Appendix B
Estimation Methodologies
Introduction
Statistics
concerning vehicle miles traveled (VMT), vehicle fuel efficiency
(given in terms of miles per gallon (MPG)), vehicle fuel consumption,
and vehicle fuel expenditures are presented in this report. The
methodology used to estimate these statistics relied on data from
the 1993 Residential Energy Consumption Survey (RECS), the 1994
Residential Transportation Energy Consumption Survey (RTECS), the
U.S. Environmental Protection Agency (EPA) fuel efficiency test
results, the U.S. Bureau of Labor Statistics (BLS) retail pump price
series, and the Lundberg Survey, Inc., price series for 1994.
The
estimation of these four statistics (VMT, vehicle fuel efficiency,
vehicle fuel consumption, and vehicle fuel expenditures) occurred
in several steps (Figure B1). First, for each RTECS vehicle, the
VMT were determined from two actual odometer readings or imputed
using data from the 1993 RECS. Second, the annual on-road fuel efficiency,
given in terms of MPG, was estimated using the questionnaire responses,
decoded Vehicle Identification Number (VIN) data, EPA fuel efficiency
test results, and the months that the vehicle was in use. The MPG
were adjusted to account for the difference between EPA test values
and on-road values. Third, estimated vehicle fuel consumption was
derived by dividing the VMT by the estimated MPG. Finally, the estimated
vehicle fuel expenditures were derived by multiplying the vehicle
fuel consumption by the fuel price. The 1994 RTECS, like the 1988
RTECS, did not collect vehicle fuel prices via fuel purchase diaries.
Instead each RTECS vehicle was assigned a price based on reported
fuel type used in each vehicle. Gasoline prices were obtained from
the BLS 1994 Retail Gasoline Pump Price Series. Diesel fuel prices
were obtained from the Lundberg Survey, Inc. (See "Other Fuel
Types" in this appendix for a discussion of the gasohol and
propane prices.)
The
following sections of this appendix describe the estimation procedures
used for calculating the VMT, the MPG, the vehicle fuel consumption,
the vehicle fuel prices, and the vehicle fuel expenditures. Also
described in this appendix are the sources of data that were used
in the estimation procedures.
The
following terms are used throughout this appendix:
Terms |
Definitions |
EPA
Composite MPG |
The
EPA dynamometer test procedure, performed on preproduction prototype
vehicles, yields separate test values for EPA city and highway
MPG. These city and highway MPG are often combined to form the
"composite" MPG. |
On-Road
MPG |
A
composite MPG that was adjusted to account for the shortfall
between the test value and the fuel efficiency actually obtained
on the road. The adjustment did not take into account the driving
patterns of individual drivers and seasonal differences. |
In-Use
MPG |
MPG
that were adjusted for seasonal differences and annual miles
driven. Vehicles that are driven relatively few miles during
the year are assumed to be driven mostly on short trips that
involve frequent stops. Vehicles that are driven relatively
many miles are assumed to be driven mostly on long trips where
few stops are needed. |
MPG
Shortfall |
A
measure of the difference between actual on-road MPG and the
EPA laboratory test MPG. Expressed as the ratio of test MPG
to on-road MPG. |
Vehicle
Miles Traveled
When
possible, VMT were determined for a sample vehicle by taking the
difference between two odometer readings, which spanned a period
of time. This method was used to determine VMT for 2,648 (48 percent)
of the 5,553 RTECS sample vehicles. Attempts were made to obtain
odometer readings during the RECS interviews, the End-of-Year (E-O-Y)
RTECS interview and any time a vehicle was acquired or disposed.
A "span" of odometer readings was the difference between
two odometer readings. In most cases, this span was a B-O-Y to E-O-Y
span, although due to an occasional nonresponse, only shorter spans
were obtained, such as RECS to B-O-Y. Odometer spans of less than
a full year were also obtained for vehicles that were either acquired
or disposed of during the survey year.
