-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Nitrogen Oxides
Nitrogen oxide (NOX) is an important pollutant in the generation of ozone and in some cases of
PM. Although NQX emissions are a product of gasoline-powered engines for vehicles, diesel
engines are the most significant producers of NOX emissions. Diesel engines power heavy-duty
trucks and also a variety of off-road sources, including rail, air, and marine transportation. Buses
are also powered by ,diesel engines, and contribute heavily to NOX emissions. HDD trucks
contribute as much as 35 percent to 50 percent of regional mobile source NOX emissions and 10
percent to 25 percent of total regional NOX.7
While NOX emissions from passenger vehicles .have been substantially curtailed by catalytic
converters, commercial diesel vehicles are still significant NOX emitters. Given progress in
reducing VOCs for the 1996 15 percent reduction demonstration, NOX has now become a problem
focusjn many areas. ~ - . .
\ , ""''•.'••
NOX emissions obey different travel relationships than .VOCs. NOX emissions are high at speeds
under 10 mph, fall slightly to speeds of about 20-25 mph, rise slowly to speeds of about 45 mph
and then increase rapidly with speeds above 45 mph. Hence, strategies that relieve congestion and
improve operating speeds to over 10 mph are generally effective in reducing CO, VOCs and NOX
emissions;.programs that increase speeds above 25 mph will reduce CO and VOCs but may
increase NOX emissions (depending on pre-program speeds); and speeds over 45 mph result in
substantial NOX emissions, though VOCs and CO emissions continue to decline until speeds of 55
mph are reached. Current models indicate that PM emissions are not much affected by speed,
except that higher speeds lead, to more travel and therefore more PM.
Selection of an emissions-reduction strategy based on travel patterns and speed must accommodate
divergent effects across pollutant types. Moreover, the ratio of NOX to VOC reductions is
important. As the ratio between a region's VOCs and NOX increases, the appropriate strategy for
controlling ozone changes. At a VOC/NOX ratio of 10 or less, for example, VOC controls are
more effective, but at VOC/NOX ratios of 20 or more, NOX control is more effective. In between
these extremes, NOX and VOC policies are more comparable in their effectiveness. These
variations also must be taken into account in program design and evaluation of program impacts.
7 "Air Quality Issues in Intercity Freight," see footnote 1.
U. S. Environmental Protection Agency
2-19
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Particulate Matter
EPA estimates show that about 26.4% of PM-10 sources are due to transportation. This excludes
miscellaneous sources and is based on direct emissions, but road dust kicked up by traveling-
vehicles is another significant source. Tailpipe emissions also create secondary PM-10 emissions
from the oxidation of NOx and SOx to nitrates and sulfates in the atmosphere; also, some of the
VOC compounds are also oxidized to secondary particulates.
PM-2.5, or fine particulates, result from two major types of sources: 1) the incomplete combustion
of fossil and biomass fuels in motor vehicles, boilers, furnaces and open burning sources and 2)
gaseous emissions such as SC>2, NOx, and VOCs from boilers and other fossil fuel combustion
processes that transform in the atmosphere to form sulfate and nitrate particles. The resuspended
soil component of PM-2.5 is generally less than 5-15%. EPA is continually reviewing and
improving estimates of the emission factors and inventories of PM-2.5.
Revisions to the particulate matter standard were announced on July 18, 1997. The review of
hundreds of peer-reviewed scientific studies, published since the original PM-10 standards were
established, provided evidence that significant health effects are associated with exposure to
ambient levels of fine particles (PM-2.5) allowed by the previous standard. These new standards
are available in 62FR 38652 or at the following world wide web address:
http://www.epa.gov/ttn/oarpg/rules.html.
Pricing strategies to reduce sources of particulates would probably vary based on the sources
within a region and various climatic factors. In general, reductions in VMT, rather than specific
congestion reduction strategies or emissions fees tailored for other pollutants, may be more
effective.
2.2.4 Overall Advantages and Disadvantages of Market-Based Measures
Consumers who have grown accustomed to subsidized prices (i.e., prices that do not reflect the full
social and economic costs of their decisions) are likely to oppose any change in the status quo
unless it can be demonstrated that they will be made better off (or not worse off) as a result of the
change. Travelers are no different from other consumers and thus the introduction of market-based
measures to date has been unpopular with the public and their elected representatives.
2-20 . U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Demonstrating the advantages of market-based pricing measures is critical to gaining their ultimate
acceptance. Some important advantages of market-based measures, which may be used to increase
support for their introduction, include the following:
• Applying market-based measures to achieve program objectives (in contrast to
regulatory actions that compel governments or individuals to assume particular
courses of action), provide considerable latitude for .individual freedom of choice.
• By reducing driving subsidies and shifting costs to those who are responsible for
, them, market-based measures give individuals more information about their
transportation costs and choices.
• Consumers can make more rational trade-offs between how much they want to pay
and how and when they want to travel under a pricing system based on true costs;
the current system of indirect taxes and subsidies obscures the information
necessary for such choices. ' .....':
• Rather than locking in current institutional and service concepts, market-based
measures support creativity to develop new and more efficient solutions to travel and
access needs currently met mainly through private vehicle travel.
• Market-based measures can be applied to all types of travel,and trips, not just
employment-related travel. This is in contrast to TGMs, which.have been criticized
aS having a limited (and perhaps inequitable) impact because they affect only a
portion of travel activity.
- • Market-based measures are typically self-financing, and in fact may provide revenue
to support their implementation, as well as, other transportation improvements.
• Market-based measures can substantially reduce the cost of transportation (both
direct and indirect). Reduced vehicle dependency can result in improved travel time
and reliability in movements of people and goods, reduced construction and
maintenance costs, lower taxes, and fewer accidents in _ addition to lower health
costs, less environmental damage and clean up across a range of media: water,,
climate, noise, solid and hazardous wastes, and habitat. •
• Market-based measures can help states and local areas delay or offset the need for
expensive new transportation capacity, help finance that capacity when it is needed,
and result in more efficient usage. They also can help cut costs for maintenance and
government services.
• Longer-term, structural changes induced by market-based measures are critical to
the sustainability of transportation and air quality plans, and can help areas avoid
recurring updates of their SIPs in search of new and stronger controls to offset VMT
growth. • . '.-••'
U.S. Environmental Protection Agency ' ','..-' " 2-21
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Market-based measures are not, of course, without their shortcomings. Perceived disadvantages
include:
• Political resistance, as noted above, to any form of direct transportation pricing,
elimination of subsidies, or imposition of new taxes.
• Technological and administrative challenges and uncertainties in implementation.
• Concerns about fairness and equity, including, for example, whether travelers have
reasonable alternatives and whether they impose increased burdens on the poor or
other disadvantaged groups.
These concerns are important to consider in program design, and should not be dismissed by local
decision-makers. Political consensus on complementing policy changes at the local level will be
required. This document will allow decision-makers to evaluate the costs and benefits of
alternative market-based policy options. (Issues of implementation and practical acceptance are
discussed in Chapter 4).8 " '
Issues related to public involvement, public acceptance, and equity are also discussed in: USEPA,
Opportunities to Improve Air Quality through Transportation Pricing Programs (EPA 420-R-97-004),
September 1997. This document .is also available at the following world wide web address:
http://www.epa.gov/OMSWWW/gopher/Market/pricing.pdf
2-22
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
2.3 Evaluation of Market-Based Measures
Real-life examples that demonstrate how pricing can influence transportation choices exist: transit
ridership clearly fluctuates with fare changes; bridge and tunnel usage demonstrate sensitivity to
toll changes; HOV. usage responds to exclusion from tolls; and automobile owners respond to fuel
price levels by changing their driving behavior. Currently, planners and decision-makers do not
have system-wide implementations of pricing initiatives at their disposal (such as those discussed
in this guidance document) to furnish detailed empirical evidence of their effectiveness and related
.impacts. DOT's Value Pricing Pilot Program (formerly the Congestion Pricing Pilot Program) is
generating valuable experience and research into the potential effectiveness of pricing policies.
For the immediate future, however, estimation of the impacts of these measures will likely come
from simulation tools and other analytical approaches described in this section. There also are
examples from other countries upon which evaluations can be based.
2.3.1 Analytical Process
Most major metropolitan areas •perform transportation planning analysis through some variation of
the "four step process," which derives its name from the four basic analysis modules through
which it assesses travel demand: Trip Generation, Trip Distribution, Modal Split, and Traffic
Assignment. This process, which is pictured in Figure 2.1, methodically translates,spatial activity
patterns into trip flows, which are then represented on a transportation network to reflect specific
, travel volumes on facilities, with respective speeds and levels of congestion.
Supporting these major elements are important " upstream" functions^ such as land use/activity
allocations, detailing of the transportation network, and vehicle ownership forecasts, which
provide essential starting inputs to the process. Although these sub- functions may not be thought
of as demand estimation procedures, they have an important role in profiling travel demand.
Allocation of trips to peak and off-peak time of day is also an important function.
Emissions are calculated as an end product of this process. Analysts extract information on
vehicle trips, miles of travel and speeds from the transportation model and link it with information
on the regional vehicle fleet mix and emissions characteristics. This typically is accomplished
through application of EPA's MOBILE model (version 5a), or in California, the EMFAC model
developed by the California Air Resources Board (CARS). PM-i-o emissions-are calculated using
EPA's PART 5 Model. ,
This section describes the steps in the analysis hierarchy and issues associated with these analyses.
This serves as a prelude to the subsequent discussion on analysis of specific market-based
measures.
U. S. Environmental Protection Agency • 2-23
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Figure 2.1 Schematic Overview of Four-Step Planning Process
Precursor Activities
Regional Growth
Forecasts
Land Use
Allocations
Vehicle Ownership
Four-Step
-
-
1 wooes? ^ , ^^
Trip Generation
v^5
* x ' "'
1 r,/, "
/- ~ >^ '
Trip Distribution
^ ^^^ "^
' 1
1' , ("-
f
Mode Split
•*- " r, - -
\ y ,(
V ^
/v
-'r
//
•& *
ft-'
/, 1
- 1
-W - i
' '' •-> -.t" s'f .-.,:
~ - " ~~ ~~ ~~ ~~ ^ ^^^
.; Time of Day/ -tr ^-
Peaking - *-'
i ° , , x«««»
/ f
' \.
~~ -> ^
%•'
r .
' " * 1, '
Traffic Assignment
- x-,!~
1
.
r
Emissions
>„%; *^
V
/• s
'**), ' -
\ « ^ V
//
/
"> '.-
^"'^"
-•*1
^
'J^
j1
^^
^ 'H'«J
-
^ ^ ^Li!
s 5- j
- ^'
2-24
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
2.3.1,1 Steps in the Analysis Hierarchy
Nine steps comprise the analysis hierarchy, including three precursor activities (regional growth
forecasts, activity allocations, vehicle ownership), the components of the " four step process" (Trip
Generation, Trip Distribution, Modal Split, and Traffic Assignment), as well as its implied time of
day/peaking component, and, last, the emissions evaluation.
Regional Growth Forecasts
The regional planning process begins with forecasts of population and economic growth. Usually,
economic growth, stated in terms of basic^ employment, is estimated first, followed by estimation
of the population growth that would be stimulated by those jobs. Then, population-serving
employment and attendant population increases are estimated. Frequently, these estimates are
keyed to federal and state control totals.
Activity Allocations
Once jobs arid population information is compiled, they are located spatially across the region,
based on development trends, master plans, incomes/economic criteria, or a variety of other
factors. Specific allocations of jobs and households are made to geographic units, generally traffic
analysis zones that loosely conform to groupings of census tracts. This process may involve
analytic techniques (so-called "land use models" which reflect transportation access and cost
issues), but most commonly are the result of an iterative negotiating process. /
Vehicle Ownership
Travel activity forecasts are driven by the economic resources available to households* Most
travel models represent this through the nuniber of vehicles owned by a household, which are then
used throughout the process.. Some model systems rely directly on income figures^ and do not
estimate vehicle ownership. Vehicle ownership estimates are typically 'done through simplistic
cross-classification models, reflecting differences by income, location arid other socioeconomic
characteristics. The estimates usually ignore land use and alternative mode availability as explicit
factors. , ,,
Trip Generation .
From the information on population and employment activity, actual trip making is estimated
.through trip generation analysis techniques. Factors are applied to the population/economic
activity levels to estimate "productions" and "attractions," where households are assumed to
produce trips and employment and commercial activity are assumed to attract trips. These
estimates of household trip making are generally cast in terms of person trips, though in most
models these are motorized vehicle person trips only and exclude walk/bike trips. Some model
systems are even more limited and forecast only private vehicle trips. Trip estimates also are
developed for each major trip purpose (work, shopping, personal business and social/recreational).
Trip production rates are based on such: factors as household size, income/vehicle ownership,
location, etc. Trip attractions are based on the type of employment activity and other factors.
Again, land use, density, and transportation facilities and price/source levels are frequently not
entered into the forecasting process.
U. S. Environmental Protection Agency . 2-25
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions . '
Trip Distribution
The next step is to match the trip productions and attractions to create connected trips for each
zone-to-zone (origin-destination) pair. This usually is done using a gravity model, where
productions and attractions for each trip purpose are matched based on the " size" of the relative
production/attraction and inversely to the "separation" or travel difficulty between the two zones.
This step tends to represent destination choice. '
Mode Choice
The next step in the process projects a traveler's choice of desired travel mode for the given trip.
Generally, this entails choices among automobilej automobile passenger, and transit, since trips
which would be candidates for walking or bicycle have been separated out at the trip generation
stage. A small number of models do include non-motorized modes among the choice options.
Typically, this assessment is made for each trip purpose and each zone pair using a statistical
model that considers the relative attractiveness of each alternative, expressed in terms of its travel
time and cost, and weighted by the characteristics (income, vehicle ownership, etc.) of the traveler,
and estimates the probability with which each mode is likely to be chosen. Some models relate
mode choice and destination choice, but this is not common at present.
Time of Day
Most travel models attempt to account for the time of day that travel occurs, which is obviously
critical to portraying peak versus off-peak travel conditions. However, rather than treat this as an
active travel choice, the models handle this phenomenon through peaking factors. These factors
simply proportion a certain percentage of trips into the peak and off-peak periods, based on survey
observations. A separate travel network and assignment is necessary to reflect peak and off-peak
conditions. It should be noted that a fair number of areas do not have a peak (i.e., a "constrained"
flow) network, and work only with 24-hour conditions.
Traffic Assignment
Once blocked into time-of-day groups, trips are then placed into the actual transportation network
using a statistical procedure that assigns them to routes that represent the most desirable path to
their chosen destination (generally the " least time" path). This simulates the choice of travel route
by the traveler, although like time of day, it is a decision that the process "makes for them"
through a procedure that shifts trips around until the resulting volumes on the links best represent
observed traffic levels. Traffic assignment also is important as the part of the process where level
of service and congestion is determined, from which travel speeds are calculated for use elsewhere
in the process.
The realism of the forecasts from this multi-step procedure depends on the extent to which the
various steps are operating with the same information. Thus, a fair amount "of cycling of
information among the steps' is required. It is necessary to pass information from later steps (e.g.,
travel times as calculated during traffic assignment) to earlier steps (e.g., trip distribution or modal
split) as inputs to the calculations (i.e., traveler de,cisionmaking) at those stages,. Thus, preceding
steps contribute inputs to succeeding steps, with the process continuing until a defined state of
2-26 If. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
.to Reduce Transportation Emissions
equilibrium is reached. Model systems vary greatly in the degree and sophistication with which
they "feedback" information among steps and reach a satisfactory equilibrium.9
Emissions Estimation
This step is not part of the conventional transportation modeling approach, but is necessary when
estimates of vehicle emissions are required for air quality analysis. Typically, this procedure is
performed through a special model, such as MOBILE or EMFAC, which takes information on
vehicle trips, vehicle miles of travel, and speeds, by facility type, combines it with information on
the composition of the local vehicle fleet and its emissions characteristics, and generates estimates
of emissions by type of pollutant. MOBILE is used to estimate VOC, GO; and NOX emissions,
while a similar model, PART5, is used to estimate PMiQ. EMFAC also includes PMio for tail
pipe, brake wear, and tire wear emissions, while PARTS is used for re-entrained dust emissions.
Future. EMFAC versions may incorporate re-entrained dust emissions (It should be noted that
MOBILE produces emissionjorctor estimates rather than estimates of emissions amounts).
2.3.1.2 Issues
Two sets of issues are raised concerning, the suitability of these analytic tools in performing
evaluations of market-based mechanisms. The first issue is that the four-step process has various
limitations as a tool for examining the impacts of market-based approaches in general, and pricing
measures in particular. The second issue is that the analytic capabilities of candidate regions and
agencies vary widely.
N • ' i
It is generally acknowledged that the four-step process was not developed as a tool for performing
sensitive policy testing. It was originally developed and has been primarily used for generating
vehicle traffic volumes for highway planning and sizing purposes. Over time, enhancements have
been made that increase its capabilities and sophistication, but a number of shortcomings still exist
that raise concerns when examining pricing actions:
• Pricing is often not rigorously included in the model's structure. Typically, the only
. choice dimension that is sensitive to cost is mode choice.
• The accuracy of the pricing relationship in the models is challenged by the heavy
emphasis on travel time in model development, and the relative lack of good variation in
pricing conditions in most U.S. urbanized areas when compiling input data,
• ; It is expected that travelers will respond differently to pricing actions depending on their
income status, although this is not explicitly accounted for in most models
• Based on a structure that approximates behavior through zonal averages, and the
specification of the models themselves, it often is necessary to restrict consideration of a
measure to the most basic definition.
Work is underway in several regions under the U.S. DOT/FH WA's Travel Model Improvement Program to
develop a new generation of. travel models that more effectively integrate these steps, considering
individual activity scheduling, the changing of trips ipto tours, and the potential for trip sequencing and
time of day of travel, as well as recognizing the effects of transportation on growth patterns. Many of these
improvements are already available to local agencies through the Texas Transportation Institute.
U. S. Environmental Protection Agency ' _ 2-27
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
• Commercial and "through" (external-external) travel is generally not addressed in the
application of policy actions, especially pricing.
• Whereas pricing actions would be expected to have a pervasive effect on travel
behavior, ranging from changes in destination, route, time of day, or even location
choice, most models consider price only in relation to mode choice. Also, these
decisions would be expected to have some interdependency, whereas the models usually
represent behavior as a sequential-step process.
Added to these inherent limitations, there is the further complication that overall capability varies
widely by location. The model systems of areas like Portland, Oregon or San Francisco are fairly
progressive and have many of the features that would raise confidence in their use for pricing
measure analysis. However, there are many other major metropolitan areas whose models are
fairly rudimentary or which have acknowledged flaws or limitations. Added to this are a,number
of areas that might choose to look at pricing measures, but which have little or no modeling
capability at all. These might be counties or urban areas that are too small to have sophisticated
transportation planning capabilities. The planning process for many of these areas, for example,
may be only a "three-step" process (excluding mode choice) or less.
This guidance document has been developed with appreciation for the demand that pricing
measures place on the analytic process and the wide range of analytic capabilities that exist. It is
EPA's desire that areas that need additional help in reaching their attainment goals consider
market-based measures. In this regard, the candidate area should endeavor to perform the most
accurate and realistic assessment possible, within its capabilities, when submitting such measures
for emissions credit. EPA's decision as to how much credit to grant will be based on: (1) whether
the analysis supports the requested credit; and (2) whether the steps leading to implementation are
realistic. Obviously, the greater the significance and contribution of the measure to the area's
attainment goals, the more carefully the analysis procedures must be reviewed when deciding to
grant credit. However, EPA will work cooperatively with the performing agency to support its
efforts and help achieve an accurate determination of credit.
2.3.2 Identifying and Assessing Areas of Impact
This section presents how each measure is expected to affect travel and emissions. Discussion is
organized by area of impact, presented in the same order as the analysis hierarchy. This section
also describes the analytical tools that may be used to assess those impacts. -
Table 2-4 lists each of the five market mechanisms, as well as an assessment of where in the travel
demand hierarchy they would be likely to have impact. Regional growth forecasts have been
excluded as a step, since their value is almost always exogenously determined. Vehicle ownership
appears in two ways: (1) conventional vehicle ownership, which represents the number of vehicles
owned by households as a factor in travel decisionmaking for many model systems; and (2) vehicle
mix (shown as a component of the emissions generation step), which represents the types of
vehicles owned and their composition in the vehicle fleet mix.
Table 2-4 indicates where a pricing measure would have an impact in this structure, and the
expected level of importance of that impact. Impact importance is rated using the following scale:
2-28 U.S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
• to Reduce Transportation Emissions
* Primary: The impact is major and basic to the measure; every effort'should be
made to evaluate its effects.
• Secondary: The impact is judged to be an important: effect which may even exceed
the primary effect and thus must be considered; however, while its formal analytical
evaluation (i.e., a complete regional travel demand modeling analysis) may not be
possible, an area should provide documentation to show that the effect has been
considered in the most reasonable manner possible.
• Trace: Some impact would be expected, but its implications for emissions would
be minimal and no evaluation would be required.
• None: For practical purposes, no impact would be expected.
A draft version of the impact hierarchy outlined above was originally developed as the result of a
meeting that included EPA and DOT staff and contractors involved in the production of this report.
Subsequently, the draft hierarchy was reviewed by EPA and FHWA staff and modified. EPA and
FHWA review and modification led to the hierarchy outlined above.
Table 2-4 should be used with several caveats in mind. First of all, level of importance valuations
in the table are not linked to time periods. For example, while the land use impacts are often, long-
term the distinctions among primary, secondary and trace effects are generally linked to the
ultimate level of impact expected, rather than how long it takes for that impact to fully materialize.
U. S. Environmental Protection Agency
2-29
-------
I
.c*
^M
2
o
if
CO
• MM
(0
"to
c
^f
.^
(0
3
to
re
«E
*5
4-»
O
(0
Q.
E
73
0)
•»-*
(0
Q.
'c
to
3
(0
1-
'<
LI, , ,„
;,:- '
0
1
o
£
1
s
en
"o
•c
a.
SK
1
i
c
0
1
tu ix
*~«
to <»
^
*s
jg
zs
CO
i
~=
nil
o
£
ra
JC
1
«t-«
Areaofimpac
^,
•a
£=
O
O
CD
CO
..^
CO
"0
o
o
CD
CO
•a
1
CD
CO
1
t—
CO
•a
8
QJ
CO
Land Use/Activity
Allocations
^,
CO
•D
1
CD
CO
8-
r
CD
h^
CD
O
s
1—
CO
•a
1
CO
Q-
Vehicie Ownershil
8
CO
cb"
•a
3
CO
CD
JE *
Stn CD
||
D~ "o
1
CO
•a
CO
*
-
Q
.1
£
CD
0
CD
1
CD
O
CO
^ §> ^ .1 -CO-
CD c 0 W ™
£ co ^ .22 S
_CD
L^jt
"1
c5 §>
Hi
CD
O
CD
O
,
^ '
X --
S ' -
*o
i
.2
"25
u>
uu
^
o
S
"Q
M
O
S
i
§
a
1
K
8
-§
K
•c
&.
K
&
.§
1*3
o
I
CN
-------
Technical Methods for Analyzing Pricing Measures
! to Reduce Transportation Emissions
Second, level of importance valuations do not account for major differences in price levels. • For
any of the measures to have an effect on behavior, the level of the price applied would have to, be
. sufficient to be felt. At low price levels, all of the instruments would be expected to function
primarily as revenue generators, and have only marginal effects on travel behavior. However, the
relative level of impact implied by the importance valuations would be expected to stand across a
wide range of price levels.
The analytical tools or approaches described in this section are categorized as basic analysis or
advanced analysis, according to how they are rated by EPA submission requirements:
• Basic Analysis: Procedures that EPA suggests be used when formal analysis is
required if the measure is to be considered for emissions credit (that is, when the
measure is expected to have a primary and direct impact or a significant indirect
effect). ' ..-.'.•'••
• Advanced Analysis: Suggested procedures beyond the basic analytical procedures
that EPA believes would result in better estimates of the travel and emissions impact
of the measure if applied. Use of Advanced Analysis Techniques will enhance the
acceptability of an application for SIP credits. L.
In many cases, the same analytical technique is considered basic in one circumstance (e.g., when
the measure will have a primary impact), but advanced in another analytical context (e.g., when the
measure will have a secondary or trace impact); these techniques are thus categorized as "basic or
advanced analysis:'' This consolidated discussion of analytical techniques helps avoid undue;
repetition in Section 2.3.3, below, on analytical requirements for individual market-based
measures. - , • "
2.3.2.1 Land Use/Activity Allocations
Some pricing measures can affect long-term locational decisions and land use patterns and both
short and long-term individual decisions. These measures might either induce more concentrated
growth patterns that might be favorable for longer term transportation and emissions management,
or could produce divergent trends toward dispersed land use patterns and greater automobile
dependence. , '
Effects of Market-Based Measures on Land Use/Activity Allocations
Much depends on the level and type of cost impact of the particular measure, but most of the
measures could conceivably have an impact on this underlying aspect of travel behavior. Despite
the potential impact of a change in land use/locational choice, these impacts are listed as secondary
in Table 2-4 because of the uncertainty associated with projecting these changes with current
modeling tools. - -
Parking pricing would be expected to induce a locational effect, with greater preferences for
locations that provide more concentrated amenities and services and better transportation
alternatives. If applied differentially, businesses and households might be expected to show
. preferences for locations with lower parking rates. .For example, should the policy be to raise
downtown rates relative to suburban rates, this would be expected to increase the attractiveness of
suburban development. If .parking were implemented uniformly, and resulted in some areas 'having
. U. S, Environmental Protection Agency . • .'• , 2-31
-------
better design/amenity or offering better access through non-SOV modes, the measure also might
induce locational shifts to these areas.
Modal subsidies would be expected to have only a trace effect on land use decisions, unless they
were of significant size and the public perceived them to be permanent (in availability and in
relation to the cost of other modes). At high levels, dedicated subsidies might shift locational
preferences to areas where these price advantages could be exploited, that is, where transit or- other
subsidized modes are more available and useful in reaching desired destinations.
Pump charges would be expected to have a concentrating (secondary) effect on land use, since
they function to visibly raise the cost of driving for all travel purposes. Thus, households might be
expected to locate in areas where less dependence on private vehicle travel v/as'required.
Emissions fees, if scaled primarily to emissions output and not VMT, would have little more than a
trace effect on locational decisions. However, an emissions fee that is substantial, or based heavily
on VMT, could encourage households to own fewer vehicles and thus prefer areas with mixed land
uses and which otherwise provide better access to desirable locations without private vehicles.
Roadway pricing might have an important (secondary) effect. Widespread pricing would tend to
encourage locating where there is less reliance on private vehicles. Assuming only selected
facilities would be priced, households and businesses would be expected to locate so as to
minimize long-term reliance on these facilities. The impact of this on air quality is uncertain. Area-
wide implementations of road pricing would tend to reduce VMT and encourage a greater mixing
of uses. Congestion pricing would be limited in influencing location choice, particularly if fees
were peak-only. To the extent that congestion pricing leads to travel time decreases, it could lead
to more dispersed new development, could allow better access to currently congested centralized
developments, or both. Area cordon fees may lead to more transit-oriented development, provided
the priced area remains attractive relative to other locations.
Analytical Techniques for Land Use/Activity Allocations ,
Advanced Analysis
Two approaches are generally applicable for evaluating the impacts on land use/activity allocation
of market-based measures:
• Estimate impacts using, a " land use model" that is sensitive to transportation costs
(through incorporation of" generalized cost," for example);10 or
• Develop one or more growth scenarios to depict the alternate ways in which regional
growth might respond to changes in transportation costs of the type proposed. This
might be done in conjunction with application of land use models, and involve
10
'The DRAM/EMPAL model, for example, developed by University of Pennsylvania's dr. Stephen Putman,
is commonly used in the United States'to model land use changes.
2-32 [/ s. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
.•''.-'. . • to Reduce Transportation Emissions
development and integration of alternative land use/urban design configurations
supportive of alternatives (e.g., subsidized) modes.1"
Because land use/activity allocation effects are identified as secondary impacts, no formal analysis
is called for, although a statement of anticipated impacts should be included in most cases*12
2.3.2.2 Vehicle Ownership '. .
Effects of Market-Based Measures on Vehicle Ownership and Vehicle Mix
Pump charges would encourage drivers to shift to owning and/or using more fuel-efficient
vehicles. This would have an important primary effect on vehicle types, which should be captured
in the emissions modeling step, in terms of the fleet mix used to calculate regional emissions
factors. These more efficient vehicles might or might not be lower-emitting.
' ' " • * ' *
Pump charges that serve as a medium for pay-as-you-go insurance might have an effect on vehicle
ownership, although the direction of the effect would be hard to predict. Such charges would
significantly boost the cost of vehicle ownership and use for groups that do not carry insurance (a
legal violation in most cases, acting to discourage ownership). Pay-as-you-go charges, however,
may be more economical for owners currently insured and could encourage ownership.
Emissions fees, if significant enough, could affect the number of vehicles owned (hence a primary
impact). Emissions fees would likely cause households to shift ownership to (and possibly use)
newer and/or lower-emitting vehicles. This would be an important primary effect that would need
to be reflected in the emissions modeling step through adjustments to the fleet mix.
An outstanding example of this approach is Portland, Oregon's "Region 2040" planning process, and a
closely related analysis process by 1000 Friends of Oregon, the "Land Use, Transportation, and Air Quality
Connection" (LUTRAQ). Region 2040 information is available from Portland Metro, and LUTRAQ
information from 1000 Friends of Oregon See, for example, 1000 Friends of Oregon, "Making the
Connections: A Summary of the LUTRAQ Project, February 1997. ;
2 In a relevant recent case, a US District court ruled that as part of planning for a major new highway, Illinois
DOT would need to. develop a separate set of socioeconomic/land use forecasts based on what would likely
occur with vs. without construction of the road. A single set of socioeconomic/land use assumptions for both
build and no-build scenarios for evaluation of ozone precursors was deemed to fail the legal requirements of
the National Environmental Policy Act. US District Court (Northern District of Illinois, Eastern Division),
Case No. 96 C 4768. Sierra Club, Illinois Chapter; South Corridor Against the Tollway, Inc.; Environmental
Law and Policy Center of the Midwest; and Business and Professional People for the Public Interest vs. US
DOT/FHWA, Illinois FHWA, Illinois DOT, and Illinois State Toll Highway Authority.
U. S. Environmental Protection Agency ' 2-33
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
No other market-based measures would be expected to have a major direct effect on the number of
vehicles owned by households. The primary effect on vehicle ownership levels for these other
measures would be a result of induced shifts in location (land use/activity allocations) to areas
where fewer vehicles were required.
Analytical Techniques for Vehicle Ownership and Vehicle Mix
Basic or Advanced Analysis
Analytical tools used for evaluating ownership patterns apply primarily to vehicle mix because that
is the factor expected to be most affected by market-based measures. Approaches include:
• Analytic procedures such as hedonic models to forecast shifts in ownership patterns;
• Research studies on shifts in vehicle ownership in response to price, and application
of factor methods to estimate shifts; and
• Special analyses, such as stated preference surveys, to investigate likely responses to
ownership that might result from alternative fee structures.
2.3.2.3 Trip Generation
Market-based measures can have a range of impacts on the generation of trips from households.
Probably the key factor in terms of the impact on trip generation is the overall scope of the measure
and its comprehensiveness in" increasing the overall cost of travel by single-occupant vehicles.
Ideally, trip generation should account for all person-trips. Non-motorized trips are generally not
accounted for in current modeling systems, which could constitute an important limitation when
evaluating pricing systems.
Effects of Market-Based Measures on Trip Generation
Many analysts assume that the number of trips generated by a household is not particularly
sensitive to transportation factors, including cost. They suggest that a household's propensity to
travel is determined by its need to satisfy specific activities, which are dictated by its size and
socioeconomic composition. Rather than forego activities and make fewer trips in response to
higher prices for vehicle travel, households would be expected to make less severe changes in their
pattern of trip making — for example, more trips to closer/different destinations or, shift to other
modes. These are important effects, but are generally dealt with elsewhere in the travel analysis
hierarchy.
There are growing challenges to this view, however, particularly in light of new information and
telecommunications systems that are reshaping our economy and communities. For example, an
increasing number of workers are opting to work at home under telecommuting arrangements;
many others have shifted to a modified work schedule. In these cases, the number of trips made to
a physical work site is reduced and/or the time of travel changed, suggesting that the household
might make fewer total trips as a result.
Similarly, shopping and entertainment needs are being increasingly satisfied through cable TV, the
Internet, video movie rentals, deliveries, etc. There also are other ways in which households could
shift their activity patterns to group trips together into trip "chains," possibly accomplishing
j , • • '
2*34 -U. S. Environmental Protection Agency
-------
- Tecfiniqai Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
multiple functions in a single trip. Such adjustments to activity patterns could show up in overall
trip rates in response to higher travel costs, although their impact is presently not well understood.
Parking pricing could influence trip generation in many different ways and depends heavily on the
details of the pricing changes. For example,.requiring employers to charge employees for parking
(or instituting a "cash-out" program) could have primary impacts on work trips but, probably,
virtually no impacts on non-work trips. If non:work parking were priced, one could expect far
larger impacts, though their character might differ. Impacts of non-work parking pricing might
emphasize long-term land use changes by retail locations. Since land use effects are captured
elsewhere in .the modeling system, trip generation would probably only have a trace impact for
non-work trips. " '•''".
Modal subsidies would only have a trace impact on trip generation. This would probably tend to
affect work travel more than non-work travel. Minor increases in carpooling and transit use would
be likely. • '
Pump charges could have a secondary impact on trip generation. In the short run, this would be
reflected by better scheduling of activities and vehicle usage within households (often a household
with more than one vehicle will increase use of the more fuel efficient vehicle).
Roadway pricing could have a primary impact on trip generation if applied area-wide. An increase
in the overall cost of travel would be expected to result in household activities being more
efficiently scheduled and coordinated. Limited roadway pricing, such as tolls on individual
facilities or limited congestion pricing would only be expected to have a trace impact.
Emissions fees are not expected to have more than a trace impact on trip generation. Note that if
these fees are mileage-based, arid if the mileage-based component of fees is significant, they could
have a greater effect.
Analytical Techniques for Trip Generation
Basic Analysis . . .
Aside from uncertainty regarding the actual impacts, practical concerns have been raised regarding
the modeling conventions used to assess impacts. Frequently, trip generation models estimate only,.
those household person trips that may be served by vehicle. They do not include trips that would
be made by walking or bicycle. Therefore, in situations where a household may respond to a
change in travel conditions by shifting some trips to pedestrian or bike travel, this shift would not
be captured accurately in the trip generation analytical step. .
Some regions employ models that are even more restrictive; they consider only trips made in
vehicles, not person trips, typically, these are smaller areas where the private vehicle is the
overwhelmingly dominant mode and the planning process is less complex in that they are not
concerned with transit or non-motorized alternatives. Thus, if a region's trip generation model
estimates all person trips, the effect of pricing on trip generation would be expected to be at a trace
level, since allowance would be made for vehicle trips to shift to non-motorized modes.
However, if trig .generation analysis is based only on vehicle or vehicle-eligible trips, then pricing
would be expected to show at least a secondaryeffect, at least in the case of pump charges, but
U.S. Environmental Protection Agency • ' , 2-35
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
possibly also for parking pricing and modal subsidies, because vehicle trips would shift to non-
motorized modes would " disappear" from the inventory.
Advanced Analysis
In general, if the trip generation analysis deals only with motorized-vehicle trips, then an
adjustment should be made to the vehicle trip generation estimates to reflect "shifts to non-
motorized or non-automobile-driver options. Wherever possible, trip generation analysis should be
performed separately for each purpose and income strata, to support subsequent analysis of trip
distribution and mode choice.
Moreover, if land use analysis suggests shifts in activity or land use patterns, then the vehicle
ownership and trip generation analytical steps should be repeated to reflect relocation of activity
and its association with potentially different rates of trip making and vehicle ownership.
2.3.2.4 Trip Distribution
Effects of Market-Based Measures on Trip Distribution
All market-based measures^ except emissions fees, would be expected to influence choice of
destination, that is, trip distribution. As noted earlier, if emissions fees are mileage-based, and if the
mileage-based component of these fees is significant, they could have a greater effect. The same
caveat applies to all references to the effects of emissions fees throughout this section. Parking
pricing would be expected to influence destination choice at a secondary level. Increased parking
rates would increase the attraction to destinations that provide better access to desirable locations
without private vehicles. If parking rates were uneven in their application, travelers would be
attracted to comparable destinations with zero or lower parking rates, unless the priced area also
offered better access, more services or a higher level of amenities.
I ' /' _-,•"'
Modal subsidies would be expected to have a secondary effect. Lowered costs of selected non-
SOV modes (through subsidies) would be expected to increase the attractiveness of those
destinations accessed by subsidized modes.
Pump charges would have a secondary effect on destination choice. By raising the comparative
costs of private vehicle travel, this measure would be expected to increase the attractiveness of
closer destinations and/or destinations easily reached by alternative modes.
U. S. Environmental Protection Agency
-------
•••^••^•^^•^•^•^^••1
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Emissions fees could have a trace effect on choice of destination. If the mileage-based component
of the charge were significant, then it could encourage travel patterns that would reduce VMT,
although the imposition of this fee on a periodic/annual basis would be expected to dampen its
effect as a signal to daily travel. .
Roadway pricing would be expected to have the largest effect of all pricing measures on
destination choice, constituting & primary impact. Prices would tend to induce consumers to reduce
vehicle trips, to reduce the length of trips, and also to consolidate destinations. If prices were
imposed on select facilities, travelers would be expected-to re-examine their options (particularly
for non-work travel) and choose to:
• Travel to alternative (possibly more distant) destinations not subject to pricing;
• Travel to alternative destinations served by priced facilities that offer good
alternatives, special amenities, or are closer than their original choices; or
• Travel to the same or alternative destinations, where they obtain quicker access at an
• acceptable price using priced facilities (more likely among higher income travelers).
Area-wide road pricing would result in shifts to closer destinations. Isolated road-pricing systems
could, in some cases, result in increased travel.
Analytical Techniques for Trip Distribution
Basic Or Advanced Analysis : > ' •
Areas that have trip distribution models that incorporate price sensitivity directly through a
generalized cost procedure can account for the effects of market-based measures explicitly. Many
trip distribution models are sensitive to travel time but not cost. Some agencies address this in their
models by incorporating a generalized cost formula. This characterizes the full cost of travel by
private vehicle from each origin to each destination in terms of both the travel time "cost" and the
monetary cost (use of travel time equivalents is discussed in Section 3.3.5.3.). Some models also
consider the time and money cost of transit, where it is available, and average it into the overall
measure of generalized cost. - . ."
If an agency has a model system with a generalized cost procedure that can be made to reflect the
cost of market-based measures, this should be used to perform a revised trip distribution analysis
that incorporates the proposed cost measures. Moreover, trip distribution analysis should be
performed separately by trip purpose and income strata.
U. S. Environmental Protection Agency 2-37
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
2.3.2.5 Mode Choice
Effects of Market-Based Measures on Mode Choice
All of the pricing measures, with the exception of emissions fees, would be expected to have a
primary effect on the choice of travel mode.
Parking pricing would. clearly serve to discourage driving, and, in particular, driving alone, in
favor of other alternatives, and will shift some .shorter trips to walking or biking if suitable
alternatives and access exist.
Modal subsidies would create a direct attraction to the mode or modes being subsidized.
Pump charges would encourage use of alternative modes that would allow the traveler to avoid or
reduce the higher cost of private vehicle travel.
Emissions fees could have & trace effect on mode choice through the mileage-based component of
the charge. Again, however, 'the effect would be greatest if fees are collected frequently and/or
there are frequent education and reminder efforts.
»
Roadway pricing would encourage use of alternative modes to avoid or reduce the cost of priced
facilities, although this effect would be in combination with consideration of alternative
destinations, routes and times of day at a much higher level than with the other measures.
Analytical Techniques for Mode Choice
Basic or Advanced Analysis
With the exception of emissions fees, all market-based measures would be expected to have a
primary effect on mode choice. The performing agency can conduct the basic analysis with any of
the three following options:
• Accepted mode choice model;
• Accepted " quick response" methods; or
• Elasticity methods.
An agency with access to a mode choice model is expected to use that model to conduct any
required analyses of the impact of market-based measures. It may be permissible to adjust the
factors included in these models or their coefficient values, provided adequate documentation is
supplied on the nature of and need for the adjustment.
Areas without a mode choice model but with access,to person trip tables may consider acquiring
one of several "quick response" methods, which are widely available. These include: the TDM
Evaluation Model from the Federal Highway Administration, the QRS planning system, and
various methodologies developed to support CMAQ project analyses. Information on these tools is
given in Section 3.3.6.9.
2-38 U.S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
If an agency does not have reasonable access to either a mode choice model or one of the above
surrogate procedures, analysis may be based on application of elasticities from research studies or
from the mode choice model of similar urban areas. Section 33.6.9 provides references on sources
for such elasticities and guidance toward their proper application.
Advanced Analysis • . ., . '
Advanced analysis of impacts on mode choice would use current generation, more sophisticated, -
mode choice models such as "nested" multinomial logit models, which permit testing of a wide
array of alternative modes. The range of transportation service and socioeconomic variables that
they incorporate offers greater flexibility in relating strategies and more realistic estimates of travel
response. Some models have separate coefficients for categorically different cost'items, such as
transit fares, parking subsidies or parking costs, and HOV costs,, and should result in more" accurate
estimates of response to corresponding policies. Models that reflect the interaction of mode
choices and destination choices are desirable and EPA encourages their development.
More realistic analysis will result from application of separate models for each primary trip
purpose, including home-based work, .home-based other, and non-home based trips. Many
agencies, however, do not have mode choice models for non-work travel. Accuracy would be
further enhanced by accounting for effects by income strata, as well as trip purpose, in order to
reflect differences in cost sensitivity by income level. This would also help analyze the equity-
related issue of differential impact by income group. ,
Finally, model accuracy can increase if models can capture the attractiveness of an area to other
travel options, including non-auto travel. Just as the design of the road system increases or
decreases the " impedance" a driver experiences, so the walking and cycling infrastructure helps
determine whether, how often, and how far people walk or bicycle. And just as impedance for
drivers can be quantified, new work has significantly improved the quantification of determinants
of walking and bicycling. Recent work on Pedestrian Environment Factors and, Urban Density
Factors has helped improve modeling in Portland, for example.13
2.3.2.6 Time of Day
Effects of Market-Based Measures on Time of Day
Only three of the described market-based measures would be expected to have any effect at all on
travel time of day; neither pump charges nor emissions fees would influence time patterns since
assessment is not linked to specific trips.
Parking pricing could have a secondary effect on time of day if the differential between peak rates
and off-peak rates were sufficient to shift users with flexibility to off-peak travel. In many, cases,
however, impacts would be at a trace level.
Modal subsidies that favor alternatives (mode and/or specific service within that mode) used
primarily in peak periods could shift users from unsubsidized modes/services used off-peak to the
13
See, for example, 1000 Friends of Oregon, Making the Land Use, Transportation, Air Quality Connection:
Volume 4A, the Pedestrian Environment. 1993. Portland: 1000 Friends of Oregon. - ' -.
U. S. Environmental Protection Agency . ' . , ' 2-39
-------
Technical Methods for Analyzing Pricing Measures •
to Reduce Transportation Emissions
subsidized service during peak periods. Nonetheless, only trace effects would be expected for the
level of subsidies generally proposed unless subsidies vary by time of day.
Roadway pricing in the specific case of congestion pricing would be expected to have & primary
effect on time of day choice. Higher peak-period prices would be expected to shift lower-valued
trips to off-peak travel and would draw some trips into the peak if a higher level of service becomes
available as a result of the pricing program.
'i ' •
Analytical Techniques for Time of Day
Basic or Advanced Analysis
Time-of-day analysis procedures are still uncommon in transportation planning practice. Some
agencies have time-of-day models. For those that do not, analysis could be based on procedures
borrowed from another location or based on a special analysis at the affected site to identify
relationships and estimate behavioral factors. More specifically, analysis of the impacts of market-
based measures expected to have a measurable impact on time of day could be based on any of
, three approaches:
• Use of a time of day choice model acquired from another location (see Section
3.3.7);
• Consultation of references on time of day sensitivity, which could include analyses
of other area programs, transit or parking peak pricing results, or other special
analyses; or
• Performance of a special stated preference survey analysis to develop a model for
the area under investigation.
2.3.2.7 Traffic Assignment/Route Choice
Effects of Market-Based Measures on Traffic Assignment/Route Choice
While all of the pricing measures affect traffic assignment in terms of revised traffic patterns, only
roadway pricing directly targets route choice.
Roadway pricing forces travelers to make conscious decisions about spatial patterns and, as such,
has a primary impact on traffic assignment. Travelers may have a choice between routes that are
tolled but that offer a higher level of service/travel time and another set of routes that are not tolled
but that have a lower level of service/travel time. As with destinations, the choices ultimately made
depend on the specific characteristics of the pricing scheme and the available travel alternatives.
2-40 jj_ $. Environmental Protection Agency
-------
Technical Methods for Analysing Pricing Measures
, to Reduce Transportation Emissions
Analytical Techniques for Traffic Assignment/Route Choice - .
Basic or Advanced Analysis , ^
Roadway pricing explicitly forces travelers to evaluate and respond to the trade-off between tolled
. roadways that may have a higher level of service and toll-free routes'that may be more congested.
They also may choose shorter trips or forego some trips that would previously have been taken.
The anticipated close links between roadway pricing and traffic assignment (route choice) mandate
analysis of these effects. Yet, traveler route choice, is not directly handled by most traffic
assignment models. There are various options for addressing anticipated impacts, including:
• Use of procedures that represent toll and free path opportunities to travelers in an
assignment routine. These are relatively advanced techniques, however (see Section
3.3.8.5); or
• Representation of the effect of tolls for a given facility as an increase in travel time
, through use of value of time relationships to estimate the appropriate value. This
• raises the cost of a tolled facility to a traveler considering its use (also see Section
,3.3.8.5).
2.3.2.8 Emissions /Vehicle Mix
Effects of Market-Based Measures on Emissions Vehicle Mix
All market-based measures should ultimately affect the level of emissions generated. Two market-
based measures—pump charges and emissions fees—could directly affect vehicle mix and, through
that, emissions rates. .
Pump charges could induce shifts in favor of vehicles with greater fuel economy. The net effect
on emissions is the result of two factors: lower fuel consumption per mile traveled (and potentially
overall) and possibly different emissions characteristics (rate of emissions per mile). The overall
relationship between pump charges and vehicle mix/emissions could be considered primary,
though the indirect link of pump charges working through increased efficiency and lower total fuel
consumption qould be considered secondary. Pump charges would induce travelers to shift to more
efficient vehicles, partially offsetting the benefits of fu,el price increases.
Emissions fees would act similarly to the pump charge, except that by inducing a shift in the
vehicle fleet toward cleaner vehicles, it would constitute a: primary effect. Influence of a VMT
component of emissions fees would also show up in lower aggregate mileage rates within the fleet.
U. S. Environmental Protection Agency ' • 2-4.1
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Analytical Techniques for Emissions /Vehicle Mix
Basic Analysis
Agencies should use an approved .emission factor model to calculate emissions, with factors
computed to reflect any program-induced adjustments in the regional vehicle mix. Areas without
formal emissions factor models would be expected to explain how the average emissions rates
would change with the induced shifts in vehicle fleet composition and/or use.
Basic or Advanced Analysis
Analysis of anticipated shifts in vehicle fleet composition or usage can utilize the following
approaches:
• Use of hedonic models (models of consumer satisfaction with particular vehicle
attributes) to project changes in vehicle ownership in relation to cost differences;
• Consultation of research studies on shifts in vehicle ownership in response to price,
and application of factor methods to estimate shifts; or
• Performance of special analyses (using revealed or stated preference survey
methods) to forecast possible shifts in vehicle ownership and use.
These methods are further detailed in Chapter 3, Section 3.3.3.
2.3.3 Analysis Guidance
This section presents EPA's recommendations for evaluation of market-based mechanisms and
describes the types of analyses that EPA expects to see accompany submissions of the respective
measures for emissions credit. The guidance descriptions in this chapter are at a general level,
designed to communicate to managers and decision-makers EPA's general thinking; specific step-
by-step guidance suitable to direct technical staff in the actual evaluation practices is provided in
Chapter 3, further supported by examples in Appendices A and B.
The table on the following page summarizes rules of thumb for submissions to EPA; impacts are
assessed for individual market-based measures at each stage of the analysis hierarchy (refer to
Table 2.3 for description of anticipated impacts). Clearly, the more careful the analysis, the more
credit EPA can be expected to approve. -
• ' U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
; fapectejnpact .1 * Acceptable Analysis
^Primary effect-^ Fom?a/ analysis expected for impact on that step of analysis hierarchy.
Must include some treatment oi impact on that stage of analysis hierarchy, though
*• J requirements are less formal. .
v ~ Agency expected to fate nofe of possible side effects, which could occur under special
--*• applications of the measure. .
<-3/aceeffect v
In laying out its recommendations for effective analysis, EPA has taken into consideration the wide
. range of capabilities that are available to those considering market-based measures. Because of the
strong likelihood that there will be sites with minimal analysis capabilities, the guidance
recommendations have been made flexible enough to allow these states to.respond.
This flexibility does not take away from the position, however, that an. accurate, comprehensive
analysis is extremely important. Accuracy and completeness are important to EPA's review and
decision to grant emissions credit. Therefore, agencies proposing pricing measures should consult
with a wide range of parties to get as much assistance as possible in developing good analytical
approaches. Furthermore, an accurate analysis also is essential to the performing agency's ability
to assess the impacts and tradeoffs that are critical to local acceptance and implementation. EPA's
position is to support the best possible analysis, t6 make allowances for states with limited
capabilities, but to expect that states will take every reasonable step to perform the best analysis
within their ability.
The remainder of this section presents analysis recommendations for each of the five measure
groups: parking pricing, modal subsidies, pump charges, emissions fees^ and road pricing. To
avoid repetitive discussion of analytical approaches, EPA submission components are often noted
only briefly with reference to the more extended discussion of basic or advanced analysis organized
by impact category in Section 2.3.2. Recommendations for analysis are separated into basic
analysis and advanced analysis and, within each lof these, according to expected level of impact
from primary to less important effects.
Advanced analysis is optional from-the standpoint of EPA guidance. This reflects an
acknowledgment of current limitations, but also willingness on EPA's part to work with all areas in
exploring advanced strategies to deal with long-term air quality concerns. Nonetheless^ EPA
expects that those areas with advanced, planning tools will use those tools to their best ability.
Areas that do not have such tools or capabilities are encouraged to strive to acquire those advanced
capabilities, both to serve their own current needs and to support other air quality and planning
requirements that are likely to be part.of future planning and funding tests.
In addition to attention to analysis tools and procedures, EPA places importance on the quality of
data used to conduct the analyses, particularly with regard to data age, coverage, and accuracy.
The level of scrutiny applied by EPA to analyses that accompany submissions of market-based
measures will be tailored to each circumstance. In particular, the level of credit claimed for the
measure and its importance in achieving a given area's attainment or maintenance goals will be a
U. S. Environmental Protection Agency . 2-43
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
factor considered by EPA in its review. More rigorous scrutiny can be offset through a more
accurate and defensible analysis.
2.3.3.1 Parking Pricing
Suggestions for basic analysis and advanced analysis pertaining to parking pricing are illustrated in
Figure 2.2a.
Basic Analysis for Parking Pricing Measures
Mode Choice: Toe primary effect of parking pricing on travel and emissions would be through its
impact on mode choice. A submission that seeks emissions credit for parking pricing must pay
particular attention to the shift of travelers from private vehicles, especially single-occupant
vehicles, to transit, ridesharing or non-motorized alternatives.
Given the importance of parking pricing on mode choice, submitting agencies should use the most
sophisticated procedures available to them, following the order of preference described in Section
2.3.2.5 on analytical techniques for mode choice—basic or advanced analysis (i.e., mode choice
models, "quick response" techniques, elasticity analysis).
In performing appropriate analysis, the performing agency must address, either explicitly in its
analysis or through a listing of assumptions, how it has accounted for the following types of
questions:
• What percentage of parking situations and consumers are affected by the proposed
policy, and how is this reflected in the tested rate?
• What travel groups are affected by the proposed price (e.g., commuters, shoppers,
etc.) and how is this reflected in the trip populations to which the analysis is applied?
2-44
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
.••'•'.. . to Reduce Transportation Emissions
Figure 2.2a Elements of Analysis for: Parking Pricing
ELEMENTS of Basic Analysis ANALYSIS STEP RECOMMENDATIONS for Advanced Analysis
___•_,_ (Type of Impact)
Formal analysis not required;
Statement of possible effects if rates are
applied with geographic differences.
None
None
Formal analysis not required;
Statement of possible effects if differential
rates are applied. '
Analyze through mode choice model, if
available; •
^Elasticity factor" or "quick response" analysis
if no model;
Document sources and assumptions.
None ' . '
Standard assignment analysis, if capability
exists.
None.
Standard analysis, if emissions model
available;
If no model, apply emissions factors to the
change in vehicle trips and VMT.
LAJiDUSE
ie i
^ #£'.•
TRIP'
p MO|>E CHOICE,
^(Primary) •*, l
:(Secoijdary)
. ^ASSIGNMENT
.-•; • (Trice)
RATKAflEHICLE"'
• "" .wlx \v
" " (None) \ -
£x*
T*f <.-V «/ " ,*„
« TRIP,,, -
GENERATION .
^^fr^o^ - >
, •£ * * ^
^-' (S^oondtary) "-;
Develop alternative growth scenarios or apply land use models
capable of considering changes in transportation access and
cost; ...
Integrate alternative land use/urban design supportive of
alternative modes into land use models, or adjust subjectively to
evaluate: ' , ', . '
Re-run if shifts in land use (above). •
Re-run if shifts in land use .or vehicle ownership;
Try to incorporate change in cost; ' ,
Run separately by purpose and income;
If using motorized vehicle trip tables only, adjust trip rates to
reflect shifts to non-motorized and non-automobile driver modes
using market share analysis or other model.
Run separately by time of day if time-variant rates; ',
Run separately by purpose and income; •
Explicitly model to reflect impact of change in parking cost.
Use separate models for each purpose; -
Run separately by purpose and income;
Separate coefficient for parking cost;
Account for pass-through effect and HOV exemption/ discount;
Estimate shift to non-motorized/ non-travel modes.
Try to model if parking prices vary by time of day or short/ long
term;
Model by trip purpose and income.
Standard analysis
None
.ANALYSIS^
Standard analysis.
U. S. Environmental Protection Agency
'2-45
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
• What proportion of the tested fee is assumed to be passed through to the traveler
(affects tested rate)?
• Does the policy provide for exclusion of or parking discounts to HOVs?
• Is the dollar value of the policy being tested consistent with the year in which the
elasticity was developed?
• What other factors are likely to be introduced or changed during the evaluation
period that could affect the implementation or traveler response to the proposed
application?
Traffic Assignment and Emissions Analysis/Vehicle Mix: Subsequent to the required mode
choice analysis, areas with full model systems are expected to perform traffic assignment analyses
and apply an emissions factor model to determine the resulting change in emissions. If these
procedures are not available, emissions impacts may be estimated by applying emissions factoring
methods to estimated net changes in vehicle trips and VMT, taking cold start effects into account if
they are considered significant.
Land Use and Trip Distribution: Parking pricing that is applied evenly in an area could have
important secondary impacts on land use/activity allocations and trip distribution. Mode shifts
would tend to shift people to locations with more available services. Parking pricing that is
applied unevenly in an area also could have important secondary impacts on land use/activity
allocations and trip distribution. If the proposed parking pricing strategy will not be uniform.
across different geographic and travel purpose markets, then the performing agency should address
this concern in its submission.
While a formal analysis of these effects is not required, the submitting agency should provide at
least a statement of the potential impact from the proposed strategy on near-term choice of
destination and longer-term land use trends, and the anticipated implications of these shifts for air
quality. This statement should include acknowledgment by decision-makers of the anticipated
impact.
Suggestions for Advanced Analysis of Parking Pricing Measures
Use of the following advanced analysis techniques could produce a more comprehensive and
accurate analysis of the impact of parking pricing; these are listed in approximate order of
presumed importance of effects (most important first).
Mode Choice: More advanced evaluation of mode choice impacts would utilize more
sophisticated models (described in Section 2.3.2.5 as advanced analysis for mode choice) such as
" nested" multinomial logit models, incorporating parking cost coefficients (where possible), and
differentiating among income strata and trip purpose.
Trip Distribution: Enhanced analysis would be based on use of a trip distribution model sensitive
to trip costs such as parking pricing, or on a model that uses time-equivalencies to account for such
costs (see Section 2.3.2.4 on advanced analysis for trip distribution).
