Prepared for:
Environmental Protection Agency
Ann Arbor, MI
Prepared by:
Energy and Environmental Analysis, Inc.
1655 North Fort Myer Drive, Suite 600
Arlington, Virginia 22209
September 30, 1993

1.	Introduction 		1-1
1.1	Purpose of the Report 		1-1
1.2	Limitations of the Report		1-1
1.3	Summary of Forecast Structures 		1-2
2.	Results		2-1
2.1	Results Summary 		2-1
2.2	Conclusion		2-3
Bibliography 		B-l

The last twenty years has seen the advent of the computer greatly ease the task of
economists in their attempts to construct ever more realistic mathematical models and
forecasts of the economy, and the transportation sector. The perceived realism of the
models, however, only translates somewhat into more accurate forecasts, since the
fluctuations of important model input variables (e.g. petroleum prices) are difficult in
themselves to predict.
This study, prepared for the EPA, compares the historical projections of several different
types of models that forecast vehicle miles traveled (VMT), with a focus on the differing
methodologies and results. The projections examined date from around 1970 to 1985, at
roughly five year intervals, and are compared with actual VMT performance (as
measured by the FHWA) at five year intervals, ending in 1990. The projections
examined include forecasts from the EPA MOBILE model, the Motor Fuel Consumption
Model (EEA), the Department of Energy/Energy Information Agency (DOE/EIA), Jack
Faucett Associates (JFA), Wharton Econometric Forecasting Associates (WEFA), and
Argonne National Laboratory (ANL). Forecasts were collected from documents dating
from the time of the projection, with the exception of the MOBILE model which was re-
run using only data available at the time of the supposed "projection".
One difficulty in preparing this document was that many of the forecast were not directly
comparable, in that some of the forecasts predict VMT of all types of vehicles, while
others only predict a subset of that. Scarcity of older documents and the changing
structure of some of the performing organizations presented another difficulty. Thus,
this report is meant to be a "quick look" at the results and methods of certain types of

models and projections, and not an encyclopedic compilation or analysis of forecasting
tools. Also, it must be noted that the very nature of VMT data makes it difficult to
estimate, and the FHWA data is based on widely varying state-level traffic data, and may
contain inaccuracies.
U.l Argonne National Laboratory
ANL has periodically produced transportation forecasts for DOE since 1979, most
utilizing the Transportation Energy and Emissions Modeling System (TEEMS), which
was developed fully by 1983. TEEMS is a group of interconnected behavioral models
sensitive to economic and demographic variables, as well as more specific transportation
sector attributes. The first forecast (ANL79) projected energy demand using a "bottom-
up" methodology, by summing results from many modal and submodal projections based
on the SEAS input-output model. The 1981 forecast (ANL81) used the earliest versions
of the models currently incorporated into TEEMS, and all subsequent versions have
employed the TEEMS model as it exists today. For the purposes of this report, EEA
compared the output of ANL79 and ANL86, which each incorporate different
methodologies and achieve dissimilar results.
ANL79 is a baseline projection which incorporated continuations of then-current trends,
and conservative estimates for the major input variables. The report based its
projections of automobile activity on the demand for passenger-miles, with local,
commuter, and intercity travel calculated separately. Local passenger-miles were
assumed to grow proportionally with per capita disposable income. Intercity auto travel
was calculated using a more complex demand model, and commuter activity was simply
assumed to be a constant share of auto travel. VMT were estimated by subtracting
estimates of travel by competing modes, and then using average occupancy rates to
translate travel to VMT. Personal light trucks VMT were assumed to be proportional to

their share of the light vehicle fleet. Results from the forecasts are presented graphically
in the next section.
ANL86, aside from using the TEEMS model structure, presented a forecast for
transportation activity under persistently low petroleum prices, based on the Autumn
1986 Data Resources, Inc. (DRI) projection of world petroleum prices. In this scenario,
given the structure of the TEEMS model, personal VMT are mainly influenced by fuel
prices (though demographic changes are important), and commercial vehicle activity
relies more clearly on economic activity variables such the rates of growth for specific
economic sectors.
1J.2 Department of Energy / Energy Information Administration
Since the late 1970s, EIA has been developing forecasts of transportation activity for its
Annual Report to Congress and Annual Energy Outlook. These projections are largely the
result of simple econometric algorithms in which VMT is a function of fuel price and
personal income. This report examines the 1980 and 1984 VMT forecasts. The 1984
was chosen over the 1985 due to the exclusion of specific VMT data in the 1985 Annual
Energy Outlook.
UJ Energy and Environmental Analysis. Inc.
EEA is the corporate author of the Motor Fuel Consumption Model (MFCM), developed
for the DOE, which was first created in 1978 to include only light-duty vehicles, but has
since been expanded to include heavy-duty vehicle and off-highway fuel use. The MFCM
calculates the average vehicle's VMT for each of seven different vehicle classes, taking
into account for each class the age distribution of vehicles, fuel type, and year-to-year
shifts in economic and demographic patterns. Total VMT is simply the aggregate of
average VMT for each vehicle type multiplied by the vehicle stock for each type. This
study focuses on the May 1981 (the first to incorporate heavy-duty vehicles) and the
November 1985 output of the MFCM. The structure of the MFCM has remained fairly
1 ^

