Analysis of Light-Duty Vehicle Fuel Switching in the NPD Data Base

ANALYSIS OF LIGHT-DUTY VEHICLE
FUEL SWITCHING IN THE NPD DATA BASE
EPA Prime Contract 68-01-6558	j
ENERGY AND ENVIRONMENTAL ANALYSIS, INC.
1655 North Fort Myer Drive
Arlington, Virginia 22209
(703) 528-1900

-------
EPA-420-R-84-100
ANALYSIS OF LIGHT-DUTY VEHICLE
FUEL SWITCHING IN THE NPD DATA BASE
EPA Prime Contract 68-01-6558
Subcontract 130.109
Work Assignment No. 30, Task 4
Work Assignment No. 30A, Task 2
Prepared for:
ENVIRONMENTAL PROTECTION AGENCY
Motor Vehicle Emissions Laboratory
Ann Arbor, Michgan
Prepared by:
ENERGY AND ENVIRONMENTAL ANALYSIS, INC.
1655 North Fort Myer Drive
Arlington, Virginia 22209
September 1984

-------
TABLE OF CONTENTS
Page
PREFACE 		1
1.	THE NPD DATA BASE 		2
1.1	Selection of Che Panel 		2
1.2	Coverage of Leased Vehicles 		3
1.3	Determination of Fuel Type Requirement		3
l.A Trucks In NPD 		3
1.5	Non-Catalyst Cars With Unleaded Fuel Requirements ..	4
1.6	Confidence Intervals 		4
2.	VALIDITY OF THE DATA			5
3.	ASSUMPTIONS PERTAINING TO THIS ANALYSIS 		14
3.1	Unknown Fuel Purchase Assumption 		14
3.2	Reporting Errors Assumption 		15

-------
LIST OF TABLES
Page
1	Composition and Location of Households: NPD Versus
RECS and NFO 		 6
2	Age and Economic Status of Households: NPD Versus
RECS and NFO 	 7
3	Household Income (1981$) NPD Versus U.S. Census 	 8
4	Comparison of Alternate Assumptions About Unknown Fuel
Type Purchases Using 1982 NPD Data 	 16
5	Comparison of Vehicle Involvement in Misfueling in 1982
Reset Maximum-of-Two Leaded Purchases to Unleaded
Versus No Reset 	 18
6	Comparison of Vehicle Involvement in Misfueling in 1982
Reset Maximum-of-Two Leaded to Unleaded Versus No Reset 	 19
LIST OF FIGURES
Page
1	Monthly Trends in Household VMT 	 9
Notes for Figure 1 Comparability of Household VMT Sources .... 10
2	Annual VMT Per Car By Age 	 12
3	Annual VMT Per Light Truck By Age 	 13

-------
PREFACE
This is the final report on misfueling work for EPA. Misfueling is
defined as the use of leaded gasoline in vehicles certified for the use
of unleaded gasoline. EPA is publishing the results to stimulate public
discussion of the subject but the results do not represent EPA conclu-
sions.
The work on this project was done under EPA Prime Contract 68-01-6558,
Subcontract 130.109. This report combines the results from two work
assignment/tasks:
Work Assignment 30, Task 4 "Misfueling of Light-Duty Vehicles
in 1981: Interim Report"
Work Assignment 30A, Task 2 "Misfueling of Light-Duty Vehicles
in 1982: Interim Report"
Each of these task reports is included as a section in this final
report.
Part I of this report, "Misfueling of Light-Duty Vehicles in 1981:
Interim Report," quantifies the rate of misfueling of light-duty vehicle
(LDV) owners/operators in 1981. The report presents highlights of the
misfueling trends observed in a time series data base of fuel purchasing
behavior. Findings are presented without any attempt at interpretation.
Part II of this report, "Misfueling of Light-Duty Vehicles in 1982:
Interim Report," repeats the work done for Part I using data for 1982.
-1-

-------
1. THE NPD DATA BASE
The survey Information in the NFD data base is derived from the NPD
Petroleum Marketing Index (PMI), a diary panel survey of over 5,000
households conducted by NPD Research, Inc. Panel members are chosen on
the basis of demographic characteristics and geographical location. The
panel does not include singles or non-family households.
In order to ensure that consistent demographic information is available
from all households, NPD requires an adult female be present in all
families selected for participation. The rest of Section 1 of this
preface discusses important characteristics of the NPD data base which
should be kept in mind when examining the results presented In this
report.
1.1 Selection of the Panel
Possible respondents for the PMI panel are selected from the American
Shoppers Panel (ASP). Each candidate is sent a letter asking about his
or her interest in participating in a vehicle use diary panel. Between
60 and 65 percent of those asked respond and return their initial
questionnaires. This percentage includes those people who own no
vehicle. The questionnaire"requests information including the VIN and
other engine characteristics for a maximum of five separate vehicles per
family. Since the respondents are already in the ASP, no demographic
data needs to be collected. NPD selects a subset of these respondents
on a demographic basis in order to maintain a balanced sample for the
PMI survey. Those selected are sent a monthly diary, a visor holder,
and an introductory letter. There is a 75 to 80 percent response to the
first monthly diary. Respondents are guaranteed anonymity and at no
time are told who will be using the data, although they do know that
-2-

-------
companies buy Che data for gasoline brand market share studies.
Respondents are given a hotline number to call if they have any ques-
tions, but the people who staff the phones do not know themselves who
the clients are. Participants in the ASP are recruited from a variety
of mailing lists. The response rate at this stage varies from 2 to 25
percent, depending upon the scope of a particular recruitment effort.
1.2	Coverage of Leased Vehicles
Survey respondents were asked to include "all leased cars whether leased
by a company, a business, or privately by any family member."* Unfortu-
nately, it is impossible to determine from the available data which
vehicles are leased by a company or business.
1.3	Determination of Fuel Type Requirement
Although the respondent was asked if the vehicle required unleaded fuel,
the answer to this question was not used in this report for classifying
the vehicle fuel requirements. Rather, the fuel requirements were
determined on the basis of make, model, model year, engine data provided
by the respondent, and, when available, were confirmed with engine data
obtained from the VIN.
1.4	Trucks in NPD
In any calendar year, nearly 2,000 light-duty trucks participate in the
NPD survey. Due to difficulties in determining truck fuel requirements,
only about 150 of these trucks may be positively identified as having
catalysts. Because of the small sample size for trucks, only cars were
*Statement from the letter mailed to each potential participant.
-3-

-------
included in this analysis. The exclusion of trucks must be kept in mind
when considering the results since the tampering rates for trucks have
been reported to be substantially different from those for cars.*
1.5	Non-Catalyst Cars With Unleaded Fuel Requirements
Certain vehicles are required to use unleaded fuel even though they do
not have a catalyst. Since it has been assumed that the ultimate use,
if any, of this analysis will be for estimating the effect of misfueling
on catalyst vehicles, a misfueling rate among catalyst vehicles is
sufficient. While it can be assumed that the misfueling rates among
non-catalyst unleaded cars are either the same or different, this issue
is not relevant. Misfueling among non-catalyst unleaded cars does
affect estimates of leaded gasoline consumption and lead emissions.
However, the number of such vehicles and their contribution to leaded
gasoline consumption and lead emissions is small and errors will be
small if equal misfueling rates are assumed. Furthermore, the data base
has few of these cars, so any separate estimates would have great
uncertainty.
1.6	Confidence Intervals
Selected tables in this analysis include a statistic termed "Estimated
Errors" to denote the reliability of reported misfueling rates. Inas-
much as the NPD data base is derived from a quota sample, it may be
argued that no statistic can reflect the "error of estimate" as applied
In the strict sense of a random sample. Nevertheless, it is important
to realize that estimates derived from a quota sample are subject to
variability and that the analyst must consider the variability of
*"Motor Vehicle Tampering Survey - 1982," U.S. Environmental Protection
Agency, Office of Enforcement and Legal Counsel, Publication No.
EPA-330/1-83-001, April 1983, p. 20.
-4-

-------
derived estimates in interpreting the findings. A more detailed expla-
nation of the derivation of these estimates may be found in Appendix 1
of the analysis.
2. VALIDITY OF THE DATA
Since the NFD data is from a diary panel survey it is important to
examine the make-up of the panel and to determine if observed trends in
vehicle use behavior are consistent with results from other surveys.
Tables 1 and 2* present some of the demographic distributions observed
in the NPD data base and compare them to distributions seen in the
Residential Energy Consumption Survey (RECS) and in the National Family
Opinion (NFO) gasoline diary survey. Table 3** presents a comparison of
Household Income distribution as observed in NPD and as reported by the
U.S. Bureau of the Census. Owing in large part to the exclusion of
singles from NPD, the demographic profile is not strictly representative
of the U.S. as a whole. This being the case, it is crucial to compare
trends in driving behavior in NPD with those seen in other sources.
A comparison of monthly trends in household Vehicle Miles of Travel
(VMT) is presented in Figure l.+ Notes on the pages following the
figure describe the sampling and estimation techniques used in each of
the studies. NPD is consistent with these other data sources with
*Fuel Purchasing Patterns and Vehicle Use Trends From the NPD Research
Gasoline Diary Data Base: Data Display, Energy and Environmental
Analysis, Inc., prepared for the U.S. Department of Energy, September
1982.
**Ibid.
+Ibid.
-5-

-------
TABLE 1
COMPOSITION AND LOCATION OF HOUSEHOLDS:
NPD VERSUS RECS AND NFO
(Percent)
Family Size
RECS
NFO
NPD
One
15.8
17.1
00.0
Two
36.6
38.5
42.7
Three
Four
18.1	16.1
16.8	16.2
20.5	23.3
Five or More
13.4
11.4
13.5
Number of Vehicles
RECS
NFO
NPD
One	Two	Three	Four or More
40.2	42.8	12.3	4.7
51.3	39.2	8.1	1.4
32.0	40.7	17.1	10.2
RECS
NFO
NPD
Census Regions
Northeast	North Central	South	West
21.0	27.2	31.9	19.9
21.3	29.2	31.7	17.8
20.4	26.8	32.2	20.6
-6-

-------
TABLE 2
AGE AND ECONOMIC STATUS OF HOUSEHOLDS:
NPD VERSUS RECS AND NFO
(Percent)
Age of Head of Household
RECS
NFO
NPD
Under 30
16.9
3.2
7.8
30-39
20.5
19.5
22.8
40-49
17.7
18.0
16.8
50 and Over
44.9
59.3
52.6
RECS
NFO
NPD
Own
74.9
84.5
87.5
Rent
23.8
13.4
11.2
Rent Free
1.3
0.5
1.3
Other
0.0
1.6
0.0
-7-

-------
lADLif. J
HOUSEHOLD INCOME (1981$)
NPD VERSUS U.S. CENSUS
(Percent)
Census*
NPD
Under
10,000
25.4
17.2
10,000
14,999
14.4
18.2
15,000
19,999
12.3
16.2
20,000
29,999
21.1
27.2
30,000
39,999
13.1
13.1
40,000
49,999
6.6
4.7
50,000
or More
7.1
3.3
*U.S. Bureau of the Census, Current Population Reports, Series P-60,
No. 134, "Money Income and Poverty Status of Families and Persons in
the United States: 1981 (Advance Data from the March 1982 Current
Population Survey)," U.S. Government Printing Office, Washington, D.C.,
1982.
Note: NPD households include families only, while U.S. Census data uses
a broader definition that includes singles.
-8-

