EPA-AA-CAB-77-1
       Comparison of EPA Measured Fuel
       Economy with the Mileage Guide
             James A. Rutherford
                 August  1977
  Characterization and Applications Branch
    Emission Control Technology Division
Office of Mobile Source Air Pollution Control
     Office of Air and Waste Management
    U.S. Environmental Protection Agency

-------
                                -2-

Background

The Gas Mileage Guide presented by EPA in conjunction with FEA is a tool
for comparing new cars on the basis of fuel economy.  Test sequences
which produce the figures for the guide are precisely defined in an
effort to provide uniformity in evaluation and more scientifically
comparable results.  Prototype vehicles at 4,000 miles are driven on a
dynamometer by professional drivers in a 75° F environment.  Starting,
stopping, acceleration, and deceleration within the city and highway
cycles are intended to be representative of these modes of operation by
consumers.

It would not be expected that an owner calculating gas mileage for his
car would get the exact figure shown in the guide although the discrepancy
should not be too great.  The difference between an owner determined gas
mileage and the guide value may be considered to contain two major
components.  The first is the difference between the owner's determination
of gas mileage and the figures that would result if his car were put
through the test sequences used by EPA.  These differences include the
specific type of driving, the ambient temperature, the vehicle engine
temperature, etc.  The second is the difference between these tests run
on the consumer's in-use vehicle and the published figures in the guide
for that specific type of vehicle. These differences include prototype/
production differences as x^ell as differences in specific vehicle
configuration such as axle ratio, test weight, tires, etc.  This report
will focus upon the second of the two components.

Data

Data utilized in this report come from the Fiscal Year 1975 Emission
Factor Program.  The program includes testing of 2200 vehicles from 1966
through 1976 model years.  Consumer owned, in-use vehicles were selected
in seven cities based upon sales weighting for the determination of make
and model and based upon vehicle miles traveled for the determination of
model year characteristics.  Information in this report is based upon
the model years 1975 and 1976.

City fuel economy is calculated for each vehicle from data obtained in
the 1975 Federal Test Procedure via the carbon balance method.  Highway
fuel economy results were obtained on a subset of the vehicles via the
Federal Highway Fuel Economy Test.  Various classification parameters
(e.g., engine size, transmission type, etc.) were recorded at reception
of the vehicle for testing.  The vehicle owners were asked to complete a
questionaire which included information about vehicle use and maintenance.
Of 842 model year 1975 and 1976 vehicles included in FY75 EFP, 235 had
Highway Fuel Economy Tests performed.  City and highway fuel economies
as they appear in the 1975 and 1976 Gas Mileage Guides were used for all
but thirty of the vehicles which could not be located in the guides.

-------
                                -3-

Approach

In examining the discrepancies between EF values and guide values for
fuel economy, the prima facie approach would be to consider one value
minus the other.  However, a difference of 3 miles per gallon would
probably be more important when dealing with values around 8 miles per
gallon than around 30 miles per gallon.  This would lead to a consideration
of some relative measure.  In this report it is assumed that both absolute
and relative measures are of interest.  The analyses are performed with
both types of measures thus providing the possibility of determining
whether they lead to consistent conclusions.

Absolute differences were calculated as Emission Factor values minus Gas
Mileage Guide values (i.e., the result from HFET minus the guide highway
fuel economy and the result from FTP minus the guide city fuel economy
for each vehicle).  Relative measures were calculated as EF value as
percent of guide value (i.e., HFET divided by guide highway fuel economy
multiplied by 100 and FTP divided by guide city fuel economy multiplied
by 100).  Thus, if the EF value were less than the guide value the
difference would be negative and the percent would be something less
than 100, if the EF value x^ere equal to the guide value the difference
would be zero and the percent 100, and if the EF were greater than the
guide value the difference would be positive and the percent something
greater than 100.
     j-
That the resultant differences were not all zero and that there was a
great deal of variability will be presented later.  Beyond the overall
results it is of interest to determine whether the differences show any
systematic relationship to various vehicle classification and maintenance
factors.  Due to the nature of the measurements being utilized, normal
theory statistical approaches do not seem appropriate.  The non-parametric
method of choice for determining whether vehicle classification and
maintenance factors have significant statistical effects upon the
EF/guide fuel economy differences and percents is the analysis of
variance test applied to ranks  (termed the Kruskall-Wallis test).

