United States
Environmental Protection
Agency
AiT
Region 10
1200 Sixth Avenue
Seattle WA 98101
EPA 910-9-80-078
Incidence of Automobile
Fue! Switching in the
Pacific Northwest
1979
-------
United States
Environmental Protection
Agency
Region 10
1200 Sixth Avenue
Seattle WA 98101
Air
EPA 910-9-80-078
Incidence of Automobile
Fuel Switching in the
Pacific Northwest
1979
Final Draft
-------
EPA 910/9-80-078
INCIDENCE OF AUTOMOBILE FUEL SWITCHING
IN THE PACIFIC NORTHWEST
by
W. Douglas Smith
Surveillance Branch
ENVIRONMENTAL PROTECTION AGENCY
REGION X
SURVEILLANCE & ANALYSIS DIVISION
1200 Sixth Avenue
Seattle, Washington 98101
December 1980
-------
ACKNOWLEDGEMENTS
We gratefully acknowledge the cooperation and assistance of the
Washington State Department of Motor Vehicles and the Idaho and Oregon
Offices of Vehicle Licensing'. Their help was a prerequisite to the
success of this study.
The author offers personal thanks to a team of highly skilled and
motivated EPA employees that served many hours beyond their normal work
loads to insure accurate data, and offer viable suggestions, and
innovative solutions to unique situations. These persons include Gayle
A. Smalley, Kirt Palmer, and Cheryl L. Stewart, field team leaders and
auditors; Bruce R. Cleland, computer programmer; Janice Gedlund, Janice
Noel, and Joyce M. Crosson field observers; and William B. Schmidt
Surveillance & Analysis Division, Air Surveillance & Investigation
Section Chief. Their tolerance for work under continuing pressure
deserves the highest praise.
Thanks to Barry Nussbaum of MSED in EPA Headquarters for assisting in
gathering the financial support and management backing for this study.
Finally, appreciation is expressed to Leo Breiman, Charles Stone, and
Peter Bickel of Technology Service Corporation for their extensive and
exhaustive support.
-------
TABLE OF CONTENTS
Item Page
INTRODUCTION ii
EXECUTIVE SUMMARY 1
METHODOLOGY 4
Survey Design 4
Field Methods 5
Quality Assurance 7
Field Observations 7
Transfer of Data 9
Motor Vehicle Registration Data 9
AUDIT RESULTS 11
MAJOR RESULTS 12
-1-
-------
ABBREVIATIONS AND SYMBOLS
OHV Department of Motor Vehicles
TSC Technology Service Corporation, EPA's statistical methodology
contractor.
EPA Environmental Protection Agency
MSED Mobile Sources Enforcement Division
QA Quality Assurance
I&M Inspection and Maintenance Program
""
A" Symbol used to signify that an alternate gasoline station was
observed when the scheduled station was closed to the general
publ ic.
P Symbol used to signify leaded premium gasoline.
PU Symbol used to signify unleaded premium gasoline.
R Symbol used to signify leaded regular gasoline.
RU Symbol used to signify unleadea regular gasoline where a leaded
regular grade was unavailable; as found in some Union Oil
stations.
ill Symbol used to signify unleaded gasoline.
-------
INTRODUCTION
National health standards for carbon monoxide, nitrogen dioxide, and
ozone are presently being exceeded in many cities throughout the nation,
including several in the Pacific Northwest and Alaska. Motor vehicles
contribute nearly 82% of the total amount of carbon monoxide found in the
urban atmosphere. They also emit over 42% of the hydrocarbons and nearly
44% of the oxides of nitrogen, two major causes of ozone (EPA 1974-75).
Through the Clean Air Act and its subsequent amendments, the U.S.
Government has placed various restrictions on motor vehicle emissions
since 1963. The Clean Air Act amendments of 1970 and 1977 required
further vehicle emission requirements, such as the development of air
pollution control strategies by the state pollution control agencies
through State Implementation Plans. Under the amendments, the automobile
industry developed their own technology for reducing these emissions.
The U.S. and some foreign automobile manufacturers chose the catalytic
converter as a control device. The catalytic converter, which requires
the use of unleaded gasoline, was first installed on some 1975 model year
automobiles. Since that time, most new automobiles were equipped with
catalytic converters. The amendments prohibit tampering with motor
vehicle emission controls by businesses engaged in repairing, servicing,
leasing, or selling motor vehicles. To insure the use of unleaded fuel,
retail gasoline stations, fleet operators, service centers, and wholesale
distributors were prohibited from fueling the vehicles designed for
unleaded fuel (those having catalytic converters) with leaded gasoline.
Additional safeguards were added to the vehicles and the unleaded
gasoline pumps to insure compliance with the Act.
-ii-
-------
However, neither the Act nor subsequent regulations addressed tampering
by individual auto owners. Some states have laws which prohibit
tampering by individuals, but few are adequately enforced.
The Environmental Protection Agency is responsible for assuring
compliance with the Act by monitoring and enforcing where violations
resulting from fuel switching and unleaded fuel contamination exist.
During the course of routine service station inspections, it was noted
that in spite of prohibitions concerning the introduction of leaded fuel
into automobiles designed for unleaded fuel, varying degrees of fuel
switching were observed at many of the gasoline service stations.
-m-
-------
EXECUTIVE SUMMARY
The Surveillance & Analysis Division of EPA Region X, Seattle,
Washington has completed a survey which was designed to: (1) assess the
relative percentage of fuel switching in the six major urban centers
within EPA Region X (Alaska, Idaho, Oregon, and Washington), (2)
establish both quality assurance and statistical methodology for field
observation and evaluation of fuel switching data, and (3) determine
possible motives for switching such as price, octane level, or gasoline
brand.
During the period April to September, 1979, approximately 9,000
vehicles were observed refueling in six cities throughout the Region,
excluding Alaska. Alaska did not represent a significant vehicle
population. Of these 9,000 vehicles, 3,000 were identified as requiring
unleaded gasoline. The fuel requirements for another 1,600 vehicles
could not be identified for various reasons (Table 4).
The cities of Boise, Eugene, Portland, Seattle, Spokane, and Tacoma
were chosen, because they represented the major population centers where
air quality problems related to vehicle emissions had been identified.
Vehicles were observed while participating in routine refueling
operations at randomly selected gasoline stations. The stations were
selected so that observations reasonably represented the total population
of vehicles and fueling practices for each specific city studied.
Employees of the Surveillance Branch of the Surveillance & Analysis
Division of Region X worked with an experienced consultant (Technology
Service Corporation). This consultant was used in close collaboration in
-1-
-------
each city to assure consistent application of predetermined study
criteria. A total of 432 service stations were observed with an average
of 21 vehicles monitored at each station. Observation periods were
statistically set to average 1.5 hours per station.
The proportion of all service station visits by catalytic-converter-
equipped vehicles in the urban area during the observation period that
resulted in the purchase of leaded gasoline (fuel switching) is defined
as "R*". The R* can be estimated as a statistical mean with 95%
confidence intervals from the observed service station visits. Table 1
shows the estimated fuel switching for each city studied.
Table 1
Fuel Switching Rates
City R* Estimate (% of fuel switching)
Boise 18.1
Eugene 8.4
Portland 6.0
Seattle 7.0
Spokane 7.3
Tacoma 4.8
With the exception of Boise, there was no significant difference
between the fuel switching rates in any of the remaining five urban
areas. The average fuel switching rate for these five areas was R* -
6.9%. This survey included a comprehensive field methodology compatible
with strict statistical techniques to determine R* and 95% confidence
intervals.
-2-
-------
Finally, the effects of the following service station characteristics
were analyzed; location, price differential, type of service, and trade
name. Although no statistically valid conclusions were reached, the
overall qualitative conclusions that might indicate parameters for
further examination are:
1. There was strong evidence that the degree of fuel switching
depends on the city under study, or at least that Boise is
different from other cities within Region X.
2. There is weak evidence that, in addition, the proportion of
switchers is higher at self, rather than full service stations.
3. At the time of the study, there was no evidence that price
difference or major/minor trade name classification played a
role. A substantially larger, carefully stratified survey would
be needed for further analysis. The 1979 fuel shortage may have
made gasoline availability more important than price, octane,
brand, or any other factor.
-3-
-------
METHODOLOGY
The methodology for this study was jointly developed by regional
office staff, representatives from EPA's Mobile Source Enforcement
Division (MSED), a statistician from EPA's Planning and Management
Office, and a contractor-consultant represented by Technology Service
Corporation of Santa Monica, California. The field, data handling, and
quality assurance (QA) procedures were agreed upon by MSED, the
consultant, and the regional office staff. Methodology was divided into
the survey design, field methods, and quality assurance procedures.
Meeting the objectives of the study was complicated by a fuel
shortage which occurred during the period of the survey. Field methods
had to be subtly altered on a continuous basis because of irregular
gasoline availability requiring greater coordination so that results
would remain statistically valid.
Survey Design:
Six urban areas were chosen to be studied in Region X on the basis of
representative population and violations of air quality standards for
pollutants related to vehicle emissions. These cities were Seattle,
Spokane, Portland, Tacoma, Eugene/Springfield, and Boise. Portland is
unique, in that it is the only city in this study with a vehicle
inspection and maintenance program (I&M). The I&M program is operated by
the Oregon Department of Environmental Quality under EPA guidelines. The
metropolitan area of Eugene/Springfield, Oregon was selected for
-4-
-------
comparison to Portland since National Ambient Air Quality Standards are
exceeded there also, but that area does not have an I&M program.
In each of the cities, the purpose was to estimate the fraction of
catalytic-converter-equipped vehicles that were using leaded gas. The
study design was to select a representative number of service stations
listed in the telephone directory yellow pages of each study area using
statistically defensible procedures, and to observe motor vehicle filling
operations at each location. Detailed survey design procedures are
presented in Appendix A (Breiman, 1979).
Field Methods:
Only trained observers were used. Control and coordination was
provided by a field supervisor. Observers positioned themselves close
enough to the fueling operations to gather accurate data without creating
suspicion or altering the normal activity of the station. The team
developed innovative observation techniques for a wide variety of
circumstances.
A Field Observation Form was supplied that contained information
about the time, location, trade name, price per gallon, vehicle license
numbers, make, model, and type of vehicle, service category, and whether
the fuel was premium, leaded, or unleaded (Appendix B). Each form also
indicated a unique station identification number. On the back of the
form, the observer would construct a map or diagram of the station
describing the location of the offices, garages, and pump island
configuration.
-5-
-------
The individual pumps were marked and identified a P (Premium), PU
(Premium-unleaded), R (Regular), RU ( Regular-unleaded), UL (Unleaded),
or D (Diesel) on the diagram. RU designated that no regular leaded
exists, as in Union stations in Washington and Oregon. If it was
impossible to determine the type of gasoline initially, the pumps were
numbered and later the type of gasoline was noted and recorded
correspondingly.
The observation period for each station was set at 1.5 hours. There
were circumstances when it was impossible to continue the observation,
and a shorter time resulted. As long as a minimum of 0.5 hours was
completed, the remaining time could be pro-rated to 1.5 hours based upon
the vehicles already observed before leaving the site. Pro-rating was
used on less than 5% of the observations and did not significantly alter
the statistical procedure or results.
If the assigned station was closed, out of gas, or restricted in
consumer service (i.e., commercial vehicles only), it was agreed that a
nearby station could be substituted. However, every effort was made to
observe the original designated station, even if it obviously had a much
lower volume than nearby stations. If selection of a nearby alternate
was necessary, the field form identification number for the designated
station was given an "A" to signify "alternate" station. Alternate "A"
stations were considered as the same as the designated stations. This
procedure was necessary as the result of the 1979 gasoline crisis, and
was under the close coordination of the field leader and consulting
contractor from Technology Service Corporation.
-6-
-------
Quality Assurance:
Quality assurance procedures were considered critical to the validity
of this study. Specific attention was given to the following study areas
where error was most likely to occur: (1) identifying and recording
license plates and other pertinent information at gasoline stations; (2)
transferring data from Field Observation Forms to computer data files;
and (3) processing data from the state Department of Motor Vehicles (DMV)
for comparison with EPA information on catalytic converter automobiles.
A system of audits and cross checks was devised for both field and office
procedures.
A. Field Observations: Accuracy in observing and recording license
plate numbers in the field was audited in three ways: (1) dual
observations; (2) dual audits; and (3) independent random audits.
(1) Dual observations were made by two observers at the same
time and location. Observers agreed on the license number
and fuel pumped prior to recording the information. If
they could not agree, both license numbers were recorded.
In this case, the description of the vehicle was
particularly important as DMV records would indicate which
number was correct by its description.
(2) Dual audits occurred when a field leader would assist in
the usual dual observation but occasionally leave for a
"rest break", but instead continue to record fueling
operations without the partner's knowledge. A minimum of 6
vehicles were recorded using this procedure at each dual
audit station.
-7-
-------
The 6 audited license numbers were compared with the
partner's original list and gave an indication of the
reliability of that observer. Any conflicts were recorded
as a "miss". This method of monitoring observer
reliability was used on a minimum of 20% of gasoline
stations included in the total study sample.
(3) Independent random audits were introduced to help maintain
high accuracy standards when observers were working alone.
Each morning, team members were assigned a list of stations
to observe, as well as the probable route and observation
sequence for the gasoline stations. This gave the field
leader some idea where each team member might be at any
given time. The leader would then adapt his/her own route
to include one or more of the other observer's assigned
stations without their knowledge. The team leader would
make a 15 minute duplicate record while the other
observation was in progress; preferably without the other
observer's knowledge. This was then compared with the
designated field form at the day's end. Conflicts were
thrown out and recorded as a "miss". This audit procedure
was done for a minimum of 6% of the total station
observations.
By using these three methods, it was possible to determine the
percent error, the type of error, and the relative accuracy of
each observer. Observers with the highest reliability were
assigned more time to work independently and accompany assistant
observers as team leaders or dual auditors. (Audit results in
appendix C.)
-8-
-------
B. Transfer of Data
All field data were audited in every phase of the transfer
process as follows:
(1) Following the field activity, each Field Observation Form
was checked for completeness and legibility and finally
initialed by the observer responsible for completing that
form. Each form was then rechecked and initialed by one or
more other people to assure data was ready for entry into
the computer data file.
(2) All of the data transcribed from the file forms to the
computer file were audited at least twice. A printout of
the original data entry was checked against the original
forms and initialed by the individual who entered the
data. The printout was again verified and initialed by one
other person. This verification procedure was completed
again if any additional errors were found. Where license
numbers or fueling information was not complete resulting
from some uncontrollable event in the field, a "miss" was
recorded.
C. Motor Vehicle Registration Data
(1) A compatible computer tape containing vehicle license
numbers was hand-carried to the respective state's
Department of Motor Vehicles office (DMV) where the tape
was run through their computers. The resulting DMV
information required by the study was again hand-carried
back to the EPA office. Procedure for the Boise portion
-------
of the study was an exception to the procedure discussed
above. An explanation of the unique measures required for
Boise is discussed in succeeding sections.
(2) Information from the DMV tapes was entered into EPA's
computer system. A printout was obtained that listed each
vehicle observed by license number, make, model, vehicle
identification number (VIN) and model year. This printout
was then rechecked against the original Field Observation
Form data for agreement. "Misses" were recorded when there
was no DMV record of the license number observed or where
the license number was protected by the DMV (Table 4).