-------
FIGURE 1
MONTHLY TRENDS IN HOUSEHOLD VMT
HUES PER MONTH

-------
NOTES FOR FIGURE 1
COMPARABILITY OF HOUSEHOLD VMT SOURCES
•	Data Sources
-	Energy Information Administration: Residential Energy Con-
sumption Survey Household Transportation Panel (June 1979 to
September 1981)
-	Federal Highway Administration (FHWA): Traffic Volume Trends
(April 1978 to December 1981)
-	National Family Opinion Poll: NFO/Auto-Facts Gasoline Diary
Panel (May 1978 to February 1981)
-	NPD Research, Inc.: Petroleum Marketing Index Diary Panel
(August 1978 to December 1981)
•	Sampling Techniques
-	RECS: Systematic random sample of households (includes single
people and unrelated persons sharing a dwelling)
-	FHWA: Does not sample individual vehicles
-	NFO/Auto-Facts: Quota sample survey of households (includes
single people and unrelated persons sharing a dwelling)
-	NPD Research: Quota sample survey of families (single people
and unrelated persons sharing a dwelling are not included)
•	Estimation Techniques
-	RECS: Odometer readings — data weighted to national level on
the basis of demographic characteristics of household
-	FHWA: City and highway traffic flow counts conducted by State
highway departments — estimate of total travel scaled down by
g
factor of 10 for directional trend comparison to household
estimates; Census data show approximately 68 million vehicle-
operating households in the U.S.
-	NFO/Auto-Facts: Odometer readings — data are sample-
weighted.	'
-	NPD Research: Odometer readings — data weighted to national
level on the basis of demographic characteristics* of household
•	Coverage
-	RECS, NFO/Auto-Facts, NPD Research: Report on all vehicles
driven (owned/operated) by a household
-	FHWA: Includes trucking and commercial travel
-10-

-------
respect to monthly trends in household VMT. Further evidence of NPD's
consistency may be found in comparisons of annual vehicle miles of
travel by vehicle age.
Figures 2 and 3 plot the relationship between annual VMT (Vehicle Miles
of Travel) and vehicle age for cars and light-duty trucks. The figures
show the well-known trend in decreasing VMT with age as observed in NPD
and three other surveys. The other surveys are the Nationwide Personal
Transportation Study (NPTS) conducted by the U.S. Department of Trans-
portation, Office of Highway Planning; the Residential Energy Consump-
tion Survey (RECS) conducted by the Energy Information Administration;
and the NFO/Auto-Facts (NFO) national panel diary survey conducted by
Auto-Facts, Inc. As can be seen from the two plots, the vehicle-age
dependent declines in VMT found in NPD are consistent with those report-
ed by RECS and NFO. The NPTS survey, which consistently reports higher
annual VMT than the other data sources, was collected several years
earlier than the other studies. Furthermore, NPTS respondents were
simply asked to recollect their prior year's mileage accumulation; no
effort was made to corroborate the response with odometer records. The
RECS survey, which did ask for odometer readings, is a systematic random
sample of households, including single people and unrelated persons
sharing a dwelling. Since the NPD results are not markedly different
from those in RECS, it appears that NPD does not have a serious non-
response bias vis-a-vis a random sample with respect to vehicle travel
characteristics. In addition, the exclusion of singles seems to have
little effect on observed aggregate vehicle use behavior.
The preponderance of evidence suggests that NPD is valid and appropriate
for studies of vehicle use behavior in the U.S. Besides having proven
itself to be reliable, NPD is also the only currently available source
of extensive time series data for the U.S. personal transportation
-11-

-------
19119 ~
I
18999 ~
I
17999 ~
1*999 ~
I
15999 ~
I
14999 ~
I
13999 ~
I
12999 ~
I
11199 ~
I
19999 ~
I
N3
I
9999 ~
I
8999 ~
I
7991 ~
I
*999 ~
I
3999 ~
I
4999 ~
I
3999 ~
I
2999 ~
FIGURE 2
ANNUAL VHT PER CAR BY AGE
1980 NFO
LEGEND,


I-	% =
1969
NPTS
S—— & =
: 1980
NFO
$	$ =
1980
RECS
#~—-# =
= 1981
RECS

= 1977
NPTS
+ —+ =
= 1981
NPD
2 3
4
5
1969 NPTS
1980 RECS
1981 NPD
19
AGE
11
12 13 14 15 1* 17 18 19

-------
FIGURE 3
17HI ~
ANNUAL vw per light truck by age
lilll ~
I
15111 ~
1MH !
1»U ~
12111 1
mil 1
I
urn ~
I
9111 ~
Sill ~
I
7111 ~
ill! ~
I
SIM *
I
4111 ~
I
3111 ~
I
2111 ~
I
111! ~
I
I ~
1980	t
1981	tl
LEGEND
&	& =
1980
NFO

«	ft =
1981
RECS

II


1980
RECS
* 1980 RECS
+ + =
1981
HPD

9
AGE
II
II
12
IS 14 15 14
17

-------
fleet. Vehicles typically stay in the panel for 10 to 12 months, thus
making it possible to take a detailed look at the behavior of many
individual vehicle owners over an extended period.
3. ASSUMPTIONS PERTAINING TO THIS ANALYSIS
In processing the NPD data for this analysis, assumptions have been made
about unknown fuel type purchases and about reporting errors. These
assumptions are discussed in this section.
3.1 Unknown Fuel Purchase Assumption
In approximately 1.6 percent of all purchases reported by the catalyst-
equipped cars in NPD, the respondent failed to report whether the fuel
purchase was leaded or unleaded. For purposes of assigning misfueling
involvement categories the unknown fuel volume is divided between leaded
and unleaded fuel on the basis of the ratio between known leaded volume
and known unleaded volume for the vehicle making the unknown purchase.
Assumptions that unknown is always unleaded or always leaded have also
been examined to determine the sensitivity to this approach. The
resulting.involvement rates are displayed in both the 1981 and 1982
sections of this analysis.*
For purposes of determining the maximum number of successive leaded
purchases, unknown purchases are treated as if they were unleaded. It
is not practical to randomly assign individual purchases as either
leaded or unleaded since on average a vehicle reports only 0.7 unknown
purchases during a year. In order to provide bounds for the effect of
*The results may be found in Table 4 "Comparison of Involvement Under
Alternate Assumptions About Unknown Fuel Type." This" table is Includ-
ed in both the 1981 and the 1982 sections.
-14-

-------
this methodology two alternate cases have been tested. The first case
assumes unknown purchases are unleaded; the second case assumes unknown
purchases are leaded. Table 4 compares these two cases. Percentages on
the diagonal represent vehicles not affected by the manner in which
unknown fuel is classified. Percentages to the right of the diagonal
represent vehicles that would move into a higher successive purchase
category if unknown fuel were assumed to be leaded. When the maximum
number of successive leaded purchases is at least three, assuming
unknown fuel to be leaded has little effect on the distribution of
vehicles.
In summary, the unknown fuel type volume is assigned to leaded or
unleaded but no attempt is made to correct each individual purchase. If
individual purchases were reassigned, the effect on the repeated mis-
fueling statistics is not expected to be large.
3.2 Reporting Errors Assumption
In any large data collection effort there is a potential for recording
or transcribing errors.* To avoid overstating the incidence of misfuel-
ing, only those vehicles recording at least three leaded purchases
during the time they were in the sample are counted as misfuelers. If
no more than two leaded purchases are reported, a data error is assumed
and the fuel type designator is changed to unleaded.
The rationale for this screening criterion is based on the fact that
even with a probability of reporting error as low as I percent, there is
about one chance in three that one or two misfuelings would be reported
*H.T. McAdams, Analysis Memorandum to R. Dulla (EEA), "Reporting Errors
in Fuel Purchase Records," under Letter of Agreement No. 026003-1,
Contract No. B-F6895-AZ, February .17 and 20, 1984.
-15-

-------
TABLE 4
COMPARISON OF ALTERNATE ASSUMPTIONS ABOUT UNKOWN FUEL TYPE
PURCHASES USING 1982 NPD DATA
(Using Weighted Data)
Maximum Number
of Successive
Leaded Purchases
Assuming That
Unknown is Unleaded
0
1
2
3
4
5
6	or more
Number of
1982 Vehicles
So Classified
3,050
10
34
106
100
55
339
Classification Under Alternate Assumptions
(Unknown = Leaded) (Percent of Vehicles)
1
74.6
11.0
90.4
2
3
4
5
6
7.6
2.4
1.3
0.8
2.2
9.6
0.0
0.0
0.0
0.0
81.5
4.4
4.3
0.0
9.7

87.7
3.3
3.0
6.0


98; 4
1.3
0.4



97.6
2.4




100.0

-------
during the course of the survey, even though no leaded purchases were
actually made. On the other hand, actual misfuelers who misfuel to a
significant degree would seldom report as few as two mlsfuellngs during
the survey. Thus, the rule is structured to strike a compromise between
the two types of errors to provide a refined estimate of the actual
vehicle involvement rate in the context of the study.*
Table 5 displays the effect of this methodology on the vehicle involve-
ment rate using data from 1982.** Using the criteria, the overall
vehicle involvement rate is 18.0 percent. Without the criteria, taking
all leaded designations at face value, the involvement rate is 30.0
percent. Table 6 displays the effect of the methodology on the amount
of leaded fuel purchased by the catalyst car fleet. Using the criteria,
7.7 percent of the fuel purchased by the catalyst car fleet was leaded.
Without that criteria, leaded fuel purchase volume rises to 8.1 percent
of the fuel bought by the fleet. Thus, although the methodology reduces
the apparent vehicle involvement in misfueling by 40 percent, the change
in fuel volume is less than 5 percent.
The criteria results in a conservative lower bound estimate for vehicle
involvement and has very little effect on the reported volume of mis-
fueling. An 18 percent vehicle involvement rate, although a lower
bound, is not insignificant. By comparison, the 1982 EPA tampering
survey reports that 10.58 percent of vehicles sampled show at least one
*H.T. McAdams, Analysis Memorandum to R. Dulla (EEA), "Vehicle Involve-
ment Rate and Its Dependence on Sample Size," under Letter of Agree-
ment No. 026003-1, Contract No. B-F6895-A-Z, March 30, 1984.
**A total of 445 cars meet the maximum-of-two leaded purchase criteria
in 1982.
-17-

-------
TABLE 5
COMPARISON OF VEHICLE INVOLVEMENT IN MISFUELING IN 1982
RESET MAXIMUM-OF-TWO LEADED PURCHASES TO UNLEADED VERSUS NO RESET
Vehicle Involvement Misfueling Rate
Reset		No Reset
Model
Year
Number of
Vehicles
Rate
(% of Fleet)
Estimated
Error
Rate
(Z of Fleet)
Estimated
Error
1975
280
25.0
5.1
37.2
5.7
1976
465
22.4
3.8
32.1
4.2
1977
586
22.0
3.4
32.3
3.8
1978
632
" 17.1
2.9
29.1
3.5
1979
590
14.8
2.9
26.9
3.6
1980
450
16.0
3.4
26.3
4.1
1981
485
13.0
3.0
26.9
4.0
1982
185
16.6
5.4
39.6
7.1
1983
21
16.3
15.8
23.2
18.1
Overall
3,694
18.0
1.2
30.0
1.5
-18-

