models).
FY74 Emission Factor Testing - In comparing FY74 EF city fuel economy to
the Mileage Guide, two of the possible differences between the EF vehicles
and the prototypes can be accounted for. First, the EF values are com-
pared to the specific certification vehicle test results, thus correctly
matching transmission types and vehicle weight. Second, a trend of fuel
economy improving with accumulated vehicle mileage is accounted for on a
model specific basis, thus correcting the emission factor results back
to a common 4000 mile basis. (Mileage Guide values are determined from
prototype vehicles which have accumulated an average 4000 miles, although
individual vehicle mileages as high as 10,000 are allowed and occasionally
occur. Since the EF vehicles had an average mileage of 9000 miles, the
effect of the mileage correction is to reduce the EF fuel economy values
from their measured values).
The data set chosen contains 463 cars and excludes California and high
altitude vehicles. Only city fuel economy results are compared. On the
average, in-use 1975 model year vehicles, when tested by the Federal
Test Procedure, achieved a 6 percent lower city fuel economy than the
comparable pre-production prototype vehicles used to derive the Mileage
Guide values, i.e:
EF mP§ = 0.94 city cycle, 1975 models
Cert mpg
The discrepancy is evident for virtually all manufacturers and for all
inertia weight categories.
Restorative Maintenance Testing - EPA's Restorative Maintenance Program
(RM) tests domestic in-use vehicles in an as-received condition and
after being restored, if necessary, to manufacturer's specifications. ^J
The FY76 and FY77 programs have concentrated on equal-sized samples of
the three major domestic manufacturers. The results of testing restored
1975 and 1976 model year vehicles indicate a 3 percent fuel economy
discrepancy on the city cycle and 2 percent loss on the highway cycle,
compared to the Mileage Guide. For the subset of vehicles achieving
highway guide values of 25 mpg or better, the discrepancy grows to 5 and
4 percent respectively. The average mileage on the vehicles was 8000.
_5/ Restoration includes correcting any disabled or damaged equipment,
adjustments to manufacturer's specifications and,if necessary,a complete
tune-up.
-------
10
A preliminary look at the results of the 1977 model year RM testing
(sample of 60) indicates a 6 percent shortfall from city Mileage Guide
values, and a discrepancy of 5 percent for the highway results. Average
vehicle mileage is approximately 2500.
75-76 RM mpg = .97 city cycle
Mileage Guide mpg = .98 highway cycle
77 RM mpg = .94 city cycle
Mileage Guide mpg = .95 highway cycle
Union Oil Tests - Organizations other than EPA have performed fuel
economy testing on a large number of passenger cars. Union Oil Company
in 1975 purchased the most popular new cars and accumulated 2000 miles
on each vehicle before performing both track and dynamometer fuel economy
tests. All of the cars met applicable emission standards.
The Union Oil city dynamometer test results, using the Federal Test
Procedure, have been compared to the EPA testing of similar prototype cars.
The average of two Union Oil tests on each vehicle was compared to the
results of the EPA certification counterpart; 93 matches were made. The
Union Oil vehicles exhibited an 8 percent lower fuel economy than the
prototypes used to derive the Mileage Guide values, i.e:
Union mPS = 0.92 city cycle, 1975 MY
Cert mpg
Lower Union Oil measured fuel economies are evident for virtually all manu-
facturers and all inertia weight categories.
California Assembly Line Testing - Another source of fuel economy data
using the Federal Test Procedure is the California assembly line test
program (CALT). The California Air Resources Board tests vehicles taken
from the assembly line. Approximately 100 miles are accumulated on the
vehicle before the test is performed. The selection of vehicles is
primarily based on those models that come close to failing the EPA
certification tests and is not a sales weighted sample.
Matches to the EPA certification data were obtained for 106 of the CALT
tests on 1975 model year vehicles. The CALT vehicles showed a 14
percent lower city fuel economy than the EPA certification tests used to
develop the California Mileage Guide values.—'
CALT mpg = 0_86 ci cycie 1975 models
Calif Cert mpg
6^/ California has different emission standards than are applicable to
vehicles sold in the other 49 states. A separate Mileage Guide has
been published since 1975 for California cars.