Land Use/Activity Allocations: Programs that introduce major changes in the level or pattern of
parking prices could affect longer term land use and activity patterns. Advanced analysis may be
warranted, using procedures such as a " land use model" sensitive to cost and/or growth scenarios
2-46 U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
'-••:- to Reduce Transportation Emissions
(see Section 2.3.2.1 above on advanced techniques and, for a more technical discussion, Section
3.3.1 of the following chapter). ,
Vehicle Ownership and Trip Generation: If the trip generation procedure deals only with
motorized-vehicle trips, then an adjustment should be made to the vehicle trip generation estimates
to reflect shifts to non-motorized or non-automobile-driver options; the analytical approaches
described in Section 2.3.2.3 ar,e appropriate for advanced analyses of parking pricing effects.
Time of Day. Unless the parking policy results in major changes in cost structure between peak
and off-peak periods, it is unlikely that a sufficient time-of-day travel shift would occur that would
be discernible by analysis. Advanced time of day analysis, if warranted,- would follow those
approaches described in Section 2.3.2.6, above (see also the more technical discussion in Section
3.3.7); these include choice models, other references, or special surveys.
2.3.3.2 Modal Subsidies
Suggestions for both the basic analysis and the advanced analysis pertaining to Modal Subsidies
are illustrated in Figure 2.2b and described below. •
Basic Analysis for Modal Subsidy Measures
Mode Choice: Because the only primary effect of modal subsidies on travel and emissions is
through mode choice, to be acceptable, an analysis should demonstrate a convincing evaluation of
the modal split effect of the proposed subsidy. (Some effects on trip distribution may be likely
with mode shifts).
The options for completing the basic analysis include, in order of preference, use of: accepted
mode choice model; accepted "quick response" methods; or elasticity methods (see basic or
advanced analysis, Section 2.3.2.5).
The performing agency also, is expected to relate how it has accounted for the following types of
questions:
What percentage of travelers will be affected by the proposed policy, in terms of receiving a
subsidy, and how is this reflected in the tested rate?
• What travel groups are affected by the subsidy (e,g., commuters, shoppers, etc.) and
how is this reflected in the travel base which has been used for analysis?
• What is the actual subsidy that is expected to be passed through to the traveler? '
• Does the policy reward different groups at different levels (e.g., ridesharers versus
transit users) and how does the analysis represent the competition among modes?
• Is the dollar value of the policy being tested consistent with the year in which the
elasticity was developed? • , -
U. S. Environmental Protection Agency . ~'! ~~' ~ \ ! -2-47
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Figure 2.2b Elements of Analysis for: Modal Subsidies
ELEMENTS of Basic Analysis Jiype of Implct)' RECOWIMENDATIONS for Advanced Analysis
If significant impact on inter-mode cost relationships, deve
alternative growth scenarios or run land use models with
consideration of cost;
Formal analysis not required;
Statement of possible effects if subsidy levels
are significant.
None
None
Formal analysts not required;
Statement of possible effects if subsidies are
at significant level.
Analyze through mode choice model, if
available; ' .
"Elasticity factor" or 'quick response" analysis
if no model;
Document sources and assumptions.
Formal analysis not required;
Statement of possible effects if subsidies are
significant.
Standard analysis if have assignment
procedure.
None
Standard analysis if have emissions model;
If no model, apply emissions factors to the
change in vehicle trips and VMT.
LAND USE
ALLOCATION
(Trace)
.« VEHICLE
tj-" OWNERSHIP
(Trace)
TRIP
GENERATION j
(Trace)
TRIP
i rvic
DISTRIBUTION
(Secondary)
MODE CHOICE
(Primary)
•TIME OF DAY/
PEAKING
(Secondary)
TRAFFIC
ASSIGNMENT
(Trace) fe
EMISSIONS
RATE/ VEHICLE
MIX
(None)
EMISSIONS
ANALYSIS"
ihips, develop
aiiGiiiauve yiuwin sueiidnus ui iuii tana use models with
consideration of cost;
Integrate alternate land use/ urban design supportive of
alternative modes.
Re-run if shifts in land use. (if significant mode shift, ownership
should drop).
Re-run if shifts in land use or vehicle ownership;
Try to incorporate change in subsidies;
Run separately by purpose and income;
If vehicle-based trip generation, account for non-motorized/ non-
automobile driver shifts.
Run separately by purpose and income strata;
Try to reflect impact of change in cost among alternatives and
mode shift.
Use separate models for each purpose;
Run separately by purpose and income;
Make certain models appropriately sensitive to the different
subsidized modes/ instruments; ,
Perform supplemental analysis to assess effects on use of
subsidized or non-motorized modes.
Perform supplemental analysis to account for peak/ off-peak
effect of subsidies.
Standard analysis
None
Standard analysis
2-48
U. 5. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
• • to Reduce Transportation Emissions
•, What other factors are likely to be introduced or changed during the evaluation
period that could affect the implementation or traveler response to the proposed
application?
Traffic Assignment and Emissions Analysis/Vehicle Mix: Subsequent to the require'd mode
choice analysis, areas with full model systems'are expected to perform traffic assignment arialysis
and apply an emissions factor model to determine the resulting change in emissions. If these
procedures are not available, emissions impacts may be estimated by applying emissions factoring
methods to estimated net changes in vehicle trips and VMT, taking cold start effects into account if
they are considered significant. "'•'',.-.
Land Use, Trip Distribution and Time of Day: Modal subsidies also may have secondary impacts
on land use/activity allocations, trip distribution, and time of day. Impacts on land use trends and
trip distribution would be anticipated only from significant levels of subsidy. Time of day impacts
might occur if subsidized rates drew travelers into or out of peak periods. In such cases, although
a formal analysis would not be required, EPA expects a statement of anticipated impacts.' ^
Suggestions for Advanced Analysis of Modal Subsidy Measures
Expanding the evaluation to include the following components would generate a more
comprehensive and accurate analysis of the impact of modal subsidies; these suggestions are listed
in descending order of importance.
Mode Choice: More extended analysis would incorporate those approaches described as advanced
analysis in Section 2.3.2.5 above (i.e., mode choice models that specifically break out and allow
for interactions among each of the candidate travel, modes), may include separate coefficients for
parking costs, transit fares, etc., and are preferably run for each primary trip purpose, including
home-based work, home-based other, and non-home based trips. Mode choice should be
performed separately for each purpose and income strata.
Trip Distribution: Extended evaluation would incorporate those procedures described under
advanced analysis in Section 2.3,2.4 on trip distribution (i.e. models that are sensitive to cost
considerations and that distinguish among trip purposes and income strata).
Time of Day: Only if the subsidy policy results in major changes in cost structure between peak
and off-peak should a formal analysis be attempted of time of day travel shifts. Such analyses are
described in Section 2.3.2.6 (and Section 3.3.7) and include models borrowed from other locations,
previous research, or conducting special surveys.
Land Use/Activity Allocations: If significant levels of modal subsidy suggest a different cost
structure for regional travel, estimating this effect may be important. Because modal subsidies .
would result in across-the-board changes in the cost of travel, significant subsidies could lead to
longer-term shifts in residential and employment locations to areas that require less automobile
dependence. Approaches for extended evaluation include those described in Section 2.3.2.1 on
advanced analysis for land use/activity allocations, including land use models sensitive to such
costs and growth scenarios. - • '-
Vehicle Ownership and Trip Generation: On their own, modal subsidies would not be expected
to cause households to own fewer vehicles or make a different number of trips. If alternative land
U. S. Environmental Protection Agency , , 2.49
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
use patterns are determined in the previous step, then the vehicle ownership and trip generation
steps should be re-run to reflect the relocation of activity and its association with the respective
vehicle ownership and trip rates in the new location.
If the trip generation procedure deals only with motorized-vehicle trips, then an adjustment should
be made to the vehicle trip generation estimates to reflect shift to non-motorized or non-
automobile-driver options.
Wherever possible, trip generation analysis should be performed separately for each purpose and
income strata, to support the subsequent analysis of trip distribution and mode choice.
2.3.3.3 Pump Charges and VMT Fees
Suggestions for both the basic analysis and the advanced analysis pertaining to pump charges are
illustrated in Figure 2.2c. The primary effects triggered by pump charges would be on mode
choice and vehicle mix. Higher fuel prices would be expected to reduce the demand for private
vehicle travel and increase ownership and use of higher-mpg vehicles or alternative-fuels/electric
cars.
Basic Analysis for Pump Charges
Mode Choice: Basic analysis includes those techniques described as basic or advanced analysis in
Section 2.3.2.5 on mode choice impacts, that is (in order of preference): mode choice models,
"quick response" methods, or elasticity methods.
The performing agency also is expected to indicate how it has accounted for the following factors
or concerns:
• What percentage of regional travelers will be affected by the increased pump price,
versus those who would be expected to evade the price increase by purchasing fuel
outside the area; how is this reflected in the assumptions regarding the tested price?
• What average cost per mile is being assumed, and upon what average mpg is it
based?
• Is the dollar value of the policy being tested consistent with the year in which the
elasticity was developed?
2-50 • U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Figure 2.2c Elements of Analysis for: Pump Charges
ELEMENTS of Basic Analysis ANALYSIS STEP
' ': • • , • ' . • (Type of Impact) REC°MMENDATIONS for Advanced Analysis
Formal analysis not required;
Statement of possible impacts.
. None
If trip generation for vehicle trips only, account
for non-motorized and non-automobile driver
modes. : .
Formal analysis not required;
Statement of possible effects.
Analyze through mode choice model, if
available; '
"Elasticity factor" or'quick response" analysis
if no model;
Document sources and assumptions.
None
Standard assignment analysis if have
assignment procedure.
Indicate assumptions regarding change in
vehicle mix to reflect shift in mpg anticipated.
Standard analysis if have emissions model; .--
If no model, apply emissions factors to change
in vehicle trips and VMT.
LAND USE
ALLOpAftpN,,
' VEHICLE
OWNERSHIP
Develop alternative growth scenarios; or run land use models with
change in transportation access and cost considered; . '
Integrate alternative land use/urban design supportive of , '.
alternative modes into land use models.
Re-run if shifts in land use.
* -- - TRIR, .
GENERATION'
* a J ««•£ ^
*~ - '* *~ ^V~
' '(Track) ':-
Standard analysis
EMISSIONS <"'
RATBP/EHICLE
, - MIX ~
-^(brimaryyc '.
Adjust vehicle fleet mix to reflect changes in ownership and use.
EMISSIONS
ANALYSIS.,
Adjust vehicle fleet mix to reflect changes in ownership, and use.
U. S. Environmental Protection Agency
• . 2-51
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Trip Generation: If the trip generation process forecasts only vehicle trips, an attempt should be
made to estimate the shift to non-motorized or non-automobile-driver modes as a result of higher
fuel prices, following the recommendation for advanced analysis in Section 2.3.2.3 on analytical
techniques for trip generation.
Traffic Assignment: Following mode choice analysis, areas with appropriate 'capability are
expected to perform a revised traffic assignment for input to the emissions analysis.
Emissions/Vehicle Mix: The performing agency should relate its assumptions on changes in
average mpg due to shifts in vehicle mix, the impact this would have on the effective cost per mile
used in each of the analysis steps, and whether this would be expected to result in important
differences in impacts for different market groups, particularly by income level. Analysis
requirements can be met with use. of an emission factor model that incorporates vehicle mix
impacts or, if that is not available, a narrative description that details program-induced changes in
emissions rates in response to increased fuel prices.
Suggestions for Advanced Analysis of Pump Charges
A more comprehensive and accurate analysis of the impact of pump charges could be achieved
with more extended analysis, using the following techniques (listed in descending order of
importance).
Mode Choice: Appropriate evaluation is described under advanced analysis in Section 2.3.2.5 on
analytical techniques for mode choice (i.e., mode choice models such as "nested" logit models,
with attention to differentiation among trip purposes and income levels).
Emissions/Vehicle Mix: Extended evaluation of emissions/vehicle mix effects is described as
basic or advanced analysis in Section 2.3.2.8, and includes hedonic models, research studies,
and/or special surveys.
Trip Distribution: Extended evaluation of trip distribution effects would incorporate the advanced
analysis approach described in Section 2.3.2.4; that is, a trip generation model sensitive to cost,
used to perform a revised trip distribution analysis (performed separately by trip purpose and
income strata if possible), which also reflects changes in attractiveness of destinations (and
purposes) due to fuel-price based overall increases in travel costs.
Land Use/Activity Allocations and Trip Distribution: Because fuel price increases would result in
across-the-board changes in the cost of private vehicle travel, significant increases in price could
lead to near-term shifts to closer trip destinations and longer-term shifts in residential and
employment locations to areas that require less automobile dependence. Because these are
secondary impacts, no formal analysis is required, but the performing agency would be advised to
perform such an analysis if it has the capability, or at a minimum, state its assumptions as to the
anticipated impacts this price change would have on trip distribution and land use/activity
allocations. ' ;
If significant levels of fuel prices suggest a different cost structure for regional travel, it may be
important to estimate this effect on long-term land use locational preferences. Two options are
described under advanced analysis in Section 2.3.2.1 on land use/activity allocations, that is: land
use models sensitive to cost, and growth scenarios to depict the alternate ways in which the growth
2-52 u 5; Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
patterns might adjust to an overall change in pricing. Scenarios might be developed in conjunction
with application of the land use model, and involve development and integration of alternative land
use/urban design configurations supportive of subsidized modes.
Vehicle Ownership and Trip Generation-. If shifts in land use patterns are determined in the step
above, then the vehicle ownership and trip generation steps should be re-run to reflect the
relocation of activity and its association with potentially different rates of trip making and vehicle
ownership. Wherever possjble, trip generation should be performed separately for each purpose
and income strata, to support the subsequent analysis of trip distribution and mode choice.
In general, an increase in fuel price would not be expected to have a-maj.or, impact on the overall
number of trips made by households. However, if the local planning process considers vehicle-
serviceable trips rather than person trips in trip generation, then the evaluating agency should
investigate any impact of higher fuel prices on shifts to non-motorized or non-automobile-driver
modess and to closer destinations.
2.3.3.4 Emissions Fees
Suggestions for both the basic analysis and the advanced analysis pertaining to emissions fees are
illustrated in Figure 2.2d and discussed below. (If there is a significant VMT component; then
address also the Pump Charges guidance).
Basic Analysis for Emissions Fees,
The primary^ effects of emissions fees would be on vehicle mix and emissions. If the proposed fees
imposed a significant cost on vehicle ownership, then impacts on land use and vehicle ownership
also could be important. Because emissions fees (as defined here) would be assessed on a periodic
basis, they would not be expected to affect daily travel choices.
Vehicle Mix/Emissions Analysis-. Fees levied on vehicles in relation to their age and/or emissions
production (emissions rate, perhaps cpupled with VMT) would have the primary effect of
encouraging replacement of higher emitting/older vehicles and/or encouraging higher use of lower-
emitting vehicles in multi-vehicle households. ,
Evaluation required in submissions to EPA includes the basic or advanced analyses techniques
described in Section 2.3.2.8,' above; that is, hedonic models, research studies, and/or special
surveys (these methods are further detailed in Chapter 3, Section 3.3.3). The agency also should
make an effort, to indicate emissions fee impacts on relative usage patterns of particular vehicles,
accompanied by a statement on assumptions relating impacts to market segment (income). For
fees based on actual (measured) emissions rates, evaluation should include an estimate .of the
effects on improved maintenance. , .
U. S. Environmental Protection Agency
"
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Figure 2.2d Elements of Analysis for: Emissions Fees
ELEMENTS of Basic Analysis ANALYSIS STEP „«-/»«»«.=•,•«.««««.« --,
(Type of Impact) RECOWIMENDATI°NS for Advanced Analysis
If measure affects vehicle ownership/ use,
submit statement that assesses effects on
tocational patterns.
LAND USE
ALLOCATION
(Secondary)
, If price significantly affects vehicle ownership or VMT, then
consider development of alternate growth scenarios or application
of land use models to reflect change in costs on number of
vehicles.
Assess if measure affects number of vehicles
owned, following procedures for vehicle mix
analysis.
VEHICLE
OWNERSHIP
(Primary)
Re-run if shifts in land use;
Possible supplemental analysis to estimate change in vehicle mix
and possibly usage patterns favoring cleaner vehicles.
Investigate impacts if land use or vehicle
ownership affected.
TRIP
'(' GENERATION
if ' (Trace)
Re-run if shifts in land use or vehicle ownership;
If trip generation does not include non-motorized, non-automobile
driver modes, adjust to assess impact;
Run separately by purpose and income.
Investigate impacts if land use or vehicle
ownership affected, or if VMT fee significant;
Otherwise, no analysis.
TRIP
DISTRIBUTION
(Trace)
Run separately by purpose and income;
If significant VMT cost effect, incorporate cost modeling.
Investigate impacts if land use or vehicle
ownership affected, or if VMT fee significant;
Otherwise, no analysis.
MODE CHOICE
(Trace)
Use separate models for each purpose;
Run separately by purpose and income;
Supplemental analysis to assess any effects of shift toward higher
mpg vehicles;
Assess demand for non-motorized modes, or impact of
periodically-assessed charge on daily travel.
None
TIME OF DAY/
PEAKING
(Ndne)
None
Investigate impacts if land use, vehicle
ownership, distribution or mode choice
changes;
Otherwise, no analysis.
ASSIGNMENT
(None)
Standard analysis
Indicate assumptions to account for shift in
vehicle mix to cleaner vehicles.
EMISSIONS '
RATE/ VEHICLE
MIX
(Primary)
Adjust vehicle fleet mix to reflect change in ownership and use;
Adjust emissions rates for better vehicle maintenance, if a feature
of measure.
Standard analysis if have emissions model;
If no model, apply emissions factors to change
in vehicle trips and VMT.
EMISSIONS
ANALYSIS
Adjust vehicle fleet mix to reflect change in ownership and use;
Adjust emissions rates for better vehicle maintenance, if a feature
of measure.
2-54
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
• . • .,;f ; to Reduce Transportation Emissions
Areas with emissions factor models are expected to use these to assess the result of the proposed
strategy on emissions and to .incorporate the effect of emission fees on vehicle mix and/or
improvements in'maintenance.
In the absence of an emissions factor model, agencies may apply emissions factor relationships to
changes in vehicle trips and VMT, with appropriate supporting documentation of their analysis and
assumptions for emissions factors that reflect anticipated shifts in fleet mix and maintenance
levels. , . . , , '
Vehicle Ownership: If proposed emissions fees lead to significant cost increases in automobile
ownership, household ownership rates may be reduced, particularly ownership of high-emitting
vehicles. Evaluation would follow that described under vehicle mix/emissions analysis above: .
Land Use/Activity Allocations: If vehicle ownership is affected by a sizable .emissions-based
increase in registration fees, or if the VMT component of emissions fees is significant, households
could shift to locations where fewer vehicles and/or less dependence is required. This effect
should be investigated. '
Suggestions for Advanced Analysis of Emissions Fees
Listed in declining .order of importance, the following analyses could help generate a more
comprehensive and accurate evaluation of emissions fee impacts.
>* . ;""'-•' '
Vehicle Mix and Emissions Analysis: Because the primary impact of emissions fees is expected
to be on vehicle ownership patterns, enhanced evaluation of emissions/vehicle mix effects could be
achieved through more sophisticated application of the same techniques described above under
basic analysis (i.e., hedonic models and/or secondary analyses, such as stated preference surveys to
explore vehicle ownership and usage). In addition, the agency could use the above techniques to
investigate whether the VMT aspect of the fee produces any change in daily travel behavior, how
large a fee would have to be applied, and how frequently it would have to be assessed.
The most accurate emissions analysis will probably use an accepted emissions factor model
adjusted to account for these changes, rather .than application of emissions factors to net trip and
VMT changes.
Vehicle Ownership and Land Use: Vehicle ownership would be affected by significantly higher
emissions-based registration fees. Agencies proposing such fees should estimate vehicle
ownership patterns, by income group if possible. In turn, revised ownership rates should then be
incorporated in a revised land use analysis that reflects the impact of fewer vehicles and higher
costs, using models, if available, or best professional judgment.
Mode Choice, Trip Distribution, Trip Generation: Generally, emissions fees collected through
infrequent periodic assessments would not be expected to lead to major, adjustments in daily travel
behavior. If the VMT component of the fee is either significant or controlling, however,
investigate its effects on mode choice, trip distribution, and trip generation by consulting research
studies or performing a supplemental analysis of relative impact on daily behavior of annual (or
periodic) fees. ."'.'••-
U. S. Environmental Protection Agency , ,2.55
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions . '
Any changes in vehicle ownership and household location also would have an indirect effect on
trip generation, distribution, and mode choice, and should be assessed. Trip generation analysis
should reflect total person trips, and allow for shifts to non automobile-driven and non-motorized
modes.
2.3.3,5 Roadway Pricing
Suggestions for both the required.and advanced analysis pertaining to roadway pricing are
illustrated in Figure 2.2e and discussed below.
Basic Analysis for Roadway Pricing
Because it is a fairly complex measure in terms of travel behavior impacts, there are more primary
effects associated with roadway pricing than any of the other measures. Roadway pricing would
be expected to have a primary impact in relation to mode choice, trip distribution, and traffic
assignment, and, if peak period or congestion pricing is instituted, time of day choices as well.
Trip Distribution: Travelers would be expected to respond to the pricing of specific transportation
facilities by favoring comparable destinations which can be reached without payment of a toll
(where such a substitution can occur). There are several options for approaching this analysis.
The more rigorous approaches are described in Section 2.3.2.4 as basic or advanced analysis
techniques for trip generation, which include cost-sensitive trip generation models and
"generalized cost" procedures.
Sites that do not have formal trip distribution modeling capabilities should formulate a statement
that reflects a careful assessment of the anticipated impacts of the road pricing scheme on shifts in
travel patterns.
Traffic Assignment (Route Choice): Traffic assignment becomes particularly important in the
case of road pricing since travelers can choose between tolled routes that may have a higher level
of service, and toll-free routes that may be more congested.
This choice is not directly handled by most traffic assignment models. The options for addressing
anticipated impacts, described in Section 2.3,2.7 on basic or advanced analysis techniques for
traffic assignment, include assignment routines for toll and free paths, or use of travel time
equivalency estimates (see also Section 3.3!8.5).
If an area cannot perform traffic assignment analysis, or if the tolled facility can be easily avoided,
then it may not be possible to consider the emissions impacts of roadway pricing because the
impacts would be too uncertain.
Mode Choice; Mode choice is still an important factor in considering road pricing. Analytical
options include those detailed above in Section 2.3.2.5 as basic or advanced analysis techniques for
mode choice (i.e., mode choice models, "quick response" methods, or elasticity methods).
2-56 • . U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Figure 2.2e Elements of Analysis for: Roadway Pricing
ELEMENTS of Basic Analysis ANALYSIS STEP RECOMMENDATIONS for Advanced Analysis
============^^^^^^^ (Type of Impact)
, Formal analysis not required;
Statement of possible effects due to pricing of
select facilities.
^ALLOCATION,
Develop alternative growth scenarios or run land use models, to
reflect cost related to priced network;
, Integrate alternative land use/urban design supportive of
alternative modes. : -
None
OWNERSHIP,''
Re-run if shifts in. land use-
None
Jp'
Analyze if have trip distribution model;
If no model, provide estimates of expected
impacts;
Account for non-motorized trips if vehicle trips
only.
DISTRIBUTION;
;/~ (Primary)',
Re-run if shifts in land use;
Run separately by purpose and income;
Consider shifts to non-motorized alternatives, if area-wide pricing
and vehicle^only trip generation.
Run separately by purpose and income;
Explicitly model to reflect change in cost and travel time changes.
Analyze through mode choice model, if
available;
"Elasticity factor" or 'quick response" analysis
if no model;
Document sources and assumptions.
MODE CHOICE
, (Primary) •
"*_ "'<13>v
Use separate models for each purpose; .
Run separately by purpose and income;
Directly represent pjice for trip (rather than travel time equivalent);
Possible supplemental analysis to ascertain response to
electronic/periodic charge assessment.
Must analyze if apply time of day pricing;
Document sources and assumptions.
PEAKING ,T,.
•• (Primary'lt ' r
Acquire/ develop model to simulate response to time of day
pricing;
Possible supplemental analysis to gauge time of day shifts.
Must analyze if pricing applies to selected
facilities; •
Can use elasticity factor methods; Document
sources and assumptions. r
** TRAFFIC
Develop and use toll diversion model to reflect toll/free path
options; ,
Possible side analysis to gauge route choice shifts.
None
.EMISSIONS
WE/VEHICLE,,
: :MIX,'(Norie) :
None
Standard analysis if have emissions model;
If no model, apply emissions factors to ihe
change in vehicle trips and VMT.
EMISSIONS^
ANALYSIS -
Standard analysis;
Possibly adjust emissions rates for major speed changes on
facilities. .
U. S. Environmental Protection Agency
2-57
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions •
Most likely, an area considering road pricing would have a reasonably sophisticated modeling
system, complete with a mode choice model. Application of elasticity techniques would only be
expected to be valid where simple pricing-schemes are being considered (i.e., a bridge or tunnel
with few/no parallel routes).
Important issues that should be addressed in the analysis include:
• Will the toll be paid to toll booth operators, to exact change machines, or
electronically through advanced vehicle identification (AVI) transponder
technology? What are the queuing and delay impacts of the chosen method?
• Are HOV users exempt from the toll, or charged discounted rates?'
• Is the dollar value of the policy being tested consistent with the year in which the
. elasticity was developed?
• What other factors are likely to be introduced or changed during" the evaluation
period that could affect the implementation or traveler response to the proposed
application?
Time of Day: Peak-period "congestion pricing" versions of road pricing would be expected to
affect time of day choices as well as route choice and, if such strategies are being considered, the
agency should attempt to estimate the impact of the fee on peak versus off-peak travel demand by
trip purpose and income group.
Options for conducting this analysis are described above in Section 2.3.2.6, basic or advanced
analysis techniques for time of day, and include: time of day choice models, references to other
research, and survey analysis. Under extenuating circumstances, the area may be allowed to use
assumptions to suggest the degree of change, but it is expected to provide documentation as to the
assumptions and their degree of impact on the result.
X
Land Use/Activity Allocations: Implementation of toll pricing on selective transportation facilities
would be expected to cause a shift in locational trends, which could reduce use of these facilities.
Because these are secondary impacts, no formal analysis is required. Nonetheless, the performing
agency is expected to state its assumptions (and show the concurrence of decision-makers)
regarding the anticipated impacts that the system of road pricing would have on land use/activity
allocations. This is particularly important in the context of partial-network pricing.
2-58 U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Suggestions for Advanced Analysis of Roadwiy Pricing
In descending order of importance, the following analyses would contribute to a more
comprehensive arid accurate evaluation of roadway pricing impacts.
Traffic Assignment (Route Choice): The best analysis would use a procedure including both toll
and free paths as options to travelers, taking into consideration the fact that traveler choice will be
sensitive to type of trip and traveler income level (see Section 2.3.2.7 for a more 'extended
discussion, and Section 3.3.8 for a more technical presentation).
Time of Day: Time of day choice flexibility is particularly important in assessing the impacts of
congestion pricing schemes. Shifts in time of day would be expected to vary significantly by trip
purpose and income level. Time of day transportation modeling is a relatively .uncommon
practice; options described above in Section 2,3.2.6 include time of day choice models, references
to other research, and surveys (Section 3.3.7 describes appropriate models).
Trip Distribution: Shifts to alternative destinations induced by road pricing could be important,
particularly with partial-network pricing and numerous travel or route alternatives; impacts are
likely to vary by trip type and traveler income levels. Include distribution effects of reduced
congestion.
Application of a trip distribution model which is sensitive to cost through a generalized cost
relationship to individual trip purposes and income strata would be essential to an effective
analysis. , .
Mode Choice: Mode choice analysis would be'most accurate if it were accompanied by the
described advanced analyses of trip distribution, route choice, and time of day alternatives. It also
would be most accurate if performed with separate models of each trip purpose and evaluated
separately for each income strata.
Land Use/Activity Allocations: Road prices' would be expected to make some areas more
attractive than others in terms of future locational preferences. Options for approximating andv
accounting for these trends include use of land use models that incorporate sensitivity to cost,
and/or the development of alternative growth scenarios. It would be particularly meaningful to
develop land use and urban design scenarios which would complement travel alternatives to those
sites with priced facilities/access. *
Vehicle Ownership and Trip Generation: If alternative land use patterns are anticipated, the
vehicle ownership and trip generation steps should be re-run to reflect the relocation of activity and
its association with potentially different rates of trip making and vehicle ownership. When
possible, trip generation analysis should be performed separately for each purpose and income
strata, to support the subsequent analysis of trip distribution and mode choice.
U. S. Environmental Protection Agency 2.59
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
• 2.4 Packaging Measures for Best Effect
The preceding discussion of market-based measures and guidelines for their evaluation necessarily
provide a somewhat one-dimensional view of the nature and impact of these strategies. In reality,
such measures would seldom be deployed in isolation, as described here. Rather, they would be
deployed as part of a carefully considered "package" of different market measures and other
actions to complement and support the market-based measure(s). Market-based measures serve as
signals and incentives to change; other actions provide the appropriate setting and options
necessary to complete the process.
There are two major benefits from packaging market-based measures into balanced programs:
• Balanced programs offer synergy that will generally produce a. greater impact than a
market-based measure acting alone; and
• Balanced programs that offer tangible and attractive choices are more likely to foster
public acceptance of the proposed pricing measure component of the program.
This guidance "manual cannot offer hard and fast rules detailing the most effective packaging
strategies to accompany pricing initiatives, given the wide variation in settings, needs, and
circumstances. In general, however, pricing measures are expected to work best if the following
general principles are applied:
1. Match actions to consumer responses— Anticipate the types of responses consumers
are likely to make in response to a particular pricing measure. Analysts can use the
guidance presented in the preceding sections, coupled with the performing agency's
knowledge and experience. Analysts also should evaluate reciprocal actions that
could be mated with these responses.
2. Identify realistic alternatives— Where a measure is likely to raise interest in an
alternative mode of travel, identify alternatives that will be attractive and appropriate
in that setting. For example, consider transit service improvements if a high density
corridor or downtown area is involved. If the impact area is more suburban, transit
may still have a role, but actions to stimulate carpooling or vanpooling may be more
suitable.
3. Consider non-motorized modes— Broaden the analysis to include non-motorized
alternatives such as walking and biking, which are often overlooked when options
are considered. A fair portion of shorter trips may be amenable to walking or biking
under the appropriate circumstances; shifting these trips from private vehicles to
non-motorized modes could eliminate substantial emissions associated with vehicle
" cold-starts." More specifically, consider enhancing circumstances associated with
bike and pedestrian modes such as: improving sidewalks; and pedestrian access to
buildings; providing frequent, safe pedestrian crossings; alleviating
vehicle/pedestrian design conflicts; separating vehicles from pedestrians and
reducing speeds; and providing bicycle lanes, paths, suitable lockup/storage
facilities, and lockers or change facilities at destinations. Also consider approaches
that encourage mixed use land development, such as reduced taxes for development
that reduces transportation costs.
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
"..•-" " to Reduce Transportation Emissions
4. Consider positive incentives for desirable alternatives—Identify and evaluate
incentives for users of the more environmentally efficient modes. Incentives such as
fare subsidies to transit or vanpool users, parking subsidies for HOVs, and equipment
rebates can provide a powerful economic complement to market-based measures that
primarily increase the cost of travel for driving or use of high-emitting vehicles.
5. Consider barriers to counter-productive traveler choices—Articulate strategies to
limit 'or direct the choices of travelers who might select "counter-productive"
alternatives in response to the market-based measure. Depending on who they are
and the purpose of their travel, many travelers have flexibility in their choice of
route, destination, and time of day. When faced with market-based incentives, these
travelers would be expected to consider and use that flexibility in their own best
interest. Analysts should anticipate these responses and, unless these secondary
responses support emissions reduction objectives, should propose complementary
actions that reinforce desired behavior patterns. Mechanisms include: limitations on
off-street parking coupled with site parking charges; and limitations on use of
secondary/neighborhood roads as alternatives to a priced facility.
6. Explore outreach and indirect support—Consider complementary actions in
addition to financial incentives and physical alternatives that support travelers faced
with changes in their travel context. For example, information outreach, education,
and marketing can ease barriers to use of alternative travel modes They can also
explain and demonstrate the problems caused by driving and the virtues of walking,
bicycling, and telecommunications. Effective matching programs can put people
into acceptable ridesharing arrangements. Employers can play a very valuable role
in supporting employee travel choices, including: guaranteed rides home in event of
late work nights or family emergencies; assistance with dependent care; preferential
or covered parking for ridesharing employees; and assistance with vanpool formation
arid operation. Employers also can offer opportunities to work at .home through
telecommuting arrangements, or introduce modified work schedules to reduce the
number of physical trips to worksites. ,
7. Promote land use planning that supports alternative travel—Recognize the
importance of an environment that supports the use of alternative travel patterns. If
persons are expected to travel to targeted destinations without their personal vehicles,
they must be confident that they can achieve their objectives at that location without
a car at their disposal This calls for innovation in urban design that allows people to
circulate and accomplish activities pleasantly through walking, biking, transit or
through shuttle services. This also means designing those centers so that the mix .of
uses is varied, complementary and attractive. Street space should be managed in
" these areas to provide pedestrians with safety from motor vehicles, .freedom of
movement, attractive visual surroundings, and comfortable settings for conversation,
rest arid shopping. , ' •
8. Use program revenues to support transportation and emissions goals— Clearly and
explicitly link the revenues that are collected through application of a pricing action
with the Use of those revenues. If the public perceives that the revenues extracted
from them are used for the direct enrichment of their environments, rather than
- diverted to other uses, they should show stronger support for those programs since
they are getting something back from their contribution. Revenues from pricing
U.S. Environmental Protection Agency • 2-61
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
programs are often a convenient and adequate source to finance a variety of
supporting strategies.
For more information on techniques for effective packaging, the reader is referred to other
information sources, such as Implementing Effective Travel Demand Management Actions, which
is recommended as a comprehensive source for packaging and impact information and contains
considerable information on a wide variety of strategies and packaging tips, as well as a significant
number of references to other sources on specific related topics.14
COMSIS Corporation et al., Implementing Effective Travel Demand Management Measures: Inventory of
Measures and Synthesis of Experience, prepared for the Federal Highway Administration and the Federal
Transit Administration, U. S. Department of Transportation, Washington, D.C.,1 September 1993 (DOT-T-
J/*T~l/^ ),
2-62
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions,
3.0 Guidelines for Analysis of
Market-Based Measures
3.1 Introduction
This chapter provides guidelines and suggestions for those responsible for performing the actual
evaluation of market-based measures for emissions credit. Whereas the preceding chapter was
designed for managers and those without a technical background,, this chapter was developed for
those versed in the basic principles of transportation modeling and analysis. Expanding on topics
discussed in Chapter 2, this chapter employs terminology and concepts that'assume greater
famijiafity with transportation and emissions modeling tools and procedures. The intent is to
provide fairly specific language as to EPA's expectations when reviewing credit submissions.
The chapter is divided into two main sections. The first section, Section 3.2, presents guidance on
each of the five basic groups of market-based measures (parking pricing, modal subsidies, pump
charges, emissions fees, and road pricing). The second section of the chapter, Section 3.3, is a
reference compendium on methods and procedures suitable for the types of analyses suggested in
Section 3.2. The reference compendium is organized according to steps in the analysis process.
The methodological guidelines in Section 3.2 refer extensively to the methodological discussions
of Section 3.3, many of which also list external references for further details.
I , '
This document has been prepared more as guidance than traditional requirements. EPA recognizes
that that no single approach can work for all sites and that areas vary widely with respect to
analytic tools, data, and staff capabilities. It is EPA's intention to suggest analytic approaches for
best practicable evaluation of the proposed market-based measures.
EPA will use this guidance when evaluating an area's submission of a measure for credit.
However, EPA also desires to work cooperatively with any area that is seriously considering
market-based measures and will support reasonable .best efforts to introduce such measures. EPA
expects those areas with more sophisticated analytical tools to use them to the best possible
advantage in performing an accurate assessment. At the same time, areas with limited analytic
capability should not be discouraged from investigating market-based measures, and are
encouraged to use this interest as a basis for upgrading their analytical tools for future
investigations.
This chapter was designed to give direction.to technical staff and, as such, goes beyond basic
description of analytical approaches. Nonetheless, some practitioners may want additional
information, such as clear-cut examples or instructions on how to make particular changes to their
U. S. Environmental Protection Agency 3-1
-------
Technical Methods for Analyzing"Phcing Measures
to Reduce Transportation Emissions
models or how to apply their models in specific circumstances. Appendices A and B address these
needs by providing detailed technical procedures on model applications.
Appendix A describes a "Case Study Model," which involves practical modification of a
conventional four-step model system to improve its sensitivity to pricing. In addition to a
description of the modification itself, the appendix demonstrates model applications to test market-
based measures. Appendix B describes a somewhat different approach that uses household data
and sample enumeration methods to obtain increased flexibility and potentially greater accuracy in
evaluating pricing measures.
,!' ' ' '
Performing sites are encouraged to contact their respective EPA regional offices for more
information. In the course of undertaking an evaluation of market-based measures, agencies also
are advised to:
• Establish and maintain a dialogue with EPA and other organizations with approval
authority;
• Confer with appropriate agencies to coordinate assumptions and data;
• Consult with state and local air agencies and EPA on planned analytic approaches to
obtain agreement on the proposed approach; and
• Consult with transportation agencies, environmental and community groups, and
possibly consultants or academics to make sure that analyses and assumptions are
reasonable and accepted.
3-2 , U. S. Environmental Protection Agency
-------
Technical Methocfs for Analyzing Pricing Measures
to Reduce Transportation Emissions
3.2 Evaluation Guidance for Individual Pricing Measures
This section provides step-by-step, guidance for evaluation of each of the following five types of
market-based measures:
• Parking Pricing (Section 3.2.1);
• Modal Subsidies (Section 3.2.2);
• Pump Charges (Section 3.2.3); , . '
• Emissions Fees (Section 3.2.4); and
' . t
• Road Pricing (Section 3.2.5).
The discussion of each follows a common four-part format. Each of the five segments begins with
an introductory table that summarizes the relationship between the given measure and each of the
nine steps in the analysis hierarchy with regard to the following: , •
• Form of impact expected at that step— that is, how the market-based measure will
work to affect transportation factors analyzed in that step of analysis;
of impact expected— -primary, secondary, trace, none and derivative (i.e., it
must be addressed as a result of impacts in other steps); and '
. • Impact indicator— that is, measures of performance that should be employed in the
evaluation of each step.
Discussion begins with a statement of Anticipated Impacts, or those effects that are expected to be
important in evaluating the specified measure. This is followed by a short summary statement of
the Basic Analysis. The third component is Guidelines for Analysis, which presents step-by-step
procedures or considerations that EPA would expect to see applied in evaluating the given
measure. Analysis recommendations progress from "basic analysis5' to "advanced analysis,"
which is more likely to result in the granting of maximum emissions credit.
U. S. Environmental Protection Agency
3-3
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
3.2.1 Parking Pricing
3.2.1.1 Anticipated Impacts
The primary impact of parking pricing would be on Mode Choice. Parking prices raise the cost of
vehicle trips and thus increase the attractiveness of alternative.modes, including transit, fidesharing
and non-motorized options. They may also affect trip distribution, shifting destination to those
more accessible without private vehicles. The degree to which other options become more
attractive depends very much on-
Table 3.1 Impacts of Parking Pricing
attendant factors, such as:
• Importance of reaching the
particular destination;
• Amenity/design characteristics of
that location; and
• Service/quality of the options that
serve that location.
Parking pricing collected as a daily fee
may have a greater effect than other
pricing measures, since it is directly
associated with the cost of the
particular trip. Parking pricing is
expected to have a greater effect on
shorter trips, for which it represents a
higher proportion of overall trip costs.
This has important implications for
affecting cold starts and micro-level
traffic problems.
Parking pricing policies also raise
practical concerns about distributional
impacts. It often is difficult in practice
to:
Analysis
Step
Land Use/
Activity
Allocation
Vehicle
Ownership
Vehicle Mix
Trip
Generation
Trip
Distribution
Mode Choica
Time of Day
Traffic
Assignment
Emissions
" Impact Form
Greater attractiveness
of areas with lower
prices (uneven
application) or with
better alternatives &
supportive land
use/density
None ,
None
Possible effect on
attractions
Diversion to
destinations with
lower prices or better
alternatives
Decreases private
vehicle use
Shift time of day
Shift in trips and VMT
as result of mode
choice & distribution
Reduced vehicle trips
and VMT
lm£act>v '
TsrwT
Secondary
None
None
Primary on
work-trips,
trace overall
Secondary
Primary
1 */
Secondary
(if time of
day differs)
Derivative
effect
Derivative
effect
impact indicator
<;•- '=-
Population and/or
employment, by
• zone
Vehicles per
household per zone
Percent of vehicles
by VMT
Number of person
trips to/from each
zone
0/D patterns
(average trip length)
Parson trips by ^ ,
mode; % transit; ''•*
% walk/bike;
# vehicle trips;
average vehicle L; „
occupancy-
Trips by time period
Link volumes; VMT, '
Tons or kg. of VOC,
NOX, CO, PM
• Apply parking pricing uniformly
with regard to geography (e.g., suburb versus city), type of site (site size and accessibility, on-
street versus lots or spillover options, etc.), and
• Eliminate employer or merchant discounting.
Travelers with flexibility in choice of destination may shift away from locations with increased parking
costs. This could apply to many non-work trips, and also to work and other fixed- location destinations
in the longer run. In turn, this could have an effect on land use as the comparative
advantages/disadvantages of various locations shift in response to pricing levels and uniformity, quality
of alternatives, amenities, etc. ,
If parking pricing is applied at different rates to different travel markets (e.g., commuters versus non-
commuters, or peak versus off-peak), then travelers with time flexibility, can be expected to shift their
chosen time of travel.
3-4
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing, Pricing Measures
- to Reduce Transportation Emissions
\
3.2.1.2 Basic Analysis for Parking Pricing
Aside from the standard analysis of emissions, parking pricing analysis should include a careful
Assessment of the impacts on mode choice (person trips by mode) and the resultant effects of mode
choice changes on overall vehicle trips, .VMT, and emissions. If parking rates are to be
implemented non-uniformly across the region, the effect on-shifts in trip distribution (particularly
for non-work travel), land use, and time of day (if applicable) should be assessed, either through
formal analysis/assumptions or through a discussion that describes the anticipated impacts on
travel and emissions.
• . / • • .•'"..'
3.2.1.3 Guidelines for Analysis
1. Parking Pricing and Land Use/Activity Allocations
Parking pricing can affect long-run locational patterns by decreasing vehicle trips, particularly to
locations affected by the policy. EPA does not require a formal analysis of parking pricing's
effects on land use, although it would be more important to consider land use impacts if pricing is
not uniform. If parking prices are not applied uniformly across the region, businesses and
employers might shift to areas with lower charges or areas with better travel alternatives (e.g.,
transit) and supporting density. Households are expected to make long-term shifts in residence to
areas that support lower cost travel patterns—that is, access to work/non-work destinations that
have parking with low or no fees or that offer effective travel alternatives.
If an area's planning process includes a land use allocation model that incorporates transportation
cost through a measure such as composite impedance* the area may wish to consider using this
model to investigate parking price impacts on long-term development trends. However, these
results should be interpreted with caution given the limited accuracy of such models and the
existence of other factors that can affect location trends. Section 3.3.1.2 provides more
information on the nature and use of land use models.
Agencies may wish to consider formulating alternative growth Scenarios based on various
anticipated effects that development shifts might have on regional travel and emissions. This
approach is particularly relevant for agencies that do not use models for land use analysis. This
option, detailed in Section 3.3.1.1., involves defining alternative land use/urban design schemes
that support the alternative modes and/or travel patterns encouraged by the proposed parking
pricing policy. '
If the proposed parking policy entails major disparities in regional coverage or level, and the
agency cannot perform a formal analysis of its impacts, the agency should provide a statement that
reflects local decisionmakers' appraisal of (1) the long-term consequences such a policy would be
expected to have on regional development, travel trends, and emissions; and" (2) what impact, if
any, this will have on other land use and transportation investment analyses. ' "
Composite impedance is a measure that combines the time and cost of highway and transit (and perhaps
other) modes into one statistic; see Section 3.3.5.1. • ••'•..
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing'Pricing'Measures'
to Reduce Transportation Emissions
2. Parking Pricing and Vehicle Ownership '
Because parking pricing is expected to have little or no direct impact on vehicle ownership, no
formal analysis is required. Indirect impacts on vehicle ownership could occur as a result of
household location shifts described in (1) above; in this case, advanced analysis would include re-
running the vehicle ownership analysis.
3. Parking Pricing and Vehicle Mix
Parking pricing is not expected to have a direct impact on ownership vehicle mix, thus no formal
analysis is required. Given that parking pricing is likely to affect short trips more than long trips,
however, parking pricing could affect vehicle usage patterns; this should be taken into
consideration in determining vehicle distribution used for the emissions estimation step.
4. Parking Pricing and Trip Generation
It is unlikely that parking pricing will have a measurable impact on the overall number of person
trips made by households as forecast in trip generation analyses, thus formal analysis is not
required.
Household vehicle trip rates could be affected by parking pricing in the following ways:
• Higher numbers of workers involved in telecommuting or compressed work weeks;
• Increased errand chaining to reduce the number of stops at priced destinations;
and/or
• More use of telecommunications for non-work activities.
While these responses may affect the pattern of household tripmaking, it is unclear whether they
affect the overall number of trips or simply trip characteristics, such as destination, trip length,
mode, or time of day. Because there are no formal existing methods for dealing with these
responses, however, no formal analysis is required (see Section 3.3.4.5).
There is some concern regarding what types of trips are depicted by the trip generation process
used for analysis—total person trips or only those trips that are vehicle serviceable. Because the
latter approach excludes walking and bicycle trips, important shifts in household trips from
vehicular to non-motorize^ modes are likely to be underestimated; Section 3.3.4.4 discusses this
aspect in greater detail.
••••'' • t
Areas that do not currently forecast total person trips are encouraged to upgrade to such a
framework as a future goal; for the present, these areas should try to consider these effects in their
analysis as best they can. Section 3.3.6.6 suggests ways in which areas currently unable to project
non-motorized modes can account for these effects later in the mode choice step.
Smaller urban areas without mode choice model.s may have an even more restricted definition of
trip generation, considering only, vehicle trips (i.e., only those person trips that are made in private
vehicles, excluding transit). Trip tables limited to private vehicles inhibit analysis of pricing
3-6 . U.. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
'. . to Reduce Transportation Emissions
strategies, although options for overcoming this constraint are discussed in the mode choice
segment below.
Parking pricing may have an effect on trip attraction rates, although there are no known models
that are directly sensitive to parking pricing in this way. For example, suburban shopping sites
tend to have higher vehicle trip attraction rates than downtown areas, although this may simply
derive from a failure to account for transit and non-motorized trip's. Areas are advised to be
sensitive to these differences in their trip generation analyses.
\ . ' - - - , ' ' .
Parking pricing is not expected to have a major (or predictable) impact on freight/commercial
traffic, except in the event that parking.supply itself also..becomes more restricted, for example,
due to limitations on curb parking. It also is unlikely that parking pricing would have a tangible
effect on through (or "external-external") traffic—that is, trips originating from and, destined for
locations outside the area of the pricing scheme.
5. Parking Pricing and Trip Distribution
Parking pricing is expected to have only a secondary effect on trip distribution; thus, no formal
analysis is required. If parking pricing is applied differentially, however, a statement of possible
effects should be included. •. : ,
Trip distribution analysis can be a particularly important element in assessing the travel/ emissions
response to parking pricing in the following ways:
• Trip purpose: Travelers who have the flexibility to shift to alternative destinations
for some trip purposes, such as shopping and social/recreation, would be expected to
consider and perhaps shift to destinations which serve more needs and/or are
accessible without a private vehicle and if parking charges increase in only some
locations to those with free or lower parking charges (unless there are offsetting
amenity or utility circumstances). Travelers are less likely to shift destinations on
trips with more rigidly determined destinations, "such as work, school o'r personal
business, at least in the short run.
• Trip length: Parking charges will have a higher proportionate effect on shorter trips,
which also are strong candidates for transit or non-motorized alternatives (see mode
choice discussion below).
Advanced analysis for areas with appropriate modeling capability includes evaluation of the effect
of .parking pricing on trip distribution, both for non-uniform and uniform pricing applications.
This is best done with a gravity model that incorporates cost in its measure of separation through a
feature such as composite impedance (see Section 3.3.5.1). If parking price is added to overall cost
of travel to the respective destination, a different pattern of destination choices in the trip table can
be expected to result. If the separation measure in the gravity model does not directly incorporate
cost, then it may be reasonable to transform the parking cost into a travel time equivalent (see
Section 3.3.5.3). ' . ,
This analysis should be done independently for each trip purpose and income strata^ or done with
an equivalent income-sensitivity procedure (see Sections 3.3.6.3 and 3.3.6.4). If parking pricing
varies by time of day, trip distribution analysis should be performed separately for each period.
U. S. Environmental Protection Agency • 3.7.
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Users are cautioned that in typical gravity models, which are doubly constrained, the model tries to
allocate (through iterations) the same number of trip attractions to each zone that the trip
generation model has estimated would be attracted to that zone. If the relative attractiveness of a
zone decreases because of an increased parking price, the gravity model will still try to force a
certain number of attractions to go to that zone. This could cause unintended distortions in trip
patterns. That problem could be minimized by including some type of price-based variable (like
composite impedance) in the attraction model. Another option is to switch to a logit destination
choice model to allocate attractions, as is done in Portland, Oregon or San Francisco, for example
(see Section 3.3.5.4). »
Areas without trip distribution modeling procedures will have difficulty accounting for the impacts
that would result from differential application of parking pricing. In this case, there are two
options:
1. Assume only applications of parking pricing that would affect all areas at the same
level, and avoid applications to non-work travel; or
2. If the above option is not realistic, then clearly state assumptions on how the
proposed implementation is expected to affect the pattern of regional trip origins
and destinations and how this is reflected in the estimated emissions reductions.
6. Parking Pricing and Mode Choice •
Changes in choice of mode is the principal impact expected from parking pricing. All areas
considering parking pricing are expected to furnish estimates of:
• Probable impact on number of person trips by mode; and
• Change in number of vehicle trips and average vehicle occupancy.
Areas -with Mode Choice Models
Areas with mode choice models in their planning systems are expected to (1) use them for this
analysis, or (2) explain to EPA their reasons for not doing so. In the absence of appropriate models,
agencies should use " elasticity factor" or " quick response" analysis. Advanced analysis includes
disaggregating analysis by purpose and income, applying separate coefficients for parking costs,
and estimating shifts to non-motorized modes and non-travel.
EPA offers the following guidelines for the best possible mode choice analysis results:
• Type of Model: Multinomial logit (or similar probability-based) models are
expected to provide the most realistic estimates of impact; models with a " nested"
choice structure are among the most advanced of these (see Section 3.3.6.5 for more
information). An ability to represent different automobile occupancy levels as
"modes" increases accuracy in estimating the response to parking charges,
particularly if special exceptions in parking fees are made for HOVs. An ability to
account for non-motorized modes and multi-modal (e.g. park and ride), options will
increase realism of modeling, subject to the capabilities and information furnished by
the overall model system.
3-8 U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
. • -to Reduce Transportation Emissions'
• Parking Cost Coefficient: Some models carry a separate coefficient value for
- parking cost, Which suggests that parking cost carries more weight than routine
operating cost (typically by a factor of 1.5 to 2.0). Agencies may wish to incorporate
* • such a. coefficient in their models (see Section 3.3.6.1), .
• Non-Work Mode Choice Models: There should be a mode choice model for each
separate trip purpose. Some areas model only work trip mode choices and estimate
non-work mode choice through a "factoring" process; this is not likely to yield a
credible result for non-work trips (see Section 3.3.6.4).
• ' Income Stratification: Mode choice analysis should be performed separately for
different income levels (strata), both to allow for different cost sensitivity and to later
account for distributional impacts. An acceptable alternative approach would be to
: base the analysis on mode cost prices divided by traveler income (see Section
3.3.6.3):
• Program versus Effective Rates: Agencies should not take it for granted that
travelers will bear the full weight of program-based increases in parking prices.
Effective rates may be altered by .employer subsidies or other discounts. Unless the
policy explicitly assumes full rate pass-through, some assumption should be made
regarding effective rates paid per zone. It also is desirable'to develop models that
,' . consider the distribution of parking costs by site, rather than using zonal averages.
The STEP approach discussed in Appendix B offers such a capability. •
• Cost Basis: Mode choice modeling is typically based on the comparison of costs for
a one-way trip. Thus, it is generally appropriate to apply half of the daily parking to
each one-way work trip. For non-work trips, where rates may be hourly, or for trip
chains with multiple destinations, the relevant price per trip may be less clear and
analysis will require explicit assumptions.
• Adjustment for Inflation: The,dollar value implied in the suggested parking price
must be consistent with the base year for which the^ model coefficient was developed.
Generally, this requires deflating the tested parking price with a time-inflation factor
that makes parking costs compatible with'the coefficient in real terms (see Section
3.3.6.2). . :
" • HOV Special Rates: Few, if any, mode choice models permit testing of different
parking prices for different automobile occupancy levels. Such .distinctions are
important if the proposed pricing strategy provides incentives^ to carpools or
vanpools. The TDM Evaluation Model (discussed in Section 3.3.6.9) may be useful
. to areas that wish to perform such an analysis'.
r . ' • • • - .
Areas without Mode Choice Models • • '
Areas that do not have mode choice models will have a more difficult time assessing parking
pricing impacts, but may still do so. Suggestions for evaluations by these areas include: ;
1 • Areas with Person Trip Tables: Areas with person-trip tables (by individual trip
purpose), can obtain an estimate of modal shifts by applying one of the following
methods:
-.',••'' ~ Quick Response Methods: Mode choice analysis can be achieved with a
number of techniques that use sketch planning or pivot-point methods.
U. S. Environmental Protection Agency , : o_o
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
These include the FHWA's TDM Evaluation Model and the QRS software
system (see Section 3.3.6.9 for more information).
- "Elasticities: Mode choice analysis can be approached through application of
pricing elasticity estimates from empirical studies or from mode choice
models of similar areas (see Section 3.3.6.9 for more information and
references). It must be stressed that elasticities will vary widely from region .
to region, depending on base costs, travel times, urban-design factors, and
. mode shares as well as socio-economic factors.
• Areas with Vehicle Trip Tables (not Person Trip Tables): Some areas perform
analysis using only vehicle trip tables. It is still possible,to evaluate mode choice
responses to parking pricing if the agency first generates an estimate.of person trips
and then uses the derived estimates with either of the methods identified above.
Person trips can be derived by (1) dividing vehicle trips by average vehicle
occupancy, or (2) using a technique such as the TDM Model to produce a starting
person trip table (note that the TDM model does not include non-motorized trips,
however).
• Areas with No Trip Tables: Areas without trip tables have fewer analysis options,
but can develop at least a rough estimate of mode split response to parking pricing if
they can estimate total person trips, total vehicle trips (or average vehicle
occupancy), and VMT through HPMS or another source. Elasticity methods can
then be used to factor the base trip rates to reflect the change in parking. Segmenting
the base population by average trip length, if possible, may increase accuracy.
Generally, if an area's planning tools and data are limited to trip tables or gross trip and VMT
measures, the range of policy options that can be evaluated is also limited." Only the most
simplistic definitions can be considered, such as a uniform charge on all parking, which also raises
questions as to realism of the assumptions and assessment of the impact (both locally and to EPA).
7. Parking Pricing and Time of Day
' '"••••"
Parking pricing effects on time of day of travel are likely to be minimal in most cases; thus, no
formal analysis is required. Shifts in travel time of day are anticipated only if parking rates varied
t substantially by time of day, in which case the analysis should be treated more like congestion
pricing (Section 3.2.5-7). Presumably, most parking schemes directed at work travel would be in
effect all-day, and thus traveling off-peak would offer no advantage. Restructuring daily rates vis-
a-vis hourly rates away from all-day discounts for commuters and toward more favorable short-
term rates, however, could induce non-work travelers to shift time patterns. Analysis of such
effects is quite complicated and requires either fairly specialized tools or off-line methods (see
Section 3.3.7).
8. Parking Pricing and Traffic Assignment/Route Choice
Parking pricing is likely to have limited or no direct effect on traffic assignment. Parking pricing
applied in a special geographic pattern (e.g., downtown peripheral parking zones)-may, in some
cases, generate micro-level traffic concerns. If these threaten to be. significant, then traffic
microsimulation procedures should be considered (see Section 3.3.8.3 for more information).