-------
TABLE 6
COMPARISON OF VEHICLE INVOLVEMENT IN MISFUELING IN 1982
RESET MAXIMUM-OF-TWO LEADED TO UNLEADED VERSUS NO RESET
Leaded Fuel Purchased as Percent of Total
Reset
No Reset
Model
Year
1975
1976
1977
1978
1979
1980
1981
1982
1983
Overall
Number of
Vehicles
280
465
586
632
590
450
485
185
21
3,694
Percent Leaded
15.3	(4.2)*
11.0	(2.9)
13.4	(2.8)
7.4	(2.0)
4.8	(1.7)
4.2	(1.9)
1.6	(1.1)
2.2	(2.1)
3.5	(7.8)
7.7	(0.9)
Percent Leaded
15.7	(4.3)
11.3	(2.9)
13.7	(2.8)
7.8	(2.1)
5.2	(1.8)
4.5	(1.9)
2.0	(1.3)
3.0	(2.5)
4.5	(8.9)
8.1	(0.9)
*Values In parentheses are estimated errors.
-19-

-------
positive indication of misfueling.* Since the EPA survey is a random
sample and does include trucks, this comparison to NFD is necessarily
approximate. However, since EPA reports a higher tampering rate for
trucks than for cars a combined sample might be expected to yield a
higher rate than for cars alone. A key to the results might be found in
the different manner in which the two surveys collected information.
The NPD data was collected over a long period of time from individuals
who thought they were simply providing marketing information. The EPA
data is a compilation of single observations on a random selection of
vehicles. One advantage to the EPA method is that classification of a
misfueler is based on a direct examination of each vehicle by the survey
team. The examination includes a Plumbtesmo test for lead in the
exhaust pipe, a check of the filler neck restrictor to see if it has
been tampered with, and chemical analysis of a gasoline sample to see if
lead in the gas tank is above a threshold of 0.05 grams per gallon.
There is very little chance that a regular misfueler could escape
detection. At the same time, there is some chance that an infrequent
misfueler might be overlooked. For example, an individual purchasing
leaded gasoline every five or six tankfuls, who uses a funnel to bypass
the filler neck restrictor, would show no obvious tampering and might
easily have less than 0.05 grams of lead per gallon of fuel in the tank
at the time of survey. Furthermore, as noted in the tampering survey, a
hastily field-administered Plumbtesmo tailpipe test is unreliable when
negative.** Thus, while a positive Plumbtesmo test is reliable evidence
of lead in the tailpipe, a negative test means only that lead was not
detected — the possibility remains that a repeat test under more ideal
*Motor Vehicle Tampering Survey - 1982, U.S. Environmental Protection
Agency, Publication No. EPA-330/1-83-001, April 1983, p. 28.
**Ibid.
-20-

-------
circumstances would yield a positive result. An additional negative
bias is associated with the non-compulsory nature of the survey. Since
the EPA survey is openly conducted for a government agency, misfuelers
may be extremely hesitant to participate.
In conclusion, while it is difficult to make an exact comparison between
NPD and EPA results, each has sources of downward bias and each has
strong points. EPA uses a random sample and, through actual examination
of the vehicles, has a high probability of identifying regular misfuel-
ers. NPD samples a wider geographic range, including rural areas, and
provides demographic information, as well as detailed time-series
purchase data. Preference for one type of survey over another is
ultimately dependent upon the analysis to be performed and it is the
analyst's responsibility to judge the suitability of a particular data
base to the task at hand.
-21-

-------
PART I:
MISFUELING OF LIGHT DUTY VEHICLES
IN 1981:
INTERIM REPORT
EPA Prime Contract 68-01-6558
Subcontract 130.109
Work Assignment No. 30, Task 4
Prepared for:
ENVIRONMENTAL PROTECTION AGENCY
Motor Vehicle Emissions Laboratory
Ann Arbor, Michigan
Prepared by:
ENERGY AND ENVIRONMENTAL ANALYSIS, INC.
1655 North Fort Myer Drive
Arlington, Virginia 22209
September 1984

-------
LIST OF TABLES
Page
1	Highlights of Misfueling Frequency Distributions 			6
2	Distribution of Misfueling By Degree of
Involvement Aggregate Across All Model Years 		8
3	Comparison of Weighting Methods 				10
A Comparison of Involvement Under Alternate
Assumptions About Unknown Fuel Type 					12
5	Vehicle Involvement Misfueling Rates by Model Year 		14
6	Repeated Misfueling 			16
7	Length of Survey Participation 		18
8	The Distribution of Successive Misfueling Purchases
Cross-Tabulated With The Distribution of Misfueling
By Degree of Involvement 				20

-------
SUMMARY
The purpose of this task Is to quantify the rate of misfueling by light-
duty vehicle (LDV) owners/operators in 1981. At the direction of EPA,
light-duty trucks are specifically excluded from the analysis. This
report presents highlights of the misfueling trends observed in a time
series data base of fuel purchasing behavior. The survey information is
derived from the NPD Petroleum Marketing Index (NPD), a diary panel
survey of over 5,000 households conducted by NPD Research Inc. The
tables and accompanying descriptive notes in this paper present findings
without any attempt at interpretation.
lisfueling may be measured in a variety of ways; the appropriate method
lepends upon the questions to be answered. In this study the misfueling
rate is measured by fleet involvement, i.e., the proportion of all
atalyst vehicles which are misfueled. Vehicles are categorized on the
asis of whether or not they are ever misfueled, on the ratio of leaded
uel purchased to total fuel purchased, and on the maximum number of
uccessive leaded purchases made during the survey period. Most of the
ables presented here are aggregated across all model years. An appen-
ix to this report contains a computer printout with more detailed
esults on specific model years.
nis study is not intended to answer questions about why people misfuel.
ither, it quantifies the behavior observed in a representative sample
E individuals during the course of a recent calendar year. Most
revious studies of misfueling have sampled a cross-section of the
:hicle population at one point in time. A major advantage of using a
.ary panel survey Is that individuals may be followed through time.
-1-

-------
Such a survey provides more complete information about the frequency of
misfueling and could allow for detailed studies of the demographic
characteristics of misfuelers or motivational factors.
The most common reservation about the use of diary surveys is that they
depend upon consistent and truthful self-reporting. Despite concerns
about respondents' potential unwillingness to Incriminate themselves,
the panel participants were quite open about their purchasing behavior
and freely indicated the purchase of leaded fuel. The participants know
their responses are being collected for gasoline brand market share
studies and they are accustomed to reporting detailed Information about
what they have purchased. The participants do not know that government
agencies purchase the raw survey data for studies such as this one.
Furthermore, the participants are guaranteed anonymity by NPD Research,
Inc. when they agree to participate.
DATA BASE PREPARATION
The NPD data base contains fuel purchase histories for over 12,000
privately operated vehicles. The data, collected during 1981, contain
detailed information about fuel purchases, including date, gallons, type
of fuel, and total cost. An example of the purchase logs filled in by
respondents is shown in Appendix 2. There is body style/engine informa-
tion as well as household demographic data associated with each vehicle
purchase history. This data base has been used extensively by the U.S.
Department of Energy (DOE) to examine trends in fuel consumption,
on-road fuel economy, and vehicle miles of travel. During the course of
this previous work the data were cleaned and established as a SAS data
-2-

-------
set. As part of this work for DOE, the engine description information
provided by the survey respondent was verified (and corrected when
necessary using information extracted from the vehicle identification
number (VIN).* Based on this engine information, each vehicle has bee
classified as to whether or not it has a catalyst.
The only additional processing of the data in the current work was to
re-examine and verify existing catalyst information. Identification o;
a catalyst equipped vehicle is made on the basis of the VIN-augmented
data for make/model, model year, CID, number of cylinders, fuel system
and type of transmission.
As noted previously, this task is based on a study of the 3353 catalyst
equipped LDVs from NPD. A small number of LDVs manufactured during the
late seventies required unleaded fuel but were not actually equipped
with a catalyst. We have assumed that the ultimate use, if any, of thi
analysis will be for estimating the effect of misfueling on catalyst
vehicles. For this purpose, a misfueling rate among catalyst vehicles
is sufficient. Therefore, those vehicles not having a catalyst but
requiring unleaded were not included in the study. Another group of
LDVs excluded from the study were those participating for less than two
nonths. For the most part, respondents with only one month of partici-
)ation have very poor record-keeping practices and incomplete purchase
listories. Frequently only one or two purchases are reported and
.ypically consist mostly of missing information. A total of 127 vehi-
cles were deleted from the survey for participating less than two
tonths. An additional three vehicles were deleted because none of theii
urchases were identified as either leaded or unleaded.
The VIN is reported by the owner, along with the engine description
information, when a vehicle first enters the survey.
-3-

-------
In any large data collection effort there is a potential for recording
or transcribing errors. To avoid over-reporting the incidence of
misfueling, only those vehicles recording at least three leaded pur-
chases during the year are counted as misfuelers. If no more than two
leaded purchases are reported, a data error is assumed and the fuel type
designation is changed to unleaded. A total of 524 vehicles meet this
maximum-of-tvo leaded purchases criteria. A total of 1,110 "purchases"
showing' 0.0 gallons of fuel bought were deleted. In general, these
records are null entries representing months when a diary was returned
but no fuel was purchased.
The tables in this report highlight misfueling behavior in the catalyst
fleet. They also provide information pertaining to the manner in which
the data are weighted and to the way in which unknown fuel type pur-
chases are handled. Each table is prefaced with explanatory notes to
assist the reader in interpreting the information presented.
-4-

-------
Table 1 - Highlights of Misfueling Frequency Distributions
•	Data in this table are weighted on the basis of the NPD
projection factors. For information on the effect of alterna-
tive weighting methods, see Table 3.
•	Total fuel purchased includes leaded, unleaded, and type
unknown. For more detail on the treatment of purchases with
unknown fuel type, and its effect on the findings, see Table
4.
•	Leaded fuel as a percent of total fuel purchased by the
catalyst fleet measures misfueling on a gallons purchased
basis.
•	The leaded fuel under 11 percent and 91-100 percent of total
fuel purchased categories measure misfueling on a vehicle
basis. Each vehicle's degree of involvement is judged on the
basis of how much of their purchase volume is leaded. The
vehicle is then assigned to an appropriate category. So, for
example, 5.6 percent of the vehicles in the catalyst fleet
were misfuelers whose leaded purchases amounted to less than
11 percent of the fuel they purchased during the year. By
comparison, 3.5 percent of the vehicles in the catalyst fleet
purchased 91 to 100 percent leaded fuel by volume.
•	Catalyst fleet involvement in misfueling includes any catalyst
vehicle that ever purchased leaded fuel, regardless of quanti-
ty or percentage of total fuel purchased over the year.
•	There is a small number of vehicles (0.2 percent) of the
catalyst fleet who purchased leaded fuel, but less than 10
gallons worth. These vehicles are included in the "Leaded
Fuel Under 11% of Total" category regardless of actual per-
centage.
-5-

-------
TABLE 1
HIGHLIGHTS OF MISFUELING FREQUENCY DISTRIBUTIONS
Number of Catalyst Equipped Vehicles	3353
Leaded Fuel as Percent of Total Fuel Purchased	6.4 (0.8)*
by the Catalyst Fleet
Leaded Fuel Under 11% of Total Fuel Purchased by	5.6 (0.8)
Vehicle (percent of catalyst fleet)
Leaded Fuel 91-100% of Total Fuel Purchased by	3.5 (0.6)
Vehicle (percent of catalyst fleet)
Catalyst Fleet Involvement in Misfueling	14.3 (1.2)
(percent of catalyst fleet)
Purchased at least 10 gallons of Leaded Fuel	14.5 (1.2)
(percent of catalyst fleet)
Purchased leaded fuel at least 3 times from January	3.0 (0.6)
to June 1981 and not at all from July to December
1981 (percent of catalyst fleet)
Purchased leaded fuel at least 3 times from January 10.0 (1.0)
to June 1981 (percent of catalyst fleet)
*Values in parentheses are estimated errors in the percent of catalyst
fleet.
-6-