(3) Once the make and model year of each vehicle surveyed was
determined through the DMV, the fuel requirements could be
assessed by referring to EPA reference materials and from
the manufacturer through EPA Headquarters. The number of
observed automobiles requiring unleaded gasoline in each
city was then compared with the fuel actually seen being
used and the fraction of switchers was then extrapolated.
-10-
-------
AUDIT RESULTS
It was projected that an error rate of 2% or less would meet the
criteria for field sample accuracy. Audit data indicated that field
sample accuracy was held to less than 0.8% error overall. The greatest
single error rate occurred in Eugene, where an independent audit revealed
one error in 27 observations, or 3.7%, but this seems to have resulted by
chance. Double observations and double audits for Eugene were one error
(the same error) in 180 vehicle observations and one error (the same
error) in 212 respectively, or 0.6% and 0.5%. Complete audit data may be
found in Appendix C. The following table is a city-by-city quality
assurance profile.
City
Table 2
Quality Assurance Audit Data
City-By-City
Double Observation
Disagreements/Total Double Audits/Total Independent Audits/Total
Boise 0/90 = 0% error
0/207 = 0% error
0/30 = 0% error
Portland 1/190 = 0.52% error 1/434 = 0.2% error 0/59 = 0% error
Eugene 1/180 = 0.6% error 1/212 = 0.5% error 1/27 = 3.7% error
Seattle 4/987 = 0.4% error 3/1082 = 0.3% error 0/6 = 0% error
Spokane 4/647 = 0.6% error 6/751 = 0.8% error 1/99 = 1% error
Tacoma 5/355 = 1.4% error 11/677 = 1.6% error 0/15 = 0% error
TOTAL 15/2449 = 0.6% error 22/3363 = 0.65% error 2/236 = 0.8% error
-11-
-------
MAJOR RESULTS
Technology Service Corporation of Santa Monica, California was
responsible for the final evaluation of field data in this study. They
were contracted to: (1) determine the percent of fuel switching and
confidence intervals from field data taken in each of six Region X
cities, (2) through collaboration with the field personnel, insure that
field techniques satisfied survey design criteria that would be
statistically defensible, and (3) indicate if and to what level of
confidence statements might be made regarding causes for switching, such
as price differential, gasoline trade name, type of vehicle, octane
level, or city. Several of the tables used in this section were
extracted from the Technology Service Corporation report. Their complete
report appears in Appendix D.
A total of 432 stations were observed in the six cities. Table 3
describes the number of stations observed in each.
Table 3
Station Observations
City No. of Stations Observed
Boise 87
Eugene 80
Portland 95
Seattle 61
Spokane 31
Tacoma 78
Total 432
-12-
-------
Spokane and Seattle represent the smallest number of observations for
two reasons: (1) Spokane was the first city studied, and various
observation techniques and survey design criteria had to be negotiated
with the contract consultants. Spokane therefore, represents the minimum
acceptable representative sample. (2) Seattle is the largest urban area
studied. The 1979 gasoline crisis caused extensive lines at the pumps
and extremely high fueling volume at open stations. While only 61
stations were observed, each represents a larger average number of
observed vehicles.
In Boise, a total 115 stations were observed, though only 87 were
finally determined to be within the central commuting area. Boise's
sample represents 100% of all gasoline retail outlets meeting study
criteria. This was done to compensate for a unique problem encountered
with the Idaho DMV, therefore, Boise is not a random but a complete
station audit.
Idaho DMV does not have vehicle registration computerized; license
numbers were retrieved from the DMV files manually by EPA field
observation personnel. In Idaho, the license plate is registered to the
owner and not to the vehicle. There were cases where a positive
identification was made at the gasoline station, but it failed to match
the vehicle described on the registration form. State DMV personnel said
that this was a common problem when the owner failed to notify the DMV of
the sale of the original vehicle or purchase of another. Where this
occurred, the team member retrieving the information recorded a "miss" on
the Field Observation Form, and it was not counted as an observed
fueling. This variation from the computerized system of other states is
not felt to have altered the validity of the information gathered.
-13-
-------
Each license was double checked with the corresponding registration, and
100% rather than a random cross section of gasoline stations listed in
the Boise Yellow Pages were observed.
There were approximately 9,000 vehicles observed in the survey of
which 4,500 were determined to accept leaded gas, 3,000 required unleaded
gas, and the fuel requirements of approximately 1,600 could not be
identified. The following table lists the major causes for failing to
identify fuel requirements.
Table 4
Unidentified Vehicles
Reason No. of Vehicles Not Identified
1. Mo file in DMV 407
2. Unable to determine fuel requirements because of
inadequate DMV records 556
3. Out of state license plate not checked 222
4. Could not see license plate (damaged, missing,
blocked, etc.) 393
TOTAL 1,578
Of the 407 that were returned "No file", it was learned that some may
have been unmarked police or classified vehicles whos identity was
protected. The exact number of this description was undetermined.
There were 556 vehicles observed whos fuel requirements could not be
classified. Many of these were utility, recreational, or special purpose
vehicles that were ordered with custom engine modifications. This number
represents those vehicles even the manufacturer could not classify after
being given the vehicle identification number.
-14-
-------
There were 222 unidentified out of state vehicles. Not all states
were receptive to this study or charged enormous sums to run DMV record
searches. A national police source was used but in many cases failed to
offer adequate data on make, engine classification, model or model year
to fulfill our requirements for determining fuel classification.
A survey average of 21 vehicles was observed fueling at each
station. "The proportion of all service station visits by catalytic-
converter-equipped vehicles in the urban area during the observation
period that resulted in the purchase of leaded gasoline...", is defined
as R* (TSC, 1979). Then R* can be estimated as a statistical mean with
95% confidence intervals from the observed service station visits.
Table 5 is a city-by-city break out of those means and confidence
intervals.
Table 5
City R* Estimate (%) 95% Confidence Interval]
Boise 18.1 (13.7, 22.5)
Eugene 8.4 (5.2, 11.6)
Portland 6.0 (4.3, 7.7)
Seattle 7.0 (5.2, 10.6)
Spokane 7.3 (3.5, 11.1)
Tacoma 4.8 (2.3, 7.2)
(TSC, 1979)
1. A 95% Confidence Interval indicates that there is a 95% chance that
the true percentage of switchers falls between these two numbers. There
is overlap in all the cities studied except Boise. Because of this
overlap, it is possible that the true fuel switching rate is the same.
Boise is the only city whos interval is beyond the range of any other
city. Therefore, there is no significant statistical difference between
any of the other Oregon or Washington cities studied, and Boise does
represent a significantly higher rate of switching.
-15-
-------
It was found throughout the survey that the majority of switching was
for vans and pick-ups. Boise field data indicated a higher than average
percentage of pick-ups in its total observed vehicle population. This
may have influenced the elevated rate of switching found in this city,
but further evidence is required to make any direct ties. The van
population did not differ significantly from other cities monitored.
The R* for Portland and Eugene was 6.0 and 8.4 respectively. This
suggests an apparent positive impact of Portland's I&M program, however,
because there is overlap in their confidence intervals, it would require
further documentation to be conclusive.
An overall estimate of fuel switching for Oregon and Washington's
five urban areas was:
Estimated R* (%) 95% Confidence Interval (%)
6.9 (5.7, 8.1)
The format of the study did not allow sufficient data for conclusive
analysis of the influences of price, octane, trade name, or self versus
full service. In different states, the same trade name and
classification of gasoline had different octane ratings. While
Washington and Idaho offer self-service, Oregon has full service only.
In some cities, the availability of gasoline was far more significant
than price. Each parameter has a motivating affect on switching but in
itself represents a unique and complex statistical problem. Trends may
be suggested however, and appear in the contractors report. (Appendix D.
-16-
-------
Technology Service Corporation Appendix A
2811 WILSHIRE BOULEVARD • SANTA MONICA, CALIFORNIA 90403 • PH. (213) 829-7411
SAMPLE SURVEY DESIGN
REGION X, EPA
June 1979
Leo Breiman
Charles Stone
-------
SURVEY DESIGN FOR ESTIMATING FUEL SWITCHING IN AN URBAN AREA
We deal with the problem of estimating the fraction of catalytic
converter equipped vehicles that are using leaded gas. One approach is to
conduct a survey of vehicles passing through service stations. Adopting
this approach, suppose an urban area is selected and we propose to observe
a number of service stations in the area in order to make our estimate.
What we want to come up with is the fraction P of catalytic coverter cars
passing through these stations that are buying leaded gas. This type of
survey leads to the following questions:
1. What area-wide number is the survey attempting to estimate and
is the fraction F observed as above a reasonable estimate
of it?
2. How is the urban area defined?
3. How are the stations selected within the area?
4. How many cars should be observed at each station?
5. How will the catalytic converter cars be identified?
6. What are the sources of systematic error and how can
these be reduced or compensated for?
7. What confidence statements (statistical error bounds)
can be made regarding the accuracy of the estimate?
Following our discussions with the Region X personnel (William Schmidt
and W. Douglas Smith) and after going over their draft design, we propose
a some different design we discuss below, together with some answers to the
above questions. The main alteration to the Region X design is a random
mechanism for selecting the stations to be observed. Although this may
seem like a minor change, it does permit the use of standard statistical
-------
methodology to the computation of confidence intervals. We also propose
two substudies for the purpose of estimating the size of systematic errors
in the survey.
SAMPLE SURVEY DESIGN
The essential ideas of the sample survey are these:
1. Service stations are drawn at random from the yellow pages
of the phone book (or books) covering the desired metropolitan
area. Each of the stations is then assigned a random number.
The number of stations drawn depends on the sample size
requirements which, in turn, depend on the accuracy requirements.
This will be discussed in more detail below.
2. The selected stations are then arranged by locale, i.e., are
grouped together geographically. This is to reduce travel
time between stations.
3. In each locale, the stations are visited in order of the
assigned random numbers. If there are gas lines and the
stations are only open a few hours each day, then at the
beginning of the week all stations to be surveyed should be
called and their open hours determined. If the visit plan
by random number is not feasible because of the open hours -
the order of visiting the stations should be rearranged to
make a feasible schedule.
4. Each station should be observed for a fixed time period.
We suggest two hours as a reasonable compromise time length.
All cars passing through the station during that time should
be observed.
-------
In Appendix A we go through the selection of stations in the Seattle
area using a random number table as applied to the yellow pages of the
Seattle phone book. All the details will be illustrated there.
In Appendix B we do a preliminary computation of confidence intervals
based on a reasonable statistical model. This computation gives us some
ball park figures for sample size requriements for adequate estimates of
switching in large urban area. We suggest a sample of 1000 catalytic
converter cars with 40 stations being observed. If 4 stations a day are
observed and an average of 100 cars per day surveyed, this gives 10 working
day or two week survey per urban area (another possibility is a 1:10 survey).
Finally, we address the question of systematic error. In the Region X
survey, the strategy is to note the license plate number of all vehicles
which may be equipped with a catalytic converter. Because of a slow change
in some models and makes, this may include some vehicles which are older
than 1975, say, some 1972, 73 and 74 cars. These license number are then
given to the automobile registration agency which produces make, model
and year. If we assume that almost all cars without catalytic converters
take leaded gas, then identification errors can bias our results upward.
That is, suppose a 1974 car, because of a mistake either in recording the
license number or in the registration data, is incorrectly identified as a
catalytic converter car. Then, if it was observed taking leaded gas, it was
incorrectly identified as a fuel switching car. In Appendix C we derive an
estimate of the amount of upward bias B in the estimate due to a wrong
identification of an older model car. The formula is
B - QPn/N
-------
where
Q = proportion of all registered cars that have
catalytic converters
n = number of recorded license plates registered to
cars without catalytic converters
N = number of recorded license plates registered
to cars with catalytic converters
P = proportion of the sample with incorrectly recorded
license numbers plus proportion of cars with
registered license number not corresponding to
the vehicle.
For instance, suppose Q = .5, P = .1 + .1 = .2 and n/N = .3,
then
B = .03
so that our estimate is 3% too high. The values Q, n, N are easily found.
However, the value of P poses more difficulty.
The registration error can be estimated by taking an unhurried sample
of stationary vehicles, carefully noting year, make, model, color, license
number, and checking against the registration output data for these vehicles
The proportion of licenses incorrectly recorded is harder to get at.
One possiblity is to compare independent observers over a test period.
The difficulty is that performance over the test period can be significantly
better than under normal operating conditions. This issue needs to be
pursued with the Region X personnel.
-------
APPENDIX A
RANDOM SELECTION OF SEATTLE SERVICE STATIONS
This section illustrates the use of a table of random number to select
40 service stations from the Seattle area. The population sampled consisted
of all service stations listed in the yellow pages of the Seattle telephone
book. A copy of these pages is in the back of this section. There were
742 listings. Many of these were double listings. For instance Al's Arco
Station appeared both under its name and under Arco stations.
Also in back of this section are listed 10,000 random digits. One
can start on any page. Our procedure was as follows: we used the random
digits appearing on page 546. Look at the first column of 5 digits. The
first number is 32847. Use the first three digits, 328, to locate the
station. Looking at the telephone pages, the 328 listing is the home
office of Gull Oil Co. Since this is not a service station, it is discarded.
The next number in the column is 16916. The 169 listing is Rainier Chevron
Service. But the station is double listed. To see whether to accept a
double listed station, look at the 4 digit of the random number. If it
is even, retain the station as a selection. If it is odd, do not accept.
In the present case the 4U digit is odd, and Rainier Chevron is not
selected.
In the telephone pages, the underlined stations are the ones corresponding
to the first 3 digits in the column of random numbers. The E or 0 signify
whether the 4 digit is even or odd. If even, the station is selected. If
it is odd and not double-listed, it is selected, and if double-listed, not
selected. The check by a station indicates that it was selected.
We sent down the 1s five-digit and 2n five-figit columns until 40
stations were selected. The names, addresses and telephone numbers of these
-------
selected stations are listed in Table 1. The number following the station
rd
consists of the first two digits of the random number in the 3 column.
These numbers will be used to establish the order in which the stations will
be visited.
-------
TABLE 1 SERVICE STATIONS
Name
Madison Park Arco
Dippy Duck Car Wash
Bill's Arco
Bergie's Arco
Bob's Exxon Service
Mills Chevron Service Station
Rainier Self Serve
Wedgewood Chevron
Chevron Service Station
Dick's 76 Union
K & M Exxon Products Service
Tree Exxon Service Station
Exxon Self Service Station
Gene's Mobil Service
Genesee Automotive Sales
& Service
Glendale Arco Service
Greenwood Mobil Service
Center
Gull Oil Co
Hegge Ted Chevron
Howard's Exxon Products
Service, Stn No 207
Joe's Auto Service
Ken's Shell Service
Address
4000 E. Madison
16627 Bottell Way, NE
7810 SE 27th Mercer Is!
•
523 15th E
18205 Des Moines Way S
1554 NE 145th
801 Rainier S
7300 35th NE
Brooklyn NE & NE 47th
6956 Empire Wy S
7833 SE 28th Mercer Isl
4603 S 196th
16850 Pacific Hy S
4580 Fauntleroy Wy SW
3611 Genessee
11215 8th S
405th & Greenwood
3404 4th S
7580-35th SW
101 NE 50th
6759 15th NW
150001 Bothe!! Wy NE
Tel.