3-10 ' ' JJ. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
:' to Reduce Transportation Emissions
Areas with traffic assignment as part of their transportation analysis process are expected to
perform a new assignment to reflect the impacts of mode choice and trip distribution analyses on
vehicle trip patterns and their corresponding VMT levels in the travel network. These inputs are
necessary for application of the emissions model in the next step. Areas lacking traffic assignment
capability must rely on,net changes in vehicle trips and VMT to perform emissions analysis.
9. Parking Pricing and Emissions
All analyses of parking pricing should conclude with an estimate of the net changes in VOC, NOX,
and CO. Areas with PMio attainment problems also should evaluate changes in anticipated PMio'
emissions. Parking pricing is not'expected to alter emissions rates or emissions analysis factors
related to baseline vehicle mix. However, because shorter trips are likely to be most affected by
parking pricing, emissions analysis cold start factors should be adjusted or evaluated separately to
account for changes.
Areas with traffic assignment capability are advised to use results from the traffic assignment
process in an emissions factor model, such as MOBILE or EMFAC (PARTS where appropriate), to
generate estimates of emissions (see Section 3.3.9). Areas without traffic assignment capability
are limited to application of emissions factors to net changes in vehicle trips and VMT (also see
Section 3.3.9), again accounting for possibly disproportionate changes in cold starts.
U. S. Environmental Protection Agency
3-11
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
3.2.2 Modal Subsidies
3.2.2.1 Anticipated Impacts
Modal subsidies are expected to have their
primary impact on Mode Choice. The
reduced cost of the subsidized mode
increases its attractiveness relative to all
other modes, particularly driving alone.
Subsidies could include reductions in cost
for transit users, rideshare travelers,
walkers, and cyclists. Modal subsidies
also might include incentives for
alternative work arrangements (e.g.,
telecommuting, compressed work weeks).
Impacts depend on the subsidy level (and
its net value to the recipient), mode or
modes subsidized (modal competition),
and quality of the alternatives
(service/coverage). Subsidies can
increase the attractiveness of destinations
where the alternatives are best deployed,
which could affect trip distribution and
long-term land use decisions.
Subsidies that affect peak/off-peak
differences could trigger time of day
shifts. Significant subsidies could
possibly influence vehicle ownership
decisions, but probably more as a
secondary result of household relocation.
Table 3.2 Impacts of Modal Subsidies
Analysis
, Step - „
Land Use/
Activity
Allocation
Vehicle
Ownership
Vehicle Mix
Trip
Generation
Trip
Distribution
Mode
Choice
Time of Day
Traffic
Assignment
Emissions
Impact Form
Areas where subsidy
applied can be more
attractive
May eliminate need
for multiple vehicles
None
Could induce non-
motorized trips or
alternate work
schedules
Destinations
reachable by
subsidized mode
more attractive
Encourages use of
subsidized modes
i ~ f*'
Shift trips to periods
with most subsidy
benefit
Shift in trips and
VMT as result of
mode choice &
distribution
Reduce vehicle trips
and VMT
Impact
Type
Trace
(Secondary
level if
significant)
Trace
None
Trace
Secondary
Primary -
^ "*
Secondary
(if time of
day differs)
Derivative
effect
Derivative
effect
Impact Indicator
Population and/or
employment by zone
Vehicles per
household per zone
Percent of vehicles by
VMT
Number of person
trips to/from each
zone
O/D patterns (average
trip length)
Person trips by
mode; Sfrtransjt;^ ^
no vehicle trip*;*
Average vehicle
occupancy
Trips by time period
Link volumes; VMT
Tons or kg. of VOC,
NOX, CO, PM
3.2.2.2 Basic Analysis for Modal Subsidies
In addition to the standard analysis of emissions, agencies should carefully assess the impacts on
mode choice (person trips by mode) and, in turn, mode choice effects on overall vehicle trips,
VMT, and emissions. If subsidies are mode-oriented, submitting agencies should investigate the
effects on other modes. If subsidies favor particular locations, in terms of modal access or other
amenities, agencies also should address the indirect impacts on trip distribution and land use:
3-12
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
3.2.2.3 Guidelines for Analysis
1. Modal Subsidies and Land Use/Activity Allocations
Modal subsidies are unlikely to affect long-run locational patterns unless they are substantial,
considered permanent, and linked to concurrent investment in alternative mode and/or land
use/urban design strategies that complement the subsidy; these latter also would add to the sense of
"permanence." EPA does not require a formal analysis of subsidy impacts on land use unless the
nature of the subsidy suggests an obvious impact on long-term trends; in this latter case, EPA
requires a statement of possible effects. ''
2. Modal Subsidies and Vehicle Ownership ;
Modal subsidies are.expected to have little or no direct impact on vehicle ownership and, thus, no
formal analysis is required. If an agency feels its subsidy program will directly affect vehicle
ownership levels, the agency should present appropriate assumptions or analysis to support this
claim. ,
There may be an indirect effect on vehicle ownership, however, if modal subsidies are expected to
affect land use through household shifts to areas with lower rates. Advanced analysis could then
include a re-run of vehicle ownership evaluation to reflect anticipated land use impacts.
3. Modal Subsidies and Vehicle Mix
" ' _'-("•,*
Modal subsidies are expected to have little or no impact on vehicle mix and, thus, no formal
analysis is required. Since mpdal subsidies can. have differential effects on different trip types and
lengths, however, shifts in usage patterns may be important in determining the regional fleet mix
used in the emissions estimation step '
4. Modal Subsidies and Trip Generation
It is unlikely that modal subsides will have a measurable impact on the overall number of person
trips generated by households. However, an analytical problem arises if trip generation is only
performed for vehicular trips since the overall vehicle trip reduction effects will be understated if a,
subsidy draws travelers from private vehicle to transit, non-motorized or telecommuting modes.
Modal subsidies might. influence household-trip generation, and estimates of trip generation,
through: . • • . •
• , Shifts to or from non-motorized modes, if those modes are not reflected in trip
generation estimates; and/or - .
• Attraction of non-work trips to subsidized transit that may have not been made
previously, or were made by npn-motorized mode (and not included in the trip
generation rate). -."-_.
' U. S. Environmental Protection Agency
3-13
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
* * ' ' '
While a formal analysis is not required, the performing agency may wish to estimate modal
subsidy impact on trip generation if the proposed subsidy:
• Stimulates alternative work arrangements;
• Involves transit or non-motorized modes; and
• Is set at a significant level (e.g., free off-peak transit).
While there are no known formal methods (i.e., methods typically included in a complete regional
travel demand modeling analysis) for dealing with these impacts, the user may consider off-line or
sketch planning techniques such as described in Section 3.3.6.6 (Special Modes) and 3.3.6.9 (Off-
line Methods). If the agency does not perform an analysis, it should offer an assessment of the
anticipated impact of the subsidy on vehicle trip generation.
/
5. Modal Subsidies and Trip Distribution
Subsidies that favor travel to particular destinations can affect destination choices. The effects on
destination chqice for work-related trips would only be realized in the longer term; the effects for
non-work trips, however, could well be realized in the shorter term. Trip lengths also may be
affected, depending on the nature of the subsidy (whether it is fixed or distance-based).
As these effects are likely to be secondary, a formal analysis is not required. If subsidies are
significant, however, agencies should provide a statement of possible effects on trip distribution.
Agencies with trip distribution models can perform ah analysis of modal subsidy effects on
destination choices.
Models that incorporate a composite impedance measure (see Section 3.3.5.1) can estimate the
effect of most subsidies directly, by altering the respective generalized travel costs for each
affected origin-destination pair. This approach is generally effective for motorized modes.
Current techniques, however, do not deal well with subsidies that greatly influence the choice
between non-motorized and motorized mode destination choices.2
If subsidies vary by destination, a different pattern of destination choices in the trip table would be
expected to result. Analysis should be disaggregated by trip purpose and should account for
differences in income though income stratification or other techniques (see Section 3.3.5.2).
Areas without trip distribution models of the type described here will not be able to perform this
analysis. These areas are advised to exercise caution when interpreting the modal choice effects of
subsidy policies and should state how failure to account for changes in trip distribution might
affect their final analysis.
1 It would be necessary to incorporate measures of bike/pedestrian generalized cost measures into the models.
Initially, such models would need to be estimated where there is a large difference in non-motorized mode
utility and use, with a transfer of coefficients to regions lacking such diversity, using choice model
coefficient scaling (see section 3.3.6.8).
3-14 • Ui S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
6. Modal Subsidies and Mode Choice
Because modal subsidies are expected to have their greatest impact on mode choice, formal
analysis should be used. All areas considering modal subsidies should furnish an estimate of the
change in the number of person trips by mode, the change in number of vehicle trips and average
vehicle occupancy, and the change in the percent of travelers using transit as a result of the
particular policy.
Modal subsidies alter the relative cost and attractiveness between subsidized and unsubsidized
modes, thus altering competition among the modes for person trips. Modal subsidies are expected
v to have limited or no impacts on commercial travel or through travel. Because the relationships
are complex, areas with mode choice models in their planning systems are encouraged to use them
for this analysis. -
EPA offers the following guidelines to help agencies develop the best possible mode choice
analysis:
Areas-with Mode Choice Models
• Type of Model: Multinomial logit (or similar probability-based) choice models will
provide the best estimates of impact, provided the relevant modes are included in the
modelstructure. Nested models are expected to provide the most accurate estimates
(see Section 3.3.6.5). .
• Suitable Modes/Coefficients: Since most mode choice models include cost
relationships for transit fare and automobile operating cost, they can accommodate
many subsidy policies. Coefficients' suitability is addressed in Section 3.3.6.1. If
subsidies for HOV travel involve parking discounts, then use of a parking coefficient
may be appropriate (see Sections 3.3.6.9 and 3.2.1.6).
• Accommodation of Modes: In order to test the full range of modal pricing options,
the mode choice model should incorporate different costs for each sub-mode (e.g.,
walk to transit, drive to transit, drive alone, 2-person carpool, 3-person carpool, 4-
persori carpool, and perhaps even vanpool, walk and bike). Not all models
accommodate such refinements. Alternatives include use of the TDM Evaluation
Model (Section 3.3.6.9), which allows subsidies to be targeted individually to various
carpool levels, vanpools, and transit, and distinguishes between direct out-of-pocket
and less direct per-mile costs. Most model systems (including the TDM Model) do
not accommodate non-motorized modes. Areas considering subsidies to these modes
will need to consider off-line methods for their evaluation (see Sections 3 366 and
3.3.6.9).
• Non-Work Mode Choice Models: Mode choice analysis of modal subsidy impacts
should be performed separately for each separate trip purpose; use of factoring
methods is discouraged,(see Section 3.3.6.4).
' • Income Stratification: The performing agency should perform mode choice
analysis separately for each income stratum, or use an alternative approach to
address income group differences in sensitivity and impact (see Section 3.3.6.3).
• Cost Basis: Costs should'be specified on the basis of a (daily) one-way trip.
U. S. Environmental Protection Agency ' 3_I5
-------
Technical Methods for Analyzing Pricing i
to Reduce Transportation Emissions "
• Adjustment for Inflation: The dollar value implied by the proposed subsidy should
be deflated for consistency to model coefficient's year of calibration (see Section
3.3.6.2).
• Special Policy Definitions: Certain subsidies may be difficult to test with existing
models because they may require specialized information; for example, parking cash
out requires knowing whether the commuter currently has free parking, whether the
employer leased the parking, and what the value would be to the employee back as a
subsidy. Policies of this nature can sometimes be addressed through sample
enumeration methods (see Section 3.3.6.7) or use of stated preference methods (see
Section 3.3.6.8).
Areas -without Mode Choice Models
Areas that do not have mode choice models will have a more difficult time assessing subsidy
policies, but may still do so. Suggestions for these areas are as follows:
• Areas with Person Trip Tables: If agencies have access to trip tables based on
person trips (by individual trip purpose), they can estimate the impact of modal
subsidies on modal shifts through application of one of the following methods:
- Quick Response Methods: Mode choice analysis can be achieved with a
number of techniques that use sketch planning or pivot-point methods.
These include the FHWA's TDM Evaluation Model and the QRS software
system (see Section 3.3.6.9 for more information).
- Elasticities: Mode choice analysis can be approached through application of
pricing elasticity estimates from empirical studies or from mode choice
models of similar areas (see Section 3.3.6.9 for more information and
references). It must be stressed that elasticities will vary widely from region
to region, depending on base costs, travel times, and mode shares as well as
socio-economic factors.
' '" '.- '
• Areas with Vehicle Trip Tables (not Person Trip Tables): Some areas perform
analysis using vehicle trip tables only. It is still possible to evaluate mode choice
responses to parking pricing if the agency first generates an estimate of person trips
and then uses the derived estimates with either of the methods identified above.
Person trips can be derived by (1) dividing vehicle trips by average vehicle
occupancy, or (2) using a technique such as the TDM Model to produce a starting
person trip table (note that the TDM model does not include non-motorized trips,
however).
• Areas with No Trip Tables: Areas without trip tables have fewer analysis options,
but can develop at least a rough estimate of mode split response to parking if they
can estimate total person trips, total vehicle trips (or average vehicle occupancy), and
VMT through HPMS or another source. Elasticity methods can then be used to
factor the base trip rates to reflect the change in parking. Segmenting the base
population by average trip length, if possible, may increase accuracy.
Generally, areas that are limited to trip tables or gross trip and VMT measures can evaluate only a
limited range of policy options and consider only the most simplistic policy definitions.
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
In all cases, more advanced analysis includes the use of separate models for each purpose,
disaggregated by income levels, and an accompanying supplemental analysis to assess the effects
on the use of subsidized or non-motorized modes.
7. Modal Subsidies and Time of Day
Modal subsidies can have secondary effects on time of day of travel, but, in most cases, the effects
are expected to be minimal and formal analysis is not required.
Subsidies that entail major peak/off-peak differentials may trigger migration into or out of peak
periods, particularly for non-Work trips. Since ridesharing is almost exclusively a commuter mode,
ridesharing subsidies are not expected to produce much of a shift in time of day unless time of day
differentials include off-peak periods close to conventional work day beginning and end.
Subsidy-induced differentials in transit pricing by time of day, however, could produce travel time
shifts—lower off-peak fares would be expected to attract non-work riders and a small number of
workers, while relatively lower peak fares might draw some non-work trips into the peak, as well
as some work-based transit trips. .
If time of day impacts are considered possible, methods such as those described in Section 3.3.7
may be of use. In the particular case of transit, time of day demand information may be available
through studies (possibly local) of transit peak/off-peak fare adjustments.
8. Modal Subsidies and Traffic Assignment
Modal subsidies are not expected to exert any special influence on traffic assignment. Areas that
have traffic assignment as part of their transportation analysis process, however, should perform a
new assignment analysis to reflect the impacts of mode shifts on vehicle trip and VMT levels, and
to accommodate these changes in the travel network pattern. These inputs are necessary for
application of the emissions model in the next step.
Areas which do not have traffic assignment capability must rely on net changes in vehicle trips and
VMT to perform emissions analysis.
If HOV subsidies are high enough, demand may justify new or expanded HQV facilities.
Regarding analytical procedures, this would require some additional effort in network coding and
assignment procedures; however, this would be a categorically different policy from a. straight
HOV subsidy and would argue for a formal HOV analysis.
9. Modal Subsidies and Emissions
: - •• /
All analyses of modal subsidies should conclude with an estimate of the net changes in the major
pollutants for which they are claiming credit. Modal subsidies are not expected to affect emissions
through emissions rates or through changes in regional vehicle fleet distribution; thus, no special
attention need be paid to these impacts in the emissions analysis process.
U. S. Environmental Protection Agency : ., 3-17
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Areas with traffic assignment capabilities are strongly encouraged to use an emissions factor
model, such as MOBILE or EMFAC or PARTS, to furnish estimates of emissions, using the results
from the traffic assignment step (see Section 3.3.9).
Areas without traffic assignment capability are limited to application of emissions factors to net
changes in vehicle trips and VMT (see Section 3.3.9), accounting for the possibly disproportionate
changes in cold starts.
3-18
U. S. Environmental Protection Agency
-------
3.2.3 Pump Charges
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
3.2.3.1 Anticipated Impacts
Pump charges (or similar VMT-based fees)
increase the per-mile cost of driving. This
is expected to reduce the demand for
.private vehicle travel. The expected
primary impact is on mode choice, though
significant price increases also could affect
trip distribution,, long-term locational
decisions (land use), and possibly trip
generation. Households also are expected
to'react to higher fuel prices by shifting to
ownership and/or use of higher-mpg
vehicles, thus affecting vehicle mix.
Table 3.3 Impacts of Pump Charges
', Analysis -
'*"-Step
Land Use/
Activity
Allocation
Vehicle
Ownership
Vehicle Mix
Trip
Generation
Trip
Distribution
,«ode'r; --
Choice
,, ,^.
Time of Day
Traffic
Assignment
Emission
Analysis
-• Impact Form "
Causes location in
less auto-dependent
areas
If shifts in location
encouraged
Shift to- higher mpg -
vehicles;:' "
Overall higher cost of
travel
Causes shorter
trips/closer
destinations
Decreases private '
vehicle use ~ *'/ *
*> V"
. J^j? ^ -
None
Change in trips and
VMT; may affect
speeds
Reduces vehicle
trips and VMT; may •
affect'speeds
'jJmpsicfl'l
- Ty|ie\-
Secondary
Trace
^Primary -
1/1 >,
Secondary
Secondary
Primary ' ,
None
Derivative
effect
Derivative
effect
(possible
change in
vehicle mix)
Impact Indicator
V" , ~
Population and/or
employment by zone
Vehicles per
household per zone
Percent of Vehicles
JSyVMT " ';,<•,,
Number of person
trips to/from each
zone
O/D patterns (average
trip length)
Person'trips hy*~ ,
mode; % transit} •
no vehicle trips;
Average vehicle '
occupancy^ ^ ^"f
Trips by time period
Link volumes; VMT
Tons or kg. of VOC,
NOX, CO, PM
3.2.3.2 Basic Analysis for Pump Charges
Aside from the standard analysis of emissions, analysis of pump charges should include mode
choice (person trips by mode) and the effect of mode choice shifts on overall vehicle trips, VMT,
and emissions. Agencies also should address the implications of a potential shift to higher-mpg
vehicles on effective operating cost and vehicle mix (which, in turn, affects factors used in
emissions analysis).
3.2.3.3 Guidelines for Analysis
1. Pump Charges and Land Use/Activity Allocations
Pump charges-(or VMT fees) influence long-run locational patterns by favoring locations that
support reduced dependence on private vehicles (e.g., better walk access, availability of transit,
etc.). Because this is designated as a secondary impact, a formal analysis is not required; however^
EPA expects some acknowledgment of the long-range impacts. • •
U. \ Environmental Protection Agency
3-19
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
If pump charges (or VMT fees) are significant, they may influence land use/locational effects.
Although a formal analysis is not required, any agency anticipating major changes in pump prices
should submit at least a statement of likely long-term impacts on regional development trends,
travel, and emissions. The agency may or may not support this statement with a more rigorous
analysis of anticipated impacts.
Areas with land use allocation models (see Section 3.3.1.2) that incorporate transportation cost
through some measure like composite impedance (see Section 3.3.5.1) may wish to use these
models to investigate potential land use impacts as part of their advanced analysis. Another
approach would be to consider one or more alternative growth scenarios that: (1) are based on a
system of higher travel costs, (2) incorporate alternative transportation investments, and (3)
include various land use/urban design elements (Section 3.3.1.1).
2. Pump Charges and Vehicle Ownership
Pump charges (or VMT fees) are unlikely to directly affect overall levels of vehicle ownership
and, thus, no formal analysis is required. If shifts in land use/locational patterns are projected in
Step 1, however, then these changes should be carried forward into a revised assessment,of vehicle
ownership. •
3. Pump Charges and Vehicle Mix
Given the primary impacts anticipated, submitting agencies should indicate their assumptions
regarding the effects of pump charges on vehicle mix.
Assuming that program-based pump charges are proportional to gallons of fuel purchased (rather
than set as fixed fees, per transaction or based on recorded VMT), households can be expected to
respond by gradually shifting ownership and/or use to higher-mpg vehicles.3 Shifts to higher-mpg
vehicles could have two effects:
1. Offset some of the gains from pump-based charges by reducing cost-per-mile of
surcharges; and
2. Shift the mix of vehicles reflected in the emissions model, thus the effective
emissions rate.
If the increase in fuel price will be significant, the performing agency should assess the shift in
ownership and usage to higher-mpg vehicles. This assessment could employ:
. • Forecasting tools, such as hedonic or other vehicle share models (see Section
3.3.3.1);
• Scrappage models and studies (see Section 3.3.3.2);
• Empirical/research studies or trend data (see Section 3.3.3.3);
3 If ongoing emissions charges are pure VMT-based charges (using smart-card or other mileage tracking
technology) where fuel consumption is not used as a proxy or intermediate variable, there is no incentive to
shift to higher-mpg vehicles.
3-20 . U. S. Environmental Protection Agency
-------
. Technical Methods for Analyzing Pricing Measures
'to Reduce Transportation Emissions
• Stated preference analyses (see Section 3.3.3.4); and
• Elasticity methods (see Appendix A discussion related to Vehicle Registration Fees).
Clearly, this assessment has applicability beyond those agencies that employ vehicle ownership as
a formal part of their planning system, since this assessment gauges the changes in average mpg,
cost per mile of driving, and vehicle distribution variables used for emissions analysis.
4. Pump Charges and Trip Generation
.Because pump charges increase the overall cost of vehicle usage for all trips regardless of purpose,
they.could have an impact on the number of activities that households support with trips, i.e., trip
generation. More likely, however, households would adjust trip characteristics without altering
the number of trips, for example, by favoring other modes, shorter trips, and increased trip
chaining. Thus, formal analysis of the impact of pump charges (or VMT fees) on trip generation is
not needed.
In general, EPA encourages agencies to do the following:
• Re-assess trip generation 'if land use or vehicle ownership shifts are projected in
earlier steps and the agencies have trip generation models based on these inputs;
and/or ,
• Perform an assessment of the impact of increased pump charges (or VMT fees) on
trip generation if the agencies have trip generation models able to reflect cost
through a measure such as composite /mpedowce,(see Section 3.3.5.1).4
Agencies basing trip generation on vehicle serviceable or vehicle only person trips rather than all
person trips may face concerns regarding accuracy in this step since higher fuel prices can shift
trips to non-motorized modes. These agencies may use off-line methods to approximate mode-
choice shifts (see discussion below in sub-section 6). EPA encourages these agencies to upgrade
their models to deal specifically with non-motorized trips in the future.
5. Pump Charges and Trip Distribution
Vehicle per mile charges will have a secondary effect on the distribution of trips. By increasing
the effective separation between zones, longer-distance trips will have a higher price and thus be
discouraged. The impact on work trips is expected to be long term, while the impact on non-work
trips could be realized in the near term.
A difficulty in modeling the influence of an increase in the cost per mile of automobile travel is that at the
trip generation stage of a four-step model, there is no specific trip length from which to gauge the actual
increase in the user cost of travel. Also, for transit users and bicyclists, there would be no increased user.
cost. One way around this is to use the mode choice model to calculate a composite accessibility variable
which reflects changes in the generalized price of reaching all destinations via all modes. There are few
current trip generation models that are.sensitive to user cost (see.Section 3.3.4.1.).
U. S. Environmental Protection Agency
-------
Technical ffethods for Analyzing'Pricing Measures
to Reduce Transportation Emissions
Although formal analysis is not needed, submitting agencies should submit a statement of possible
effects. Agencies that anticipate significant increases in pump prices as a result of the proposed
policy should perform a trip distribution analysis in relation to the higher fuel prices if they have
trip distribution models that:
• Accommodate costs through measures such as composite impedance (Section
3.3.5.1); or
• Incorporate travel time approximation techniques (Section 3.2.5.3.).
Where possible, this analysis should be performed separately by trip purpose and income level. In
addition, if the price per gallon is likely to trigger a shift to higher-mpg vehicles, agencies should
consider the effect of mis shift on cost per mile.
Areas that do not have trip distribution model capability, or that anticipate only insignificant changes
in prices, should document their assumptions and effects on final results.
.r
6. Pump Charges and Mode Choice
Because pump charges (and VMT fees) are expected to have a primary effect on mode choice, a
formal assessment of anticipated effects should be conducted.
Pump charges (or VMT fees) increase the per-mile cost of private vehicle travel both absolutely
and relative to other modes; this should increase the demand for alternative modes. Transit and
non-motorized modes would be exempt from price increases, thus improving their comparative
cost position. Although rideshare modes would be assessed as private vehicles, this mode would
still enjoy a cost advantage over solo driving since the increased price would be distributed among
several travelers. Although the comparative cost positions of bike and walk modes also are
improved, these modes may be less affected because of the short trip length (small total cost
impact for competing automobile trips). However, significantly higher fuel prices could generate
more non-motorized trips to closer destinations. Commercial vehicle trips could be affected
depending on the nature of the business activity.
Areas -with Mode Choice Models
Areas with mode choice models should use them to evaluate mode choice effects of pump prices
(see Section 3.3.6 for model characteristics). In using these models, agencies should increase
operating costs for all private vehicle modes by the increase in cost per mile times the origin-
destination trip length as determined from the highway distance matrix. The following guidelines
should be applied to this analysis:
• Calculation of Cost Impact: For a VMT fee, the increase in cost per mile is direct.
For pump charges, the incremental cost per mile must be determined by first dividing
the change in price per gallon by the average mpg of the regional vehicle fleet. In
the event of significant price increases yielding a shift in the vehicle mix toward
. higher mpg vehicles, the revised average mpg should be used to estimate the tested
per-mile cost impact.
3-22 . U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
• • to Reduce Transportation Emissions
• . Control for Inflation: Agencies should ensure that the cost increase input to the
. model has been inflation-adjusted to match the year of calibration of the model
coefficient (see Section 3.3.6.2).
• Income Stratification: The mode choice analysis should be performed separately
' byincome strata, or with an acceptable alternative approach (see Section 3.3.6.3).
• Trip Purpose Distinctions: Separate mode choice models should be applied for
each trip purpose (see Section 3.3.6.4). ,
Areas without Mode Choice Models ,
Areas without mode choice models may estimate mode choice impacts'from pump charge or VMT
fee policies using less formal methods. Suggestions for these areas include:
• Quick Response Methods, which include computer software methods like the
. FHWA TDM Model and the QRS system (particularly useful for areas with trip
tables); and/or
• Elasticity and Sketch Planning Methods, which include various pivot-point
methods and elasticity estimates that may be applied to either a trip table base or
' ' more aggregated travel estimates.
These techniques and their application are detailed in Section 3.3,6.9. Areas that do not formally
account for non-motorized modes in mode choice may wish to do so by considering one of the off-
line methods listed in Section 3.3..6.9. . / •
Areas without formal mode choice models or trip tables are cautioned that their analyses may have
questionable accuracy with respect to differences among travelers regarding trip purpose, trip
length, modal alternatives, and income level. , .
7. Pump Charges and Time of Day
There could be a second^order effect on peak/off-peak travel distribution to the extent that total
vehicle trips, VMT, and congestion are reduced by pump charges (or VMT fees). This effect,
. however, is not expected to be measurable; and no analysis is needed. '
8. Pump Charges and Traffic Assignment
Most of the impact from pump charges (or VMT fees), on traffic assignment is likely to come from
reductions in length and number of vehicle trips.- However, increases in the per-mile cost of
driving also may affect route selection for given origin-destination pairs. If the additional cost is
high enough, it will encourage drivers to select paths that minimize cost as well as time (e.g.,
selecting shorter but less speedy routes). In general, these effects are expected to be at the trace
level, and standard assignment analysis is expected only of those agencies with assignment
procedures.
Agencies with appropriate capabilities can estimate traffic assignment impacts of pump charges by
building highway paths based on a combination of travel time and user cost, using the per-mile
U. S. Environmental Protection Agency ' ' ,~! ' ~'. 2-23
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
user cost of vehicle operation and an assumed average value of travel time to motorists. All major
software packages are believed to provide this capability.5
Areas without traffic assignment capability must rely on net changes in vehicle trips and VMT to
perform emissions analysis.
9. Pump Charges and Emissions
jN-
All analyses of pump charges (or VMT fees) should conclude with an estimate of net changes in
VOC, NOX, and CO (and PMjo for areas with PMjo problems). The impacts on emissions are
expected to derive from projected changes in vehicle trips and VMT (without shifts in emissions
rates), and shifts in mpg and possibly speeds from the traffic assignment step (with concomitant
changes in emissions rates).
Areas with traffic assignment capability and with emissions factor models, such as MOBILE or
EMFAC, are expected to use these procedures to estimate the emissions changes. PART 5 should
be used if particulates are a problem. If analysis in Step 3 indicates a change in the regional
vehicle fleet mix, then the agency should use the new vehicle distribution results to alter the
distribution in the MOBILE or EMFAC model and resultant emissions factors (see Section 3.3.9).
Areas without emissions model, capability, that is, areas that rely strictly on emissions factor
estimates, should detail their assumptions and the procedures they used if they have modified their
base emissions rates to reflect the shift in vehicle fleet mix.
It may be possible to make this analysis sensitive to income groups by first building (skim and assign)
separate paths for each income group, so that low-income drivers are assumed to use routes that minimize
vehicle operating cost and high-income drivers to use routes that minimize travel time.
3-24 . U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
3.2.4 Emissions Fees
3.2.4.1 Anticipated Impacts
Emissions fees, as defined for this guidance, consist of increases in vehicle registration fees based
on some combination of vehicle emissions characteristics (age and/or emissions rate) and usage
(odometer reading). The emissions element rather than VMT is presumed to be dominant; that is
emissions would account for the greatest variation in the level of the fee: (-VMT-based fees are
addressed as a variation of pump charges in Section 2.2.3.-)
The primary impact anticipated is
increasing ownership and greater usage
of newer, low-emitting vehicles. If
emissions fees are significant, they also
could affect the number of vehicles
owned. In turn, this could influence
land use/locational patterns, which then
" affects trip generation.
If emissions fees are assessed
infrequently, it is unlikely that they
would have direct effects on travel
unless education campaigns make people
aware of the charges.
Table 3.4 Impacts of Emission Fees
^Analysis *
- '$t»p :
Land Use/
Activity
Allocation
Vehicle '71
* **a £, 4? <
Ownership «
Vehicle Mix
V !?»•;, "
Trip
Generation
Trip
Distribution
Mode
Choice
Time of Day
Traffic
Assignment
, Impact Form
r* f, '"£:•* '!**
Lowers demand for
auto-dependent
' environments
Reduces demand
for vehicle " <'-c """
ownerslftp' , ~i'
^Increases ** !
ownership, & use
of newer/ tow^ >
Remitting vehicles •*/
If vehicle ownership
affected
If vehicle ownership
affected
If vehicle ownership
affected '
None
None
'.:ChaijjeSi®t6le'-
SrroffemaSWancC..
impact
., ... TSP*"
Secondary
(if vehicle
ownership
is affected.)
, Primary
s Primary
•V ^
Trace
Trace
Trace -
None
None
Primary "•> -~
^ Impact Indicator „
^ -> ~ *=v®^
Population and/or
employment by zone
Vehicles per household ,f
jparzone J " - -
^ 'V ^^
Percent of vehicles by
«MT- , '
" <* ". ™Z-
Number of person trips
to/from each zone
O/D patterns (average
trip length)
Person trips by mode;
% transit; no vehicle
trips; average vehicle
occupancy
Trips by time period
Link volumes; VMT
Tons or, kg. of VOC,
NOXt COJPM - - - „
3.2.4.2 Basic Analysis
Agencies proposing to introduce
emissions fees should analyze the
effects of this measure on regional vehicle mix and the subsequent effects on emissions
attributable to a change in vehicle mix. Additional assessments should be used if significant
impacts are anticipated for vehicle ownership and, from that, for land use and other secondary
effects.
U. S. Environmental Protection Agency
3-25
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
3.2.4.3 Guidelines for Analysis
1. Emissions Fees and Land Use/Activity Allocations
Two instances in which emissions fees could trigger land use/locational impacts, are:
1. Major VMT component in fee: If the emissions fee has a major VMT component, the
cost per mile for vehicle use could deter vehicle use. In this case, analysis should
follow the guidelines for pump charges as well (Section 3.2.3).
2. Significant increase in automobile ownership costs: If the emissions fee significantly
increases the cost of vehicle ownership, it may deter vehicle ownership. A reduction
in household vehicle ownership could increase the demand for location-efficient
areas.
If the nature or magnitude of emissions fees are considered sufficient to affect vehicle ownership
rates, EPA suggests that the performing agency address the potential impact on land use/locational
patterns with a land use allocation model, if one is available and suitable, or through scenario
testing. Agencies should prepare a statement that reflects the agency's assumptions of the
potential impacts on long term location patterns and the overall impact on emissions if the agency
(1) does not have appropriate analysis tools, or (2) elects to declare this impact to be insignificant.
2. Emissions Fees and Vehicle Ownership
. Emissions fees can trigger a reduction in vehicle ownership if the fees cause a significant increase
in costs of vehicle ownership. The performing agency should account for shifts in vehicle
ownership if the agency (1) thinks that its proposed emission fee may have an impact on vehicle
ownership levels and (2) uses vehicle ownership in its modeling process. The agency may use
analytic methods or the best possible assumptions. Techniques suitable for this analysis include:
• Forecasting tools, such as hedonic or other vehicle share models (see Section
3.3.3.1);
• Scrappage models and studies (see Section 3.3.3.2);
• Empirical/research studies or trend data (see Section 3.3.3.3);
• Stated preference analyses (see Section 3.3.3.4); and
• Elasticity methods (see Appendix A discussion related to vehicle registration fees).
3. Emissions Fees and Vehicle Mix
Emissions fees are expected to have their primary effect on vehicle mix. Households would be
expected to either replace high-emitting vehicles with cleaner ones, and/or shift use within multi-
vehicle households in favor of low-emitting vehicles.
In some circumstances, such as in a revenue-neutral feebate approach, an emissions fee program
could actually increase the number of motor vehicles by encouraging diversification in favor of
low-emissions vehicles. This would be in contrast to recent trends toward vehicles with high
3-26 ; U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
• to Reduce Transportation Emissions
weight/power/performance characteristics. If the agency's proposed strategy is likely to encourage
net expansion in vehicle fleet size, the agency should address the effect of its proposed emissions
.fee program on overall vehicle ownership levels as well as on distribution.
'Emissions fees also may affect the composition or.use of commercial vehicle fleets registered in
the affected jurisdiction. Emissions fees would not, however, affect through traffic or external
travel vehicles.
Section 3.3.3 provides options that may be used to estimate shifts in vehicle mix.
4. Emissions Fees and Trip Generation
In general, emissions fees are expected to have only trace effects on trip generation. (Recall that
infrequent assessment of emissions fees would mean limited direct impact on daily travel
decisions.). Nonetheless, emission fee'effects on the costs of vehicle ownership could lower the
number of vehicles owned by households. In turn, this could affect trip generation analysis in
jurisdictions that use models based on vehicle ownership. These agencies may wish to estimate
the impacts on trip generation of changes in vehicle ownership.
Higher annual costs of ownership and operation of high-emitting vehicles also could discourage
their use in favor of low-emitting vehicles in multi-vehicle households, thereby altering household
interactions and travel decisionmaking. This could generate more errand chaining, and/or use of
non-motorized modes, thus affecting trip generation for motorized trips and trip characteristics. It
also could induce various other effects that influence mode choice, destination, or trip timing.
Formal procedures to evaluate these impacts are not in common use; off-line techniques are the
only realistic approach for addressing them (see recommendation under emissions fees and vehicle
ownership above).
5. Emissions Fees and Trip Distribution
Emission fees are expected to have only trace effects on trip distribution. Although formal
analysis is therefore not needed, some investigation of effects should be undertaken by agencies
anticipating impacts on land use and vehicle ownership and by agencies contemplating fees with a
significant VMT component.
Reduced vehicle ownership could have an effect on trip distribution results in models that employ
vehicle ownership to estimate trip generation. If earlier steps in the analysis indicate changes in
automobile ownership, areas with such models are advised to repeat the trip distribution step to
account for ownership changes.
There are no known techniques, however, that permit direct assessment of trip distribution effects
in response to changes in vehicle "availability" (i.e., vehicle ownership). One possible technique
is to use a composite impedance approach (as defined in Section 3.3.5,1), stratified by vehicle
availability.- If the coefficients are arranged so that households with fewer vehicles are more
sensitive to trip cost (as a surrogate for distance), then such households' .trips would be more likely
to be distributed to cjqser destinations, yielding a revised pattern of trips by origin-destination.
U. S. Environmental Protection Agency . 3-27
-------
Technical Methods for Analyzir^PncingMeasures
to Reduce Transportation Emissions
Trip distribution also might be affected by the VMT component of an emissions fee if that
component is significant enough. It is difficult, however, to analyze the VMT-based impact
because the relationship between aperiodic—typically annual—fee and daily travel is uncertain.
In the absence of documented relationships, areas that believe the VMT component of their
emissions fee to be significant are advised to use the same methods as they would for pump
charges (Section 3.2.3.5), which are VMT-sensitive. Alternatively, such areas might consider one
of the off-line techniques described in Section 3.3.6.
'"*'." i '
,i' . • • ;
6. Emissions Fees and Mode Choice
Emissions fees are not expected to produce a major impact on mode choice. If the proposing
agency believes the VMT component of a proposed fee will be a substantial element in mode
choice, however, the agency should undertake a mode choice analysis similar to that specified for
pump charges, incorporating the cost from the emissions fee as a change in operating cost.
Caution should be taken in treating the emissions fee as a cost element if the fee is administered
infrequently (i.e., annually), since there is little information on travelers' translation of periodic
fees into daily/per mile costs. If the performing agency chooses to evaluate a mode choice effect
in this instance, its assumptions regarding periodic fee assessment and its effect on behavior
should be clearly documented.
Emissions fees can affect mode choice indirectly through changes in vehicle ownership. If prior
analysis indicates an impact on vehicle ownership, agencies with a mode choice model that
includes vehicle ownership in its specification may revise their estimates of mode distribution of
person trips.
7. Emissions Fees and Time of Day
Emissions fees are not expected to have any measurable effect on the time of day of travel.
Therefore, no analysis is needed.
8. Emissions Fees and Traffic Assignment
In general, emissions fee measures are not expected to exert any special effect on traffic
assignment and, thus, no formal analysis is needed.
If travel changes are calculated in any of the preceding steps, for example, land use, vehicle
ownership, trip distribution, or mode choice, then agencies with assignment procedures should use
them to determine changes in vehicle trips and VMT. Areas without traffic assignment capability
must rely on net changes in vehicle trips and VMT to perform the subsequent emissions analysis.
9. Emissions Fees and Emissions Calculations
Given the anticipated impact of emissions fees on emissions, EPA requires proposing agencies to
(1) indicate how the vehicle fleet mix and VMT distribution will be changed by the particular fee
structure, and then (2) revise the appropriate parameters in the emissions model (MOBILE or
•*"2° U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
' .. ' ' to Reduce Transportation Emissions
EMFAC; PARTS for PM areas) to reflect the impact on average emissions factors (see Section
3.3.9 for information on procedures to effect this change).
Emissions-fees based on direct measurement of vehicle emissions output by inspection may result
in improved maintenance and reduced emissions rates. EPA's Office of Mobile Sources has
developed a methodology for estimating emission reductions from inspection/maintenance (I/M)
programs. The methodology is based upon the observed effectiveness of different I/M short tests
for identifying high-emitting vehicles and the observed emission reduction associated with
commercial repairs resulting from I/M programs. • ' : ''
Areas without emissions model capability, which rely strictly on emissions factor estimates, must
detail their assumptions and the procedures they use to modify existing emissions rates to reflect
the shift in vehicle fleet mix. ' -. • •
If the proposed emissions fee policy is expected to generate changes in travel, resulting in a new
travel assignment in Step 8 (or revised vehicle trips and VMT for areas without models), the
submitting agency should calculate the revised levels of VOC, NOX, CO, and PMio (where it is an
issue) due to these changes in travel and in fleet mix/emissions factors.
. S. Environmental Protection Agency
3-29
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
3.2.5 Roadway Pricing
3.2.5.1 Anticipated Impacts
The impact of roadway pricing on travel
and emissions is complex. Traveler
responses to roadway pricing are
conditioned by roadway coverage (i.e., the
extent and number of facilities that are
priced), availability of alternatives, and
trip purpose. Traveler responses include
changes in destination, route, mode, and,
under congestion pricing, time of day of
travel. Activity planning may be effected
to schedule trips more efficiently to
minimize costs.
The impact on emissions depends on
reduction in VMT and changes in flow
conditions (reduced congestion).
Location and land use can be affected in
the long term by the pricing scheme.
Table 3.5 impacts of Roadway Pricing
Analysis
Step
Land Use/
Activity
Allocation
Vehicle
Ownership
Vehicle Mix
Trip
Generation
Trip
Distribution
Mode
Choice
Time of Day
Traffic
Assignment
Emissions
Impact Form
** * j^j
Changes relative
attractiveness based
on pricing and
alternatives to driving
Depends on type of
location shift
None
May reduce vehicle .
trips if location shift
or "area" fees
Pricing affects
access (cost/
time) of particular
destinations
Use alternatives to
avoid road price
Congestion pricing
cause* time shifts
Compound:
reduces trips & „
VMT; cauuos rout* > x
shift&v '
Reduces vehicle
trips and VMT; also
congestion & speeds
/Impact
Type
Secondary
Trace
None
Primary if
area-wide —
otherwise,
trace
Primary
•/
Primary
•f ^ /
Primary
(» , C.
connection
pricing)
Primary
/ /
Derivative
effect
, Impact Indicator
Population and/or
employment by zone
Vehicles per household
per zone
Percent of vehicles by
VMT
Number of person trips
to/from each zone
0/D pattermt (average
trip length)
Person trips, by mode;
% transit; %wsMbikes
no vehicle trips;
average vehicle
occupancy
xTHp* by time period
/
•>> \ *
Link volume*; VMT
„ f
/
Tons or kg. of VOC,
NOX, CO, PM
3.2.5.2 Basic Analysis
Submitting agencies should analyze the impacts of roadway pricing on mode choice, trip
distribution and route choice through traffic assignment. Agencies considering congestion pricing
forms of roadway pricing also should evaluate the anticipated impact of the measure on time of
day of travel. Emissions analysis should reflect changes in vehicle trips, VMT, and speeds.
Roadway pricing is expected to have a secondary, rather than primary, impact on land use; thus,
formal analysis is not- needed. However, submitting agencies should include a statement that
describes their assumptions and the measure's anticipated impacts on long-term land use.
3.2.5.3 Guidelines for Analysis
1. Roadway Pricing and Land Use/Activity Allocations
In the long term, roadway pricing is likely to affect locational patterns, particularly if roadway
pricing is applied only to selected facilities. However, if a large number of facilities are priced
3-30
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
, . - . • to Reduce Transportation Emissions
then it becomes very difficult to make longer trips without paying. In that case there is a strong
incentive to avoid vehicle trips, particularly longer ones. In turn, businesses and households have
an incentive to shift to locations that either (1) are not dependent on priced roadways or (2) will
benefit from higher levels of service on the priced roads. "Areaywide" pricing reduces or even
eliminates incentives to re-locate. The degree and pattern of relocation depend on roadway pricing
rates and scope of the coverage. Substantial improvement of alternatives and supportive land use
planning together with roadway pricing, would positively reinforce these trends.
Although a formal analysis of roadway pricing effects on land use is not needed, submitting
agencies should include either (1) an assessment of the implications of the proposed policy on
travel and emissions or (2) a statement that describes the area's assumptions and interpretation of
potential impacts. This statement should reflect the appraisal of area decisionmakers as to the
impact on long-term land use and transportation plans and programs.
Performing agencies with access to land use allocation models (Section 3.3.1.2.) that incorporate
transportation cost through a measure like composite impedance (see Section 3;3.5.1) may use such
models to investigate roadway pricing impacts on long-term development trends. Results should
be treated with caution, however, in view of the analytical limitations of these procedures.
Submitting agencies (particularly those that do not use models for land use analysis) might
consider formulating alternative growth scenarios (Section 3.3.1.1) to account for the effects on
regional travel and emissions from roadway-pricing induced development shifts. These scenarios
also should reflect any complementary actions with regard to land use alternatives or urban design.
* - " - • ' -
2. Roadway Pricing and Vehicle Ownership
No direct impact is anticipated on vehicle ownership as a result of roadway pricing and, thus, no
formal analysis is needed. If land use and household location shifts are projected in Step 1,
however, these shifts may translate to revised vehicle ownership levels; agencies should account
for these effects.
, - \ - - '
3. Roadway Pricing and Vehicle Mix '
Roadway pricing is expected to have only limited impacts on the types of vehicles used by
households and, therefore, regional vehicle fleet distribution; no formal analysis is suggested.
4. Roadway Pricing and Trip Generation '
Roadway pricing can have a measurable direct impact on household person trip generation. This is
particularly true when an area-wide pricing policy is pursued. EPA suggests, when an area-wide
pricing policy is pursued, that analysis of trip generation should be conducted. Areas with trip-
generation models that incorporate a measure of generalized transportation cost (i.e., composite
impedance) are encouraged to use those methods to assess the impact on trip generation (see
Section.3.3.5.1.). Use of these techniques is likely to show significant impacts on total trip
generation, and it would be beneficial for areas to develop this capability if they are considering
area-wide pricing. Shifts in land use and vehicle ownership should be carried through to the trip
generation stage when area-pricing is being modeled.
U. S. Environmental Protection Agency ' 3.31
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Roadway pricing that focuses on a single corridor or facility is less likely to have major impacts on
trip generation. In these cases, most of the impact will be from mode and route shifting, changes in
destinations, and time-of-day (if congestion pricing is used). At a minimum, areas modeling single
corridor pricing effects should document assumptions used for any anticipated changes in trip
generation.
5. Roadway Pricing and Trip Distribution
Roadway pricing is likely to have a primary impact on trip distribution; thus, a formal analysis of
anticipated effects is suggested. In the short term, roadway pricing is likely to affect only non-
work trips through shifts in destinations selected in response to pricing patterns. In the long term,
work as well as non-work trips could be affected as workers change their work and/or home
locations to avoid the additional road price (These latter effects may be more appropriately
modeled through the land use allocation step.).
Areas with trip distribution models that can account for the effect of cost through a measure such
as composite impedance (Section 3.3.5.1) are encouraged to use those models to assess the impact
of roadway pricing on the origin-destination patterns of non-work trips (both short-run and long-
run) and work trips (long-run). Agencies should perform independent analyses for each trip
purpose by income strata or use some other approach to account for income differences (Section
3.3.5.2).
Areas with trip distribution models that do not incorporate cost directly in their measure of
separation may consider using a travel time approximation for the increased link cost (see Section
3.3.5.3).
Areas without trip distribution models will have difficulty accounting for roadway pricing impacts;
these agencies may not be able to perform a realistic analysis given the complexity of roadway
pricing programs. Pricing applications considered by these agencies would likely be limited to
situations where chances for toll-avoidance are minimal (e.g., pricing of a strategic bridge or
tunnel).
In contrast to some of the other market-based measures considered, road pricing can have a fairly
important effect on commercial and through traffic. For commercial users, the existence of a road
user charge may actually prove to be an advantage, given anticipated impacts on road congestion
on priced roadways and the importance of travel time to commercial users. The ultimate net
benefits depend on the nature of the goods or services being transported.
When travel time is of higher value than the anticipated increase in cost, local commercial traffic
(as well as personal traffic) will be drawn to the priced facility if pricing results in higher levels of
service. If pricing-induced increases in travel cost are considered important, however, trips will
shift to unpriced facilities; this, in turn, may generate local traffic congestion and emissions
problems.
Like local commercial traffic, through traffic will also weigh the toll costs against the value of
time saved on priced facilities. Price sensitive through-travelers will shift to non-priced alternative
routes or be inclined to pick an alternative route around the given region unless they simply know
3-32 U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing'Measures
.-.:.-. ,. to Reduce Transportation Emissions
of no alternatives. Traffic response to trade-offs between time and money are difficult to assess
with currently available methods, although new methods are under development.6 Agencies
wishing to assess these impacts are encouraged to do so within the limits of their analytic tools and
data. All assumptions should be clearly documented.
6. Roadway Pricing and Mode Choice
Roadway pricing is likely to have a primary impact on mode choice. Agencies should submit a
formal assessment of the impact of roadway pricing on person trips by mode, change in vehicle
trips, and average vehicle occupancy and change in transit trips.
Areas with mode choice models in their planning systems are encouraged to use them for this
analysis. However, it should'be noted that the analysis of roadway pricing presents several
important modeling issues and special requirements:
• Network Representation of Tolls: Modeling the pricing of specific rpadways
requires a concerted effort to represent the price in the highway network. This is
equivalent to properly coding a toll value on all affected, individual links. When
coding is completed, the path-building and skimming software must be alerted to the
presence of the toll and must be capable of creating a matrix that reflects the total toll
cost of traveling on all origin-destination pairs. It is, somewhat simpler to model
area-wide pricing since the matrix manipulation program in most modeling software
packages can build a cost matrix representing the additional cost of traveling into the
affected area from any origin zone.
• Mode Cost Sensitivity to Tolls: Few mode choice models are set up to accept toll'.
costs explicitly; most deal only with automobile operating cost and possibly parking
cost. This leaves open the-issue of what coefficient value to assign to tolls. If the
toll is paid physically at the point of charge, it may be reasonable to treat it as an out-
of-pocket cost comparable to parking and apply a parking cost coefficient (see
; Section 3.3.6.1).' If the toll cost is determined electronically, with billing issued
periodically, then it may be more appropriate to treat it as an operating cost. The
agency should state the assumptions underlying its choice of coefficient. The agency,
also may .wish to consider stated preference methods to ascertain the relevant
sensitivity (see Section 3.3.6.8). ' ' . •
• Queuing and Delay: If tolls are collected through automatic identification methods
(i.e., no toll booths), delays related to toll payment would be minimal. When tolls
6 Three recent studies Tiave directly addressed air quality and travel impacts of freight transportation.
Cambridge Systematics, Sierra Research, and Jack Faucett Associates, "Air Quality Issues in Intercity
Freight," prepared for the FRA, the FHWA, and U.S. EPA, draft final report, December 1996
Cambridge Systematics et al., "A Guidebook for Forecasting Freight Transportation Characteristics
Demand," NCHRP Report 388, National Academy Press, Washington, DC, 1997.
Cambridge Systematics, COMSIS Corporation, and Alan Horowitz, "Freight Planning Manual," developed
for me Federal Highway Administration. . .
U. S. Environmental Protection Agency • - 3-33
-------
Technical Methods for Analyzing'Pricing j
to Reduce Transportation Emissions
are collected manually, however, toll booth operations may well produce queue
delays that affect (1) travel time on the facility and (2) micro-level traffic effects that
contribute to emissions. Agencies may consider using traffic microsimulation
methods to assess these effects (see Section 3.3.8.3).
• Type of Model: Multinomial logit, especially nested logit models, provide the most
realistic estimates of mode choice impacts, particularly when toll treatment differs
across modes (see Section 3,3.6.5).
, • Non-Work Mode Choice Models: There should be a mode choice model for each
separate trip purpose (see Section 3.3.6.4).
• Income Stratification: The mode choice analysis should be performed separately
by income strata, or through an equivalent method that accounts for income
differences (see Section 3.3.6.3).
• Adjustment for Inflation: It is important that the dollar value of the toll price be
• adjusted for compatibility with the base year for which the model coefficient was
developed. This generally means deflating the tested price by a time-inflation factor
that makes the costs compatible with the coefficient in real terms (see Section
3.3.6.2). .
• HOV Special Rates: If the policy incorporates toll exemptions or discounts for
HOVs, few mode choice models will provide sufficient analytical support. Off-line
methods, such as the TDM Evaluation Model, are suggested for assessing this effect
(see Section 3.3.6.9).
Areas -without mode choice models will have a difficult time adequately assessing a measure as
complex as roadway pricing. Off-line methods, which include quick-response techniques and
sketch-planning/elasticity approaches (see Section 3.3.6.9), may be used with caution; agencies
using such methods must document their assumptions regarding potential limitations of the
analysis.
7. Roadway Pricing and Time of Day
If roadway pricing is applied evenly throughout the day, then only limited impacts on travel time
of day would be expected. However, if the implementation involves peak/off-peak congestion
pricing differentials, time of day shifts are anticipated; and these should be analyzed.
Current time of day analysis methods that employ "peak factors" are probably not sufficient to
capture time of day effects. Modified procedures, such as those described in Section 3.3.7, may be
considered. Analysts should estimate the specific sensitivity of travelers' start times to price
differentials at different times during the day. Non-work trips are likely to exhibit greater
sensitivity than work trips because of their greater flexibility (and perhaps cost sensitivity),
although some workers may be able to shift hours outside the peak. The sensitivity of travelers'
trip times to price differentials is uncertain in current practice, but it may be possible to make an
assessment that can be refined subsequently using observed data. Such initial estimates may be
based on empirical sources or on stated preference surveys (see Section 3.3.7.4).
All applications of roadway pricing that involve time of day price differentials should make an
estimate of the time of day shift in travel, using approximation methods such as those described
'^ U. S. Environmental Protection Agency •
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
above for broad applications of the policy. However, agencies introducing more complex
applications are expected to use more sophisticated analysis methods.
^ . I "
8. Roadway Pricing and Traffic Assignment
Roadway pricing has a special effect on traffic assignment because pricing of specific links in the
highway network is likely to induce some shifting to. alternative routes, particularly by cost
sensitive travelers with some flexibility (lower income and non-work travelers). This jvould
probably constitute the policy's primary effect and should be evaluated. Evaluating agencies
should be careful to ensure that all alternative paths to the tolled network are represented in order
to address the diversion issue.
Areas with traffic assignment models may perform an analysis of roadway pricing impacts on
revised vehicle trips, VMT, and speeds using one of the following methods:
• Toll Diversion Model: Some areas have developed models that specifically offer
travelers a choice of a toll or free path during assignment. This approach is
described in Section 3.3.8.5.
• Travel Time Equivalent: A less accurate, but acceptable, approach is to translate
the impact of the toll into a travel time equivalent (also described in Section 3.3.8.5).
This approach, however, disregards the likelihood that travelers with a higher value
; of time will opt to use a tolled facility even at higher cost simply to take advantage of
the improved travel time. That is, this method assumes away traveler willingness to
pay for time saved.
Areas without traffic assignment capabilities will probably not be able to evaluate roadway pricing
options unless alternative route choice options are greatly limited (e.g., isolated bridge or tunnel).
9. Roadway Pricing and Emissions
" ' i l ' ' " < ~
All analyses of roadway pricing should conclude with an estimate of the net changes in the major
pollutants for which credit is being claimed. '
If roadway pricing has an effect on emissions rates, it would be through changes in congestion
levels, rather than through impacts on vehicle mix. That is, roadway pricing implemented as
congestion pricing can affect emissions rates if it significantly affects level of service and travel
speeds.
Areas able to perform traffic assignment are advised to use an emissions factor model, such as
MOBILE or EMFAC (PARTS for particulate emissions), to furnish emissions estimates, using the
results from the traffic assignment process (note that MOBILE produces emission factor estimates
rather than estimates of emissions amounts). ' • :
Areas without emissions models may apply emissions factor estimating methods to the net changes
in travel projected from traffic assignment. .
Areas without traffic assignment capability are not expected to be able to perform this assessment,
except in limited circumstances as earlier described.
U. S. Environmental Protection Agency '..'-.' 3.35
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
3.3 Review of Methods and Procedures
Clearly, the most reliable basis for projecting the impacts of changes in pricing levels—especially
major changes—would be to statistically track and measure consumer response to such actions
under (and in relation to) actual conditions. Before-and-after studies, possibly involving consumer
panels, can provide great insight into the expected and unexpected nature of response.
Unfortunately, there are relatively few studies (in the United States) in which such measurements
have been made. It will be important to institute such monitoring systems in coming years and as
new policies are tested to obtain a fuller understanding of consumer response.
In the interim, we must rely on existing planning and emissions analytic procedures and develop
new ones. As indicated in Chapter 2, current capabilities vary widely across the U.S. Virtually all
major urbanized areas and many smaller areas use some variation of the "four-step" planning
process for all transportation-related analyses. This process portrays transportation decision-making
as a series of sequential steps. The four core steps are trip generation, trip distribution., modal choice,
and traffic assignment. These systems work from trip tables that describe travel movements between
origins and destinations defined by traffic analysis zones. These trips are "assigned" to a coded
transportation network to simulate actual traffic flows and conditions on actual facilities. Emissions
analysis is then performed as a separate step, utilizing basic travel outputs (vehicle trips, VMT) from
the four-step process.