-------
11
No manufacturer or inertia weight class showed less than a 10 percent
discrepancy. This relatively large discrepancy is partially attributed
to the fact that the engines are not fully broken-in due to the low
mileage on the vehicles (100 miles or less).
Subcompact Testing - In response to complaints from the public that EPA
fuel economy values are higher than can be achieved in-use, and because
the subcompact class showed the highest slippage between prototype and
production vehicles in other programs, EPA obtained and tested three
1977 model year production vehicles of each of the fuel economy leaders
in the subcompact class. On the average these cars had accumulated 5500
miles. Each car was a near identical match to a certification counterpart.
Each car was tuned to the manufacturer's specifications, usually with a
manufacturer's representative present. Of the thirty-one cars, rep-
resenting 11 models tested, all but four met all emission standards
after being tuned to manufacturer specification. Test conditions such as
road load, inertia weight and manual transmission shift patterns were
identical to those used for testing the certification counterpart.
Three repeat tests on each car were run.
Compared to certification fuel economy, the fuel economy for this group
of 1977 model year production subcompact cars was 6 percent low for the
city cycle and 7 percent low for the highway cycle. For vehicles with a
combined Mileage Guide fuel economy of at least 30 mpg, the discrepancy
increased to 9 percent city and 11 percent highway, i.e:
0.91 city 1977 models with
*,rod !,E combined Mileage
= 0.89 highway Guide FE greater
than 30 mpg
The shortfall tended to get larger as the vehicle's Mileage Guide fuel
economy increased.
These results are similar to those from the 1975 and 1976 subcompacts
tested in the FY75 EF program in which a 6 and 12 percent discrepancy in
city and highway fuel economy respectively was found. ,
Notable is that three of the 1977 production models, all with less than
30 mpg combined fuel economy, exceeded the EPA certification fuel economy
values by 3 percent city and highway. These vehicles are the VW Rabbit,
the 2.8 litre Pinto, and the 2.0 litre Gremlin. None of the other 23
units tested exceeded their certification fuel economy. A common denominator
of these three vehicles is their engines are wholly or partially built
in Germany.
-------
12
A possible explanation for this result may be the varying build practices
of the different manufacturers. VW, for example, claims to utilize
engines removed directly from the assembly line. Other manufacturers
disassemble and inspect the engine to assure all components meet nominal
tolerances. More analysis of the build practices for test vehicles
among the various manufacturers, and quantification of the effects of
such practices, is needed before definite conclusions can be drawn on
the degree to which the shortfall can be attributed to this phenomenon.
Mileage Effect on Fuel Economy - A trend of improving city fuel economy
with vehicle accumulated mileage can be identified using the FY75 emission
factor data. The FY74 emission factor data show the same trend. Figure
3 shows an estimate of this effect. The solid line was derived from the
FY74 emission factor data on 1975 model year cars. The other data are
the results of the test programs previously discussed. These data
support the trend that city fuel economy improves with higher accumulated
vehicle mileage.
This relationship may explain why the city fuel economy discrepancy
determined from the numerous data sources varies from 14 percent at 100
miles (CALT) to -3 percent for 1975 model year cars in the FY75 EF
program (22,400 average odometer mileage). From Figure 3, the city fuel
economy discrepancy at the 4000 mile point is 5 to 6 percent.
The small sample of highway fuel economy results currently available
does not indicate a trend with accumulated mileage.
State-of-Tune and Maintenance - Tests on older, pre-catalyst cars have
shown an average 8 percent benefit in fuel economy due to regular tune-
ups. However, more recent studies on relatively new, low mileage catalyst
cars have shown that the fuel economy benefit for such cars is smaller.
The distinction for the newer low mileage cars is that the primary cause
of failure to meet emission standards is disabled parts and maladjustments,
as opposed to worn out parts. On the average, correction of maladjustment
(except idle mixture and speed) or disablements resulted in a one percent
improvement in fuel economy. For those vehicles requiring an adjustment
to idle mixture and/or speed, fuel economy improved another 3 percent.