324-7360
232-9977
324-2256
244-6286
362-9746
329-5398
523-9470
525-9969
725-8522
232-3007
433-9887
242-5663
938-0836
723-6300
243-3500
783-7294
624-5900
938-3333
545-9666
782-8414
362-9940
No
3
30
21
12
45
92
91
98
13
69
52
51
8
20
56
5
23
7
40
27
64
18
-------
Name
TABLE 1 SERVICE STATIONS (Cont'd)
Address
Mac's Chevron Rainier
Magnolia Union 76
Stubbs Highland Service
Benaroya Texaco
Bob's Texaco
Simpson Texaco
Angle Lake Union Service
Broadway & Maidson Union
Service
Madison Street Union 76
North City Union
Queen Anne Union
Spokane Street Union
Wayne's Shell Service
Werries 76 Service
Pat & Tom's Texaco Service
Rex's Phillip 66
Second & Union Parkade
Tom's Shell Day & Night
Towing
5 & Henderson
3301 W'McGraw
N 175th & Aurora Av N
600 S Michigan
1935 N Northgate Wy
6230 Rainier S
2065 Pacific Hy S
1100 Broadway
1700 E. Madison
1211 ME 175th
218 W McGraw
3460 1st St
S 128th & Des Moines Wy S
7900 Greenwood N
2806 S 188th
Wilson & S Dawson
1400 2d
118 NE 45th
Tel.
722-9390
283-7600
545-3191
762-8012
363-8817
723-1262
873-3829
322-9694
325-3356
362-9575
236-9962
522-7600
433-9047
782-9657
433-9941
725-5944
624-1170
No.
84
79
62
29
93
9
67
38
85
63
67
39
2
95
82
7
86
633-5400
25
-------
1230 Service
SEATTLE
For INDEX To Classifications
7
ELECTRONIC TUHE-UPS
BRAKES - DISC & DRUMS
WHEEL ALIGNMENT
COMMERCIAL & FLEET ACCHTS WELCOME
TIRES • SHOCKS • MUFFLERS
PICK-UP & DELIVERY
CLOSE TO SEATTLE CENTER
MOTELS - DOWNTOWN - FREEWAY
OPEN TODAY
623-20 !O
6th. 81 Danny Wy. •
Blanton's
Senvtee
FULL SERVICE & SELF SERVICE
INCLUDING TUNE-UPS • AUTO AIR CONDITIONING
• TIRES • BRAKES
"WE BUY & SELL GOOD USED TIRES"
DOWNTOWN
107 FAIRVIfW N
623-OOU7
NORTHGATE
1 I 5.15 15lh N I
363-270O
LAKE CITY
8001 ISlh N E.
524-8151
OPEN 7 DAYS
24 HOUR SERVICE
&
TOWING
MINOR REPAIRS
24 HOURS
MAJOR REPAIRS
7AM 5PM
AIU roNIIITIUNINU SKIIVKK
WHEEL ALIGNMENT
NEW < AR WARKANTY
SHELL A AAA TIRES
GU COOPER'S SHELL SERVICE
WOODLAND PARK ROOSEVELT DISTRICT
632-8812 & 525-8812
4r,0b FREMONT N 7501 ROOSEVELT WY.
u/'C/v '*>•> noum; N.I
COMPLETE ENG3NE WORK
FOREJGN & DOMESTIC
* FRONT END * TUNE-UPS
* BRAKES * ELECTRICAL
* AIR CONDITIONING
SHELL TIRES
AMD PRODUCTS
TIM SCOTT'S LflSlE CITY Sil
9O17 LAKE CITY WAY NE
DISTRICT
'ITEXACOI
GENERAL REPAIRS-BRAKE SERVICE
ELECTRONIC TUNE-UPS-TIRES
WHEEL BALANCING-STEAM CLEANING
• TOWING-ROAO SERVICE
1 AM 10 PM
OPfN I (JAYS A WtKK
6230 RAINIER S
Service Stations-Gasoline & Oil
Ace 4 Speed Service 10215 Greenwood N — 784-2474
Admiral Shell 2347 California SVV 937-J500
Airport Arco 15001 Pacific Hy S :—-—246-3035
Airport Chevron 18514 Pacific Hy S 246-561?
AIRPORT TEXACO
Broken—Tune-LJps—3 Wy Alignment
Auto—Electrical—Service
Bank Carda Accepted
17010 Pacific Hy S 433-9927
Alaska Native Petroleum Products
14211 Pacific Hy S-
ALKI BURTON'S EXXON PRODUCTS
-242-5100
2504 Alki SW-
-932-9845
AL'S ARCO 801 NE 65th 524-6381
(Plc.ne Spt Advfrli\fmerit rhn I'jqet
Anderson & Wedlund Texaco Service
15th NW & NW Market 782-3550
Andy's Service Center 148 SW 160th 243-7300
ARCO PETROLEUM PRODUCTS
SERVICE STATIONS—
BALLARU
CHUCK'S BALLARO ARCO
5615 24th NW 783-9900
CENTRAL
OERRY'S ARCO b23 lllh L
-324-22J6
HAROLD'S OLIVE WAY SERVICE
Pickup & Delivery-Tune Up-Minor
Mechanical Repairs-Lubrications-Srakes-
Undercoatingj Emergency Service Capitol
Hill & Downtown
Sundays 9 AM-4 PM
1611 E Olive Wy —-324-12U
£AST
MADISON PARK ARCO
4000 E Madison
—324-7360
MERCER ISLAND
DAN'S AUTOMOTIVE & ARCO
SERVICE
7810 SE 27th Mercerlsl --- 232-6280
NORTH
ACE * spcrn scHvicr
111,-]'. l.rrfiiwui»l N IW /«(«
BOB'S ARCO MINI MART
14424 Greenwood N -------- 363-1700
-364-0350
BRESNAN'S NORTHGATE ARCO
SERVICE
Electronic Scope Tune-Ups
Wheel Balancing-Brake Service <
Alignment
2101 N Northgate Wy
DON TAIE'S ARCO
General Repairs-Tune Ups & Brake
Service
14507 Aurora N 363-444J
FREMONT ARCO
3526 Fremont PI N . 634-0574
HUSTON'S ARCO MINI MART
10504 Aurora N
525-1234
NORTHEAST
AL'S ARCO 801 NE 65th
BALLINGER WAV ARCO
18503 Ballmger Wy NE
BILL WATERS ARCO SERVICE
General Repairs-Tune-Uos & Brake
Service
9500 35th NE
CREST ARCO SERVICE
509 NE IdStn
-524-6381
-362-7290
•525-3347
• 363-6247
Continued Next Page
PLETE CAR SERVICE
r-,1. 1955
.MAJOR ,J. MINOR CAR REPAIRS
! UIU'.IGN • DOMESTIC
WARRANTY WORK • BRAKES
TUNE-UPS • AIR CONDITIONING
ROADSIDE SERVICE
"REE PICK-UP i DELIVERY
ATLAS TIRES i BATTERIES
"WE NEVER SAY NO, IF WE CAN'T DO IT,
WE KNOW WHO CAN"
• ON TOP OF QFC"
525-0303
4530 25TH NE
UNIVERSITY DISTRICT
AMERICAN & FOREIGN
« ELECTRONIC TUNE-UPS
• BRAKE SERVICE
• WHEEL ALIGNMENT
ARCOM*
TIRES • BATTERIES • ACCESSORIES
AL'S A&CO
524-6331
801 N.E. 65TH
FOREIGN & DOMESTIC
• Front End Work
• Valve Jobs
• Brake Service
• Diagnostic Tune-Ups
• Infra-Red Emission Diagnosis
• Electrical Repair • Exhaust Systems
• Road Service • Qualified Mechanic
CHUCK'S
WEDGWOOD SHELL AUTO CARE
2501 N.E. 75th
-------
Please See Back Of Yellow Pages
SEATTLE
Servica 1231
Service Stations-Gasoline & Oil-
(Cont'd)
itCO PETROLEUM PRODUCTS
SERVICE STATIONS—?
RON'S SERVICE 3511 Nt 4Mn 524-1811
SANO POINT ARCO
Complete Car Service
9702 Sand Point Wy NE 5Z5-2726
TROTTER'S ARCO
Tune-Up & Repair
3418 NE 65th 522-9758
UNIVERSITY MINI MARKET
4106 Brooklyn NE 632-0908
NORTHWEST
EARL AND FRED'S ARCO SERVICE
STATION
Tune-Ups & Orakes-Mmor Repairs
Art's Highland Union Service
14056 Greenwood N
Astro Oil 1710 E Madison
Astro Oil Co 10645 16th SW
Aurora Exxon Self-Service Station
1504 Aurora N •
365-3369
324-9916
242-3376
Burien Exxon Products 14807 1st S 246-8093
Burien Gull 14302 1st S 243-2186
Burien Texaco Service 14605 1st S 433-9955
Campus Shell 700 12th 324-4886
Carl's Shell Service 511 5 Dearborn 223-9155
CHAPMAN FRED CHEVRON
11750 Lake City Wy NE 364-7180
Charley's Shell 15041 Des Momes Wy S 242-1868
CHEVRON PRODUCTS
283-9960
Aurora Service Center 12815 Aurora N 365-8540
Baldwin's Srrvke 'i4',0 S.inil Point Wy NE--5?3-1>l1)?
ll.ill.lrd Chevron .'u.'l NW Market 709-1219
Bdllard Exxon Products Service
6500 15th NW 783-7100
Ballard Union 5409 15th NW 783-3103
Ballinger Way Arco
18503 Ballinger Wy NE 1 362-7290
Barnecut Admiral Way Service
4100 SW Admiral Wy 935-7588
BEACON AVE SHELL
Complete Automotive Air Conditioning
Tunc-U|w—Hniki-a— Disc & Drum
All Makes—Bankcards Accepted
2424 Beacon S 322-7861
7724 24th NW -
-782-9774
ERIC'S GREENWOOD ARCO
Complete Automotive Repairs
7 AM-t> PM Daily-Closed Sunday
7.1 in Grcrnwond N - 783-6767
HAitvcY's AUTO scnvicc
'U.'l 1101111.111 Kd NW VU2-66U6
JOE'S AUTO SERVICE
6759 15th KW 782-8414
REARDON & JOHNSON AUTOMOTIVE
Hydramatic SDecialists-Complete Major
Overhaul
4420 Ucary Wy NW—: 783-0352
SIGH'S ARCO 6501 32d NW 789-3151
QUEEN ANNE
Bell & Eaton Service 5100 25th NE-
Benaroya Texaco 600 S Michigan —
Bert's Gull Service 17704 15th NE •
-525-0550
-762-8012
-367-0111
Bill Hall's Chevron Service 1424 NE 125th—362-9923
Bill Waters Arco Service 9500 35th NE 525-3387
BILL'S SHELL SERVICE
SUKLL
AUTO CARE
N.I.A.S.E. CERTIFIED MECHANICS
15th & 145th NE
Engine4 & Electrical Repair
Brake Wheel Alignment
Tune-Ups Air Conditioning
All Makes
1513 NE 145th 362-1223
QUEEN ANNE ARCO BIRCHARD & AGEE SERVICE &
11 -»1 nirrrrt fl nn. N ^fl^-ArtA? MAR1NF ?A?^ NF ft"}fh -_, .— — 571-9400
SOUTH Bl
BOB'S ARCO SERVICE 5304 1st S — 762-9871 B'
BURIEN ARCO
Complete Car Maintenance-Specializing g.
In Electrical Winng-Hours-7 AM-11 PM
GENESEE AUTOMOTIVE SALES 4
GLENOALE ARCO SERVICE
IMPERIAL RICHFIELD CO INC
JBM SERVICE
12025 DCS Momes Wy S 246-6352
KOHL & KOHL ARCO SERVICE
l *i(Mtllir -878-791 I
BALLARD
BALLARO CHEVRON
2021 NW Market -
-789-1219
BELLEVUE
MAIN STREET CHEVRON
10812 Mam Bellevue
-454-7468
BURIEN
CROMBIETRED CHEVRON SERVICE
15804 Des Momes Wy S 433-9969
CAPITOL HILL
BOREN & MADISON CHEVRON
Complete Automotive Service-Open 24 Hrs
1101 Madison — 623-0833
BROADWAY CHEVRON STATION
915 e Hoy 322-0195
RENTAS LOU CHEVRON STATION
General Repairs Road Side Service
1531 Broadway 329-3331
DOWNTOWN
BOREN & MADISON CHEVRON
Complete Automotive Service-Open
1101 Madison
BOREN & STEWART CHEVRON
1024 Stewart •
Colman Building Garage
809 Western Av
GENE SNYDER'S CHEVRON
1024 Stewart ——
KNECHTEL CHEVRON SERVICE
2900 1st Av-
MIKE BEASLEY'S CHEVRON
SERVICE Oenny Wy & Stewart-
RED CARPET CAR WASH
1164 Oenny Wy
24 Hrs
623-0833
-623-7990
—682^991
-623-7990
-682-0737
-623-5723
-624-3317
CHEVRON PRODUCTS—(Cont'd)
DEALEnS-(Cenl'd)
EASTLAKE
KELLY TERRY CHEVRON STATION
2727 Eastlake E 329-1240
FEDERAL WAY
FEDERAL WAY TRUCK VILLAGE
S 348th & 16th S Federal Way
DES MOINES TEL No 838-9014
FIRST HILL
OOREN & MAOISON CHEVRON
Complete Automotive Service-Open 24 Hrs
1101 Madison 623-0833
!APHAEL FRED CHEVRON STATION
Foreign & Domestic Minor Repairs Front
Cnd Alignment-Tuneups Brakes Hallmark
Award Station
914 James 624-1108
CREENLAKE
GREENLAKE CHEVRON
Auto Repairing & Service Brakes & Tune-
Up
MIDO I r.rren L.ike Wy N 522-O408
DUWAMISH
DUWAMISH CHEVRON
Lubes-Brakes-Alignments-Atlas Tires-
8attenes-Tune-Ups-Electronic Wheel
Balance
South Of Boeing-E Marginal
10655 E Marginal Wy 5 762-7077
Continued Next Column
GREENWOOD
RAY'S CHEVRON SERVICE
10415 Greenwood N-
-782-9717
BOBBINS' CHEVRON SERVICE
Ryder Truck & Trailer Rentals
Propane Uollles Filled
14504 Greenwood N 362-9715
TEX SALMON'S CHEVRON SERVICE
CENTER
Tune-Ups-Front End Alignments
355 NW85th 789-1661
KENMOKE
BOTHELL WAY CHEVRON
Tune-Ups Brakes Mufflers Minor Repairs
6504 NE Bothell Wy Kenmore —485-5500
Kenmore Chevron
lUtllirll My A JU.IIKI.I Ur llullii-tl 4U6 W/0
K1RKLAND
JUANITA CHEVRON SERVICE
11601 98th NE Kirkland 823-6766
LAKE CITY
CHAPMAN FRED CHEVRON
11750 Lake City Wy NE
-364-7180
LAKE FOREST PARK
FOREST PARK CENTER CHEVRON
SERVICE
Complete Lubrication-Brake & Tuneup
17017 Bothell Wy NE 365-7565
MAD/SON PARK
Kelly Terry Chevron Station No 2
3115 E Madison 325-2500
MAGNOLIA
KELLEY'S CHEVRON SERVICE
384834th W
MAGNOLIA VILLAGE CHEVRON
3300 W McGraw
-284-2330
-285-1761
MERCER ISLAND
CLYDE'S CHEVRON SERVICE
2800 Island Crest Wy Mercerlsl-232-9772
SOUTH MERCER CHEVRON
8407 SE 68th Mercerlsl 232-5555
MOUNT £-\K£ TEHllACK
MIKE'S CHEVRON SERVICE
Towmg-Tune-Up-Brakes-Steam Cleaning-
Pickup & Delivery-Minor Repairs
20330 15U1 NE 363-0066
NORMANDY PARK
Hickman Jim Chevron Service
Electronic Tune Ups-Brake Service
17651 1st S-
243-7484
NORTH CITY
TOM'S NORTH CITY CHEVRON
Tuneup-Brakes-Mufflers & Tail Pipes
17508 15th NE 364-1840
Continued Next Page
Skiing this month? Don't forget to
phone ahead for reservations.