The
VMT that were assigned to each RTECS vehicle corresponded to the
period of time that the vehicle was in possession by the sample
household. In most cases, however, this period of possession did
not correspond exactly with the beginning and ending dates for the
odometer span. This was true even for vehicles with a complete B-O-Y
to E-O-Y odometer span; because odometer cards were mailed to respondents
in several distinct waves at the beginning and end of the RTECS
survey; and because the exact dates of odometer readings were often
left to the convenience of the respondents. Therefore, all VMT obtained
from odometer spans were adjusted to correspond to the period of
time that the vehicle was in possession by the sample household.
A 2-step adjustment procedure was used. STEP 1 adjusted the odometer-span
VMT to a standard annualized mileage covering 365 days, and STEP
2 readjusted the annualized VMT to correspond to the exact period
of time that the vehicle was in possession by the household. These
adjustments took into account a typical distribution of VMT fractions
among the different months of the year. Step 2 was performed only
for vehicles that were not in the possession of the household for
the entire calendar year 1994.
STEP
1:
This
step adjusted the odometer-span VMT to a standard annualized VMT
covering a full year, regardless of whether the span of odometer
readings covered approximately 1 year or only a short span of time.
Annualized VMT for vehicle i were computed as:
Where:
Fj
= Monthly VMT fractions from the standard distribution in Table
B1
soi
= Month of starting odometer readings for vehicle i
eoi
= Month of ending odometer readings for vehicle i.
The
starting and ending Fj were prorated according to the
exact day of the month for the odometer readings. For example, if
a final odometer reading was taken on September 15, then (14/30)
x FSEP was used.
Table
B1.Distribution of Average Monthly Vehicle Miles Traveled Fractions
Month
j |
Average
VMT per Vehicle |
Fj |
January |
688 |
0.0728 |
February |
697 |
0.0738 |
March |
771 |
0.0816 |
April |
783 |
0.0829 |
May |
832 |
0.0880 |
June |
847 |
0.0896 |
July |
868 |
0.0919 |
August |
872 |
0.0923 |
September |
800 |
0.0847 |
October |
802 |
0.0849 |
November |
756 |
0.0800 |
December |
734 |
0.0777 |
Total |
9,450 |
1.0000 |
Source:
1984 Petroleum Marketing Index (PMI) Survey, NPD Research Inc. The
survey is a demographically and geographically balanced-quota sample
of 4,100 households. Respondents maintained fuel purchase diaries
for an average of 10 months. As part of the survey, information
was collected on the characteristics of trips taken in vehicles
during a designated day. Trip lengths were recorded as respondent
perception rather than from odometer readings. The distribution
of monthly mileage fractions has been obtained from this survey.
STEP
2:
Once
an annualized VMT was obtained from STEP 1 as described earlier,
it was adjusted to correspond to the time period that vehicle i
was in possession by the sample household as:
Where:
Fj
= Monthly VMT fractions from the standard distribution in Table
B1
spi
= Month starting possession of vehicle i by the household, or January
1994, whichever is later
epi
= Month ending possession of vehicle i by the household, or December
1994, whichever is earlier.
If
vehicle i was in the household for the entire year then spi
= JAN and epi = DEC. If a vehicle was acquired or disposed
of during the survey, the starting or ending Fj was prorated
according to the appropriate day of the month.
To
ensure that the distribution of average monthly vehicle miles traveled
fractions given in Table B1 reflected 1994 driving patterns, a study
of Federal Highway Administration (FHWA) Traffic Volume data for
1994 was conducted. This study resulted in FHWA VMT fractions being
constructed for 1984 and 1993. FHWA 1993 data were used since the
1994 data for the entire year was unavailable. Annual VMT calculations
were completed using both the NPD and FHWA fractions. The differences
in average annual VMT per vehicle between using the NPD and FHWA
VMT fractions ranged between 1 and 18 miles and were less than the
standard errors of the average annual VMT. Therefore, in 1994 the
NPD VMT fractions given in Table B1 will be used to compute annualized
VMT since the differences in annual VMT between using the NPD and
FHWA fractions were minor (RTECS Technote 5[3]).