In practice, there are many variations within each analytical stratum with respect to:
• Level of detail used to characterize travel by purpose, time of day, income and other
socioeconomic differences;
• Detail in the travel networks; and
• Sophistication of individual models.
Smaller urbanized areas, where transit service is minimal or non-existent, may not have a formal
mode choice capability (these systems are characterized as "three-step" models). In some areas,
capabilities are even more limited. With this wide range of capabilities among areas that might
consider market-based measures, it is difficult to prescribe use of particular methods or " standard"
techniques. Rather, the objective of this guidance is to encourage areas to perform the best
possible analysis using their existing tools and data, and to recommend special procedures that
may help them to improve their analysis. Those with more sophisticated analytical capabilities
will be rewarded with greater opportunity to receive emission credits.
It is commonly acknowledged that current-generation travel forecasting tools are not sensitive to
pricing; rather, price sensitivity has had to be back-fit into their structure. These forecasting tools
were developed primarily to guide the location and design of highway facilities; cost was not
included in their structure because of an underlying assumption that the price of transportation
3-36 . • U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
, to Reduce Transportation Emissions
would change at the same rate as inflation, thus would not be an important future criterion.7 The
gasoline crises of the 1970s and 1980s; together with federal requirements related to transit system
funding, caused many areas to initiate upgrades to their models to formally include pricing.
Pricing sensitivity, however, is limited tp mode choice in most models. It is reasonable to believe
that the effects of many pricing measures will extend beyond mode choice to include destination
choice, route choice, and even time of day; some of these may have even more significant .impacts
than mode choice. It also is reasonable to expect that longer-term effects on household or business
locational decisions can affect land use, vehicle ownership, and trip generation
/ ' • ' ' "
Current models may be unable to deal with these broader impacts. In many areas, model
capabilities are in a state of flux as they (1) undergo change and enhancement in response to the
requirements of the CAAA and ISTEA and (2) respond to new data available from the 1990
Census and numerous urban area regional travel surveys.
Until agencies have greater experience with actual implementation of major pricing applications,
there will be limited empirical evidence upon which to base or validate projected impacts from
these measures. Hence, it is necessary to extract the best possible analysis from existing travel
forecasting tools. There are numerous ways that these tools can be adapted to improve their
sensitivity and accuracy in the evaluation of pricing measures. Adaptive procedures include:
• Accounting for cost throughout the travel demand choice hierarchy through a
generalized cost measure such as composite impedance or accessibility;
. • Sharpening estimates of price coefficients through special surveys, or transferring
coefficients or models from other sites;
• Improving sensitivity to differences in response by income level through income
•'., stratification or similar techniques;
• Qaining initial insight into relationships between special pricing instruments and
travel behavior response through stated preference surveys;
• Using sample enumeration methods to achieve greater specificity for particular
measures as they are experienced by households and possibly greater accuracy in
' estimating response; and
• Use of sketch planning or off-line methods to support formal methods where tools or
capabilities do not currently exist.
This section provides a brief review of the steps in travel forecasting that may be involved in
assessing pricing actions and evaluating their impacts on travel and emissions. The intent is not to
provide a full review of the transportation planning process, but rather to draw attention to those
aspects of the process that are important in relation to pricing. This section describes modeling
concepts that offer improved accuracy for analyzing pricing and presents references for additional
information on their development and use.
7 Transportation cost also was a difficult element to include in model structures because of the.models' heavy
reliance on travel time and the close correlation between [per-mile] cost and travel time for most trips in
most U.S. urban areas.
U. S. Environmental Protection Agency *-. . . 3-37
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
'••'"'' ' '' '
This section is designed to address the range of needs and capabilities that may exist among areas
desiring to perform pricing analysis. Larger urbanized areas may include many of these concepts
in their modeling systems, while other areas seek to develop such capabilities as part of a longer-
term model enhancement process. Other areas with more limited capabilities may need help
identifying more basic techniques, as well as help prioritizing future model development activities.
The reader is directed to two primary resources that serve as important guides for agencies
involved in performing transportation air quality analysis:
1. Manual of Regional Transportation Modeling Practice for Air Quality Analysis*
(hereafter referred to as the "NARC Manual"), developed in 1993 for the National
Association of Regional Councils, is a major resource on current model capabilities
and deficiencies in relation to air quality applications. It provides a thorough review
of the requirements of the Clean Air Act and their impact on travel forecasting tools,
the nature and capability of current tools, and suggestions toward most effective
methods. Numerous references to example sites and literature sources are provided.
2. Travel Model Improvement Program (TMIP)\ which is being jointly sponsored by
the U.S. Department of Transportation, the U.S. Department of Energy, and the U.S.
Environmental Protection Agency to foster both near and long-term improvements in
travel forecasting tools. An October 1994 report on Short Term Model
Improvements10 provides a valuable update on model enhancements that are
underway or in place for addressing air quality and other closely related modeling
issues.
Both of these sources were used extensively in preparing the synthesis of analytic procedures in
this section.
In addition to these two primary references, the reader should be aware of several key research
projects now underway or recently completed which are charged with developing major
enhancements and new approaches to the analysis of transportation-related emissions. These
include:
• National Cooperative Highway Research Program (NCHRP) Project No. 8-33:
Quantifying Air Quality and Other Benefits and Costs of Transportation
Control Measures: This study is undertaking a comprehensive review of the
analytic techniques used for evaluating TCMs. It is also charged with designing a
new framework that addresses the overall linkages between transportation,
emissions, and atmospheric models. Computational limitations in each of the steps
8 Greig Harvey et al., "A Manual of Regional Transportation Modeling Practice for Air Quality Analysis,"
National Association of Regional Councils, 1993.
9 Cambridge Systematics, Inc., "Travel Model Improvement Program: Short Term Model Improvements,"
prepared for the Federal Highway Administration, U.S. Department of Transportation, 1994.
10 B. Spear, "New Approaches to Travel Forecasting Models: A Synthesis of Four Research Proposals,"
Travel Model Improvement Program, Volpe National Research Center for the U.S. Department of
Transportation and the U.S. Environmental Protection Agency, January 1994.
3-38 U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing^ Pricing Measures
., ' to Reduce Transportation Emissions
are to be addressed individually, as are those features that affect accuracy in the step-
to-step linkages.
Phase 1 of NCHRP Project Number 8-33 is complete, and copies of Phase 1 reports
are available on loan from the NCHRP. Research papers based on Phase 1 work are
expected to, be available for review in Fall 1997. Working titles for these papers are
" Relationships between Implemented Transportation Control Measures and
Measured Pollutant Levels," and "Development of an Improved Framework for the
Analysis of Air Quality and Other Benefits and Costs of Transportation Control
Measures."
NCHRP Project No. 25-11: Modal Emissions Factor Development: Undertaken
by researchers at the University of California at Riverside, this study is charged with
developing a new emissions-factor model that addresses the deficiencies of the
current-generation MOBILE and EMFAC models with regard to sensitivity to
vehicle and transportation inputs. Specifically, current models are insensitive to
variations that occur in vehicle emissions under normal driving conditions; where
acceleration, idling, steady state, etc., produce very different emissions rates. Current
models smooth over these variations by using emissions factors based on an average
speed. This project is obtaining substantial new vehicle activity and emissions data.
This data will be used to develop a modal emissions modelthat (1) is compatible
with the types of traffic simulation models used to design and evaluate transportation
operational improvements, and (2) accurately reflects the impacts of speed, engine
load and start conditions on emission.
Research has been completed for Phase 1 of NCHRP Project Number 25-11.- This
phase called for 1) review of the existing literature on factors in the vehicle operating
environment; 2) definition of the domain and distribution of the modal parameters of
each vehicle operating mode; 3) evaluation of extent to which existing models meet
analytical needs; 4) design of testing protocol to measure vehicle modal emissions;
5) preliminary testing of a sample of vehicles currently in use; and, 6) development of
a working model for analysis.
NCHRP Project No, 25-7: Improving Transportation Data for Mobile Source
Emissions Estimate: This study by the University of Tennessee and SAIC,. et al., is
specifically directed at appraising the data requirements needed for transportation/air
quality modeling. Key elements of the research include: 1) identifying transportation
variables that are available or heeded for preparing air quality, forecasts;
2) identifying techniques for quantifying these variables; and 3) describing the inter-
relationships between transportation and emissions rates.
Project number 25-7 has been completed, and the final report will be available on a
loan basis pending a publication decision by the NCHRP.
Air Quality Issues in Intercity Freight: Sponsored by the U. S. DOT. and EPA, this
project has developed methods for evaluating the emissions impacts of intercity
freight operations, especially truck'and intermodal rail. Extensive work has been
done to develop practical planning procedures to (1) realistically assess intercity
freight travel characteristics,and reactions to policy actions (TCMs), and (2) refine
emissions factor estimates for these modes.
U. S. Environmental Protection Agency . ' ' . 3-39
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Finally, it also should be noted that this guidance document is followed by two technical
appendices that may be of great relevance to reviewers: '
• Appendix A: A Case Study Adaptation of a Regional Model to Increased Price
. Sensitivity. Appendix A provides an illustrative, step-by-step example of how a
typical four-step modeling process was strategically upgraded and' enhanced to
improve its ability to analyze pricing measures, and then applied to test a range of
market-based measures. The base model that was transformed is the Greater
Metropolitan Washington Council of Governments (MWCOG) model. The
improvements are hypothetical; because the adaptation was done as a
"demonstration.," typical steps to thoroughly validate the revised model were not
performed, and the revised model is not in use by MWCOG. However, the "case
study" model is useful in demonstrating how enhancements such as those suggested
in this chapter could be made.
• Appendix B: An Example of the Application of the STEP Process to the Analysis
of Pricing Strategies. Appendix B illustrates how sample enumeration techniques
can be used to link travel survey data and conventional modeling methods, which
would provide additional flexibility and even accuracy to analysis of pricing
measures' impacts on air quality. These methods have, been used to analyze pricing
TCMs in Seattle, Washington and several California cities.
Neither of the approaches set forth in these appendices represent an analytic procedure prescribed
by EPA. They are provided for illustration only, to show analysts how specific adjustments and
innovative applications are being made to address the complex range of impacts that may derive
from market-based measures. The procedures described also are useful in demonstrating two very
different ways of approaching analysis of market-based measures — through concerted direct
modifications to the existing four-step process, and through a "parallel" approach that moves
outside the conventional process for key analytic assessments. Both approaches have distinct
advantages and disadvantages, which must be evaluated by the performing agency.
3.3.1 Land Use/Activity Allocation Methods
Most regional planning processes start with a procedure that allocates household and employment
activity to specific geographic locations across the region. The pattern and level of these
allocations then becomes the basis for all subsequent estimates of trip levels and flow patterns.
'' „',.,!
A key question for analysis of pricing measures is whether these locatipnal decisions might be or
should be sensitive to levels and variations of transportation pricing effects. This question has two
levels: (1) "real world" sensitivity of household and business locational decisions to pricing
measures and (2) analytical sensitivity of the transportation planning process to pricing measures.
This is perhaps one of the least precise steps in the current transportation analysis process, though
it represents what can be a very significant factor in long-term travel trends and effectiveness of air
quality and other management strategies. Currently, most land use allocation decisions are made
through a process of negotiation, with considerable political involvement and concern. Market
realism may or may not be prominent in this process.
£/.. S. Environmental Protection Agency
-------
Technical Methods:for Analyzing Pricing Measures
, - to Reduce Transportation Emissions
There are two basic ways in which areas attempt to account more explicitly for the effects'of
market forces or transportation policy on growth allocations: scenario approaches and land use
allocation models. .
3.3.1.1 Scenario Approaches
Generally, population and employment allocations based upon hard data (including population and
employment trends, land availability and price, land use occupancy and rent, zoning and land use
decisions, economic development plans and market assessments, measures of current and
anticipated transportation accessibility, etc.) are more realistic and sensitive to market forces than
those based purely on "fair share" allocations or political aspirations. Some areas address their
uncertainty in both level and distribution of growth by formulating one or more scenarios that are
then carried forward into the analysis. These scenarios can be. used to reflect major differences in
expectations linked to pricing, urban design and zoning/growth management changes, or other
policies. Analysts should take care that eventual emissions results that stem from these scenarios
reflect realistic and implementable land use assumptions."
11 Greig Harvey et al., "A Manual of Regional Transportation Modeling Practice for Air Quality Analysis "
' 1993, Sec. 3.3.2, pp. 3-13 to 3-18.
U. S. Environmental Protection Agency
3-41
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
3.3.1.2 Land Use Allocation Models
Formal mathematical models for land use allocation exist and are in use in a small number of
regions. Options include DRAM/EMPAL12 (most commonly used land use model in U.S.),
POLIS13 (currently used in the San Francisco Bay Area), the Herbert-Stevens Model14 (used in the
Penn-Jersey Study), MEPLAN15, METROSIM16, TRANUS'7, and TOPAZ18. A recent draft report
prepared for EPA's Office of Mobile Sources provides a comprehensive review of the various land
use modeling tools currently available.19
The models mentioned above generally use zone-level time series data on population,
employment, land availability and accessibility to allocate population and employment forecasts to
subareas. Use of these models has been limited, however, both by the effort required to run them
and by questions about their predictive accuracy.
Factors that appear to limit the accuracy of current generation land use models include: a tendency
to extrapolate past trends, rather than incorporate behavioral decision-making; a tendency to base
household location decisions substantially on the work trip; limited accounting for market
responses (real estate, speculative investments, ethnicity) and time-lags; and weak representation
of transportation accessibility and cost.
12 S.H. Putman, "DRAM/EMPAL - ITLUP: Integrated Transportation Land-Use Activity Allocation
Models: General Description," S.H. Putman Associates, Philadelphia, Pennsylvania, 1991.
13 P. Prastacos, "Urban Development Models for the San Francisco Region: From PLUM to POLIS,"
Transportation Research Record 1046, 1985.
14 J.P. Herbert and B.H. Stevens, "A Model for the Distribution of Residential Activity in Urban Areas,"
Journal of Regional Science 2(2), 1960, pp. 21-36.
15 MEPLAN was developed by staff at the Center for Land Use and Built Form Studies at Cambridge
University, at the firm of Applied Research of Cambridge, and at the firm of Marcial Echenique and
Partners. MEPLAN allocates exogenous forecasts of basic employment to analysis zones and then allocates
non-basic employment and population to zones based on costs of travel, floor space, and the availability of
other goods and services. Demand for accessibility generates rent levels, and land uses are segregated from
one another based on capacity to pay rent.
16 METROSIM models regional impacts on land use patterns and transportation systems that may result from
project development, and was developed by Alex Anas of the University of Buffalo. The model can analyze
impacts for up to 3,500 analysis" zones. For each zone,. METROSIM models employment, real estate
demand, household demographic characteristics, and travel patterns.
17 TRANUS shares many characteristics with MEPLAN. Development of the TRANUS model is outlined in
T. de la Berra," Integrated Land Use and Transport Modeling," Cambridge University Press, 1989.
18 J.F. Brotchie et al, "Alternative Approaches to Land Use Modeling," Commonwealth Scientific and
Industrial Research, United Kingdom, 1981.
19 Systems Applications International, Inc., "Evaluation of Modeling Tools for Assessing Land Use Policies
and Strategies," draft report prepared for Office of Mobile Sources, April 1997.
3-42 JJ. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
. to Reduce Transportation Emissions
These models can aid land use impact evaluations, provided they are applied with careful
judgment. Similar to typical trip distribution mbdels, land use models generally employ travel
time m, their allocation process. One way to enhance their utility for pricing analysis is to
introduce pricing directly as a variable through a measure such as composite impedance (see
Section 3.3.5.1). The Bay Area model system utilizes such a formulation, and would be worth
further study by areas considering such enhancements and/or where the rate or pattern of growth is
a major issue.
Readers are encouraged to consult the NARC Manual (Section 3.3,2) and the TMIP's review of
Short-term Travel Model Improvements (Section 3.0) for rnqre information on land use models
and potential enhancements.20 , .
3-3.2 Vehicle Ownership
Vehicle ownership is often used as an input for U.S. urban area model systems. Agencies in some
areas use vehicle ownership rates (or a measure of vehicle availability) as a measure of
households' propensity to travel in trip generation and other modeling steps. Very few areas
explicitly estimate a policy-sensitive model of vehicle ownership rates as part of their overall
transportation modeling system. Most planning professionals believe that explicit representation of
vehicle ownership results in more accurate and realistic transportation analysis. Evidence suggests
that vehicle ownership influences person trip generation in ways that extend beyond the effect of
income. Vehicle availability may have a key effect on discretionary trips and on the level of
errand chaining. Households with fewer vehicles per licensed driver are more.pressed to work out .
vehicle use priorities, which affects vehicle sharing decisions and use of non-motorized modes.
Where residential densities and levels of transit service are high enough to affect vehicle
ownership levels, explicit prediction of vehicle ownership would clearly be expected to increase
forecasting accuracy.
Model systems that rely exclusively on household income and not vehicle ownership can still offer
good accuracy, depending on how well income is specified as an explanatory variable. Generally,
this means that other modeling steps must exhibit strong relationships between travel and income. '
20
Throughout the remainder of this section, the Greig Harvey et al., publication, entitled "A Manual of
Regional Transportation Modeling Practice for Air Quality Analysis," National Association of Regional
Councils, December 1993, will be referred to as " the NARC Manual."
The report by Cambridge Systematics, Inc., entitled " Travel Model Improvement Program: Short Term Model
Improvements," (prepared for the Federal Highway Administration, U.S. Department of Transportation 1994)
will be referred to throughout the remainder of this report as " the Short-term Model Improvements Report."
U. S. Environmental Protection Agency ' > ' 3 43
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
There are several types of vehicle ownership predictive models. These include:
•• ' * • - ' •
• Cross-Classification Models: tables of factors derived from Census or travel
survey data (generally have no explicit linkage to transportation conditions).
• Nomographs or Curves: empirical curves that reflect the fraction of households
owning a particular number of vehicles, based on household size and income level.
''•'!"
• Mathematical Models: actual equations that calculate the number of vehicles
owned, based on the value of specific equation variables.
Generally, the form of model is less important than the sensitivity with which it reflects and
responds to key socio-demographic and policy variables, especially pricing. Aggregation of key
variables may affect accuracy levels. For example, some models produce an estimate of average
vehicles per household rather than preserve the distribution of number of households by number of
vehicles. The reader is encouraged to consult the NA&.C Manual21 for a more detailed discussion'
of the attributes of vehicle ownership models.
In order to analyze the effect of pricing measures such as registration fees, these models must be
somewhat sensitive to the price of purchasing a vehicle and the annual cost of owning one. In the
context of travel modeling, increases in ownership cost should be reflected in decreased numbers
of multiple-vehicle households and increased numbers of one-vehicle households (possibly even
an increase in zero-vehicle households). The inclusion of ownership cost .is not a common feature
in current vehicle ownership models, presenting an area for research, along with composition and
use of vehicles (discussed below under vehicle mix). At least one vehicle ownership model has
been documented that is sensitive to user cost.22
3.3.3 Vehicle Mix
Vehicle mix, which is a special area of concern in modeling pricing impacts on vehicle ownership,
is not accounted for in conventional vehicle ownership analyses. Vehicle ownership models
account exclusively for the number of vehicles owned by a household, and do not deal in any way
with the characteristics or use patterns of those vehicles. Yet, vehicle mix is-an important factor
for policies such as emissions-based registration fees or even fuel prices, where the respective
economic forces may be more likely to influence the type of vehicle owned or its relative use,
rather than the number of vehicles owned.
21 Greig Harvey et al., "A Manual of Regional Transportation Modeling Practice for Air Quality Analysis,"
National Association of Regional Councils, December 1993
K. Train, Qualitative Choice Analysis, Theory, Econometrics, and an Application to Automobile Demand,
The MIT Press, Cambridge, Massachusetts, 1986.
3-44 U. S. Environmental Protection Agency
-------
, . Technical Methods for Analy2ing Pricing Measures
to Reduce Transportation Emissions
The accuracy of emissions analyses based on an emissions factor model depends on the
distribution of the regional vehicle fleet and its average emissions rates, which derive from vehicle
distribution. Shifts in the age or type of vehicles owned would obviously affect this relationship,
a,s would different patterns of use, since the distribution is weighted by VMT for analysis.
Typically, vehicle distribution is not directly considered by conventional travel analysis tools.
Hedonic models, scrappage studies, empirical studies, special studies, and sample enumeration
offer approaches to considering fleet vehicle mixes. It should be noted that there may be overlap.
between these categories of approaches. For example, empirical studies may include some
examples of special studies. The bullets below provide brief descriptions of these analytical tools
and some illustrative examples .--'". .
Hedonic Models—These models analyze the relationship between consumer
demand for characteristics of goods rather than for goods themselves. These models
can be used, for example, to project price-induced shifts in ownership of vehicles
with different characteristics. Hedonic models are complex and may only be an
option where staff have a strong economics background. Works by Train and by
Golub and Kitamura offer two examples of hedonic vehicle choice models.23*24
Sample Enumeration—Sample enumeration involves, in the vehicle mix context,
taking a sample of households within a region arid extrapolating to the regional
population base from the sample's vehicle ownership and/or usage responses to
pricing measures. Survey samples have the advantage that they can be re-surveyed
in follow-up analyses to study actual responses to changes in price signals (see
Section 3.3.6.7 below for more information).
Scrappage Studies—Scrappage studies have been performed in conjunction with
programs to buy back or otherwise accelerate the retirement of older, high-emitting
vehicles from the regional fleet. These studies offer estimating methods that may be
of value to many areas.2526
Empirical Studies—Periods of major change in fuel prices relative to income during
the 1970s and early 1980s, and major changes in the price of vehicles and their
characteristics relative to real incomes, have produced trends in ownership that are
reflected in various research studies. These studies could be used to develop
estimates or assumptions regarding shifts in ownership or use patterns that might
occur in response to price signals.27'28
23 K. Train, Qualitative Choice Analysis, Theory, Econometrics, and an Application to Automobile Demand
The MIT Press, Cambridge, Massachusetts, 1986.
24 T. Golob and R, Kitamura, for the California Energy Commission. It should be noted that the analysis
conducted by Golob and Kitamura is not publicly available as of this writing. ,' ' ' -
25 U.S. Environmental Protection Agency, "Accelerated Retirement of Vehicles," in Transportation Control
Measure Information Documents, March 1992.
26
27
Beth Deysher and Don Pickrell, Emissions Reductions from Vehicle Retirement Programs, U.S.
Department of Transportation, John A. Volpe National Transportation Systems Center, 1996.
David L. Greene, "Vehicle Use and Fuel Economy: How Big is.the Rebound Effect?" The Enerzv
Journal 13, No. 1,1992, pp. 117-143. '' ^
,U.S. Environmental Protection Agency ' 3.45
-------
Technical Methods for Analyzing Pricing Measures
to_ Reduce Transportation Emissions
/ ' i f
• Special Studies—Locally-performed studies involving stated preference techniques
could produce estimates of consumer response to particular types of vehicle-based
pricing policies (see Section 3.3.6.8.below for more information). Recent work
sponsored by the California Energy Commission has used such methods to ascertain
the demand for electric vehicles. The study used travel data from panels of
households to establish current travel and vehicle use patterns, and investigated how
these patterns might change under different pricing and vehicle attribute
conditions.29
3.3.4 Trip Generation
This step of the planning process estimates the number of daily trips that are produced by
households {productions] or that originate in a zone [attractions]. These trips usually are
estimated separately by trip purpose. The simplest classification is home-based work, home-based
other, and non-home based trips; more complex systems split home-based non-work trips into
shop, school, and "other" and disaggregate non-home-based trips into work-related and other.30
Trip productions and attractions are estimated through separate procedures. These may be either
cross-classification relationships, matching a particular trip " rate" with a particular combination
of determining characteristics, or regression models, which use a functional relationship expressed
as an equation. The cross-classification models are the most common approach, and can be
reasonably accurate depending on the variables used for stratification, the degree of stratification,
and the reliability of the trip rates themselves. Regression models are generally regarded as less
accurate (unless they are of a fairly sophisticated non-linear form).
Significant references exist on trip generation practices. The reader is advised to consult Section
3.3.5 of the NARC Manual or the TMIP Short-term Travel Model.Improvement report for more
information and references on the general subject.31
There are a number of issues related to trip generation methods that are important when
considering the suitability of models to evaluate pricing actions; these include:
• Person trips versus vehicle trips;
• Inclusion of transportation and cost variables;
28 Don Pickrell, "Automobile and Gasoline Demand Revisited," Volpe National Transportation Systems
Center, U. S. Department of Transportation, 1992.
39 C. Kavalec, "CALCARS: The California Conventional and Alternative Fuel Response Simulator,"
California Energy Commission, April 1996.
30 Some models, such as those of the Maryland National Capital Parks & Planning Commission, have
evaluated trips generated by time of day and purpose, (Montgomery County, Maryland) and applied
distribution, mode choice and parking factors.
31 Cambridge Systematics, Inc., "Travel Model Improvement Program: Short Term Model Improvements,"
prepared for the Federal Highway Administration, U.S. Department of Transportation, 1994.
Greig Harvey et al, "A Manual of Regional Transportation Modeling Practice for Air Quality Analysis,"
National Association of Regional Councils, December 1993
3-46 U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
• Disaggregating household by vehicles (income) and size and workers; and.
.• Trip chaining and other trip adjustments.
Procedures exist within current practice or are being looked at in research efforts that offer to
enhance trip generation capabilities in dealing with pricing.
3.3.4.1 Person Trips versus Vehicle Trips
Most trip generation models furnish estimates of daily person trips; however, these trips are
commonly limited to vehicle trips and exclude non-vehicle trips such as walking and bicycle
trips.32 Smaller planning agencies may even restrict trip production estimates to person trips made
in private vehicles (eliminating transit entirely). ,.
Exclusion of non-motorized or non-automobile trips from trip generation analysis generally stems
from capabilities elsewhere in the model system, especially mode choice and trip distribution
analyses. Including non-automobile trips in vehicle trip-only systems implies a separate need for a
mode choice model, which greatly increases the system's complexity. Yet, agencies that use
vehicle trip-only approaches seldom deal with non-automobile travel. Consideration of non-
motorized modes requires an approach that:
• Looks at a finer-grained pattern and mix of land uses, street space allocation and
design; and
• Considers proximity, as well as mobility.
- . ' . • « i
The availability and consideration of these modes as travel options in response to pricing (or land
use) actions is clearly important; their inclusion in the analytic process may represent a future
standard for progressive model systems and planning agencies. Areas that have developed true
person trip generation capabilities include Portland, Oregon and Montgomery County, Maryland.
The EPA and the FHWA are currently developing a manual for MPOs to use to upgrade their
micro-scale design features modeling practice. This manual should be available in 1998 thrpugh
the Travel Model Improvement Program.
As a near-term goal, agencies that limit trip generation to Vehicle trips should endeavor to expand
their systems to full person trips.
3.3.4.2 Inclusion of Transportation and Cost Variables
Few trip generation models relate the effect of transportation level of service or cost to household
trip productions. More research is needed to determine whether a major change in transportation
cost or level of service would affect the overall number of person trips made by a household,
32
A vehicle-trip is the one-way movement of a vehicle between two points. A passenger- or person-trip is
defined as one passenger making a one-way trip from origin to destination. Person-trips are measured as the
sum of the number of passengers, added across all vehicle-trips. Thus, person-trips are equal to vehicle-trips
only if the average vehicle occupancy is one passenger per vehicle.
U. S. Environmental Protection Agency 3.47
-------
Technical Methods for Analyzing Pricing Measures"
to Reduce Transportation Emissions
particularly in light of non-motorized travel options and growing opportunities for
telecommunications or delivery as substitutes for travel.
Transportation characteristics and costs might be incorporated through the development of a
measure of composite impedance (see Sec. 3.3.5.1 below). Composite impedance is a measure
which combines the time and cost of highway and transit (and perhaps other) modes into one
statistic. In the trip generation context, composite impedance is the unit of measure used to
determine how far travelers are willing to go from a defined origin. More specifically, the question
is phrased as "How many jobs (or other activity variable) in each zone can be reached within X
units of composite impedance from that zone?" The value of X, the tolerance limit for composite
impedance, would be determined in the calibration phase.
Analytically, an increased road price (or reduced level of service) would increase the composite
impedance associated with trips along that road, thus reducing the destinations that could be
reached within a fixed impedance limit from zones in the area. As composite impedance increases
and accessibility decreases, trip frequency should also decrease because in general it becomes
more difficult Jo go anywhere, particularly by vehicle.
The Bay Area model system includes composite impedance (termed accessibility) in its trip
generation model for home-based shopping, and is one example of an attempt to include such a
relationship in trip generation (see the NARC Manual, Sec. 3.3.5, pp. 3-36-37 for more
information).
Trip generation also can respond indirectly to pricing. If pricing applications are evaluated with
models or procedures that affect an earlier-step change in land-use/activity location, or in vehicle
ownership levels through some type of composite impedance relationship, then trip generation
estimates would be subsequently affected. The Portland, Oregon vehicle ownership model, for
example, is sensitive to transit service accessibility (and pedestrian environmental quality) and, in
turn, trip generation is sensitive to vehicle ownership.
3.3.4.3 Disaggregating Households by Vehicles (Income) and Size
Most trip generation and distribution models are handicapped because they must use zonal-average
data. One of the more important recent modeling improvements is disaggregation of certain zonal
data by characteristics that influence travel. A common practice is to split households jointly by:
• Household size and vehicles available, used by Washington, DC, Detroit,
Philadelphia, San Francisco, and Portland, Oregon; or
• Household size and household income, used by New Orleans, Atlanta, Northern
New Jersey, Dallas, Denver, Phoenix and Minneapolis-St. Paul.
Such disaggregation of household data produces cells that are more homogeneous in nature,
reducing the variance in trip rates, Using income or vehicles as one of the dimensions also allows
the explicit identification of trips by some measure of wealth, which can be useful for modeling the
differential effect of pricing on various groups.
Typically in these submodels, analysts use Census data to establish the relationship between zonal
averages and the distribution of households by number of persons (I,12, 3, 4 and 5+) and number
3 "48 U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
1 to Reduce Transportation Emissions
of vehicles (0, 1, 2+). From these individual distributions, the region's joint distribution of
households by size and vehicles is used to estimate the equivalent joint distribution for each-zone
The reader is encouraged to consult Appendix A for a discussion of this procedure, demonstrated
by the case study model.
3.3.4.4 Trip Chaining and Other Trip Adjustments
Trip chaining is one likely response to an increase in the cost of travel (or reductions in the number
of vehicles available to a household); that is, individuals group trips to make more efficient use of
vehicles o^ reduce the cost associated with a given trip. This could have an effect on total VMT.
and potentially an impact on cold starts. -
No conventional models deal explicitly with trip chaining33, although researches ongoing under the
Travel Model Improvement Program, and improved operational models that deal with this activity
should be available in the next few years.34, 3S, 3V37 Data from the 1974 and 1980 gasoline
shortages might be investigated to gauge the impact of chaining responses to travel constraints.
Current efforts to develop activity-based models represents one important approach to addressing
methodological difficulties posed by trip chaining and other trip adjustments. In this approach, the
number of trips generated by a household is. not considered to be a function of household
characteristics directly, but of activities that generate travel. This approach is expected to be useful
because respondents typically can provide better information about the number of travel-
generating activities they are involved in, and can provide better information about the number of
trips associated with eaph activity.
Households can affect overall trip rates through other trip generation adjustments, such as '
telecommuting and other telecommunications applications and alternative schedules (compressed
work weeks), which reduce the physical number of home-based work trips. The TDM Evaluation
Model allows for some estimation of the change in person trips due to these strategies (see Section
3.3.6.9. below). However, there is some question as to whether households that engage in
telecommuting or alternate schedules actually, reduce overall household trip rates or simply -
33 Montgomery County, Maryland: TRAVEL2 Model deals with chaining in a form that is sensitive to land
use (Maryland National Capital Parks and Planning Commission, 1991).
34 Erniq Cascetta (University of Naples - Italy), Agostinq Nuzzolo (University of Tor Vergata - Rome - Italy)
Vito Velardi (ELASIS - Salerno'- Italy), "A System of Mathematical Models for the Evaluation of
Integrated Traffic Planning and Control Policies.^ Unpublished .research paper, Laboratorio Richerche
Gestione e Controllo Traffico, Salerno, Italy, 1993. . '•''',
3? Moshe Ben-Akiva, John L. Bowman, and Dinesh Gopinath, "Travel Demand Model System for the
Information Era"-Transportation, vol. 23, pp. 241-266, 1996. , . '
Yoram Shiftan, "A Practical Approach to Incorporate Trip Chaining in Urban Travel Models," Fifth
National Conference on Transportation Planning Methods Application, a Compendium of Papers, 1995.
37 Konstadinos Goulias, Ram Pendyala, and Ryuichi Kitamura, "Practical Method for the Estimation of Trip
Generation and Trip Chaining," University of California Transportation Center, University of California
Berkeley, 1991. , • " . '
U. S. Environmental Protection Agency •: ! ' 3*49"
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
substitute other travel. Research is still necessary on this issue, before these pattern shifts can be
assumed to lower household trip rates.38
3.3.5 Trip Distribution
',"''• *' , ' *-'
The trip distribution analytic step links together the "trip ends" (i.e., productions and attractions)
that are estimated in trip generation, into specific origin-destination trip movements, or "trip
tables." , •
In most instances, the distribution of trips from an origin zone to a potential destination zone is
performed using a gravity model. This model distributes trips as a function of the number of trip
attractions relative to the degree of separation among the zones. Almost all areas measure
separation by highway travel time.
Researchers have long known, however, that factors other than highway time play a role in the
allocation of trips among destinations. For example, there is considerable evidence that the
presence of good transit service between two zones will increase the number of person trips
between those zones. A logical extension of this concept is that other dimensions of separation,
such as prices that users pay for transportation, also influence destination choice. The significance
of these factors, which is supported by intuitive logic and empirical observation, should be taken
into account.
This section discusses several topics that affect analytic accuracy of the trip distribution step:
composite impedance, income stratification, travel, time equivalency, over-allocation of attractions,
trip distribution as "destination choice," and importance of short trips.
3.3.5.1 Composite Impedance
Gravity modeling that uses composite impedance as its measure of separation is one approach that
is being tested and used as a way of accounting for pricing effects in trip distribution. Composite
impedance, a measure that combines the time and cost of travel modes into one statistic, is
generally computed as the natural logarithm of the sum of the exponentiated disutilities for all
available modes (including non-motorized), i.e., the natural log of the denominator of a logit mode
choice model, also called the " log sum."
This approach has been used in the model systems of New Orleans, San Francisco, Boston, Atlanta
and Denver.39 Most areas that use this formulation use it only for work trips, but theoretically it is
applicable to all trip purposes. The Case Study Model in Appendix A also illustrates such an
application.
38 P. Mokhatarian et al., "Effectiveness of Telecommuting as a Transportation Control Measure," University
of California, Davis, 1994.
39 Greig Harvey et al., "A Manual of Regional Transportation Modeling Practice for Air Quality Analysis,"
National Association of Regional Councils, December 1993, Sec. 3.3.6, pp. 3-47,48.
3-50 • U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
, to Reduce Transportation Emissions
3.3.5.2 Income Stratification
The New Orleans model adds another dimension to trip distribution that merits consideration:
income, stratification. The composite impedance calculation accounts for income level of the
traveler, as well as time and cost, of all modes. This is practical because all elements of the New
Orleans model (from trip generation to mode choice) are income-stratified.
The benefit of income stratification in this context is that the effect of price on destination choice
is.differentiated according to traveler [household] income level. Since pricing changes will
generally influence low income travelers more strongly than high income, stratifying by income
allows the trip destination choice to vary realistically in discontinuous fashion relative to income.
Other models provide similar sensitivity by dividing price variables by average household income
for travelers expressed as cents per minute.
" ' . -' * ', ' • '•
3.3.5.3 Travel Time Equivalency
Many agencies include travel time only in their measure of separation for trip distribution and
mode choice analyses.' These agencies may approximate the effect of an increase in cost for
certain measures by translating cost into an "equivalent" value of travel time. Such agencies
incorporate a cost increase from market-based measures as an increase in travel time impedance
for a given origin-destination pair. The increment in time is cpmputed by using traveler value of
time to convert cost to minutes of equal value, which can be interpreted from the coefficients in the
mode choice model.40
, Time equivalents can vary by situation and type of market measure to reflect variations in traveler
price sensitivity. For example, a pricing action that affects automobile operating cost would
probably translate as a lower value of, time than an out-of-pocket cost such as a parking charge.
Also, non-work travel would be expected to exhibit a "lower value of time/higher value of cost"
ratio than work travel.
3.3.5.4 Over-Allocation of Attractions
With a doubly-constrained trip distribution procedure, such as the standard gravity model, it may
prove necessary for trip attraction estimates to, be sensitive to user costs. Otherwise, trip pattern
distortions could occur under certain pricing scenarios as the gravity model tries to reallocate
attractions in response to cost changes, but the attraction model would maintain the same
attractions in a zone. This could result in over-allocation of attractions.
This problem can be minimized by including some type of price sensitive variable, such as
composite impedance, in the attraction model. Another option is to switch to a logit destination
choice model to allocate attractions, as discussed below.
40 In the Washington, D.C. Council of Governments model, for example, the ratio of the home-based work
auto running time coefficient to the auto operating cost coefficient = 0.0173/0.0035, or about $0.05 per
minute. . . • • r
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
3.3.5.5 Trip Distribution as "Destination Choice"
Trip distribution analysis attempts to represent the process of travelers choosing destinations.
Ultimately, it approximates behavior by matching origin and destination " potentials," offset by the
degree of difficulty in traveling between zone pairs.
Some areas have implemented a more advanced procedure that explicitly models destination
choice as a behavioral process, much like mode choice. These models estimate the probability that
a traveler will choose a particular destination based on the "utility" or attractiveness of that place
relative to other places; attractiveness is defined by characteristics of the destination, the traveler,
and travel conditions, including cost.
The behavioral approach offers a more realistic evaluation of destination choice behavior,
particularly with regard to cost factors, than conventional gravity models, which require
considerable manipulation with adjustment factors to match "behavior" with observed origin-
destination trip volumes. Portland, Oregon uses such a model, with separate formulations for
home-based work, school, college, and other; non-home based work-related; and non-work.41
Some areas also have found it advantageous to model destination choice and mode choice
simultaneously, particularly for non-work travel where these decisions are so interrelated. The
Bay Area exemplifies this approach with its logit probability model for shopping trips.42
Greig Harvey et al., "A Manual of Regional Transportation Modeling Practice for Air Quality Analysis,"
National Association of Regional Councils, December .1993, Sec. 3.3.6, pp. 3-45,47.
42 Greig Harvey et al., "A Manual of Regional Transportation Modeling Practice for Air Quality Analysis,"
National Association of Regional Councils, December 1993, Sec. 3.6, pp. 3-47,48.
3-52
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
• to Reduce Transportation Emissions
3.3.5.6 Importance of Short Trips
Pricing measures place a great premium on being able to effectively model short-distance trips,
since short vehicle trips contribute substantial emissions from cold starts. These are the trips that
are perhaps most amenable to diversion to transit or non-motorized modes. The importance of
short trips in emissions policies raises concerns about the existing level of detail in analytical
representation of transportation highway and zonal networks.
In fact the importance of short trips could justify an increased level of network detail, down to the
level of minor collectors and perhaps local streets in applicable areas, accompanied by smaller
traffic analysis zones to overcome problems with non-representation , of " internal" trips.
Moreover, this same concern might argue for better data on walk and bicycle trips and
infrastructure, transit accessibility, and information on the mix'of land uses.
3.3.6 Mode Choice
The mode choice step estimates the share of trips made using each mode of travel. Within the
four-step process, mode choice models generally stand out as being the most rigorously developed
and directly price-sensitive analytic procedure. Many larger urban areas have developed
sophisticated discrete-choice logit models that estimate the share of person trips by mode, based on
the socioeconomic level of the traveler, and the time and user cost attributes of the various modes.
Models typically address the major vehicular modes—autos and transit—with shared-ride often
treated as a separate mode. More advanced models also include non-motorized modes (walk and
bike), which is an important consideration for pricing.
There are a number of characteristics or application procedures that can maximize the sensitivity
and accuracy of mode choice models with regard to pricing. Topics in this section include:
• Satisfactory model coefficients; . , , ,
• Effects of inflation; . - '
• Income stratification;,
• Differentiation by purpose;
• Integrating mode and destination choices;
• Nested versus multinomial models;
• Special modes; ' ' ' ' " ; ' . ' .
• Sample enumeration methods;
• Stated preference surveys; and "
• Off-line methods. .
3^3.6.1 Satisfactory Model Coefficients
i ' • - . ' f
The ultimate product from a mode choice analysis depends most directly on the value and validity
of the coefficients that are used to indicate traveler sensitivity tos pricing. An area can take one of
U. S. Environmental Protection Agency 3 53
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
several approaches if it is not comfortable with its coefficients, wishes to confirm its coefficients,
or does not have particular coefficients:
• Develop new coefficients/models: Development of a new mode choice model is a
data- and time-intensive task. Yet, in conjunction with the 1990 Census, many areas
commissioned new regional travel surveys that, taken together with Census data, serve
as an excellent opportunity for model update or upgrade. A number of references can
provide guidance to areas interested in developing new coefficients and/or models (see
NARC Manual Section 3,3.7. for information on models, or Section 3.6.2 for
information on data sources and requirements43). Areas .interested in developing new
models or greatly modifying current models should consider incorporating such
features as: nesting, income stratification, non-motorized modes, and special pricing
. variables; these topics are discussed elsewhere in this section.
,,„'' ' ' ! ' , ' 'i'1 ' ,»'•. i _
• Conduct surveys: Coefficients can be estimated from formal (cross-sectional or
longitudinal) travel surveys of revealed preference. Assuming these surveys correctly
capture the appropriate travel conditions, they may be fairly accurate in describing
travel behavior response.44 However, revealed preference surveys must be conducted in
response to actual pricing changes or signals.
Stated preference surveys, which derive estimates of response sensitivity from
reactions to carefully staged "hypothetical" pricing situations (or from a combination of
real and stated preference data), may be an acceptable alternative or complementary
approach. Stated preference surveys are discussed in more detail in section 3.3.6.8
• Transfer coefficients: An area that is not comfortable with its cost coefficients for
pricing analyses may consider transferring model coefficients from another urban area.
The disaggregate nature of logit models generally supports transferability. Analysis of
coefficients from different areas indicates that the variation among these coefficients is
not as great as might be expected, suggesting that careful transference is possible.
The most critical issue is ensuring that independent variables are consistent within the
same model, and that the " donor" model itself is the result of a well-documented
development and application process. Inflation adjustment of price coefficients (to .
reflect the year in which the original model was developed) is another major
consideration in transferring pricing coefficients (inflation adjustment is discussed
below). For example, if the original model was calibrated using 1980 prices, the new
model must either: (1) have its prices converted to 1980 dollars; or (2) have its original
cost coefficients adjusted to reflect the effect of inflation since 1980.
• Synthesize coefficients: If it is not possible to derive a particular cost sensitivity from
direct sources, it may be acceptable to synthesize a coefficient value and then perform a
series of sensitivity tests on it The model would be run for a series of parameter values
above and below the hypothesized value; the results of these simulations (mode choice,
trip length, VMT, etc.) would be charted for each run.
43 Greig Harvey et al, "A Manual of Regional Transportation Modeling Practice for Air Quality Analysis,"
National Association of Regional Councils, December 1993.
Cambridge Systematics, Inc., "Travel Model Improvement Program: Short Term Model Improvements,"
prepared for the Federal Highway Administration, U.S. Department of Transportation, 1994!
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
.. . . . •• to Reduce Transportation Emissions
Comparing the variation in results with theoretical values, coupled with judgment from
experienced practitioners, can often help identify an acceptable value for a parameter.
•Although such an analysis cannot conclusively identify the proper value for an unknown
parameter (even though carefully reasoned and conducted), it may provide a suitable
interim result until better information can be found.
• Develop specialized coefficients: Some models have different coefficients for parking
cost than for regular automobile operating cost; these may be higher by a factor of 1.5 to
2.0. This difference may reflect a higher sensitivity to parking as an out-of-pocket cost,
or it may be a function of higher density/more restricted parking supply in areas that are
priced. Parking cost coefficients may be synthesized from other models. Separate
. coefficients'for roadway tolls are not in widespread use. Both parking and toll
coefficients might be approximated initially by such methods as stated preference
surveys.
3.3.6.2 Effects of Inflation
Traveler sensitivity to a given dollar price, reflected in model coefficients, will change over time
with the general rate of inflation. As a general rule, the monetary value of a pricing policy should
be deflated to reflect the change in prices between the forecast year and the year in which the
model is calibrated. Extreme care should be taken in forecasting what the value of a price will be
in a. future year. It is commonly assumed that all travel prices will increase'in the future at the
same rate as inflation, and, thus, that there is no change in the relative price among travel choices.
This may be an acceptable assumption, though it should be noted that automobile operating costs
have generally been dropping in real terms since WWII.45 '
3.3.6.3 Income Stratification
One of the biggest issues raised in evaluating pricing strategies is how impacts vary with the
income level of the traveler. Two concerns are raised: assuring an accurate estimate of traveler
response in light of real income differences and accounting for the distribution of winners and
losers. Since mode choice is such a significant step in travel demand analysis of pricing policies
(in terms of impact and cost specificity), it is important to account directly for income differences
when applying mode choice models.
• ^ I - .
-f Generally, income differences can be accommodated by applying purpose-specific models
independently to each income stratum (usually quartiles or quintiles). It is essential that the trip
tables brought into mode choice from trip distribution already have this segmentation. Readers are
encouraged to consult both appendices for examples of how this procedure is performed. Another
It should be noted, however, that the environment for non-auto modes like walking, biking and transit, has
steadily eroded, further decreasing the utility of these modes relative to the auto which has been increasing
, in terms of vehicular and environmental design features. This suggests that the generalized cost for auto
may continue to decrease in its rate over time, le'ading to greater demand than would occur under the
assumption of constant pricing.
U. S. Environmental Protection Agency 3.55
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
way income differences can be accommodated is by developing mode choice models with
coefficients that are directly sensitive to traveler income.46
3.3.6.4 Differentiation by Purpose
Traveler response is likely to vary by trip purpose and mode choice. Coefficients valid for one
purpose may be invalid for another type of trip. Some model systems include a true mode choice
model only for home-based work (HBW) trips. Non-work trips, usually home-based other (HBO)
and non-home based (NHB) are calculated as a distance-based factor multiplied by ,the HBW
shares. This implies that the non-work mode shares move up or down in the same proportion as
the HBW mode shares, which is unlikely.
Evidence suggests that mode choice in non-work trips is more sensitive to price and less sensitive
to time than for work trips. Therefore, it is preferable to develop separate mode choice coefficients
for work and non-work trips. This is particularly important if a composite impedance measure is
developed from the mode choice model to be used in trip distribution and possibly elsewhere.
While the Census of Transportation (CTPP) has been a robust source of calibration data for work
, trips, more extensive home interview surveys (perhaps as large as 1.0 to 1.5%) may be needed to
furnish enough observations of non-work trips to permit calibration of mode choice coefficients.
The NPTS micro-sample provides one source for non-work model estimation. Montgomery
County, MD has such a model system, developed from about a 0.5% to 1.0% sample.47
3.3.6.5 Nested versus Multinomial Models
Most first-generation mode choice models have been multinomial, models, where all modes in the
alternative set are assumed to represent independent options in the model. However, when
alternatives are closely related, straight multinomial logit models can pose problems because they
assume that the alternatives represent realistic independent choices (e.g., a red bus versus a blue
bus on the same service schedule).
Probit model formulations are one way of addressing this issue, but they are highly complex and
have difficulty incorporating more than two modes. The preferred approach has become the nested
logit, which is easier and more practical to develop and apply. The nested model splits the choice
process into tiers, where the choices at any given level are relatively independent. This ensures
that two similar modes, for example, do not capture an independent share of the total, but rather
split the subtotal after the " generic" share for that type of mode is determined.
. There are many resources on logit and probit models for the interested reader (consult the NARC
Manual, Section 3.3.7, or Ben-Akiva and Lerman for references48); examples of nested logit mode
46 See M. Ben-Akiva and S. Lerman, Discrete Choice Analysis: Theory and Application to Travel Demand
MIT Press, Cambridge, MA, 1985.
> >,> | |
47 "Montgomery County, Maryland: Travel2 model deals with chaining in a form that is sensitive to land
use," Maryland National Capital Parks and Planning Commission, 1991.
45 Greig Harvey et al., "A Manual of Regional Transportation Modeling Practice for Air Quality Analysis,"
National Association of Regional Councils, December 1993.
(Footnote continued on next page...)
3~*° ' U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
, to Reduce Transportation Emissions
choice models are used in Portland, Oregon and San Francisco. The reader also is advised to )
consult the illustration in the Case Study Model in Appendix A, in which a nested logit model was
adapted and applied for mode choice. '
3.3.6.6 Special Modes
Some modes are not well accommodated by convention^mode choice procedures; this is a
function of the models as well as the data available. This is a second-order, but important, issue
for pricing because these modes are often key options. Examples include special transit services or
improvements that depart from the coded network, including auto-access/park-and-ride, and non-
motorized modes. Bike and walk modes are considered in only a small number of metropolitan
area models, with Portland being an example. • • "
Modeling bike and walk is made even more challenging because many of the trips are intra-zonal
(and hence not well represented), may not be well represented in the base data on trip rates (field
and subsequent trip generation), and require special descriptors in the models. Portland, Oregon
includes non-motorized modes in their mode choice model,49 as do Montgomery County, Maryland
and Sacramento, California.
Most areas will have to analyze these modes "off-model," usually relying on evidence from other
implementations and adjusting regional trip tables accordingly. 50 Another possible way of
representing these modes might be through sample enumeration methods, discussed below.
Another example of a "special mode" is any mode used to access transit. Access modes can
include walking, bicycling, auto travel, and transit itself. Many of the difficulties associated with
modeling transit-access modes are described above, with the additional complication that transit
access points may need to be treated as trip attractors themselves. Mode of access to transit has
been modeled in some cases as part of the nested logit mode choice model. In the mode choice
model, walk and drive access are modeled at the same level or nest. The Washington Council of
Governments (WashCOG) mode choice model is an example of transit access modeled in this
way. , '. •
3.3.6.7 Sample Enumeration Methods
Some analysts have suggested that they can realize a higher level of flexibility and realism, and
possibly accuracy, if they evaluate pricing actions through a sample of households, rather'than
M. Ben-Akiva and S. Lerman, Discrete Choice Analysis: Theory and Application to Travel Demand MIT
Press, Cambridge, MA, 1985 . /
49 Cambridge Systematics, Inc., Calthorpe Associates, and Parsons Brinckerhoff Quade & Douglas, Inc
Making the Land Use Transportation Air Quality Connection: The Pedestrian Environment Volume 4A*
for 1,000 Friends of Oregon, December 1993.
50 Delaware Valley Regional Planning Commission, "An Analysis of Potential Transportation Control
Measures for Implementation in the DVRPC Region," May 1994.
51 FY 94 Development Program for MWCOG Travel Forecasting Models; Volume A: Current Applications
June 1994 '
U. S. Environmental Protection Agency •, -„
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
work with the zonal averages of the conventional 4-step model. In this approach, called sample
enumeration, a sample of households (generally derived from a regional home interview survey,
although Census data may be a source) is used to represent the regional travel base, and becomes
the level at which the models are applied.
While this method is being introduced in connection with mode choice, in fact it may be applied to
various other steps in the travel demand hierarchy. The household sample is drawn to represent
the range of geographic, socioeconomic and travel situations that would occur in the regional
population. Generally, the sample is related (by zonal location) to the 4-step modeling process
through the regional trip tables and the corresponding travel times and costs of the transportation
alternatives.
Policies are then "applied" directly to the households; evaluation takes advantage of important
information on household composition, vehicle ownership, income, and proximity to activities and
alternatives to generate fairly robust estimates of response to the given policy. These results are
then "weighted" back to reflect the behavior of the population.
1 /
The process may or may not involve actual modification of regional trip tables and performance of
traffic assignment in estimating emissions. In the view of practitioners, this approach implies a
tradeoff between the accuracy and flexibility that can be derived from a household-level analysis
and the intricacies of a network-based traffic assignment, though traffic assignment can be
performed.
This technique has been applied in a number of locations, including California and Washington
state, and has shown good potential-52* 53» 54 It offers the additional advantage of flexibility in
analyzing certain alternatives or behaviors that can be traced at a household level, but not at a
zonal aggregation level. For example, analysis at the household level is particularly useful for
tracing short bike/walk trips, access to special alternatives, automobile ownership/fleet
composition/usage issues, and greater specificity of particular pricing issues (such as parking cash
out, etc.). The reader is encouraged to consult Appendix B, which reports on an application of this
approach.
3.3.6.8 Stated Preference Surveys
ii ' '
Stated preference surveys represent an interesting option for gaining insight into mode choice
response to policies, alternatives, and situations when solid empirical information does not exist.55
52 Greig Harvey, Curbing Gridlock: Peak Period Fees to Relieve Congestion, Transportation Research
Board Special Report 242, Vol. 2, National Academy Press, Washington, D.C., 1994.
53 Greig Harvey, "Transportation Pricing and Travel Behavior," presented at 69th Annual Western Economic
Association International Conference, June 1994.
Greig Harvey et al., "Transportation Pricing for California: An Assessment of the Air Quality, Congestion,
Energy, and Equity Impacts," Vol. 1: Summary Report, Draft, California Air Resources Board, June 1995.
Cambridge Systematics, Inc. and Barton-Aschman Associates, Travel Model Improvement Program: Short
Term Travel Model Improvements, U.S. Department of Transportation, U.S. Environmental Protection
Agency, U.S. Department of Energy, October 1994, pp. 1-7,1-8.
3-58 ' if. S. Environmental Protection Agency
54
55
-------
Technical Methods for Analyzing Pricing Measures
• "" to Reduce Transportation Emissions
Stated preference surveys might be used, for example, because no data has yet been generated on a
new type of travel mode or a special type of pricing instrument with unique characteristics.
In a stated preference approach, it is possible to derive statistical estimates of,"tradeoff rates
between various alternatives or their attributes by! making respondents choose from among them in
measured ways that indicate the relative importance of key attributes. These rates can then be
evaluated in relation to the type of traveler and his/her circumstance.
The validity of the derived statistical relationships relies on how well the alternatives are portrayed
to (and understood by) the respondent, and itheir comparison to known " standards." While stated
preference surveys rely on hypothetical situations, comparison of "elasticity" relationships
derived from stated preference with more conventional revealed preference surveys or models
have shown surprising corroboration. The results from these surveys should be used with caution,
but they offer'an important interim tool for agencies to estimate relationships between pricing
instruments and travel behavior response, not just in mode choice but in relation to destination,
time of day, route choice, etc.
Stated preference methods were developed by the private market research industry, and have been
used successfully for many years to aid companies in identifying the critical attributes of their.
product, and maximizing those attributes to gain market share over competitors. Use of the
techniques in transportation is a fairly recent development, but there are numerous examples of its
successful application-56, 57, 58, 59, 60 Portland and Denver have both utilized stated preference
methods in relation to pricing. Portland has used stated preference methods to explore time of day
.choice^ conducting its survey in conjunction with its recent home interview'survey, .while Denver
used such a survey to assist in development of a route choice model (described under traffic
. assignment, below).
3.3.6.9 Off-Line Methods
A number of planning agencies may have limited ability to perform mode choice analysis, either
because they do not have mode choice models or the models they have are insufficient for
evaluation of the particular measure. The NARC Manual cites a number of "off-line" methods
that may be valuable as substitute or complementary approaches. Examples include:
• TDM Evaluation Model: The Federal Highway Administration is distributing a special
software product that can substitute for or enhance a conventional mode choice model.
M. Ben-Akiva and S. Lerman, Discrete Choice Analysis: Theory and Application to Travel Demand
MIT Press, Cambridge, Massachusetts, 1985. , ;'
57 M. Ben-Akiva et al., "Combining Revealed and Stated-Preference Data," prepared for publication in
Marketing Better. .
Cambridge Systematics, Inc. and Hague Consulting Group, VFT Feasibility Study: Market Analysis Final
Report, July 1988. , ' ' * > .
David A. Hensher, P. Truong, and P.O. Barnard, "The Role of Stated Preference Methods in Studies of
Travel Choice," Journal of Transport Economics and Policy; January 1988, pp. 45-57.
J. Bates, "Econometric Issues in Stated Preference Analysis," Journal of Transport Economics and
Policy, January 1988,
U: S. Environmental Protection Agency
58
59
60
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
The TDM Model is a spreadsheet approach that incorporates a "pivot-point" procedure
to estimate mode choice changes. The model forecasts the change from an initial mode
share based on information regarding changes in decision variables associated with the
policy action. A wide variety of pricing and other strategies can be considered,
individually or grouped into programs, with great flexibility to vary .the levels of the
strategies and target their application.