Telepl one service is a friendly service.
-------
1282 Service
SEATTLE
For INDEX To Classifications
Service Stations-Gasoline & Oil-
(Cont'd)
CHEVRON PRODUCTS—(Com'di
DEALERS^Canfdl
NORTH SEATTLE
BILL'S CHEVRON SERVICE
SS-V'i.Auror,! H 542-2020
MILLS' CHEVRON SERVICE STATION
Tune- (Jo-Brake Service Minor Repairs-Air
ConoM.onmg-Front End Alignment
1554 NE 145th 362-9746
tjllKKN ANNK
MERCER STREET CHEVRON
150 Mercer 285-9823
QUEEN ANNE CHEVRON
Don Maxwell-Dealer
Ml 7 Queen Anne H 284-1090
RAINIER DISTRICT
RAINIER CHEVRON SERVICE
2BOO Rainier S
723-3033
RAINIKR V ALLEY
CHEVRON PRODUCTS—(Confd)
DEALERS-tCont'dl
WEST SEATTLE
DEAN MOON'S CHEVRON
Tune-Ups-8rakes-Atlas Tires 4 Batteries
230i California SW 332-9822
HECGE TED CHEVRON
7580 35in Sw 938-3333
Ev's Mercer Avenue Shell
Mercer 4 Boren N—•-
EXXON PRODUCTS
Stromberg's Chevron Station
3720 California SW —
WHITE CENTER
WHITE CENTER CHEVRON
Chevron Service Stations
Qrookl/n NE & Hi 47th
East Marginal Way S 4 1st S
5940 E Mjroinal Wy S— 763-1567
Chin Brothers Service 2901 17tn S 324-4646
CHUCK'S EXXON PRODUCTS & SERVICE
Minor Repairs & Tune-Ups
Tires—Brakes— Mufflers—Shocks
Steam Cleaning—Pick-Up At Delivery
5620 Empire Wy S 722-0500
SMITH'S CHEVRON SERVICE
lowing Service Av.lHdble
6061 Empire Wy S -------- 725-3200
RICHMOND BEACH
NELSON CHEVRON
fiMir-ilp^-Qr.i •!••,- Air Co'iditiOnmy- Minor
Hc[i.iir\- r.rcs-b.ltlenci
1HO N i75th --------- 546-2626
KIVKKTON HEIGHTS
RON'S CHEVRON
Tune 'Jps i Qrjke Service
I utu.r.HKin .Wncu-i Aliijnm?nl
I62oo Minury Hd b - ......... -242-4222
SANDPOINT
SANDPOINT CHEVRON
CHUCK'S WEDGWOOD SHELL AUTO
CARE 2501 NE 7Mh 522-8284
(Please See Advertisement Page 1280)
Cliff Housed 76 11845 Oes Moines Wy S — 248-0470
CLYDE'S CHEVRON SERVICE
MERCER ISLAND
WRECKER SERVICE
DAY OR NIGHT
AIN-CONDITJONING
AAA EMERGENCY SERVICE
2000 Isi Crest Wy Mercer Isl——232-9772
>.<,-! -...1..0 HUII.I wy 1L 524-2463
SOUTH I'AHK
RUSS'S CHEVRON SERVICE
^D-Jir>-Or iUs-Tune-Ups
£JcclrDnic Wheel Balancing
ClO'.e To Boeings Mam Piant
3700 i4tn S 762-3233
SOUTH SEATTLE
Chevron Service Stations
5140 £ Marginal Wy S —763-1567
SOim/E.VD
MURRAY'S TRUCK TERMINAL
.,:-,., tir, . -682-0800
RAY HALL'S CHEVRON 2740 tst S 223-9758
TUKWIL,\
SOUTHCENTER CHEVRON
Tunc-Uoi A Wheel Alignments
220 Stranaer Bv 575-0360
U.\ / VERSITV DISTRICT
Chevron Service Stations
Jio,-.u/n NE .v NE 47th 525-9969
MANLEY CHEVRON
Electronic Wieel Balancing
LuDncjiion-Tuneups-Brdke*
J31*! Poo5cveit Wv f;E 633-5534
UNIVERSITY VILLAGE CHEVRON
A.r i-.ind-ar.ites-Rd Serv-Tunf-i.lpi
Jbiu .'t>tn NE 525-0303
WALLI.\C,rORD
OIEN'S CHEVRON SERVICE
Minor RepJi'J-TL,nc-UPS-yr.)ires
in;., ,'i..jge --H, H 632-2005
WEBB LARRY CHEVRON SERVICE
•Vw Cjr Warranty Se^ice
,'mcori i Domestic Repjir-Serv.ce
1-1:0 v -"in 633-1665
VIEWRIDGE
Electronic Tune-Ups
Wheel Alignment
Specializing In
Drakes & Aulumntic Trnn»>mis*>ona
Frit Pu-k-Up 4 Delivery
Tires—Batteries—Road Service
7347 35th NE— 525-5925
South Of Bocmit On E. Marpnal
Lubes—Brakes—Minor Repair
Alignment—-Alias Tires & Accessories
10655 E f.Mrnm.ll Wy S — 762-7077
URRY ROE'S CHEVRON
WtDCEWOOO CHEVRON
.'•ev.--Tji.s i Oe'-.e'»
-..00 J5:- •.£ -523-3470
Continued Nfit Column
ERIC'S GREENWOOD ARCO
Complete Automotive Repairs
Open 7 A.M-ti I'M Djily — Clo>«d Sunday
,'« 18 Greenwood S 7B3-6767
-623-0954
-932-9835
762-6226
-525-9969
Clyde's Chevron Service
2800 Island Crest Wy Mercerlsl -232-9772
Coastal Oil Co Inc 13515 Ambaum Bv SW —242-9009
Collins' Service Station
inOOO Empire Wv S 722-6696
CuluintHAn funn PrcMlucU
.'IJMJ ', (..,lull,l.i.iii Wy /6J-1366
Counter Balance Union Service
700 Queen Anne N — 284-1076
Crest Arco Service 509 NE 165th 363-6247
Crombie Fred Chevron Service
15804 Oes Moinej Wy S • 433-9969
Crossroads Shell 2121 Empire Wy S 723-9970
Dale's Texaco Service & Towing
210 NE 45th 632-1404
DAN'S VIEWRIDCE SHELL
Dean Moon's Chevron 2301 California SW—932-9822
Dean's Texaco Svc 7301 15th NW 782-2111
Delndge-Way Exxon
7301 Oeindge Wy SW 762-O448
Perry's Arco 523 ;5th E 324-2256
DICK'S 76 UNION 6956 Empire Wy S-723-8522
Dick's Towing Inc 13038 Inleruroan S 243-1647
Discovery Part Automotive
J317 W Government Wy 282-0500
Don Taie's Arco 14507 Aurora N 363-4464
Don's Arco Service
10022 Roosevelt Wy NE 365-6226
Don's Interbay Arco 3201 20th W 284-5427
Don's Union 76 Service Station
2-115 Deacon S 322-5722
Doug's Union 2501 HE 5Mn 524-5559
DUWAMISH CHEVRON
Earl And Fred's Arco Service Station
7724 2-Un NW 782-9774
Edwards Bras Service Center
5400 Sin NW— —-789-1414
Eg.in 4 Sons Exxon '.4433 M.hur* Rd S —246-0441
EMPIRE TEXACO SERVICE
6600 Ema,-e Wy S 722-9642
525-2626 Empire Way Cull 183-1 Emp.re Wy S 722-7338
Empire Way Mobil 2801 Ervc.rc .Vy S 722-49V5
EXXON Gasolines;
Umflo Motor Oil,
A Complete Une of
Atlas ® Tires,
Batteries 4 Accessories
Enon Self-Serv 420 NE 45th S?"9233
Exxon Self-Serve Station 6408 Aurora N—525-478!
Exion Self Service Station
16850 Pacific Hy S 242-5661
Fast Gas
4001 California SW — —932-9965
3810 S Morgan -723-2111
14711 15th NE 362-9779
Favors Service 932 19th 322-0680
Federal Way Truck Village
S 348th 4 16th S Federal Way
Oes Moines Tel No 838-9014
"WHERE TO GET SERVICE" FENNER'S TEXACO SERVICE
DISTRICT OFFICE
EXXON COMPANY USA Seiievue - 453-4500
W HOUR STATIONS & DEALEKS
BALLARD EXXON PRODUCTS
SERVICE
Tune-Ups-8rakes-Alignment General
Repairs
6500 15th NW 783-7100
SERVICE STATIONS & DEALERS
AUO BURTON'S EXXON PRODUCTS
2504 Alki SW .
-932-984S
BROWN'S EXXON PRODUCTS
Uuen 7-U (.'very DJy
1426 34th 323-9858
BURIEN EXXON PRODUCTS
Complete Mechanical Service
14807 ist S 246-8093
CHUCK'S EXXON PRODUCTS &
SERVICE 5620 Empire Wy S 722-0500
Egan & Sons Exxon
14438 Military Rd S 246-0441
EXXON SELF-SERVE STATION
8408 Aurora N 525-4788
EXXON SELF SERVICE STATION
16850 Pacific My S 242-5668
HIGHLINE EXXON
Corner Of 0«l Moines Way S A 8th S
Maior & Minor Overhaul-Electronic Tune-
Ups
18205 Des Momes Wy S 244-6286
Holman Road Exxon Products Service
Tune-(Jps-8r,ikes-Afiinment
•I/Ill Hull,i.tn IM NW -702-3030
HOWARD'S EXXON PRODUCTS
SERVICE
Utility Trailers-For Rent
3400 California SW 937-6600
Jerry's Exxon Service Center
2851 SW Roxbury
John's Exxon Service
12911 Empire Wy S -
937-6155
—772-6512
-232-3007
Specializing In Tires 4 Batteries 4
LuDncations
•>»i03 S !88th
VERDI'S EXXON SERVICE
1M02 Military Rd S
433-9887
244-7085
Continued Next Column
Keep a phone by your oedside.
CROWN HILL
Minor Repairs—Tune-ups—Goodyear Tires
Pick-Up & Delivery
1701 NW 85lh 783-6199
FIII-Em-Fast
4115 SW Admiral Wy —
14656 Ambaum Sv SW
•935-1082
246-7733
•345-9401
Flajole Bros Service 2201 4th S
FOREST PARK CENTER CHEVRON
SERVICE 17017 Botheii Wy NE 365-7555
Forrest's Shell Service 6419 15tn NW 784-8877
Fourth Ave Arco 2200 4th S 624-2677
4th Ave South Shell 6185 4th S 763-0505
Franklins Service 5406 California SW 935-4123
Frank's Self-Service 12603 Senton S 772-1380
Fremont Arco 3526 Fremont PI N 634-0574
Freeway CarJge lil? Ulh 623-3806
Gary's Union Service 159 Denny Wy 622-6683
Gary's Wesuide Arco Service
3901 SW Alaska 938-0800
Gas Station The 21449 Pacific Hy S 824-5600
Gasco Inc 14805 Interurban S 241-0486
Gene Snyder"! Chevron 1024 Stewart 623-7990
Gene's Mobil Service
4580 Fauntleroy Wy SW 938-0836
Geoffrey's Arco Station 12354 15th NE 364-8200
GEORGETOWN SHELL 6200 Corson S- 767-6200
Glen- ale Arco Service 11215 8th S 243-3500
Glen, ale Union 76 806 S 112th 243-5228
Glem.'s R/C Automotive 2201 4th S 623-5584
Grant's Gas 'N' Go
1945 Aurora H 283-1693
13435 Intergrban S 243-1103
7219 Rainier S 723-3169
2805 SW Roxbury 938-3656
6451 42d SW 932-9637
GREENLAKE AUTOMOTIVE
umj wiuvii.i»ti NP - --- - 522-9108
(ireenljke Chevron
6800 E Green Lake Wy N 522-0488
Greenlake Gas & Auto Repair
6501 Aurora N 782-6094
— 362-2236
623-0033
K & M Exxon Products Service
7833 SE 28th Mercerlsl
LEN & CLEV'S EXXON PRODUCTS
& SERVICE 2841 S math 244-8610
MURPHY JERRY EXXON PRODUCTS
SERVICE STATION
Major Automobile Repairs And Service
7301 5th NE 522-0507
PETZOLO'S EXXON SERVICE
2137 N Northgate Wy 362-9983
REBEL JACKSON'S BURIEN EXXON
PRODUCTS 14807 m S 246-8093
REBEL JACKSON'S VILLAGE EXXON
Complete Mechanical Service
17956 1st S
246-8053
—244-1622
Riverton Heights Exion Products
Office 14415 Pacific Hy S
Products & Service
14415 P.K.fic Hy S 433-9958
SKYWAY EXXON PRODUCTS
SERVICE
Complete Service-Free Pickup-Delivery
Brakes-Front End Alignment Tuneups
11655 Renton S 772-0660
SMITH & McLAUCHLAN EXXON
PRODUCTS 4004 NC ssth 522-fl709
SPENCER'S EXXON PRODUCTS
SERVICE
Open 7.00 AM -6:00 PM Daily
SW i07th 4 isth SW 244-2697
STEVE'S EXXON SERVICE
Tune-Ups-Srakes-i-ubncation-Emission
Control
6056 Empire Wy S 723-1000
TERMINAL SERVICE 2401 4th S -623-4675
TERRACE EXXON SERVICE NO 2
20010 Ballmger Rd NE 364-6729
Tyee Exxon Service Station
Grinhagen's Wally Shell Service
11346 Lake City Wy NE
Grosvenor House Garage 505 Vine —
Gull Industries 2005 E Madison 329-3700
GUU. SERVICE STATIONS
QUALITY
GASOLINE
FOR LESS
"WHERE TO BUY IT"
HOME OFFICE
GULL INDUSTRIES INC 3404 4th S 624-5900
BROADWAY
Broadway Gull 1500 Broadway 324-0919
ROOSEVELT
Roosevelt Gull 6417 Roosevelt Wy NE—524-4199
SOUTH INDUSTRIAL AREA
Pingrey's Gull 4115 4th S 623-5320
WHITE CENTER
While Center Gull 11007 16th SW 248-1570
-763-8676
Gull Station Georgetown
Corson S 4 S Michigan
GUS COOPER'S ROOSEVELT SHELL
7501 Roosevelt Wy N£ 525-8812
GUS COOPER'S SHELL SERVICE
4605 Fremont N 632-8812
(Please See Advertisement Page 12301
HadfieW's Garage 3127 E Madison 322-4965
HAL'S SHELL SERVICE
Expert Brake Service & Tune-Up
2244 Eastlake £ 323-9438
HANCOCK—
I U-SAVE OIL CO
9076 Holman Rd NW 762-4930
Continued Next Page
S«e the Yellow Pages . . - first!