Incomplete
Odometer or VMT Data
For
1,927 sample vehicles (35 percent), no odometer span was available,
although an estimate of annual VMT had been obtained from the respondent
during the RECS interview. VMT for these vehicles were imputed from
a regression on the estimated VMT obtained from the RECS. For another
978 sample vehicles (17 percent), no odometer span was available
and a VMT estimate was not obtained during the RECS interview. VMT
for these vehicles were imputed using a multiple linear regression
model, where the independent variables were number of drivers, household
income, age of household head, type of vehicle, and use of vehicle
on the job. This regression was also used for imputing VMT for vehicles
that were imputed as being acquired or disposed. Both of the regression
models described above yielded estimates of annualized VMT. The
STEP 2 adjustment described previously was then used to adjust this
VMT to correspond with the time the vehicle was in the possession
of the household.
Vehicle
Fuel Efficiency
Fuel
efficiency (MPG) must be estimated for each RTECS sample vehicle
in order to estimate each vehicle's fuel consumption for the survey
year. (Fuel consumption is estimated by dividing the VMT for time
of possession, by the MPG.) Prior to 1988, the RTECS obtained actual
fuel consumption data and on-road MPG from fuel purchase diaries
maintained by the respondents. However, no fuel purchase diaries
were used in the 1988 thru 1994 RTECS. Instead, the 1994 MPG were
estimated using EPA laboratory test MPG that were adjusted to account
for differences between actual on-road MPG and the EPA test MPG.
This difference is known as MPG "shortfall." The feasibility
of using shortfall-adjusted MPG in an RTECS survey was investigated
by Lax, 1987[6]. That study verified that the method yielded unbiased
MPG, when using a data base from a 1984 fuel purchase diary study
performed by NPD Research, Inc. The adequacy of current shortfall
adjustment methods is sufficient for late 1980 through early 1993's
motor vehicle model years also (RTECS Technote 5[3]).
The
RTECS sample vehicles were assigned EPA test MPG from the EPA Emissions
Certification Files. Each record of the Certification Files contained
EPA test MPG for each unique combination of vehicle attributes within
a given make, model, and year. These attributes were (1) number
of cylinders, (2) cubic inches of engine displacement (CID), (3)
type of transmission (manual or automatic), (4) gasoline or diesel
fuel, and (5) whether the vehicle's emissions control package met
Federal or California standards. Each record of the Certification
Files also contained the number of vehicles sold for each unique
combination of attributes. The vehicle attributes needed to assign
a test MPG for sample vehicles were obtained from the Vehicle Identification
Number (VIN), and/or from the RTECS questionnaire responses when
the VIN was unavailable. The VIN was decoded to yield the vehicle
attributes, by use of the Highway Loss Data Institute's "Vindicator"
software.
In
addition to assigning test MPG, the EPA Certification Files were
used to impute for missing vehicle attributes. Based on the nonmissing
vehicle attributes obtained from the questionnaire and VIN, several
records from the EPA Certification Files were usually found as potential
"matches" to a given sample vehicle. A matching record
was chosen from among the several applicable ones, with probability
proportional to sales, using the sales figures on the EPA Certification
Files. Once chosen, a record provided EPA test MPG (city and highway),
as well as any vehicle attributes that were missing.
The
1994 RTECS used a sequential adjustment procedure where the EPA
Composite MPG was adjusted first to an on-road MPG, and then to
an in-use MPG.
The
EPA Composite MPG
Beginning
in the early 1970's, EPA measured fuel efficiency from tests that
were conducted on a dynamometer to simulate actual driving conditions.
By 1975, EPA had incorporated separate "city" and "highway"
driving cycles into the test. The city and highway MPG were combined
to form a "composite" MPG, that was then weighted according
to sales of the production vehicles in order to assess compliance
with Corporate Average Fuel Economy (CAFE) standards. The EPA Composite
MPG is based on the assumption of a "typical" vehicle-use
pattern of 55 percent city driving and 45 percent highway driving,
and has become a convenient single fuel efficiency measure for analytical
and regulatory purposes.