The starting base may be trip tables from an existing planning process, special trip
tables formed from survey data, or simply aggregate estimates of person, vehicle and/or
transit trips. The coefficients in this model have been "synthesized" from national
experience and can be altered by the user to use local or other coefficient estimates.
Results are both aggregate and in trip table format. The trip table results can be returned
to the four-step process (if applicable) for traffic assignment and emissions!6'
The TDM Model has special capabilities for trip table manipulation or creation. The
model interfaces directly with the common planning software packages (TRANPLAN,
MinUTP and EMME/2), enabling easy interactions with the core model system. The
TDM Model allows analysts to alter starting trip tables, or create a trip table from
scratch (e.g., for transit or vanpool trips). Trip tables may be created entirely from
scratch using survey data, if desired, or more simplistic estimates of total tripmaking
and starting mode shares may be used to fuel the model. For areas with vehicle-only
trip tables, it may be feasible to use the model "in reverse," knowing average
occupancy rate, to create a person trip table suitable for performing mode choice
analysis. (Unfortunately, the TDM model does not incorporate non-motorized modes,
and this should be adjusted for in its application.)
Quick Response System: The quick response system procedure (QRS) consists of a
collection of short-cut, parametric methods that may be used for transportation analysis
in the absence of a local model system. Procedures exist to cover the steps of trip
generation through traffic assignment, and include mode choice. It exists in either
microcomputer or manual form, and allows users with minimum data and computer
capability to form reasonable estimates of travel response.62;63
Elasticity Factoring Methods: Agencies without access to computerized planning
methods or to either of the above techniques may consider using factoring methods to
derive estimates of mode choice responses to pricing strategies. Such an analysis
requires the following travel estimates:
- Person trips
- Vehicle trips (or average occupancy rate)
7 Transit trips (if applicable)
- Vehicle miles of travel
61 COMSIS Corporation,' Users Guide: Travel Demand Evaluation Model, for the Federal Highway
Administration, 1993 (For information contact McTRANS, distributing agent for FHWA at 904-392-0378).
62 COMSIS Corporation, "NCHRP Report 187: Quick Response Urban Travel Estimation Techniques and
Transferable Parameters," 1978.
63 COMSIS Corporation, ORS User's Manual, for the Federal Highway Administration, January 1984.
3-60 . U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Ideally, these estimates would be in the form of trip tables, separated by'trip purpose
(work, non-work). Given information on trip length for each origin-destination trip
combination and starting modal shares, analysts should be able to estimate the change in
mode shares triggered by the respective pricing action.
. The report " Transportation Air Quality Analysis Sketch Planning Methods" M provides
a convenient spreadsheet method for performing this analysis. The reader may also
consult various sources for estimates of elasticities, but should be careful to choose a
value or values appropriate to local conditions.65* 66>61
In the event that the agency does not have trip tables, it may be acceptable to estimate
policy responses from a more aggregate estimate of regional trip making. Baseline
estimates can come from: , .
- Local travel survey data; , ;
- Census CTPP estimates; '
- Population or activity estimates factored by measured or ITE vehicle trip
rates; and
- HPMS estimates.
Obviously, the more coarse the background estimate, the less accurate the estimated
response and the more simplistic must be the strategy definition. The above references
are recommended as sources for elasticities and application methods.,
3.3.7 Time of Day
The ability of a model system to simulate differences in travel volumes and conditions at different
times of the day is critical to realistic estimates of speeds, congestion levels, and emissions,
particularly in larger urban areas with pronounced travel peaks. Yet, time of day allocation of
travel is done in a very approximate way, using peaking factors derived from local surveys to
apportion a certain percentage of daily trips into one or more peak hours or periods.
Time of day is an important dimension for many TCMs, since changes in the level of congestion
and time distribution of travel affect important calculations regarding speeds and travel volumes,
which fuel the emissions model. Time of day takes on special importance in relation to pricing
measures that vary by time of day, for example, roadway congestion pricing. With major price
differentials between peak and off-peak, it is likely that travelers who are sensitive to cost and/or
who have flexibility in trip timing will consider shifting their departure to another time period;
alternatively, persons without flexibility and/or with a high value of time might pay a Higher price
to travel in the peak (and possibly even enjoy a higher level of service).
64 Cambridge Systematics, Transportation Air Quality Analysis Sketch Planning Methods, Volume I:. Analysis
Methods, prepared for the U.S. Environmental Protection Agency, 1979.
COMSIS Corporation, Users Guide: Travel Demand Evaluation Model, for the Federal Highway
Administration, 1993 (For information contact McTRANS, distributing agent for FHWA at 904-392-0378).
R.H: Pratt Associates, Traveler Response to Transportation System Changes, prepared for the Federal
Highway Administration, Second Edition, 1981. ,
Apogee Research, Inc., Costs and Effectiveness of Transportation Control Measures: A Review and
Analysis of the Literature, prepared for the National Association of Regional Councils, January 1994.
. U. S. Environmental Protection Agency ' 3-61
65
66
67
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
This is a fairly complex choice that is not well handled by conventional peaking factor approaches.
Research is underway in this area; sites interested in time of day forecasting (because of their
interest in congestion pricing types of measures) might investigate the following activities or
options:
• Approximating methods;
• Time of day choice models;
• Time of day factors; and
• Stated preference surveys
" ' ' ' i
3.3.7.1 Approximating Methods
Current travel models do not attempt to accommodate potential traveler response to pricing
measures that vary by time of day. Such analysis requires procedures that are sensitive to travel by
time period, which in turn would require input data on price by period. This type of information
would be cumbersome to create, but would not be outside the reach of current software.
"' , ' ' i
For example, consider a model to estimate the percent of daily trips that begin in the morning peak
three hours, which could be applied in two phases. The first phase would calculate one percentage
split for all O/D pairs, separately by trip purpose. The second phase would consider the cost of
making each trip during the morning peak and during the off-peak, and then adjust the initial
percentage split based on the peak/off-peak cost differential for that O/D pair.
Non-work trips would be expected to exhibit greater sensitivity to time of day price differentials
than work trips because of greater traveler flexibility. The exact sensitivity of trip start times to
price differentials is speculative in light of current knowledge, but it should be possible to make a
reasoned estimate using stated preference methods (see discussion below), which can be refined
later using observed data.
3.3.7.2 Time of Day Choice Models
Domestic modeling practices have not featured formal modeling of time of day as a travel choice.
Several researchers have devoted particular attention to this element and performed work that may
serve as a basis for agencies that wish to advance their capability.68>69
3.3.7.3 Time of Day Factors
1 „ ' », , i
If trip generation models can be made to include a measure ofaccessibility (composite impedance)
that incorporates cosfand travel conditions at different times of day, it may be possible to reflect
time of day preferences through the trip rates themselves, prior to trip distribution and mode
68 Kenneth A. Small, "Fundamentals of Pure and Applied Economics," Urban Transportation Economics,
Vol. 51, Harwood Academic Publishers, 1993.
69 P.R. Stopher and A.H. Meyburg, Urban Transportation Modeling and Planning, Lexington Books, 1975.
3-62 , u_ S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
" . to Reduce Transportation Emissions
choice.70 If overly simple measures of accessibility are used, however, the value of such an
approach would be minimal. It is also possible to develop separate trip generation models by trip
purpose by time of day, followed by separate trip distribution and mode choice analysis; this has
been done by Montgomery County, Maryland.
3.3.7.4 Stated Preference Surveys
Traveler response to time of day pricing/service differences may be approximated through stated
preference surveys. Sites also might consider synthesizing time of day relationships from other
travel contexts that employed differential time of day pricing, such as bridge tolls or peak/off-peak
transit fare differentials. ' •
3.3.8 Traffic Assignment/Route Choice
The traffic assignment step estimates the traffic volumes for each link in the transportation
network, for each time period that is considered. Traffic assignment is critical to the transportation
analysis and emissions process; it produces the estimates of link volumes, speeds/travel times, and
vehicle miles of travel (VMT) that are key inputs to earlier planning steps, particularly trip
distribution and mode choice, and are the primary determinants of emissions when input to the
emissions factor model.
Traffic assignment consists of assigning trips from a trip table to a transportation network
according to specified criteria, almost exclusively the minimum travel time path. Travel cost is
generally not included as a criterion, although some models account for tolls.71 It is feasible to
include cost in assignment, and therefore extend its effect more broadly through, the other steps of
the modeling process (see discussion below).
The details of the traffic assignment process are not of central concern in this manual. It is clear,
however, that the accuracy of travel speeds generated by traffic assignment are of key importance:
First, travel speeds are critical to effective analysis of all transportation alternatives, including
those affected by pricing mechanisms. Second, travel speeds are a direct input to emissions
analysis. ' ,
Additionally, there is concern about how well conventional assignment models deal with the issue
of route choice when a roadway pricing system creates competing Opportunities.
Techniques in use or under development that may be of interest in addressing these issues include:
• Incremental loading and equilibrium assignment methods; , , "
• -Dynamic assignment; . •
70
Greig Harvey et aL, "A Manual of Regional Transportation Modeling Practice for Air Quality Analysis,"
National Association of Regional Councils, December 1993, p. 3-58.
Cambridge Systematics, Inc. and Barton-Aschman Associates, " Travel Model Improvement Program:
Short Term Travel Model Improvements," prepared for U.S. Department of Transportation,' U.S.'
Environmental Protection Agency, U.S. Department of Energy, October 1994, Sec. 4.0, p 4.1.
• £/. S. Environmental Protection Agency ' 3.53
-------
Technical Methods for Analyzing Pricing Measures '
to Reduce Transportation Emissions
• Traffic microsimulation;
• Speed feedback and recursion; and
• Path choice.
3.3.8.1 Incremental Loading and Equilibrium Assignment Methods72
Incremental loading and equilibrium assignment methods are largely replacing the all-or-nothing
methods and provide much more realism in allocation of trips to paths. Among other things, these
new methods greatly increase the realism with which level of service variations are represented
and, in principle, provide a means for incorporating the cost of highway travel as -well as travel
time, which can then be used in the earlier steps of the modeling process.
3.3.8.2 Dynamic Assignment73
Dynamic assignment, which is an improvement on conventional "static" assignment procedures,
addresses problems with demand variations within an analysis time period and begins to account
for queuing impacts on speed.
3.3.8.3 Traffic Microsimulation74'75'76
The outputs of a typical traffic assignment analysis reflect average link speeds, and do not
document variations in those speeds that would result from queuing at intersections, entrance
ramps, or under congested stop/go conditions. Such variations in speed can significantly affect
vehicular emissions; valid estimation can be critical to proper computation of emissions. Research
is ongoing in this area to improve the linkage between the assignment models and emissions
analysis through use of traffic microsimulation techniques. The current NCHRP Project No. 25-7:
"Improving Transportation Data for Mobile Source Emissions Estimates," is developing emission
profiles that can be linked to a microsimulation to account for these effects.
72 Greig Harvey et al, "A Manual of Regional Transportation Modeling Practice for Air Quality Analysis,"
National Association of Regional Councils; December 1993, Section 3.3.9, p. 3-66.
Cambridge Systematics, Inc. and Barton-Aschman Associates, "Travel Model Improvement Program:
Short Term Travel Model Improvements," prepared for the U.S. Department of Transportation, U.S.
Environmental Protection Agency, U.S. Department of Energy, October 1994, Section 4.0, pp. 4-1 to 4-7.
Cambridge Systematics, Inc. and Barton-Aschman Associates, "Travel Model Improvement Program:
Short Term Travel Model Improvements," prepared for the U.S. Department of Transportation, U.S.
Environmental Protection Agency, U.S. Department of Energy, October 1994, Sec. 5.3, pp. 5-4 to 5-7.
75 Patrick DeCorla-Souza et al., "A Trip-Based Approach to Estimate Emissions with EPA's MOBILE
Model," Paper No. 940376, Transportation Research Board, 1994.
76 E. Ruiter, "Highway Vehicle Speed Estimation Procedures," prepared for the U.S. Environmental
Protection Agency, 1991.
3-64 . . U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
3.3.8.4 Speed Feedback and Recursion77
Modeling accuracy would benefit from a formal recursive procedure that cycles information from
assignment back to the earlier steps, thus ensuring compatible values among the steps when
equilibrium is reached. If this is not done the inconsistency between the assumptions that are
inputs .to different modeling steps raises serious questions about the .reliability of the final outputs.
If pricing is introduced as a factor in more of the modeling steps through a measure like composite
impedance, then it. becomes even more important to ensure passage of this information through the
; relevant steps. Methods are currently under development by the Federal Highway Administration
in conjunction with the Travel Model Improvement Program, and are in use in various places.78
3.3.8.5 Path Choice .
Few conventional models are able to adequately model the effect of the cost of driving on route
choice, which becomes a particular issue in evaluating roadway pricing. A common procedure is
to convert the toll value to an equivalent number of extra minutes by dividing it by some value of
travel time. Thus, when a route is considered as a potential path, the effect of the toll would be
reflected in an increased travel time. This may be acceptable as a default analysis, but it greatly
clouds the issue of choice discrimination among classes of travelers, some of whom would
recognize the toll as a signal to avoid the facility and some of whom might be expected to pay the
toll in exchange for a higher level of service (shorter travel time).
Toll diversion models are a more realistic way to account for route choice responses to pricing
measures. Toll diversion models split vehicle trips for each O/D pair into "toll" and "free" trips.
This split is based on a comparison of toll and free path characteristics—mainly travel time and
cost (at least one model also includes "number of toll stops" as a variable). The toll path is the
one that would be taken if the trip were forced to use the toll facility and the free path is the one
that would be taken if the trip were prohibited from using the toll facility. Popular planning
software packages such as TRANPLAN, MinUTP, and EMME2 are able to create and skim these
paths. , "• •
Simplified versions of these models have been used for years by toll road planners, most often by
applying manual adjustments (based on rough estimates of travel time savings) to estimated toll
road traffic volumes. The, process has been implemented within a computerized traffic forecasting
model for use in planning a toll road in the Denver region.79
This same procedure can be adapted to analysis of priced roadways. Unfortunately, this method is
subject to the same limitation as methods to deal with the mode choice and time of day effects of
78
77 Greig Harvey et al. "A Manual of Regional Transportation Modeling Practice for Air Quality Analysis,"
National Association of Regional Councils, December 1993, Sec. 3.3.9, pp. 3-70, 71.
8 Montgomery County, Maryland has documented the effects of incorporating full feedback in a composite
equilibrium trip distribution/mode choice/network assignment model (Levinson & Kuman, 1992)
Incorporating Feedback in Travel Forecasting: Methods, .Pitfalls, and Common Concerns, Travel Model
Improvement Program, March 1996. ; '
79 J. Heisler et al., "Estimating Toll Diversion Using Existing Transportation Planning Software," presented
at the Second Conference on the Application of Transportation Planning Methods, April 1989. "
U. S. Environmental Protection Agency . .'•''< 3-65
-------
Technical Methods for Analyzing Pricing Measures *
to Reduce Transportation Emissions
congestion pricing: lack of observed data with which to calibrate a model. The Denver model is
based on a small-sample stated preference survey of Denver area residents, most of whom had
never used a toll road. The model has undergone extensive reasonableness checks and its results
have been accepted as plausible.
The reader is referred to the case study model application detailed in Appendix A for an example
of this procedure.
3.3.9 Emissions Modeling
The final step in the analysis process is the translation of the transportation effects estimated in the
other analytic steps into quantities of the respective mobile source pollutants: VOC, NOX and CO,
and PM. Areas with comprehensive regional model systems generally perform this step by
applying EPA's MOBILESa model or, in California, the EMFAC7F model, and PART5 for
reentrained dust emissions. These models accept input from the traffic assignment process on
vehicle trip ends, VMT and speeds, sorted by facility type, geography and time of day. These
inputs are then translated into emissions through the application of emissions factors, which are
based on the regional vehicle fleet mix, average annual distance traveled by class, average travel
speeds, ambient temperatures, inspection and maintenance programs, fuel policies and other
factors.
This manual does not detail the development and specific operating features of the MOBILE or
EMFAC models or the PART 5 model. Readers who desire more information can consult existing
reference documents on these models.80*81*82 What is of particular concern here is the effect that
particular pricing actions may have on the vehicle mix that is included in the models. Pricing
measures such as fuel taxes and emissions fees would have their principal effect on changes in the
composition of the regional fleet. It is this change in vehicle mix, separate from (or in addition to)
any change in travel level, that would alter emissions levels. Section 3.2.2.3 addressed issues
associated with changes in vehicle mix and suggested methodologies and sources for estimating
these changes.
Areas that analyze fuel price or registration fee measures should use their analysis of the probable
changes in regional vehicle, fleet mix to alter the base distribution in MOBILE (or EMFAC). The
Case Study Model application in Appendix A presents an example of how this can be done.
Users should recognize that changes in vehicle mix may be the compounded result of two shifts:
• Vehicles may shift in proportion by model year; and
80 Sierra Research, Inc. and Jack Faucett Associates, "Evaluation of MOBILE Vehicle Emission Model,"
prepared for Volpe National Transportation Systems Center, U.S. Department of Transportation, December
1994.
81 California Air Resources Board, "Methodology for Estimating Emissions from On-Road Motor Vehicles
Volume I: EMFAC7F," Draft Report, June 1993.
32 U.S. Environmental Protection Agency, Office of Mobile Sources, "Draft User's Guide to PART5: A
Program for Calculating Particle Emissions from Motor Vehicles," February 1995 (EPA-AA-AQAB-94-2).
1 3-66 , , U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
. . , to Reduce Transportation Emissions
• Vehicles may also shift by relative use (VMT).
Historical trends, such as reflected in the Nationwide Personal Transportation Survey arid the
Motor Vehicle Manufacturers Association statistics, indicate a tendency to use new cars more than
older ones; this is reflected in the annual VMT assigned to each vehicle age-category in the
analysis. Thus, an analytic process that re-apportions fleet rnix by age (to favor newer vehicles),
yet maintains mileage rates (VMT) in each age-stratum at their prior levels implicitly assumes that
a shift to newer vehicles means more VMT each year — an unlikely result. Analytically, if not
corrected, this would result in newer, cleaner models in the MOBILE distribution being given an
unrealistically high VMT weighting.
Unless some type qf analysis or survey research is performed to support adjustments by year and
mileage, users should normalize the distribution of mileages to account for this effect. In the
example below, the ratio of pre-shift [VMT*share] to post-shift [unadjusted VMT* share] is 0.97
(i.e., 6.69/6.89). Adjusted VMT is obtained by applying this ratio to the original VMT figures for
each age stratum.
i&Efcunpte of VMT Adjitstnieirt
Age
(yrs)
1
2
3
4
5 •
Pet of
Vehicles
15
12
10
8
5
' 50%
Annual
VMT
000
15
14
13
12
10
Revised
Veh. Pet
20
13
9
5
3
50%
Effect.
VMT
000
15
14
13
12
10
Adjusted
VMT (*)
000
14.6
13,6
.12.6
11.6
9.7 :
Sum Original Veh. Dist. x Ann VMT = 6.69k;
Sum Revised Veh. Dist. x Ann VMT = 6.89k
Adjustment = 6.69/6.89 = 0.97
U. S. Environmental Protection Agency
3-67
-------
-------
Technical Methods for- Analyzing Pi-icing Measures'
to Reduce Transportation Emissions
4.0 Implementation Issues
4.1 Introduction
The ability of market-based measures to achieve their intended objectives depends on:
(1) measures that are- technically effective; and (2) an implementation program that is
administratively and politically practicable. Applying sound analysis to the estimation of travel
and emission impacts of market-based measures is the necessary first step in receiving emissions
reduction credit for a proposed program. Submitting agencies "should also provide complementary
documentation to assure reviewing agencies that reasonable progress can be expected in program
implementation.
In short, agencies proposing to incorporate and receive credit for market-based measures should
address the implementation issues outlined in this section to receive EPA approval. The
significance of specific implementation issues will vary from one area to another, and from one
measure to another; this chapter presents generic guidelines that may hot apply in some instances.
For example, this section notes that any needed technology must be shown to be available, yet.
some measures require no technology deployment at all.
"Implementation" covers a broad array of qoncerns. It includes substantive issues such as legal
authority and administrative capacity. It also includes more " politically-oriented" issues; clearly,
the method and form in which market-based measures are packaged and introduced influences
public acceptance and ultimate program success. A sampling of implementation questions might
include:
• What organization will bear responsibility for ^implementation, administration and
enforcement?
• What other measures will be combined with the subject measure? Would these other
measures offer attractive and meaningful choices? ,
• How significant are technical and financial uncertainties? How would public acceptance
and program goals be affected if a "partial" program were implemented?
• How are the resulting revenues allocated — back to complementary transportation
services, or used for other societal purposes?. Should fees be set so revenues equal the
cost of these complementary uses/services, .that is, a revenue-neutral policy?
• How will equity issues be handled?
The way in which issues such as these are addressed will determine: (1) how the measure .itself is
defined and its impact quantified; and (2) the likelihood that the measure will actually be
implemented as proposed. In turn, these factors will influence the emissions credit determination.
U. S. Environmental Protection Agency . 4_j
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
The Clean Air Act requires EPA to recognize and approve SIP credit for any quantifiable,
permanent, and enforceable emissions reduction; it does not give EPA authority to advise on
implementation issues that do not"(directly) affect air quality. Thus, although all implementation
issues contribute to ultimate program success, this section discusses only those implementation
issues that affect an implementing body's ability to produce predicted air quality benefits. Several
research institutions are available to advise cities and states on the non-air quality aspects of
implementing market-based measures. EPA and the Department of Transportation also are able to
provide research and related assistance.1
Experience with implementation of market-based measures is growing but is still limited. EPA is
committed to working with states and their agencies to resolve any implementation uncertainty and
to respond to reasonable assumptions where experience has not yet provided a sure guide. This
chapter of the Guidance summarizes EPA's position on supporting documentation to address the
following types of implementation issues:
• Public policy context;
• Legal authority;
• Administrative capacity;
• Technology availability; and
• Revenue generation and reinvestment.
1 Much of the information in this chapter.has been summarized from Special Report 242, Volume 2, Curbing
Gridlock: Peak-Period Fees to Relieve Traffic Congestion, Transportation Research Board, Washington
D.C., 1994.
4-2
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
4.2 EPA Position on Documentation
States and their subsidiary agencies and implementing authorities requesting emissions credit for
market-based measures should provide a narrative summary of critical implementation issues
along with their technical analysis of travel and emissions impacts.
While there are no standard analytical approaches for examining implementation issues, EPA
needs enough information to (1) be assured that the intended market-based measures can be
implemented over the timetable proposed and (2) understand uncertainties that may affect the
implementation timetable and total emissions reductions.
The discussion of implementation issues identified below should, at a minimum, describe:
• Current circumstances likely to affect implementation in each issue area;
• Actions currently underway to advance implementation of proposed measures;
• Planned or proposed future actions, with projected timetables; and
• Implementation issues as yet unresolved, remaining uncertainties, and their potential
implications for the implementation of proposed measures.
In each case, the submitting agency should include relevant empirical or quantitative information
or reference the appropriate sources. In short, if pricing measures are to be credible and creditable,
sponsors should provide reasonable evidence that the actions underway will lead to
implementation within the specified timetable.
4.2.1 Public Policy Context
Like any transportation or emissions policy, successful implementation of proposed market-based
measures will depend, in part, on how well they can be introduced and integrated into the city's or
state's general priorities. It is important to acknowledge that public relations and public
participation may be as influential as any technical analysis in the implementation of pricing
mechanisms. F6r instance, the current consumer costs of travel may be widely perceived as the
actual market costs. An increase in consumer costs would consequently be regarded in a negative
light, rather than as a movement toward a more equitable cost structure. Therefore, educating and
involving the public is a very important element of any pricing strategy.
Sponsors should briefly describe the public policy setting in their region, priorities among
competing objectives, and the extent to which implementation,of market-based measures has been
integrated with major public .policy objectives. Sponsors also should address issues related to
U. S. Environmental Protection Agency 4.3
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
public involvement and education regarding planned implementation of market-based
mechanisms.2
4.2.2 Legal Authority
Because this class of measures is new to many areas, and may involve multiple agencies, questions
of legal authority may be raised that were not raised for other kinds of emissions-control strategies.
Local units of government sponsoring, endorsing, or implementing market-based measures should
document that they have the full legal power to do so, based on statutory or delegated authority.
The credibility of a measure depends on the availability of such authority. Sponsors should review
and confirm legal authority to ensure that they are able to enact and enforce the proposed market-
based measures.
Sponsoring agencies should answer the following questions:3
• What is the nature and source of the legal authority?
• Are the proposed measures a reasonable exercise of this authority?
• Are the revenue-raising provisions and pricing structures consistent with existing
taxing authority? '
If any measures affect multiple jurisdictions, sponsors should detail the mechanisms to be used
and the basis for crafting agreements among multiple units of government.
If implementation requires legislative action, sponsors must detail the process, prospects and
timetable for any required legislative action to establish or clarify legal authority.
4.2.3 Administrative Capacity
In addition to legal authority, sponsoring agencies should demonstrate that capacity exists to
effectively administer all aspects of the proposed measure.
Sponsors of market-based measures should describe the administration that will support each
measure, noting areas in which increased competence or capability may be needed and outlining
the approach, and timetable for making necessary improvements. Depending on the measure,
implementation will require adequacy in some or all of the following areas:
• Staffing levels and skills;
• Computer resources;
For additional discussion of issues related to public involvement, public acceptance, and equity, see:
USEPA, Opportunities to Improve Air Quality through Transportation Pricing Programs(EPA 420-R-97-
004), September 1997. This document is also available at the following world wide web address:
http://www.epa.gov/OMSWWW/gopher/Market/pricing.pdf
See articles by Olson and by Pietrzyk in Transportation Research Board Special Report 242, Vol. 2, 1994.
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to, Reduce transportation Emissions
• Budget support;
• Audit capability; , . . ' ' •
• Accounting capability;
• Ability to process revenues;
• Contracting capability; and
- • Leasing and procurement capability.4
4.2.4 Technology Availability
Some measures, such as congestion pricing, require technology deployment. In these cases,
sponsors should do the following in their submission:
• Document access to the required technology, either available currently off-the-shelf
or available in the future in time for scheduled implementation;
• Evaluate the technologies to be deployed to ensure that they (1) are capable of
achieving the objectives established for each measure, and (2) operate effectively
and efficiently;
• Summarize the results of technical analyses and assessments that have been carried
but for the proposed measures; and
• Note constraints, areas of uncertainty, and areas needing further development in the
deployment of new technologies. • • / . '
Assessments for congestion pricing, for example, would cover the following technology areas: in-
. vehicle units, vehicle to roadside communication systems, vehicle detection and classification
systems, and payment and accounting systems.
4.2.5 Revenue Generation and Reinvestment
In a number of cases, pricing measures generate new revenue. This revenue could be used in two
basic ways: (1) to reduce other fees or taxes; or (2) to invest in public programs, particularly in
transportation. In the case, of fee or tax offset, sponsors do not need to detail revenue use.
In general, sponsors should describe:
• Anticipated programs for which new revenues will be used;
• 'Anticipated impacts of those inyestments on travel behavior and emissions;
• Analysis used to estimate travel and emissions impacts, including consideration of
synergism and overlaps;
i • i
See Olson in Transportation Research Board Special Report 242 Vol. 2, 1994.
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
* Act'ons anc* timetables anticipated in implementing the reinvestment strategy and
programs; and
• Anticipated difficulties or uncertainties in executing the reinvestment strategy and
how they will be overcome. .
If revenues are to be reinvested in any way that affects transportation demand or emissions,
sponsors should detail demand and the final emissions impacts. Reinvestments with such impacts
include: '
• Investment in transit facilities: generates shifts in mode share and reductions in
Vehicle trips and VMT;
'": , ••' '•,•'<• ' ' • ' i
• Investment in traffic flow improvements or preferential facilities for high-occupancy
vehicles: generate increases in travel speed and shifts in mode share;
• Revenue used to support transit fares: produce shifts in mode share, reductions in
Vehicle trips and VMT;
• Investment in vehicle buy-back programs: induce changes in fleet mix and emission
characteristics;
• Financial support of local government and business trip reduction programs:
support reductions in vehicle trips and VMT; and -
• Enhancement of pedestrian and bicycle environment, or non-motorized access to
public transportation: support reductions in vehicle trips and VMT.
» • ' • • . • j .
Many market-based measures gain effectiveness when supported by non-market measures, for
example, expanding the diversity of transportation choices through increased transit options or
enhanced pedestrian environment. Land use strategies may likewise be supportive. These are
mentioned here because some may be potential recipients of program-generated revenue.
Emissions effects can be established by iterating through the modeling process described in earlier
sections, adjusting the processes and values for effects of the reinvestment, where they can be
determined or estimated. Alternatively, effects can be estimated independently.
If planned reinvestment — whether in fixed capital, or operations and maintenance — is expected
to generate these types of impacts at a significant level, additional iterations in the modeling
analysis may be required to capture the full effects and claim full emissions credit.
'*"" U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Appendix A
Modeling Pricing Measures in a Traditional Modeling Environment
A Case Study Approach
A.1 Introduction
Almost all existing travel models were developed not for policy analysis, but for planning major
transportation facility improvements: One of the unintended consequences of this is that most
models are not well suited for analyzing the full range of travel impacts of different transport
pricing strategies. Although these models are generally not structured for assessing the impacts of
such public and private policy actions, they still represent the best available data base for travel
forecasting within most urban areas. Thus, there is a strong need either to modify existing models
or develop new models that can meet a wider range of analysis requirements.
It was not the goal of this research effort to develop and calibrate a new model set, because that
would have required time and data resources that were not available. Instead, this project took the
approach that many good models exist in different urban areas and that it should be possible to
create a model set that could serve the study's purposes by " borrowing". appropriate parts from
various models and assembling them into one model set that would exhibit reasonable sensitivities
towards pricing measures. A side benefit of this approach is that the resulting model would not be
biased towards any one urban area, but would represent a composite picture of a variety of areas.
The model would, Jiowever, b& applied and tested using actual data from one metropolitan area, in
order to produce more realistic results than an analysis using a hypothetical urban area. The
. resulting model set is referred to as the Case Study Model, because it,was used to analyze the
different market-based case studies for this project. This model is described further in the next
section.
It should be noted that the Case Study Model is not intended to be one that can be picked up and
applied "as is" to any particular area for use in analyzing market-based TCMs. It was only
intended for use in this national-level overview analysis of the potential emission reductions of
market mechanism;?. However, the model documentation is presented here as part of the technical
guidance for urban transportationTjlanning staffs who may seek to adapt their local models to better
analyze pricing strategies. As such, the following is intended to assist such staffs to develop and
use similar modeling advances so as to conduct more credible forecasts of the effects of market
mechanisms in their areas. ',
A.1.1 Case Study Model Specification
As noted above, the Case Study Model set is adapted from various four-step travel forecasting
models from urban areas across the U.S. Despite some of the problems associated with the four-'
step process, it is still the most widely used and readily understood modeling approach. It was an
explicit premise of this project that most of the problems with the four-step process are related to
•••^ , -_ •• ' * ;
U. S. Environmental Protection Agency • ' ', , ^_j
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions , ' '
the way in which the steps are applied and other deficiencies in the input data, rather than flaws
that are inherent in the process itself. The project's researchers hypothesized that an enhanced
model set could be developed that would satisfy the need for specific sensitivity to pricing
measures within the context of the four-step process.
Figure A-l is a flowchart depicting the overall structure of the Case Study Model set. The
following sections describe how each component of the model was crafted., with emphasis on the
nature of the sensitivity to pricing. It bears emphasis that although each component of the Case
Study Model set might not be considered the most advanced " state of the art" technique, the way
they are combined, as shown in Figure A-l, is unique and is the most important feature of the
model set.
The Case Study Model is applied to data representing 1996 conditions in the Washington, D.C.
metropolitan area. This city and scenario were selected because the data were readily available,
courtesy of the Metropolitan Washington Council of Governments (MWCOG). However, neither
the Case Study Model nor any results of this analysis are endorsed or supported by MWCOG. It
also bears emphasis that this model has not been fully calibrated to any set of data and cannot be
said to represent the specific sensitivity of Washington area travelers (or any other travelers) to
pricing. Its purpose is to demonstrate the structure of a pricing-sensitive model and to produce
results that are reasonably illustrative of the range of emission reductions that might be associated
with the assumed levels of transport pricing changes. Neither EPA nor the authors warrant that the
results estimated by this model would be achieved in actual practice.
A.1.2 Submodels
The Case Study Model set includes two important submodels: households by vehicles and size, and
parking cost.
Households by Vehicles and Size Submodel: Most existing trip generation and distribution
models are handicapped in many cases by working with zonal average data. One of the most
substantial modeling improvements in recent years has been the. increasingly popular practice of
subdividing certain items of zonal data by those characteristics which influence travel. A common
practice is to split households jointly by size and vehicles available (or household income).
Washington, D.C., Detroit, Philadelphia, San Francisco, and Portland (Ore.) are examples of urban
areas which use vehicles, while New Orleans, Atlanta, Northern New Jersey, Dallas, Denver,
Phoenix, and Minneapolis-St. Paul use income. This produces " cells" of households which are
more homogeneous in nature, reducing the variance in trip rates. Using income or vehicles as one
of the dimensions also allows the explicit identification of trips by some measure of wealth, which
can be useful for modeling the differential effect of pricing on the different groups.
The submodel operates by using Census data to establish the relationship between zonal averages
and the distribution of households by integer values. The size submodel splits households by
number of persons: 1, 2, 3, 4, and 5+, and the vehicle submodel splits households by number of
vehicles: 0, 1, 2+. For the Case Study Model, data on households by vehicles available were
provided by MWCOG, and so' only a size submodel was needed. Figure A-2 presents the size
submodel.
A-2 , ' (7. S. Environmental Protectipn Agency
-------
Technical Methods for Analyzing Pricing Measures
, to Reduce Transportation Emissions
Once the individual distributions of households by size and by vehicles are known for a zone, the
region's joint distribution of households by size and vehicles is used to estimate the joint
distribution for each zone. The size and vehicle submodels and the joint distribution submodel
used in the Case Study Model were developed from a 1987-88 home interview survey in the
Washington, D.C. metropolitan region. The regional joint distribution is shown in Table A-l. This
distribution is used as a "seed" for each zone and then modified to match the desired distributions
of households by size and by vehicles for that zone.
Parking Cost Submodel: The Case Study Model's parking costs were estimated using the
MWCOG parking cost model. This model estimates average daily parking cost in a zone as a
function of the number of home-based work person trip attractions per square mile in that zone. A
threshold value is included, so that no parking cost is estimated for most zones.' The parking cost
submodel is implemented as a look-up table, shown graphically in Figure A-3. The values in this
chart represent the average cost of parking between those who pay something and those who pay
nothing and are expressed in cents per day, in 1980 year dollars. (Throughout the Case Study
Model set, all costs are expressed in 1980 dollars.) The average parking cost value estimated by
this submodel can be overridden by the user for any zone.
U. S. Environmental Protection Agency
A-3
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Figure A-1
Case Study Travel Forecasting Model Flowchart
highway network
transit times
and feres,
parking cost
households and
employment
Trip generation
by purpose
Build peak and off-peak highway toll
and free path skims
Apply toll path choice model to calculate
weighted peak highway skims
Calculate peak and
off-peak composite
impedance
. Trip distribution by
vehicle
ownership data
transit times
and fares,
parking cost
toll/free
highway skims
Mode choice by purpose
I
Toll path choice model by purpose
vehicle
ownership data
Misc. PPAQ input files:
Traffic patterns
District equivalencies
etc.
Combine, balance, and assign vehicle trips
\
Calculate final VMT, speeds,
and emissions
Report of daily mobile source emissions
Spreadsheets for
estimating fee impacts
on vehicle age mix
A-4
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Figure A-2
Household Size Submodel
Proportion of Households by Size vs. Average Persons/Household
0.00%
1.05
1.55 .2.05 2.55 3.05 3.55
Average Persons per Household
4.05
4.55
5.05
Size
1
2
3 ,
4
5+
Total
Table A-l
Regional Joint Percentage Distribution of Households by Size and Vehicles
Vehicles
0
5.7%
33.0%
Source: 1987/88 MWCOG Home Interview Survey.
2+
4.2%
1.1%-
0.3%
0.1%
0.0%
18.6%
9.7%
2.6%
1.5%
0.6%
. 2.1%
' 24.7%
14.9%
• - 13.4%
6;2%
61.3%
Total
24.9%
35.5%
17.8%
15.0%
6.8%
. 100.0%
U. S. Environmental Protection Agency
A-5
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
A.1.3 Trip Generation Specifications
This component is the most traditional part of the model set. Daily home-based person trips
produced by households are estimated as a function of the joint number of households by size and
vehicles ownership. Daily person trips attracted to a zone are estimated as a linear function of
households and employment by type. Four trip purposes are used: home-based work (HBW),
home-based other (HBO), non-home-based (NHB), and truck. NHB and Truck vehicle trips are
calculated as trip ends only; i.e., origins = destinations = trip ends. The trip rates include all trips
and the split of trips that begin or end outside the region (external) are estimated as a percentage of
total trip ends. External trips are then split off as a separate trip purpose. Internal/external
productions are distributed to external stations in proportion to the 1990 traffic count at the external
station. ,
Figure A-3
Parking Cost Submodel
Parking Cost Submodel
350
so
100 150 . 200 250 3QO 350 400
Home-Based Work Trip Attractions per Square Mile (Thousands)
450
500
Source: MWCOG. . •
Work trip ends are balanced to the attraction total and trip ends for the other purposes are balanced
to production totals. The trip rates and attraction equations were derived from the models of
Washington, D.C., Dallas, and Minneapolis-St. Paul. The trip generation model is presented in
Tables A-2 to A-4. This is a very simplistic model, not so much calibrated as synthesized, whose
main focus was on matching the Washington, D.C. region's total number of trips by purpose. This
trip generation model is not sensitive to pricing.
A-6
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Table A-2
Trip Production Model
HBWtrips/HH
Size
1
2
3' •'• ' ' '
4
5+
HBO trips/HH
1
2
3
4 ' • •• ~
5+ ' .
0
0.372
1.105
1.912
1.912
3.038 ,
Vehicles
1
• 1.147
1.540
2.347
: . . ' • 2.347
3.187
2+
1.583
2.581
3.367
3.367
3.951
-.- . .
3.074
4.437
4.930
6.293
8.207
, 3.190
5.220
8.120
12.180
15.950 ,
3.335
5.452
8.497
12.644
19.372
• '
HBW Attractions =
HBO Attractions =
NHB Trip Ends =
Truck Trip Ends =
TabIeA-3
Trip Attraction Model
1.484 * total employment
0.45 * office employment + 13.50 * retail employment + 3,00 * other
employment + 0.75 * households
0.555 * office employment + 2.59 * retail employment +1.11'* other
employment + 0.'3 7 * households -
0.011 * office employment + 0.212 * retail employment + 0.212 * other
employment + 0.042 * households
Table A-4
External Trip Percentages
HBW
HBO,
NHB'
Truck
5.9%
1.5%
2.0%
5.9%
U. S. Environmental Protection Agency
A-7
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
A.1.4 Trip Distribution Specifications
In most areas, the distribution of trips from an origin zone to potential destination zones is
performed using a gravity model. This model distributes trips as a function of the number of trip
attractions and a measure of the separation of the zones. Almost all areas use highway time as this
measure of separation. However, researchers have long known that factors other than highway
time play a role in the allocation of trips to destination zones. For example, there is considerable
evidence that the presence of good transit service between two zones will increase the number of
person trips between those zones. A logical extension of this concept is that other components of
this separation, such as prices that users pay for transportation, should also influence destination
choice and thus be accounted for. This effect appears to have an intuitive and empirical basis that
cannot be ignored.
So far, very few urban areas have developed distribution models that are sensitive to the prices
travelers pay and service levels of all travel modes (also referred to as composite impedance).
Some examples include San Francisco, Boston, New Orleans, Atlanta, and Denver. Most areas that
use this formulation use it only for work trips, but in theory it should be applicable to all trip
purposes. This has become a well documented process and was adopted for the Case Study Model.
The Case Study Model distributes trips for all purposes with a gravity model that uses composite
impedance (CI) as its measure of zonal separation. This impedance is defined as the log sum from
the mode choice model, i.e., the natural logarithm of the inverse of the sum of the exponentiated
disutilities of all available travel modes (the denominator of the mode choice equation). This
method was adopted from the New Orleans regional model. This version of composite impedance
includes the following travel costs: auto operating cost, tolls, transit fares, and parking fees.
The use of this function makes the allocation of trips to destination zones sensitive to differences in
those user costs. For example, if transit fares were to decrease in a certain corridor, not only would
the transit share increase for those trips (from the mode choice model), but the number of person
trips in that corridor would increase slightly, because the decrease in fare causes a decrease in the
separation between zones in that corridor. Further, using information from the mode choice model
addresses the need for model
A-8
,. U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
connectivity, even though the models are still applied sequentially. Equation A-l presents the log
sum equation as used in this model.
1.0
where:
Cly =
U(tr) =
U(da) =
U(cp) =
e~U(da) + e~U(cp}
composite impedance from zone i to zone j
disutility of Transit mode
disutility of Drive Alone mode
disutility of Carpool mode
(A-l)
The disutility equations are presented in the discussion of the mode choice model, below. Peak
(congested) travel times and user costs were used for HBW trips; off-peak times and user costs
were used for the other purposes. A separate CI value, was calculated for each of three vehicle
ownership groups and three transit access areas (walk to transit, drive to transit, no transit access).
A weighted average of these nine values was then calculated. The value of 2 in equation A-l is to
ensure that the resulting value is greater than zero and the value of 25 is to spread the CI values
over the range of 1 to 254. '
-''"'/
In the New Orleans model, composite impedance is used to distribute work trips, but not non-work
trips. In the calibration of that model, it was found that travel prices and transit service were
uncertain influences on non-work destination choice. However, it seems intuitive that travel cost to
users should affect non-work trip patterns, in some way and so the Case Study Model uses
composite impedance to distribute home-based other, non-home-based, truck, and external trips as
well. .
Special friction factors (" F-factors" ) were developed for use with the composite impedance values.
These were estimated such that the resulting trip tables would have the same average trip length
and average CI as the "known" trip tables from the Washington, D.C. area. The F-factor curves
are shown in Figure A-4. In order to adhere to MINUTP requirements, the F-factors are scaled to
be integers between the values of 1 and 999,999. .
Various trip distribution calibration statistics are shown in Table A-5. The "observed" data are
based on 1996 trip tables previously estimated by MWCOG. These results suggest that the CI-
based gravity models reproduce known trip patterns fairly well. "'.'..
U. S. Environmental Protection Agency
A-9
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Figure A-4
F-Factor Curves
F-Factors
1,000,000,
so
100 150
^ •
Composite .Impedance
200
Purpose
Table A-5
Trip Distribution Statistics
Avg. Composite Imped.
'Observed" Estimated
Intrazonal Trips
Observed"
HBW
HBO
NHB
Truck
I/X75.24
; r— • ;
68.10
45.58
45.81
61.88
75.29
67.90
47.52
47.04
61.84
'
1 16,300
538,500
314,100
10,000
133,300
581,500
269,800
11,200
The resulting trip tables were found to sharply overestimate travel across the Potomac River and so
K factors were developed to force a better correspondence with the traffic counts on the bridge
crossings. HBW trips across the river were factored by 0.20, while all other purposes were factored
by 0.25. The need for such factors is supported by the existing MWCOG model, which uses extra
time penalties on the Potomac River bridges to accomplish essentially the same function.
A.1.5 Mode Choice Specifications
*
Within the four-step process, current mode choice models stand out as being the most rigorously
developed and properly price-sensitive component. Many larger urban areas have developed
sophisticated logit models which estimate the share of person trips by mode, based on the
socioeconomic level of the traveler and the time and user cost attributes of the various modes. The
Case Study Model mode choice model is based on the approach used in the Washington, D.C. area,
which is considered fairly representative of good practice.
A-10
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
• • to Reduce Transportation Emissions
•' ' ' i • • '
The HBW mode choice model incorporates a Carpbol model which estimates the split among 2-
person carpools, 3-person carpbols, and 4+-persoh carpools. Those percentages are used to
estimate the "average" attributes of the, Carpool mode, allowing the model to then split total
person trips into Drive Alone, Carpool, and Transit modes. Separate walk-access and drive-access
"markets" are used to calculate the Transit split. This split is sensitive to various auto mode
attributes, including terminal time, driving time, auto operating cost, tolls, and parking cost, as well
as transit attributes, including walk time, initial wait time, transfer time, in-vehicle time, and transit
fare. A special high-occupancy vehicle (HOV) feature allows the modeler to define HOVs as
having either 2, 3, or 4+ persons per vehicle and uses special travel times and costs for such trips.
The Washington model's sensitivity coefficients were replaced with those representing an average
of experience from logit models in numerous urban areas around.the. country (which-turned out to
be quite similar to the Washington area's coefficients).
Equation A-2 presents the logit mode choice model structure.
p =
where:
>-U(ma)
e-U(tra)+e-U(daa)
+e
~U(cpa)
(A-2)
U(daa).
U(cPa)
- probability of choosing mode m for a trip from zone i to zone j by ' •
travelers with auto availability a
= disutility of Transit mode for travelers with auto availability a
= disutility of Drive Alone mode for travelers with auto availability a
= disutility of Carpool mode for travelers with auto availability a
= disutility of mode m (Transit, Drive Alone, or Carpool) for travelers with
availability a *
auto
The disutility of the Carpool mode is a weighted average of the disutility of each integer Carpool
occupancy mode:
U(cp) = P2*U(2) + P3*U(3) + P4*U(4) ,
where: •
U(2), U(3), U(4) = disutility of 2-person, 3-person, and 4+-person carpools
P2> ?3 = P4 = probability of the trip being made in a 2-person, 3-person, or 4+-
person carpool, given that the traveler is in a carpool, calculated
as:
(A-3)
•'ijoa
(A-4)
U. S. Environmental Protection Agency
A-ll
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
where:
Pijoa = probability of the trip being made in a vehicle with occupancy o for a trip from zone i to
zone j by carpoolers with auto availability a
U(Oa) = disutility of carpool occupancy p for travelers with auto availability a
This is considered a "quasi-nested" logit structure in that the carpool model (equation A-4) is not
used to directly calculate the carpool share, but rather to calculate the average carpool disutility
(equation A-3), which is then used in the primary model (equation A-2). -Table A-6 presents the
disutility equations for the HBW primary model and the carpool model.
For home-based other and non-home-based trips, the MWCOG model does not use the logit
equation to calculate a mode share. Instead, it multiplies the estimated HBW mode shares by a
factor that varies with distance (and, for HBO trips, by vehicle ownership). A similar approach was
used in the Case Study Model, except that the non-work mode shares are based on off-peak
highway and transit times. The factors are shown in Table A-7.
A very important point to note about the cost coefficients is that they were originally calibrated
using prices expressed in 1980 year dollars. Thus, when this model is applied, all prices must be
adjusted to represent the equivalent price in 1980 dollars. For example, prices in 1993 dollars must
be deflated to 1980. This is done by multiplying them by a deflation factor of 0.57, which is the
ratio of the U.S. Urban Consumer Price Indices of 1993 to 1980 (= 82.4/144.4). Neglecting to
account for this effect will result in the model being improperly sensitive to price.
This model was not calibrated, per se, but the bias coefficients were adjusted so as to achieve
approximately the same transit trips and auto occupancy that had been forecasted for the
Washington, D.C. region for 1996. Those statistics are shown in Table A-8,
Prices were carefully specified in this model. All prices were expressed in cents per one-way trip,
in 1980 year dollars, because that is the year for which the original coefficients were, developed!
Auto operating cost was calculated as 11.0 cents/mile, multiplied by the minimum path distance for
each O/D pair. This is the incremental cost of using a vehicle for each trip and does not include
any of the fixed costs of vehicle ownership. Parking cost to users was calculated using the density-
based model shown above in Figure A-3 and divided by two to represent half the daily price. For
non-work trips, the average parking cost to users is assumed to be 20% of the daily parking price.
The toll (roadway price) input represents the weighted average toil for each O/D.
A-I2
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
table A-6
Home-Based Work Mode Choice Disutility Equations
Prime Mode'Choice
U(tr) = 0.05 * (WLK + WT1 + WT2) + 0.02 * RUN + 0.004 * FARE + 0.02 * AAC + ACBIAS
U(da) = 0.1001 * TERM + 0.02* RUN + 0.004 *OPCST + 0.0095 * (PKCST + TOLL) + DABIAS
U(cp) = 6.0519.* TERM +,0.02•* (RUN+'l.l*(CPOCC-l)) + 0.004 * OPCST/CPOCC + 0 0095 *
(PKCST + TOLL)/CPOCC + CPBIAS
CPOCC =
1.0
2.0
Carpool Share
P(2) P(3) P(4)
-) 1
3.0 4.394
U(2) = 0.1823 * TERM + 0.07 * (RUN + 1.1) + 0.014 * OPCST/2 + 0.0333 * (PKCST + TOLL)/2
U(3) = 0.1823 * TERM + 0.07 * (RUN + 2.2) + 0.014 * OPCST/3 +0.0333 * (PKCST + TOLLV3
3PBIAS ',,'••
U(4) = 0.1823 * TERM + 0.07 * (RUN + 3.733) + 0.014 * OPCST/4.394 + 0 0333 * (PKCST +
TOLL)/4.394 + 4PBIAS
Bias Constants by Vehicles.Owned
Vehicles 0
1
2+
ACBIAS
DABIAS
CPBIAS
3PBIAS
4PBIAS
2.1333 0,9360
3.3077-0.0439
2.9882 1.0989
•1.2951 1.7186
2.1106 2.5532
0.9501 (if auto used to access transit system)
-1.0552,
0.4200 , • . ! ' '• ••
'•1.6952 ..' •.-.-.•
2.1909
Abbreviations:
WLK = transit walk time
WT1 =? transit initial wait time
WT2 = transit transfer time
RUN = transit in-vehicle time
FARE== transit fare
AAC = auto access time to transit
TERM = highway terminal time
OPCST = auto operating costtp the user
PKCST = daily parking cost to the user (divided by
TOLL = average toll paid
CPOCC = average occupancy of Carpool mode
All times are in minutes and costs are in cents, expressed in 1980 year dollars.
U. S. Environmental Protection Agency
A-13
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Table A-7
Non-Work Mode Share Factors
Distance (mi.)
0-1
1-2
2-3
3-4
4-5
5-6
6-7"
7-10
10-30
30-50
Vehicles
Owned
0
V
2+
0
1
2+
0
1
2+
0
1
2+
0
1
2+
0
1
2+
0
1
2+
0
1
2+
0
1
2+
0
1
2+
HBO
Transit
Factor
0.5261
0.2374
Q.2374
0.6428
0.1807 -
0.1807
0-6428
0.1807
0.1807
0.6428
0.1807
0.1807
0.6428
0.1807
0.1807
0.6236
0.1607
0.1607
0.6236
OJ607
0.1607
0.6236
0.1607
.0.1607
" 0.5156
O.'l514
0.1514
0.0
0.0
0.0
HBO
Auto Pass.
Factor
2.1790
1-5515 .
2.7676
2.1790
1.5515
2.7676
3.7090
2.0419
3.3290
2.9104
2.0419
3.3290
2.1036
2.0419
3.3290
2.0978
1.9051
3.3290
1.8674 .
1.9051
3.3290
1.7769
1.9051
3.3290
1.5292
1.8150
3.3290
1.5.057
1.7791
3.3290
Table A-8 Summary of Mode Choice
-
Transit Trips*
Percent Transit
Vehicle Trips
Auto Occupancy
* Linked trips.
HBW
675,900
15.4%
3,309,700
1.125
HBO
460,300
4.5%
7,513,700
1.314
NHB
Transit
Factor
0.1334
0.1334
0.1334
0.1334
0.1334 •
0.1334
0.1334
0.1334
0,1334
0.1334
0.1334
0.1334
0.1334
0.1334
0.1334
0-1334
0.1334
0.1334
0.1334
0.1334
0.1334
0.1334
0.1334
0-1334
0.1274
0.1274
' 0.1274
0.0
0.0
0.0
Results
NHB
110,300
2.8% ]
3,052,600
1.239
•,
NHB
Auto Pass.
Factor
2.0081
2.0081
2.0081
2.0081
2.0081
2.0081
2.0081
2.0081
2.0081
2.0081
2.0081
2.0081
2.0081
2.0081
2.0081
2.0081
2.0081
2.0081
2.0081
2.0081
2.0081
2,0081
2.0081
2.0081
2.0081
2.0081
2.0081
2.0081 '
2.0081
2.0081
Total
1,246,500
6.7%
13,876,000
1.252
A-14
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
pair, which is the skimmed toll value multiplied by the percent of drivers which are estimated to
use the toll path for that trip.
The'mode choice model outputs vehicle trips by LOV and HOV occupancy categories. The HOV
occupancy criterion can be 2, 3, or 4 persons/vehicle; in this project it was set to 3. The
Washington area does have HOV highway facilities and their impact was included in the analysis.
A;L6 Path Choice and Traffic Assignment Specifications
Studies of toll roads focus on drivers' trade-off between paying a toll and saving time.
Traditionally, relatively less attention has been paid to drivers' path choices,.as planners have
relied mainly on traffic assignment software to handle that task. However, recent toll road studies
have discovered that such software is inadequate for modeling complex toll vs. time trade-offs and
have developed more sophisticated models of path choice. These models determine the best "free"
path for each zone-to-zone pair, i.e., the best path that does not use the toll facility. They then
determine the best "toll" path, which is the best path that includes the toll facility. Those paths are
analyzed to determine their time and toll difference, which is then used in a logit model to estimate
the split of trips between the two paths. This is done for every zone-zone pair in the network.
Separate toll and time sensitivities are used for work and non-work trips. Recent advances in
assignment software permit the two resulting trip tables to be assigned simultaneously, but each to
its own set of paths. Within the multiple iterations of assignment, trips are allowed to "migrate"
between paths to a limited degree in response to congestion. The result is a more realistic
assignment of trips to toll facilities, in a manner that is sensitive to the level of toll as well as to the
capacity of the alternative non-toll routes. "
This kind of process has been recently used in toll road studies in Denver and New Jersey, and was
adapted for use in the Case Study Model, under the assumption that it is suitable for analyzing
roadway pricing measures. The resulting toll, values affect not only the path choice, but also the
choice of mode, which uses toll as an input, as noted above. Because toll is part of the composite
impedance calculation, toll values affect the distribution of trips as well.
The toll diversion model is a logit formulation that splits all vehicle trips for each O/D pair into
"toll" and "free" trips. The equation is as follows:
P(toll), = •
1.0
xAU(toII-free)
(A-5)
where:,
P(toll)jj= probability of choosing the toll path for a trip from zone i to zone j -
AU(toll-free) = difference in disutility between toll and free path
= 0.301 * (toll time - free time) + 0.0127 * tollpath toll + 1.5 (HBW.trips)
= 0.191* (toll time - free time) + 0.0272 * toll path toll + 0.3 (all other trips) toll
is in cents and time is in minutes , • , .
The toll time is computed based on paths that are built by temporarily reducing the time on all
tolled links by 10%. This encourages any trip which is near the toll link to use it. The free path
time is computed by prohibiting the use of all tolled links. Throughout the Case Study Model,
U. S. Environmental Protection Agency
A-15
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
highway paths for skimming and for assignment are built on the basis of minimizing impedance,
where impedance is calculated in equivalent cents as:
impedance (in cents) = 10.0 * time (in minutes) + 11.0 * distance (in miles)
The 10.0 factor on time is derived from an assumed effective value of travel time of $6.00 per hour
(roughly 22% of the Washington, D.C. area's mean 1990 household income). The 11.0 factor on
distance represents the user cost of operating an automobile, as noted above. This formulation
ensures that user cost influences highway paths.
The traffic assignment procedure uses four iterations of incremental., capacity-restrained
assignment, with 25% of the trips assigned on each iteration, following current MWCOG practice.
Thus, the assignment of trips is sensitive to roadway capacity in an incremental fashion: some trips
"see" an open roadway, while others see a congested one. The input daily vehicle trips are split by
four categories: low-occupancy vehicle (LOV) free path, LOV toll path, HOV free path, and HOV
toll path, with each category of trips being assigned to its specific set of paths, respecting the
presence of priced roadways and HOV roadways. The output of this process is a loaded network
with daily traffic volumes on each link.