Continued Nut Column
-------
543
TABLE A 1
TEN THOUSAND RANDOMLY ASSORTED DIGITS
00
01
02
03
04
05
06
07
OS
09
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
00-04
54463
15389
85941
61149
05219
41417
28357
17783
40950
S2995
96754
34357
06318
62111
47534
98614
24856
96887
90801
55165
75884
16777
46230
42902
81007
68089
20411
58212
70577
94522
42626
16051
08244
59497
97155
98409
45476
89300
50051
31753
79152
44560
68328
46939
83544
91621
91896
55751
85 156
07521
05-09
22662
85205
40756
69440
81619
98326
94070
00015
84820
64157
17676
88040
37403
52820
09243
75993
03648
12479
21472
77312
12952
37116
43877
66892
00333
01122
67081
13160
42866
74358
86819
33763
27647
04392
13428
66162
84882
69700
95137
85178
53829
38750
83378
38689
86141
00881
67126
62515
87689
56898
10-14
65905
18850
82414
11286
10651
87719
20652
10806
29881
66164.
55659
53364
49927
07243
67879
84460
44898
80621
42815
83666
84318
58550
80207
46134
39693
5I11I
89950
06468
24969
71659
85651
57194
33851
09419
40293
95763
65109
50741
91631
31310
77250
83635
63369
58625
15707
04900
04151
21108
95493
12236
15-19
70639
39226
02015
88218
67079
92294
35774
83091
85966
41180
44105
71726
57715
79931
00544
62846
09351
66223
77408
36028
95108
42958
88877
01432
28039
72373
16944
15718
61210
62038
88678
16752
44705
89964
09985
47420
96597
30329
66315
89642
20190
56540
71381
08342
96256
54224
03795
80830
88842
60277
20-24
79365
42249
13858
58925
92511
46614
16249
91530
62800
10089
47361
45690
50423
89292
23410
59844
98795
86085
37390
28420
72305
21460
893SO
94710
10154
06902
93054
82627
76046
79643
17401
54450
94211
51211
58434
20792
25930
11658
91428
98364
56535
64900
39564
30459
23068
46177
59077
02263
00664
39102
25-29
67382
90669
78030
03638
59888
50948
75019
36466
70326
41757
34833
66334
67372
84767
12740
14922
18644
78285
76766
70219
64620
43910
32992
23474
95425
74373
87687
76999
67699
79169
03252
19031
46716
04894
01412
61527
66790
23166
12275
02306
18760
42912
05615
85863
13782
55309
11848
29303
55017
62315
30-34
29085
96325
16269
52862
84502
64886
21145
39981
84740
78258
86679
60332
63116
85693
02540
48730
39765
02432
52615
81369
91318
01175
91380
20423
39220
96199
96693
05999
42054
44741
99547
58580
11738
72882
69124
20441
65706
05400
24816
24617
69942
13953
42451
20781
08467
17852
12630
37204
55539
12239
35-39
69831
23248
65978
62733
. 72095
20002
05217
62481
62660
96488
23930
22554
48888
73947
54440
73443
71058
53342
32141
41943
89872
87894
03164
60137
19774
97017
87236
58680
12696
05437
32404
47629
55784
17805
82171
39435
61203
66669
68091
09609
77448
79149
64559
09284
89469
27491
98375
96926
mn
07105
40-44
47058
60933
01385
33451
83463
97365
47286
49177
77379
88629
53249
90600
21505
22278
32949
48167
90368
42846
30268
47366
45375
81378
98656
60609
31782
41273
77054
96739
93758
39038
17918
54132
95374
21896
59058
11859
53634
48708
71710
83942
33278
18710
97501
26333
93842
89415
52068
30506
69448
11844
45^9
08186
26927
15345
77455
75577
30976
76305
75779
90279
37231
27083
71113
80182
11551
13491
34770
44104
94771
18106
41067
85436
10620
59337
13119
49037
21546
33848
63700
03283
13163
62S80
60631
72655
83864
82859
41567
22557
03887
33258
22716
48805
68618
65747
91777
55349
23466
60142
09808
87530
01117
-------
544 Appendix Tables
TABLE A 1—(Continued)
00
01
02
03
04
05
06
07
08
09
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50-54
59391
99567
10363
86859
11258
95068
54463
16874
92494
15669
99116
15696
97720
11666
71628
40501
22518
75112
80327
60251
57430
73528
25991
78388
12477
83266
76970
37074
83712
20287
74261
64081
05617
26793
65988
27366
56760
72880
778S8
28440
63525
47606
52669
16738
59348
12900
750S6
99495
26075
13636
55-59
58030
76364
9751S
19558
24591
88628
47237
62677
63157
56689
75486
10703
15369
13841
73130
51089
55576
30-185
02671
45548
82270
39559
65959
16638
09965
32883
80876
65198
06514
56862
32592
49863
75818
74951
72S50
42271
10909
43338
38100
07819
94441
93410
45030
60159
11695
71775
2^537
51434
31671
93596
50-64
52098
77204
51400
64432
36863
35911
73800
57412
76593
35682
84989
65178
51269
71681
78783
99943
98215
62173
98191
02146
10421
34434
70769
09134
96657
42451
10237
44785
30101
69727
86538
08478
47750
95466
48737
44300
98147
93643
03062
21580
77033
16359
96279
07425
45751
29845
49939
29181
45386
23377
65-69
82718
04615
25670
16706
55368
14530
91017
13215
91316
40844
23476
90637
69620
98000
75691
91843
82068
02132
84342
05597
00540
88596
64721
59980
57994
15579
39515
68624
78295
94443
27041
96001
67814
74307
54719
73399
34736
58904
58103
51459
12147
89033
14709
62369
15865
60774
33595
09993
36583
51133
70-74
87024
27062
98342
99612
31721
33020
36239
31389
03505
53256
52967
63110
03388
35979
41632
41995
10798
14878
90813
48228
43648
54086
86413
63S06
59439
38155
79152
98336
54656
64936
65172
18888
29575
13330
52056
21105
33863
59543
47961
47971
51054
89696
52372
07515
74739
94924
13484
38190
93459
95126
75-79
82848
96621
61891
59798
94335
80428
71824
62233
72389
81872
67104
17622
13699
39719
09847
88931
86211
92879
49268
81366
7588S
71693
33475
48472
76330
29793
74798
84481
85417
08366
85532
14810
10526
42664
01596
03280
95256
23943
83S41
29882
49955
47231
87832
82721
05572
21810
975S8
42553
48599
61496
80-84
04190
43918
27101
32803
34936
39936
83671
S0827
96363
35213-
39495
53988
33423
81899
61547
73631
36584
22281
95441
34598
66049
43132
42740 '
39318
24596
40914
39357
97610
43189
27227
07571
70545
66192
85515
03S45
73457
12731
11231
25878
13990
58312
64498
02735
37875
32688
38636
2S6I7
68922
52022
42474
85-89
96574
01896
37855
67708
02566
31855
39892
73917
52887
09840
39100
71087
67453
07449
18707
69361
67466
16783
15496
72856
21511
14414
06175
35434
77515
65990
09054
78735
6004S
05158
80609
89755
44464
20632
35067
43093
66598
83268
23746
29226
76923
31776
50803
71 153
20271
33717
17979
52125
41330
45141
90-94
90464
83991
06235
15297
80972
34334
60518
82802
01087
34471
17217
84148
43269
47985
85489
05375
69373
86352
20168
66762
47676
79949
82758
24057
09577
16255
73579
46703
72781
50326
39285
59064
27058
05497
03134
05192
50771
65938
55903
23608
96071
05383
72744
21315
65128
67598
70749
91077
60651
46660
95-99
29065
51141
33316
28612
08188
64865
37092
84420
66091
74441
74073
11670
56720
46967
69944
15417
40054
00077
09271
17002
33444
85193
66248
74739
91871
17777
92359
98265
72606
59566
65340
07210
40467
33625
70322
48657
83665
81581
44115
15873
05813
39902
88208
00132
14551
82521
35234
40197
91321
42338
-------
545
TABLE A I—(Continued)
50
51
52
53
54
55
56
57
53
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74.
75
76
77
78
79
SO
81
82
83
84
85
S6
87
88
89
90
91
92
93
94
95
96
97
98
99
00-04
64249
26533
05845
74897
20872
31432
66890
41894
11303
54374
64852
16309
42587
40177
82309
797S8
40538
64016
49767
76974
23854
68973
36444
03003
17540
38916
64288
86809
99800
92345
90363
64437
91714
20902
12217
45177
28325
29019
84979
50371
53422
67453
07294
79544
64144
90919
06670
36634
75101
05112
05-09
63664
44249
00512
68373
54570
96156
61505
57790
87118
57325
34421
20384
37065
98590
76128
68243
79000
73598
12691
55108
08480
70551
93600
87800
26188
55809
19843
51564
99566
31890
65162
32242
53662
17646
86007
02863
90814
28776
81353
26347
06825
35651
85353
00302
85442
11883
57353
93976
72891
71222
10-14
39652
04050
78630
67359
35017
89177
01240
79970
81471
16947
61046
09491
24526
97161
93965
59732
89559
18609
17903
29795
85983
25098
65350
07391
36647
47982
69122
38040
14742
95712
32245
48431
28373
31391
70371
42307
08804
56116
56219
48513
69711
89316
74819
45338
82060
58318
86275
52062
85745
72654
15-19
40646
48174
55328
51014
88132
75541
00660
33106
52936
45356
90849
91588
72602
41682
26743
04257
25026
73150
93871
08404
96025
78033
14971
11594
78386
41968
42502
39418
05028
08279
82279
04835
34333
31459
52281
53571
52746
54791
67062
63915
67950
41620
23445
16015
46471
00042
92276
83678
67106
51583
20-24
97306
65570
18116
33510
25730
81355
05873
86904
08555
78371
13966
97720
57589
84533
24141
27084
42274
62463
99721
82684
50117
98573
25325
21196
04558
69760
48508
49915
30033
91794
79256
39070
55791
33315
14510
22532
47913
64604
26146
11158
64716
32048
68237
66613
24162
52402
77591
41256
26010
05228
25-29
31741
44072
69296
83048
22626
24480
13568
48119
28420
10563
39810
89846
98131
67588
04838
14743
23489
33102
79109
00497
64610
79848
00427
00781
61463
79422
28820
19000
94889
94068
80834
59702
74758
03444
76094
74921
54577
08815
82567
25563
18003
70225
07202
88968
39500
28210
46924
60948
62107
62056
30-34
07294
40192
91705
17056
86723
77243
76082
52503
49416
97191
42699
30376
37292
62036
40254
17520
34502
45205
09425
51126
99425
31778
52073
32550
57842
80154
59933
58050
53381
49337
06088
31508
51144
55743
96579
17735
47525
46049
33122
91915
49581
47597
99515
14595
87351
34075
60839
18685
60885
57390
35-39
84149
51153
86224
72506
91691
76690
79172
24130
44448
53798
21753
76970
05967
49967
26065
95401
75508
87440
26904
.79935
62291
29555
64280
57158
90382
91486
72998
16899
23656
88674
99462
60935
18827
74701
54853
42201
77705
71186
14124
18431
45378
33137
62282
63836
36637
33272
55437
48992
37503
42746
40-44
46797
11397
29503
82949
13191
42507
57913
72S24
04269
12693
76192
23063
26002
01990
07938
55811
06059
96767
07419
57450
86943
61446
18847
58887
77019
19180
99942
79952
75787
35355
56705
22390
10704
58851
78339
80540
95330
34650
46240
92978
99878
31443
53809
77716
42833
00840
03183
19462
55461
39272
45-49
82487
58212
57071
54600
77212
84362
93448
21627
27029
27928
10508
35894
51945
72308
76236
76099
86682
67042
76013
55671
21541
23037
24768
73041
24210
15100
10515
57849
59223
12267
06118
52246
76803
27427
20839
54721
21866
14994
92973
11591
61130
51445
26685
79596
71875
73268
13191
96062
71213
96659
-------
546 Appendix Tables
TABLE A 1— (Continued)
^50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
SO
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
50-^54
32847
16916
66176
46299
22847
41851
28444
47520
34978
37404
32400
89262
86866
90814
19192
77585
23757
45989
92970
74346
87646
50099
10127
67995
26304
81994
59537
51228
31089
38207
88666
53365
89807
18682
63571
68927
56401
24333
17025
02804
08298
59883
469 S2
31121
97867
57364
09559
53873
35531
28229
55-59
312S2
00041
34037
13335
47839
54160
59497
62378
63271
80416
65482
86332
09127
14833
82756
52593
16364
96257
94243
59596
41309
71038
46900
81977
80217
41070
34662
10937
37995
97938
31142
56134
74530
8 1 038
32579
56492
631S6
95603
84202
08253
03879
01785
06682
47266
56641
86746
26263
55571
19162
88629
60-64
03345
30236
21005
12180
45385
92320
91586
98855
13142
69035
52099
51718
98021
08759
20553
56612
05096
23850
07316
•40088
27636
45146
64984
18984
84934
56642
79631
62396
29577
93459
09474
67582
38004
85662
63942
67799
39389
02359
95199
52133
20995
82403
62864
07661
63416
08415
69511
00608
86406
25695
65-69
89593
55023
27137
16861
23289
69936
95917
83174
82681
92980
53676
70663
03871
74645
58446
95766
03192
26216
41467
98176
45153
06146
75348
64091
82657
64091
89403
81460
07828
75174
89712
92557
90102
90915
25371
95398
88798
72942
62272
20224
19850
96062
91837
02051
17577
14621
28064
42661
05299
94932
70-74
69214
14253
03193
38043
47526
34803
68553
13088
05271
49486
74648
11623
27789
05046
55376
10019
62386
23309
64837
17896
29988
55211
04115
02785
69291
31229
65212
47331
42272
79460
63153
89520
11693
91631
09234
77642
31356
46287
06366
68034
73090
03785
74021
67599
30161
49430
75999
91332
77511
30721
75-79
70381
76582
48970
59292
54098
92479
28639
16561
08822
74378
94148
29834
58444
94056
88914
29531
45389
21526
52406
86900
94770
99429
33624
27762
35397
02595
09975
91403
54016
55436
62333
33452
90257
"'2~'''3
94592
54913
89235
95382
16175
50865
13191
03488
89094
24471
87320
22311
44540
63956
24311
16197
80-84
78285
12092
64625
62675
45683
33399
06455
68559
06490
75610
65095
79820
44832
99094
75096
73064
85332
07425
25225
20249
07255
43169
68774
42529
98714
13513
06113
95007
21950
57206
42212
05134
05500
91588
98475
91583
97036
08452
97577
57868
18963
12970
39952
69843
37752
15836
13337
74087
57257
78742
85-89
20054
86533
22394
63631
55849
71160
34174
26679
44984
74976
69597
73002
36505
65091
26119
20953
18877
50254
51553
77753
70908
66259
60013
97144
35104
45148
86197
06047
86192
87644
06140
70628
79920
80774
76884
08421
323
-------
Appendix B
COMPUTATION OF CONFIDENCE INTERVALS
Model:
Assumptions: j driver has probability q. of gassing up on any day.