-------
Table 2 - Distribution of Misfuellng by Degree of Involvement
•	In the table, vehicle involvement in misfuellng is measured by
the ratio of leaded fuel to total fuel purchased. A vehicle
purchasing a total of 400 gallons of fuel (all types) during
1981, of which 30 gallons are leaded, has a ratio of 30 to 400
or 7.5 percent. This vehicle is placed in the under 11
percent leaded category. Had the same vehicle purchased 350
gallons of leaded, out of 400 gallons total, the ratio would
be 87.5 percent leaded and the vehicle would be placed in the
81-90 percent leaded category.
•	5.7 percent of the catalyst fleet, or 39.6 percent of the
misfuelers, have leaded fuel purchases totaling less than 11
percent of their annual fuel purchases. By comparison, 3.5
percent of the fleet, or 24.3 percent of misfuelers, purchased
91-100 percent leaded fuel by volume.
-7-

-------
TABLE 2
DISTRIBUTION OF MISFUELING
BY DEGREE OF INVOLVEMENT
Aggregated Across All Model Years
Leaded Fuel Purchased
by Vehicle as Percent
of Total Fuel Purchased
Percent of Catalyst
Fleet In Category
Estimated
Error
Percent of
Mlsfuellng
Vehicles
Under 11
5.7
0.8
39.6
11-20
2.2
0.5
15.3
21-30
0.6
0.3
4.2
31-40
0.3
0.2
2.1
41-50
0.3
0.2
2.1
51-60
0.2
0.2
1.4
61-70
0.3
0.2
2.1
71-80
0.8
0.3
5.6
81-90
0.5
0.3
3.5
91-100
3.5
0.6
24.3
-8-

-------
Table 3 - Comparison of Weighting Methods
•	Each vehicle contributes one observation to the misfueling
analysis. When calculating the overall misfueling rate it is
helpful if an individual vehicle's contribution can be weight-
ed to account for its importance relative to other vehicles in
the fleet. This table compares three methods of weighting.
•	Sample weighting, with each vehicle assigned a weight of one,
does not distinguish among vehicles.
•	The NPD projection factor weights are assigned to each house-
hold on a monthly basis by NPD Inc. The factors are designed
to weight the sample, demographically, to the national level
based on income, race, region, and the educational level and
occupation of the female head of house. As respondents enter
and leave the survey, each household projection factor is
adjusted to maintain the national level weighting scheme. The
weight used in this study is the sum of these factors over
each month a vehicle participates in the survey.
•	The Months in Survey method assigns a weight to each vehicle
solely on the basis of the number of months a vehicle partici-
pates in the survey.
•	Overall the three weighting methods produce similar results,
although on a model year specific basis there are some differ-
ences. This is particularly true for model year 1982 where
small sample size is a problem.
•	Since the NPD projection factors were designed to weight the
survey to a national level on the basis of household demo-
graphics, these factors are used in reporting all results
except those in this table.

-------
TABLE 3
COMPARISON OF WEIGHTING METHODS
Vehicle Count
NPD
Vehicle Involvement
Misfueling Rate (percent)
NPD
Model
Year
Sample Projection
Height Factors
Months
in
Survey
Sample Estimated Projection
Weight Error* Factors
Mon ths
in
Survey
1975
266
366,200
2,664
21.1
4.9
20.7
21.4
1976
444
613,373
4,642
17.3
3.5
16.9
17.1
1977
598
835,144
6,056
16.4
3.0
17.1
16.5
1978
631
841,370
6,380
13.9
2.7
13.0
14.0
1979
520
714,709
5,246
12.7
2.9
12.6
13.0
1980
485
663,866
4,848
10.3
2.7
9.0
9.9
1981
387
402,127
2,649
12.9
3.3
12.7
14.4
1982
22
10,011
75
18.2
16.1
16.4
16.0
TOTAL
3,353
4,446,800
32,560
14.6
1.2
14.3
14.8
^Estimated errors would all be based on the unweighted vehicle count,
hence they would be nearly identical across the weighting methods. In
order to simplify comparisons of the fleet involvement percentages the
error estimates have been included only for the sample weight
calculation.
-10-

-------
Table 4 - Comparison of Involvement Under Alternate Assumptions About
Unknown Fuel Type
•	This table compares three methods of treating unknown fuel
types. An unknown fuel type purchase is one in which the
respondent has failed to check either the leaded or the
unleaded column on the monthly diary log.
•	If unknown fuel type is assumed to be leaded fuel, the overall
fleet involvement in misfueling is 42.0 percent of the cata-
lyst fleet.
•	If unknown fuel type is assumed to be unleaded fuel, the
overall fleet involvement in misfueling is 14.3 percent of the
catalyst fleet.
•	If the unknown fuel for each vehicle is allocated between
leaded and unleaded, based on the percentages of known leaded
and known unleaded bought for that vehicle, the overall fleet
involvement in misfueling is 14.3 percent of the catalyst
fleet.
•	On average, each vehicle in the survey made 1.3 purchases of
unknown fuel type during 1981. This fuel, roughly 12.6
gallons per vehicle, typically represents approximately 2.3
percent of the year's total fuel purchases. Including all of
these purchases in the leaded category increases fleet in-
volvement in misfueling by 190 percent.
•	The assumption that unknown fuel purchases actually represent
leaded fuel is made to test the belief that consumers do not
wish to implicate themselves in misfueling. While there may
be a handful of respondents whose behavior fits this pattern,
most individuals appear to be extremely forthcoming about
their misfueling habits. Given the wide distribution of
unknown fuel type purchases and people's willingness to report
buying leaded fuel, it is likely that most, though not all,
unknown fuel purchases are the result of recording error
rather than of half-hearted deception.
•	The assumption that unknown fuel is unleaded is the most
conservative method of allocating unknown fuel. As may be
seen in the table, the results are almost identical to those
obtained by allocating the unknown fuel between leaded and
unleaded.
•	All of the tables in this report are based on the assumption
that unknown fuel may reasonably be allocated between leaded
and unleaded on the basis of the percentages of known leaded
and known unleaded bought for an individual vehicle.
-11-

-------
TABLE 4
COMPARISON OF INVOLVEMENT UNDER ALTERNATE
ASSUMPTIONS ABOUT UNKNOWN FUEL TYPE
	Vehicle Involvement Misfueling Rate	
Unknown is Unleaded	Unknown is Leaded	Unknown Allocated
Model
Year
Number of
Vehicles
Rate
(% of fleet)
Estimated
Error
Rate
(% of fleet)
Estimated
Error
Rate
(% of fleet)
Estimati
Error
1975
266
20.7
4.9
40.8
5.9
20.7
4.9
1976
444
16.9
3.5
39.1
4.5
16.9
3.5
1977
598
17.1
3.0
45.4
4.0
17.1
3.0
1978
631
13.0
2.6
45.7
3.9
13.0
2.6
1979
520
12.6
2.9
40.7
4.2
12.6
2.9
1980
485
9.0
2.5
42.5
4.4
9.0
2.6
1981
387
12.7
3.3
34.5
4.7
12.7
3.3
1982
22
16.4
15.5
45.3
20.8
16.4
15.5
OVERALL
3,353
14.3
1.2
42.0
1.7
14.3
1.2

-------
Table 5 - Misfueling Fleet Involvement by Model Year
•	Number of vehicles is a count of the actual, unweighted,
number of LDVs In each model year.
•	The categories reported here are identical to the third,
fourth and fifth items in Table 1. For example, for model
year 1975, of which there are 266 catalyst equipped LDVs in
the survey, 5.1 percent were misfuelers whose leaded purchases
amounted to less than 11 percent of the fuel they purchased
during the year. At the same time, 9.0 percent of the model
year vehicles purchased 91-100 percent leaded fuel by volume.
Overall, 20.7 percent of the model year 1975 vehicles mis-
fueled at least part of the time.
•	In general it is assumed that misfueling will increase with
vehicle age. With some slight deviations the data presented
in Table 5 shows misfueling involvement remaining relatively
flat for vehicles from model years 1978 through 1981. Vehi-
cles from model year 1977 and earlier show misfueling in-
creasing with age. Even for model years 1978 to 1981, the
percentage of vehicles in the 91-100 percent leaded category
increases with vehicle age. The involvement rate for model
year 1982 is suspect due to relatively small sample size.
-13-

-------
TABLE 5
VEHICLE INVOLVEMENT MISFUELING RATES BY MODEL YEAR
Model
Year
1975
1976
1977
1978
1979
1980
1981
1982
Number of
Vehicles
266
444
598
631
520
485
387
22
Fuel Under
11% Leaded
(% of Fleet)
5.1
4.9
7.2
5.0
5.4
4.7
7.5
0.0
Fuel 91-100%
Leaded
(% of Fleet)
9.0
4.5
5.1
3.6
2.0
0.7
0.0
10.8
Overall
Vehicle
Involvement
(% of Fleet)
20.7
16.9
17.1
13.0
12.6
9.0
12.7
16.4
Estimated
Error in
Involvement
4.9
3.5
3.0
2.6
2.9
2.5
3.3
15.5
Overall 3,353
5.6
3.5
14.3
1.2
-14-

-------
Table 6 - Repeated Misfueling
•	This table displays the Incidence of successive misfueling for
vehicles at each end of the age spectrum (model years 1975 and
1981), and for the catalyst fleet as a whole. Model year 1982
was excluded because of relatively small sample size. Vehicle
Involvement rates are percents of the model year fleets.
•	Vehicles having made at least two leaded purchases In a row
are assigned to one of five purchasing categories. The
assignment Is based on the longest string of leaded purchases
made by that vehicle during 1981.
•	Vehicles making only singleton purchases of leaded gasoline
will not appear in this table. Thus, although 20.7 percent of
the model year 1975 vehicles misfueled at least once (see
Table 5), only 20.6 percent (the sum of the five purchasing
categories for 1975 vehicles) of the fleet is represented in
Table 6. The remaining 0.1 percent of the fleet that mis-
fueled never purchased leaded twice in a row.
•	Percent of leaded purchases is calculated on a model year
specific basis. For example, 95.4 percent of the leaded
purchases made by catalyst equipped model year 1975 vehicles
were made by vehicles that have purchased leaded at least 6
times in a row.
•	Category assignments are exclusive. A vehicle making two
leaded purchases in a row on several occasions, and four
leaded purchases in a row on one occasion will be assigned
only to the category for vehicles having made four successive
leaded purchases.
-15-

-------
TABLE 6
REPEATED MISFUELING
Maximum Number
of Successive
Leaded Purchases
During the Year
2
3
4
5
Percent of
Model Year
Catalyst Fleet
Vehicles Involved
Estimated
Error
MY 75 Fleet
6 or more
0.8
0.7
2.2
0.4
16.5
1.0
1.0
1.8
0.7
4.5
MY 81 Fleet
2
3
4
5
6	or more
0.3
5.2
3.2
0.7
3.4
0.5
2.2
1.8
0.8
1.8
Overall Fleet
2	1.0	0.3
3	2.5	0.5
4	1.7	0.4
5	1.0	0.3
6	or more	7.8	0.9
Percent of
Leaded Purchases
By Fleet
0.8
0.5
2.7
0.4
95.4
1.1
24.7
22.4
4.5
47.4
1.2
2.5
3.2
2.1
90.4
-16-