In light of the stipulated purpose of the Fuel Economy Guide being the
comparison of vehicles, it is also of interest to consider the question
of how the ranking of the fuel economies of vehicles compares between EF
and  guide values.  For this purpose a nonparametric correlation measure
is utilized.  Due to the large number of tied observations the Goodman -
Kruskal Gamma is used rather than the more frequently seen Kendall's
Tau.  The G-K Gamma is similar  to a normal theory correlation coefficient
in that its possible values range from -1 to +1.  As in normal theory a
value of zero indicates independence while values approaching +1 indicate
strong agreement.  From the G-K Gamma an estimate of the probability of
concordance is calculated where the probability of concordance is
defined as the probability that, for two vehicles drawn at random from
the  appropriate stratum, if one of the measures ranks one vehicle above
the  other then  the other measure will rank them in the same order.

-------
The interaction of  fuel economy and regulated emissions has been a topic
of much interest as a result of increasing demand for fuel and more
stringent regulations.  The analysis of fuel economies in this report
are duplicated for  that subset of vehicles which are within the national
stipulated  limits for new cars on HC, CO and NOx for model years 1975
and 1976.

Results

Of the 842  1975 and 1976 model year vehicles in the data, 398 were
within the  emission limits for HC, CO and NOx.  Thirteen of these passed
vehicles were among the original thirty not appearing in the Gas Mileage
Guides.

Percents  (EF as percent of guide value) and differences (EF minus guide
value) were calculated for each available vehicle for highway and city
fuel economies.  A  rough presentation of the results of these calculations
appears in  Table 1.  Using the arbitrary figures of a difference of less
than two or within  ten percent as non-significant differences it is seen
that about  thirty percent of these vehicles have significantly lower EF
values than guide values on the highway cycle and 10 to 20 percent on
the city cycle.  Less than ten percent of the vehicles have significantly
higher EF values on both cycles.

As will be  seen throughout there is general concurrence of results both
between percents and differences and between all vehicles and passed
vehicles.   However, as seen in Table 2, there is quite a bit of disagreement
between city and highway cycles for the same vehicle.  For example, of
the 67 vehicles in  the less than 90% group on the highway cycle, only 37
of these vehicles are in the less than 90% group for the city cycle.

Tables 3 and 4 give the medians and means for the differences and
percents by the various factors under investigation.  Table 3 includes
"all available data  while Table 4 presents results for those vehicles
which were  within the limits set by Federal regulations on HC, CO and
NOx emissions.  The asterisks  (*) indicate the Kruskal-Wallis tests
which achieved the  nominal 0.01 significance level for testing whether
the factor  in question has any relationship to the measure in question.
For example, in Table 3 the asterisks following "Site" indicate that
"site" has  a significant effect upon both percents and differences for
the city  cycle but  for neither on the highway cycle.  That is, on the
city cycle, the city in which a vehicle was operated affected the vehicle's
relative  in-use/guide fuel economy.  These two tables will be summarized
by factor headxngs  below:

     All Vehicles:  The overall medians and means appear for all vehicles
           in Table  3 and for all passed vehicles in Table 4.

-------
                           —5—


Site:  For both groups of vehicles site was a significant factor
     on the city cycle but not on the highway cycle.  It appears
     that vehicles in Houston had lower EF values for the city
     cycle relative to the guide than other cities.

Model Year:  The city cycle showed significant differences for
     model year for both groups of vehicles.  Model year 1975
     vehicles had higher EF values relative to the guide than 1976
     model year vehicles.  There was an insufficient number of
     model year 1975 vehicles with Highway Fuel Economy Tests to
     make an adequate comparison on the highway cycle.

Model Size;  All tests were significant based upon model size.
     Subcompacts had lower EF values relative to the guide than
     other vehicles.