J
Given that he gasses up, he gets gas at station s with probability p.(s)
J
where s denotes the station I.D.
Assume that
Zp,Cs) = 1
s J
That is, we havp included in our model all stations that have a non-zero
probability of use by a driver in our population.
Let
{1, j driver is switcher
0, otherwise
Let J = total number drivers driving catalytic
converter cars.
The proportion of switchers is
P -
Observe stations s,, s_, ..., s, selected at random. At station s. observe
total of n. catalytic converter cars, and m. switchers. At each station the
duration of observation is the same. Consider the ratio
-------
k
£ -i
The question is: what is R an estimate of: To answer this, define
q(SW) = av. q. for switchers
J
q(NS) = av. q. for non-switchers
J
We allow the possibility that
q(SW) / q(NS).
That is, the average probability that a switcher gets gas on a given day
may be different than the average probability for a non-switcher. Intuitively,
one may suspect that a switcher may be a larger gas consumer and be likely
to make more frequent visits to gas up. We will show that R is an asymptotically
unbiased estimate of
R* = P q(SW)
Pq(SW) + (l-P)q(NST
Thus, we get, not an estimate of P, but an estimate of P weighted by
(approximately) q(SW)/q(NS). That is, if we assume P is small, then
-------
If we assume q(SW) = q(NS) then R* = P.
The fact that R* gives a weighted estimate of P is not surprising. If
switchers turn up more frequently at service stations than non-switchers,
they will be observed more often.
To derive the above result and also to derive the distribution of the
estimate R, we assume that the stations are sampled with replacement.
This is a good approximation if the number of stations surveyed is small
compared to the total number of stations. Let S be the total number of
stations, let n , m be the random variables defined as the total number
of catalytic converter cars and switchers we would observe if station s
were observed at a fixed duration period (i.e., two hours) randomly
selected during the day on a randomly selected day. Let N , M be the
numbers of catalytic converter cars and switchers we would observe at
stations s if observed over the full duration of randomly selected day.
Assume that
Ens = a ENS
Ems = a EMs
where E denotes expectation.
Now, note that
-------
Let s* be a random selection from the set {1, 2, ...,S} of all service
stations.
Then
ENS* =
-------
30 k
ECS "U
-Jr-!- • R*
ECE .,)
The next thing we want to derive is the value of
L = E(R-R*)2
or, at least, an upper bound for L. Denote
k
Y = £ n.
• i i
Then
L2 = E(* - f*f
= C/X(EY) - Y(EX)V
\ /
Denote X = X - EX, Y = Y - EY
Then
,2 _ ../X(EY) - Y(EXT2
\ YEY
Use the inequality
E(U + V)2 < (vEV2 +
-------
to get
/~9
< vi(x/Y)2
—
Since Y is a sum of K identically distributed independent random variables,
to a first order, we assume
Y ^ EY
to get
i/ < KvTx)
where V denote variance.
Now
EX = kEm* , EY =
V(X) = kV(m J V(Y) = kV(ns,
Thus
L<
Em
En
En
a(n
-------
Once the survey is finished, the values in this bound for L may be replaced
by sample estimates, i.e., R* may be replaced by R, the standard deviations
a(m *) and a(n *) may be replaced by the sample standard deviations, etc.
To get a rough a priori estimate of the accuracy of the survey, we assume
that all n., m. are Poisson variables with means y , A not depending on i
and A= yP. Then we get, assuming P small, say P <_ .3,
, : R* _ R*
and assuming, R* = P, we get
where n is the number of catalytic converter cars observed. This bound is
low because of the assumption of the equality of the A-. Take the bound
fp
to be 2y — even though this is probably too high. To get some idea of
sample size requirements, suppose P = .25, then the bound is l//n" . If we
want the 95% confidence interval to be
.25 + .05
then 2//rf= .05 or
n = 1600
This is probably too large a sample size to go for and our opinion is that
a sample size of 1000 cars is sufficient. If 100 catalytic converter cars
are observed per day at 4 different stations - this gives a 10 day survey
survey covering 40 stations. This seems reasonable and feasible.
-------
Appendix C
BIAS COMPUTATION
Define
P. = probability of incorrect registration
identification
PQ = probability of incorrect license
number recording
Q = proportion of all vehicles that have
catalytic converters
To compute: given a non-catalytic converter car, what is the
probability P that it is identified as a
catalytic converter car?
Two possiblities: registration wrong
recording wrong
Even if a non-catalytic car is incorrectly identified for either of the
two above reasons, it can still be identified as a non-catalytic converter
car. Thus, we will assume that if a non-catalytic converter car is
incrorectly identified, the probability that it is identified as a catalytic
converter car is equal to the proportion Q of catalytic converter cars.
in the entire population. Therefore
PCC = Q Pr (incorrect identification).
Now, assuming independence,
Pr (correct identification) = (1 - PL)(! - PQ)
so
Pr (incorrect identification) = 1 - (1 - PL)(! - P )
• PL + P0 - PLP0 '
' PL * P0
-------
The bias B is the proportion of cars identified as catalytic converter cars
that are actually non-catalytic converter. Denote
n = true number of non-catalytic cars
observed
N = true number catalytic cars
observed.
Then the number of cars identified as catalytic cars that are non-catalytic
cars is given by n P
L. L*
Let N be the apparent number of catalytic cars observed. Then
B = n PCC/N
Now n is unknown, but if P is reasonably small, then n = n, where n is the
apparent number of non-catalytic cars observed. Therefore
B * n QP/N
For P not small, a more exact value for B can be derived.
-------
Appendix B
United States
Environmental Protection
Agency
Region 10
1200 Sixth Avenue
Seattle, WA 98101
FUEL SWITCHING SURVEY
Region T1rne_ Inspection Number
Beginning End
STATION INFORMATION Inspection Date ^ o>v ywr
Name of Station Brand of Gasoline
Address of Station Urban Residential Rural
Strwt
Person in Charge
CSV
Gasoline Price Regular Unleaded Premium Diesel Manager
County [Zip Code Business Telephone Office of Manager Telephone
SURVEY INFORMATION
License Number
State
Make
Mode!
Passengers
Full Serve
Self Serve
Gas Type
-------
Appendix C
Audit Results
TACOMA
DOUBLE OBSERVATIONS
Dual Obsv./ Audited
Error Vehicles
1 1
4
2
2
2
2
1
4
1
7
3
1
3
1
7
2
6
20
25
4
21
3
2
2
2
4
2
6
52
19
18
1 2
27
5
18
16
12
1 13
22
6
6
Sta. ID. No.
WA-TA-0010
WA-TA-0019
WA-TA-0020
WA-TA-0021
WA-TA-0023
WA-TA-0025
WA-TA-0026
WA-TA-0032
WA-TA-0033
WA-TA-0035
WA-TA-0036
WA-TA-0037
WA-TA-0041
WA-TA-0042
WA-TA-0043
WA-TA-0045
WA-TA-0051
WA-TA-0054
WA-TA-0057
WA-TA-0061
WA-TA-0064
WA-TA-0067
WA-TA-0071
WA-TA-0075
WA-TA-0089
WA-TA-0092
WA-TA-0095
WA-TA-0097
WA-TA-0101
WA-TA-0104
WA-TA-0105
WA-TA-Oin
WA-TA-0124
WA-TA-0133
WA-TA-0134
WA-TA-0148
WA-TA-0161
WA-TA-0172
WA-TA-0182
WA-TA-0186
WA-TA-0231
Total
Vehicles
23
27
20
18
17
11
5
15
32
48
3
6
15
22
35
11
19
20
25
18
21
6
2
6
17
4
2
6
52
19
18
9
27
5
18
16
12
13
22
6
6
Dual
Audi t/Error
1
1
2
1
1
1
1
1
1
1
INDEPENDENT RANDOM AUDITS
No.
1
1
5
3
2
3
T5
WA-TA-0002
WA-TA-0003
WA-TA-0006
WA-TA-0129
WA-TA-0147
WA-TA-0223
0.0% error
0
0
0
0
0
0
677
TT
1.4% error
1.6% error
-------
Audit Results
DOUBLE OBSERVATIONS
Dual Obsv./ Audited
Error Vehicles
30
4
41
6
0
26
3
20
14
16
7
39
14
26
29
8
15
14
16
21
44
41
33
2
1 20
27
1 46
8
16
35
2 26
4 647
Sta. ID. No.
WA-SP-02
WA-SP-07
WA-SP-08
WA-SP-33
WA-SP-40
WA-SP-50
WA-SP-65
WA-SP-67
WA-SP-68
WA-SP-75
WA-SP-76
WA-SP-78
WA-SP-82
WA-SP-83
WA-SP-85
WA-SP-107
WA-SP-108
WA-SP-109
WA-SP-114
WA-SP-128
WA-SP-134
WA-SP-145
WA-SP-148
Wa-SP-177
WA-SP-188
WA-SP-194
WA-SP-203
WA-SP-207
WA-SP-210
WA-SP-222
WA-SP-226
SPOKANE
Total Dual
Vehicles Audit/Error
30
4
41
25
85
26
3
20
14
16
7
39
14
26
29
8
15
14
16 1
21
44
41
33
2
20 1
27
46 1
8
16
35
26 3
751 6
INDEPENDENT RANDOM AUDITS
No.
Observ.
6
6
41
46
99
Sta. ID. No. Errors
WA-SP-02
WA-SP-08
WA-SP-145
WA-SP-203 J_
1
1% error
0.6%
0.8%
-------
Audit Results
DOUBLE OBSERVATIONS
Dual Obsv./ Audited
Error Vehicles
60
48
16
8
16
38
48
2 49
69
56
8
62
14
2 25
49
54
67
37
22
32
4
39
1
23
4
25
13
43
21
32
4
4 987
Sta. ID. No.
WA-SE-16
WA-SE-22
WA-SE-23
WA-SE-27
WA-SE-34
WA-SE-38
WA-SE-60
WA-SE-62
WA-SE-67
WA-SE-74
WA-SE-79
WA-SE-83
WA-SE-102
WA-SE-104
WA-SE-109
WA-SE-131
WA-SE-132
WA-SE-139
WA-SE-147
WA-SE-154
WA-SE-159
WA-SE-170
WA-SE-188
WA-SE-190
WA-SE-192
WA-SE-194
WA-SE-195
WA-SE-288
WA-SE-289
WA-SE-500
WA-SE-501
SEATTLE
INDEPENDENT RANDOM AUDITS
Total Dual No.
Vehicles Audit/Error Observ. Sta. ID. No. Errors
60 6 WA-SE-67
48
16
12
20
38
48
49 2
69
56
8
62
14
25 1
49
54
67
37
22
32
42
39
30
23
10
25
27
43
21
32
4
1W2 3
0.4% error
0.3% error
-------
Audit Results
DOUBLE OBSERVATIONS
Dual Obsv./ Audited
Error Vehicles
8
12
1
16
9
39
r-\
o
6
14
16
10
1 8
4
13
16
Sta. ID. No.
OR-EU-0002
OR-EU-0027
OR-EU-0041
OR-EU-0042
OR-EU-0055
OR-EU-0062
OR-EU-0063
OR-EU-0065
OR-EU-0068
OR-EU-0076
OR-EU-0078
OR-EU-0079
OR-EU-0082
OR-EU-0094
OR-EU-0098
EUGENE
Total Dual
Vehicles Audit/Error
8
12
1
16
9
39
8
38
14
16
10
8
4 1
13
16
180
212
1
INDEPENDENT RANDOM AUDITS
No.
Observ.
3
1
2
5
5
5
_6
27
Sta. ID. No. Errors
OR-EU-0011
OR-EU-0018
OR-EU-0031
OR-EU-0033
OR-EU-0066 1
OR-EU-0088
OR-EU-0089
0.6% error
0.5% error
-------
Audit Results
DOUBLE
Dual Obsv./ Audited
OBSERVATIONS
Error Vehicles Sta. ID. No.
4
3
5
9
6
47
13
17
17
8
5
3
5
2
9
1 28
5
6
OR-PO-22
OR-PO-42
OR-PO-78
OR-PO-101
OR-PO-107
OR-PO-110
OR-PO-124
OR-PO-159
OR-PO-187
OR-PO-199
OR-PO-201
OR-PO-220
OR-PO-271
OR-PO-333
OR-PO-335
OR-PO-374
OR-PO-380
OR-PO-412
PORTLAND
Total Dual
Vehicles Audit/Error
48
21
58
9
6
43
12
17
17
8
24
13
29
26
21
30
28
192
INDEPENDENT RANDOM AUDITS
No.
Observ. Sta. ID. No. Errors
5 OR-PO-00
4 OR-PO-02
5 OR-PO-05
5 OR-PO-17
2 OR-PO-47
4 OR-PO-69
1 OR-PO-70
5 OR-PO-76
3 OR-PO-80
5 OR-PO-139
5 OR-PO-176
6 OR-PO-205
3 OR-PO-295
3 OR-PO-340
3 OR-PO-346
3 OR-PO-382
_2 OR-PO-415
74
0.0% error
1.04 % error
-------
Audit Results
BOISE
DOUBLE OBSERVATIONS
Dual Obsv./ Audited
Error Vehicles
10
6
6
6
10
10
21
7
6
2
6
0 90
Sta. ID. No.
ID-BO-48
ID-BO-72
ID-BO-79
ID-BO-96
ID-BO-102
ID-BO-107
ID-BO-m
ID-BO-217
ID-BO-277
ID-BO-1050
ID-BO-79
Total
Vehicles
10
16
6
12
10
10
21
57
56
2
7
207
INDEPENDENT RANDOM AUDITS
Dual No.
Audit/Error Observ.
2
2
1
5
8
6
2
3
1
30
0
Sta. ID. No.
