EPA-AA-IMS-81-23
Evaluation of I/M Effectiveness Using Emission Factors Data:
Phoenix vs. Other Low-Altitude, Non-California Sites
October, 1981
James Rutherford
Inspection and Maintenance"Staff
Emission Control Technology Division
Office of Mobile Source. Air Pollution Control
Office of Air, Noise, and Radiation
U.S. Environmental Protection . ^ency
Ann Arbor, Michigan
-------
1.0 INTRODUCTION
1.1 Purpose
The Emission Factors (EF) Testing Program is a continuing project in which the
Environmental Protection Agency (EPA) contracts with several independent
laboratories to perform emission testing of samples of in-use vehicles in
several cities in the United States. Phoenix, Arizona has been a site for EF
testing for several years both before and after the advent of its Inspection/
Maintenance (I/M) program. Federal Test Procedure (FTP) data recorded from EF
testing in Phoenix and several non-I/M cities provide an opportunity to
evaluate the effectiveness of the I/M program in reducing emissions from
in-use light duty vehicles.*
This report presents an analysis of EF data directed toward estimating the
effects of the Phoenix I/M program.
1.2 Summary
Preliminary evaluations of the means of the raw data indicate a definite
reduction in emissions due to I/M. More sophisticated analyses of the data
using analyses of covariance allow for statistical testing and quantification
of the differences due to I/M in Phoenix. As a result of the analyses of
covariance, the mean emissions of FTP HC and CO are found to be statistically
significantly lower after I/M in Phoenix than in the other non-I/M,
low-altitude, non-California EF testing sites while there was no significant
difference before I/M began. Within about two years of the start of the
Phoenix I/M program, reductions due to I/M are about twenty percent for both
HC and CO. Analyses of test results from vehicles tested in more than one EF
program provide qualitative corroboration of the reductions due to I/M.
* EPA has taken advantage of several other opportunities previously; reports
documenting this previous work are listed in the bibliography at the end of
this report.
-------
2.0 ANALYSIS
2.1 Data Inclusion
Emission Factors testing included Phoenix, Arizona in the Fiscal Year (FY)
1974, 1975, 1977 and 1979 programs.* Other low-altitude, non-California sites
in these programs were Chicago, Houston, St. Louis, and Washington, D.C.
These four sites do not have I/M programs. For this analysis, all as-received
FTP tests from the four non-I/M, low-altitude, non-California sites on model
year 1972-1978 light duty passenger cars were used. For Phoenix, all,
as-received tests for these model years were used from the FY 1974 and 1975
programs. For the FY 1977 and 1979 programs, the tests for cars of these
model years in Phoenix were used if there was indication that the vehicle had
participated in the I/M program. Some vehicles were too young when tested to
have participated yet; others may have immigrated to Phoenix too recently to
have since participated.
The model years included in the analysis provide two distinct emission control
technologies. The 1972 through 1974 model years are pre-catalyst vehicles
subject to the same new car emission standards for HC and CO. The 1975
through 1978 vehicles are first generation catalyst vehicles all subject to a
more stringent set of new car standards for HC and CO. In this report, these
two groups of vehicles are referred to as Technology I and Technology II
vehicles respectively. Although the EF programs included vehicles of model
years before and after the above mentioned groups, they are not included in
this analysis since they are sufficiently different in emission control
technology and they are not included in enough of the FY programs to provide
before and after I/M comparisons. Model years before 1972 were subject to
less stringent new car tailpipe emission standards defined on an obsolete test
procedure. Although 1979 model year vehicles were subject to the same new car
HC and CO standards as the 1975-1978 model year vehicles, the majority of the
sample for this model year was General Motors vehicles utilizing sealed idle
mixtures. They have emission characteristics sufficiently different from the
rest of the sample to cause variability problems in the analysis if included
with the 1975 through 1978 model years.
2.2 Analysis of Covariance
The individual FY programs of the EF testing provide convenient discrete test
points for comparison of test results between Phoenix and the non-I/M cites.
One FY program was carried out at approximately the same point in time at all
the sites. There was a lapse of time between FY programs. The FY74 testing
was performed before the advent of the Phoenix I/M program. Based on the
January 1, 1977 start-up date for mandatory Phoenix I/M and test dates from
the FY75 EF testing in Phoenix, it is estimated that roughly one-fourth of
those tests were on vehicles which had participated in the I/M program. For
* See bibliography at the end of this report for EF testing documentation.