constant over this period, with the exception of the addition and subtraction of certain
vehicle and fuel classifications.
1J.4 EPA Motor Fuel Consumption Model (MOBILE4.1 MFO
The EPA developed the MOBILE model in the early 1980s, primarily to measure the
amounts of pollutants emitted by transportation, with the MFC being an offshoot of that
model. Similar in structure to EEA's MFCM, the EPA MFC model computes VMT in
the same way, with the total VMT per year being simply the product of the average
VMT for each vehicle and the total number of the vehicles stock. The number of miles
traveled per vehicle is dependent on the vehicle's age as well as its class. The
MOBILE4.1 MFC model computes VMT distributions for eleven vehicle classes, and
both gasoline and diesel vehicles. For this study, the EPA is using the current model to
re-CTeate what the then-current version of the model would have produced in 1982.
1J.5 FHWA/Jack Faucett Associates
JFA has projected transportation activity for the FHWA and DOT since the early 1970s.
This report examines the JFA projections from 1971 and 1980. 1975 and 1985 estimates
were unavailable.
The 1971 projection calculated VMT for three different categories of automobiles
(business, personal, and government), as well as school and other buses. VMT for
business automobiles (those used in business fleets, whether company-owned, employee-
owned, or leased) was assumed to grow in proportion to industry output. Personal VMT
are forecast as a function of personal consumption expenditures in the corresponding
National Income Division category, "User Operated Transportation". Personal
consumption expenditure forecasts were produced by the Bureau of Labor Statistics.
Government vehicles (inclusive of all levels of government) are calculated to remain
proportional to the growth rates of expenditures in that category. Bus VMT are related
to the school system and industrial output.

Tne 1980 report by JFA uses the DOT Long-Range Forecasting Model developed for
DOT by JFA, Transportation summaries cover 31 for-hire and private transport modes,
including VMT forecasts for personal and business automobiles, buses, and certain types
of truck activity (freight activity is handled on a ton-mile basis, as opposed to VMT).
Automobile VMT are a function of the vehicle stock, income and wealth, and operating
costs. Vehicle stock and operating costs are forecast based on a detailed analysis of both
consumer demand and an industry supply simulation component, as well as gasoline
1J.6 Wharton Econometric Forecasting Associates
WEFA had produced forecasts of vehicle activity since the mid-1970s, but ceased in
the early 1980s. For this study EEA examines the 1975 output of the WEFA Automobile
Demand Model. The model, a long-run econometric equiJibrium model, is primarily
concerned with the size, composition, and attributes of the automobile stock. The model
derives its relationships from analysis of state-level data from 1972, and uses the "fanuly-
unit" as the basic variable unit (as opposed to households or individuals). The VMT
equation in the model estimates VMT per family as a fjnction of fleet MPG. fleet age
distribution, income variables, and gasoline prices.

In this section, the different forecasts of VMT are directly contrasted with the historical
FHWA data (Table 1). FHWA VMT data is disaggregated into passenger cars,
motorcycles, buses, 2-axle-4-tire trucks, other single-unit trucks, and combination trucks.
These figures generally do not correspond directly to the output of any of the models
examined, and caution must be exercised when comparing the forecast outputs to the
FHWA category. For example, many states reporting vehicle registrations (from which
total VMT activity is stratified by the FHWA) classify passenger "mini-vans" as
automobiles, while others place them in the truck category, creating a situation wherein
some forecast estimates may differ by definition. In addition, LDT comparisons are
further complicated by differing defintions, with industry defining LDTs as trucks under
10,000 lbs. and other entities (e.g. U.S. EPA) defining LDTs as trucks under 8,500 lbs.
Where information was available, these differences preface the caomparisons. Forecasts
are examined against FHWA figures for each of the output vehicle classes for which
analagous FHWA classifications exist.
2.1.1 Argonne National Laboratory
The ANL projections of VMT includes automobiles and light trucks (Table 2) with
motorcycles and buses included in the 1979 projection. The 1979 forecast did not
provide 1980 figures, nor did the 1986 output include 1985. The 1986 forcast was chosen
over the 1985 projection, due to the contrast it provides versus previous ANL projections.
While both versions of the model also include other highway modes, commercial freight
and passenger performance are measured in ton-miles and passenger miles. Light truck
(LDT) VMT are compared with FHWA 2-axle-4-tire trucks data for all forecast