ID-BO-01
ID-BO-04
ID-BO-17
ID-BO-43
ID-BO-138
ID-BO-388
ID-BO-1015
ID-BO-004
ID-BO-09
0.0% error
Errors
0
0.0% error
0.0% error
-------
Tec/mofogy Service Corporation Appendix D
2811 WILSHIRE BOULEVARD • SANTA MONICA, CALIFORNIA 90403 • PH. (213) 829-7411
STATISTICAL DESIGN AND METHODOLOGY
FOR THE ANALYSIS OF THE 1979
REGION X FUEL SWITCHING SURVEY
TSC-PD-A223-4
March 1980
Peter Blckel
Leo Breiman
Contract Mo. 68-01-5086
Task 4
Submitted to:
U.S. Environmental Protection Agency
401 M Street, S.W.
Washington, D.C. 20460
Task Officer: Mel Kollander
-------
CONTENTS
1. INTRODUCTION 1
2. BACKGROUND AND DESIGN 3
3. SOME SUMMARY CONCLUSIONS 6
Appendix
A. Computation of Estimates, Biases, Variances and Confidence
Level s '. 10
B. Analysis of Service Station Location and Characteristics
on Fuel Switching 25
iii
-------
1. INTRODUCTION
Under contract number 68-01-5086, Technology Service Corporation
was requested to provide statistical support to the personnel of the
Environmental Protection Agency, Region X, in the design and analysis of a
fuel switching survey to be carried out by gas station observations during
1979. The personnel involved in this project were Drs. Peter Bickel and
Leo Breiman. Leo Breiman assisted in the design of a randomized survey and
conferred with Region X personnel on methods of auditing for data quality.
The design is described in Section 2. A careful statistical analysis of
the data was carried out to determine:
1. Confidence intervals for the fuel switching rates. (Leo Breiman)
2. The effect of factors such as city, type of service,
price differential, etc., on fuel switching rates. (Peter Bickel)
Some brief summary conclusions of this analysis are given in Section 3,
By the very nature of the phenomenon being studied, the survey
design did not fit into the mold of one of the usual sampling designs.
Therefore, the statistical techniques necessary to analyze the results were
not standard. Appropriate techniques for the analysis had to be specially
derived. The description of these techniques, the derivation of the
equations, and the details of the analysis are given in Appendices A and B.
The results may be important if future fuel switching surveys are planned.
This present report is deliberately limited in its scope to cover
only those areas in which Technology Service Corporation was involved. A
-------
detailed description of the field study, the survey methods used, the practi
cal aspects of the survey, and a detailed data description will be released
by Region X. After many hours of checking into their data quality control
methods and general field operations, we are satisfied that excellent data
quality standards have been observed by the survey personnel. We want to
compliment the Region X personnel involved in the survey, especially
William Schmidt and Douglas Smith, for their open and cooperative attitude
which made it easier for us to contribute to the quality of the survey.
Also, for their dedicated concern that the survey be the best possible
within the limits of their resources.
-------
2. BACKGROUND AND DESIGN
In the usual fuel switching surveys, a team or teams visit an urban
area, make observations at service stations, and determine how many of the
observed cars that should be taking unleaded gas are actually taking leaded
gas. This is cenarally Jona by noting the type of gas purchased and the
license numbers of the vehicles. The license number is submitted to the
state Department of Motor Vehicles (DMV) to get a description of the
vehicle; that description is used to determine whether the vehicle is
designed to take unleaded gas.
The population being sampled from needs a careful def ini tn'cn.
The urban area has to be defined, as do the length of the team visit and
its hours of observation. The underlying population is the set of all
fueling visits to service stations in the area made by cars designed for
unleaded gas during the designated observation period of the team. The
population parameter to be estimated is the proportion R* of these visits
that resulted in a switch to leaded gas.
There are two major types of errors arising in such a survey:
1. Systematic errors including observer error, DMV
errors and license-plate switching.
2. Sampling errors, bias in sampling plan, random sampling
error.
The extent of the systematic errors can be estimated by procedures
which include:
» Independent observers.
• Observer descriptions of make, model, year of vehicle to
check against the DMV description.
-------
3. Descriptions of stationary vehicles, i.e.. parking lot
descriptions checked against OMV returned descriptions.
Some of tnese procedures were used in the Region X survey and are
described in their report. Either the systematic errors are kept to
within negligibly small bounds, or they need to be reliably estimated and
compensated for in the final switching estimate.
In prior surveys, teams drove around until they encountered service
stations and then observed. This made the sampling plan difficult to
analyze and introduced unknown sampling biases. In the present survey, a
more systematic plan was used. Phone books covering the areas defined were
obtained. At Limes, as in Seattle, mere than one phone bock was required.
Service stations were selected at random without replacement from the
yellow pages, using a table of random numbers. At times a service station
was listed twice, once individually and once under the name of a major oil
company. In this case, a fair coin was tossed (using the table of random
numbers) and the station included in the selection if the result was heads,
otherwise not.
The survey team went down the list of selected stations and visited
four a day, spending 1.5 hours observing at each one. At least, this was the
theoretical framework. Some modifications were made to make the plan more
feasible. To keep the travel time within reason, the randomly selected
stations were grouped into 3 or 4 subgroups by geographic proximity. The
daily route observed -i stations within the same subgroup. In addition, in
t.ne Seattle area, there was a continuing gas shortage, with long lines and
stations open only a few hours a day. >ihsn closed stations //ere encountered,
-------
they were later revisited, with the exception that if there was an open
station on the same intersection as the closed station, it was substituted
for the station.
A few stations were observed for time periods shorter than 1.5 hours.
When this occurred, the relevant counts were multiplied by the ratio (1.5/
hours observedJ.
The design of the survey was not perfect, but it was a step in the
right direction. Any design has to compromise between the limitations
of survey resources and theoretically more desirable goals. We must
emphasize that probably the most important potential source of error is not
in the sampling design, but in observer error. Unless all possible steps
are taken to minimize or estimate observer error the most sophisticated
design is worthless in terms of statistical validity.
Further improvements in the present design can be made. We suggest
a standardized two-week (10-day),30-station survey plan for urban areas.
Use standardized survey times such as 8:00AM to 10:30AM, 11:30AM to
2:OOPM, 3:OOPM to 5:30PM. Draw the thirty stations at random from a frame
of stations for the area and sequentially assign them (as drawn) to the
30 time slots in the survey period. Leave enough travel time to go between
any two pairs of stations. However, if, in addition to estimating R , a
study of the effects of various factors is desired, then the number of
stations sampled will have to be increased.
-------
3. SOME SUMMARY CONCLUSIONS
During 1979, the survey team visited six areas. Using the survey
methodology referred to in the preceding section, service stations were
selected for observation. The numbers of stations observed were:
C1ty No. of Stations Observed
Boise 87
Eugene 80
Portland 95
Seattle 61
Spokane 31
Tacoma 78
TOTAL 432
In all, a total of about 9,000 vehicles were observed, or about 21 vehicles
per station. Of these 9,000, about 3,000 were identified as requiring
unleaded cas. About 4,500 vehicles were identified as not requiring unleaded
gas. Fuel requirements for another 1,600 vehicles could not be identified
for the following reasons:
Mo. Mot Identi'~1ed_ Reason
407 No file in DMV
556 Unable to determine fuel
requirements from GMV records.
222 Cut-of-state license pi 2-5 not
checked yet.
393 Could not see license olate.
-------
In addition to a description of each observed vehicle, some character-
istics of each observed service station were recorded, i.e., price
differentials between various types of fuels, type of station ownership,
and service offered (i.e., self service, full service, etc.).
Technology Service Corporation was given the completed data file
and asked by Region X to do two tasks:
First: For each urban area, estimate the rate of fuel switching
and computer confidence intervals for the estimates.
Second: Examine the effects of a number of factors on the fuel
switching rates.
Regarding the latter analysis, the relevant factors were in two
groups:
1. Service station characteristics
2. Vehicle characteristics
The sorting-out of the effects of vehicle characteristics was sized up as
a difficult task- For instance, a description of a vehicle as a Ford
Sedan coulc include either a sedate family car or a high-performance V8
tire squealer. Furthermore, the individual vehicle descriptive data tended
to be spotty. Therefore, the analysis of the association of fuel switching
with individual vehicle characteristics was not undertaken. If desired, it
can be done at a future date-
As previously mentioned, neither of these two above tasks could
be done by straightforward applications of known statistical models.
Therefore, a technical framework had to be specially built. The derivation
-------
of the confidence Intervals is given in Appendix A. In particular, page
13 gives the formula used for the 95% confidence interval computation.
*
With these formulas, the data was used to compute estimates of R and 95%
confidence intervals for each of the six urban areas surveyed.
It is important to Keep in mind the definition of the parameter R*
being estimated. It is NOT the proportion of drivers that are fuel switchers
neither is it the proportion of vehicles that are using the wrong "uel . R"*
is the proportion o~ all service-station visits by cata lyt-'c-converter-
equipce-d vehicles in the urban area during the observation period that
results in t.ie purchase of "leaded gasoline. With this definition R* can
be estimated by the survey. "Table 1 below gives a summary of the resulting
st^-ates -KG v^.-
Ci ty
Soise
Eugene
Portland
Seattle
Spokane
Tacoma
con- iaen
-
R* Es:
13
3
6
7
-j
a.
ce ^terv
-3 La '. •
timateU )
.1
.4
.0
.9
.3
.3
95'o Confidence Intervals
(13.7, 22.5)
( 5.2, 11.6)
( 4.3, 7.7)
( 5.2, 10.5)
( 3.5, 11.1)
( 2.3, 7.2)
i ne Region X Assort v/i ; j^'/e more oackgrour.G c
Boi^e. W"'th t1"!0 ;2xc2nt i en o~ Boise, al ccMT^c
t.",e "i''.if"! estimated
-------
Because of this, we decided to merge the Oregon-Washington data and produce
an overall estimate and confidence interval based on the five urban areas.
The result was:
Estimated R*(iX) 95% Confidence Interval (%)
6.9 (5-7, 8.1)
Next, we analyzed the effects of the following service-stations
characteristics:
Location: Which of the six sites the stations were located in.
Price Differential: The difference in price per gallon between
unleaded and regular gasoline.
Service: Self service or full service.
Major/Minor: Classification according to membership of the oil
company in the seven major oil companies or as a
minor oil company.
Because of possible interactions between these factors (i.e., no self
service in Portland), it might not be appropriate to analyze the effect
of each factor separately. Therefore, a method was developed for looking
at the effects of various combinations of factors in a four-dimensional
2
X analysis. The discussion of the statistical model and derivations
of the various statistical tests are given in Appendix B. The overall
qualitative conclusions of the analysis are:
1. There is strong evidence that the switching proportion depends
on the city under study or, at least, that Boise is different.
2. There is weak evidence that, in addition, the proportion of
switches is higher at self- than full-service stations.
3. There is no evidence that price difference or major/minor
classification plays a role, but a substantially larger, care-
fully stratified survey would be needed for further analysis.
-------
10
Appendix A
COMPUTATION OF ESTIMATES, BIASES,
VARIANCES AMD CONFIDENCE INTERVALS
by Leo Breiman
The statistical framework is as follows: Let there be a population
S of service stations 5 in a given urban area. Denote by
-------
11
proportion of fuel switching in the travel time periods differed from the
proportion in the observed periods. Therefore, a reasonable interpretation
of R* is that it is the proportion of daylight visits to service stations
in the frame during the period of the team's visit to the area that
resulted in switching.
Conceptually, the sampling plan can be modeled by assuming that J
stations were selected, without replacement, from the frame S, and the j n
* t H
station selected s. was observed in the j" interval. Let
J
xj - Vsj> and YJ • Vs?
be the number of observed switches and visits, respectively, in the j
period and denote
J ,1
X = £ X Y = £ Y
so that X is the total number of observed switches, and Y the total
number of observed visits by cars designed for unleaded gas. The
*
estimate we use for R is
R = X/Y
which is a ratio estimate and is generally biased. However, letting
the total number of stations in the frame be N, note that
EX = 1 E£x.(s)
s j J
-------
12
so that
R" = EX/EY
where E denotes expectation.
In the following section, approximate bounds for the variance and
bias of R are derived. The bounds for the bias are derived under two dif-
ferent sets of assumptions. When the two bounds were computed using the
sample data, both resulted in values so small as to be negligible. "The
approximate'bound for the variance is
Var(R) < -y Z (X. - RY.)2
r j J J
This result is similar to the standard variance computation for ratio
estimators. See, for example, W. C. Cochran, Sampling Techniques, Wiley,
1953, pp. 22-23.
In fact, the survey structure is similar to a household survey.
Suppose there is a population of H households (service stations); ] of
them are selected at random without replacement, and the ju household has
"lemcers of .vrcm '(.. a^e s.-.": tehees. The jur.cse :s to estimate tne
overall fopcrtion of sv/itcners.
The ci ^ference oet'.vesn tne oresen~ s'j^'/ey anc a stancard ncusenol-
;jv--;3V -;; --a- - - r ^ = ^-~ " t u > a" c 1 c . tne values r"7 '^ ar-c ' . rnance over
-------
13
the 0 observation periods. Therefore, the present survey can be conceived
of as a household study in which a household is selected from the population
of all households and then a period j is selected without replacement
from the set J of all observation periods to observe the selected
household.
/\
Making the usual assumption that the ratio R has an approximately
normal distribution, the 95?o confidence intervals were taken to be
and the bound given above was used in the computation.
A.I VARIANCE AND BIAS COMPUTATIONS
Now the problem becomes to derive bounds for the bias and varianc
^.
of the estimate R. First, we work on the variance. Let
V = E
X r/X
— c —
Denote the bias by b. Then we note that
/Y *\'L 9
MSE = E £ - R = V + b
It is easier to compute
- R
-------
14
2
and subtract b than to estimate V directly. Hence, we will work on
getting bounds for the MSE. Write
^2
/ : v "
MSE = E A '
'Jsinc a customary accrcximation
,-.-:, v ^2
MSE = h;X"!V
(EYT
As 'jSLa": , we replace cv by the sarnole vai-je l/, and work 3n the numerator
and define
Z,(s) - X.(s) - R*Y,(s), Z. - Z.(s*
J «J J J o J
V He 9
(X - YS )"
-
-------
Now, for j r k
15
Zj(S)Zk(,-)
N-l
SO
Using the fact tha.t EZ. •= 0 gives
E Z (s) Z (s)
jfk J K
Define
Z(s) =
so that
-------
16
= I>r(Z ) -^
v^
Evar(Z.-
< JL y^z 2
i,-, y*i
Replace 2J EZ-1" with "he samole value
•i ^
j
E(X. RY
J J
~o get tne approximate bound
% E(X, - RY.)2 - (A-2)
v'- -; J J
Ins acc't^cna; zcss^z^- S33^mo~";':n ~s "hat cne J values
-------
17
[hen an estimate for
is given by
Using this in (A-2) gives the more familiar bound
V ±~- Z(X. - RY.)2 (A-3)
Y i
where
f = J/N-1
However, this assumption is not an altogether comfortable one, and our
opinion is that (A-l) is a sounder value.
The bias computation is more complicated. Write
X. = X = _X_ A Y-EY
Y EY+(Y-EY) EY V EY
-------
18
"akirc expectations gives
X \ _ .,* r ( X, Y)
" (EY)2
where r denotes covariance. So, to a first approximation
o =
As before, use the aporcximation
0 —
(Y)
2
Mow
E
r(x., Yk.)