A.I.7 Emissions Calculation Specifications
The estimation of mobile source emissions requires two basic data items from the traffic
assignment: VMT and speed. U.S. EPA's MOBILESa emission factor program is applied to
calculate emission rates in grams/mile for the: criteria pollutants (the EMFAC7F program is used in
California). These rates are a function of the mix of vehicles by eight types, the average distance
they travel per year, average travel speeds, ambient temperatures, inspection and maintenance
(I/M) programs, fuel policies, and other factors. The PPAQ (Post-Processor for Air Quality)
program developed by Garmen Associates is used to read the loaded network, revise the speed
calculations by facility type and time period, summarize VMT by facility type and time period, and
apply the MOBILESa emission factors. PPAQ requires a series of input tables that reflect the mix
of vehicle types by roadway type, the percentage of traffic by hour, and other parameters that
describe traffic patterns in more detail. These parameters have been adopted from work recently
performed in the Philadelphia region. The result of a PPAQ run is an estimate of total daily tons of
HC, CO, and NOX from on-road mobile sources.
Changes in most of the pricing measures under study, such as roadway pricing, transit fares, and
parking costs, will be reflected in the assigned link volumes. The exceptions are the measures
involving registration fees that are based on age or emission level. It is assumed that such
strategies have no measurable impact on the amount of VMT, but will affect the mix of vehicles by
age. Strategies that make it more expensive to own an older vehicle should result in fewer older
vehicles on the road. Since older vehicles were generally built to less stringent emission standards
and are usually less well maintained, a reduction in such vehicles can be expected to reduce the
emission rates calculated by MOBILESa.
. ,J "* * . ! ' • \
Spreadsheets have been developed in this study to analyze age-based fees and emission-based fees.
These spreadsheets estimate the impact of different fee structures on the default MOBILESa
vehicle mix by year. The spreadsheets use a very simplistic relationship between the annual user
cost of owning a vehicle and the percentage of vehicles in each age category. For each age
A-16 , U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
category, a,value is computed representing the annual fee divided by the estimated market value of
the vehicle. This value is the change in fee as a percent of the vehicle's value. An elasticity of -
1.0, derived from literature on the price elasticity of new'car sales, is applied to that change in fee
The resulting value is the estimated percentage change in fleet share for that age category - This is
done separately for each of the eight MOBILESa vehicle types and MOBILESa defaults are used
for the initial age mix. It is assumed that none of the registration fee policies would affect the total
number of vehicles; only the age mix. Thus, all of the vehicles removed from older age categories
are shifted to newer age categories, in their existing proportions.
The output of these spreadsheets is a revised.set of vehicle age mixes by vehicle type that can be
input directly into MOBILESa. Table A-9 shows an example spreadsheet, analyzing a fee schedule
for light-duty gasoline vehicles (passenger cars) that varies from $100 for 4 year old vehicles to
$900 for vehicles of 20 years or more.
. ' '• r '
The revised age mix is entered into PPAQ, along with the loaded highway network from the
assignment step. PPAQ splits the link volumes by time period, adjusts the speeds on each link
summarizes VMT and average speed by time period, facility type, and jurisdiction, applies
MOBILESa to each "scenario", multiplies the resulting emission rates by the VMT, and produces
a concise summary of the regional mobile source emissions of interest. Table A-10 is a sample of
the summary page of PPAQ output.
U. o. Environmental Protection Agency
A-17
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Table A-9
Sample Vehicle Age Mix Adjustment Spreadsheet (Age Fee)
ILDGV |
year
25
24
23
22
21
20
19
18
17
16
IS
14
13
12
11
10
9
S
7
6
5
4
3
2
1
estimated
value
$400
$500
$600
5700
$900
51,000
SI ,200
$1,500
51,800
$2,100
$2,500
$3,000
$3,500
54,100
54,900
$5,700
$6,700
57,900
59,100
510,500
512,000
513,500
514,700
514,800
517,922
elasticity:
change in fee
5900
5900
5900
$900
5900
5900
5800
$800
$700
$700
$600
$600
$500
5500
5400
$400
$300
$300
5200
5200
$100
5100
$0
$0
SO
-1.00
fee change
as % of value
225.0%
180.0%
150.6%
128.6%
100.0%
90.0%
66.7%
53.3%
38.9%
33.3%
24.0%
20.0%
14.3%
12.2%
8.2%
7.0%
4.5%
3.8%
2.2%
1.9%
0.8%
0.7%
0.0%
0.0%
0.0%
avg age:
MOBILESa registration mix by year (starting 1 July 90) new/base
base % by yr
1.10%
0.30%
0.40%
0.50%
0.60%
0.80%
1.10%
1.50%
1.40%
1.90%
2.40%,
3.70%
4.70%
5.40%
5.00%
5.10%
5,00%
5.60%
7.70%
8.10%
8.40%
8.20%
'8.30%
,7.90%
4.90%
100.0%
8.2
new % by yr
0.00%
0.00%
0.00%
0.00%
0.01%
0.10%
0.41%
0.77%
0.93%
1.38%
1.98%
3.23%
4.41%
5.23%
5.09%
'5.28%
5.33%
6.03%
8.44%
8.93%
9.39%
9.19%
9.39%
8.94%
5.54%
100.0%
7.1
%left
1.10%
0.30%
0.40%
0.50%
0.59%
0.70%'
0.69%
0.73%
0.47%
0.52%
0.42%
0.47%
0.29%
0.17%
-0.09%
-0.18%
-0.33%
-0.43%
-0.74%
-0.83%
-0.99%
-0.99%
-1.09%
-1.04%
,-0.64%
0.0%
cum % left
1.10%
1.40%
1.80%
2.30%
2.89%
3.58%
4.27%-
5.01%
5.48%
, 6.00%
6.41%
6.88%
7.17%
7.34%
7.26%
7.08%
6,75%
6.32%
5.58%
4.75%
3.76%
2.77%
1.68%
0.64%
0.00%
avg fee:
ratio
0.0%
0.0%
0.0%
0.0%
., 2.4%
13.0%
3.7.1%
51.2%
66.5%
72.6%
82.6%
87.3%
93.9%
96.8%
101.7%
103.5%
106.6%
107.6%
109.6%
110.3%
111.8%
112.1%
113.1%
113.1%
113.1,%
$235
Note: "LDGV" = light-duty gasoline vehicle
A-18
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Table A-10
Sample PPAQ Output
EPA Market-Based Measures Study: 1996 Base
Page 36
• Total Emissions by Scenario: • '. ' . [ ~ ' ~~- — : —
Scenario
A 67F
A 67F
A. 67F
A 67F
A 67F
• : A '67F
A 67F
A 67F
• A 67F
• A 67F
A 67F
"A 67F
A 67F
A 67F
A 67F
A- 67F
A 67F
A 67F
A '-67F
A 67F
A 67F
A 67F
A 67F
A .67F
TOTAL -
IT
IT
IT
IT
2T
2T
2T
2T
3T
3T
3T
3T
4T
4T
4T
4T
5T
5T
5T
5T
6T
6T
6T
6T
kg
1
2
3
4
1
2
3
4 ,
1'
2
3
4
1
2
3 '
4
1 '
2
3 .
4
1 •
2
3
4
VMT
(veh-mi)
6959230
9368194
17085802
11197222
1 819904
1285529
. 2004210
• 951494
6361328
9973934
15.549907
7382279
4687831
7350054
•11459147
5440198
1564851
1 2745873
3572584 .-
1958525
2094284
3674875 .
4781289
1 2621150
140889696 ' -
VHT Avg Sp'd -
.(veh-hr)
1 141448
324159. ,-
'411706
1 189783 '
20345 '
76520 .
49856 ,
17818
340178
1049888 .
822746
232147
393935
1080890
962954
256613
• 115915
450143,
. 24'4698
, . 99925
66910
117784
152757
83743
.7702861
(mph)
49.
.28.
• 41.
59.
40.
' 16.
40.
53.
18.
9.
18.
'31'.
11.
6.
11.
-' 21,
13.
, 6.
. 14.
.19.
31.
31.
31.
31.
18.
2,
9
5
0
3 •
8
2
4
7
5
9
8'
9
8
•9
2
5
1
6
6 '
3
2
3
3
3
- tons ,
VOC HC
(kg.)
7289
16466
22108
12909
942
3435
2563
926
12872
42272
35269
9719
13371
'40505..
" 38330
9549
40.71
17039
10154
3616
- 2903
5964
7395
- 3457
323124-
•356.179
, CO
, , (kg) -
49933
107271
1.42505
13i713 •
6434
• , 22543
16282
• ' 6571
99395
273395
• 249739
70717 •
102474 '
'267663
260217
74506
. 31003
110003
' 69201
28669
20651 "
,37679
.48753 •
• ' 25087
• 2252406
2482.827
NOx
• (kg)
13.690 •
1,6541
31823
32602
1404
2085
3452'
2052
10507
17313
• 25847
12602 ;
8178
13550
:19982
8818
2687
5153
6048
3129
3575'
6094 ,
8159 •'
' 4355
259645
286.207
HC Emissions Breakdown: . • > '
Exhaust 186606 kg 205.696 tons. 57.8%
Evaporative 33070 kg 36.453 tons 10.2%
Refueling 0 kg 0.000 tons 0.0%
Running, 9li80 kg 100.507 tons 28.2%
Resting .. 12268 kg '13.523 tons 3.8%
PPAQ/MOBILE5 ' . . .
04-21-1994. / dO : 5.7:18' •
Note: . A 67 refers to the summary jurisdiction, of which there is only one.
F x refers to the facility type.
T y refers to the time period: 1=AM peak, 2=PM peak, 3=midday, 4=night.
U. S. Environmental Protection Agency .
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Post-processor programs such as PPAQ are commonly used in air quality modeling. This is mainly
because historically, the constrained speeds that result from most traffic assignments are not
sufficiently accurate. These speeds represent just one particular value and they tend to include
other elements of impedance, as needed to obtain accurate assigned volumes. Many planning
agencies have developed their own speed processor programs to adjust these constrained speeds to
be more realistic and to summarize VMT by various stratifications, in preparation for running
MOBILESa.
A.1.8 Model Application
The case study model set is applied in a series of 15 program steps. The MINUTP planning
software system is used for most of these. Custom FORTRAN programs were written to prepare
the land use data and to apply the mode choice model. PPAQ and MOBILESa are "stand alone"
programs and the age mix spreadsheets are in Microsoft Excel. The full model set requires about 9
hours to apply using an 80486-based computer running at 66 MHz.
The case study model set was applied using basic land use and network data from the Washington,
B.C. area (1,478 zones), representing approximate 1996 forecast conditions. However, since the
model incorporates components from various cities, the results do not reflect actual (or forecasted)
conditions in Washington and cannot be compared to the results from the Washington area's own
model set
The Washington area is projected to comprise about 1.8 million households, 4.0 million persons,
and 3.0 million jobs. The area has an extensive Interstate system, including a Beltway around the
District of Columbia and the close-in Maryland and Virginia suburbs (I-95/I-495) .. In addition to
an extensive bus network, the area is served by the Metrorail system and four commuter rail lines.
Major HOV facilities exist in Virginia on the Shirley Highway (1-395 and 1-95 to the south) and I-
66 to the west. There is one existing toll road, connecting the Beltway to Dulles International
Airport to the west.
*
As Figure A-l shows, the model set is applied " backwards":
• The path choice model is applied first, to derive weighted average highway time,
distance, and toll values. Peak period values are used for work trips and off-peak
values are used for all other purposes.
• _The mode choice program is then applied to calculate the composite impedance
value by zone-zone pair.
''.,•• ' .' . l '
• Trip generation and distribution are applied to estimate person trips by purpose.
• The mode choice program is applied again, this time to split person trips by mode.
• The path choice model is applied again to split vehicle trips by toll vs. free path.
• The vehicle trips are assigned to the highway network.
• The age mix spreadsheets are applied to determine changes in the age mix.
• PPAQ and MOBILESa are applied to compute emissions from the loaded network.
The case study model has not been " calibrated" in the true sense, since it was developed from data
representing several urban areas. However, the results of the various components were checked for
A-20 . U. £.' Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
, • . - to Reduce Transportation Emissions
internal consistency and to ensure an approximate level of correspondence with 1996 total VMT
estimates for the Washington, D.C, area.
A.1.9 Summary of Advanced Features of the Case Study Model
As noted above, the Case Study Model set is hot based on new research and does not represent any
breakthrough in-the-state of the practice in travel demand modeling. Its advancement is that it was
created from the best pricing-related features from several other well-documented model sets that
have been extensively tested throughout the years. It is very likely the first time that these various
components have been assembled in this exact manner and demonstrates one way in which the
four-step process can be enhanced to be sensitive to policy issues such as pricing.
The most noteworthy features of this model set are:
• Use of composite impedance for trip distribution for all trip purposes
• Integration of a toll/free path choice mode] within a four-step process
• Mode choice model, including a nested carpool occupancy model
• Parking price sub-model within the mode choice model that permits separate
specification 6f,parking prices for LOVs and HOVs
• Assignment procedure that simultaneously handles LOV/HOV and toll/free path
trips -. - . •
• Ability to easily calculate effects of vehicle age mix changes
• Integration of mobile source emission calculations within a four-step travel model
A.2 Enhancements to the Case Study Model
The resource constraints of this project limited the development of the Case Study Model. As this
work progressed, several shortcomings became apparent. These are described here as guidance for
enhancing this particular model set, and other models in general.
Income Stratification: It would be preferable to stratify the entire model set by income level. This
would permit the identification of differential price sensitivities by income level and would
facilitate the examination of the differential effect of pricing policies by income level.
Unfortunately, time did not permit the development of such a model, nor was the 1990 Census
Transportation Planning Package (CTPP) available at the time this work was conducted. Now that
CTPP files have become more widely available, income stratification should be pursued more
vigorously. This is important because there is considerable evidence which suggests that people of
different income levels respond to pricing measures differently, and public policy considerations
suggest the need to analyze how the implementation of such measures affects each income group.
More investigation into the combined effects of income and vehicle ownership on traveler behavior
should also be conducted. Traveler surveys need to collect information on household income (or
income category, such as quartile). Progress is being made in the difficult area of estimating the
stratification of trip attractions by income.
Vehicle Ownership: There is growing evidence that the effect of vehicle ownership on person trip
generation cannot be discounted, even after considering the effect of income. Vehicle availability
/. S. Environmental Protection Agency , . _
• ~ '
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
may well have a measurable effect on discretionary trips and on the level of trip chaining.
Considerable attention is now being paid to developing vehicle ownership models that are sensitive
to income, locational, density, accessibility, and/or transit service variables.
In order to analyze the effect of registration fees or one-time purchase surcharges/rebates, these
models must also be at least marginally sensitive to the price of buying a vehicle and the annual
user cost of owning one. In the context of travel modeling, increases in ownership cost should be
reflected in decreased numbers of multiple-vehicle households and increased numbers of one-
vehicle households. There might even be an increase in zero-vehicle households, but that is much
less clear. At least one vehicle ownership model sensitive to user cost has been well documented
[5]. . .....
Speed Feedback: The Case Study Model now begins with peak and off-peak speed values, the
peak speeds having been derived from previous MWCOG model runs. It would be preferable to
run the entire model set at least one more time in each application, using the speeds from the first
run (modified as necessary to more closely match observed data) as peak speeds in the second run.
This is particularly important for estimating congestion pricing scenarios. However, in this model,
two iterations would require a total of 18 hours (this is one of the disadvantages of an
interconnected model set — changing one thing changes everything and so the entire model set
must be iterated to ensure equilibrium). Doing this would have severely limited the ability to meet
this project's schedule and so was not done.
In addition, numerous studies have documented the sensitivity of emission rates to average speeds.
Thus, it is vital to have reasonably consistent and accurate speeds for emission modeling. In this
study, the PPAQ program was used to adjust assignment speeds by facility type and time period.
Auto-Access to Transit Trips: V.ery few models include such trips in their vehicle trip table. To
do so requires access to transit network data (such as park-and-ride [PnR] lot locations) that was
unavailable for this project, as well as additional processing steps and time. Still, this phenomenon
should not be overlooked, because some improvements to transit service can increase emissions by
enticing some who carpool or who walk to transit to switch to driving to transit. Although most
drive-access trips are short, they almost always involve a cold engine, and therefore cold start
emissions. Increasing the number of such trips might not compensate for the passenger car VMT
that is removed from the network.
1 . • • '
« " .
Parking Price Measurement: The cost of parking to the user is one of the most important
variables in determining mode choice. Thus, it would be preferable to measure it more carefully,
such as specifically modeling the proportion pf travelers who have free parking (instead of
accounting for that effect in the average value, as the Case Study Model does). This would require
some additional data, which is difficult to obtain, but it would improve the model's ability to
respond directly to policies which change the percent of travelers who park for free. It would also
be preferable to include variables other than density in the parking cost submodel, to increase its
accuracy.
In addition, if strategies to change non-work parking pricing are to be properly modeled, an
improved method of estimating non-work parking prices is needed. This submodel would probably
be based on employment density and locational variables. The cost of all-day parking to
commuters is relatively stable and closely related to work trip end density, as suggested above in
A-22 ' . U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
'to Reduce Transportation Emissions
Figure A-3. Of course, non-work travelers have a shorter parking duration time and are even more
likely to park for free than commuters5 but otherwise little is known about their average parking
costs.
Effect of Price on Trip Generation: In this model set, the number of person trips per household is
completely insensitive to the incremental user cost of travel by any or all modes. As was noted
above, that might be remedied by- including some kind of composite accessibility measure at the
zonal or household level, which would require additional research. This would make travel
estimates sensitive .to area-wide pricing measures such as a gasoline tax increase. It would also
address the phenomenon of "induced travel" that is a concern to some planners by making trip
rates sensitive to the general level of development in a region and the general quality of highway
and transit service. ,
* • . " • . ' •
With a doubly-constrained trip distribution procedure such as the standard gravity model, it would
probably prove necessary for the trip attraction estimate to be sensitive to user cost (if the gravity
model uses composite impedance). Otherwise, trip pattern distortions could occur under certain
pricing scenarios as the gravity model tries to reallocate attractions in response to cost changes, but
the attraction model would maintain the same attractions in a zone.
Time of Day: This model set estimates.total daily travel only. Although separate .peak and off-
peak impedances are used to represent those periods, the model does not account for the possible
"migration" of trips from one period to the other. Such migration might occur due to congestion,
pricing, or employer policy. ^A basic model that splits daily travel into three or four periods is not
conceptually difficult; such models are currently used in several cities. Further refinements to
make that model sensitive to peak/off-peak time savings should be readily feasible.* However, to
analyze peak-only pricing or congestion pricing measures* the model also needs to be sensitive to
the peak/off-peak travel, cost differential and it is uncertain whether the data to define such
sensitivity exists. However, it may be necessary to synthesize. such a sensitivity (or perhaps
estimate it from stated preference surveys or the experience of transit systems with peak/off-peak
fares) and re-examine it in the context of an actual congestion pricing experiment. This is
important, because a likely major response to peak-only pricing measures is to shift the time of
travel. s .,
Trip Chaining: As in most other travel forecasting models, this model does not directly estimate
any effects on trip "chaining". Many researchers believe that a logical and likely response to
increases in the price of travel (or reductions in the number of vehicles available to a household) is
for people to use their vehicles more efficiently: This could have some effect on VMT and
potentially a greater impact on cold starts. It would be helpful if the model set could reflect this
phenomenon. . ... '
Nested Logit Model: Although the multinomial logit model is the most widely used formula for
mode choice modeling, it is starting to be replaced by the nested logit model. This is because the
true nested model is more adept at handling sub-mode splits (e.g., bus vs. rail, within the transit
mode). There is some evidence that the nested logit structure more closely represents travelers'
trade-offs of attributes when selecting a travel mode. The nested logit also eliminates the
" independence of irrelevant alternatives" property, which can cause mode share estimation errors
for travel modes with similar characteristics.
U. S. Environmental Protection Agency . , .. ,--.-'-' A-23
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Separate Mode Choice Models by Purpose: The Case Study Model includes a true mode choice
model only for HBW trips. HBO and NHB mode shares are calculated as a distance-based factor
multiplied by the HBW shares. This implies that the non-work mode shares move up or down in
the same proportion as the HBW mode shares, which is unlikely. There is evidence from around
the country that the mode share of non-work trips is more sensitive to price and less sensitive to
time than for work trips, which is logical. It would be preferable to develop separate mode choice
coefficients for work and non-work trips. This is particularly important, given that the mode
choice coefficients influence trip distribution in this model as well.
One of the reasons why there is so much attention to work trips is because the CTPP provides a
potentially robust source of calibration data for commuting. More extensive home interview
surveys, perhaps as large as 1.0 - 1.5%, may be needed to capture enough observations of HBO and
NHB trips (especially using transit) to permit the calibration of usable coefficients.
Improved Vehicle Age Mix Submodel: The Case Study Model's vehicle age mix submodel is
overly simplistic and based on rough approximations. For example, its assumption that the age mix
is sensitive to the ratio of the annual fee divided by the value of the vehicle is open to question.
The submodel that estimates the average value of the vehicle by age and type was not rigorously
calibrated due to time constraints but should be redeveloped based on known data. Further, for the
• emission-based fee, some provision in the.submodel should be made to allow motorists the option
of repairing the vehicle to improve its emissions, thus avoiding some or all of the fee. Additional
research is necessary to determine if an annual or quarterly fee would truly affect vehicle
ownership decisions or day-to-day travel choices (or both). s Finally, some research should be
conducted into the effect of leasing vs. buying on vehicle replacement decisions.
Long-Term Land Use Allocation Effects: The Case Study Model, as with most travel demand
models, simulates travel choices that are made in the short-term; i.e., hours, days, weeks.
However, those pricing strategies -which travelers perceive to be permanent can also have long-
term influences; i.e., months or years. It would be very desirable for the travel model to be
supplemented by a land use model that could adjust the allocation of households and jobs in
response to such pricing changes. For example, a significant change in the toll on a bridge might
be expected ultimately to persuade some existing travelers, over time, to change their location of
residence, work, shopping, etc. Such tolls might also influence the locational decisions of residents
and/or employers who are new to the region. Modifying future land use in response to pricing
changes is one way to respond to the issue of " induced, demand". The few land use allocation
models now in use employ travel time in their allocation process, just as typical trip distribution
models do. Introducing pricing variables, such as composite impedance, might be one way for land
use allocation models to be sensitive to the travel cost of users. •
It should also be noted that the sensitivity of long-term locational decisions to pricing is probably
quite different than short-term decisions on where to travel or by what mode. This long-term
sensitivity is probably lower, but additional research is needed to quantify this difference.
Emissions Estimate Sensitive to Trips: The Post-Processor for Air Quality (PPAQ) program used
in this project greatly facilitates the calculation of the emissions burden from each scenario's
loaded network. However, PPAQ uses only VMT in its calculations. Although the additional
emissions from cold starts, and hot soak and diurnal VOCs are included in the MOBILESa
emission rates, these should more properly be modeled as a function of the number of trips rather
A-24 U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
• , to Reduce Transportation Emissions
than VMT. This would more accurately differentiate the emissions impact of those strategies that
.reduce vehicle trips from those that reduce VMT.
• • • > '••'..'
This can also be partially addressed through an alternative technique for estimating cold start
emissions [8]. That method uses purpose-based factors to split the vehicle.trips into cold- and hot-
start trips. The two trip tables are then assigned and the assignment software keeps track of where
the cold start VMT occurs on the network. This provides an estimate of cold start VMT on each
link, which improves the estimation of cold start emissions without having to calculate the change
in vehicle trips.
A.3 Application of the Case Study Model
In this project, market-based measures, to reduce mobile .source emissions were classified into
several groups. Within each group, specific variations of the measures were identified for detailed
analysis. The following sections discuss how the Case Study Model was used to evaluate each
measure. This information is presented primarily to provide guidance to travel demand forecasters
in the development and use of similar techniques for evaluating market-based measures in their
own areas.
<* . , ' ' , '."'•. ' .
A.3.1 Parking Pricing Measures
IN i • ' ' ' •
Analysis of parking pricing measures involves adjusting parking costs for a specific part of the
region (e.g., the CBD), and/or a specific time period (peak vs. off-peak), and/or a specific segment
of the travel market (e.g., work trips). Parking pricing measures are represented by a fixed
incremental change or a percentage adjustment to existing parking costs. Parking costs are
represented in the model by specifying the average daily cost for parking in each traffic analysis
zone. A separate cost can be supplied for each zone for both LOV parking and HOV parking. This
allows the specifiqation of strategies with differential pricing based on vehicle occupancy.
/ . ^ - . . . • . \
Parking costs are used to compute combined impedances and to estimate modal shares. This
affects the estimation of traveler choice of destination and mode of travel (with the potential to
affect travel by time of day). Five parking scenarios were examined in two broad categories:
• Increase in average cost of parking (all spaces)
^ (a) $4 per day increase in the core area and $2 increase elsewhere
(b) $ 10 per day increase in the.core area and $5 increase elsewhere
(c) $5 per day increase in the core area and no change elsewhere
• Increase in average cost of parking (employee parking only) ".-
(a) same as (a) above but only applied to home based work trips
(b) same as (b) above but only applied to home based work trips
Each of the scenarios are represented by adjusting the zonal data file used in the combined
impedance step and the mode choice step. All incremental cost adjustments are converted to 1980
dollars by using the 0.57 factor prior to adjusting the parking costs in the zonal data file. Table A-
11 summarizes the impacts of this scenario. - ' . .' .
U. S. Environmental Protection Agency ~~' " ~~' : A~25
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Table A-ll
Parking Cost Impacts
Model Step Impact
Trip Distribution The composite impedance changes for paths destined to zones that have
modified parking costs. This does not change the total number of trips that
are attracted to these zones since that value is fixed from the trip
generation step. It does change the distribution of trips to zones with
modified parking costs.
Mode Choice Changes in parking cost affect both the split between auto and transit and
the split between LOV auto trips and HOV trips.
Assignment VMT decreases due to fewer vehicle trips.
Emissions HC, CO and NOX emissions decrease due to the decrease in VMT.
A.3.2 Roadway Pricing Measures
Roadway pricing involves charging a fee (or toll) for vehicles to use certain roadways or to cross a
border into a designated area. The toll can vary by time of day, by vehicle occupancy, and/or by
the level of congestion on the roadway. Most such proposals would have the toll collected through
automatic vehicle identification so that no delay would be imposed in collecting the toll.
It was assumed that this toll would influence travelers' choice of 1) destination zone, 2) travel
mode, and 3) highway path. The Case Study Model does not include a time of day submodel, so
influences on the time of travel could not be estimated. In effect, the toll is assumed to be assessed
to all vehicle trips (LOV and HOV) on the selected roadways at all times of the day. In this
scenario, the selected roadways were those specific freeway and expressway links in the
Washington, D.C. region that had been shown, through previous trip assignments, to have an
estimated peak hour volume/capacity (V/C) ratio of 0.9 or higher. These were almost all of the
Interstate and major through Primary roads in the region (727 one-way links out of a total of 18,400
links). Although these priced roadways represent 4% of the total coded network, they carry about
31% of the total daily VMT.
One level of toll was analyzed: $0.10/mile. This is a typical high-side value of the toll being
charged in newly proposed toll roads in various areas. The first step was to convert this into 1980
year dollars by multiplying by 0.57, as noted above. The equivalent 1980 toll level is then
$0.057/mile. The increased cost of using certain roads was first reflected in the path skimming
step, in which path-building impedance is defined as the sum of auto operating cost (cost/mile
multiplied by distance), and the money value of travel time. Priced roadway paths and non-priced
("free") roadway paths were then built and input to the toll diversion model, which estimated the
priced route/free route share for allO/D pairs. The average toll paid is then the toll on the shortest
path, multiplied by the proportion of drivers who are estimated to use that path. A weighted
average of the priced and free travel time and the priced and free distance were also computed.
A-26 U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
The average time, distance, and toll for each O/D pair was then input to the mode choice program
in order to calculate composite impedance. That impedance value was used in the gravity model to
distribute trips from origins to destinations. The average time, distance, and toll were again input
to the mode choice program, this time to compute the split among travel.modes. The toll diversion
model was applied a second time to split the vehicle .trips by priced/free path. The assignment
program assigned priced trips to the priced paths and free trips to the free paths. Within the four
assignment iterations, priced trips were allowed to "migrate" to the free paths (and back) in
response to changing relative congestion levels between priced and free roadways; however, free
trips were required to stay on a free path. Table A-12. summarizes the impacts of this scenario on
the travel choices.
Table A-12
Roadway Pricing Impacts
Model Step
Impact
Highway Skims
Trip Generation
Trip Distribution
Mode Choice
Toll Diversion
Assignment
Emissions
Highway paths are.modified to select paths that minimize the total cost.
Average time, distance, and toll values are estimated that reflect the impact
of the priced roadways. -
No impact. .-,.-'
Composite impedance increases for paths using the priced roads due to the toll.
It also increases slightly for paths using free roads because they are longer
(drivers are accepting increased distance, and operating cost, in order to
avoid the toll).
Cost of driving increases; driving alone decreases, and carpooling and
transit increase slightly.
Presence of tolls causes a share of the remaining vehicle trips in affected
O/D pairs to switch to the free paths.
VMT decreases due to fewer vehicle trips. However, congestion increases
on the free roads, resulting in much slower overall average speed.
HC and CO emissions increase, and NOX emissions decrease due to the
lower average speed. -
These results fit with the initial expectations and the results of other research, up through the toll
diversion model. The assignment results, however, bear further explanation. Although some
switching to carpools and transit occurs, this does not occur in significant numbers. Some travelers
switch destinations to one which can be reached via a free route. The majority of travelers are
those who dp not change their destination or mode and the primary method for them to "escape"
tolls in this model is by switching to a free route. In this scenario, the free routes are generally the
arterials, which already had slower average speeds to begin with, and those speeds are made even
slower with the additional traffic from the priced freeways. Reducing average network speeds
generally has a negative effect on HC and .CO (although this depends on the existing average speed
values, as noted above). - . ' F
U. S. Environmental Protection Agency
A-27
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Part of the problem may lie in the definition of " congestion". In this scenario, the priced roads
were those freeways and expressways with a V/C of 0.9 or more. Although V/C is a good indicator
of a roadway's congested speed compared to its free-flow speed, that is a false indicator of
emissions. That is, MOBILESa's emission rates are not based on congestion, per se, but absolute
average speed. Consider the case of a freeway section with a free-flow speed of 60 mph. With
congestion, the speed might drop to 40 mph, a difference of 33%. But according to MOBILESa,
vehicles on that roadway will still have lower emission rates than vehicles on a nearby arterial'
whose congested speed might typically be 35 mph. This is true even through the arterial's free-
flow speed might be 40 mph, in which case the arterial is experiencing less congestion than the
freeway (i.e., a speed drop of 13%).
Still, this example highlights some of the complexity of roadway pricing analysis. Even with the
enhancements made to the Case Study Model, the following shortcomings are still seen:
• No price-sensitive time of day model.
• Toll diversion model not specifically calibrated to Washington, D.C. area.
• No speed feedback in the model set.
the above problems notwithstanding, a key observation of this analysis is that a network-based
procedure is essential to the complete analysis of roadway pricing measures. The interactions
among traveler choices of destination, mode, and path are too complex to model without
performing a specific assignment of vehicle trips to a roadway system.
A.3.3 Vehicle-Based Measures
A.3.3.1 VMTFee
A fee would be assessed for each vehicle-mile driven. It was assumed that this fee would be
collected in such a manner that travelers would be aware of it on a daily basis and therefore the fee
would influence their short-term travel decisions. Thus, the VMT fee was modeled in terms of an
increment to the automobile operating cost parameter. No impact on vehicle ownership levels, land
use, or long-term locational choices was assumed.
Three levels of VMT fee were analyzed: $0.04/mile, $0.10/mile, and $0.40/mile. The first step was
to convert thesp into 1980 year dollars by multiplying by 0.57, as noted above. The equivalent
1980 fee levels are then $0.023/mile, $O.Q57/mile? and $0.228/mile. These values were then added
to the base automobile operating cost of $0.1 I/mile to yield new automobile operating costs of
$0.133/mile, $0.167/mile, and $0.338/mile. The mode choice/composite impedance application
program includes a specific user-coded parameter for auto operating cost per mile and this
parameter was adjusted for each run,
In addition, the increased cost of driving was reflected in the highway network path building and
skimming steps, performed using the MINUTP PTHBLD program. Since paths are built based on
an impedance that includes time and operating cost, changes in the operating cost will affect paths.
In PTHBLD, this change is achieved by coding DCOST as 133, 167, or 338, for the three pricing
A-28 U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
• • to Reduce Transportation Emissions
levels. Table A-13 presents a summary of the travel choices affected by this change, as reflected in
the Case Study Model. ,
It should be noted that the Table A-13 shows no impacts of VMT fees on trip generation. This is
because the Case Study Model's trip generation module is not sensitive to pricing changes.
Model Step
Table A-13
VMT Fee Impacts
Impact
Highway Skims
Trip Generation
Trip Distribution
Mode Choice
Toll Diversion
Assignment
Emissions
Highway paths are nTodified to select paths that are shorter in distance,
to minimize the cost of travel.
No impact. , '
Composite impedance increases with distance (operating cost) and goes
up relatively faster for longer-distance trips; thus, travelers tend to
choose destinations that are slightly closer. Average trip distances
decrease slightly. ,
Cost of driving increases; driving alone decreases, carpooling decreases
slightly, and transit use increases. '
No impact.
VMT decreases due to fewer vehicle trips and shorter trip lengths;
reduced VMT means less congestion, resulting in faster average speed.
Emissions of all pollutants decrease due to less VMT and higher average
speed.
A.3.3.2 Fuel Tax
f , . - ' '
The tax on all motor fuels would increase. It was assumed that this increase would-be perceived as
simply more expensive gasoline and that this would influence the short-term decisions of travelers.
Thus, the fuel tax was modeled in terms of an increment to the automobile operating cost
parameter. , .. "
Three levels of fuel tax were analyzed: $0.20/gallon, Sl.OO/gallon, and $1.50/gallon. These were
converted into 1980 year dollars: $0.114/gallon, $0.570/gallon, and $0.855/gallon. These were
then converted into cents/mile by dividing by an average value for on-road fuel economy. A value
of 19 mpg was calculated from, the 1990 Nationwide Personal Transportation Study (passenger cars
only), but this was increased to 20 mpg for all scenarios by assuming that a short- to medium-term
(3-12 months) impact of higher fuel prices would be to encourage the purchase of more fuel-
efficient vehicles. The resulting additional cost per mile is then: $0.0057/mile, $0.0285/mile, and
$0.0428/mile, making the new automobile operating cost parameter: $0.1157/mile, $0.1385/mile,
and $0.15287mile, respectively. . . •
U. S. Environmental Protection Agency
A-29
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions . .
This new cost of driving was then input to the model in exactly the same manner as for the VMT
fee, as described above. The impacts were the same as noted in Table A-13, although at a lower
level due to the lower level of operating cost increase. ^
A.3.3.3 Age-Based Registration Surcharge
•' ,1 .
A surcharge would be placed on the annual registration cost of all vehicles and would vary with the
age of the vehicle. It was assumed that this fee would be collected annually and that travelers
would not be aware of it on a daily basis and therefore the surcharge would not influence their
short-term travel decisions. This surcharge was assumed to affect only the age mix of the vehicle
fleet by encouraging drivers to trade-in their older vehicles fo^newer ones. Such older vehicles
were presumed to be removed from the fleet (i.e., scrapped) and replaced with newer vehicles, in
roughly the same proportion as newer vehicles currently exist in the fleet. The MOBILESa default
age mix by vehicle type was used as the initial age mix.
The analysis methodology presented estimates only the impacts due to changes in vehicle mix by
age and assumes that no changes in VMT occur due to the shift from older vehicles to newer
vehicles. Since the fee is levied solely based on vehicle age, no changes in travel characteristics
are assumed except for changes in vehicle mix by model year. The fee can be structured in various
combinations of fee variation by each separate model years, or can be charged by certain ranges of
vehicle ages, or it can be varied among different vehicle types.
The changes in registration distribution by model year are estimated through a process in which
certain assumptions about the value of the vehicles in each age group and changes in car ownership
behavior are made to quantify the change in number of vehicles under each model year. The basic
principle in the analysis methodology is that the vehicles are dispensed or scrapped at a rate which
is proportional to the ratio of the registration fee to the market value for that particular model year
of vehicle.
General Assumptions: '
,.'•••. i .
• It is assumed that market value for a particular model year of vehicle is the same for
all the vehicles under that year.
• No drop in the latest model year vehicle registration is assumed because the market
value for these vehicles is much higher than the estimated registration fee.
• LDGT2, HDGV, LDDT and HDDV are assumed to have lower price elasticity than
other lighter duty vehicles because they are mostly used for business purposes and
• are less likely to be scrapped or replaced (see Table A-14) •
Motorcycles are assumed to have lower price elasticity because of their unique
nature.
Model Specific Assumptions:
(these are assumptions that could be replaced by more accurate data if available to the user)
• The same estimated market value is assumed for both LDGVs and LDGTls.
• The estimated market value for a late model year HDDV and Motorcycle are
assumed to be $60,000 and $7,500 respectively.
A-30 • , [/. 5. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Table A-14
Price Elasticities by Vehicle Type
Type
Elasticity
LDGV
1 -1.0 .
LDGT1
-1.0
LDGT2
-0.9
HDGV
-0.5
LDDV
-1.0
LDDT
-0.9
HDDV
-0.5
1 • MC
-0.75
The two major inputs to the analysis are the existing vehicle mix by model year and the average
market value for each model year of vehicle. In the Case Study Model,, the. MOBILESa default
vehicle mix is assumed for the base conditions, but these can be replaced by more accurate local
data if available or from any other updated versions of MOBILE. "
The other key input to the analysis is the average market value for each model year of vehicle
which could also be replaced with more accurate local data. For the Case Study Model, the average
market value for the latest model year (1993) light duty gasoline vehicle is assumed to be $17 922
(source: Automotive News, 26 May 1993, p. 61) and the value for older models is estimated from
this number assuming a price deterioration rate which is proportional to the age of the vehicle The
following summarizes the specific steps of the analysis:
Step 1 Percent Registration Fee Cost of the Total Market Value:
The percentage of the registration fee cost of the total market value for that model year
of car is calculated. This percentage of registration fee cost of the total value is assumed
to be equal to the percentage of, vehicles scrapped or replaced.
Percent Registration Fee = Registration Fee * 100 / Total Market Value
my
• my
Step 2 Estimation of Vehicle Scrappage or Replacement:
It is assumed that the vehicle scrappage or replacement percentage due to the registration
fee is same as the percentage of registration fee of the total market value. If the
registration fee exceeds the market value of the vehicle (as in the case of older cars) then
all the vehicles under that model year are replaced by newer model years. For example,
if the registration fee is 1% of the market value, then 1% of those model year vehicles
are assumed to be replaced by newer cars. This approach is applied to all the vehicle
model years except for the latest model year which is assumed to be independent and no
scrappage or replacement for this year is calculated.
To account for the variation among different vehicle types, especially the 'differences
between lighter and heavy duty vehicles, different price elasticities as'given in Table A-
14. are assumed for different vehicle types. Lower elasticities are assumed for heavy
duty vehicles since most of these vehicle types are generally used for business purposes.
The estimated, scrappage percentages as calculated above are corrected for these price
elasticities for each vehicle type. . • ', .•
U. S. Environmental Protection Agency
A-31
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Example: Assume a market mechanism with a $900 registration fee for cars between
ages 15 to 25, $600 for ages 5 to 15, and $300 for ages 0 to 5 years, the estimated market
value of a 24 year old car is $400, the registration fee for this model year vehicle from
the above description is $900, the percentage of registration fee is more than 100% of the
total market value, hence all of these model year cars are assumed to be replaced by
newer cars. Similarly, for a 7 year old model year car with a market value of $9,100, the
registration fee for this vehicle is 600, which is about 6.6% of the market value. Hence
6.6% of these model year cars are assumed to be replaced with newer cars.
Step 3 Estimation of Replacement Vehicles-
In the analysis all the scrapped vehicles are replaced with newer model year vehicles in
the same proportion as their existing vehicle distribution and weighted by appropriate
values to make1 the total of the distributed percentages to 100.
Step 4 Results from the Analysis
All the above described steps have been included in a spreadsheet which performs the
calculations and estimates the resulting vehicle registration distribution. ' The resulting
revised age mix was used in the PPAQ/MOBILE5a run instead of the default age mix.
Three levels of maximum surcharge were assumed: $100/year, $500/year, and
$900/year. The maximum surcharge was assessed on vehicles 20 years old or older, no
surcharge was assessed on vehicles 3 years old or newer, and an intermediate surcharge
was assessed on vehicles between 3 and 20 years old, varying with age (see Table A-9
for an example, showing the $900/year surcharge). Table A-15 summarizes the impacts.
Table A-15 s
Age-Based Surcharge Impacts
Model Step Impact ;
Emissions Emissions decrease due to shift in age mix of fleet from older to newer
vehicles.
A.3.3.4 Emission-Based Registration Surcharge
• ., " •'. . '•
A surcharge would be based on the estimated total annual emissions of all vehicles. An estimate
would be made of the annual emissions of each vehicle (by the 8 MOBILESa vehicle types), using
the default MOBILESa emission rates by age and the default MOBILESa annual mileage by age.
1993 was assumed as the base year, so a 25-year-old vehicle is of model year 1969, for example.
It was assumed that this fee would be collected annually and that travelers would not be aware of it
on a daily basis and therefore the surcharge would not influence their short-term travel decisions.
This surcharge was assumed to affect only the age mix of the vehicle fleet by encouraging drivers
to trade in their older vehicles for newer ones. Such older vehicles were presumed to be removed
from the fleet (i.e., scrapped) and replaced with newer vehicles, in roughly the same proportion as
newer vehicles currently exist in the fleet. In the definition of this measure, it was assumed that the
surcharge would be based on the estimated emissions for all vehicles of a certain age, not measured
, • ' ' " j _
A-32 • U.S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
emissions. Vehicle owners would not have the option of repairing and retesting the vehicle. The
MOBILESa default age mix by vehicle type was used as the initial age mix.
The emissions fee is similar to the age-based registration surcharge, the only difference being that
the fee is levied on the total emissions emitted by the vehicle. The emissions fee has certain
advantages over the registration surcharge in terms of equity, since this measure changes the price
based on total emissions rather than on age. Empirical evidence shows that newer cars travel much
more than the older cars, although this is more than compensated for by the net emissions per mile
(emissjon factors) being much lower for newer cars. This type of measure takes these
characteristics into consideration and distributes the price,uniformly over all vehicle age groups
based on their total emissions. Equity is also maintained across different vehicle, types since the
heavy-duty vehicles which have higher emission rates and VMT accumulation are assumed to pay
a higher fee compared to the lighter duty vehicles.
The changes in registration distribution are based on essentially the same technique as described in
the previous section on age-based registration surcharges. The differences in assumptions are as
follows: .
General Assumptions: .'•_'-.
• An average emission factor and VMT accumulation is used for all the vehicles under
a particular model year.
U. i. environmental Protection Agency
A-33
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions ,
Model Specific Assumptions:
(these are assumptions that could be replaced by more accurate data if available to the user)
• MOBILESa estimated average emission levels by each model year and default
registration distribution by age are used in the analysis.
The three major inputs to the analysis are the vehicle emission levels by model year, the annual
mileage accumulation by model year, and the average market value for each model year of vehicle.
In the Case Study Model, MOBILESa default annual mileage accumulation rates and emission
rates were usecl for the base conditions, but these can be replaced by more accurate local data if
avai lable or from any other updated versions of MOBILE.
The following two steps are used to estimate the change in annual user cost for this measure. From
that point, the analysis proceeds exactly as in Steps 1 - 4 of the age-based registration surcharge,
described above.
„ " ",''*,' ' ' ' ' ' ' , , 'I, , '
!;i ' • ' '•' ' .''',';. ; , • • •
Step 1 Estimation of Pollutant Tonnage:
The initial step in the analysis is the estimation of pollutant levels for each model year of
vehicle. The emission rates (g/mile) for each model, accounting for the appropriate
control measures such as inspection and maintenance, reformulated gasoline, etc., as
well as local conditions such as temperature and altitude, • are estimated using'
MOBILESa. A-single index of pollutant tonnage is calculated by combining the
estimated levels of HC, CO and NOX, using equivalency factors for each pollutant based
on previous EPA studies. The factors used in this project are shown in Table A-16. The
level of each pollutant multiplied by its equivalency factor is summed to get to the
combined emission rate for each model year.
Combined Emission Rate my(gms/mi)
= EJIC* HC level + ECQ* CO level + E^Qx* NOx level "
Table A-16
Pollutant Equivalency Table
Type
Equivalency
EHC
1.0
ECO
0.1
ENOx,
0.9"-
A-34 , . U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
The combined emission rate by model year multiplied by the annual mileage accumulation for that
model year vehicle gives the annual total tonnage of pollutants:
'»_ . -
Combined Annual Tonnage of Pollutantsmy . ,
= Combined Emission Ratemy * Annual Mileage Accumulation mv
Step 2 Estimation of Emissions Fee by Model Year:
The combined annual tonnage of pollutant as calculated from the previous step for each
model year is multiplied by the emissions fee set for the market mechanism. This gives
the additional price paid by each model year's vehicles due to the emissions fee.
Price due to Emissions Feemy
= Combined Tonnage of Pollutantmy(tons) * Emissions fee ($/ton)
(From this point, continue with Step 1, from Section A.3.3.3, described above.)
Application of the Methodology:
The calculations involved in the estimation of the changes in vehicle registration distribution due
to emissions fee are performed in a spreadsheet, an example of which is shown in Table A-17
These revised distributions go into the MOBILE input section of the PPAQ set up file.
This methodology is a sketch planning technique involving numerous assumptions and factors in
the process which are based solely on engineering judgment and are not substantiated by
empirical results. The methodology itself is in a rudimentary stage and requires further
enhancement with respect to the factors and values assumed in the process. The key assumption
regarding changes in vehicle ownership in response to external costs such as emissions fees and
registration surcharges should be verified. The assumed independence between changes in
vehicle ownership and changes in the annual vehicle mileage also bears further examination.
'}. S. Environmental Protection Agency
A-35
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
• _ '•; ' : ; .'. TableA-17 ' / .'. _ ' |
Sample Vehicle Age Mix Adjustment Spreadsheet (Emission Fee)
|UH»V|
r»<
2>
24
23
22
21
20
19
IS
17
16
IS
14
13
12
11
10
9
g
7
6
5
4
3
2
1
•xnK
n,KU
J«"
IK6»
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
I9S5
1936
1987
1988
1919
1990
1991
1992
1993
ntngfalarr
MOBILE
4,305
4,305
4,305
4,305
4,305
4,305
4,559
4.829
5.114
5,416
5,737
6,076
6,435
6.815
7,218
7,645
8.096
8.575
9,082
9.619
10.187
10.7S9
11,427
12.102
12.818
1.0 0.1 0.9 elasticity: -1.00
MOBILE MOBILE* MOBILES MOBILES combined combined
earn, mile* othllCtvl CO level NOx level level lons/yr
IV8,36y
174.064
169.759
165.454
161,149
156,844
152,539
147,980
143,151
138.037
132,621
126,884
120.808
114,373
107,558
100.340
92.695
84,599
76.024
66,942
57,323
47.136
36,347
24,920
12.818
8.89
9.44
9.28
6.17
6.10
6.03
5.19
s.oi
4.94
4.80
4.65
1.63
1.70
1.59
1.50
. 1.50
1.31
1.22
1.11
1.00
0.89
0.72
0.60
0.47
0.34
101.82
96.65
95.30
81.15
80.10
79.05
57.74
56.56
55.31
53.98
52.58
15.35
25.08
22.94
21.53
21.76
,19.21
17.62
15.33
13.63
11.83
9.72
7.76
5.68
3.47
M5
4.35
4.35
4.35
3.51
2.91
2.90
2.88
2.99
2,94
2.89
2.39
1.56
1.54
. 1.42
1.43
1.36
1.29
1.29
1.20
1. 10
1.00
0.89
0.78
0.66
22.99
23.02
22.73
18.20
17.27
16.55
13.57
13.32
13.16
12.84
12.51
5.31
5.61
5.28
4.93
4.97
4.46
4.14
3.81
3.45
3.06
2,59
2.18
1.74
1.28
07IT
0.11
0.11
0.09
0.08
0.08
0.07
0.07
0.07
0.08
0.08
0.04
0.04
0.04
0.04
0.04
0.04
0.04
0.04
0.04
0.03
0.03
0.03
0.02
0.02
fee/ton;
cAirnmed
— sw
$500
$60,0
$700
$900'
$1,000
S1.200
$1,500
51,800
$2,100
$2,500
$3,000
$3.500
$4,100
• $4,900
$5,700
$6,700
$7,900
$9,100
$10,500
$12,000
$13,500
$14,700
$14,800
$17,922
SI 0.000
ehungc
into
$1,089
$1,090
$1,076
$862
$818
$784
$681
$707
S740
$765
$790
$355
$397
$396
4392
'$418
$397
$390
$380
$365
$343
$308
$274
$232
$181
Icechg
„•/,„!
vulue
272.1%
218.0%
179.4%
123.1%
90.9%
78.4%
56.7%
47.2%
41.1%
36.4%
31.6%
11.8%
11.3%
9.6%
8.0%
7.3%
5.9%
4.9%
4.2%
3.5%
2,9%
2.3%
1.9%
1.6%
1.0%
avg age:
MOBILESa rcgiKlratiim mix by year
leaning 1 July 1990}
base-/. new 54 eum.%
by year by year %lcft loll
1.10%
P.30%
0.40%
0.50%
0.60%
0.80%
1.10%
1.50%
1.40%
1.90%
2.40%
3.70%
4.70%
5.40%
5.00%
5.10%
5.00%
5.60%
7.70%
8.10%
8.40%
8.20%
8.30%
7.90%
4.90%
100.0%
8.2
0.00%
0.00%
0.00%
0.00%
0.07%
0.20%
0.52%
0.86%
0.89%
1.31%
1.79%
3.53%
4.53%
5.33%
5.06%
5.22%
5.22%
5^93%
8.25%
8.79%
, 9.23%
9.13%
9.34%
9.01%
5.79%
100.0%
7.1
1.10%
0.30%
0.40%
0.50%
0.53%
0.60%
0.58%
0.64%
0.51%
0.59% .
0.61%
0.17%
0.17%
0.07%
-0.06%
-0.12%
-0.22%
-0.33%
-0.55%
-0.69%
-0.83%
-Q.93%
-1.04%
-1.11%
-0.89%
0.0%
1.10%
1.40%
1.80%
2.30%
2.83%
3.43%
4.02%
4.66%
5.17%
5.76%
6.36%
6.53%
6.70%
6.77%
6.71%
6.59%
6.37%
6.05%
5.50%
4.81%
3.98%
3.05%
2,00%
0.89%
0.00%
avg fee:
-new/buse
— CT?T
0.0%
0.0%
0.0%
11.5%
24.5%
46.9%
57.1%
63.9%
69.2%
74.8%
95,4%
96.4%
98.8%
101.1%
102.4%
104.4%
105.8%
107.1%
108.5%
109.9%
111.3%
112.6%
114.1%
118.2%
$360
Nqte: "LDGV" = light-duty gasoline vehicle
A.3.4 Modal Subsidies
Analysis of modal subsidy measures involves reducing cost to travelers who use a mode that is
promoted by subsidy. Typical subsidies involve incentives for transit (fare reduction, on-site
transit stops), cost incentives for ridesharing (cost reimbursement, reduced parking costs,
preferential parking locations, toll reductions) and transportation allowances (parking " cash out").
Two levels of transit fare reduction were studied:
• 50% reduction in transit fares
• 100% reduction in transit fares (i.e. free transit service)
A reduction in transit fares influences a traveler's choice of destination zone, travel mode and
(indirectly) highway path. Destination zone choice is affected by changing the computation of
combined impedance which is an input to the trip distribution model. Travel mode choice is
affected in the mode choice model by making transit more attractive and thereby increasing transit
share. Highway paths are affected by assigning fewer vehicle trips to the highway network which
, .reduces congestion effects and causes path changes to occur in the traffic assignment model.
There are two "matrix" files that reflect zone to zone transit fares in 1980 dollars. These files
. (walk to transit fare matrix and drive to transit fare matrix) are modified with the MINUTP
program MATRIX. For the 50% reduction, all cells are factored by 0.50. For the 100% reduction,
all cells are replaced with zero. These revised fare matrix files are specified in the model setup and
A-36 , U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
used during both the combined impedance calculation and the mode choice estimation steps. Table
A-18 summarizes the impacts of this scenario on the travel choices:
Model Step
Table A-18
Modal Subsidy Impacts
Impact
Trip Distribution
Mode Choice
Assignment
Emissions
The composite impedance decreases for those paths that show a reduction
in transit fare under this scenario. This modifies the distribution of trips in
favor of transit accessible zone pairs.
Transit costs decrease causing transit shares to increase and vehicle trips to
decrease.
VMT decreases due to fewer vehicle trips. This can also result in some
increases in speed for certain highway facilities.
HC, CO and NOX emissions all decrease due to the decrease in VMT.
A.4 Other Model Enhancement Issues
A.4.1 Addressing the Unknown Sensitivity to Pricing
• The sensitivity of urban travel demand to pricing has not received much attention in the past. The
best known analysis has been the "Curtin Rule", which states that a 10% increase in transit fare
leads to a 3% drop in transit ridership. In almost all other applications, the price of transportation
to the traveler has often been assumed to change at the same rate as inflation and thus be of no
more or less importance in the future than today.
The rapid rise in gasoline prices in the mid-1970's and early 1980's, combined with a flurry of
activity in rail transit planning created a need for mode choice models which were sensitive to auto
operating cost to the user. There is now over a decade of experience with such models, and
transportation planners have some confidence that adjusting this parameter to account for increases
in the price of driving, as described in sections A.3.3.1 and A.3.3.2, produces reasonable results
that are not too great an extrapolation beyond present experience. A similar comment can be made
for the price of parking, which has been identified in the past ten years as a significant influence on
mode choice.
However, when it comes to more innovative pricing strategies, such as congestion pricing, planners
in the U.S. have little or no experience on which to draw. Moreover, there is an intuitive and
logical belief that in some way, strategies such as congestion pricing will have a travel impact far
beyond the choice of travel mode. But the problem remains: without actual experience from which
to observe and compare, how can demand modelers estimate the impacts of such measures without
falling back on theory?
A powerful argument can be made that " before-and-after" studies offer the best evidence of
travelers' response to pricing measures. The problem, of course, is that this assumes that a strategy
U. S. Environmental Protection Agency
- A-37
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
has actually been implemented, in a relatively controlled manner, and without other confounding
factors that could also influence the outcome. Careful analysis of the last gasoline crisis might
yield some insight into motorists' response to steep and rapid increases in the user cost of driving,
assuming one could filter out the impact of the apparent shortage of supply and the psychological
(and real) effects of long lines at the pump. Analyses of the ridership effects of existing differential
peak/off-peak transit fares, before and after implementation, might produce some insight into the
cost sensitivity of trip start time.
- . '• / ' ' -!•''
Another criticism of the pricing parameters currently used in models is that they are based on
cross-sectional surveys. Such parameters are not based on what a fixed group of people did over
time, but on what many people did at one time. Current model calibration theory says that if you
can take a "snapshot" survey of enough people with different enough travel conditions and
choices, you can generalize enough to draw conclusions as to how one traveler would respond to
changes over time. Although this is how almost all mode choice models are presently calibrated,
there is some sentiment among planners that it is an insufficient basis for identifying true traveler
sensitivity to pricing.
One alternative which is gaining popularity is the panel survey, which follows a fixed group of
households and individuals over time and attempts to measure changes in their travel behavior in
response to actual changes in their travel conditions. In theory, this should produce more accurate
cost sensitivities, but in practice such surveys are very expensive and difficult to conduct. Worse,
they generally require 18-24 months or more, to allow for three or more survey "waves". Still,
such surveys have been conducted in Seattle, Washington and Montgomery County, Maryland and
hold considerable promise for the future.
Another technique that may be useful is the stated preference survey, in which people are asked
how they would respond to a series of hypothetical situations. Such surveys are not as reliable as
surveys of observed behavior, because of the large gap between what people say they will do
(especially under hypothesized conditions which may be far outside their normal experience) and
what they actually do. Still, if performed under carefully controlled conditions, stated preference
surveys can produce estimates of pricing sensitivity for travel decisions, when no other option is
available.
Some mode choice models have been developed by transferring model coefficients from another
urban area. The disaggregate nature (i.e., modeling individual choices) of logit mode choice
models generally supports this. Analyses of coefficients from different areas indicates that the
variation in these coefficients is not terribly great, suggesting that careful transference is possible.
Perhaps the most critical issue here is to be aware of how the independent variables were developed
for the original model and to maintain consistency in the new model.