Table 1
(Million VMT)

Other Single

* - 2 Axle, 4 -Tire	rji
Source: Highway Statistics

Table 2
ANL Forecast Data
(Million VMT)
ANL (8/79)
Year	Cars	% Error Motorcycles % Error
1980	N/A	N/A
1985	1,363,600	8.2% 45,140 396 8%
1990	1,517,900	H.2% 68,480 653.7%
Source: Projections of Direct Energy Consumption
ANL (86)
Year	Cars	%Error LDTs % Error
1985	N/A	N/A
1990	1,407,840	-7.1% 576,240 23.4%
Source: Transportation Energy Outlook Under Conditions
of Persistently Low Petroleum Prices
LDTs	% Error
394,840	5.8%
459,690	-1 5%

The ANL79 forecast proves to be quite accurate, excluding the motorcycle estimates,
which are an order of magnitude too large. The 1986 forecast however, includes all
miruvans m the light truck category, accounting for some of the difference between the
ANL estimate and the FHWA data.
2.12 DOE/EIA
The DOE/EIA forecasts are from 1980 and 1984, and include cars and trucks, with
varying degrees of disaggregation (Table 3). The forecast methodology and level of
detail have varied over time, but have remained essentially a econometric analysis of
VMT as a function of fuel price and income. The 1980 estimates underestimate both
passenger car and single unit truck VMT, the latter by a quite substantial margin. The
1984 estimate is taken from the middle-world-oil-price and middle-economic-growth
scenario, displays much improved accuracy.
2.1 J Energy and Environmental Analysis. Inc.
The MFCM forecasts automobile, LDT, and medium and heavy truck (HDT) VMT, with
the May 1981 and November 1985 outputs shown here (Table 4). The 1981 estimate
underestimated growth across the board, while the 1985 estimate fared considerably
better. The LDT classification of the MFCM includes only trucks 8,500 GVW and
under, accounting for some of the differential seen in the truck classifications, especially
in the 1985 projection.
2.1.4 EPA MOBILE4.I Motor Fael Consumption Model
The output provided for the M4FC lists VMT by vehicle class, and are aggregated here
to correspond to the FHWA classes (Table 5.). The "forecast" was obtained using input
data that would recreate the output of the M4FC model during 1982. The uncalibrated
output was used for comparison. The output shows considerable shortfall in the
passenger car and HDT categories for all years, but the LDT class proves quite accurate.

Tabic 3
DOE/EIA Forecast Data
(Million VMT)
Year Cars* % Error Trucks	% Error
1980 N/A N/A
1985 1,184,000 - 6.7% 314,000	- 25.2%
1990 1,364,000 -10.6% 366,000	- 29.7%
* - includes motorcycles
Source: EIA Annual Report to Congress 1980
EIA (84)
Year	Cars	% Error	Trucks	% Error
1985	1,315,400	4.3%	498,500	10.1%
1990	1,501,000	-0.9%	571,800	1.5%
Source: Annual Energy Outlook 1984
Trucks % Error
80,000 0.5%
95,000 -1.5%

Table 4
EEA Forecast Data
(Million VMT)
EEA (5/81)
Year	Cars	% Error
1980	1,132,460	1.9%
1985	1,249,250	-0.9%
1990	1,414,840	-6.6%
LDTs	% Error	HDTs
262,430	- 9.8%	91,410
315,490	-15.4%	110,210
360,260	- 22.8%	136,110
% Error	TOTAL	% Error
-15.7%	1,486300	-1.6%
-12.9%	1,674,950	-4.8%
-9.3%	1,911,210	-10.4%
Source: Highway Fuel Consumption Model 4th Quarterly Report

Table 5
EPA Forecast Data
(Million VMT)
CPA '"82
Source: Mark Wolcott, EPA
Cars % Error
1,067.790 -15.3%
1,167,280 -23.0%
LDTs % Error
385,970 3.5%
471,770 1.1%
HDTs % Error
115,910 -8.4%
132,020 -12.0%
Buses % Error
340 -93.0%
380 -93.4%
TOTAL % Error
1,570,010 -10.8%
1,771,450 -16.9%