A full tune-up, performed after the disablement and maladjustment repairs,
had a negligible effect.
The conclusion reached for the low mileage car data included in this
report is that the state-of-tune, on the average, does not explain the
differences between the prototype and production vehicle fuel economy.
As previously discussed, the FY75 emission factor data also support this
conclusion (see Table 1).
The above discussion should not be construed to conclude that vehicle
maintenance is not needed. The data indicate that for maladjusted low
mileage cars a 3 to 4 percent improvement in fuel economy is possible.
-------
13
° Tested as received
D Restored to meet emission
standards
1.05
w
•H
3
O
0)
00
td
0)
w
c
o
•H
4J
O
3
1.00
.95
.90
,85
Certification
kPrototypes _
I
01975 Models from
FY75 EF Program
1975 Models
FY7A EF Program
gl976 Models from
FY75 EF Program
D/ Upion Oil
/ 1975 Models
/ i
/
D1975 Models CALT
I I
10
I
20 30
Accumulated Miles (1000)
FIGURE 3
Production Fuel Economy vs. Odometer Mileage
(City cycle only)
-------
14
For higher mileage cars with worn parts (e.g., misfiring spark plugs) a
tune-up can restore substantially more lost fuel economy. In addition,
EPA restorative maintenance programs have shown average reductions in
tailpipe emissions of 60 percent CO, 32 percent HC, and 5 percent NOx
due to corrective maintenance.
Findings: Comparison of Mileage Guide Fuel Economy to Dynamometer
Tests of In-Use Cars - The large quantity of data from the numerous data
sources presented above strongly support a conclusion that the prototype
vehicles tested by EPA are not fully representative, from a fuel economy
standpoint, of production vehicles. In-use cars, when tested on a dyna-
mometer, achieve an average 5 percent lower city fuel economy and 6
percent lower highway fuel economy. For the high mpg subcompact class
this discrepancy grows to 9 and 11 percent respectively.
Since many of the results (RM, Union Oil, EPA Subcompact Program) are
based on cars that met the applicable emission standards, it is unlikely
that the maintenance condition or state-of-tune of these relatively new
in-use cars explains any of the observed production vs. prototype fuel
economy differences. The conditions under which mileage has been accu-
mulated has been suggested as a possible cause of this discrepancy.
(Mileage is accumulated on the certification prototype vehicles in a
relatively short time with few periods with the engine off). Although a
small sample upon which to discount this argument, the fact that seven
of the eight VW Rabbits, 2.8 litre Pintos, and 2.0 litre Gremlins exceeded
the Mileage Guide fuel economy while none of the other 23 vehicles (8
models) did, would suggest that the fuel economy discrepancy is attribu-
table to the vehicle and not the method of mileage accumulation.
The discrepancy would thus appear to be caused by physical differences
in the prototype vehicle when compared to its production counterpart.
These differences have not been identified, but may include such items
as atypical component tolerances and fuel system calibrations, or the
use on prototypes of unrepresentatively low rolling resistance tires.
B. Comparison of On-Road Fuel Economy to Dynamometer Tests of In-Use
Vehicles
Section A identified a fuel economy shortfall between prototype and
production vehicles when both vehicle types were tested on a dynamometer
using the same test procedures. The next question is whether on-road
fuel economy differs from the dynamometer test results for the same
production car.
Owner Estimates - Owners who participate in the EPA emission factor
programs are asked to estimate the city and highway fuel economy they
obtain in-use. These owner estimates have been compared to the dynamometer
-------
15
test results of the owner's car. — The conclusion drawn is that far more
people estimate poorer highway in-use fuel economy than is obtained by
testing their car using the Federal Test Procedure. For the city test
the results are more evenly distributed, as shown below.
Owner Estimate Percent of Sample
Minus EPA Test City Highway
-2 mpg or poorer 18%£/ 45%
-2 to +2 69% 48%
+2 mpg or better 13% 7%
On the average the owners estimated city fuel economy 2 percent lower
than measured by city dynamometer tests on their car. For the highway
test the average estimate was 11 percent low.