-------
Table 7 - Length of Survey Participation
•	In order to be included in this misfueling study a vehicle
must have provided data for at least two months. Vehicles
reporting for only one month generally provide purchase
records with much missing or inconsistent information. In
order to reduce the effect of missing data, the minimum
reporting requirement was adopted. A total of 127 vehicles
were eliminated as a result of this requirement.
•	The majority of vehicles contributed a full 12 months of data.
-17-

-------
TABLE 7
LENGTH OF SURVEY PARTICIPATION
Number of Months In	Number of	Percent of
Survey During 1982	Vehicles	Catalyst Fleet
2	92	0.6
3	102	1.0
4	109	1.3
5	317	6.0
6	73	1.3
7	114	2.2
8	86	2.2
9	100	2.8
10	92	2.7
11	161	5.8
12	2,107	74.2
-18-

-------
Table 8 - The Distribution of Successive Misfueling Purchases
Cross-Tabulated with The Distribution of Misfueling by Degree
of Involvement
•	This table shows the degree of misfueling involvement,
measured on a volume basis as in Table 6, for the vehicles in
each of the successive misfuel purchasing categories shown in
Table 5.
•	PBMAX identifies the maximum number of successive leaded
purchases: 0 (includes single isolated purchases), 2-5, and 6
or more.
•	PBCAT identifies categories of leaded fuel as a percent of
total fuel purchased by a vehicle during 1981.
•	FREQUENCY is the NPD projection factor weighted vehicle count.
•	PERCENT is the percentage of the catalyst fleet.
•	Cumulative values are reported for both FREQUENCY and PERCENT.
•	Since only non-zero percentages are reported, some values of
PBCAT are not printed for some values of PBMAX.
•	Due to the labeling limitations of the SAS statistical package
the work "LEADED" is usally truncated to "LEA" in this table.
In a final indignity the word is further reduced to the letter
"L" for the under 11 percent category.
-19-

-------
TABLE 8
THE DISTRIBUTION OF SUCCESSIVE MISFUELING PURCHASES
CROSS-TABULATED WITH
THE DISTRIBUTION OF MISFUELING BY DEGREE OF INVOLVEMENT
PBMAX PBCAT
0	NO LEADED
1	FUEL UNDER 11% L
1	FUEL	11- 20% LEA
2	FUEL	UNDER 11% L
2 FUEL	11- 20% LEA
2 FUEL	21- 30% LEA
2 FUEL	31- 40% LEA
FUEL	UNDER 11% L
FUEL	11- 20% LEA
FUEL	21- 30% LEA
FUEL	31- 40% LEA
FUEL	41- 50% LEA
FUEL	71- 80% LEA
FUEL	81- 90% LEA
FUEL	91-100% LEA
FUEL	UNDER 11% L
FUEL	11- 20% LEA
FUEL	21- 30% LEA
FUEL	31- 40% LEA
FUEL	41- 50% LEA
FUEL	51- 60% LEA
FUEL	61- 70% LEA
FUEL	UNDER 11% L
FUEL	11- 20% LEA
FUEL	21- 30% LEA
FUEL	41- 50% LEA
FUEL	51- 60% LEA
FUEL	UNDER 11% L
FUEL	11- 20% LEA
FUEL	21- 30% LEA
FUEL	31- 40% LEA
FUEL	41- 50% LEA
FUEL	51- 60% LEA
FUEL	61- 70% LEA
FUEL	71- 80% LEA
FUEL	81- 90% LEA
FUEL	91-100% LEA
FREQUENCY CUM FREQ PERCENT CUM PERCENT
3812663
25085
1007
35199
8223
765
798
79856
16183
749
1106
180
' 234
1259
1533
49901
20905
6578
2640
1112
385
606
21903
14463
2068
566
3662
38369
35115
14669
10025
9521
6964
12030
35243
22776
152459
3812663
3837748
3838755
3873954
3882177
3882942
3883740
3963596
3979779
3980528
3981634
3981814
3982048
3983307
3984840
4034741
4055646
4062224
4064864
4065976
4066361
4066967
4088870
4103333
4105401
4105967
4109629
4147998
4183113
4197782
4207807
4217328
4224292
4236322
4271565
4294341
4446800
85.739
0.564
0.023
0.792
0.185
0.017
0.018
1.796
0.364
0.017
0.025
0.004
0.005
0.028
0.034
1.122
0.470
0.148
0.059
0.025
0.009
0.014
0.493
0.325
0.047
0.013
0.082
0.863
0.790
0.330
0.225
0.214
0.157
0.271
0.793
0.512
3.429
85.739
86.304
86.326
87.118
87.303
87.320
87.338
89.134
89.498
89.514
89.539
89.543
89.549
89.577
89.611
90.734
91.204
91.352
91.411
91.436
91.445
91.458
91.951
92.276
92.323
92.335
92.418
93.280
94.070
94.400
94.625
94.840
94.996
95.267
96.059
96.571
100.000
-20-

-------
APPENDIX 1
ESTIMATED ERRORS
Selected tables within the body of this report have included a statistic
termed "estimated error" to denote the reliability of key misfueling
rates. The purpose of this appendix is to discuss briefly the calcu-
lation of this quantity and the considerations that led to its use.
Inasmuch as the NPD data base is derived from a quota sample, it may be
justifiably argued that no statistic can reflect the "error of estimate"
as applied on the strict sense of a random sample. Nevertheless, it is
important to realize that estimates derived from a quota sample are
subject to variability and that, as a matter of pragmatism, the issues
of bias and variability should be decoupled. The analyst must exercise
due caution in selecting a quota sample, considering the purposes of the
study, comparison of sample composition and observables (estimates of
known quantities) with independent reference sources, and the avail-
ability of alternatives to the quota sample's use.
Given that the quota sample is accepted for the purposes at hand, the
analyst must consider the variability of derived estimates in interpret-
ing the findings. The estimated error statistic is used in this report
to reflect the variability of estimates in the sense described above.
This calculation follows that of a standard error of estimate derived
from a random sample. For sufficiently large samples of size N, the 95
percent confidence limit of an observed proportion p is given by:*
CI95 = ±1.96 / p(l-p)/N

-------
In this study, the proportions p are calculated as ratios of vehicles
falling within a defined misfueling category to the total number of
catalyst vehicles in the survey. Where noted, the proportions are
weighted by the NPD projection factors (thereby incorporating both
survey participation and control of the sample's demographic balance).
In all instances, the sample size is taken to be the (un-weighted)
number of catalyst vehicles in the sample.
The resulting estimated error is an approximation to the variability
that is present in sample estimates. A more exacting calculation would
need to consider the time-series nature of the data (i.e., extended
observations of vehicles across many purchases) and the implications of
weighting factors for determining the "effective" sample size. These
extended considerations are not germane, however, to the use of the
estimated errors as an order-of-magnitude guideline to estimate
variability.
*Engineering Statistics (Second Edition) by Albert H. Bowker and Gerald
Lieberman, Prentice-Hall Inc., 1972, pp. 466-467

-------
APPENDIX 2
This appendix contains an example of the purchase logs filled in by NPD
panel participants.
-23-

-------
I
to
«p-
I
AMERICAN SHOPPERS PANEL
P.O. BOX 5401, NEW HYOE PARK, NY 11040
INSTRUCTIONS
1.	Slip this diary in the visor holder you attach to the
sun visor.
2.	Writ! in ODOMETER (Speedometer) READING
whan you racaiva this diary (but no soonar than
tha date indicated) and again before you return
it Thare is space for thesa readings above.
NOTE: If you replace this vthide. be sure to
write in Odometer readings for BOTH vehicles
3.	Enter all GASOLINE/MOTOR FUEL PURCHASES
on tha ravena side of this diary.
4.	Enter ALL OIL AND ANTIFREEZE PURCHASES
on this side of tha diary. Make sure purchases at
wtvitas stations are included.
5.	Be wre to tad me all the necessary details if an-
other vehide is edded (see back of return envelope).
Tell ma about the change i.i this vahide (sold, no
longer used, disposed of. treded-in, etc.) to tha right.
6.	All your diaries are important, so please be sure to
mail this diary even if no gasoline or oil was pur-
chased this month.
Was the vehicle described on the ID label (top left) SOLD. RETIRED FROM USE, TRAOEO-IN
or otherwise disposed of during tha month?	DYES ~ NO
If YES. and vehicle was TRADEO-IN or OTHERWISE REPLACEO, fill in below for new vatiida:
If NO.please do not fill in below.
MAKE 	—	WHAT TRANSMISSION DOES IT HAVE? M On.
(ftwoln,Ford. ToyaU, Dodji,Bit)	.	.
(Z)3 Spatd Mtmial CJFourSgytdMtnud OS Spttd ftUmui
—;		~Automatic OAutomitit ml Overdriva
OMta, a 2ia CmlM. flittu. TR I. Mel
HOOEl YEAR tfl		BWBER OF CYLINDERS M One
~Four OFme OSix GEi(ht ~Rotvy
BOOYSTYLE	 nm U\n
U*.mtm.4«.Kdw. 1 hadiap.kttcktack. w.u.twn	Bl" <*'un8
—y. pit* «p. comma*. umpm. «t.i	LJTurbochan^d
OOES THIS VEHICLE	P )nt«^d
REQUIRE UNLEAOEO FUEL? (4 0n«	HoD	URejdator Catburitor
REQUIRE PIESEl FUEL? M On# V« ~ NoD VEHICLE MAINTENANCE-DOES FAMILY MEMBER	
HAVE AIR-CONDITIONING? (4 On. Y„D	CHANGE THE OIL? M On#	*,~ N.O
VEHICLE IDENTIFICATION NUMBER to*	TIMK-lIP THE CART M Ooa Y« ~ No ~
H M I I H I I I I I I I H I . TftWilEU	 SEX .~ FD
ENGINE DISPLACEMENT	(. teofatahKiw. uimi WHICH CREDIT CARDS OOES THIS DRIVER CARRY?
|l]lM Off	n
Suniber of Cubic Indies "	jfumbm^TUteri

-------
GASOLINE/MOTOR FUEL PURCHASES	AMERICAN SHOPPERS PANEL	Enter OIL and ANTIFREEZE (Coolant) Purchases on reverse lidi of card.
GAS
COPY THIS INFORMATION DIRECTLY FROM GAS PUMP
METHOD OF PAYMENT
DIO STATION
HAVE	
l^lal! (hit apply
DID YOU USE..
(~) til that apply
DIO
YOU
FILL
UP
THE
TANK?
(/) one
SPECIAL OFFER
WAS
PUR-
CHASE
MTHIN
IDLES
OF
YOUR
HOME?
M«»e
NAME OF
BRAND
NAME OF
GRAOE
IS THIS
GAS...-
(~1 one
NUMBER OF
GALLONS/
LITERS
PURCHASED

TOTAL
AMOUNT
Wss then
a special
offer?
M one
IF YES, describe
offer such as:
Price-Off, Free
Gift (describe
Gift), Discount
for Cash, etc.
DATE
BUYER
Such ai:
Shell.
Texaco,
Exxon,
etc.
Such as:
Regular,
Premium,
Super,
Diesel,
Gasohol, etc.
(/) one
PAID
(gasoline
purchase
only)
U
01
«e
If credit card
was used—
NAME OF
CREDIT
CARD
UJ
u
>
GC
UJ
Ik
-
K
UJ
Mm
Si
UJ3
MO.
z
3
£
ae
<
o
I5
Aw
it
u a
u
LJ
>
E
UJ
33
u. b
UJ
u
>
oc
ui
MM
5 a
SS
X
2
*
c
<
u
|i
II
tao
Year
of
bnth
19L_
Sat
o
UJ
o
<
UJ
.J
o
UJ
o
<
UJ
z
3
CO
z
o
<
(9