Cylinders:  All tests were significant for number of cylinders.
     Four cylinder vehicles had lower EF values relative to the
     guide than other vehicles.  On each measure it appears that
     the vehicles generally are in the ordering four cylinders, six
     cylinders, eight cylinders from least to greatest.

Garb Venturis:  Only the test for highway differences on passed
     vehicles turned out significant.  This might be considered a
     spurious result since the same test on all vehicles which
     showed the same general trend with a larger sample did not
     show significance.

CID;  All tests were significant for engine size as measured in
     cubic inches displacement.  The smallest engine size category,
     0-150, showed lower EF values relative to the guide than
     larger engine size vehicles.

Transmission:  All tests were significant when comparing manual
   »  transmission vehicles with automatics.  Manual transmission
     vehicles showed lower EF values relative to the guide than
     automatics.

Manufacturer:  Manufacturer was a significant factor by all tests.
     Vehicles in the "Other" category  (not AMC, Chrysler, Ford, or
     GM) had lower EF values relative  to the guide than vehicles in
     the other categories.

Catalyst;  For both groups of vehicles presence of catalyst only
     appeared as a significant factor  for highway differences.
     Those vehicles which had no catalyst showed lower EF values
     relative to guide values for highway differences than vehicles
     with  catalyst.

Primary Use;  None of the  tests showed significance for usual load.

-------
                              -6-


     Maintained According to Manufacturer's Recommendation?;   None of
          the tests showed significance for this factor.

     Satisfied with Engine Performance7:   This factor turned  out significant
          for both percents and differences on the highway cycle for the
          group consisting of all avaliable vehicles.  Those  vehicles
          for which the owner answered "yes" to the question  were in the
          middle with "no" below and "most of the time" above for EF
          values relative to guide values.

     How Often Tuned?:   This factor showed significant effect for all
          vehicles on the city cycle.  Vehicles with owners answering
          that their vehicles were tuned less often than once per year
          had higher EF values relative to guide values.

     Last Tune7:  This factor turned out significant for both groups of
          vehicles on the city cycle for both percents and differences.
          Vehicles owned by people who answered that the last tune was
          more than a year previous had higher EF values relative to the
          guide than other vehicles.  Note that this result as well as
     ,     the above result concerning the frequency of tune could be
          confounded with the effect of model year.

     Who Tuned?:  Only the percents on the city cycle for the group
          consisting of all available vehicles showed significant effects
          from this factor.  This should probably be considered non-
          meaningful significance since there appears to be little range
          in the medians and means.

     Mileage Group:  Both groups of vehicles showed significance on both
          tests for the city cycle while none of the tests for the
          highway cycle were significant.  It appears that for the city
          cycle vehicles with low mileage have low EF values  with
          respect to guide values and as mileage increases EF values
          approach guide values.  No such trend is observed on the
          highway cycle.

Although all four measures (city percents, city differences,  highway
percents and highway differences) appear to be slightly lower for the
group consisting of passed vehicles than for the entire group of vehicles,
Kruskal-Wallis tests for differences between the group of vehicles which
passed and the group which didn't were far from significant for all four
measures.  As observed above, the tests on the two groups of  vehicles
lead to generally consistant results.

Tables 5 and 6 present the results of the calculations of the Goodman-
Kruskal Gamma and an estimate of the probability of concordance for
appropriate groups.  These measure the agreement in ranking between
Emission Factor test values and Gas Mileage Guide values.  The Goodraan-
Kruskal Gamma is similar to a correlation coefficient ranging from -1 to
+1 and the probability of concordance is the probability that two

-------
                              -7-

vehicles drawn at random from the appropriate group would be ranked in
the same order by the two determinations of fuel economy.  Thetables
stratified by vehicle size groups since these groups correspond to the
current organization of Gas Mileage Guide and they form a breakdown
which would be practical for a consumer considering the purchase of a
new vehicle.  It is seen that for larger groupings (all vehicles, all
1976 vehicles, etc.) the probability of concordance and the G-K Gamma
is larger than for the more discrete breakdowns.  This is reflective of
the fact that the smaller groups are fairly homogeneous within while a
large  degree of heterogeneity exists between the groups.  These measures
are of the same order of magnitude and generally show the same trends
when comparing the two groups of vehicles (passed vehicles and all
vehicles).