Owners of subcompacts estimated city fuel economy 3 percent better than
the test result, however for highway driving the result slipped to 9
percent low. Unlike the production versus prototype test result, this
highway discrepancy was typical for all car classes from subcompact to
full size, .i'
]_/ Data are for 1974-1976 model year vehicles in the FY75 emission factor
program. Sample size is 105 for the highway test and 565 for the city
test.
J3/ Read as 18% of car owners estimate city fuel economy more than 2 mpg
lower than the results of EPA's dynamometer test of their car.
9_/ The owner estimates are usually not based on detailed records, but
instead on a general perception of the vehicle's fuel economy. The
only measure of fuel economy is based on refueling the vehicle's tank,
and it is unlikely the entire tank would have been consumed by driving in
a manner similar to the EPA city test. The estimates for the highway
test, if based on long trips, may be more accurate, however they are
likely to be low because of the fuel economy penalty associated with op-
eration at highway speeds typically higher than represented by the
EPA test. These limitations should be considered when interpreting
these data.
-------
16
C. Overall Comparison; Mileage Guide to On-Road Fuel Economy
The results presented in the previous sections each addresses a part of
the shortfall of the EPA fuel economy numbers. By combining these
results, a comparison of the Mileage Guide to in-use, on-road fuel
economy can be made. This is schematically presented below.
Data Source
Reason for
Difference
Causes
high
F.E.
Mileage Guide
Fuel Economy
Dyno Tests of
In-Use Cars
low
F.E.
Owner Measured
or Perceived
Fuel Economy
Prototype to
production
slippage
Test procedures
not representative
of how owners
drive
Prototype vehicle
is different than
production
vehicles
Road forces not
fully simulated
Ambient conditions
different (cold,
wind, rain,
altitude)
Driving conditions
different (high
speed, short trips,
rough road, hard
driving)
OVERALL DIFFERENCE =(B/A) times (C/B)
The average fuel economy difference between the Mileage Guide and dyna-
mometer tests of in-use cars was shown to be 5 percent for the city test
and 6 percent for the highway test. Owners estimated fuel economy 2
percent lower for city driving and 11 percent lower for highway driving
compared to dynamometer tests of their cars. Combined these results
suggest an overall discrepancy between the Mileage Guide and on-road
fuel economy of 7 and 16 percent for city and highway driving respectively.
For combined city and highway driving the overall discrepancy is 11
percent.
-------
17
Overall In-Use Fuel Economy Discrepancy
(percent)
Best Estimate City Highway Combined
production vs prototype 5% 6%
on-road vs dyno tests 2% 11%
Total (Mileage Guide ,
vs On-Road) 7% 16% —' 11%
For the subcompact class the highway discrepancy increases.
Overall Subcompact In-Use Fuel Economy Discrepancy
(percent)
Best Estimate City Highway Combined
production vs prototype 9% 11%
on-road vs dyno tests j^%_. 9%
Total (Mileage Guide vs
On-Road) 6% 19% 11%
In terms of absolute miles per gallon, the discrepancy for a typical
vehicle in the fleet is approximately 1 mpg city and 4 mpg highway.
For subcompacts the shortfalls are 1.4 and 7 mpg respectively. If this
discrepancy continues the average mpg shortfall for 1985 (assuming
.an average fuel economy equal to EPCA's 27.5 mpg standard) would be
approximately 2 and 5 mpg respectively. These results are summarized
below.
Estimated Overall Fuel Economy Discrepancy, mpg
City Highway Combined
1976 average vehicle 1 mpg 4 mpg 2 mpg
1976-77 subcompacts only 1+' 7 3
projected 1985 vehicle 2 53
10/ Totals calculated from (production FE/Mileage Guide FE) times (owner-
estimated on-road FE/production FE); for example, 0.94 times 0.89
equals 0.84 or a 16 % discrepancy.
-------
18
GM Owner Survey - A check on the reasonableness of the above results
can be made by asking owners to record mileage and fuel consumed over a
period of time. General Motors asked its 1975 car owners to record
odometer readings and gallons of fuel purchased for four fill-ups. The
resulting data were screened to eliminate illogical data. EPA's analysis
of these data concentrated on those vehicles nominally six months old
for which the data recorded were obtained in the summer months. These
conditions come closest to matching the EPA certification test condition.