The
EPA Composite MPG is defined as:
where:
MPG(EPA
55/45) = the composite MPG
MPG(EPA
city) = the fuel efficiency when vehicle use pattern is city
driving only
MPG(EPA
highway) = the fuel efficiency when vehicle use pattern is
highway driving only.
Fuel
Efficiency Shortfall
Fuel
efficiency shortfall occurs when the fuel efficiency that is actually
obtained while using the vehicle is lower than the EPA test results.
Reasons for this shortfall are (1) a result of the differences between
EPA test vehicles and the vehicles actually in use and (2) the differences
between EPA procedures for simulated driving conditions and actual
driving conditions. For example, EPA test vehicles are prototypes
that do not contain the wide variety of power-consuming accessories
often found on vehicles sold to consumers. The test procedures also
do not simulate the actual driving conditions that affect fuel efficiency
such as speed and acceleration of individual drivers, road conditions,
weather, and traffic. In the 1994 RTECS, adjustments for this fuel
efficiency shortfall were made to the composite MPG (MPG(EPA
55/45)) that were assigned to the sample vehicles.
Fuel
efficiency shortfall was expressed in terms of the "Gallons
per Mile Ratio" or GPMR:
Where:
GPMRi
= Gallons per Mile Ratio for vehicle i
MPGi
= On-road MPG or in-use MPG for vehicle i, depending on the analysis
MPG(EPA
55/45) = EPA Composite MPG applicable to vehicle i.
If
GPMRi = 1 then there is no shortfall. If GPMRi
> 1 then there is a shortfall for vehicle i (That is, the on-road
or in-use fuel efficiency is less than the fuel efficiency indicated
by the EPA Composite MPG.) Note that GPMRi can represent
shortfall with respect to either the on-road or in-use MPGi,
depending on the analysis being performed. GPMRi is commonly
chosen as a measure of shortfall as opposed to MPGi for
the following reasons:
A
shortfall adjustment is most often thought of as a correction factor,
or multiplicative constant, rather than as an additive correction.
GPMRi satisfies this convention.
Shortfall
is usually dependent on a vehicle's fuel efficiency level. That
is, shortfall is usually higher at high levels of MPG(EPA 55/45)
than at low levels of MPG(EPA 55/45). Therefore, it is
more informative to express the amount of shortfall relative to
MPG(EPA 55/45) rather than as an absolute quantity.
GPMRi
is a linear function of MPG(EPA 55/45) and can be modeled
using ordinary least squares linear regression.
GPMRi
is a transformation that stabilizes error variances for the purposes
of least squares linear regression.
The
On-Road MPG
On-road
MPG is a composite MPG that was adjusted to account for the shortfall
between the EPA fuel efficiency and the actual fuel efficiency obtained
on the road.
The
EPA developed two general procedures for adjusting MPG(EPA
55/45) to an on-road value. One procedure bases the size of
the adjustment on specific technology features of the vehicle. The
other procedure uses just two MPG discount factors, one to adjust
the EPA highway estimate, the other to adjust the city estimate.
These two factors are used for all vehicles, regardless of technology
class. Either of these procedures could be used to adjust MPG(EPA
55/45) to an on-road MPG value for use in the 1994 RTECS.
Since both procedures were unbiased for trucks, the choice as to
which to employ in the 1994 RTECS should be based on their performance
with cars. The adjustment based on discount factors seemed to be
less biased than the Technology-Specific Adjustment. The discount
factors are also less expensive since they do not require collection
or imputation of information on fuel delivery system and drive-train.
Because of these reasons the Discount Factors Adjustment Method
was selected.
Shortfall
Adjustment Based on Discount Factors
EPA's
discount factors have widespread appeal because of their simplicity
(Hellman and Murrell, 1985[4]; Hellman and Murrell, 1984[5]). The
factors are .10 percent for city MPG and .22 percent for highway
MPG. That is, for any vehicle i,
These
discount factors are the ones used to produce the "sticker"
MPG figures seen on vehicles on dealer lots, and are used to produce
the DOE/EPA Gas Mileage Guide. The analysis behind the
development of these factors was performed on a conglomerate data
base with data from Ford Motor Company, General Motors, Chrysler
Corporation, DOE, and EPA. The data base contained approximately
38,000 vehicle records with model years from 1979 through 1981 with
some 1982 models included. The data base contained predominately
American-made vehicles, but also included foreign vehicles as well.