OB
UJ
CAR WASH
ii
i|
SI
ISO
Year
of
tilth
la-
ss
LEAOEO
UNLEADED
GALLONS
LITERS
.18
-* .C —.
Ill
ICASH/CHECI
CREDIT CARD
Zuj
Oc
55
Gal/Lt.
lUths
$
4
VES
NO
YES
NO
write in
YES
NO

-------
PART II:
MISFUELING OF LIGHT DUTY VEHICLES
IN 1982:
INTERIM REPORT
EPA Prime Contract 68-01-6558
Subcontract 130.109
Work Assignment No. 30A, Task 2
Prepared for:
ENVIRONMENTAL PROTECTION AGENCY
Motor Vehicle Emissions Laboratory
Ann Arbor, Michigan
Prepared by:
ENERGY AND ENVIRONMENTAL ANALYSIS, INC.
1655 North Fort Myer Drive
Arlington, Virginia 22209
September 1984

-------
LIST OF TABLES
Page
1	Highlights of Misfueling Frequency Distributions 		6
2	Distribution of Misfueling By Degree of
Involvement Aggregate Across All Model Years 		8
3	Comparison of Weighting Methods 		10
4	Comparison of Involvement Under Alternate
Assumptions About Unknown Fuel Type 		12
5	Vehicl Involvement Misfueling Rates by Model Year 		14
6	Repeated Misfueling 		16
7	" Length of Survey Participation 		18
8	The Distribution of Successive Misfueling Purchases
Cross-Tabulated With The Distribution of Misfueling
By Degree of Involvement 				20

-------
SUMMARY
The purpose of this task is to quantify the rate of misfueling by light-
duty vehicle (LDV) owners/operators in 1982. At the direction of EPA,
light-duty trucks are specifically excluded from the analysis. This
report presents highlights of the misfueling trends observed in a time
series data base of fuel purchasing behavior. The survey information is
derived from the NPD Petroleum Marketing Index (NPD), a diary panel
survey of over 5,000 households conducted by NPD Research Inc. The
tables and accompanying descriptive notes in this paper present findings
without any attempt at interpretation.
Misfueling may be measured in a variety of ways; the appropriate method
depends upon the questions to be answered. In this study the misfueling
rate is measured by fleet involvement, i.e., the proportion of all
catalyst vehicles which are misfueled. Vehicles are categorized on the
basis of whether or not they are ever misfueled, on the ratio of leaded
fuel purchased to total fuel purchased, and on the maximum number of
successive leaded purchases made during the survey period. Most of the
tables presented here are aggregated across all model years. An appen-
dix to this report contains a computer printout with more detailed
/¦
results on specific model years.
This study is not intended to answer questions about why people misfuel.
Rather, it quantifies the behavior observed in a representative sample
of individuals during the course of a recent calendar year. Most
previous studies of misfueling have sampled a cross-section of the
vehicle population at one point in time. A major advantage of using a
diary panel survey is that individuals may be followed through time.
-1-

-------
Such a survey provides more complete information about the frequency of
misfueling and could allow for detailed studies of the demographic
characteristics of misfuelers or motivational factors.
The most common reservation about the use of diary surveys is that they
depend upon consistent and truthful self-reporting. Despite concerns
about respondents' potential unwillingness to incriminate themselves,
the panel participants were quite open about their purchasing behavior
and freely indicated the purchase of leaded fuel. The participants know
their responses are being collected for gasoline brand market share
studies and they are accustomed to reporting detailed information about
what they have purchased. The participants do not know that government
agencies purchase the raw survey data for studies such as this one.
Furthermore, the participants are guaranteed anomymity by NPD Research,
Inc. when they agree to participate.
DATA BASE PREPARATION
The NPD data base contains fuel purchase histories for over 12,000
privately operated vehicles. The data, collected during 1982, contain
detailed information about fuel purchases, including date, gallons, type
of fuel, and total cost. An example of the purchase logs filled in by
respondents is shown in Appendix 2. There is body style/engine informa-
tion as well as household demographic data associated with each vehicle
purchase history. This data base has been used extensively by the U.S.
Department of Energy (DOE) to examine trends in fuel consumption,
on-road fuel economy, and vehicle miles of travel. During the course of
this previous work the data were cleaned and established as a SAS data
-2-

-------
set. As part of this work for DOE, the engine description information
provided by the survey respondent was verified (and corrected when
necessary using information extracted from the vehicle identification
number (VIM).* Based on this engine information, each vehicle has been
classified as to whether or not it has a catalyst.
The only additional processing of the data in the current work was to
re-examine and verify existing catalyst information. Identification of
a catalyst equipped vehicle is made on the basis of the VIN-augmented
data for make/model, model year, CID, number of cylinders, fuel system,
and type of transmission.
As noted previously, this task is based on a study of the 3694 catalyst
equipped LDVs from NPD. A small number of LDVs manufactured during the
late seventies required unleaded fuel but were not actually equipped
with a catalyst. We have assumed that the ultimate use, if any, of this
analysis will be for estimating the effect of misfueling on catalyst
vehicles. For this purpose, a misfueling rate among catalyst vehicles
is sufficient. Therefore, those vehicles not having a catalyst but
requiring unleaded were not included in the study. Another group of
LDVs excluded from the study were those participating for less than two
months. For the most part, respondents with only one month of partici-
pation have very poor record-keeping practices and incomplete purchase
histories. Frequently only one or two purchases are reported and
V
typically consist mostly of missing information. A total of 148 vehi-
cles were deleted from the survey for participating less than two
months.
* The VIN is reported by the owner, along with the engine description
information, when a vehicle first enters the survey.
-3-

-------
In any large data collection effort there is a potential for recording
or transcribing errors. To avoid over-reporting the incidence of
misfueling, only those vehicles recording at least three leaded pur-
chases during the year are counted as misfuelers. If no more than two
leaded purchases are reported, a data error is assumed and the fuel type
designation is changed to unleaded. A total of 445 vehicles meet this
maximum-of-two leaded purchases criteria. A total of 1439 "purchases"
showing 0.0 gallons of fuel bought were deleted. In general, these
records are null entries representing months when a diary was returned
but no fuel was purchased.
The tables in this report highlight misfueling behavior in the catalyst
fleet. They also provide information pertaining to the manner in which
the data are weighted and to the way in which unknown fuel type pur-
chases are handled. Each table is prefaced with explanatory notes to
assist the reader in interpreting the information presented.
-4-

-------
Table 1 - Highlights of Misfueling Frequency Distributions
•	Data in this table are weighted on the basis of the NFD
projection factors. For information on the effect of alterna-
tive weighting methods, see Table 3.
•	Total fuel purchased includes leaded, unleaded, and type
unknown. For more detail on the treatment of purchases with
unknown fuel type, and its effect on the findings, see Table
4.
•	Leaded fuel as a percent of total fuel purchased by the
catalyst fleet measures misfueling on a gallons purchased
basis.
•	The leaded fuel under 11 percent and 91-100 percent of total
fuel purchased categories measure misfueling on a vehicle
basis. Each vehicle's degree of involvement is judged on the
basis of how much of their purchase volume is leaded. The
vehicle is then assigned to an appropriate category. So, for
example, 7.0 percent of the vehicles in the catalyst fleet
were misfuelers whose leaded purchases amounted to less than
11 percent of the fuel they purchased during the year. By
comparison, 4.3 percent of the vehicles in the catalyst fleet
purchased 91-100 percent leaded fuel by volume.
•	Catalyst fleet involvement in misfueling includes any catalyst
vehicle that ever purchased leaded fuel, regardless of quanti-
ty or percentage of total fuel purchased over the year.
•	There is a small number of vehicles (0.1 percent) of the
catalyst fleet who purchased leaded fuel, but less than 10
gallons worth. These vehicles are included in the "Leaded
Fuel Under 11% of Total" category regardless of actual per-
centage.
-5-

-------
TABLE 1
HIGHLIGHTS OF MISFUELING FREQUENCY DISTRIBUTIONS
Number of Catalyst Equipped Vehicles	3694
Leaded Fuel as Percent of Total Fuel Purchased	7.7 (0.9)*
by the Catalyst Fleet
Leaded Fuel Under 11% of Total Fuel Purchased by	7.0 (0.8)
Vehicle (percent of catalyst fleet)
Leaded Fuel 91-100% of Total Fuel Purchased by	4.3 (0.6)
Vehicle (percent of catalyst fleet)
Catalyst Fleet Involvement in Misfueling	18.0 (1.2)
(percent of catalyst fleet)
Purchased at least 10 gallons of Leaded Fuel	17.9 (1.2)
(percent of catalyst fleet)
Purchased leaded fuel at least 3 times from January	2.0 (0.5)
to June 1982 and not at all from July to December
1982 (percent of catalyst fleet)
Purchased leaded fuel at least 3 times from January	8.0 (0.9)
to June 1981 (percent of catalyst fleet)
*Values in parentheses are estimated errors in the percent of catalyst
fleet.

-------
Table 2 - Distribution of Misfueling by Degree of Involvement
•	In the table, vehicle involvement in misfueling is measured by
the ratio of leaded fuel to total fuel purchased. A vehicle
purchasing a total of 400 gallons of fuel (all types) during
1982, of which 30 gallons are leaded, has a ratio of 30 to 400
or 7.5 percent. This vehicle is placed in the under 11
percent leaded category. Had the same vehicle purchased 350
gallons of leaded, out of 400 gallons total, the ratio would
be 87.5 percent leaded and the vehicle would be placed in the
81-90 percent leaded category.
•	7.0 percent of the catalyst fleet, or 38.7 percent of the
misfuelers, have leaded fuel purchases totaling less than 11
percent of their annual fuel purchases. By comparison, 4.3
percent of the fleet, or 23.8 percent of misfuelers, purchased
91-100 percent leaded fuel by volume.
-7-

-------
TABLE 2
DISTRIBUTION OF MISFUELING
BY DEGREE OF INVOLVEMENT
Aggregated Across All Model Years
Leaded Fuel Purchased
by Vehicle as Percent
of Total Fuel Purchased
Percent of Catalyst
Fleet In Category
Estimated
Error
Percent of
Mlsfuellng
Vehicles
Under 11
7.0
0.8
38.7
11-20
3.4
0.6
18.8
21-30
0.8
0.3
4.4
31-40
0.4
0.2
2.2
41-50
0.4
0.2
2.2
51-60
0.4
0.2
2.2
61-70
0.1
0.1
0.5
71-80
0.4
0.2
2.2
81-90
0.9
0.3
5.0
91-100
4.3
0.7
23.8
-8-