Because transmission type was not available, the data were compared to
the automatic transmission certification results. (A small minority of
cars has manual transmissions). In addition, no city or highway data
were indicated, so the comparison to certification data was made using
the 55/45 combined fuel economy values.
The results of the GM postcard survey suggest a 9 percent lower in-use
fuel economy compared to the EPA Certification values. This compares
well to the 11 percent combined fuel economy shortfall estimated above.
The average vehicle mileage in the GM survey was 6000.
Ill TEST PROCEDURE INFLUENCES
In the previous section a production vehicle fuel economy shortfall was
identified relative to the prototype vehicles used to establish the
Mileage Guide fuel economy values. This slippage was determined by
testing both prototype and production vehicles using identical test
procedures and test conditions. The in-use vehicle fuel economy shortfall
was attributed to a physical difference in vehicles.
To determine the overall difference between on-road fuel economy and the
Mileage Guide,the difference between owner estimated fuel economy and
dynamometer tests of the owner's car was also determined, These data
indicated a lower on-road fuel economy. The possible causes for this
part of the overall discrepancy are related to the difference between
real world driving conditions and the conditions simulated by the EPA
dynamometer tests.
A. Driving Factors
The test cycles cannot represent all types of driving. For example, the
city cycle is not representative of heavily congested downtown driving.
At an average speed of 5 mph, fuel economy will be less than one half
that represented by the typical 20 mph urban trip. Trip length also has
a significant effect. From a cold start, fuel economy for a 2 mile trip
may be 25 percent less than the typical 7.5 mile trip. A rough road can
reduce cruise fuel economy 15 percent, and underinflated tires can
account for a 7 percent loss. In- use vehicles may have high brake drag
or misaligned front tires (preliminary data indicate poorly aligned
front tires can cause a significant fuel economy penalty). Both of these
effects can increase the work required to propel the vehicle, thus
reducing in-use fuel economy. These data illustrate why no cycle can
represent all driving conditions.
-------
19
The extent to which these driving conditions account for the lower owner
perceived fuel economy has not been quantified. In several cases,
however, certain EPA test conditions tend to optimize measured fuel
economy. For example, the dynamometer surface simulates a smooth road.
All other road qualities tend to reduce fuel economy. Another example
is the average speed of the city trip (20 mph).
Change in F.E.
Average Speed (mph) Relative to a 20 mph trip
5 mph -57%
10 -30
15 -12
20 0
25 +10
30 +16
40 +25
Clearly, a slower speed city trip extracts a larger fuel economy penalty
than the benefit gained from the higher speed trip. To this unquantified
extent, the EPA test driving conditions bias the Mileage Guide results
to the high side.
Probably the most prevalent cause of lower In-use highway fuel economy
is vehicle speeds which exceed the 55 mile per hour speed limit. Average
cruise fuel economy for six cars, ranging from a Rabbit to a Continental,
indicates 70 mph speeds reduce fuel economy an average of 24 percent
from the fuel economy at 45 mph. These data are shown below:
Cruise Speed Fuel Economy Reduction from
(mph) (mpg) 45 mph, %
45 25 0
60 21 16
70 19 ' 24
80 16 36
These test results show that speeds higher than represented by the high-
way test cycle will result in on-road fuel economy significantly lower
than estimated by EPA.—'
Shift Patterns - Another test procedure aspect which can affect fuel
economy as measured by the Federal test procedure on manual transmission
cars is the shift patterns and speeds. Historically EPA has recommended
shift points for manual transmission cars, although allowing manufacturers
117 Average speed for the highway cycle is 48 mph, however since it is
a transient cycle, lower fuel economy can be expected than for a 48 mph
steady speed cruise. Portions of the highway cycle, however, represent
cruise operation at 45 and 55 mph.
-------
20
to use different values if these values are included in the owners
manual. The purpose of this provision is to allow for representative
shift points for cars which have unusual engine and gearing characteristics.