Conclusions

Based on all avilable data, 68% and 88% of vehicles had Emission Factor
fuel economies within 2 miles per gallon of Gas Mileage Guide fuel
economies for the highway cycle and city cycle respectively.  In terms
of relative fuel economies, 67% (highway) and 74% (city) of these vehicles
had Emission Factor fuel economies within ten percent of the guide
values. Twenty-eight percent on highway and ten percent on city had EF
fuel economies more than 2 miles per gallon less than the guide values.
Twenty-nine percent and eighteen percent for highway and city respectively
had EF fuel economies that were less than ninety percent of guide values.
The results for the group of vehicles which were within emission standards
were very close to the above.

Those vehicles which generally have relatively high fuel economies
(e.g., vehicles with engines smaller than 150 CID) showed EF fuel
economies lower than other vehicles relative to the guide values.
Vehicles which had the least maintenance appeared to have higher EF
economies relative to guide values.  However, the significance of this
is uncertain since the quality of the received maintenance is not known.
It could imply that the "maintenance" consisted of carburetor adjustment
which was detrimental to fuel economy.  On the city cycle vehicles with
low mileage have low EF fuel economies relative to guide values and as
mileage increases EF values approach guide values.

The difference between the total group of vehicles in the analyses and
that subset of vehicles which were within emission standards is essentially
insignificant.

-------
                 -8-
              Table 1

  Number and Percent of Vehicles
 in Difference and Percent Groups
AH Vehicles
Passed Vehicles
Highway
N7
/>
Total
Differences
< -2
-2 to +2
> +2
Percents
< 90
90-110
> 110
231
65
157
9
67
154
10
100
28
68
4
29
67
4
City
N %
812
78
712
22
143
601
68
100
10
88
2
18
74
8
Highway
N %,
132
41
88
3
40
89
3
100
31
67
2
30
67
3
City
N %
385
46
328
11
72
279
34
100
12
85
3
19
72
9

-------
                                     -9-
                                 Table 2
                   Number of Vehicles  Cross-Classified
                       by Highway and  City Groups
                                             Highway
                            All Vehicles
                                 Passed Vehicles
Differences
< -2
-2 to +2
> +2
<-2
37
30
0
-2 to +2
3
153
1
>+2
0
9
0
<-2
, 23
17
0
-2 to +2
2
85
1
>+2
0
3
0
Percents
< 90
90 - 110
> 110
< 90

 29
90 - 110  > 110
                     <  90    90  -  110    >  110
 24
126
  4
             0
             5
             5
22
17
 1
14
71
 4
0
1
2

-------
-10-
Table 3
Medians and Means of Percents and Differences by Factors
All Vehicles (* indicates nominal significance for
appropriate Kruskal-Wallis test)
Highway City
N Percents Differences N Percents Differences
All Vehicles
Site
Chicago
Denver
Houston
Los Angeles
St. Louis
Washington
Phoenix

Model Year
1975
1976
Model Size
Full Size
Intermediate
Compact
Subcompact
Truck
231
43
43
43
43
13
44
2

1
230
29
49
51
49
51
Median
94
99
94
93
90
92
96
102

102
94
96
97
95
89
93
Mean
94
97
94
93
90
92
96
102

102
95
*
96
97
97
88
92
Median
-1.3
_ o
-1.5
-1.4
-2.0
-2.0
-1.0
.7

.5
-1.3
~~ * /
- .6
-1.0
-3.7
-1.4
Mean
-1.7
— 9
-1.7
-1.9
-2.5
-2.4
-1.3
.7

.5
-1.7
*
- .8
- .7
- .7
-4.1
-1.9
- Median Mean Median
812
208
75
77
78
104
79
191

273
539
150
197
178
196
89
97
*
98
99
93
98
95
96
98
i
100
95
97
99
98
94
99
98 -.5
98 - .3
100 - .1
94 -1.0
101 - .2
95 - .7
97 - .6
98 - .3