The technology mix was dominated by rear-wheel drive and carbureted
vehicles, but contained some vehicles with front-wheel drive or
fuel injection. Vehicle records contained make, model, year, vehicle
characteristics, the MPG as measured on the road, MPG(EPA city),
and MPG(EPA highway). The data base also
included
the driver's perceptions of the proportion of their travel that
was mostly urban (so called "city fraction"), and their
average miles driven per day (AMPD).
Fuel
economy shortfall is affected by the vehicle use pattern: city-driving
pattern is characterized by frequent starts and short trip lengths,
while highway-driving pattern is characterized by infrequent starts
and long trips. AMPD is a good surrogate variable for representing
these different driving patterns.
The
city-driving pattern was characterized by AMPD from 5 to 22 miles
per day, while the highway-driving pattern was characterized by
AMPD's from 15 to 105 miles per day (Hellman and Murrell, 1984).
City fraction and AMPD were used to split the data into two sets,
one for development of the city discount factor, the other for development
of the highway factor. The "city" and "highway"
data sets were each stratified by vehicle technology classes. Linear
regression was performed within each stratum. GPMR was regressed
on city fraction, AMPD, MPG(EPA 55/45), odometer reading,
and average temperature. The fitted models were then weighted and
combined across vehicle technology strata, to produce a single "city"
shortfall model and a single "highway" shortfall model.
The weights were used to increase the influence of those models
that represented technology mixes expected to become more prominent
in the future (e.g., front-wheel drive and fuel-injected vehicles).
The discount factors were derived from the two weighted models set
at average or typical values of the independent variables.
For
each RTECS vehicle, discounted city and highway on-road MPG were
computed and then combined to form an on-road 55/45 composite as
follows:
A
shortfall ratio based on EPA discount factors was computed for each
RTECS vehicle as follows:
The
In-Use MPG
In-use
MPG are MPG that are adjusted for individual driving circumstances.
The on-road adjustments to MPG(EPA 55/45) discussed in
the previous sections were "general" in that they did
not take into account any effects on fuel economy that are due to
the driver's individual circumstances. They, instead, utilized general
attributes such as the technology features of the vehicle and average
driving conditions. Fuel economy shortfall estimates can be refined
for an individual vehicle by taking into account the following "in-use"
effects.
Urban
versus rural driving pattern. That is, frequent starts and short
trips as opposed to infrequent starts and longer trips. As mentioned
in the previous section, a useful single variable for representing
this effect is AMPD. High AMPD's usually represent mileage accumulated
on the highway.
Traffic
congestion, which increases with population density.
Seasonal
temperature variations, especially for gasoline-carbureted vehicles.
Humidity,
which together with temperature, affects air-conditioner use.
Differences
among geographic areas of the country.
Altitude.
Wind.
Road
gradient and road surface conditions.
In
general, the first four items are considered the most significant
in-use influences (Crawford, 1983[1]). In the cited study, shortfall
variations as high as 25 percent or more occurred over the range
of typical AMPD. Shortfall was 16 percent higher in urban areas
than in completely uncongested areas, and was 12 percent higher
in suburban areas. Shortfall varied seasonally (i.e., monthly) by
7 percent in the South and by 13 percent in the North.
Regression
models were developed (Crawford, 1983) for use in adjusting GPMRi(on-road)
to an in-use shortfall employing measurements of several in-use
effects as the independent variables.
The
regressions yielded a shortfall adjustment that was an additive
one, as follows:
where:
GPMRij(in-use)=
the in-use shortfall ratio estimated for vehicle i and month j (j
= 1...12), GPMRi(on-road) = the on-road shortfall ratio
estimated for vehicle i, from the above equations, and
ij
= an adjustment calculated for vehicle i and month j, from a regression
model.