-------
the FY77 and FY79 programs, information was available indicating whether the
vehicles had experienced I/M and only data from vehicles in Phoenix which had
experienced I/M were included in the analysis.
A preliminary indication of I/M effects can be gleaned from examination of
simple mean emissions and mileage at the test points. As will be seen, this
examination was performed. However, variations in accumulated mileage need to
be accounted for in making precise comparisons. This was achieved through
analysis of covariance. The effects of I/M are evaluated by comparing the
non-I/M cities with Phoenix at each test point after accounting for mileage as
a covariate. The initial analyses of covariance using raw data were found to
be invalid due to non-normality of residuals as determined by the Lilliefors
adaptation of the Kolmogorov-Smirnov test. A log transformation of the
emissions data was performed. The analyses of covariance performed on the
transformed data were found to be valid. The resultant analyses performed at
each test point within each technology with groupings for Phoenix and non-I/M
cites and mileage as a covariate provided valid statistical tests of the
effects of I/M. The adjusted means for Phoenix and non-I/M cites also provide
good estimators of mean emissions at the mean mileage for total samples (i.e.,
mean mileage for Phoenix and non-I/M cites combined).
It should be noted that the validity of log-transformed data in the analyses
of covariance and the invalidity of raw data for these analyses does not
necessarily imply non-linearity of emissions deterioration with mileage. The
usual linear estimation of emissions deterioration on mileage is based on the
useful life of vehicles. This is a much larger span of mileage than was used
in each of these analyses. While it may be more appropriate to assume
exponential deterioration of emissions over the relatively short mileage spans
used in the individual analyses of covariance presented in this report, it
could yet be more appropriate to assume linear deterioration over the life of
a vehicle as is commonly done.
-------
3.0 RESULTS
3.1 Fiscal Year Results
As explained above, the individual FY programs provide convenient discreet
points for estimating the relative deterioration in emissions that occurred
over time in the EF data. For the FY74 program, none of the data come from
vehicles exposed to I/M. Approximately one-fourth of the Phoenix vehicles had
experienced I/M before the FY75 testing. From the FY77 and FY79 programs in
Phoenix, data was only used from vehicles for which there was indication that
the vehicle had experienced I/M.
3.1.1 Raw Data Results
A preliminary examination of the I/M effects is presented in Table 1 and
Figures 1 through 4. The general conclusion from means of the raw data is
that the Phoenix mean emissions are approximately equal to mean emissions from
the other cites before I/M (FY74 means). After I/M, in the later EF programs
(FY75, FY77, and FY79), the mean emissions for Phoenix are lower than mean
emissions for the other cites.
3.1.2 Adjusted Mean Results
As discussed above, the next step in this investigation was a series of
analyses of covariance performed at each test point (i.e., for each FY
program), for each technology, for each pollutant, stratified by presence or
absence of I/M (Phoenix or other sites) with mileage as a covariate. These
analyses allow valid statistical testing of I/M effects after accounting for
mileage effects and the estimation of mean emissions at the same mileage.
Table 2 and Figures 5 through 8 present the results of these analyses. As
seen in Table 2, there are no significant differences between Phoenix and the
other sites in the FY74 program (before Phoenix I/M). Progressing through the
EF programs, a significant difference develops in both HC and CO for both
technologies between Phoenix and the other sites.
The reader may notice a counterintuitive result in Table 2 and Figure 6. The
adjusted mean CO emissions for Technology II vehicles drop between the FY74
and the FY75 programs for both Phoenix and the other sites. An investigation
of this phenomenon indicated that Technology II vehicles in the FY74 program
were primarily model year 1975 vehicles while in FY75 they were primarily 1976
vehicles. The 1976 model year vehicles in addition to having a higher
percentage of catalysts and air pumps may also have been more reliable than
the 1975 model year vehicles due to one year of experience with the new
emission control technology. Thus, the drop in emissions is understandable.
It does not affect the comparison between Phoenix and other sites.
The I/M program in Phoenix became mandatory January 1, 1977. All testing for
the FY74 EF program was completed before that date. FY75 EF testing in
Phoenix was performed in the first part of calendar year 1977. The Phoenix
testing for the FY77 program was carried out in the last three months of
-------
calendar year 1977. The Phoenix testing for the FY79 program occurred in the
first part of calendar year 1979. From Table 2, it appears that within the
first two years of the implementation of I/M in Phoenix, there is roughly a
twenty percent reduction in emissions of HC and CO relative to what the
emissions would be without I/M.