In the past few model years, use of this provision has increased from
practically nothing in model year 1975 to special shift patterns being
used for testing most of the small high fuel economy cars. The trend is
;towards shifting gears at lower speeds and skipping gears when sufficient
power is available.
A comparison of the impact on fuel economy of various shift patterns has
been made for two high fuel economy cars. The baseline shift pattern
was defined at 15-25-40 mph.—' Manufacturer recommended shift patterns,
using lower speed shift points, were used during the city test and com-
pared to the results from 15-25-40 shifting. The manufacturer's rec-
ommended shift speeds produced a 10 percent improvement in fuel economy
on the city cycle.
Such an effect has several implications. If owners do not follow the
lower speed shift patterns recommended in the owners manual, a significant
discrepancy between city in-use and Mileage Guide results will occur. Of
!equal importance, the ranking ability of the guide could be affected,
putting automatic transmission vehicles or vehicles not recommending
lower shift points at a significant disadvantage. A procedural change
which will eliminate use of unrepresentative shift patterns is being
prepared for the 1979 model year.
Road Load - Similar to shift points, EPA has historically utilized a
road load power test condition which is a function of only the vehicle
weight. In the past most manufacturers have used these recommended
values. An option to demonstrate by track coast down testing that the
vehicle has a road load different than that recommended has been available,
and in recent years manufacturers have used the alternate procedure when
to their advantage. _' (Fifty-six percent of 1978 certification vehicles
used an alternate road load value). The purpose of this option is to
give proper credit for low air resistance (drag) cars, and by giving
such credit encourage the development of cleaner aerodynamic designs to
reduce in-use fuel economy.
An improved road load prediction method based on vehicle frontal area
has been developed. This method of determining recommended road load
will become effective with 1979 model year certification. On the average
the dynamometer power setting is unchanged by the new equation; however
(L2/ This is the pattern historically recommended by EPA: 1st to 2nd
gear at 15 mph, 2nd to 3rd gear at 25 mph, 3rd to 4th gear at 40 mph.
13/ A coast down test involves shifting the car into neutral at a pre-
determined vehicle speed and allowing the car to coast. From the rate
bf deceleration the forces (road load) acting on the vehicle can be
calculated.
-------
21
the ability to predict the appropriate road load for a specific vehicle
is improved. This new method will decrease the need for manufacturers
to perform as many track coastdown tests, and will provide an improved
tool for assessing the reasonableness of requested alternate road load
settings. The effect on fuel economy will be to provide a more realistic
road load test condition for higher drag cars, which previously were
tested using the inertia weight based value.
The tire to dynamometer roll interface is also receiving much attention.
This part of the road load is not simulated by the dynamometer power
absorber since it is the vehicle that must overcome tire resistances
during the test. An uncertainty lies in how well does the interface
between the tire and the twin dynamometer rolls (see Figure 1) simulate
the road surface. The assumption in the past has been that the two tires
driving,the rolls simulate the rolling resistance of all 4 tires on the
road. —'An examination of the physical constraints of the tires pinched
between two rolls suggests that simulation of a flat road surface may be
imperfect. Slippage at the tire-roll surface may occur, and because the
two driving tires are absorbing the forces of approximately four tires
on the road, the tire warm-up characteristics differ on the dynamometer.
A study of this subject, including the effects of various tire types, is
currently underway.
A method of identifying if imperfect road load simulation is occuring is
to test a vehicle on the dynamometer and on the test track, duplicating
the test conditions as closely as possible. Test programs utilizing
this approach have been performed with results indicating the dynamometer
gives from 4 to 11 percent better fuel economy than operation on the
track. Exact duplication of test conditions is difficult to obtain,
however. Corrections for ambient condition differences between the
track and lab for example, would reduce .the reported difference by one
half. The results still directionally suggest a dynamometer to road
fuel economy difference may exist, and a carefully controlled test
program to quantify any difference is planned.
Improvements identified in these road load studies could be factored
into the new road load equation. As an initial step, different frontal
area road load equations have been developed for bias and radial tire
equipped test vehicles.