One
regression model from the Crawford reference which is appropriate
for use in RTECS is as follows:
Where:
AMPDij
= Average Miles per Day for vehicle i and month j, typically 35.6
(i.e., 13,000 miles per year).
NORTH
= 1 if the household is in the North.
0
if the household is not in the North.
SOUTH
= 1 if the household is in the South.
0
if the household is not in the South.
This
regression model was chosen because the independent variables that
are important in explaining shortfall were readily available from
the 1994 RTECS data. The model had two components. One component
involved AMPDij and represented the influence of individual
driving patterns for a given vehicle and month. The other component
represented the change in shortfall that occurred throughout the
seasons, due to the annual temperature cycle. The original regression
equation also contained a minor term which accounted for the influence
of air-conditioner use during hot, humid weather. This term was
dropped in the 1994 RTECS estimations because it involved the rather
complex computation of "Discomfort Index" from NOAA weather
records, and the slight additional precision was judged insufficient
to warrant the additional processing expense. Additional terms representing
geographic regional effects, and the natural logarithm of population
density (people per square mile, to represent the influence of traffic
congestion) were not considered because of the computational cost.
Once
a GPMRij(in-use) was estimated it was used to estimate
the final in-use fuel economy for vehicle i and month j as follows:
The
regression equation had separate seasonal components for the "North"
and "South," because the difference between the winter
shortfall and the summer shortfall was greater in the North than
in the South. This difference can be seen in the model parameters.
To define the North and South geographic areas the continental United
States were divided into 97 two-digit ZIP Code regions. These regions
were grouped to form two aggregate regions ("North" and
"South") according to average winter and summer temperatures,
and seasonal shortfall trends.
Annual
Vehicle Fuel Consumption
In
the 1994 RTECS, annual consumption was calculated by dividing the
annual VMT by the annual MPG. The following is a derivation of the
annual VMT and annual MPG.
The
MPGij(in-use) shown in the above section about fuel efficiency
estimation procedures were final estimates of monthly in-use fuel
economies for vehicle i, and could have been used for estimating
monthly fuel consumptions and expenditures if monthly VMT were known.
However, RTECS collected only annual VMT, as calculated from the
B-O-Y and E-O-Y odometer readings. Nevertheless, the 1994 RTECS
still made use of the MPGij(in-use) by disaggregating
the annual VMT of sample vehicles into monthly VMT.
The
annual consumption for vehicle i can be thought of as the sum of
the individual monthly consumptions:
Where:
Ci
= Annual consumption of vehicle fuel for vehicle i, in gallons
spi
= Month starting possession of vehicle i by the household, or January
1994, whichever is later
epi
= Month ending possession of vehicle i by the household, or December
1994, whichever is earlier
cij
= Consumption of vehicle fuel for vehicle i, during month j.
Consumption
is calculated over only those months that vehicle i was reported
to be owned or used by the household. In this sense, "annual"
does not necessarily mean a full 12 months. This is an important
point since fuel economy varies seasonally. If vehicle i was in
the household for the entire year, then spi = JAN and
epi = DEC.
Consumption
for each month can be expressed in terms of monthly VMT and monthly
fuel economy:
Where:
mij
= VMT for vehicle i, month j
mpgij
= Fuel economy in miles per gallon for vehicle i, month j
so
that:
In
the 1994 RTECS Ci was estimated by substituting the estimated
MPGij(in-use) for mpgij. The mij
was estimated in RTECS by disaggregating the annual VMT from odometer
readings into monthly VMT. The disaggregation was performed as follows:
Where:
Mi
= Annual VMT for vehicle i, calculated using odometer readings and
the two-step adjustment procedure
discussed
in the section titled "Vehicle Miles Traveled"
f(i,j)
= Average fraction of "annual" VMT that was driven during
month j, estimated for vehicle i
There
is no single distribution of average monthly VMT fractions f(ij).