103 .1
95 - .7
97 -'.4
99 - .2
98 - .3
94 -1.4
102 - .2
Mean
- .5
*
- .4
- .2
-1.1
.0
-1.0
- .7
- .4

*
.2
- .9
*
-'.4
- .1
*' - .4
-1.4
- .1

-------
     -11-
 Table 3 (con't)
Highway

N
Percents
Median
Cylinders
Four
Six
Eight
Carb Venturis
One
Two
Four
Fuel Injection
CID
0-150
151-250

251-330
331-399
400
Transmission

Automatic
Manual

55
47
128

52
133
40
6

55
36

47
62
31


175
56

89
93
96

94
92
97
93

89
93

95
98
96


95
88
Mean
*
88
93
99

94
93
97
94
*
88
94

95
97
96
*

96
88
Differences
Median

-3.9
-1.6
- .7

-1.4
-1.6
- .5
-2.6

-3.9
-1.4

-1.2
- .4
- .7


- .9
-3.9
Mean
*
-4.1
-1.6
- .7

-1.6
-2.0
- .7
-2.0
*
-4.1
-1.3

-1.2
- .5
- .8
*

- .9
-4.0
N
Percents
Median

185
133
491

134
496
159
18

187
110

145
245
125


642
170

93
97
98

97
97
97
97

93
97

99
99
96


98
94
Mean
*
94
97
99

98
98
97
98
*
94
98

100
100
96
*
.
98
95
Differences
Median Mean

-1.6
- .5
9
» L.

- .5
- .5
- .4
- .5

-1.6
- .5

- .1
- .2
- .5

,
- .3
-1.1
*
-1.5
""" » J
- .1

- .5
- .5
- .4
- .4
*
-1.5
- .4
\
- .1
.0
- .5
*
- ~i -J
™" • J
-1.2

-------
      -12-




 Table 3 (con't)
Highway                               Cit}
N
Percents Differences N Percents Differences
Median Mean Median Mean Median Mean Median Mean
Manuf ac tur er
AMC 5
Chrysler 32
Ford 57
GM 101
Other 36
Catalyst
Yes 198
No 33
Primary Use
Driver only 145
Driver & 1 48
passenger
Driver & 2 28
passengers

Maintained
According to
Mfg. Rec.?
* * * *
95 98 -1.0 - .5 22 100 99 - .1 - .2
94 97 -1.2 - .7 108 98 98 - .3 - .4
92 93 -1.6 -1.7 194 96 99 - .6 - .4
97 96 - .7 -1.0 364 98 98 - .3 - .3
88 87 -4.0 -4.5 124 94 94 -1.5 -1.5
*
94 94 -1.1 -1.5 700 97 98 - .4 - .4
91 93 -3.0 -2.7 112 97 97 - .7 - .9
94 93 -1.4 -1.8 518 97 98 - .5 - .5
96 96 - .8 -1.1 186 98 98 - .3 - .5

92 91 -1.5 -2.1 76 95 96 - .7 - .8

i i
i • *. j.
i ^ —^ -.
Yes
No
Not sure
224
2
4
94
103
94
94
103
95
-1.4
.6
-1.4
              -1.7   772    97  '




                 .6    15    99




              -1.6    19   100
 97   - .5




110   - .2




 97     .0
- .5




 1.0




- .8

-------
      -13-
 Table 3 (con't)
Highway                               Cit\

N
Percents
Median
Mean
Differences
Median
Mean
N
•w-
Percents
Median
Mean
_ i — _ -
Differences
Median Mean
Satisfied with'
Engine
Performance?
Yes
Most of the
Time
No
183
31

17
*
93 93
99

91
99

90
-1.5
- .2

-1.8
*
-1.8
- .4

-2.4
658
96

58
97
98

95
98
98

95
- .4
- .3

- .8
How often tuned? *
Not yet
Mfg. Rec
6 Months
Year
Less often
Don't know
Last tune?
Too new
Due, not done