-------
Table 3 - Comparison of Weighting Methods
•	Each vehicle contributes one observation to the mlsfueling
analysis. When calculating the overall misfueling rate it is
helpful if an individual vehicle's contribution can be weight-
ed to account for its importance relative to other vehicles in
the fleet. This table compares three methods of weighting.
•	Sample weighting, with each vehicle assigned a weight of one,
does not distinguish among vehicles.
•	The NPD projection factor weights are assigned to each house-
hold on a monthly basis by NPD Inc. The factors are designed
to weight the sample, demographically, to the national level
based on income, race, region, and the educational level and
occupation of the female head of house. As respondents enter
and leave the survey, each household projection factor is
adjusted to maintain the national level weighting scheme. The
weight used in this study is the sum of these factors over
each month a vehicle participates in the survey.
•	The Months in Survey method assigns a weight to each vehicle
solely on the basis of the number of months a vehicle partici-
pates in the survey.
•	Overall the three weighting methods produce similar results,
although on a model year specific basis there are some differ-
ences. This is particularly true for model year 1983 where
small sample size is a problem.
•	Since the NPD projection factors were designed to weight the
survey to a national level on the basis of household demo-
graphics, these factors are used in reporting all results
except those in this table.
-9-

-------
TABLE 3
COMPARISON OF WEIGHTING METHODS
Vehicle Count
NPD
Months
Model Sample Projection in
Year Weight Factors Survey
Vehicle Involvement
Misfueling Rate (percent)
NPD
Sample Estimated Projection
Weight Error* Factors
Months
in
Survey
1975
280
374,844
2,646
25.4
5.1
25.0
24.5
1976
465
631,075
4,602
21.7
3.8
22.4
21.7
1977
586
813,714
5,785
20.8
3.3
22.0
21.5
1978
632
875,895
6,275
17.6
3.0
17.1
18.2
1979
590
801,696
5,821
14.8
2.9
14.8
14.7
1980
450
622,582
4,276
13.8
3.2
16.0
15.0
1981
485
718,297
4,760
12.4
2.9
13.0
13.0
1982
185
194,821
1,280
14.1
5.0
16.6
15.2
1983
21
8,002
59
19.1
16.8
16.3
18.6
TOTAL
3,694
5,040,926
35,510
17.4
1.2
18.0
17.9
^Estimated errors would all be based on the unweighted vehicle count,
hence they would be nearly identical across the weighting methods. In
order to simplify comparisons of the fleet involvement percentages the
error estimates have been included only for the sample weight
calculation.
-10-

-------
Table
4 - Comparison of Involvement Under Alternate Assumptions About
Unknown Fuel Type
•	This table compares three methods of treating unknown fuel
types. An unknown fuel type purchase is one in which the
respondent has failed to check either the leaded or the
unleaded column on the monthly diary log.
•	If unknown fuel type is assumed to be leaded fuel, the overall
fleet involvement in misfueling is 35.9 percent of the cata-
lyst fleet.
•	If unknown fuel type is assumed to be unleaded fuel, the
overall fleet involvement in misfueling is 18.0 percent of the
catalyst fleet.
•	If the unknown fuel for each vehicle is allocated between
leaded and unleaded, based on the percentages of known leaded
and known unleaded bought for that vehicle, the overall fleet
involvement in misfueling is 18.0 percent of the catalyst
fleet.
•	On average, each vehicle in the survey made one purchase of
unknown fuel type during 1982. This purchase, of roughly 7.0
gallons, typically represents less than 1.6 percent of the
year's total fuel purchases. Including all of these purchases
in the leaded category increases fleet involvement in misfuel-
ing by 100 percent.
•	The assumption that unknown fuel purchases actually represent
leaded fuel is made to test the belief that consumers do not
wish to Implicate themselves in misfueling. While there may
be a handful of respondents whose behavior fits this pattern,
most individuals appear to be extremely forthcoming about
their misfueling habits. Given the wide distribution of
unknown fuel type purchases and people's willingness to report
buying leaded fuel, it is likely that most, though not all,
unknown fuel purchases are the result of recording error
rather than of half-hearted deception.
•	The assumption that unknown fuel is unleaded is the most
conservative method of allocating unknown fuel. As may be
seen in the table, the results are almost identical to those
obtained by allocating the unknown fuel between leaded and
unleaded.
•	All of the tables in this report are based on the assumption
that unknown fuel may reasonably be allocated between leaded
and unleaded on the basis of the percentages of known leaded
and known unleaded bought for an individual vehicles.
-11-

-------
TABLE 4
COMPARISON OF INVOLVEMENT UNDER ALTERNATE ASSUMPTIONS
ABOUT UNKNOWN FUEL TYPE
Vehicle Involvement Mlsfueling Rate


Unknown is
Unleaded
Unknown is
Leaded
Unknown Allocated
Model
Year
Number of
Vehicles
Rate
(% of fleet)
Estimated
Error
Rate
(% of fleet)
Estimated
Error
Rate
(% of fleet)
Estimated
Error
1975
280
25.0
5.1
37.5
5.7
25.0
5.1
1976
465
22.4
3.8
39.6
4.5
22.4
3.8
1977
586
22.0
3.4
37.8
3.9
22.0
3.4
1978
632
17.1
2.9
38.0
3.8
17.1
2.9
1979
590
14.8
2.9
33.7
3.8
14.8
2.9
1980
450
16.0
3.4
33.6
4.4
16.0
3.4
1981
485
13.0
3.0
32.6
4.2
13.0
3.0
1982
185
16.6
5.4
32.5
6.8
16.6
5.4
1983
21
16.3
15.8
18.2
16.5
16.3
15.8
OVERALL
3,694
18.0
1.2
35.9
1.6
18.0
1.2

-------
Table 5 - Misfueling Fleet Involvement by Model Year
•	Number of vehicles Is a count of the actual, unweighted,
number of LDVs in each model year.
•	The categories reported here are identical to the third,
fourth and fifth items in Table 1. For example, for model
year 1975, of which there are 280 catalyst equipped LDVs in
the survey, 4.5 percent were misfuelers whose leaded purchases
amounted to less than 11 percent of the fuel they purchased
during the year. At the same time, 9.2 percent of the model
year vehicles purchased 91-100 percent leaded fuel by volume.
Overall, 25.0 percent of the model year 1975 vehicles mis-
fueled at least part of the time.
•	In general it is assumed that misfueling will increase with
vehicle age. With some slight deviations the data presented
in Table 5 shows misfueling involvement remaining relatively
flat for vehicles from model years 1979 through 1982. Vehi-
cles from model year 1978 and earlier show misfueling in-
creasing with age. Even for model years 1979 to 1982 the
percentage of vehicles in the 91-100 percent leaded category
increases with vehicle age. The invovlement rate for model
year 1983 is suspect due to relatively small sample size.
-13-

-------
TABLE 5
VEHICLE INVOLVEMENT MISFUELING RATES BY MODEL YEAR
Model
Year
1975
1976
1977
1978
1979
1980
1981
1982
1983
Number of
Vehicles
280
465
586
632
590
450
485
185
21
Fuel Under
11% Leaded
(% of Fleet)
4.5
6.3
6.8
6.5
6.6
9.7
7.7
7.8
0.0
Fuel 91-100%
Leaded
(% of Fleet)
9.2
7.3
8.8
4.2
1.6
1.4
0.3
0.6
0.0
Overall
Vehicle
Involvement
(% of Fleet)
25.0
22.4
22.0
17.1
14.8
16.0
13.0
16.6
16.3
Estimated
Error in
Involvement
5.1
3.8
3.4
2.9
2.9
3.4
3.0
5.4
. 15.8
Overall 3,694
7.0
4.3
18.0
1.2
-14-

-------
Table 6 - Repeated Misfueling
•	This table displays the Incidence of successive misfueling for
vehicles at each end of the age^soectrum (model years 1975 and
1982), and for the catalyst fleetf as a whole. Model year 1983
was excluded because of relatively small sample size. Vehicle
involvement rates are percents of the model year fleets.
•	Vehicles having made at least two leaded purchases in a row
are assigned to one of five purchasing categories. The
assignment is based on the longest string of leaded purchases
made by that vehicle during 1982.
•	Vehicles making only singleton purchases of leaded gasoline
will not appear in this table. Thus, although 25.0 percent of
the model year 1975 vehicles misfueled at least once (see
Table 5), only 24.3 percent (the sum of the five purchasing
categories for 1975 vehicles) of the fleet is represented in
Table 6. The remaining 0.7 percent of the fleet that mis-
fueled never purchased leaded twice in a row.
•	Percent of leaded purchases is calculated on a model year
specific basis. For example, 95.2 percent of the leaded
purchases made by catalyst equipped model year 1975 vehicles
were made by vehicles that have purchased leaded at least 6
times in a row.
•	Category assignments are exclusive. A vehicle making two
leaded purchases in a row on several occasions, and four
leaded purchases in a row on one occasion will be assigned
only to the category for vehicles having made four successive
leaded purchases.
-15-

-------
TABLE 6
REPEATED MISFUELING
Maximum Number
of Successive
Leaded Purchases
During the Year
2
3
4
5
Percent of
Model Year
Catalyst Fleet
Vehicles Involved
Estimated
Error
MY 75 Fleet
6 or more
0.9
1.0
3.3
1.0
18.1
1.1
1.2
2.1
1.2
4.5
MY 82 Fleet
2	2.3	2.2
3	5.1	3.2
4	2.1	2.0
5	0.8	1.3
6	or more	6.4	3.5
Overall Fleet
2
0.9
0.3
3
3.4
0.6
4
3.1
0.6
5
1.5
0.4
6 or more
8.9
0.9
Percent of
Leaded Purchases
By Fleet
0.5
0.6
2.5
0.9
95.2
11.9
20.9
9.7
4.0
53.4
1.0
4.0
4.4
2.7
87.6
-16-

-------
Table 7 - Length of Survey Participation
•	In order to be included in this misfueling study a vehicle
must have provided data for at least two months. Vehicles
reporting for only one month generally provide purchase
records with much missing or inconsistent information. In
order to reduce the effect of missing data, the minimum
reporting requirement was adopted. A total of 148 vehicles
were eliminated as a result of this requirement.
•	The majority of vehicles contributed a full 12 months of data.
-17-

-------
TABLE 7
LENGTH OF SURVEY PARTICIPATION
Number of Months In	Number of	Percent of
Survey During 1982	Vehicles	Catalyst Fleet
2	128	0.8
3	117	1.1
4	125	1.6
5	278	5.2
6	96	1.5
7	154	3.1
8	199	5.0
9	111	2.8
10	115	3.0
11	159	4.8
12	2,212	71.3
-18-

-------
Table 8 - The Distribution of Successive Misfueling Purchases
Cross-Tabulated with The Distribution of Misfueling by Degree
of Involvement
•	This table shows the degree of misfueling involvement,
measured on a volume basis as in Table 6, for the vehicles in
each of the successive misfuel purchasing categories shown in
Table 5.
•	PBMAX identifies the maximum number of successive leaded
purchases: 0 (includes single isolated purchases), 2-5, and 6
or more.
•	PBCAT identifies categories of leaded fuel as a percent of
total fuel purchased by a vehicle during 1982.
•	FREQUENCY is the NPD projection factor weighted vehicle count.
9 PERCENT is the percentage of the catalyst fleet.
•	Cumulative values are reported for both FREQUENCY and PERCENT.
•	Since only non-zero percentages are reported, some values of
PBCAT are not printed for some values of PBMAX.
•	Due to the labeling limitations of the SAS statistical package
the work "LEADED" is usally truncated to "LEA" in this table.
In a final indignity the word is further reduced to the letter
"L" for the under 11 percent category.
-19-