:or ; = '<, -jsing trie following express-on
1 v •/ \
^Aj>/,<; -
' V^ -1 ' \ M 1 ^
?l(N-l; ,fr_ Aj^^ 'k(i '
3 rS
•~A] ' •"' '<''
-------
19
j = k,
Therefore, defining
X(s) =
Y(S) = EY,(S)
K
r(-x,Y) -
J J
EY,
E
j
X(s)Y(s).
i E L X . (s) Y . (s)
j s
Simplifying this gives
F(X,Y) =
(EX)(EY)-
-------
2C
ihere are two sets of assumptions that can lead to estimates of the
bias. The first, and more stringent, is that
Assumption I:
Er(xn.,Y ) > o
•i *J J
The rationale for this assumption is that there usually is a positive
dependence between X^ and Y.. That is, tne larger the number of visits to
a station, then tne larger (generally) the number of switches at that
station. Under this assumction,
T(X.V) 2^ (EX)(EY) - ir7Tf-r !>( s ) Y ( s ) (A-4)
IN- I rl i, .'I- l ;
•-y
Viithout any assumptions, we have
and assume, ir accition, tnat
- '(,5 - '<;s;:c''s - r.s-: > 2
-------
21
This is a weaker assumption. Its meaning is that at a fixed station
(generally) there is a positive dependence over time between number of
visits and number of switches. The above assumption can be rewritten as
ZX(s)Y(s) £j £ X (s)Y i
s j,s J J
substituting this into (A-4), page 20, gives
r(X,Y) ^~^- (EX)(EY) - ^y
J
Replacing expressions by corresponding sample expressions gives
1 C v J
," T A I ~ / ir
and
There is another approach which substitutes a different first
assumption. If we keep Assumption II, then start with
F(X,Y) > ^ Lr(X,,Y.) + i^y (EX)(EY) -
j
-------
Put R. = EX./EY. and write
•J w \J
so that
22
- EYJ!
:
-------
23
- 1N-' (N-l)Y'
Under Assumption I1 and II
_ x. - RY.)Y.
(N-l)Y^ j J J (N-l)Y^ j J J J
The advantage of Assumption I' is that it can be tested using the day.
The service stations observed can be grouped in four subsets, depending
on whether they were observed 1st, 2nd, 3rd or 4th during the day. Denote
these subsets by C,, Co. Co, C*. For each subset define
Y, = . Y-, X =
J
Then the issue is whether the R. = X./Y. can be consisered as
approximately equal. This can be framed statistically as: Define the
four sets of variables
U. k = X,/Y,.. jsC k = 1,2,3,4
J i N J K
When do the four sets come from the same underlying population versus
shift alternatives? Although for fixed k, the LJ. , are not quite in-
j» K
dependent, independence is a reasonable first approximation and standard
tests can be applied; in particular, the Kmska!-Wall is test is an
appropriate candidate.
-------
In our computations on the bias bounds, both sets of bounds
produced the conclusion that the bias was negligible. (See Table A-l,
following. )
TABLE A-l
First Sias Sound Second Bias Sound
Boise
Ejgene
Portland
Seattle
Spokane
Tacoma
.002
.0006
.00004
.00003
.0002
.0001
.002
.0004
-.00003
-.0001
-.0002
-.00003
-------
25
Appendix B
ANALYSIS OF SERVICE STATION LOCATION AND
CHARACTERISTICS ON FUEL SUITCHIiNG
by Peter Bickel
In this appendix we analyze the effect of different characteristics
of service stations on the proportion of visits to these stations resulting
in switching. The characteristics (factors) considered were:
C1ty: Which city the stations were located in
Price Differential: The difference in price per gallon between unleaded
and regular gasoline
Service: Self service or full service
Major/Minor: Classification according to membership of the oil
company in the seven major oil companies facrn-rdina "?•
J. M. Blair, The Control of 01i , Pantheon Publishers)
The categories (levels) of the factors ana tneir numoer were:
City (6) Boise, Eugene, Portland, Seattle, Spokane, and Tacoma
Price (3) up to 5c, over Si, and difference unknown
Service (2)*Self or full service
Major/Mi nor(3) Major, minor and category unknown
As usual, we call the collection of all service stations corresponding
to a particular combination of levels a cell—for instance, all stations
in Boise, with a price difference of over 5e, offering full service and
representing a major oil company. There are a total of 69 out of
the 108 possible for which at least one representative appeared
Full- and self-service departments of the same station were
treated as separate stations.
-------
26
in the sample. Twenty-five of the missing 39 cells corresponded to "missing
value" categories ">; the price and major-minor classification, 3 to the
lack of self-service stations in Portland and Eugene. Unfortunately, in
addition to these 3 missing cells there are 11 cells in which there are
4 or fewer visits (out of a total of 2546). This lack of balance combined
with low switching rates makes conclusive analysis difficult. Our
decision is based on a simplification of the model introduced in
Aopendix A of this report by Leo Sreiman (hereinafter referred to as (B)).
(3) associated witn the sampling universe of all fueling visits to stations
in the area (made by candidate cars) a parameter R , the proportion of all
visits that resulted in switchings. Write
^ =E Ms) Ms)
s
where (in the notation of (3))
" X,(s)
* 1 = 1
* (s) -^
s j =
/ J
are, Respectively, the proportion of visits to station s and the toroorticn of
••jei sv/itcners ancnc visitors to s. The parameters we are interested in sfjdylnc
a^e, for each cell, the orooortion of switones among visits to stations in the
-------
27
R (cell) = & (s) R (s): s e cell }/£{As) : s e cell}
We prafer to interpret these quantities in a more universal framework as :
TT*(S) = probability that during a randomly selected 1.5-hour day-
light observation period a randomly selected car intends
to refuel at station s.
R*(s) = conditional probability that a car refueling at station
s will be a switcher.
Thus
R*(cell) = (conditional) probability that a car refueling at a station
in the cell is a switcher.
Let' s write:
R*(A) =y{ R*'.3) : s £ A}
^— ' J
for the proportion of switches in the stations falling in A. For
instance, R*(Boise) would be the proportion of switches in B.oise.
I. Testing: Our first aim is to study the variation of R*(cell) as we
vary the levels of the factors in each cell. Specifically, we want
to test the 15 hypotheses:
(0,0,0,0): R*(cell) = constant (none of the factors matter)
(1,0,0,0): R*(cell) depends only on city (=R*(city))
(1,1,0,0): R*(cell) depends only on city and price = R*(city, price)
-------
28
We base our tests^on (using (B's) notation):
X(cell) = Z(X.(sJ:s* z cell}
Y(cell) = Z(Y.(s*):s~ e cell}
which are the observed number of visits and of switches for stations in
the cell. (We'll also write R(cell) = Y(cell)/X(cel1), etc.) To proceed
we make two assumptions:
(3) Assumption 1': The probability that a car refueling at s during
observation period j is a switcher does not depend
on j.
Homogeneity: R*(s) = R*(cell) for all s: -: call. This is just the
hypothesis that our categorization is sufficiently fine so
that we can expect homogeneity among stations in a cell.
The second assumption is somewhat questionable, especially for the
''missing value" cells. However, it is necessary for dealing with the many
cells in which very few stations (often only 1) were represented in the
survey so that the (3) bias and variance approximations cannot realistically
be applied to Y(cell )/:<;cai1 ).
'V --\
Under tnese assumptions it follows that, given X(cell), Y(cell) has a
b'nomial (X^call); R^'call;) cistribution and the hypothesis that interests
js can, ^n orir.ciple, oe testec jsinc -r tests :"or tr.e homogeneity
2;ncrnial cistributicn.
• f ^ -
^See. fcr examola, ^. Pother and 'A. ,-ihi ttinghil 1 , '',esting for
r.cmccaneity, ' 3'cmetr-
-------
29
Such tests are of the form,
[R(cell) - RH(cell)]2
Var(cell)
•^ /x /\
where RH(cell) is the "expected" value of R(cell) under H, and Var(cell)
A
is an estimate of the variance of R(cell). The most frequently used
estimate is:
Var(cell) = X(.cell) Ru(cell)[l-Ru(cell)]
M n
We use
X(cell) R(cell) l-R(cell)*
because this is more convenient when we are looking at tests of nested
hypotheses as in this case. These denominators tend to be smaller in
2
sparse tables, such as ours. So if these x tests are not significant,
the more common ones will not be either. The statistics are referred
p
to x distributions with (# degrees of freedom) = (£ of Cfeasible) cells
(# of parameters fitted). Table B-l gives the 15 hypotheses, the
2
associated d.f., x statistics and significance probabilities.
*R(.cell) is replaced by l/2X(cell) if Y(cell) = 0.
-------
30
TABLE B-!
Hypothesis
1
1
i
1
1
1
1
0
0
0
0
0
0
0
0
f
1
1
0
0
0
0
1
1
1
1
0
0
0
0
1
0
0
1
1
0
0
1
1
0
0
1
1
0
0
0
1
0
1
0
1
0
1
0
1
0
1
0
1
0
3
1
/,
3
4
4
5
.6
7
9
0
1
4
6
7
5
4
6
9
1
LOO
-j
/
8
9
10
9
3
6
ii
9
_x
.336
- 980
-051
. 760
.538
. 739
. 725
. 791
. 4.30
. 609
. 710
. 661
. 052
. 991
-4 S 2
1
9-
4
5
2
3
9
3
9
6
7
1
9
2
S
d.
3
2
5
/,
T
5
5
6
5
6
T .
Q
0
1
2
9
2
3
i
3
60
6
6
6
6
6
6
3
7
6
3
p-val ue
0.
0.
0.
0-
0.
0-
0 .
o.
0-
0.
0-
0.
0.
0.
0.
8379
9 1 6 7 i
7465 I
7040|
8563^
7205'
7633
0658
0773
0043
OC30
0747;
0 0 7 6 ;
0065
0023
Factors are:
Column 1: Ci ty
Column 2: Pries
Column 3: Service
Column 4; Major/Minor
The simplest hypothesis that is acceptable is that R* depends only
on city (p = .76 on 63 d.f.). There is some evidence that service status
may play a role since the hypotheses that only city matters has a signi-
ficance probability of .125 on 4 d.f. under the blanket hypothesis that
only city and service type can matter. There is no evidence that any
other factors play a role.
?
ihe v~ tests have to be taken with a grain or salt, since there are
^ ^
severs: cells ror wnic.n -((cell) = 1 anc (},_, is or" the order .}}, not to
.Tier,tier, tr.e "a" 1 ibil itv of our assurnotions. -owever,
-------
31
1. The difference in significance probabilities between hypotheses
in which city (and to some extent service) is fitted and the
others is very sharp.
2. Residual analysis (see below) does not indicate any anomalies
connected with factors other than city and service.
ANALYSIS OF CITY AND SERVICE EFFECTS
In Table B-2, we give R(city) and R(city, self service), R(city, full
service) as well as the associated standard deviations of these quantities
computed according to (B) . These figures suggest:
1. That Boise is different from other cities.
2. That there is evidence for the notion that switching occurs
more frequently in self-service tnan in full-service stations.
To investigate further, we formed a.confidence interval for the
contrast, A = R*(Boise) - R* (cities other than Boise). We find:
A = .112
SD(A) = .022
To obtain a 95« confidence interval which allows for our data snooping
XV. /S /\
on the R(city) we use the Scheffe interval A - 3.31 SD(A) where 3.31 is
9
the square root of the 95% point of Xg. We obtain [.04,.182], which
implies that the proportion of switches in Boise is higher than in the
other cities.
On the other hand 95% confidence intervals for all other differences
in mean switching rates between cities cover the origin.
The values are larger than, but quite close to (within 10% of), the
values computed under the assumption that cities are homogeneous.
-------
32
TABLE S-2
Boise
Eugene
Portland
Seattle
Spokane
Tacoma
R(City)
.181
.084
.060
.079
.073
.048
SO
.022
.016
.009
.014
.019
.012
R(all cities) - .083
R(ci'ty, service) SD
Boise
full .107 .028
self .208 .028
Seattle
full .074 .021
self .084 .015
Spokane
Tacorna
full .028 .014
self .097 .024
full .035
self .057
-------
33
We list (Table B-3) the differences in estimated switching rates
R(self service, city) - R(full service, city) for all cities other than
Portland and Eugene (no self service), and the ordinary 95% confidence
intervals. These intervals should properly be interpreted at something
like the 80% level since they are being considered simultaneously (and
after data snooping!). There is weak evidence for a self-service versus
full-service effect.
Residual Analysis
~ ~ / /~ -
We examined normal plots of the differences [R(cell) R(ci ty)]/\ X(cell )
for each city. The plots revealed no overt departures from normality, al
though there was one clear outlier for Seattle, which corresponded to a
/•s
cell with unknown price difference; its omission would change R(Seattle)
negl igibly.
The residuals give no clear evidence for an association between
positive residuals and self-service stations.
The Homogeneity Assumption:
We tried to check the homogeneity assumption once we had reached
the conclusion that only city and service mattered by applying a test of
homogeneity of the service stations within each city-service cell, i.e.,
we calculated
[R(s) -
zL v
1 X(s) RH(1-RH)
-------
TABLE B-3
R(city, self service) - R(c1ty, full service) S0_
Boise
Seattle
Spokane
Tacoma
.101
.010
.069
.011
.041
.025
.029
.023
-------
35
where the summation is carried out over all stations with the city-service
2
status specified by H, and referred it to a x distribution with df =
(# of stations with the appropriate status) - 1. It turned out that the
2
X corresponding to full-service stations in all cities other than Boise
and Eugene were significant at levels ranging from .1 to .005, while those
corresponding to self-service stations were not significant. This may
just reflect the greater number of full-service stations.
/•s
In view of the excellent agreement between the variance of R as
estimated by (B) and as estimated under the homogeneity-of-cities
assumption we douot that this represents a practically as opposed to
statistically significant departure from homogeneity.
In any case, in view of the low switching rates, a substantially
larger survey with suitable stratification to ensure adequate representation
in each cell would be required to make any further analysis.
-------
TECHNICAL REPORT DATA
iPuass read Inanicnons on ;he /vi'ww before commit :ingi
i. REPORT NO.
EgA-910-9/80-078
3. RECIPIENT'S ACCESSIO^NC.
S AND SUBTITLE
"•is:," ' !
Incidence of Automobile Fuel Switching in the
Pacific Northwest 1979
S. REPORT DATE
December 1980
6. PERFORMING ORGANIZATION CODE
7.-AUTHORIS)
8. PERFORMING ORGANIZATION REPORT NO
W. Douglas Smith
9. PERFORMING ORGANIZATION NAME AND ADDRESS
U.S. Environmental Protection Agency
Region X, Surveillance & Analysis Division
1200 6th Avenue
Seattle, WA 98101
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
12. SPONSORING AGENCY NAME AND AOORESS
13. TYPE OF REPORT AND PERIOD COVERED
14. SPONSORING AGENCY CODE
15. SUPPLEMENTARY NOTES
16. ABSTRACT
the incidence of switching from unleaded to leaded gasoline for catalytic
converter equipped vehicles was observed in six (6) major metropolitan areas of
Region X. The average rate was 6.9%, with a 95% confidence interval (5.7, 8.1)
was documented.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lOENTIFIERS/OPEN ENDED TERMS
COSATi Field/Group
Fuel Switching
Unleaded Gasoline
Northwest
Region X Study
3. DISTRIBUTION STATEMENT
Unlimited
I 19. SECURITY CLASS iTills Report/
! Unclassified
21. NC. OF PAGcs
I 20. SECURITY CLASS , This p^f
Unclassified
22. PRIGS
= fft, form 2220-1 (9-73)
------- |