0-6 months

6-12 months

Over 1 year
Don't know
160
30
21
15
1
4
168
12

45

2

0
4
94
92
93
95
101
99
94
94

93

85

—
99
94
90
93
94
101
96
94
94

92

85

—
96
-1.2
-2.0
-1.4
-1.0
.2
_ 2
-1.2
-1.7

-1.9

-3.4

	
- .2
-1.5
-2.7
-2.0
-1.5
.2
- .9
-1.5
-1.6

-2.3

-3.4

	
- .9
400
118
153
106
17
18
397
. 59
v
273

42

19
22
96
97
98
99
100
96
96
98

98

100

101
98
96
98
98
100
109
98
*
96
99
i. i •%
98
„
99

109
99
- .6
- .4
- .3
- .1
.0
- .5
4
- .6
_ .3
-
- .3
i
.0
v .,
.1
- .3
- .5
- .3

- .9
*
- .7
- .5
- .4
- .1
.8
- .4
*
- .7
, - .3

- .4
-~*
- .3
r
.9
- .1

-------
      -14-
 Table 3 (con't)
Highway                              Citi

N
Percents
Median
Who tuned?
None
Dealer
Indep. Garage
Clinic
Self
Don't know
Mileage Group
< 4K
4K-10K
10K-20K
20K-30K
> 30K

180
37
4
0
6
4

53
95
66
12
5

94
92
93
—
99
99

92
95
93
97
91
Mean

94
91
93
—
95
96

93
94
94
97
94
Differences
Median

-1.2 .
-2.0
-1.5
	
- .3
- .2

-1.7
-1.1
-1.5
- .5
-1.9
Mean

-1.5
-2.5
-1.8
	
-2.0
- .9

-2.0
-1.7
-1.7
- .5
-1.7
N
Percents
Median

451
189
79
15
51 ,
27

100
233
289
, 141
49

96
98
98
96
98
99

93
96
98
99
99
Mean
*
97
99
99
97
99
104
*
93
96
99
100
100
Differences
Median

- .6
- .3
- .2
- .5
_ 2
_ 2

-1.0
- .6
- .3
- .1
- .1
Mean
-~
- .7
- .4
- .2
- .6
- .3
.3
*
-1.3
- .7
- .4
- .1
- .1

-------
-15-
Table 4
Medians and Means of Percents and Differences by Factors
Passed Vehicles (* indicates nominal significance for
appropriate Krusal-Wallis test)
Highway \ City
N Percents Differences N Percents Differences
Median
All Vehicles
Site
Chicago
Denver
Houston
Los Angeles
St. Louis
Washington
Phoenix
Model Year
1975
1976
Model Size
Full Size
Intermediate

Compact
Subcompact
Truck
132
25
25
25
24
6
- 25
2
1
131
18
32

25
35
20
93
96
93
93
90
84
95
102
102
93
96
94

95
87
98
Mean
93
96
93
92
91
85
95
102
102
93
*
95
94

98
87
96
Median
-1.5
- .7
-1.8
-1.4
-2.1
-4.5
-1.2
.7
.5
-1.6
- .8
-1.2

-1.1
-4.4
- .4
Mean
-1.9
-1.3
-1.8
-2.1
-2.3
-4.6
-1.5
.7
.5
-1.9
*
- .8
-1.1

- -5
-4.6
-1.2
Median Mean
385
95
39
40
40
41
41
89
110
275
67
82
-
84
108
42
96
*
97
99
91
97
94
94
98
*
102
95
*
95
97
- t
99
94
99
97
98
100
92
99
94
97
98
103
95
96
97

100
94
101
Median
- .6
- .3
_ j^
-1.0
_ o
- .8
- .9
- .3
.3
- .8
- .7
- .5
>
_ j^
-1.6
- .1
Mean
- .6
*
- .5
- .1
-1.4
- .3
-1.2
- .7
- .4
*
.3
— 9
*
- .5
- .4

- .1
-1.5
.1

-------
      -16-
Table 4 (con't)
Highway
Cits

N
Percents
Median Mean
Cylinders
Four
Six
Eight
Garb Venturis
One
Two
Four
Fuel Injection