Best estimates of the effect of changes in road load on fuel economy
indicate this parameter can have a substantial impact on measured fuel
economy, i.e:
Approximate Change in Fuel
Economy due to a 10% change
in Road Load
City Highway
Power absorber 1% 4%
Rolling Resistance 2% 2%
_1_4/ The tires are pressurized to 45 psig; the front tires are stationary
during the dynamometer test (for rear wheel drive vehicles).
-------
22
Road load settings derived by manufacturers from coast down tests have
varied by as much as 40 percent from the EPA inertia weight table values,
indicating the need to evaluate the representativeness of road load
values obtained by the coast down method. In addition, improvements to
the coast down methodology are currently being developed.
Air Conditioning - EPA's test procedures simulate air conditioner usage
by increasing the dynamometer road load setting by 10 percent. A com-
parison of actual operation of the air conditioners during the dyna-
mometer and track tests to the road load simulation method showed that
the dynamometer simulation of the air conditioner produces no change in
fuel economy. — Both the track and dynamometer tests with the air
conditioner actually operating showed a ten percent penalty on both the
city and highway test. The conclusion is the current test procedure
produces results which bear no resemblance to actual air conditioner
use. A test program to develop an improved simulation based on actual
air conditioner operation is planned.
B. Environmental Factors
The fuel economy test is performed at approximately 75°F. Results of
limited 50 mph tests at lower temperatures indicate a 5 percent fuel
economy penalty at 50°F increasing to a 11 percent penalty at 20°F.
Operation on hot days does not increase fuel economy to the same extent,
Thus even if a 75°F test condition is an annual daytime temperature
average in some parts of the country, its use may not yield an average
annual fuel economy. If average fuel economy data are to be achieved,
testing at lower temperatures or development of temperature correction
factors will be required. A cold weather test program on in-use vehicles
will be conducted in the winter of 1977/78 to further quantify the
temperature effect on fuel economy.
C. Ranking
A primary purpose of the EPA fuel economy program is to provide a ranking
of vehicles by fuel economy which can be used by the consumer to select
the best fuel economy car. An analysis of the ranking ability of the
Mileage Guide combined fuel economy, when compared to the actual ranking
based on the FY75 emission factor data (dynamometer tests of in-use
cars), has been made and the results are shown below.
15/ Other EPA test results indicate a 10 percent increase in dynamometer
road load setting should decrease combined fuel economy approximately
2 percent.
-------
23
Required Difference Percent
Size Class (combined mpg) of Pairs
All 4.2 58
Full 2.5 11
Intermediate 3.2 32
Compact 3.5 45
Subcompact 5.2 41
• Light Truck 4.5 36
These data can be interpreted as follows. To have a 95 percent confidence
that the Mileage Guide ranking for a pair of cars will be the same for
the production car equivalents, the Mileage Guide combined fuel economy
difference between the two cars must be at least 4.2 mpg. If the difference
in mpg between the two cars being compared is less than 4.2, the probability
of the Mileage Guide ranking being correct will decrease. For example,
at a 2 mpg difference between two cars (not shown in the table), the
probability of a correct ranking by the Mileage Guide will be 78 percent
(pure chance will give a 50 percent probability) . The "Percent of Pairs"
column shows how many of the possible comparisons in the 1976 Mileage
Guide have a difference larger than that required for 95 percent confidence.
In the subcompact class, for example, 41 percent of the possible comparisons
have a Mileage Guide difference of at least 5.2 mpg, and for these pairs
of cars, there is at least a 95 percent chance that the equivalent
production cars will rank the same way.
This ranking ability is affected by how large a range of mpg occurs in a
specific class. In 1976, cars in the full size class differed at most
by 7 mpg. Of the possible comparisons in this class, only 11 percent
had a fuel economy difference greater than 2.5 mpg. As the range of
mileages in a class increases, the number of comparisons that will be
ranked correctly also increases. However, for no class will over half
the cars rank correctly with 95 percent confidence.