3.2 Repeated Testing Results
Many vehicles in the EF programs experienced repeated tests, that is, they
were tested in more than one FY program. Although restricting analyses to
only the vehicles with repeated tests reduces sample sizes greatly and
provides less than adequate representation of the vehicle fleet, it does
provide a qualitative corroboration of the conclusions reached in the above
analyses.
Analyses analogous to those used in the results described above were carried
out on the repeated testing samples. For example, for all vehicles included
in both the FY74 and FY75 programs, an analysis of covariance was performed
for each of the FY programs, for each pollutant, and each technology. The
adjusted means are then compared for these repeated testing samples between
Phoenix and the other sites for both FY programs.
The results of these analyses are presented in Table 3 and Figures 9 through
12. The general conclusion reached from these analyses is that with I/M the
vehicles tested in Phoenix show less deterioration from one FY program to the
next than the vehicles from the other sites. Thus, the conclusion reached in
the previous section are supported by the analyses of the repeated testing
samples.
-------
BIBLIOGRAPHY
EPA I/M Evaluations
Becker, J. , "Portland Study: First Preliminary Analysis", EPA-CAB-2/POR-1,
May, 1978.
Rutherford, J. , "Portland Study Interim Analysis: Observations on Six Months
of Vehicle Operation", EPA-IMS, January, 1979.
Rutherford, J., "Portland Study Interim Analysis: Update With Nine Months of
Vehicle Operation",'EPA-IMS-004/PS-2, May, 1979.
Becker, J. , and Rutherford, J. , "Analysis of Oregon's Inspection and
Maintenance Program", presented at 72nd Annual APCA Meeting and Exhibition,
No. 79-7.3, June, 1979.
Rutherford, J. , and Waring, R., "Update on EPA's Study of the Oregon
Inspection/Maintenance Program", presented at 73rd Annual APCA Meeting and
Exhibition, No. 80-1.2, June, 1980.
Michael, B., "Evaluation of California Change-of-Ownership I/M Program",
EPA-IMS Memo, January, 1981.
Tiao, G.C., Leaolter, J. , and Hudak, G. , "Statistical Analysis of the Effect
of Inspection and Maintenance on Carbon Monoxide Air Quality in Portland,
Oregon", EPA-460/3-81-016, May, 1981.
Michael, B., "Emission Reductions from Inspection and Maintenance: Vancouver
Versus Portland Snapshot", EPA-AA-IMS-81-18, August, 1981.
Emission Factors Testing Documentation
Bereus, A., and Hill, M., "Automobile Exhaust Emission Surveillance - Analysis
of the FY 1974 Program", EPA-460/3-76-019, September, 1976.
Rutherford, J. , "Automobile Exhaust Emission Surveillance - Analysis of the FY
1975 Program", EPA-460/3-77-022, December, 1977.
"A Study of Emissions from Passenger Cars in Six Cities", EPA-460/3-78-011-
a-b, January, 1979.
"FY79 Study of Emissions from Passenger Cars in Six Cities", EPA-460/3-80-
020A-B, October, 1980.
-------
8
Table 1
Mean FTP Emissions and Mileages for Phoenix and Other Sites
by EF Program and Technologies
EF
Program
N
Other Sites
Mileage HC
Technology
FY74
FY75
FY77
FY79
470
518
889
600
8454
14058
26141
31947
1.33
1.49
1.97
2.15
Technology
FY74
FY75
FY77
415
398
432
29646
39875
56701
3.79
3.86
4.88
CO
N
II (1975-78 Model
22.73 117
22.16 200
25.94 179
28.21 135
I (1972-74
46.33
50.32
55.64
Model
82
145
112
Phoenix
Mileage
Years)
10249
15983
20349
28805
Years)
32821
44579
53023
1
1
1
1
3
3
3
HC
.30
.34
.56
.51
.53
.67
.48
CO
23
17
18
19
45
51
43
.65
.56
.68
.63
.71
.52
.02
-------
Table 2
Adjusted Mean FTP Emissions and Mileages for Phoenix and Other Sites
by EF Program and Technologies
EF Other
Program Mileage Sites
HC
Phoenix
Difference
Sig.
CO
Other Phoenix
Sites
Difference
Sig.