Rather, there was a family of distributions, the members of which
were determined by the particular months a vehicle was owned or
used by a household. According to this definition of monthly VMT
fractions, no matter which months vehicle i was in a household,
it was always true that:
The
f(i,j) were derived from the Fj in Table B1
as follows:
Otherwise
Substituting
mpgij = MPGij(in-use) and mij =
Mi x f(i,j) into Equation 13.
yields
the following estimate of annual consumption for vehicle i:
The
estimator of annual consumption in the above equation was constructed
with 1994 RTECS data.
For
vehicles that were acquired or disposed of during 1994, the estimator
took into account seasonal differences in the overall fuel economy
and the effects of these differences on the overall fuel consumption.
Substituting
MPG(EPA 55/45) in the above equation, and slightly rearranging
the terms, the estimator of consumption is:
A
single "annualized" fuel economy that is analogous to
the "annualized" MPGi from previous RTECS,
was estimated as:
Thus
Annual
Vehicle-Fuel Expenditures and Price
Vehicle
Fuel Expenditures
In
the 1994 RTECS, fuel expenditures were calculated by multiplying
the vehicle-fuel consumption by the price of the vehicle fuel. The
1994 RTECS, like the 1988 and 1991 RTECS, did not collect vehicle
fuel prices via fuel purchase diaries. Instead, each RTECS vehicle
was assigned a price based on reported fuel type used in the vehicle.
Gasoline prices were obtained from the BLS 1994 Retail Gasoline
Pump Price Series. Diesel fuel prices were obtained from the Lundberg
Survey, Inc. (See "Other Fuel Types" in this appendix
for a discussion of the gasohol and propane prices.)
Respondents
were asked if they purchased leaded or unleaded gasoline, and if
unleaded, they were asked the grade. (See Appendix D, "Survey
Forms.") The BLS prices are published by month, by Census region,
and by type and grade of fuel. In 1988, the BLS monthly prices (for
the Census region in which the household lived) were averaged across
the months that the vehicle was in the possession of the household.
This yielded for each RTECS vehicle a single fuel price, Pi,
dependent on the Census region, type and grade of gasoline, and
the months that the vehicle was in the possession of the household.
In 1988, the annual fuel expenditures in dollars for each sample
vehicle, Ei, was estimated by multiplying its assigned
average fuel price, Pi, by its total consumption in gallons,
Ci, as estimated in the previous section. However, in
1994, annual fuel expenditures, Ei, was estimated by
multiplying monthly gasoline prices by monthly consumption to produce
monthly expenditures and summing the monthly expenditures to produce
annual expenditures.
Type
of Fuel Used
Table
B2 provides the percentage distribution of RTECS vehicles by fuel
type categories. In 1994, 97.9 percent of the 156.8 million RTECS
vehicles used unleaded gasoline. The remaining 6.3 percent of vehicles
used leaded gasoline, diesel fuel or other fuel types.
Table
B2. Distribution of Residential Transportation nergy Consumption Survey
Vehicles by Type of Fuel Used, 1994
Type
of Vehicle Fuel |
Number
of Vehicles millions) |
Percent
of Vehicles |
Total |
156.8 |
100.0 |
Gasoline
Leaded
Unleaded
Regular
Premium
Intermediate |
153.4
Q
151.5
14.2
26.7
20.6 |
97.9
Q
96.7
66.4
17.1
13.2 |
Diesel |
1.8 |
1.1 |
Gasohol |
1.4 |
0.9 |
Notes:
Because of rounding, data may not sum to totals. For a discussion
of underreporting of gasohol see Appendix C,
"Quality
of the Data."
Q
= Data withheld either because the Relative Standard Error (RSE)
was greater than 50 percent or fewer than 10 households were sampled.
Source:
Energy Information Administration, Office of Energy Markets and
End Use, 1994 Residential Transportation
Energy
Consumption Survey.
Gasoline
Prices
Prices
published by the BLS survey are retail prices for leaded regular,
unleaded regular, and unleaded premium gasoline. These prices are
published monthly by Census region. The BLS Pump Price Survey is
conducted as input to the Consumer Price Index (CPI). Prices
are collected in 85 urban areas. The population covered excludes
the institutional population and households located on military
bases. The covered population includes approximately 85 percent
of all U.S. households. The BLS uses a rotating sample of approximately
1,100 service stations.