Another major trap to avoid: in transferring pricing coefficients, one must be aware of the year for
which the original model was developed and adjust such coefficients for inflation as needed. For
example, if the original model was calibrated using 1980 prices, then the new model must either
also convert its prices to 1980 year dollars, or adjust the original cost coefficients to reflect the
effect of inflation over the years. For example, the Case Study Model uses prices expressed in
1980 dollars. To convert its operating cost coefficient (0.004) to 1993 dollars, it must be multiplied
by 0.57, which is the ratio of Urban Consumer Price Index values for 1980 and 1993, resulting in a
1993 equivalent value of 0.0023.
A-38 U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
, to Reduce Transportation Emissions
In some cases it might not be possible to derive a particular cost sensitivity from any of the above
sources. In the last resort, it will generally be acceptable to synthesize a coefficient value and then
perform a series of sensitivity tests on it. .The value of the parameter would be varied above and
below the hypothesized value, the rtiodel would be re-applied, and the results (mode choice, trip
length, VMT, etc.) would be charted for each variation. Comparing this variation with theoretical
values and a great deal of common sense from a variety of external reviewers can often help
identify an acceptable value for a parameter. Although such an analysis, even though carefully
reasoned and conducted, cannot conclusively identify the proper value for an unknown parameter,
it may provide a suitable interim result until better information can be found.
A.4.2 Model Development/Application
In revising travel models to be more sensitive to pricing, several development and application
issues must be considered, including the following: '
Connectivity and Feedback: Asrthe Case Study Model demonstrates, in order for pricing to affect
each simulated travel choice, the model set must be "connected"; i.e., air travel prices must'
influence each, model step. In a four-step model, one implication of this is that the model must be
applied "backwards" and "forwards", as in the Case Study Model. This increases model
application time and resource requirements (e.g., hard disk space) but is probably unavoidable
within the current sequential model structures. Advanced programming, such as combining the
mode choice and distribution steps, can minimize this, but only at the cost of increased overall
complexity. '
A related issue is the question of achieving model equilibrium, usually defined as the consistency
of travel speeds throughout the entire model. It is unclear if or how these improvements to handle.
pricing would affect the potential and the need to iterate the model set until speed equilibrium is
achieved. It may be that price-sensitive models need to use a different definition of equilibrium,
possibly one that involves achieving consistency of equivalent prices throughout the model
(equivalent price being defined as actual price plus the traveler's cost of time).
Effects of Inflation: As noted above, the. sensitivity of travelers to price, as reflected in the
model's coefficients, is expected to change over time with the general rate of inflation. Price
coefficients that are borrowed from other areas (or from different years) must be'adjusted to
account for the effects of inflation. This means that cost coefficients can be expected to decrease
numerically.over time, as the purchasing power of money erodes, more or less steadily, over time.
Extreme care must be taken in forecasting price information. It is commonly assumed that all
travel prices will rise in the future at the same rate as inflation, meaning that there is no change in
the relative price between travel choices.. This is not always true. For example, transit fares in
many cities have not kept pace with inflation. If a forecast assumes that that trend will continue
into the future, then that must be reflected in reduced fare values for a future model run. Similarly,
as Figure A-5 shows, the trend in automobile operating cost has been dropping since after World
War II, when expressed in 1994 dollars, and the trend is for continued declines in real dollar terms.
U..S. Environmental Protection Agency A S
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Figure A-5
Trends in Passenger Car Operating Cost
Vehicle Operating Cost Trend and Forecast
18.00 .
«, 16.00 •
« 14.00 .
§ 12.00 .
te
S" 10.00 .
I
<•> 8.00 .
6.00 -
4.00 -
\
^\'
.
A
- \
\/
X
v^ —
/v
f \
^c
""-
„
•---..
— — ••"'
1950
1960
1970
1980
1990
2000
2010
2020
Historical
Forecast: DOE gasoline price- •
Forecast: gasoline price trend
Different Strategy Specifications: -Some market-based strategies have many different variations
which affect how the strategy is implemented and may affect how it is perceived by the public.
However, it is unlikely that any travel model will be sophisticated enough to be sensitive to such
nuances. It is enough of a challenge to create models which are properly sensitive to a change in
actual travel price, much less be influenced by the context in which the change is implemented.
For example, two ways of increasing the user cost of driving are: 1) increase the gasoline tax, and
2) convert the cost of motor vehicle insurance from a semi-annual payment to a pay-at-the-pump
fee, as has been proposed in California. Although these two measures are quite different from an
implementation perspective, it is questionable if motorists would perceive them as different (as
long as the increased price per gallon was the same). It would seem reasonable to estimate their
quantitative effects on travel in the same manner.
Model Refinement/Development: Some of the enhancements described above can be made fairly
easily, while others require considerable effort. An example of the former would be the Case Study
Model feature which permits the separate consideration of LOV and HOV parking price, thus
allowing the testing of differential parking price by zone by vehicle occupancy. It would be very
easy to incorporate this into a mode choice model which already uses parking price and splits trips
by occupancy level, which quite a few mode choice models already do. This would require only a
few hours of programming to revise the mode choice application procedure; no new calibration
would be required.
At the other end of the spectrum, modifying the distribution model to use composite impedance is a
much more significant task, requiring the development of a new program (or modification of an
A-40
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
• • ' ' ' to Reduce transportation Emissions
existing one) to calculate composite impedance and the complete re-calibration of the distribution
models to use the new impedance value. This could take from a few weeks to a few months.
It is important to recognize that the improvement of the pricing sensitivity of most existing travel
models is a difficult undertaking which requires time, data, computing resources, and staff
expertise which are not universally available in local planning agencies. Such a task is made all the
more difficult since agencies are usually expected to pursue this work in addition to their usual on-
going planning and forecasting duties. One way to organize the workload is to establish formal
parallel "tracks", one for on-going work and another for new model development, to ensure that
neither effort is neglected. ,
Of course, agencies which are in the process of updating their models or developing new models
anyway should take all reasonable steps to improve the sensitivity of their models to pricing. The
features of the Case Study Model and the enhancements described in Section A.5 should be
strongly considered. Once a commitment is, made to develop new models, the incremental effort to
make such models reasonably price-sensitive is not very large.
A.5 Summary
A.5.1 Summary of Desirable Model Features
The analyses of this appendix lead to the following major conclusions:
* Improve the Four-Step Process: Although many current four-step models are now
largely inadequate for modeling most pricing measures, this process can be modified
to improve its sensitivity to pricing strategies, as illustrated in this appendix. Such
modifications will address most of the criticisms of this process and enable it to
perform adequately until the next decade, at which time a new model structure may
be sufficiently developed and tested to serve as a replacement.
• Improvements are Not Easy: A few of the model enhancements described above are
fairly simple, quick, and inexpensive but most of them are not. Most of these
changes will require a recalibration of a region's entire travel model chain, which is
usually a multi-year effort with a budget in the high six figures. If a region is
planning to pursue such an effort anyway, then incorporating these improvements
does not entail a significant incremental investment of resources.
* Improvements are Necessary Anvwav: Most of the model enhancements discussed
here would help improve the sensitivity and credibility of travel forecasts in contexts
other than pricing strategies.. For example, analyses of TCMs such as changes in
transit supply, land use, ridesharing policy, and HOV lanes would benefit from these
enhancements. Thus, there are several reasons to enhance existing models.
U. S.. Environmental Protection Agency , .,
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Recognize the Unknown Sensitivities: The fact that the sensitivity of certain travel
choices to some pricing measures is still largely unknown cannot be used as an
excuse to eschew their analysis. Additional research, transferable parameters, and
sensitivity analyses can be used to make a reasoned estimate, leaving room for future
revisions to model results when better information is forthcoming.
A-42
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
References
l.
•2.
3.
5.
6.
7.
W. Allen, User's Guide for the MWCOG Mode Choice Application Program
Microcomputer Version, for the Metropolitan Washington Council of Governments, June
J. Heisler et al., Estimating Toll Diversion Using Existing Transportation Planning
Software, presented at the Second Conference on Application of Transportation Plannine
Methods, April 1989.
The Phase III Travel Demand Forecasting Model: A Summary of Inputs, Algorithms and
Coefficients, METRO, Portland, Oregon, January 1994.
W. Allen, Model Calibration for Employee Commute Options Planning, presented at the
Fifth Conference on Application of Transportation Planning Methods, April 1995.
K. Train, Qualitative Choice Analysis, Theory, Econometrics, and an Application to
Automobile Demand, The MIT Press, Cambridge, Massachusetts, 1986.
Greig Harvey and E. Deakin, A Manual of Transportation-Air Quality Modeling for
Metropolitan Planning Organizations, for the National Association of Regional Councils
Revised Draft, November 1992. '
B. Spear, New Approaches to Travel Forecasting Models: A Synthesis of Four Research
Proposals, Travel Model Improvement Program, John A. Volpe National Transportation
Systems Center for the. U.S. Department of Transportation, and U.S. Environmental
Protection Agency, January 1994.
W. Allen and G. Davies, A New Method for Estimating Cold Start VMT, presented at the
Fourth Conference on Application of Transportation Planning Methods, May 1993.
. U. S. Environmental Protection Agency
A-43
-------
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Appendix B
The STEP Analysis Package: Description and Application Examples
B.I Overview
This appendix discusses STEP, a travel demand modeling package designed for planning
applications and policy analysis, and shows how the model was used to analyze a wide range of
transportation pricing measures for four California metropolitan areas: the San Francisco Bay
Area, Los Angeles, San Diego, and Sacramento. We begin with a description of STEP then
discuss how we used the STEP models to examine the potential impacts of pricing strategies on
travel behavior, traffic volumes, energy consumption, and emissions for two analysis years 1991
and 2Q10. We also briefly describe other methods used to supplement the STEP analyses
including estimates of the elasticity of vehicle fuel economy with respect to fuel price, and detailed
regional highway network models. A series of tables presentthe analysis results.
B.2 Introduction to the STEP Package
STEP is a travel demand analysis package composed of an integrated set of travel demand and
activity analysis models, supplemented by a variety of impact analysis capabilities and a simple
model of transportation supply. STEP is based on microsimulation - a modeling technique which
uses the individual or household as the basic unit of analysis rather than dealing with population
averages, (cf. Orcutt, 1976). STEP results are aggregated only after the individual or household
analyses are completed, allowing the user great flexibility in specifying output categories. This is'
more commonly referred to as "sample enumeration" in the literature, or as "discrete choice
analysis."
STEP has been applied in a number of Bay Area studies over the years, and has been adapted for
use in studies in Los Angeles, Sacramento, Chicago, and the Puget Sound region (Seattle).
Applications can proceed with model re-estimation specifically for the region - essentially, by
creating a completely new set of models for STEP - but to date nearly all applications outside the
Bay Area have relied on extensive re-calibration of the default (Bay Area) models plus a limited
amount of re-estimation as needed to match local conditions.
Several features of STEP supported its choice as the basic modeling tool for the case studies
presented here. STEP's regional, subarea, and corridor-level analysis capabilities fit well with the
scope and scale of the policies under consideration. Its model formulations can represent a
comprehensive set of possible price effects, and its models display linkages consistent with travel
behavior and pricing theory. Its use of microsimulation makes it possible to address many of the
questions about equity and the distribution of impacts that frequently arise in debates about pricing.
J. S. Environmental Protection Agency • ' B 1
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Finally, it is far faster to calibrate STEP for a region than to upgrade the regional models to include
pricing variables, and far faster and less expensive to run STEP than to apply regional models.
STEP'S data analysis capability is another important asset in pricing studies. STEP'S
microsimulation formulation permits the package to be used as a survey tabulation technique
employing sophisticated data transforms and linkages. For example, many travel surveys contain
detailed information about the vehicles each household owns and indicate which vehicle was used
for each trip made on the survey day(s). Using STEP, these vehicle data can be tabulated so that
exact usage patterns by model year or vehicle type can be determined. They also can be related to
personal and household characteristics to yield useful information about low-income households'
dependence on old vehicles and their contributions to vehicular emissions.
STEP itself was originally developed for sketch planning analyses in the San Francisco Bay Area
(Harvey, 1978). Since that time, all of the models in STEP have been completely re-estimated and
additional models addressing location choice, time-of-day of travel choice, and congestion effects
have been added. The most recent formulations are nested logit. A number of versions of STEP
are currently available, including options that permit the analysis of activity data as well as travel
data, and versions that use either MOBILE or California EMFAC emissions data.
STEP'S models are applied using actual or forecast data on household socioeconomic
characteristics, the spatial distribution of population and employment ("land use"), and
transportation system characteristics for the selected analysis year(s). The socioeconomic
characteristics of a sample of households and its members are usually taken from a regional travel
survey or from the U.S. Census Public Use Microdata Sample (PUMS). Population, number of
households, and employment by category (type) are taken from the regional " land use" data base.
Transportation level-of-service data (times and costs) are derived from the region's travel model
system. The land use data are provided to STEP for subareas (which could be zones, districts, or
corridors) and for the region as a whole; the level-of-service data are provided in the form of large
matrices of interzonal times and costs. STEP then reads through the household sample, attaching
level-of-service and land use data to each household record as necessary. For each household,
STEP uses its models to predict a daily travel and activity pattern for each individual in the
household. Finally, household travel is summed up and household totals are expanded to represent
the population as a whole.
.'!'"'' .' I ',
STEP can analyze any change in the population or in the transportation system that 1) can be
represented in terms of the variables in its models and 2) can be associated with a specific
geographic area or grouping of households. Testing the effect of a change in conditions or policies
is a simple matter of re-analyzing the household sample using the new data values, and comparing
the results with previous outputs. For example, a new highway or new transit service can be
represented by changed travel times and costs for the areas served; a parking price increase can be
represented by an increase in out-of-pocket costs; an increase in income in a particular area or for a
particular population subgroup can be represented by editing the household file to incorporate the
revised incomes. Along similar lines, future years can be represented through proportional
factoring and re-weighting of survey observations to reflect expected regional trends, or can be
based upon a more sophisticated microsimulation of household changes based on cohort survival
and other methods of demographic forecasting.
B-2 • , U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
. . to Reduce Transportation Emissions
The sampling framework preserves the richness of the underlying distribution of population
characteristics and permits tabulation by any subgroup with sufficient observations to be
statistically significant. For example, the results can be disaggregated by income level and age
which would allow an assessment of effects for, say, various income quintiles among the retired
population. This is a significant advantage over an aggregate model, which uses zonal averages for
most socioeconomic-economic data.
STEP maintains its quick response capability while achieving great detail in representing behavior
in part by reducing its detail in representing transportation networks. STEP does not have an
internal transportation network representation and traffic assignment model, so changes in level of
service resulting from changes in demand must be calculated in another way. Both an approximate
method and a more detailed and conventional network modeling approach have been developed for
this purpose. , , ,
To approximate the effects of changes in demand on network performance and vice versa, a simple
routine for estimating level-of-service was incorporated into STEP in the early 1980s'(Harvey,
1993). The simplified level of service model uses peak and off-peak travel times and base case
demand estimates to calibrate a supply function for appropriate spatial groupings of trips (i.e., trips
in broadly-defined "corridors"). The basic form of this equation is: t=a*(l-f-[V/C]Ab), where t is
the travel time in minutes per mile; V is1 the volume in vehicles per hour; C is the "capacity" in
vehicles per hour; and a'and b are coefficients fit to each corridor, For each change in demand, the
calibrated function can be used to compute a new "equilibrium" in the corridor (This is generally
known as the Bureau of Public Roads function for determining travel time on congested links).
While the simplified level of service model is useful for many analyses, it is intended only as an
approximation of changes in network performance and is likely to be inadequate in cases where
large network perturbations could occur or where specific route choice changes are at issue. When
network questions are critical, STEP must be used in conjunction with a more detailed network
model.
In the typical application, STEP is "interfaced" wjth the region's detailed highway network.
STEP'S modal trip outputs are summarized on a district-to-district basis (a district is defined as an
aggregate of the zones for which land use data are reported; for example, in the Los Angeles region
there are 1555 zones and 55 districts defined/by the regional agency). If the policy under analysis
results in any significant differences from the base-case district-to-district trip tables, the
differences,are used to factor the zone-to-zone trip tables in the aggregate model system.' The
network models are then run using these new trip tables, and the results are fed back into STEP as a
revised set of level of service inputs. Iterations continue much as is done in a conventional travel
model system until an acceptable level of convergence is achieved. Transit networks also may
need to be run in conjunction with STEP in cases producing significant differences in highway
travel times of a sort likely to affect bus operations.
For certain transportation pricing measures, such as proposals to toll specific links or facilities in a
network, use of the detailed network models together with STEP is pf particular importance. For
the case studies presented here, we used the network models for Los Angeles and the Bay Area to
test the route choice effects of congestion pricing, interfacing in the riianner described above with
the versions of STEP developed for each region. '
U. S. Environmental Protection Agency
_
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Three major variants of the STEP model system are shown in Figures B-la, b, and c; Figure B-lb
shows the version used in our pricing studies. The basic data requirements of the STEP model are
summarized in Figure B-2. A typical sequence of activities for a STEP application is shown in
Figure B-3.
Figure B.la:
STEP Model Structure with Basic MTC Models
Household Characteristics Dependent on Travel:
Number of Autos Owned
Daily Household Trip Choices:
Trip Frequency (HBW, HBS, HBO, NHB)
Trip Destination (BS, HBO, NHB)
Trip Mode Choice (HBW, HBS, HBO, NHB)
Note:
HBW - Home-Based Work Trips
HBS - Home-Based Shopping Trips
HBO - Home-Based Other Trips
NHB - Non-Home Based Trips
B-4
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Figure B.lb:
STEP Model Structure with Enhanced MTC Models
Major Household Location Choices:
Residential Location
Primary Workplace Location for Each Worker
Household Characteristics Dependent on Travel:
Number of Autos Owned
Daily Household Trip Choices:
Trip Frequency (HBW, HBS, HBO, NHB)
Trip Destination (HBS, HBO, NHB)
Trip Mode Choice (HBW, HBS, HBO, NHB)
Time Characteristics of Household Travel:
Work Arrival Time
Transportation System Performance:
Highway Corridor Delay
Note:
HBW - Home-Based Work Trips
HBS - Home-Based Shopping Trips
HBO - Home-Eased Other Trips
NHB - Non-Home Based Trips
. Environmental Protection Agency
B-5
-------
l
•a1 {§
I
•i
T3
O
§
•s
o
S
I
s
•«-»
CO
~8
TJ
O
CO
n
w
I
>> o o < ^
f S 2 2 c
C > O (U O
ia ^ xi -o ™
2 I I I S
•g'i
Working-Age Adults i
Not Employed •
1 Is
Non-Work Chains I
Number of Chains . m
Number of Aclivilies H
ILocalion of Activities M
Mode 1
1
me
Pa
2 u
111
'SPl
Itlt
1/0 3 M fc1
1111
££££
1
ce
T3
§
_ -a
u
T3
O
E °
ttl
-
o ta
O o
-C O
,55 j
E ~ ™ >> .2 J3 .2
-— -O -^; H en e M
s
CO
.
CO ^"S o o ^
13 2
Q
3o
z 2
It
•s s
I^J Q
? S
II
S a
£, a>
1?
Il
& ?
*•
I-a
'
CD
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Figure B.2:
Primary STEP Data Requirements
Basic Data:
Regional Household Travel Survey
1990 US Census Public Use Microdata Sample
For the Survey Year:
Geography
land area, population, housing stock
for tracts, zones, and/or districts
Network Level-of-Service
highway, transit
a.m. peak, p.m. peak, off-peak
times, costs;
For Each Forecast Scenario:
Geography
land area, population, housing stock
for tracts, zones, and/or districts
Network Level-of-Service
highway, transit
a.m. peak, p.m. peak, off-peak
times, costs
Economics
expected real income growth
expected real fuel price growth
U. S. Environmental Protection Agency
B-7
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Figure B.3:
Sequence of Activities for a STEP Application
Prepare Survey Data for Initial Analysis:
Screen Survey for Unusable Observations
Reweight Survey to Match Key Census Demographic Characteristics
Reformat Network and Geographic Data to Match Database Requirements
Assemble and Test Database
Calibrate the STEP Models:
Run STEP for Base Conditions
Compare STEP Calculations with Actual Household Travel Patterns
Adjust Constants, Beginning with Upper-Level Models, and Rerun STEP
Iterate the Adjustment Process Until the Overall Fit is Acceptable
Prepare STEP for the Forecast Scenario:
Adjust Household Data to Reflect Changed Conditions
Income
Subarea Population
Household Type Cohorts
Reformat Network and Geographic Data to Match Database Requirements
Assemble and Test Database
Run STEP to Create a Base Case
Test the Policy Alternative(s) with STEP:
Alter the Database as Necessary to Represent the Policy Option
Run STEP to Estimate the Effects of the Policy Option
Post-Process the STEP Outputs
Repeat the Analysis Sequence for Variants of the Policy Option
B-8
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
' .. • to Reduce Transportation Emissions
B.2.1 Transferring STEP Models to Other Regions
Although each application of STEP could utilize models estimated specifically for the region being
studied, a less costly approach is to transfer models estimated in one region to ahother. In the case
studies presented here, STEP models originally estimated for the Bay Area were transferred to Los
Angeles, Sacramento, and Sail Diego, with detailed calibrations and a moderate amount of model
re-estimation in each case. ,
x
Procedures for transferring models and evaluating their performance are well established - in fact,
many regions routinely use one or more transferred models in their regional model systems The
procedure for transferring. STEP to a new region follows much, the,same.general sequence of
actions and so will be discussed only briefly here.
To transfer STEP to a new region, the required data first must be set up. The region's 'most, recent
household travel survey is obtained, and checked (incomplete observations are excluded), and
network data and land use data for the year of the survey are extracted from the regional modeling
data bases. The data are then linked and a trial simulation Is carried out to determine how closely
the models to be transferred match the actual travel patterns in the survey data. Invariably, a
sequence of adjustments to model constants (and sometimes to a small number of coefficients) is
necessary to achieve an acceptable replication of the base travel pattern. These adjustments serve
both to capture actual differences in behavior and to compensate for variation in the way regional
planning agencies define certain variables such as transit wait times, income ranges, and specific
categories of land use. .
Once an acceptable simulation of the survey year,(the '^base case") has been obtained in this
fashion, STEP should closely reflect travel conditions and behaviors in the region to which it is
being transferred, and consequently can be used with local data and forecasts for the full range of
modeling applications.
B.3 Applying STEP to Pricing Measures
B.3.1 Overview
x The application of travel forecasting models to specific pricing policies is rarely a straightforward
" matter. In nearly every case, both the models themselves and the available data, bases impose some
limits on the policies that can be tested. For example, the regional transportation data bases (and
models based on the data bases) typically lack information about the variation of parking price in
each zone, and may have only approximate information about the vehicle Used for a specific trip.
In cases where such details would play a large role in determining the impact of a policy being
studied, only an approximate estimate of the policy's effects can be formally estimated through
modeling: the analyst must devise a means of representing the policy as well as possible given the
models and data, and must be prepared to make off-line calculations and adjustments to improve
the realism of the analysis, or to do further analyses after gathering additional, information.
•Some discussion of implementation scenarios is necessary simply to determine how a' proposed
pricing concept should be analyzed; clearly, however, much more attention to specifics would be
needed in an actual implementation. In our analyses, for example, we implicitly assume that
evasion or outright fraud would be insignificant, hence the measures would be fully effective as
U. S. Environmental Protection Agency , . •
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
proposed. For most transportation pricing measures, monitoring, enforcement, and audits would be
needed to assure that.
" ',•.•' - i • . , . •
In our case studies we found that it generally was possible to define transportation pricing
strategies in ways that were tractable from an analysis perspective and also yielded information
which is helpful in thinking about policies as they might actually be implemented. The use of
advanced modeling capabilities, along with the availability of good data, made it possible to
explore behaviors that often would be omitted from a more conventional analysis. Nevertheless,
the analyses did require a number of assumptions, and they have certain limitations that must be
acknowledged and taken into consideration in policy evaluations.
'; . ., , , , , ' • ,';,'_ •» I ' ,
The following sections detail how the pricing concepts analyzed in our case studies were specified
and analyzed. In each case, the underlying rationale for the pricing concept is stated, a specific
pricing measure is defined, modeling assumptions to represent the pricing measure are outlined,
and key implications of the assumptions are noted.
• , . " . i
B.3.2 Congestion Pricing
Congestion occurs in the highway system when more vehicles attempt to traverse a segment of road
per unit of time than that segment can accommodate! Such a location is called a bottleneck.
Congestion pricing builds on the simple realization that travelers are sensitive to the cost of travel;
a fee levied at a bottleneck will divert some vehicles from .the traffic stream, reducing congestion.
The diversion of a specific vehicle might be to a different route, time-of-travel, mode, or
destination; it could reflect a trip foregone; or, over the long run, it might follow from a change in
residence or workplace location.
Two major design issues arise in thinking -about how to use pricing to manage congestion at a
bottleneck:
• Price level - Price can be varied over a wide range to achieve different levels of traffic
improvement. Economic theory tells us that price should be set to reflect the social cost
caused by the marginal user at a bottleneck, less the average variable cost already paid
by users. While this should be the clear goal of any congestion pricing application,
considerations of implementation and management ease may point toward a simpler
price criterion based, for example, on achieving and maintaining a conventional level of
service measure from the literature of traffic engineering. We know that the "optimal"
level of congestion reduction will be unique at each bottleneck, but it is much easier to
explain a generally applicable congestion reduction goal in the policy-making process,
and easier to implement and manage facilities based on observed performance. Hence,
the actual criterion for setting the congestion price may well be framed in terms of
standard traffic leyel-of-service metrics (e.g., B, C, D, E). For similar reasons of
simplicity and clarity, specific prices might be chosen to reduce the amount of change-
making required (rounded to the nearest 25 cents or to the nearest dollar, for instance),
although with modem road pricing technologies this would not be strictly necessary.
Periodic adjustments in price are likely to be needed to maintain effectiveness, and they
too would likely be done in simple, rounded increments of 25 cents or a dollar, unless
electronic toll collection were in place.
B-10 U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
• Period of application - Some economists have argued forcefully that congestion prices
should change dynamically in response to traffic conditions, perhaps. varying from
minute to minute to achieve the optimal reduction in congestion. However, few
seriously believe that such a dynamic scheme would be implemented any time soon, for
several reasons: 1) the practical difficulties of creating, testing, and maintaining'the
hardware and software required for such a system; 2) the unresolved theoretical question
' of whether a truly dynamic system would produce a stable set of prices; 3) the strong
revealed preference .of travelers for predictable conditions, even if the price of
predictability is a somewhat higher average time or cost; and 4) the question of how to
treat incident-related delay in a dynamic pricing environment An initial congestion
pricing scheme more likely would involve prices that can be. explained through relatively
simple signage and do not vary from day-to-day (though weekend-weekday and seasonal
variations might be both desirable and feasible). Hour-to-hour variation might, however,
be used to avoid large price increases and decreases,at the peak / off-peak boundaries,'
and might be designed as a pyramid of prices centered on each peak hour in order to be
relatively easy for the driver to remember.
In addition to these basic design issues for pricing at a bottleneck, there is a question about how
widely congestion pricing would be applied in the highway network. While pricing would be
easiest to implement on litnited access facilities, spillover from priced freeways to.unpriced
arterials and collectors could be a problem in some locations. 1 Local communities seem unlikely
to tolerate significant traffic diversion to the facilities under their jurisdiction, and could be
expected to oppose freeway pricing schemes if they created or worsened congestion on local roads.
The localities might, however, accept a broader-based pricing plan which manages traffic on a
systemwide basis, especially if part of the revenues were returned to affected jurisdictions.
Widespread implementation of congestion pricing hence could mean pricing both freeways and
parallel routes where significant delay appears.
The congestion pricing measures tested in our case studies were designed to reflect these
observations about the policy environment. We assumed that some form of electronic payment
system would be used rather than toll booths, so that there would be no stopping to pay tolls.
Prices were applied everywhere delay appeared in the highway network (as represented in each
region's model system - freeways and arterials plus some major collectors). Price levels were set to
reduce congestion to meet specific levels of service; we investigated a range of level-of-service
targets and eventually chose LOS D/E for use in all four metropolitan areas.2 Our analyses
allowed prices to vary by corridor, determined peak definitions by the extent of congestion in each
The first US congestion pricing project opened in December 1995 on State Route 91 in Orange County
California; the San Francisco-Oakland Bay Bridge is currently being studied as a second possible application!
Because of the special characteristics of these two applications, spillover to arterials. is not likely to be a
major issue. SR 91 pricing will apply only to the new lanes added in each direction, with the original lanes
left unpriced; in the Bay Bridge case there are .essentially no realistic alternative highway routes. The
extension of pricing to other facilities such as I-1Q in the Los Angeles area or 1-80 in the Bay Area would,
however, have to confront the possibility of spillovers to parallel routes.
The choice of LOS D/E was based on analyses of benefit measures from the STEP model" which indicated
that stable, near-capacity flows (about 10 percent below actual capacities) were'the most economically
efficient traffic regime. Specifically, we used delay reduction per marginal unit of price as the measure of
benefit.
U. S. Environmental Protection Agency. ,."~ B~lT
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
" '•• ' '•.:'.' I
corridor, and permitted different prices to be charged in each corridor for each hour of the AM and
PM peak periods, but we stopped short of dynamic pricing. Rather, we assumed that travelers
would face a fairly simple schedule of prices by time of day, readily, comprehensible to travelers
and influencing their travel behavior and location choices.^ -
It is important to note that under this pricing approach, users of the facilities in greatest demand
still would perceive traffic as heavy and somewhat constrained, with speeds below posted limits (at
least for the cases considered here, higher speeds would not be as efficient from an economic point
of view). Note also that we assume that prices would be maintained in constant dollars, meaning
that from time, to time price adjustments might be necessary.
•": ' ," ' • " ; '.'•''• ' , "|l'"': •'-•,. I ' ~ ' ' ' .
The STEP analyses were carried out by focusing on highway performance at the corridor level, as
follows. In the STEP calibration phase each of the metropolitan areas was divided into major
corridors based on topography and highway function. Each district-to-district trip interchange was
assigned to a corridor, and approximate volume-delay relationships (i.e., expressing travel time per
mile as a function of volume and capacity) were developed for the corridors.^ This was carried out
for both the AM and the PM peak in each region.
3 We assumed congestion prices would be in effect on non-holiday weekdays only - 250 days a year.
4 The shape of the volume/delay curve is a critical determinant of the outcome of the analysis, because it
indicates how much traffic would have to be removed from the peak in order to achieve a given LOS. To
represent volume/delay relationships for corridor-level changes, STEP uses an equation roughly analogous to
the Bureau of Public Roads (BPR) equation relating link-level volumes and travel times. The STEP equation
was initially developed in a study for the California Energy Commission and later re-estimated in studies for
the Metropolitan Transportation Commission, the Southern California Association of Governments, and the
Puget Sound Council of Governments (Seattle region). The equation expresses the relationship between the
ratio of average peak to average off-peak travel times in each "corridor" - basically a trip exchange - and the
aggregate capacity serving that corridor. Separate estimations were done using data from the detailed highway
networks of the three regions; because the coefficients of all three models were nearly identical, a single
equation was implemented in STEP. The specific functional form is t/t0 = 1 + (V/C)2.
' ' * • . ' ' !
This corridor function, derived from regional network models, shows travel time climbing rather gradually as
congestion builds. We know from highway operations research that the buildup of congestion for specific
facilities is more abrupt and steeper in the region of capacity flows than this equation indicates. However,
because the corridor function represents an aggregation of facilities of different types, it reflects the "family"
of volume-delay relationships for the freeways, arterials, and major collectors embedded in the network
models and producing their travel time estimates.
The steepness of the buildup of congestion is important in determining what the congestion price would have
to be. If the slope is steeper than our equation indicates, as it would be in a corridor with a single facility,
congestion prices could be lower for a given level-of-service improvement than we report here. This is
because a steeper slope implies that fewer vehicles would have to be priced off each corridor's facilities to
achieve a given LOS. We tested a number of functional forms in STEP, and the different forms did indeed
produce some variation in optimal prices. For example, letting the slope parameter rise to 4, the value used in
the standard Bureau of Public Roads (BPR) equation, would lower the "optimal" congestion price by about 40
percent (regional average). Since the BPR curve is for a single freeway facility, it is much steeper than any
corridor curve could be (unless the corridor consisted of a single freeway). Therefore the BPR value should
be viewed as an outer limit.
B-12 U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
The level of service target was defined in terms of the volume delay function. For the generic
functional form used in this version of STEP, level-of-service D/E corresponds to a travel time that
is about 85 percent longer than the time under free-flow conditions. In other words the target
level-of-service was represented by a l.SS.ratio of peak to uncongested travel time in a corridor.
In the Los Angeles region; about 300 aggregate " corridors" were defined in this manner, and about
220 of them - 73 percent - were sufficiently congested in the AM peak to justify congestion pricing
For the San Francisco Bay Area 150 corridors were defined, with 90 (60 percent) meeting the
cntena for pricing in the AM peak. San Diego and Sacramento were both considerably less
congested; only 15 percent of the 80 corridors analyzed in San Diego and 8 percent of the corridors
analyzed in Sacramento were candidates for pricing.
To estimate the price needed to achieve the target level of service, STEP was applied to each
sample of households and the average price per mile was adjusted on a corridor-by-corridor basis
until all corridors were at or below the 1.85 peak/off-peak travel time ratio, and no corridor had a
higher congestion price than necessary. This took approximately five iterations (model runs) for
each region and each analysis year. .
- For each region, a specific congestion price was estimated for each corridor and time period 5 For
1991 conditions, the congestion prices would vary from zero (for the uncongested exchanges) to as
much as $1.00 per mile for a very few corridors, such as the 1-80 corridor and the Bay Bridge
corridor in the San Francisco Bay Area and the 1-405 and HO corridors in the West Los Angeles -
Santa Monica area. In San Diego, the highest corridor level prices would reach about 40 cents per
mile, whereas in Sacramento the highest corridor prices would be about 20 cents per mile By the
year 2010, congestion is expected to worsen considerably in all four regions; many more corridors
would be candidates for pricing, and prices would have to be higher to maintain the LOS D/E
target.0
In the four case study regions, PM peak conditions are less sharply congested but last longer than AM
peak conditions. Hence evening congestion prices, at least initially, could be somewhat lower but would be in
effect for a somewhat longer period of time than those in the morning peak. However, congestion pricing
would flatten and spread out the AM peak somewhat, diminishing AM-PM differences in prices and hours of
application. • ,
One might ask whether the prices arrived at in this manner are the optimal prices. The issue is not simple
to resolve; in the first place it is well understood that user-optimal may not be identical to system-optimal
(Wardrop,1952). User optimality is examined here, although we note in passing that pricing also could be
used to achieve system rather than user optimality. The analysis of user-optimal prices is particularly
complex, because travelers can respond to pricing, in a.number of waysr shifting,trips among corridors and
altering their frequency and times of travel. It is necessary to account for the possibility that travelers could
switch to another route, travel at a different time of day, change modes, choose different destinations for some
trips, increase or reduce the number of trips made, move to a different residence, or change their place of
work. STEP, accounts for these phenomena, but because STEP 'is a hybrid mix of non-linear demand
functions of various types, it is not possible to mathematically prove the existence of a unique set of
congestion .prices for a given level-of-service criterion. Simulation offers an alternative approach for
assessing whether model results represent a stable and unique equilibrium, and we used it to investigate the -
optimality of our corridor prices. We applied a number of procedures designed to determine whether STEP
would produce different sets of "optimal" congestion prices. These included adopting different search
algorithms in the program code, and starting the searches from different initial corridor prices. All search
strategies that produced stable outcomes were in agreement with the initial "optimal" prices, which lends
some support to the notion of a unique equilibrium. " '
U. S. Environmental Protection Agency '~~~ ~' '. ' '. g ,,
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Estimated reductions in travel time, VMT, trips, emissions, and fuel use resulting from the
calculated congestion prices, as well as estimates of the total revenues generated, were calculated
by summing up the analysis results for each corridor. To simplify the presentation of price levels
and provide an indicator of overall price impact, a corridor-weighted average price per mile is
shown in the tables, and can be thought of as the average price peak period drivers would face
overall. It is not necessarily the price any individual traveler would experience. For example, the
price necessary to obtain LOS D/E on the San Francisco Bay Bridge in 1991 would have been
about $6, or 75 cents a mile for that corridor, in contrast to the average Bay Area AM peak price of
about 9 cents a mile J
Corridor-level results are useful for preliminary planning purposes, but for implementation
planning it is important to translate the results into specific facility charges. Within the resources
of this study, we were not able to test congestion pricing in a full network context for each of the
four case study areas. Instead, we ran STEP for the four areas, then selected two corridors for more
detailed analysis: 1-80 from the Carquinez Bridge to the Oakland-San Francisco Bay Bridge, and I-
10 from Santa Monica to Downtown Los Angeles - two of the most congested locations of all those
we studied. We ran regional network models for the Bay Area and Los Angeles to see how prices
would need to vary among facilities in the selected corridors, given the corridor prices and demand
levels produced by STEP. The Tranplan network analysis program was used, with an equilibrium
traffic assignment for the AM peak hour and price incorporated into the route choice criterion.^
'"' i „ '
Tranplan corridor analyses produce results comparable to STEP if the per-mile price is applied
equally across all facilities in the corridor. With the same price per mile on all alternate routes, the
main effect will be a reduction of overall corridor demand rather than a rearrangement of traffic
among corridor facilities (absent differential prices, traffic in a congested corridor will distribute
itself such that all routes will have about the same travel times). However, Tranplan analysis made
it possible to test link-by-link pricing to more precisely target bottlenecks in the system. We went
through five iterations in which we manually adjusted link prices in the two test corridors, each
time increasing the price on links that remained congested and decreasing the price on links with
better than D/E level-of-service and then running Tranplan to evaluate link-level impacts. Overall
corridor delay reduction tended to improve from iteration to iteration, while overall price levies
tended to fall. After the .five iterations, we judged that the effectiveness of congestion pricing, in
terms of reduced delay per dollar, might be 10-15 percent higher in these corridors than the
approximate results of the STEP analyses would suggest. This should be considered when
reviewing the average prices and/or time savings presented in the tables.
What if prices varied by location, but were set at modest prices initially and were increased only
gradually to the levels necessary to avoid stop-and-go driving? This approach would give people a
chance to adjust their travel and location behavior under prices that accurately signal the ultimate
7 The Bay Bridge congestion pricing studies underway at the time of this writing are discussing
considerably smaller prices, e.g., a $3.00 peak period toll. A $3.00 toll in 1996-1997 dollars would be the
equivalent of a $2.50 toll in 1991 dollars. Such a price increase would be sufficient to cut the queue at the toll
plaza by about a third, but would not achieve LOS D/E.
8 It was possible to use the network provided by the Southern California Association of Governments for
this part of the analysis, but for the Bay Area a new Tranplan network was created as part of the study (MTC
Uses UTPS networks and the study team did not'have access to this software).
B-14 U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
, ' . . . ' to Reduce Transportation Emissions
spatial distribution of impacts. Dynamic models would be necessary to fully explore the changes
that such a pricing approach would produce over time; STEP does not currently include such
dynamic models. However, STEP is able to evaluate lower-than-" optimal" congestion prices as
would occur in a pricing phase-in (and perhaps in many cases where prices are set on political as
well as technical grounds). We tested the impacts of lower prices by taking the final corridor
congestion prices for the Bay Area and Los Angeles and applying them in 10 percent increments
(i.e., prices at 10 percent of optimal corridor prices, 20 percent of optimal prices, etc.). The STEP
results, indicate that the shape of the aggregate demand curve is moderately convex, with slightly
decreasing effects for.each price increment. For each of the two case analyses, the first price
increment of 10 percent produced almost twice the 'impact of the final increment of 10 percent.
This suggests that implementing a constrained price can still be reasonably effective.
i i
The STEP analyses are for scenarios in which pricing is used to manage congestion wherever it
occurs on the network of highways and arterials; how congestion pricing would wprk if
implemented on a few facilities is a different question. Even if the ultimate objective is system-
wide implementation, it is likely that-initial applications would be "spot pricing" - pricing applied
to just a few facilities or corridors. As we discussed earlier, however, closely-parallel routes could
receive significant amounts of diverted traffic if a single congested facility is priced; such traffic
diversion could lead to significant congestion on the parallel routes; and opposition from affected
jurisdictions might well be enough to halt implementation, unless the parallel routes can be priced
as well. - .
Even where diversion to parallel routes is infeasible for most travelers, as is the case for the San
Francisco-Oakland Bay Bridge, or where each facility in a corridor can be differentially priced, as
our analyses x>f the 1-80, 1-405, and I-10 corridors considered, a number of concerns about "spot
pricing" remain. For example, our analyses indicate that implementation at a single .highly-
congested location or .in a single corridor will alter regional patterns of trip distribution, residential
location, and workplace location, with specific effects varying with household income level. The
result of spot pricing could lead to a distortion of the spatial structure of the region, because the
spot pricing leads to exaggerated locational impacts. Thus single facility pricing may produce a
misleading view of the eventual areawide effects of congestion pricing.
B.3.3 Employee Parking Charges
In most metropolitan areas, parking is commonly provided to its users free of charge, although
providing such parking can be quite expensive and presumably is recouped in other ways (e.g.,
through the prices charged for goods and services, for private parking, or through public tax
subsidies, for public parking). Charging for parking, whether done through private initiative or in
response to government incentives or mandates, would make the costs of parking more apparent to
travelers and would likely reduce auto use somewhat. ,
Parking could be priced for all users, and sometimes is (at many commercial garages, e.g., or by
. local governments who install on-street meters). However, proposals for the implementation of
parking pricing often focus on daytime employee parking, since the associated employee travel
typically occurs during the costly peak periods. If employees had to pay for parking, it is reasoned,
they would be more likely to use alternative commute modes such as transit, carpooling, or '
walking. In the case studies we present here we analyze only employee parking charges.
U., S. Environmental Protection Agency •~~ ^"""~~~ ' - g ,-
-------
Teffinical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
In comparison with congestion pricing, parking pricing is a relatively simple measure to analyze
using STEP. The average zonal parking price (daily, for work trips, and hourly, for non-work trips)
is a variable in each of the STEP mode choice models, and zone-lever parking price data are
available for each of the four metropolitan areas studied here. Thus, any parking scenario that can
be expressed as a change in an average zonal price can be analyzed using STEP.
1111 'i
Proposed parking price changes do not always target the average zonal parking price, however.
Consider a city in which a substantial amount (varying by zone) of the all-day parking is provided
by a private operator, who charges a daily fee for use. The operator, perhaps given an incentive by
local or state tax policy, decides to raise the fee by $1.00 per day. To analyze the impact of this
increase, it is necessary to have an estimate of the percent of all-day parking in each zone that is
provided by this operator and hence will be affected by the increase. A number of cities maintain a
parking inventory which could provide this information, although many other cities would have to
conduct a special survey to produce this estimate.
Other parking pricing proposals can be far more complicated to analyze. Consider a $3.00/day
parking surcharge which applies only at employment sites with 100 or more employees. In order to
translate this surcharge into zonal average price estimates, we would need information about the
fraction of workers in each zone who work at sites with 100 or more employees. We would need to
account for the possibility that some of those employees do not provide any parking now, in order
to figure out what share of each zone's employees would be subject, to the fee. The possibility that
some employees could avoid a fee at their workplace by parking elsewhere should already be
reflected in the calculation pf zonal average parking cost, but we also must consider the possibility
that employers will simply pay the fee themselves rather than passing it on to the employee, again
reducing the number of affected workers (note that certain implementation strategies, such as
treating parking as a taxable benefit or requiring the surcharge to be collected from the employee as
a payroll deduction, would reduce the likelihood and the impact of this latter concern). Very few
cities have an employer and parking data base organized to support such an analysis, and we have
found none that has information on likely alternative parking sites or on employer responses to
such policies. Hence, calculating the actual increase in zonal average parking charges that our
surcharge would produce could require either a great deal of data "collection. Nevertheless, for
preliminary planning purposes it usually will suffice to make some simple assumptions in
developing the data inputs or in interpreting the results. For example, we could analyze the parking
surcharge as if it applied to all employees and then factor the results downward to account for its
more,restricted reach: if regional employment data indicate that only 40 percent of the region's jobs
are provided by employers with 100 or more workers, then our impact estimates should be reduced
by about 60 percent.
For our four case studies, we utilized parking cost data files developed by the regional
transportation agencies. These files present only the estimated average employee parking price
(nominal price) by zone. Given the data we had available, we chose to model two general policy
options: a flat daily charge on all employees who drive alone and do not currently pay for parking,
as well as a daily surcharge on all employee parking, paid or not. The first option could be thought
of as a rough approximation of what prices might be like if free parking were no longer provided to
employees; or it might be thought of as the result of a policy that imposes an impact fee or tax on
free employee parking but waives the fee on parking that is already priced at or above some
threshold level. The second option would be a flat impact fee (or tax, depending on how it is
structured and applied).
B-16 . U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
r • i '
Using STEP, a range of daily employee parking charges from $1.00 to $10.00 was examined for
each of the four metropolitan areas. To model the minimum price threshold option, drive-alone
parking fees for ail workers in each sample were set to the specified minimum or to current levels,
whichever, was higher - fees in zones where existing zonal average parking fees exceeded the
threshold charge were held constant. The second option we evaluated, a flat fee or surcharge on all"
employee parking, was even easier to represent than the minimum price option; the fee was simply
added to the employee parking price in effect in each zone. In both analyses, we assumed that the
employees would personally pay the parking charges (hence we treated the charges as out-of-
pocket expenses). We also assumed that carpool and vanpools would be permitted to park for free
at their destinations, and that no charges would b,e imposed for park-and-ride parking. These latter
assumptions are generally consistent with the current treatment of HQVs. and park-and-ride in the
, four case study regions. '
STEP accounts for the full set of travel effects we would expect parking pricing to have, including
impacts on highway performance, but to verify that STEP'S simplified level-of-service functions
provide an adequate representation of the latter, the peak period trip tables from STEP were
assigned using Tranplan to the relevant networks for Los Angeles and the Bay Area, and the
resulting travel times' were cycled back through the STEP model. No significant changes from
STEP aggregate performance measures were identified.
Results for $1.00 and $3.00 parking price increases are reported here. Given the ubiquitousness of
free parking in each of the.four regions, the differences between the two policy options were
minimal: the estimated impacts of the parking fees varied by 10 percent or less (i.e., a reduction of
1 percent in VMT for the minimum price option, a 1.1 percent VMT reduction for the surcharge).
Our assumptions that prices would apply to all drive-alone vehicles? and that HOV parking would
be exempt from charges maximize the impact of-the .employee parking fees. In actual
implementation, a number of factors could reduce these impacts. For example, as our earlier
discussion pointed out, exemptions of certain employers would reduce the number of employees in
each zone who actually would pay a parking fee, with the impact varying widely among zones.
In addition, in situations where parking is differentially available to or subsidized for different
income or occupation groups, the impacts of price changes may vary from those we have shown.
Our results assume that a parking fee would be paid by all who drive alone. But under some
conditions the fee might actually be absorbed by the employer; for example, some blue collar
workers have negotiated for free parking as part of their labor agreements, and a parking surcharge
would have to be paid for by the employer or compensated through offsetting salary increases. In
cases such as these, the fee on parking could vary systematically with, income group, and hence be
disproportionate to the number of workers affected.
Finally, the impact of free parking for high-occupancy vehicles deserves special attention. Free
HOV parking is a common measure in our case study regions and might well be permitted under a
policy to charge for parking; but it is not a necessary feature of the analysis. If the parking fees
apply equally to HOVs, HOV users still experience an advantage over solo drivers because they
can split the cost among all passengers,'but the price differential between drive-alone and HOV
To calculate impacts on an annual basis, we assumed employee parking charges would apply 250 days a
year. . .
U. S. Environmental Protection Agency B 17
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
decreases - by about 40 percent on average. Based on STEP runs for all four metropolitan regions,
this diminished advantage would cut the impact of the parking fee by about 15 percent, because
fewer current drivers would switch to HOV and some of those who currently are HOV users would
decide to drive to work.
• ' •• • ' . ' • • • • " i .
• • i
B.3.4 Fuel Tax Increases
•.
A fuel tax increase would be a direct approach for reducing fuel consumption and also for reducing
greenhouse gas emissions (because CC>2 emissions are proportional to fuel consumed). Its effects
on other emissions and travel are muted, though still significant, because auto purchase decisions
and usage patterns can lead to a more efficient vehicle fleet and reduced per-mile operating costs.
The fuel tax increases analyzed here are expressed as straightforward additions to the at-the-pump
price of gasoline and diesel fuel. For our base case, vehicle fleet fuel economy is about 22 miles
per gallon (.0364 gallons per mile). Base-case fuel cost is about $ 1.20 per gallon, or 5.45 cents per
mile at average fuel economy. With no increase in fleet fuel economy, a 50 cent per gallon fuel tax
increase would add about 2.3 cents and a $2.00 per gallon tax (or other form of price increase)
would add about 9.1 cents to the average per-mile cost of driving. However, empirical evidence
and common sense suggest that the in-use vehicle fleet would become more efficient under a
significant fuel price increase. In the many households with more than one car, household members
could quickly arrange to make more use of their fuel-efficient vehicles and less use of their "gas
guzzlers," cutting fuel consumption considerably. Over time, both single-vehicle households and
multi-vehicle households could be expected to increase vehicle fuel efficiency as they replace some
vehicles and retire others.
How fast and to what degree such vehicle substitutions, replacements, and retirements might occur
in response to fuel price increases has been a matter of considerable dispute. The issue is important
to our analysis because it could significantly affect the impact of a fuel tax. Travel and location
choices are undoubtedly affected by the costs of vehicle ownership and operation, i.e., by both the
number of vehicles a household chooses to own and the type and age of its vehicle(s). Faced with
higher fuel costs, a household which for whatever reason does not reduce its per-mile fuel
consumption (by changing its vehicle holdings or changing which vehicles it uses most) will have
to devote more of its income to fuel purchases, or take steps to reduce its vehicular travel (or some
combination of the two). If on the other hand the household finds it possible to reduce the price
effect through vehicle substitution and replacement, fuel efficiency improvements will have a
smaller effect on travel.10
STEP includes a model of the number of vehicles a household chooses to own, so we were able to
capture the effects of fuel price increases on auto ownership in our analyses. However, STEP
. currently does not address the type or age of the vehicles owned, information which is needed to
estimate the cost per mile under different fuel price scenarios. We did not have direct access to a
model of household vehicle purchase decisions for this study, so to account for the broader range of
10 A household's ability to change vehicle holdings is related to its current and expected income, its current
vehicle holdings, ownership and operating costs of the alternatives, etc. The household's willingness to
change its vehicle holdings depends on many additional factors, such as vehicle seating capacity, comfort,
handling, and safety; fuel economy, an element of operating cost, is but one influence.
B-18 U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
impacts, we turned to outside sources for evidence on the elasticity of fleet fuel economy with
respect to fuel price. -
The literature from the U.S. and abroad suggests that fleet fuel economy (miles per gallon) is quite
sensitive to the price of fuel. Pickrell's recent research (Pickrell, 1993) and his syntheses for the
Presidential Commission on Greenhouse Gas Reduction (a group known popularly as "Car
Talk")(Pickrell, 1995) examine the impact of fuel prices and report findings from a wide range of
reputable U.S. and international studies in advanced economies. He cites numerous estimates of
long-run average elasticity of fleet fuel economy with respect to fuel price in the .5 - .6 range, with
estimates as low as .2 to .3 and some higher than 1.0. An elasticity of 0.5 means that a 25 percent
increase in real fuel price (e.g:, from $1.20 to $1.50) would increase, long run average fleet fuel
economy from 22 miles per gallon (mpg) to almost 25 mpg; ,a 167 percent increase in real fuel
price (e.g., from $1.20 to $3, .20) would increase long run average fleet fuel economy from 22 mpg
to about 40 mpg (82 percent). A 40 mpg fleet average sounds high for U.S. conditions, but it
cannot be dismissed outof-hand, especially for a longer-term scenario (2010 or later) and/or one in
which the price,increase was implemented nationwide or in a majority of urban states (so that
manufacturers would have sufficient time and incentive to offer more fuel-efficient vehicles).
Substantial fuel economy improvements could, in fact, be obtained through shifts in consumer
choices among the vehicles currently available for purchase: for example, by purchasing the four
cylinder rather than the six cylinder version of a midsize sedan, a consumer could obtain a 10-15
percent improvement in mpg. This percent increase in fuel economy is about what a 25-50 cents
per gallon price increase would require, at a .5 elasticity. However, for large fuel price increases,
an elasticity of .5 would imply that at least some consumers also would have, to change the type of
vehicles they own and,use, i.e., greater numbers would have to purchase and use highly efficient
vehicles and restrain their purchase and use of the least efficient ones. Currently over a dozen
vehicles are sold in the U.S. which obtain over 40 mpg, so this seems technically, feasible, and may
become more so if gradual improvements in technical efficiency, averaging perhaps 1-2 percent a
year, are forthcoming over the next decade or so, as many analysts expect (Pickrell, 1995).
Whether buying habits in fact would change in the necessary fashion could be debated.
For further evidence of how fuel prices might affect fleet composition and use, we turned to models
of the vehicle fleet. Since our case study regions were all in California, we were particularly
interested in an analysis tool known as the Personal Vehicle Model (PVM), which the California
Energy Commission has used to estimate the composition of the state's vehicle fleet by size and
age, as a function of the price of fuel and other factors.11 We asked the CEC to provide some
indication of the PVM elasticity of fuel economy with respect to fuel price, as evidence for
California fleet conditions. A run of the PVM made for this study by the CEC in January 1995
indicated that a $2.00 fuel surcharge would lead to a 2 mpg increase in fuel consumption (from 22
to 24 mpg), for an average elasticity of .05.
The PVM-estimated elasticity is much lower than the elasticities reported by Pickrell. A partial
reason for the difference is that most national and international long-term elasticity estimates allow
" The PVM was developed more than a decade ago, and at the time of our study the CEC was engaged in
a multi-million dollar project to replace it with an updated package based'on new data and state-of-the-art
modeling concepts. Hence we chose to treat the PVM as one source of evidence rather than to rely solely on
it. ' . . • ., . . . • •
U. S. Environmental Protection Agency ' • B-19
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
•, i '
for changes in the products manufacturers offer in response to fuel price increases. In contrast, the
PVM analysis assumed that the price increase would only apply in California, and that
manufacturers would not increase the fuel economies of the cars they offer in response to a change
in only one state, even a state as large as California. The PVM analysis does allow consumers to
purchase more efficient vehicles from those otherwise available. It does not. consider increased
relative use of the more fuel efficient vehicles within each'household's existing vehicle holdings.
We discussed the fuel economy - fuel price elasticity issue with a number of researchers and
ultimately settled on testing a range of assumptions about the fleet response to fuel price, expressed
in terms of the elasticity of fuel economy (miles per gallon) with respect to price. Results for three
elasticity levels are reported here: 0.5, 0.16, and 0.05. The researchers we contacted felt (and we
agreed) that the .05 PVM elasticity should be used as a lower boundary, and that a 0.5 elasticity,
i.e., the lower end of the .5-.6 estimates from the national studies, was a reasonable upper boundary
for a California-only policy. 12
1 " " i ' i I, ' ' , • ' • 'I i '
The fuel economy elasticities can be used to compute average mpg and out-of-pocket vehicle
operating costs per mile resulting from a fuel price increase. For example, consider a two dollar
per gallon increase, i.e., a fuel price of $3.20 per gallon. In comparison to the current $1.20 per
gallon, for which average out-of-pocket expenditure is about 5.5 cents per mile, the estimated mpg
and cents-per-mile costs would be:
Elasticity MPG Cents per Mile
0.00 22 14.6
0.05 24 13.3
0.16 -28 11.4
0.50 40 8.0
:t • . , •• , ' '.'•'•''': ' ••: , '• "', . . • • • •
It is clear from this table why fleet response to fuel price is such an important issue. At a .05 fuel
economy elasticity, the average fuel cost per mile increases by more than 140 percent; this would
result in large reductions in travel. By comparison, at an elasticity of .5, the average fuel cost per
mile increases by about 45 percent. In the first case, trip and VMT reductions account for most of
the drop in fuel use, while in the second case, improved fleet fuel economy accounts for most of the
drop in fuel use. Since both the incidence and the economic, implications of the fuel price increase
differ markedly between these two cases, forming a more precise understanding of fleet fuel
economy sensitivity to fuel price is of some importance.