Technology II (1975-78 Model Years)
FY74
FY75
FY77
FY79
8812
14595
25170
31370
1.07
1.19
1.53
1.62
1.05
1.06
1.23
1.32
2
11
20
19
ns
*
**
**
15.24 16.02 5 ns
13.47 11.08 18 *
17.26 13.82 20 **
19.03 15.95 16 *
Technology I (1972-74 Model Years)
FY74 30170 3.31 3.15 5 ns
FY75 41131 3.37 3.21 5 ns
FY77 55944 3.80 3.05 20 **
37.16 37.98 2 ns
41.75 43.32 4 ns
45.93 35.39 23 **
Sig. = Statistical Significance of the difference
emissions between Phoenix and other sites.
Significant at .01 level;
Significant at .05 level;
ns: Not significant at .05 level.
**
*
in adjusted mean
-------
10 .
Table 3
Adjusted Mean FTP Emissions and Mileages for Phoenix and Other Sites
from Repeated Testing Samples by EF Program and Technology
Sample Size
HC
CO
Sample*
74/75
75/77
77/79
74/75
75/77
Other
Sites Phoenix Mileage
Technology II
86 29 7743
21416
106 90 13386
28461
69 36 19735
35264
Technology I
127 42 26854
37443
87 84 38868
50577
Other
Sites
(1975-78 Model
1.09
1.28
1.16
1.59
1.25
1.74
(1972-74 Model
3.32
3.29
3.50
3.63
Phoenix
Years)
1.16
1.41
1.10
1.32
.98
1.27
Years)
3.12
3.12
3.17
3.22
Other
Sites
16.43
16.87
13.47
17.80
13.84
19.93
36.97
36.84
40.79
45.49
Phoenix
15.83
16.39
11.16
13.90
12.05
15.21
42.10
41.85
43.84
40.17
* The samples are restricted to vehicles tested in both FY programs, e.g.,
the 74/75 sample includes only vehicles tested in both the FY74 and FY75
EF programs.
-------
2.50
2.00
CD
1.50
CJ
X
£1.00
u_
0,50
0.0
11
Figure 1
Mean EC Emissions and Mileages for Phoenix and Other Sites
by EF Program: Technology II Vehicles*
.-•X
.*"'
.-X
OTHER SITES
PHOENIX
0 10000 20000 30000
5000 15000 25000 35000
MILEflGE
* X's indicate FY74, FY75, FY77, and FY79 means, respectively, from left to
right on each line. "Other Sites" includes no I/M vehicles. "Phoenix"
includes no I/M for FY74, about one-fourth I/M for FY75, and all I/M for
FY77 and FY79.
-------
12
Figure 2
Mean CO Emissions and Mileages for Phoenix and Other Sites
by EF Program: Technology II Vehicles
30
25
20
a:
e_
o!5
c_
^10
..x
.*••'
OTHER SITES
PHOENIX
0 10000 20000 30000
5000 15000 25000 35000
MILEflGE
* X's indicate FY74, FY75, FY77, and FY79 means, respectively, from left to
right on each line. "Other Sites" includes no I/M vehicles. "Phoenix"
includes no I/M for FY74, about one-fourth I/M for FY75, and all I/M for FY77
and FY79.
-------
5.0
H.O
:3.0
:2.0
1.0
0.0 I L
13
Figure 3
Mean HC Emissions and Mileages for Phoenix and Other Sites
by EF Program: Technology I Vehicles*
f.
X OTHER SITES
* PHOENIX
20000 30000 40000 50000 SOOOO
MILERGE
* X's indicate FY74, FY75, and FY77 means, respectively, from left to right
on each line. "Other Sites" includes no I/M vehicles. "Phoenix" includes no
I/M for FY74, about one-fourth I/M for FY75, and all I/M for FY77.
-------
14
Figure 4
Mean CO Emissions and Mileages for Phoenix and Other Sites
by EF Program: Technology I Vehicles *
SO
50
Q_
O
o30
o
'20
10
*"
_L
j_
X OTHER SITES
X PHOENIX
20000
30000
40000
MILERGE
50000
60000
* X's indicate FY74, FY75, and FY77 means, respectively, from left to
right on each line. "Other Sites" includes no I/M vehicles. "Phoenix"
includes no I/M for FY74, about one-fourth I/M for FY75, and all I/M
for FY77.
-------
Figure 5
15
Adjusted Mean HC Emissions and Mileages for Phoenix and Other Sites
by EF Program: Technology II Vehicles*
2.0
1.5
c.
err
a.
^~
u.