As with other forecasts, the smaller vehicle types, in this case buses, show significant
variation against the FHWA estimates.
2.1.5	Jack Faucelt Associates / FHWA
Both JFA projections (performed in 1971 and 1980) are taken from the medium growth
scenarios and both include light trucks in the automobile classification, which includes
personal, government, and business vehicles (Table 6). School and privately-owned
buses are also included, but commercial freight and passenger modal performance are
measured in ton-miles and passenger-miles, so the bus comparison is misleading. Each
projection includes only selected years, limiting its comparison with other models.
2.1.6	Wharton Econometric Forecasting Associates. Inc.
The WEFA Automobile Demand Model forecasts only the automobile sector, and while
it does so in considerable detail, only the aggregate figures are of concern in this report
(Table 7). As with the other reports, the WEFA model forecast, made in 1975,
underestimates the growth in VMT.
The low oil prices and strong growth in VMT that has occurred over the last ten years
has been largely unexpected, as can be seen from the results shown in the above
comparisons. In the passenger car category, of the projections examined, the average
error was roughly -5%, with a wide variation observed. Only one of projections
overestimated VMT in a projection year (ANL79), and that was by 0.2%. The
tremendous growth in LDT usage also appears to have been unforeseen, with a high
average error in that category, even considering the differences in classification that
make direct comparison with FHWA data difficult.

Table 6
JFA Forecast Data
(Million VMT)
JFA (7/71)
Year Cars / LDTs % Error Buses % Error
1975	N/A	N/A
1980	1,313,271 -6.4% 1,973 -67.4%
1985, 1990	N/A	N/A
Source: Transportation Projections 1970 and 1980
JFA (1/80)
Year	Cars/LDTs % Error
1980	N/A
1985	1,585,912 -2.9%
1990	N/A
Source: Transportation Projections 1985,1995.2000

Table 7
WEFA Forecast Data
(Million VMT)
WEFA (3/75)
Year	Cars	% Error
1975	1,029,700	-0.4%
1980	1,099,400	-1.1%
1985	1,216,600	- 3 5%
1990	1,283,200	-15 3%
Source: An Analysis of the Automobile Market

The following graphs (Figures 1 and 2) compare the passenger car and LDT output of
each projection for 1990, although it must be noted again that the different models use
different classification schemes, and that side-by-side comparison can be misleading.
^ 1 1

1,600 -r
1,200 --
600 --

VMT Projection Comparison
Light Duty Truck Projections - 1990

Energy and Environmental Analysis, Inc.; The Highway Fuel Consumption Model: Fourth
Quarterly Report; EEA, Inc., Arlington, VA.; 1981.
Energy and Environmental Analysis, Inc.; The Motor Fuel Consumption Model: Twelfth
Periodical Report', EEA, Inc., Arlington, VA.; 1985.
Energy Information Agency, Annual Report to Congress 1980\ ELA, Washington, D.C.;
Energy Information Agency, Annual Energy Outlook 1984\ EIA, Washington, D.C.; 1985.
Federal Highway Administration; Highway Statistics Annual; FHWA, Washington, D.C.
Jack Faucett Associates; Transportation Projections 1970 and 1980\ U.S. Department of
Transportation, Washington, D.C.; 1971.
Jack Faucett Associates; Transportation Projections 1985, 1995, 2000\ U.S. Department of
Transportation, Washington, D.C.; 1980.
Knorr, Rita and Millar, Marianne; Projections of Direct Energy Consumption by Mode:
1975-2000 Baseline; Transportation Energy Systems, Argonne National Laboratory,
Argonne, EL; 1979.
Millar, Marianne et al; Baseline Projections of Transportation Energy Consumption by
Mode: 1981 Update; Center for Transportation Research, Argonne National Laboratory,
Argonne, IL.; 1981.
Millar, Marianne et al; Long-Term Outlook for Transportation Energy Demand;
Transportation Research Record 1049, National Research Council, Washington, D.C.

Millar, Marianne et ah; Transportation Energy Outlook under conditions of Persistently Low
Petroleum Prices; Center for Transportation Research, Argonne National Laboratory,
Argonne, IL.; 1987.
Mintz, M.M. and Vyas, A.D.; Forecast of Transportation Energy Demand Through the Year
2010\ Center for Transportation Research, Argonne National Laboratory, Argonne, IL.;
Schink, George R. and Loxley, Colin J.; An Analysis of the Automobile Market: Modeling
the Long Run Determinants of the Demand for Automobiles; Wharton Econometric
Forecasting Associates, Inc., Philadelphia; 1975.
Wolcott, Mark A. and Kahlbaum, Dennis F.: The MOBILE4 Fuel Consumption Model;
U.S. Environmental Protection Agency, Ann Arbor, MI; 1991.