-------
the distribution of successive misfueling purchases
CROSS-TABULATED WITH
the DISTRIBUTION OF MISFUELING BY DEGREE OF INVOLVEMENT
PBMAX	PBCAT
0	NO LEADED
1	FUEL UNDER 11* L
1	FUEL 11- 20% LEA
1	FUEL 21- 30% LEA
2	FUEL UNDER 11% L
2	FUEL 11- 20% LEA
3	FUEL UNDER 11% L
3	FUEL 11- 20% LEA
3	FUEL 21- 30% LEA
3	FUEL 31- 40% LEA
3	FUEL 41- 50% LEA
3	FUEL 91-100% LEA
4	FUEL UNDER 11% L
4	FUEL 11- 20% LEA
4	FUEL 21- 30% LEA
4	FUEL 31- 40% LEA
4	FUEL 41- 50% LEA
4	FUEL 51- 60% LEA
4	FUEL 91-100% LEA
5	FUEL UNDER 11% L
5	FUEL 11- 20% LEA
5	FUEL 21- 30% LEA
5	FUEL 41- 50% LEA
5	FUEL 51- 60% LEA
5	FUEL 61- 70% LEA
5	FUEL 91-100% LEA
6	FUEL UNDER 11% L
6	FUEL 11- 20% LEA
6	FUEL 21- 30% LEA
6	FUEL 31- 40% LEA
6	FUEL 41- 50% LEA
6	FUEL 51- 60% LEA
6	FUEL 61- 70% LEA
6	FUEL 71- 80% LEA
6	FUEL 81- 90% LEA
6	FUEL 91-100% LEA
FREQUENCY
CUM FREQ
PERCENT
CUM PERCENT
4132638
4132638
81.982
81.982
9387
4142025
0.186
82.168
1081
4143106
0.021
82.189
737
4143843
0.015
82.204
32282
4176125
0.640
82.844
13400
4189525
0.266
83.110
119312
4308837
2.367
85.477
35829
4344666
0.711
86.188
3314
4347980
0.066
86.254
1724
4349704
0.034
86.288
497
4350201
0.010
86.298
9292
4359493
0.184
86.482
110498
4469991
2.192
88.674
32437
4502428
0.643
89.317
9484
4511912
0*. 188
89.506
252
4512164
0.005
89.511
1552
4513716
0.031
89.541
130
4513846
0.003
89.544
2146
4515992
0.043
89.587
40754
4556746
0.808
90.395
30347
4587093
0.602
90.997
1176
4588269
0.023
91.020
188
4588457
0.004
91.024
2860
4591317
0.057
91.081
942
4592259
0.019
91.099
323
4592582
0-.006
91.106
40593
4633175
0.805
91.911
57707
4690882
1.145
93.056
28025
4718907
0.556
93.612
17780
4736687
0.353
93.965
17335
4754022
0.344
94.308
17136
4771158
0.340
94.648
6188
4777346
0.123
94.771
17984
4795330
0.357
95.128
43220
4838550
0.857
95.985
202376
5040926
4.015
100.000
-20-

-------
APPENDIX 1
ESTIMATED ERRORS
Selected tables within the body of this report have included a statistic
termed "estimated error" to denote the reliability of key misfueling
rates. The purpose of this appendix is to discuss briefly the calcu-
lation of this quantity and the considerations that led to its use.
Inasmuch as the NPD data base is derived from a quota sample, it may be
justifiably argued that no statistic can reflect the "error of estimate"
as applied on the strict sense of a random sample. Nevertheless, it is
important to realize that estimates derived from a quota sample are
subject to variability and that, as a matter of pragmatism, the issues
of bias and variability should be decoupled. The analyst must exercise
due caution in selecting a quota sample, considering the purposes of the
study, comparison of sample composition, and observables (estimates of
known quantities) with independent reference sources, and the avail-
ability of alternatives to the quota sample's use.
Given that the quota sample is accepted for the purposes at hand, the
analyst must consider the variability of derived estimates in interpret-
ing the findings. The estimated error statistic is used in this report
to reflect the variability of estimates in the sense described above.
This calculation follows that of a standard error of estimate derived
from a random sample. For sufficiently large samples of size N, the 95
percent confidence limit of an observed proportion p is given by:*
CI95 = ±1.96 / p(l-p)/N

-------
In this study, the proportions p are calculated as ratios of vehicles
falling within a defined misfueling category to the total number of
catalyst vehicles in the survey. Where noted, the proportions are
weighted by the NPD projection factors (thereby incorporating both
survey participation and control of the sample's demographic balance).
In all instances, the sample size is taken to be the (un-weighted)
number of catalyst vehicles in the sample.
The resulting estimated error is an approximation to the variability
that is present in sample estimates. A more exacting calculation would
need to consider the time-series nature of the data (i.e., extended
observations of vehicles across many purchases) and the implications of
weighting factors for determining the "effective" sample size. These
extended considerations are not germane, however, to the use of the
estimated errors as an order-of-magnitude guideline to estimate
variability.
*Engineering Statistics (Second Edition) by Albert H. Bowker and Gerald
Lieberman, Prentice-Hall Inc., 1972, pp. 466-467

-------
APPENDIX 2
This appendix contains an example of the purchase logs filled in by NPD
panel participants.
-23-

-------
I
Ni
AMERICAN SHOPPERS PANEL
P.O. BOX 5401, NEW HYOE PARK, NY 11040
INSTRUCTIONS
1.	Slip this diary in the vitor holder you attach to the
sunviur.
2.	Write in ODOMETER (Speedometer) READING
when you receive this diary (but no sooner than
the data indicated) and again before you return
it Then is space for these readings above.
NOTE: If you replace this vehicle, be sure to
write in Odometer readings for BOTH vehicles.
3.	EnteraflGASOLINE/MOTOR FUELPURCHASES
on ttw reraise side of this diary.
4.	Enter ALL OIL AND ANTIFREEZE PURCHASES
on this tide of the diary. Make sure purchases at
services stations are included.
5.	Be sure to tell me all the necessary details if an-
other vehide is edded (see back of return envelope).
Tell me about the change i.i this vehicle (sold, no
longer used, disposed of, traded-in, etc.) to the right.
6.	All your diaries are important, so please be sure to
mail this diary even if no gasoline or oil was pur-
chased this month.
Wat the vehicle described on the 10 label (top left) SOLD, RETIRED FROM USE, TRADED-IN
or otherwise disposed of during the month?	DYES ONO
If YES, and vehicle was TRADED-IN or OTHERWISE REPLACED, fffl in below for new vehicle:
If NO, please do not fill in below.
MAKE-
(CtMffoUt, ford. Toyou. Oodga. ate J
MODEL NAME/SERIES.
OUtbo. 8 2I0L Udan. fl alibi I. TR 7. alLl
MODELVEAR ia	
BODY STYLE	
(2 dr. sadw. 4 dr. Mdao. 2 dr hardtop. hattfateh, van. sum.
i, pack-up, comraruttt. cjnpw. ate)
DOES THIS VEHICLE
REQUIRE UNLEAOED FUEL? M One	Ye»CD	NoD
REQUIRE DIESEL FUEL? M One	Yn ~	Ho ~
HAVE AIR-CONDITIONING? M On	Yei CD	No ~
VEHICLE IDENTIFICATION NUMBER c*»d Dmitri
i i i i i i i i i i i i i i i r n
WHAT TRANSMISSION DOES IT HAVE? (4 Om
D3 Spwd Mtraitl CZlFaur Smk) Menuil CDs Speed Mnwl
CD Automatic CDAutometic w/ Overdrive
NUMBER OF CYLINDERS M One
CDFour ~ Five (ZlSiit
(~Eight CDRatny
IS ITT (/) Om>
CDTurbocherged
~ Fud Injected
CD Regulator Carburetor
VEHICLE MAINTENANCE-DOES FAMILY MOWER.
CHANGE THE OIL? M One	Yes ~ No ~
TUNE-UP THE CAR? (4 One
PRINCIPAL DRIVER:
YEAR OF BIRTH:
Yes CD Ne CD
SEX: M ~ F ~
ENGINE DISPLACEMENT l ISO Cub« I«*n. I S uml WHICH CREDIT CAROS DOES THIS ORIVER CARRY?
Number afCubic Inches"
.OR..
"HuiiSSioT Liters"

-------
GASOLINE/MOTOR FUEl PURCHASES	AMERICAN SHOPPERS PANEL	Enter OIL and ANTIFREEZE (Coolant) Purchases on rwena side ol card.
GAS
COPY THIS INFORMATION DIRECTLY FROM GAS PUMP
METHOD OF PAYMENT
DIO STATION
HAVE—.
(J) 41 ttnl«ppl»
DIO VOU USE..
I/I an thai apply
DIO
YOU
FILL
UP
THE
tank;
Mona
SPECIAL OFFER I
m
OF
YOUR
HOMET
Mo-
NAME OF
BRAND
NAME OF
GRADE
IS THIS
GAS	
(/) ona
NUMBER OF
GALLONS/

TOTAL
AMOUNT
Was there
a special
offer?
Mone
IF YES, describe
offer such as:
Price-Off, Free
Gift (describe
Gift), Discount
for Cash, etc.
JAH

Such as:
Shell,
Texaco,
Exxon,
etc.
Such «:
Regular,
Premium,
Super,
Diesel,
Gasohol, etc.
U) ona
PAID

1 01
1—1
W	
If credit card
was used—
NAME OF
CREDIT
CARD
FULL SERVICE
PUMPS
SELF-SERVICE
PUMPS
X
3
*
DC
3
Is
u<
BUI
oo
M
LI
>
K
Mm
S5
SELF-SERVICE
PUMPS
§
*
OB
<
U
it
i>
Is
ISO
Yaw
ol
Kith
19	
1
•C"
1
LEADEO
UNLEADED
LITERS
PURCHASED
GALLONS
LITERS
(gaolina
purchaa
only)
CASN/CHECI
CREDIT CARD
Xw
Ou
55
te5
Gal/Lt
lOthi
$
«

NO
YES
NO
write in
ves
NO























































































































!








































































































































































































GASOLINE/MOTOR FUEL PURCHASES	AMERICAN SHOPPERS PANEL	Entar OIL and ANTIFREEZE (Coolant) Purchases on reverse ltd* ol card.
GAS
I COPY THIS INFORMATION DIRECTLY FROM GAS PUMP |
METHOD OF PAYMENT
010 STATION
HAVE—
WO YOU USE..
M all that apply
|P!P~1
1 SPECIAL OFFER
KB-
P
?our
HOME?
Mo-
NAME OF
BRAND
NAME OF
GRADE
IS THIS
GAS	
U) ona
NUMBER OF
6ALLONSI
LITERS
PURCHASED

TOTAL
AMOUNT

U
L
B
Wis there
especial
offer?
Mora
IF YES, describe
offer such as:
Price-Off, Free
Gift (describe
Gift), Discount
for Cash, etc
write in
BATE

Such as:
SheO,
Texaco,
Exxon,
etc.
Such as:
Regular,
Premium,
Super,
Diesel,
Gasohol, etc.
U) one
PAID

') 0
r—i
M	
If credit card
was used—
NAME OF
CREDIT
CARD
FULL SERVICE
PUMPS
u#
U 1
>
OB
Uf
S£
X
3
S
OB
2
lS
=
3<
(BUI
(SO
*1
J
>
K
ill
J*
22
SELF-SERVICE
PUMPS
§
*
ac
<
u
h
i>
&
U
oo
rrui
Yaw
of
fanth
ia_
I
1
LEAOED
UNLEADED
GALLONS
LITERS
(gaaolina
purchaa
only)
CASH/CHECI
CREDIT CARD
Zui
Ou
¦ tew
TANK1
Mono
Gal/Lt.
lOths
>
«

NO
YES
NO
ves
NO
































































































































































































































































































n
_

-------