. 0
0.5
0.0
X OTHER SITES
X PHOENIX
0 10000 20000 30000
5000 15000 25000 35000
MILERGE
* X's indicate FY74, FY75, FY77, and FY79 means, respectively, from left to
right on each line. "Other Sites" includes no I/M vehicles. "Phoenix"
includes no I/M for FY74, about one-fourth I/M for FY75, and all I/M for FY77
and FY79. The means have been adjusted relative to Figure 1 .to provide.
estimates at identical mean mileages. (See page 5 of text.)
-------
Figure 6
16
Adjusted Mean CO Emissions and Mileages for Phoenix and Other Sites
by EF Program: Technology II Vehicles *
20
15
a:
a.
olO
c_>
a.
t—
u.
X OTHER SITES
X PHOENIX
0 10000 20000 30000
5000 15000 25000 35000
MILERGE
* X's indicate FY74, FY75, FY77, and FY79 means, respectively, from left to
right on each line. ".Other Sites" includes no I/M vehicles. "Phoenix"
includes no I/M for FY74, about one-fourth I/M for FY75, and all I/M for FY77
and FY79. The means have been adjusted relative to Figure 2 to provide
estimates at identical mean mileages. The drop in mean emissions from FY74 to
FY75 appear to be due to increase in proportion of model year 1976 vehicles.
(See page 5 of text.)
-------
Figure 7
Adjusted Mean HC Emissions and Mileages for Phoenix and Other Sites
by EF Program: Technology I Vehicles *
17
4.0
3.5
3.0
o.
o
z
a.
i . o
0.5
0.0
X OTHER SITES
X PHOENIX
20000 30000 40000 50000 60000
M1LERGE
* X's indicate FY74, FY75, FY77 means, respectively, from left to right on
each line. "Other Sites" includes no I/M vehicles. "Phoenix" includes no I/M
for FY74, about one-fourth I/M for FY75, and all I/M for FY77. The means have
been adjusted relative to Figure 3 to provide estimates at identical mean
.mileages. (See page 5 of text.)
-------
Figure 8
Adjusted Mean CO Emissions and Mileages for Phoenix and Other Sites
by EF Program: Technology I Vehicles*
18
50
= 30
u
o
u
-20
u_
10
j_
X OTHER 'SITES
X PHOENIX
20000
3000.0
40000
H1LERGE
50000
60000
* X's indicate FY74, FY75, FY77 means, respectively, from left to right on
each line. "Other Sites" includes no I/M vehicles. "Phoenix" includes no- I/M
for FY74, about one-fourth.I/M for FY75, and all I/M for FY77. The means have
been adjusted relative to Figure 4 to provide estimates at identical mean
mileages. (See page 5 of text.)
-------
19
Figure 9
Adjusted Mean HC Emissions and Mileages for Phoenix and Other Sites
from Repeated Testing Samples by EF Program: Technology II Vehicles
2.0
1.5
z
Q_
ol .0
3:
o.
i—
u.
0.5
0.0
0 10000 20000 30000 40000
5000 15000 25000 35000
MILEflGE
•- OTHER 74/75
•••••••• OTHER 75/77
— OTHER 77/79
— PHOENIX 714/75
PHOENIX 75/77
PHOENIX 77/79
-------
20
Figure 10
Adjusted Mean CO Emissions and Mileages for Phoenix and Other Sites
from Repeated Testing Samples by EF Program: Technology II Vehicles
20
15
Z
0.
o
olO
0.
OTHER 74/75
:::::::::::: OTHER 75/77
OTHER 77/79
PHOENIX 74/75
PHOENIX 75/77
PHOENIX 77/79
0 10000 20000 30000 40000
5000 15000 25000 35000
MILERGE
-------
21
Figure 11
Adjusted Mean HC Emissions and Mileages for Phoenix and Other Sites
from Repeated Testing Samples by EF Program: Technology I Vehicles
a.
C3
u2
_L
OTHER 7U/75
OTHER 75/77
PHOENIX 71/75
PHOENIX 75/77
25000
35000 45000
MILERGE
55000
-------
22
Figure 12
Adjusted Mean CO Emissions and Mileages for Phoenix and Other Sites
from Repeated Testing Samples by EF Program: Technology I Vehicles
50
40
= 30
o
a
o
20
10
OTHER 74/75
OTHER 75/77
PHOENIX 74/75
PHOENIX 75/77
25000
35000 45000
MILEflGE
55000
-------
****************************************************************
ORIGINAL
DO NOT TAKE
****************************************************************
------- |