CID
0-150
151-250
251-330

331-399
400
Transmission
Automatic
Manual

36
28
67

27
67
32
6


36
25
12
1
42
17
-
98
34
*
88
93
96

95
91
97
93

*
88
93
94

97
96
ft
95
90

88
94
96

95
91
96
96


88
95
93

96
95

95
88
Differences
Median

-4.2
-1.5
- .7

-1.1
-2.0
- .6
-2.6
\
'x
-4.2
-1.4
-1.4

- .7
- .7

- .9
-3.8
Mean
*
-4.4
-1.3
- .8
*
-1.0
-2.8
- .7
-2.0

*
-4.4
-1.0
-1.6

- .7
- .7
*
-1.1
-4.2
N
Percents
Median
Mean
P^^^^B
Differences
Median
Mean
A A*
*»
102 '
69
213

61
233
79
10


102
60
55

116
52

299
86
94
98
98

98
97
96
95


94
99
100

98
94

97
94
94
98
99

98
98
96
96

*
94
98
101

98
95
*
98
95
-1.7
- .4
- .3

- .4
- .6
- .6
- .8


-1.7
- .2
- .1

""" » J
- .8

- .5
-1.2
-1.5
- .4
_ 2

- .5
- .6
- .5
- .7

*
-1.5
-*.3
- .1

- .2
- .7
*
- .4
-1.3

-------
      -17-
Table 4 (con't)
Highway
Citi
Manufacturer
AMC
Chrysler
Ford
GM
Other
Catalyst
Driver Only
Driver & 1
passenger
Driver & 2
passengers
Maintained
According to
Mfg. Rec?
Yes
No
Not Sure
Satisfied with
Engine
Performance?
Yes
Most of the time
No
N
0
10
36
56
30

80
28

17


127
1
4

106
17
9
Percents
Median Mean
ft
102 101
92 92
96 96
89 88

93 93
95 96

91 91


93 93
99 99
94 95

93 93
99 98
89 88
Differences
Median Mean
ft
.4 .2
-1.7 -1.8
- .7 - .9
-3.9 -4.4
*
-1.5 -2.0
- .8 -1.4

-1.8 -2.3
*

-1.6 -1.9
- .3 - .3
-1.4 -1.6

-1.7 -2.0
- .3 - .5
-2.4 -2.9
N
4
20
116
171
74

244
87

39


364
10
10

321
40
24
Percents
Median
*
97
99
94
98
94

97
96

94


96
99
93

96
97
94
Mean
95
99
97
99
94

98
97

95


97
102
94

97
98
94
Differences
Median
- .5
- .1
- .8
- .3
-1.6

- -5
- .5

- .9


- .6
- .1
-1.0

- .6
- .4
-1.0
Mean
*
-1.2
- .3
- .6
- .2
-1.5

- .5
- .6

-1.0


- .6
- .2
-1.3

- .6
- .4
-1.1

-------
                       -18-
N
Percents
Table 4 (con't)

Highway
     Differences
N
           City
Percents       Differences
Median
How often tuned7
Not yet
Mfg. Rec.
6 months
Year
Less often
Don't know
Last tune7
Too new
Due, not done
0-6 months
6-12 months
Over 1 year
Don1 t kom*
Who tuned?
None