Each
vehicle in the 1994 RTECS, that used gasoline was assigned a monthly
BLS fuel price. The BLS "unleaded regular" price was assigned
to all vehicles that reported using leaded gasoline..
Diesel
Fuel Prices
Diesel
fuel prices were obtained from the "Lundberg Letter-PS"
published by Lundberg Survey, Inc. The Lundberg Survey, Inc. collects
pump prices at retail service stations in approximately 80 major
metropolitan markets. The survey includes about 15,000 service stations
divided into 2 bimonthly panels. At least one city from every State
is included. Service stations on military bases and in rural areas
are excluded. Sales-weighted price data for both full-and self-service
stations are published bimonthly. Regional prices are not published.
All RTECS vehicles that used diesel fuel were assigned the same
diesel fuel prices regardless of Census region.
For
the RTECS, the following two steps were used to create diesel prices.
(1) Bimonthly diesel fuel full-service and self-service prices,
obtained from the Lundberg Letter-PS, were averaged to create prices
for each grade in the intermediate months. (2) The monthly full-
and self-service prices were then weighted and averaged to obtain
overall diesel fuel prices. The weights used to create an average
diesel fuel price from the full- and self-service prices were based
on RTECS data on "type of service" (full-service or self-/mini-service)
used when purchasing diesel fuel. For each vehicle monthly prices
were multiplied by monthly consumption to yield monthly expenditures.
The monthly expenditures were summed to produce annual expenditures.
Other
Fuel Type Prices
Approximately
1.4 million 1994 RTECS vehicles were reported using gasohol. In
the absence of applicable national estimates of the average price
paid for gasohol, the RTECS vehicles using gasohol were assigned
fuel prices using the same methodology as the most common group
of vehicles in the survey--vehicles using regular unleaded gasoline.
(See above for methodology assigning unleaded regular gasoline prices
and Appendix C, "Quality of the Data" for a discussion
of RTECS underestimation of other fuels.)
References
1.
Crawford, R. 1983. "Seasonal and Regional MPG as Influenced
by Environmental Conditions and Travel Patterns." Research
performed under contract for DOE. Energy and Environmental Analysis,
Inc., Arlington, VA.
2.
Harrison, I.M. "Retail Fuel Pump-Prices," Residential
Transportation Energy Consumption Survey Technical Note 4, unpublished
document. (Washington, DC, 1994).
3.
Harrison, I.M. "VMT 1994 Patterns," Residential
Transportation Energy Consumption Survey Technical Note 5, unpublished
document. (Washington, DC, 1994).
4.
Hellman, K.H., and Murrell, J.D. 1985. "On the Stability
of the EPA MPG Adjustment Factors." Society of Automotive
Engineers Technical Paper Series, SAE Paper No. 851216, Warrendale,
PA.
5.
Hellman, K.H., and Murrell, J.D. 1984. "Development of
Adjustment Factors for the EPA City and Highway MPG Values."
Society of Automotive Engineers Technical Paper Series, SAE Paper
No. 840496, Warrendale, PA.
6.
Lax, D. 1987. "Feasibility of Estimating In-Use Vehicle
Fuel Efficiency from Household Survey Data." Research
performed under contract for ORNL/DOE/EIA. Energy and Environmental
Analysis Inc., Arlington, VA.
7.
"Lundberg Letter-PS," Lundberg Survey Inc. 1994.
(North Hollywood, CA).
8.
U.S. Department of Energy and U.S. Environmental Protection Agency,
Gas Mileage Guide, EPA Fuel Economy Estimates. (Washington,
DC).
9.
"Average Prices for Gasoline, U.S. City Average and Selected
Areas," U.S. Department of Labor, Bureau of Labor Statistics,
1994. (Washington, DC).
File Last Modified: August 25, 1997
If you are having any technical problems with this
site, please contact the EIA Webmaster at wmaster@eia.doe.gov
|