Possible factors which can improve the ranking ability of the Mileage
Guide have been discussed previously. If one manufacturer uses production
engines in the prototype vehicle EPA tests while another uses a specially
built engine which gives unrepresentatively high fuel economy, the
relative ranking of these two vehicles may be affected. Vehicles tested
with inappropriately low manual transmission shift points can achieve a
higher ranking than automatic transmission cars or cars not shifted in
this manner, but the associated fuel economy benefit will not be realized
by the car owner. The key to good ranking ability thus lies in the
equal application of the letter and spirit of the testing procedures to
each vehicle tested. In the apparent absence of an acceptance of the
spirit, the written rules may have to become more definitive.
-------
24
IV FINDINGS AND CONCLUSIONS
1. An overall fuel economy discrepancy exists for the 1975-1976 vehicle
fleet. On-road, in-use fuel economy is approximately 7 percent (1 mpg)
lower than EPA city test results and 16 percent (4 mpg) lower than the
EPA highway test results. If this percent discrepancy remains, by 1985
the mpg shortfall will increase because of the higher projected fleet
fuel economy.
2. The 1976/77 model year highway fuel economy shortfall for the sub-
compact class is approximately 19 percent (7 mpg).
3. Sixty-five percent (0.8 mpg) of the city discrepancy and 35 percent
(1.4 mpg) of the highway discrepancy appears to be due to undetermined
differences between prototype and production vehicles. On the average,
the prototypes achieve unrepresentatively higher fuel economy. For the
subcompact class the discrepancy due to production-prototype slippage is
greater (2 mpg city and 4 mpg highway).
4. City fuel economy improves with vehicle mileage. At 15 to 25 thousand
miles, the production to prototype slippage disappears. Corrections to
the data for this mileage effect bring the shortfall reported by the
various studies into closer agreement when based on a common 4000 mile
point. Identification of an overall trend of increasing fuel economy
shortfall with each later model year, as reported by one study, is not
apparent from the production and prototype dynamometer test data.
t
5. Planned changes to the manual transmission shift point criteria and
road load prediction methods will improve the representativeness of the
measured fuel economy for some vehicles. Additional planned improvements
to the tire-road force simulation are needed. (See the Appendix for a
list of other EPA fuel economy related projects).
6. Current air conditioner simulation test methods have little effect
on fuel economy. Actual air conditioner operation results in a 10
percent fuel economy penalty.
7. Implementing test procedure changes which have a directional effect
on fuel economy relative to the 1975 test procedures may impact on V
compliance with the EPCA standards. To be implemented, major procedural
changes (such as air conditioning simulation) may require consideration
of a change in the EPCA standards.
8. Proper ranking based on the Mileage Guide can be assured only if the
two cars have at least a 2 to 5 mpg (combined city/highway) difference
in ratings, depending on the size class. In most classes less than one
half the possible vehicle comparisons have fuel economy differences
larger than this. Improvement to the ranking ability can be achieved by
assuring the test procedures are equally applied to all vehicles. This
includes shift points, road load and methods of building the prototype
vehicles.
-------
Project Title
FY77 Emission Factor Program
(Passenger Cars)
FY77 Light Duty Truck Emission
Factor Program
Proposed Rulemaking; Fuel Economy
Labeling
Report on 1978 Model Year Fuel
Economy Trends
-------
Project Title
Obj ective
Evaluation of Rolling Resistance of
Various Model Tires
Improvements to the Alternate Road
Load Procedure
Manual Transmission Shift Points
Dynamometer Evaluation
Refinements to the Fuel Economy
Test Procedure
Air Conditioner Simulation Procedure
Comparison of Dynamometer vs. Track
Fuel Economy
Quantify impact on fuel economy of various
tire models and constructions for road
and dynamometer operation.
Improvements in accuracy and reduction
of variability of the alternate road
load procedure; Issue Advisory Circular.
Develop criteria for determining re-
presentativeness of shift points used
in developing EPA fuel economy estimates;
Issue advisory circular.
Evaluate need for improvements to the
dynamometer power absorber accuracy.
Evaluate effects of various parameters
(brake drag, wheel alignment, lubricants)
on measured fuel economy.
Develop improved test method for
simulation of on-road air conditioner
fuel economy penalty.
Determine if dynamometer simulated fuel
economy and track fuel economy differ for
the same driving cycle.
------- |