Dealer

Indep. Garage
Clinic
Self
Don't know

86
19
14
10
0
3

93
8
27
1
0
3

101

23

1
0
4
3

93
93
92
94
—
100

93
94
92
102
—
100

93

93

83
—
97
100
Mean

94
91
91
94
—
95

94
94
90
102
—
95

94

91

83
—
93
95
Median

-1.5
-2.0
-2.0
-1.1
	
.0

-1.3
-1.7
-2.0
.5
	
.0

-1.4

-2.0

-5.4
	
-1.6
.0
Mean

-1.7
-2.6
-2.7
-1.6
	
-1.0

-1.6
-1.7
-3.0
.5
	
-1.0

-1.6

-2.7

-5.4
	
-3.0
-1.0
Median Mean

192
64
69
45
6
9

192
34
126
14
7
12

224

89

34
4
22
12

96
97
97
98
111
94
*
95
100
97
100
111
100

96
'
98
t
97
97
97
100
•i-i ii P
96
96
98
99
109
98
•
95
94
98
97
110
100

96
"
98

99
95
98
100
Median

- .7
- .6
- .5
- .3
1.6
_ 7
• /
*
- .8
.0
- .5
.0
1.3
* .0

- .7
'* *
- .3
i
- .3
- .4
- .5
.0
Mean
^
- .8
- .7
- .4
— 2
.9
- .3

- .9
.1
- .5
- .6
1.1
.1

-".8
,
- .4
- "'.
J_ o
— » o
- .5
.1

-------
                      -19-
N
Percents
Table 4 (con't)

Highway
     Differences
N
           City
Percents       Differences
Median
Mileage Group
< 4K
4K-10K
10K-20K
> 20K

30
50
40
12

92
93
93
100
Mean

93
93
93
97
Median Mean Median
*
-1.7 -2.0 58 93
-1.8 -2.0 103 94
-1.7 -2.0 141 97
- .1 - .8 83 99
Mean

93
96
98
99
Median Mean
*
-1.0 -1.2
- .9 - .8
- .5 - .4
- .1 - .3

-------
                                   -20-
                                  Table 5
                      Goodman-Kruskal Gamma and the
                 Probability of Concordance	All Vehicles
Model
Year
both
1 Q7*;
j-y 1 J
ti
tt
ii
n
it
Model Size
All
A1 1
riJLJL
Fullsize
Intermediate
Compact
Subcompact
Truck
HIGHWAY
n gamma p n
231 .7638 .8819 812
"I — - 97^
J. — — e.1 J
65
56
48
65
38
CITY
gamma
.8172
~IT\r\
• / / JU
.7328
.5394
.5875
.6962
.3239
£
.9086
OQfiC
. OO\)J
.8664
.7697
.7938
.8481
.6620
1976       All               230    .7632  .8816           539    .8585   .9293
           Fullsize           29    .5700  .7850            85    .6873   .8437
  "        Intermediate       49    .4513  .7065           141    .6742   .8371
  "        Compact            50    .7333  .8667           130    .6889   .8444
  "        Subcompact         49    .6927  .8464           131    .7219   .8610
  11        Truck              51    .5389  .7695            51    .8010   .9005

-------
                    -21-
                   Table 6

       Goodman-Kruskal Gamma and the
Probability of Concordance	Passed Vehicles
Model
Year
both
1Q71!
j~y / J
n
ii
n
ii
n
1976
n
ii
ii
n
n
Model Size
All
All
jTl-LJL
Fullsize
Intermediate
Compact
.. Subcompact
Truck
All
Fullsize
Intermediate
Compact
Subcompact
Truck
n
132
i
JL




131
18
32
25
35
* 20
HIGHWAY
gamma
.7853





\
.7847
.5046
.5253
.6637
.7036
.3467
£
.8765





.8924
.7523
.7627
.8319
.8518
.6734
n
385
19
10
21
37
22
275
48
72
63
71
20
CITY
gamma
.8156
777Q
• / / / y
.7458
.2500
.6647
.7539
.1034
.8699
.6758
.7709
.6446
.6870
.6508
£
.9078
QOQfl
* OO.7W
.8729
.6250
.8324
.8770
.5517
.9350
.8379
.8855
.8223
.8435
.8254

-------
                                -22-

Abstract

Data from 842 vehicles (model years 1975-1976) in the Fiscal Year 1975
Emission Factor Program are utilized to examine the differences between
fuel economies derived from EPA tests on in-use, consumer-owned vehicles
and the appropriate values for each vehicle which appear in the Gas
Mileage Guides.  The discrepancies are examined in terms of absolute
differences and percentages.  Various vehicle classification, maintenance
and utilization factors are investigated to determine their relationship
to these discrepancies.  The agreement in ranking of vehicles on fuel
economy between EF and guide determined economies is also investigated.
All analyses are performed on all available vehicles and on the subset
of vehicles which pass national standards on HC, CO and NOx emissions.

-------