12 A California-only gas tax increase seems more plausible for small to moderate tax increases (25 cents or
less) than for higher ones, especially those of a dollar or more. Of course, it is not necessary to assume that a
fuel tax or other, fuel price increase would be implemented in California only: the analyses could equally well
represent the impacts of scenarios involving federal fuel tax increases or state tax increases implemented in
many states. Also, for the analyses presented here, at-the-pump price increases implemented by sellers would
have the same effects as a fuel tax increase. A California-only interpretation of our analyses does not
necessarily require new, highly efficient vehicles to be produced for the state market (though it might make
California an attractive test bed for such vehicles, including ones currently sold overseas but not now
marketed in the U.S.). It does however presume that, of the vehicles produced for the U.S. market,
manufacturers wpuld sell a higher share of the most efficient vehicles in California. Also, the used car market
would be affected; demand for low mpg cars would decline in the state, and such cars would likely be retired
earlier or perhaps shipped to other states or countries for sale there.
B-20 , U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Using our three elasticities, we studied a range of fuel price, increases'from $0.10 to $3 00 in 10
cent increments. The results.for the $2,00 fuel price increase under different elasticity assumptions
are presented here, along .with some results for a $0.50 price increase. Results for these two price
levels are sufficient to support generalization about price effects over the full range. 13
It is worth noting that for some policy objectives, the fuel price (fuel tax) might be adjusted
penodically to maintain the per-mile cost, i.e., to reduce the impact of improved fuel economy
Such tax adjustments would make sense in terms of paying for road maintenance, since
maintenance costs do not decline proportional to fuel use. Similarly, if pay-at-the-pump insurance
policies were implemented, it would be necessary for the component of the fuel "tax" designated
for insurance to be de-coupled from fleet efficiency. If for either reason: the fuel.tax were adjusted
to compensate for revenue losses due to fleet efficiency improvements, its effects oh VMT, trip
rates, delay, and emissions would be greater than we have estimated here. Essentially such
adjustments would make the fuel tax very much like the VMT fee discussed below.
B.3.5 VMT Fees
A fee oh vehicle-miles of travel (VMT) would directly charge users for the amount of vehicular
travel consumed. A VMT fee therefore could be used to reduce VMT-related impacts.14 Such a
fee also would be a better targeted road user payment mechanism than the fuel taxes we now use
because drivers could not reduce their exposure to the fee by purchasing more fuel efficient
vehicles.1-5 , ,
Currently, the easiest way of collecting a VMT fee would be through a charge determined at the
time of vehicle registration or vehicle inspection, based on owner-reported or inspector-recorded
odometer readings. However, if one goal of a VMT fee is to reduce vehicular travel and its
negative externalities, the fee should be linked as closely as possible to day-to-day use of the
vehicle. Collecting the VMT fee as part of an annual payment for vehicle registration would
probably be less effective in reducing VMT than more frequent charges: an annual fee is remote
from individual drivers' thinking about their day-to-day driving behavior, and may be less effective
in influencing it. Also, drivers would "discount" annual payments compared to more frequent
levies.
We calculated:impacts on the basis of 250 times the average weekday rate plus 1.15 weekend-and holiday
days at 95 percent of the weekday rate.
"-* '- . ' •
VMT is roughly related to congestion, though a VMT fee would have a bigger effect on non-work travel
than on work trips, which make up the majority of VMT during the congested peak periods VMT is also
roughly related to fuel use and to hydrocarbon, NOx, and carbon monoxide emissions. In contrast, PM
emissions from on-road transportation are closely related to VMT.
15
Used as a road user payment mechanism, the VMT fee would have to be adjusted periodically or indexed
to reflect costs of road construction, operations, and maintenance, or if such road costs increase the fee's
percent cost coverage would decline. Nevertheless, costs to each user would remain proportional to use Per-
gallon fuel taxes also suffer from declining cost coverage unless adjusted or indexed, but are far less directly
related to use of the roads because of divergent vehicle fuel efficiencies. . •
U. S. Environmental Protection Agency • _«
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
There is no reason, of course, that a VMT fee tied to registration or I/M programs would have to be
paid annually. One can imagine a variety of alternative arrangements, including ones in which the
registration or I/M fee itself is paid in (monthly or quarterly .installments. One approach might
mimic the billing method used by public utilities, in which monthly or quarterly bills are based on
estimated usage, and a periodic reading (or report) is used to calculate the additional increment due
or credit earned. 16
, • . • ...-.,. | •
Recent technological developments offer other ways to frequently measure and collect a VMT fee.
It is currently feasible to put in place a VMT monitoring system using automatic vehicle
identification (AVI) technologies and covering all major facilities including freeways and major
arterials. Systems such as these are currently being deployed on tollways in many, parts of the U.S.
as well as abroad, and offer timely and accurate fee collection. In one design motorists purchase
debit cards which are displayed on their vehicles; fees are deducted from the cards electronically as
the vehicles pass AVI readers. In .another design the readers record each passing vehicle's
identification code and transmit the data to a computerized system which accumulates the charges
and periodically bills the vehicle owner.
An alternative concept currently in prototype stage would base the VMT fee on an at-the-pump
reading of an electronic odometer or a special VMT-accumulating "smart card;" the
corresponding fee would be calculated electronically and could be collected as part of the payment
for fuel, or perhaps recorded and billed separately. In one approach, scanner or microwave
technologies would automatically read the odometer or another on-board electronic device
designed to monitor VMT. In another approach, the motorist would insert the vehicle's " smart
card" into a special reader, following a sequence of actions much like those used with the
automatic credit card debiting devices now present in many fuel pumps.
The availability of approaches, high tech or low, for collecting a VMT fee at or close to the time of
road use is important, because such immediate and visible prices are likely to be treated by
travelers essentially as out-of-pocket costs similar to current fuel costs. Here we treat the VMT fee
as a pure increase in the per-mile cost of driving, with no possibility of avoidance and no
" discounting" by drivers for delayed payment. In essence, the fee defined in this way would be the
equivalent of a fuel tax increase that is indexed to vehicle fleet efficiency.
VMT fees ranging from 1 to 10 cents per mile were analyzed for each metropolitan area (at the
base case fleet fuel economy, this is equivalent to fuel price increases ranging from $0.22 to $2.20
per gallon). Results for the 2 cents per mile fee are reported here. 1? In keeping with the
methodology described earlier, all elements of the STEP model were employed, from residential
location through supply response. For Los Angeles and.the Bay Area, we further checked the
results by assigning STEP-based peak trip patterns to the highway networks. No differences were
found that would significantly alter the findings from STEP.
16 Income and payroll tax collection methods are another possible model: frequent payments are made based
on estimated amounts due and reconciliation of the amounts due is done via an annual report, subject to audit.
* ' ' ..''*' '' \
17 We calculated impacts on the basis of 250 times the average weekday rate plus 115 weekend and holiday
days at 95 percent of the weekday rate.
B-22 U- &• Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Note that because the results were produced at a regional level!, they are for within-region VMT
only. They do not include VMT generated outside each region being analyzed. A VMT fee
designed for revenue generation might, of course, be implemented on a statewide basis and could
be analyzed in that fashion.
A regional VMT fee based on-AVI monitoring of road use would be simple enough to implement.
A regional fee based on odometer readings, on the other hand, would charge the motorist for inter-
regional, interstate, and international travel (Mexico, Canada) unless some mechanism for
excluding such travel were devised. One can easily imagine ways to credit motorists for interstate
and international travel; for example, motorists who want a credit for out-of-state travel could have
their odometers read at stations along major entry and exit routes to the state, or a procedure might
be established allowing a tax credit for documented out-of-state travel, much like the, one now used
for fuel tax credits for exempt off-road vehicle use. It would be much more difficult to devise a
low-tech way to credit within state inter-regional travel without creating a major paperwork burden
for all involved. Since Caltrans periodically does statewide travel surveys which include both
within-region and inter-regional travel, one approach might be to use the survey data to create a
system of adjustments for each region to account for the average out-of-region component of VMT,
perhaps by vehicle age.
If the VMT fee were collected infrequently, e.gl, once.a year based on an odometer reading or
report, its impacts might be somewhat less than we estimate here due to discounting of future
lump-sum payments in comparison to equivalent "out-of-pocket" payments. Hence the results
reported here should be viewed as the high end of likely effectiveness.
B.3.6 Emissions Fees
Emissions fees represent a means of reducing tailpipe emissions that could give .the consumer
somewhat more flexibility than the current system of mandated performance backed by vehicle
inspection and maintenance. The basic concept is that the total pool of annual vehicular emissions
in a region would be assigned a cost (presumably pollutant-by-pollutant), and each vehicle would
be charged a fee set to reflect its contribution to the total emissions burden. Levying such a fee on
vehicular emissions arguably would be the most direct way to instill a sense of personal
responsibility for mobile source air pollution.
While the concept may be simple to state, emissions-based vehicle fees are the most difficult of the
pricing policies to define and analyze. Reasons for this are:
• the literature offers widely varying perspectives on the social costs of air pollution,
so an agreement on a monetary basis for the emissions fee is not easy to reach;
• estimates of cumulative emissions from individual vehicles are imprecise and are
likely to remain so unless and until vehicles are equipped with accurate, tamper-
proof on-board emissions monitoring devices; "
• because knowledge, about how consumers would trade off emissions fees, repair
costs, insurance, and other auto-related expenditures is not wejl developed, the
change in fleet composition resulting.from a targeted emissions fee is difficult to
estimate. .
U, S. Environmental Protection Agency B-23
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
We carried out analyses of two prototypical emissions fee strategies, each using a different type of
information about emissions, following the same line of argument as for the VMT fee, we assume
that emissions fees can be collected on a "pay as you go" basis, so that they are perceived by
drivers as an out-of-pocket expense. This could be done with a technologically advanced system
such as an on-board monitor, read and billed, e.g., at the time of fuel purchase; or by combining
some other method of fee calculation with a monthly billing system. If the emissions fee is
determined as part of vehicle registration or inspection/maintenance and is billed annually or
biennially as part of those programs, the fee may well have less influence on day-to-day travel
faehayior than we show (on the other hand, a large, infrequent fee might have a big influence on
vehicle ownership levels, vehicle age and type, and vehicle maintenance).
,',,',, j
Al] non-arbitrary emissions fee concepts rely on some assumption about the social costs of air
pollution. Accordingly, we searched through the literature for evidence that would support a
specific emissions fee in each region, and sought the advice of experts in university research groups
and air pollution control agencies. We found that the costs of air pollution had not been researched
consistently for all the case study regions, and that the sources that do exist show a wide disparity
in their damage estimates. Credible cost estimates for mobile source pollutants range from about
.25 cents per vehicle mile to about 8 cents per vehicle mile (using regional damage estimates,
reduced by the portion of emissions not attributable to mobile sources, divided by annual regional
VMT). The range reflects differences in the severity of the pollution problems of the various
regions and in the types of damage considered, as well as disagreements over specific costs in a
given region (controversy is especially acute concerning the interpretation of epidemiological
studies).
Lacking more specific estimates of the social costs of emissions in each of the California regions,
we chose to set our emissions fee to average one cent per vehicle mile. This represents a plausible,
perhaps somewhat conservative estimate of current social costs of mobile source air pollution in
these urban areas. Evidence suggests a much higher pollution cost in the Los Angeles region and
perhaps a lower pollution cost in the Bay Area. The one cent per vehicle mile average fee would
total about $1.15 million per day in the Bay Area and about $2.9 million per day in the Los Angeles
Region, under base year (1991) VMT conditions. While the amounts sound high, annual receipts
from such a fee would amount to about 0,3 percent of the gross domestic product of each region. 18
Clearly, it would be inaccurate to simply charge each vehicle the regional average per-mile
emissions fee, since vehicles' emissions characteristics vary widely. We therefore analyzed two
possible methods for assigning a per-mile emissions fee to different vehicles. Under the first
method, the per-mile emissions fee would vary by model year and would be based on data on each
model year's average emissions characteristics (i.e., using EMFAC in California). Under the
second method, the per-mile emissions fee would vary with the actual emissions performance of
each vehicle, which might be determined through emissions testing, remote sensing, or on-board
emissions monitoring. The latter approach would account for the differences in emissions among
vehicles of the same model year.
18 We used the same one cent per vehicle mile average fee for the year 2010 analyses, lacking more specific
cost data.
B-24
U. S. Environmental Protection Agency
-------
Technical (Methods for Analyzing 'Pricing Measures
• ' to Reduce Transportation Emissions
For each'household in the four regional travel survey samples,19 we knew the make, model, and
age (year) of the vehicle 'holdings for the base year, and we knew how each vehicle actually was'
used on a representative weekday. Thus, we were able to provide a well-grounded assessment of
how vehicles of different ages and types are used and who would be impacted by emissions fees 20
However, we did not have access to a model of how household vehicle holdings or vehicle usage
- patterns would change as a result of differential changes in the per-mile cost of vehicle operations,
so we had to address these issues in terms of plausible scenarios rather than modeled estimates.
Fees Based on Average Emissions by Model Year:
For the average emissions by model year approach, we began by determining, for each region the
average daily within-region VMT and emissions for every vehicle in the regional travel survey.
We extracted from the survey data the vehicle trip sequences and their characteristics, and inferred
whether the trip was a cold start, etc., based on the time between trips in the trip sequence. We also
determined the average trip speed and distance, deriving these data from the applicable highway
networks. We then used EMFAC7F data specific to each vehicle model year to compute the
emissions for each vehicle trip. , •
From the resulting samples of vehicle trips and their associated emissions, average weekday
emissions and VMT were calculated for each model year on a region-by-region basis. Annual
emissions and VMT for each region were then estimated. The annual VMT estimates were used to
calculate total emissions costs for each region at the postulated one cent per mile average.
For the year 2010 forecasts, it was necessary to describe the likely vehicle age distribution and
patterns of use for that future year. We made the simple assumption that the 2010 fleet would have
the same general characteristics (age distribution, usage profiles) as the current fleet does. We then
applied EMFAC7F 2010 emissions factors to this'hypothesized future fleet's trips to determine the
future base case (total VMT and emissions, emissions by model year, etc.).
For both 1991 and 2010, we used our calculations of emissions, by vehicle model year to apportion
the regional emissions cost estimates among model years. The annual VMT calculations by model
year then were used to determine an average emissions cost per mile for each model year. For
example, from the 1991 data for Los Angeles, the average emissions fee per mile for a 1 year old
vehicle would be about 0.4 cents, while the average emissions fee for a 17 year old vehicle (from
the pre-catalyst era) would be about 7.0 cents.
Note that the method we describe here should apply only to miles driven within each urban area,
since emissions costs are calculated and apportioned on a regional basis. If the collection scheme
^used odometer readings as the basis for the VMT portion of the fee, some vehicle owners would be
charged for miles driven in other regions or in other states. To avoid this potential inequity,
The most recent regional survey for Los Angeles did not record vehicle make and model data. However,
the Caltrans statewide survey of the same vintage included these data and had enough observations in the Los
Angeles region to support ,the analyses described here. For this policy only, then, we extracted the Los
Angeles data from the Caltrans survey and used it for our analyses.
We calculated, impacts on the basis of 250 times the average weekday rate plus 115 weekend and holiday
days at 95 percent of the weekday rate. ' '
U. S. Environmental Protection Agency • .- • B-25
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
methods could be developed to estimate in-region and out-of-region vehicle use and apportion the
fee(s) accordingly, and credits could be given for documented out-of-state travel.
We analyzed the effects of our per-mile emissions fees varying by vehicle age, assuming that
households would not alter their vehicle holdings or pattern of use in response to the fees. This
assumption is not entirely realistic, since households could lower their fees by replacing their older
cars with newer ones, and if AVI measurements or odometer readings are the basis for the VMT
component of the fee, by using their newer cars in place of their older ones for some trips.21
Nevertheless, the analysis results provide an indication of the maximum travel impact and the
minimum emissions impact that such an emissions fee could be expected to have; without fleet
changes the full impact of the fee would be passed through as an out-of-pocket cost to the driver,
and the emissions reductions would come from reductions in travel rather than from the use of
newer, presumably cleaner, cars.
A more robust analysis would consider how vehicle holdings and usage patterns might change in
response to an emissions fee. The analysis would account for the determinants of household
vehicle ownership and use and would estimate the effects of an emissions fee on the number of
Vehicles owned, the vehicle makes and model years, and VMT per vehicle. Such a comprehensive
model was not available to us, but we did have STEP'S internal auto ownership model, which
estimates whether a household will have 0, 1, or 2+ vehicles as a function of household
characteristics, travel conditions, and vehicle ownership and operating costs.
We used the STEP auto ownership model to partially account for the effects on the vehicle fleet as
follows. For each region and analysis year, the base case household fleet was used to estimate the
average annual cost of auto ownership for each household. Then, revised annual ownership costs
were computed to reflect the addition of emissions fees for each vehicle (based on model year and
the actual daily VMT revealed in, the survey). New auto ownership probabilities then were
calculated using STEP.22
While this method is an improvement over simply representing the emissions fee as an increase in
out-of-pocket costs, we feel that on balance it still is likely to overstate travel effects and understate
emissions effects: For implementation scenarios involving AVI or odometer readings, households
with more than one vehicle could shift use among household vehicles to reduce their emissions fees
without cutting back on travel. Both the revenues from emissions fees and their impact on
households are therefore likely to be lower than' what we have estimated here.
21 Alternatively, VMT could be estimated based on averages by model year taken from survey data. This
might be simpler to implement than an approach requiring odometer readings, but would remove much of the
incentive for multi-car households to reduce "older car" use by substituting their newer, presumably cleaner
vehicles for certain trips.
22 Since STEP does not predict which-autos might be disposed of or what model years added when auto
ownership levels change, we imposed a series of assumptions. We assumed that; since the per-mile emissions
fee is higher for older vehicles, households that reduce their auto ownership levels would get rid of their
oldest car(s). We assumed that households maintaining their current auto ownership levels would also
maintain the age distribution of the vehicles they own. Households that added vehicles were assumed to add
car(s) of the average age and fuel efficiency for that ownership level. These assumptions allowed us to
estimate the effects on emissions, fuel use, etc.
B-26 .'• ' U. $. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Fees Based on Measured Emissions:
*To analyze an emissions fee based on measured emissions, we first needed an estimate of how
emissions vary within each vehicle model year. One possible source of such information would be
the data from .vehicle inspection and maintenance tests, but we did not have access to these data.
Therefore we used an alternative source, a database from Professor Donald Stedman of the
University of Denver, containing in-use measurements obtained passively with his remote sensing
device at a location on Rosemead Blvd. in Southern California.23 Stedman expressed these data as
frequency distributions of emissions by model year.
We used the Stedman data to develop a frequency distribution of emissions fees per mile for each
model year in each region. Taking the fleet age distribution and the VMT by model year estimated
from the regional survey data, we used the Stedman emissions distributions both to estimate the
aggregate emissions by model year and to apportion emissions responsibility within, model years.
This approach allowed us to assess a higher fee for high-emitting vehicles, and a lower fee for
relatively clean vehicles, within each model year.24
To estimate the effects of a measurement-based emissions fee, we first made a special STEP run to
create a base case with emissions derived from the high-emitter distributions rather than from the
pure EMFAC data. Since we did not have actual emissions measurements for the vehicles in our
samples, during this run we simulated the presence of high emitters in the fleet. Each vehicle in the
sample was randomly assigned an emissions level from the distribution for its model year (and
tagged with that emissions level for use in the "after" analysis). Then, the fee policy was tested
using the same method as for the fee based on model year, averages, except that in this case the
proposed fees were based on the emissions level,assigned during the "before" run.
A fee based on measured emissions would probably require new technology of one sort or another.
Tamper-resistant on-board monitoring and recording equipment would be the preferred approach;
fees based on multiple measurements using remote sensing equipment would be a second option.'
A third approach would be to use the emissions measurements from I/M testing, though this would
raise a number of issues including whether the fee should be prospective or retrospective and
whether it should be based on before-repair or after-repair measurements.
With an emissions fee targeting super-emitters, households could be expected to adroitly.
manipulate their vehicle holdings and use to minimize the impact of the fee. This would tend to
produce lower travel impacts and higher emissions reductions than shown here.
There is some reason for concern that emissions distributions recorded for a single location and operating
environment may not reflect the full spectrum of operating conditions, and thus cannot be assumed to
represent the "high-emitter" distribution for all regimes of urban travel. A similar criticism would apply,
however, to vehicle inspection/maintenance test measurements, which are based on a single measurement and
a specified operations sequence, or to any other data set based on single measurements and conditions.
An alternative approach would be to use the EMFAC data as the estimate of the average emissions by
model year, and to use the Stedman data (or another source) to represent the underlying distribution of
emissions for that model year. Note that the overall approach does, riot produce different results if a higher or
lower total emissions burden is assumed. '
U. S. Environmental Protection Agency • g_2y
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
B.4 Impacts of Transportation Pricing Strategies
B.4.1 Basic Analysis Results
This section presents analysis results for the set of transportatiompricing measures we analyzed for
our four case study regions. We present a series of 18 tables summarizing the basic findings of our
analyses, both by measure and by region. For each pricing measure, we present the predicted
percentage changes in VMT, trips made, and travel time; fuel consumed and CC>2 emitted; ROG,
CO and NOx emissions; and annual gross revenues25, for the years 1991 (the base year) and 2010.
The results for each analysis year represent the long-term effects of pricing measures, i.e., the
impacts resulting from the pricing measures as if they had been implemented several years earlier.
- • • ' " !• ,
The results show that carefully crafted and targeted transportation pricing strategies could do much
to reduce travel times (hence congestion), cut energy use, and reduce emissions, at the same time
increasing gross revenues substantially. At the same time, it is clear that auto use and its impacts
are quite inelastic with respect to most aspects of price. This has two important implications: first,
sizable increases in revenue can be obtained with relatively little effect on travel; conversely, large
price increases are necessary to obtain sizable reductions in travel and its externalities.
The results also provide an empirical dimension to the notion that the most efficient way to use
price as a mechanism for reducing transportation externalities is to price each externality in a direct
way. Thus, as the tables detail, the most effective pricing strategy for emissions control (in the
sense of emissions reductions per dollar charged) is to target high-emitting vehicles as precisely as
possible; the most effective strategy for reducing fuel consumption (and CC>2 production) is to raise
the price of fuel; the most effective way to reduce congestion- is to impose a toll at congested
locations; and so on. Note that we refer to efficiency and effectiveness here in a purely technical
sense. Other factors - ethical, institutional, political, and social - contribute to a broader assessment
that may lead to different conclusions about policy effectiveness.
Tables B1-B5 present the results organized by pricing measure for the year 1991. Tables B6-B10
present a subset of the year 1991 results, reorganized by region. Each regional table includes
analyses of the synergistic effects of groups of pricing measures, under two scenarios:
1. "Modest Pricing" - A relatively low set of prices from each category (e.g., $1.00 per
day parking; $0.50 per gallon fuel tax), coupled with enough investment in transit only
to maintain existing levels of service.
2. "Full Pricing" - A relatively high set of prices from each category (e.g., $3.00 per day
parking; $2.00 per gallon fuel tax), coupled with investment in transit corresponding to
build-out of each region's long-range transit plan (as expressed in future network files
made available by each MPO). Note that such a transit expansion would absorb a
significant fraction of the pricing revenues.
25 Net revenues depend on the specific implementation strategy selected (public vs. private sector
implementation and administration, technologies used, scope of implementation, timing of implementation,
etc.). In general, implementation designs costing a small fraction (5-15%) of gross revenues are feasible.
B-28 U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Tables B11-B18 present the same ensemble of results for the year 2010. The percent changes
presented are changes from a year 2010 base case, created by using STEP as a forecasting tool. The
regions' forecasts of households, household income, and household size (or population) were used
to "factor" the 1991 household file to create a year 2010 household file for each region. The STEP
models then were run to create a year 2010 "base case," using the 2010 household file plus the
MPO network data for the year 2010.26 Finally, policy analyses were carried out to predict
changes from the future base case.
In.the Los Angeles region, some adjustments were made to SCAG's highway,travel times after analyses
indicated that the SCAG models then in use showed far more trips and VMT than STEP'S more complex
models would predict. Otherwise, MPO level-of-service projections were broadly consistent with STEP
internal calculations and were used as provided to form the basis of the 2010 base case.
U. S. Environmental Protection Agency
B-29
-------
o
0>
ls
i i
tyj >2i
I?
*t
••dr 3
II
11
§
I
<3
3
<0
CO -R
CO
re
m
^
en
•<-
o
u.
01
e
a
O
3
z
o
o
M
O
o:
8
0)
3
U.
>,
(!)
D
CD
1-
Q.
>-
a.
£
£1
oj Q)
g o
CD *•
> CL
5
cn
uL
Description
5?
V-
CM
^-
?
^
V)
**
in
a?
m
o>
£
£
T
CO
T
o>
o
0
OS
I
OQ
i
g
O
55
*7
as
m
^
-9
,-:
1
.
'
CO
T-
§
O
o
9
rt
o
O
3
c
£
u
cn
co
1
s?
T
?-
CM
s?
CD
CN
0
d
1
s?
in
o
S
«?
o>
°
o
T-
u>
o
o
0)
o
Q
c
(0
(O
N.
CO
§
CM
g
to
~s
10
"
^
*
^
to
CM*
CM
§
g
'y
«
CM
e
o
s
o
u
a
o
CO
The congestion pricing strategy analyzed here assumes that prices
would be assessed on a per mile basis everywhere that congestion
appears In the highway network, including on arterials and collector
streets as necessary. A technology for electronic toll collection would
be required. Roadway message signs or In-vehicle readouts would
provide information about tolls on upcoming segments, likely as
part of a broader highway information system. Prices would not vary
minute-by-minute, but would be set to reflect average conditions on
each highway link during each period of the day, perhaps with
seasonal adjustments. The results shown here are based on a
reduction of congestion to level-of-servlce DIE, defined as a volume-
to-capacity ratio of .9. Note that travelers would continue to experience
some delay under this criterion, but that greater reductions in volume
might not be justifiable in economic terms.
§
I
s
I
I
-------
I B g
O
It
H
T S
a>
o>
i
I
I
Q)
42
"3
(4
O
.
3
_ Q)
§ i
= I
< *
o
1|
i £
s
c
_o
'5>
S
Description
if
0 CM
9 <*
s«
o
So
•o
f- CO
C3 OO
s
cVi o
•7 *?
CD T^
CM v
7 •?
S?
f"
,
T «?
T-
7 CM
m u>
CM' h:
T -r
•7-7
O
"3
o
-
en*
-------
I
I
I,
l
*-
o>
0>
Fuel Tax Increases - 1
.2
1
i!
.Q TO
I— "^
ra a
3 c
C 0
<1
CO
co
ffl
T-
o>
o>
s
u_
CD
B>
nj
O
s
z
8
o
o
a:
1
O
3
U_
1"
a
Q
o
I
K
in
a.
*
Js
a.
>
P
>
""" "5
S3 "**
^ «
Hi
c
1 Tax
1 Increme
c
o
(£.
c
o
z:
£.
u
8
Q
sill
§gg.g
•???*
HP
ssssss'-s
i-- *°. *? Iq
*? r» ir- *?
^ SS 5S S?
fi eo to co
§000
•J CM tO CM
*? CM Csl •^>
„ ^5 ^ ^
§5 5 5
• T ^i* V
gsgg
•?T™^
511?
?ssa
9" 9 9 9
O O O O
to o o o
0£J M CJ
£
r
fl§5
-5S5^^
?•; **t ^1 r>
^ £2 £2 co
§ssi
gggg
0 0 0
gggg
* T *•* *r
§ggg
"^?v^
?!fl
CO CO in CM
R **i N.
*? 5- *- hr
55 g g g
'
gggg
•T S CM ?
0 ^
uj r^ CM o
g£.g§
<7'?*1*
?ll?
rs co 10 CM
»- v O CM
0 ^) «S 0
o o o o
w o «=> o
S £>£> £
o
a
CD
•Q
E
»
(O en CM co
§T- O> CO
CM en o
g-.^g
'?'?°'!:
ssSi
^SSSS^
CO CM ^ QJ
cf «2 5; ic
' '
-? ^ ss ss
^ co iq to
' '
^5 &•- o~> &-
g 0 O ID
*? K CM ?!
'
000
^^ si
"? r^ C\l tj-
o °^ **t 5s
*? *^ T^ ^r
?Ils
CO CO 10 CM
^- V O CM
*?*?*?3ES™"2j-u>c;a»a>
sIitHl!|!!i!S!!i!!
llil|ll|s=il2|x-x-S|*|
- •o-oS«o-fex£l.££8SS-SS=||a
Nicfea-oi3'Es°-Sl'iS0«3
»!?^=°>£SSSS5'f^^J5|!gJ
sf5sai|is5g;ifR ?5|sf
||l^|||||l|| g f|l|f I*
t-a.cM4»m E « a 3S^ E £2o S
"•- 1
is <° 3
I •= »,
g -2 =5
(0 0] *°
* a .2
2 . . x
1 £§
11 ?
I?!
ill
>i ^3 e
|i §
IS > 5
> •§ S
« ^e *-
— o 0
i*" O U)
* 5 £
> » S
? &*
11 -s
2 •" «!2
"So
_2 to O
1 £ if
S ° §
Hi
« 0 0
S •? 2
>t cj -n
5 o >.
S 2 *
» S .2
P> c
>> S
s •§ s>
Ions of dollars per year; VMTdenoti
weekday vehicle-trips; time is week
xide; ROG Is dally tons of reactive c
= £ °
S S f
s&|
o* 1
m » ™
I s •s
B-~ «
£ 6 g
IB » S
= = >.
Ill
* E m
* ° ~
« r o
Sou
o
•z.
%
a
o
I
CO
a
CM
*?
«
-------
,o
•il
i
o>
o>
o>
I-
i
(0
42
0) ,
•
i |
S*
I
5?
5?
o
••¥
CM
t
S?
o
ss
5
§
-15
§
«
" -
r
-
l-
i!
-------
I
•
L : » "! .,
I
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Table B.5
Analysis Results for Emissions Fees - 1991
*
I Annual |
£
1
o
£
ca
en
j=
O
a
£
[Revenue!
O
o
o
CD
o
2
0
o
I
>
G)
a
CD
£
i=
.1
£
a.
5
^
E
^
0)
8
m
e
"3>
£
IDescription
O •*
CM CO
CO Ol
*9 I?
gtij
'
BS SS
u> **!
*? 3-
*
BSBS
?- o>
•f
«? V
-5 --S
it S>
g§
T T *
BS g?
0
«#«#
So «
*? T
B? BS
r- d
cji cvi
0 o
S J=
2
CD
£
U
expenditures Is not well developed. We looked at two broad strategies:
1) the emissions fee would be calculated based on EMFAC data for
average model year emissions per mile and actual VMT (so that the
per mile fee would vary only by model year); and 2) the emissions
CO T-
T CO
§»
BSg
"? S*
*
BS *
U) *t '
1^ ^
1
BSBS
O* CO
'f »r
B? BS
U) U3
*? V
0
c» co
~«?
gg
•? v
gs s?
"
7 T
« ^"
«# «?
01 a>
^ ^
w So
p CO O
•o >. >.
QJ JO iii
a ?. »
* ° o
S 2 o
E "o c
M CO n
0 ^ S
1-s i
> • S
^» *° 13
ctTo 5.
•o 2 J=
•" ffl 0
§ > c
Notes: Revenue expressed in millions of dollars per year; VMT denotes v,
of 10 microns or less; trips are weekday vehicle-trips; time is weekday
C02 is daily tons of carbon dioxide; ROG is daily tons of reactive orga
1
o
1 *
0
•3
1 .-
1
.1
?k
^
t§
""H
S
V
m,
-------
&Q &a
| §
§ 'ES
3i
"
•s:
1
a
OS
O)
'5
I
I
I
,<2
•Q 2
21
Z I
0)
£? ffl
B O
Q) »-
> o.
15
§
S o
s g'i
Iillfliif|Sal
'IfiilliitlHl
«.t^4=«j
<=J£— 01 P in °-3 — S • - o
.2
o J2
I
-------
CM
t
U
£
0)
I
42
£
£
11
£ a.
S
Descriptio
*;>
BS s?
co ^
9 «?
S? S?
co n
.
CO 1O
•7 •?"
21
9 «?
T «>
gg
«# 9-
SS 2?
10 a>
T «?
o o
=! «!
T- O
9
9 «*'
BS B?
o iq
T f
& B?
t »
T 7
B? S?
O CO
9 «#
Q
C
e esi
5?
(O
O) U)
cy w
5? S?
T- CO
o csi
O O
o o
*- CO
1
1
: -a o oa re
to « .£ .2
3 CD i«_
O O) O
« o a
§.» •§
» § §
s a o
^ 01 w
S ». w
o- = =
5
s & s
111
~° ™
• jo o>
ill
2 «c o
g i g
&= E
•S •§ g
* i!
§> ra ra
^S "
> £ o
•is >• £•
O ^ r=
!•§ «
g „ «
III
* °
•5 g £
m ^ f?
e i
Hi'
Hi
= 2 I
g n -a.
S s
i ° =
C k. *S
o o 45
11 -
S o o
O
z
-------
11
11
f §
8?
I'
S
c
c
" T
o
2 o
- •? CO
U3 CD tO
C^ CO *^ *^ 1
CO CO U> CM
T- T- O CM
ssss
,? cn u> eo
^ C* CM V-
: ^ ss
> m o
_r eo o to"
" 1"? T,
""S.«
O O O O
to o o o
o oi oi CM
^ T- S- °*
gg
CO ^ W rj
?"•?
o -5
cy « «2
« ? g a
o o o o,
|,il,a. .».
•3ii-it!S*j!
isllflf
s 5 f | s 5 s s
o
~
«e»
-------
I
.£
I
(0
.9
0)
S .12
OQ 5
« I
o>
c S
o
O
I
g
?
fr
CQ
§
ffl
2
u
IB
CO
8
«o
s
g
o
D>
£
O
I
5
CO
§
o
IS
o
o
0
in
— S — —
oroa))o
Ilflfl
S = S * „ S i.^1
• .- ai ~ ". ^* ^r m
iiiilllllll
5-'£»'o'»
lllMllll
-------
I
ll
=n • M
a .£
S'«S
J?§
I'l
a, ?
§>!
II'
e K
13
I
1
s
I'.
2
•8
o
.8
8
* «
CQ 'a
fl) ^.
SI
§ i
c o
IT!
ro
m
o
CN
O
U.
a
Q
8
z
O
f T
C9
n
ce
o
••^
O
LL
fO
°
p
b
"~
H
EL
^
=i
to
~Z
"- S
o
O)
S
(Description .
CO CO
••t;
B?§
'i?T.
s?*5 '
•ef ^
*T
5? SS
«O
-------
I
OJ
0>
CQ
O
O
I
i
(n
o
0) ^»
•
n =
10
n
m
en
T-
£
01
c
n
O
^
^N
z
O
o
0
i
1
D
P
1-
n
£
a
H;
|
Description
Strategy
^
^
evi
BS
1
s
S
Jg
B?
•?
^
V
jo
30
An Automatic Vehicle Identification (AVI) schem
would be used to price the regional freeway and
arterial system to maintain level-of-servlce (LOS
D/E
C 0
o o
O ffi O
Reglonwide Cong
Pricing (level-of-s
D/E) - Average $0
per Mile In Peak
o
^
CO
9
ss
9
i
*
B?
f>
CM
B*
T
BS
V
BS
9O
o
!»
All workers driving alone in the region would
experience a minimum $1.00 (1991) per day chai
for parking at the workplace.
*§
Reglonwide EmpI
Parking Charge o
Per Day
o>
T"
^
^
«?
g
g
^
T
BS
r-
co
^
Cjl
s?
CO
CM
0
O)
AH workers driving alone in the region would
experience a minimum $3.00 (1991) per day chat
for parkinq at the workplace.
ffl§
>\ 4»
Reglonwide EmpI
Parking Charge o
Per Day
CM
a
o>
BS
St
>.
•7
en
s«
a
$
•9
s?
m.
JjO
3>
CO
CM
s?
V
B?
CM'
IT
g
CM
B?
2
B?
CM
CM '
O
s?
ss
CM
d.
Q.
£
i
2
Q.
S
"5
M
0
O
U.
IFuel Tax Increase
by $0.50 (1991)
eo
IFuel Tax Increase
by $2.00 (1991)
£>
CO
I
^
^
»
BS
f
•f
V
i
BS
n
B?
CM
BS
T-
g5
3
§
5?
ID
T
3?
o>
£?
CO
en
-s
T
o»
S?
V
B?
10
C H>
Fees would average 1 cent per mile, and would
be collected frequently, as for the VMT fee.
In alternative 4a, the fee would be based on
annual mileage and average model year emissio
as reflected In EMFAC7F. In alternative 4b, the
fee would be based on actual odometer reading:
and in-use tailpipe measurements.
i I
~ <*
Mileage- and Emi!
Based Fee (Range
40-400/yr)
x
a> 2
~ 5*
Mileage- and Emi!
Based Fee (Range
10-1000/yr)
CO
B?
In
•f
g
B?
CM
^
CO
BS
'
B?
M
fe
Fees would be paid often, e.g., In the same man
as fuel taxes. This implies a potentially complex
collection scheme involving real-time reading
of the odometer, perhaps each time a vehicle Is
fueled. The simpler option of billing once a year
based on the odometer reading likely would hav
less effect on travel and emissions.
I
u.
3
S.
is
o •
£
r—
in
en
CO
n
CM
RS
C*
v
g
<«•
-14.4%
18.3%
B?
S
g
CO
35
I"?
;£
<9
m
r
1 , 2a, 3a, and 4a with maintenance of current tra
service.
1
0 f=
Example of Comb
Effects: Moderate
i
ss!
^^
c "w
1, 2b, 3b, and 4b with extensive transit investme
(Revenue not reduced to reflect cost of new tran
Example of Comb
Effects: High Imp:
I
I
IIS
pa
-------
£f
fl
•S I
1-8
o>
I
I
§
1
(0
to
Q>
o
«M Q;
2
as
3
§
< £
«
c
c
I
m c
m =
E c,
s -
u. c
x: c
j :
(r
^
i
{Description
(strategy
o 3
> T-
g
5 g
:° ,
X to
3 ^
J B?
' T- .
m m
1
ss
CO
T
0
i
An Automatic Vehicle Identification (AVI) scheme
would be used to price the regional freeway and
arterial system to maintain level-of-servlce (LOS)
D/E • - •
§ S .
Regionwide Congest!
Pricing (level-of-serv
D/E) -Average $0.04
per Mile in Peak
^
en
S?
BS
V
BS
V
g
*
?
to
V
CM
*
o
§i workers driving alone in the region would
perience a minimum $1.00 (1991) per day charg
r parking at the workplace.
o
«ii O
Inegionwiae employe
Parking Charge of $1.
Per Day
•a ' •
M
O
cn
CM
B?
f .
^
5
9
1 -
D
to
r- '•
*t'
B?
?.
Ei workers driving alone In the region would
perience a minimum $3.00 (1991) per day charg
r parking at the workplace.
§
D O
4»
•8
: o
if
.c
0
= 1
,3 Q
In*
S
B?
•i
•^
?
,0
?
§
•
0
v»
CD
CO*
g
^
g
O
1*."
CO
CM
ifi
s?
!O
s
Q
d
1
CO
£;
'
3
T*
'
>
1
hees would be paid at the pump.
by $0.50 (1991)
e.,-1 T-., i
W - J
by $2.00 (1991)
?
r-
BS
1r
B?
CO
'
r^
rq
to .
i
?-
c*>
9
s|
B?
CO
1 '
M
a •
»
5 ,
\ '
I
BS
CM
B5
en
CO
g
6 }
n
eO
~*t
BS
•
'
«
i, ta, ja, ana 4a with maintenance of current trans
service.
•s
example or uomoinea
Effects: Moderate imps
CD
CO
T-
ft?
U5
fS
^
so"
7)
(O
gs
CM
O
•7
CM
9 '-'•
CJ
>
CM
-
i, ZD, ^D, ana 4D witn extensive transit investment.
Revenue not reduced to reflect cost of new transit
Effects: Hiqh Impact
T
CQ
2 £
> « S
^
= ffl S '
to
t)
-------
1
1
i
I
eo
Q>
42
3
U)
Q>
o;
o .*.
II
5 S
c o>
C 9
in
S
^
en
o
£
c
en
O
0)
X
0
Z
o
u
to
o
u.
o
o
5
j
LL.
CO
0)
Q
i
~
0)
w
s
-
s
(Description
S
33
o
B?
" '
B?
«»
§
CM
91
CM
§
0
1
r-
1
i
j.
V
IAn Automatic Vehicle Identlflcatlon (AVI) scheme
would be used to price the regional freeway and
arterial system to maintain level-of-servlce (LOS)
D/E
c o
Reglonwide Congestio
Pricing (level-of-servlc
D/E) - Average $0.06
per Mile in Peak
T-
O
0)
g
?
BS
r-
§
^J~
§
T
eg
£
1
?
B?
•^
IAII workers driving alone In the region would
experience a minimum $1.00 (1991) per day charge
for parking at the workplace.
o
IRegionwIde Employee
Parking Charge of $1.0
Per Day
n
4
S
U)
£
CM
§
g
«?
g
Cy
§
§
n
c\i
B?
D
ri
IAII workers driving alone In the region would
experience a minimum $3.00 (1991) per day charge
for parking at the workplace.
o
IRegionwIde Employee
Parking Charge of $3.0
Per Day
A
CM
S>
lO
«?
ss
«?
ss
"
£
Cr>
0
'
£
•?
i
s?
7
§
"V
o
is
£
o
$
T
M
S
°!
CO
i
i
D
3
S.
D
3
to
0
0
u_
IFuel Tax Increase
by $0.50 (1991)
t
IFuel Tax Increase
by $2.00 (1991)
CO
§
BS
*
§
o>
7
£
*
§
*?
^
csi
B?
t
T-
»
i-
B?
o>
g
?
g
hJ
§
'
BS
C\l
g
ss
Csl
Fees would average 1 cent per mile, and would
be collected frequently, as for the VMT fee.
In alternative 4a, the fee would be based on
annual mileage and average model year emissions
as reflected in EMFAC7F. In alternative 4b, the
fee would be based on actual odometer readings
and in-use tailpipe measurements.
X
i S
Mileage- and Emission
Based Fee (Range App
40-400/yr)
•3-
X
A S
Mileage- and Emission
Based Fee (Range App
10-1000/yr)
^J-
Ol
«
BS
7
jg
*
B?
7
§
T
g
'
B?
o>
•p
BS
>i
^
7"
Fees would be paid often, e.g., in the same manner
as fuel taxes. This implies a potentially complex
collection scheme involving real-time reading
of the odometer, perhaps each time a vehicle is
fueled. The simpler option of billing once a year
based on the odometer reading likely would have
less effect on travel and emissions.
5
n.
s
o
»
5 •
L
CO
*
£
55
V
0)
CO
I0
<0
o>
0
s
!*•
~9
to
1, 2a, 3a, and 4a with maintenance of current transit I
service.
r,
Example of Combined
Effects: Moderate Impa
?
CM
CM
5?
CO
«
BS
CO
£
«
*
U)
£
to.
CM
!
B?
r>
3
1, 2b, 3b, and 4b with extensive transit investment.
(Revenue not reduced to reflect cost of new transit.) |
Example of Combined
Effects: High Impact
*p
p
Q,
I
-------
11
•^ tq
bo ~
11
II.
8
I
§
X-
4
s
(4
•2
a>
§
-j
Q>
I
42
II
TO 3
Jj; lr
< s.
V)
&
o>
tn
E
t!
LL.
C
CO
U
8
z
o
U
0
n
a.
n
—
3
U.
m
"o
Q
0)
p
n
1-
i
I
(Description
£
S
w
T~
»
«*
5?
IO
•9
s?
«
IO
g
'
§
<*
»
<9
^
c\i
ro
"'
An Automatic Vehicle Identification (AVI) scheme
would be used to price the regional freeway and
arterial system to maintain level-of-servlce (LOS)
D/E
i 8
Regionwide Congesti
Pricing (level-of-servi
D/E) -Average $0.10
per Mile in Peak
en
§
T-
s?
T
s?
*7
85
1 ,
S?
"*• .
*
f
s«
*T
§
All workers driving alone in the region would
experience a minimum $1.00 (1991) per day charge
for parking at the workplace.
o
Regionwide Employe
Parking Charge of $1
Per Day
CO
CO
i5
151
SS
CM
ss
CO
CSI
g
1
5?
"2
1
1
3?
o
*?
ss
1
All workers driving alone in the region would
experience a minimum $3.00 (1991) per day charge
for parking at the workplace.
o
1, 2a, 3a, and 4a with maintenance of current-transit j
service. |
CO
Example of Combined
Effects: Moderate Imp
0>
-«-
cn
^
s?
CO
CO
S?
T-
o
CO
h-
T
s?
T
T
S
RO
?
s? '•
IO
«?
1, 2b, 3b, and 4b with extensive transit investment.
(Revenue not reduced to reflect cost of new transit.) I
Example of Combined
Effects: High Impact
X * C
> « 3
>,*> v
fl-
IS i
IT ... oj
« « r-
« tt S
L
-------
•58
I'i
tl
I
ra
2
CQ
O
o
I
I
Q>
0)
«o Q;
CQ -a
o ^«
II
~ea :
3 1
< —
11
o ;=
ir
E
a
c
_o
0
CO
Q)
Q
>s
a
1
CO
N
CM
Jj
to
55
S?
'
g.
V
5?
"
?d
°> > 1*°*
a ^g
tt S a
"
sS
9
BS
X o
3 >
Q
•O
3
^
?
s?
CO
5?
in
~S
m.
'a
•7
CM
^
»
^
3
M
C\l
5
CO
g
T
S«
I2
1
ss
f*-
s.
^
CO
«9
g
^
5?
CM
^
^f
~
s?
CD
1
i-ees would average 1 cent per mile, and would
be collected frequently, as for the VMT fee.
In alternative 4a, the fee would be based on
annual mileage and average model year emissions
as reflected in EMFAC7F. In alternative 4b, the
fee would be based on actual odometer readings
and In-use tailpipe measurements.
X
2
0.
Q.
<
3,
i
£.
Str
L. £.
I-O 0
0) C3
S3
m °
X
A> S.
c a.
o o.
•z <
CO Q)
E g
a re
o£
Mileage- ar
Based Fee
10-1000/yr)
D
CM
CM
""
g
s?
•7
~5
5?
*
o
S?
•9
~?
•7
-S
en
«?
t-ees would ue paid otten, e.g., in the same manner
as fuel taxes. This implies a potentially complex
collection scheme involving real-time reading
of the odometer, perhaps each time a vehicle Is
fueled. The simpler option of billing once a year
based on the odometer reading likely would have
less effect on travel and emissions.
_
"
^1
0
r~
s«
2
•
s«
0
in
s«
en
en
t>
D
s«
CM
CO
•^f
0
i, 'iz, 3a, and 4a with maintenance of current transit 1
service. . |
**
g,
E
0}
S
•8
i °
-------
I
to *J
II,
_.
I I
x:
3 -? §, §
• v> > """*
° = JL.E
p-Sa s
a: D- Q Q
,-
CM
""
g
B?
1-
j=
T
B?
?
S
BS
T
g
BS
o
T
IAII workers driving alone in the region would
experience a minimum $1.00 (1991) per day charge
for parking at the workplace.
o
0.0
LU H>
o «
g 0> >i
> « o
0:0.0.
CO
en
5 ,
g
5S
CO
CM
B?
DO
CM
BS
«?
S
BS
m
BS
CO
BS •
CO
•y
All workers driving alone in the region would
experience a minimum $3.00 (1991) per day charge
for parking at the workplace.
o
o> e->
§•£
L °
a» £
1 0) >
II?
J ec o
CM
^. *
t
g
"
BS
en
BS
3
T
S?
OT
b
3?
D
1O
r>
?
T
i
3
a. •
3
3
3
a
j
i
3
>
}
>
)
» T-
X 0
) **
. J3
CO
in
v
S?
10
ci
IT-
§
CM
*
BS
T
BS
S
BS
CM
CM
BS
T
g
1
BS
CM
O
*
£ m
u en
X 0
K S
o> w
£.§
eo
CD
T-
g
^
g
ui
BS
»
BS
?
BS
iq
BS
0
*
o
M
BS
t
CM
0
T-
ss '
CO
T
BS
A
1
BS
S
^ •' •
1-^
g
>i
^
t
g
S?
D
CM
Fees would average 1 cent per mile, and would
be collected frequently, as for the VMT fee.
In alternative 4a, the fee would be based on
annual.mileage and average model year emissions
as reflected in EMFAC7F. In alternative 4b, the
fee would be based on actual odometer readings
and In-use tailpipe measurements.
X
1 <
: o
" &
S g«
i £ 5-
o> -o o
is?
sm §
I
A?'
"to ^
^ o>
*° £•
? 8S
CD "- 0
c> -a o
CJ CD °
» S T
is?
ja.
S
«
CO
BS
^ .
'
-5
* . ' '
^
•f '
S '
-5
f
BS -
BS .
•?
Fees would be paid often, e.g., in the same manner
as fuel taxes. This implies a potentially complex
collection scheme involving real-time reading
of the odometer, perhaps each time a vehicle is
fueled. The simpler option of billing once a year
based on the odometer reading likely would have
less effect on travel and emissions.
£
I
CM .
*
O
O
I
to
o
BS.
5?
to
3
•7
BS
i
BS
S
T
BS
?
BS
M
a
3
s|
1 , 2b, 3b, and 4b with extensive transit investment.
(Revenue not reduced to reflect cost of new transit.) I
•o
a +j
38
U £
5*
!• S
I
I
I
I
Ǥ
-------
V»*r .'.-•• .v „
I?
I
§
I
(0
to
1
V)
Q>
•SJ ^*
•Q 2
o
3
<
m
m
o
3
u.
a>
c
m
f
3
C
Q>
>
X
z
0
o
o
n
X.
CM
O
0
0)
3
U.
re
o
Q
o
1
n
£
Q.
t-f
Ssi
jS
E
c
«
u
a
>,
05
a
1
m
>
CO
O
a
CO
4
^T
^
*9
^
CO
^a
•*
^
y
^
|^
•^
An Automatic Vehicle Identification (AVI) scheme
would be used to price the regional freeway and
arterial system to maintain level-of-servlce (LOS)
D/E
Reglonwide Congestion
Pricing (level-of-servlce
D/E) - Average $0.09 •
per Mile in Peak
r-
B
CO
»
o>
9
s
o
=.
to
S5
5
s?
3
*S
o>
o
All workers driving alone In the region would
experience a minimum $1.00 (1991) per day charg
for parking at the workplace.
Regionwide Employee
Parking Charge of $1.00
Per Day
CO
CM
D
CO
-«
«*
ib
CM
g
CM
U>
CM
ss
*r
iS
CO
CO
s?
o
CN
^0
-Cf
CM
cu
All workers driving alone In the region would
experience a minimum $3.00 (1991) per day charg
for parking at the workplace.
Regionwide Employee
Parking Charge of $3.00
Per Day
a
CM
1^
-S
n
CO
•?
CO
«?
i:
?
ss
o
^e
to
•9
s«
•7
*?
en
eo
Fees would be paid at the pump.
Fuel Tax Increase
bv $0.50 (1991)
n
rw
CM
H
T-
g
~
?5
T
CO
CM
Fuel Tax Increase
bv $2.00 (1991)
JQ
CO
N
(0
T
^
«?
>
>
^
t
o
U>
>0
CM
^
CM
X-
to
CM
Fees would average 1 cent per mile, and would
be collected frequently, as for the VMT fee.
In alternative 4a, the fee would be based on
X
Mileage- and Emissions-
Based Fee (Range Appro
40-400/vr)
n
Tt
n
N
CO
N
O> r
in
T
*!•
to
SS
cl
5?
r^
^
o>
T
annual mileage and average model year emission
as reflected in EMFAC7F. In alternative 4b,the
fee would be based on actual odometer readings
and In-use tailpipe measurements.
X
Mileage- and Emissions-
Based Fee (Range Appro
10-1000/yr)
M
CO
CO .
CO
O
-r
^
"*
4
t
lO
°P
s^O
•?
-?
3
•i
^o
CM
Fees would be paid often, e.g., in the same manne
as fuel taxes. This implies a potentially complex
collection scheme Involving real-time reading
of the odometer, perhaps each time a vehicle Is
fueled. The simpler option of billing once a year
based on the odometer reading likely would have
less effect on travel and emissions.
VMT fee of $0.02 per mill
to
o
9
O
4
eo
^
^
xO
en
•7
§
CM
3^
f
ss
*T
s^
in
«P
*j
1, 2a, 3a, and 4a with maintenance of current tran
service.
Example of Combined
Effects: Moderate Impact
o>
CO
3
o>
LJ **
1, 2b, 3b, and 4b with extensive transit investmen
(Revenue not reduced to reflect cost of new trans
Example of Combined
Effects: High Impact
I
s;
.1
I
a
-------
I
-J
I §
I
o o
I
E2;
•i
i
o
-J
o
«
5
42
3 ,
8
oo a:
2
"<5 3
ZJ I
& 0
<
y.
CO
m
o
0
.»
IRegionwide Empic
Parking Charge of
Per Day
T~
10
^-
§
*?
CM
-S
3
CM
s«
to
°?
S?
M
7 .
b
*?
i>.
IAII workers driving alone in the region would
experience a minimum $3.00 (1991) per day charge
for parking at the workplace.
o
>.£»
IKeglonwIde Empic
Parking Charge of
Per Day
^,
K
CO
ss
"
I
V-
•f
en
s?
u>
°?
5?
^_
^
5?
en
CO
«i
T
it-ees would De paid at the pump.
Ii-uei lax increase
by $0.50 (1991)
U>
CM
§
CM
g
N
s?
CO
N
*D
^
§
CM
s?
!*•
CO
S?
"
b -•
I rue i lax increase
by $2.00 (1991)
D
tf\
0
—
§
•i
i
•*
U)
•9
1
g
10
5?
CO
•»
CM
)
CM
CO
g
"
CO
CO
5?
c»
CO
s
0
*?
^
en
BS
CNI
hees would average 1 cent per mile, and would
be collected frequently, as for the VMT fee.
In alternative 4a, the fee would be based on
annual mileage and average model year emissions
as reflected in EMFAC7F. In alternative 4b, the
fee would be based on actual odometer readings
and in-use tailpipe measurements.
X
en 2
: Q-
T; <"
Mileage- and Emls
Based Fee (Range
40-400/yr)
X
CO 8
c Q.
mileage- ana tmiss
Based Fee (Range
10-1000/yr)
D ' '
3
CO
B«
CO
N
CM
-i
s?
CM
»?
52 - ,
£2
0
°fc
«?
^
T
hees would be paid often, e.g., in the same manner
as fuel taxes. This implies a potentially complex
collection scheme involving real-time reading
of the odometer, perhaps each time a vehicle Is
fueled. The simpler option of billing once a year
based on the odometer reading likely would have
less effect on travel and emissions. .
in
cs
CM
V
O
N
5?
01
£
0
i
6?
10
CM
5?
o
5
s?
o>
CO
S?
i , za, aa, ana 4a witn maintenance of current transit 1
service. • " |
CO
a.
c
txampie or uomoin
Effects: Moderate li
O
ss
«t
o>
.
i
1
•Cf
*
CO
^
^_
n
^
\|
CN
«J
1, 2b, 3b, and 4b with extensive transit investment. I
(Revenue not reduced to reflect cost of new transit.) I
o
0 *j
5 §
• E
> a
| X
if
s.^- s
is ^ ^
30
C.
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
B.4.2 Equity
! ' ' !' I ','',. ' ' i
t , I1 , j
One of the biggest contentions raised about strategies that increase the price of transportation is
that, while some people would benefit, others would be unduly hurt. Whether from an ethical or a
pragmatic political perspective, these equity concerns, which stem from the possibility of unevenly
distributed benefits and costs, are a central implementation issue for transportation pricing. Price
increases are especially a worry for low income individuals who may not be able to afford the
higher costs and hence might be priced out of certain travel options. Higher transportation prices
also are a concern for moderate income people who have little flexibility about when or where they
travel and hence might have to devote a larger share of their income to transportation.
One would not want to overstate equity issues. First, one might argue that there is nothing
inherently unfair about expecting people to pay for the services they consume, to cover the costs of
damage they do to the environment, and so on, regardless of their socioeconomic status. In fact,
this could be seen as a more equitable result, since it removes undeserved burdens from others.
Second, it is important to note that for many pricing applications, and especially for congestion
pricing, the dollar cost is higher for those who pay it, but time and other costs decline; many people
should be better off despite the higher prices. Finally, for any of the measures, use of the revenues
to improve transportation services could result in net benefits for most. In short, simply noting that
prices are higher does not mean that the result is necessarily less equitable.
Nevertheless, it is important to have good information on the distribution of costs and benefits of
various transportation pricing strategies, including the status quo, so that the social and political
ramifications can be anticipated and dealt with and so that program designs can be structured to
achieve a satisfactory level of fairness. While full treatment of the equity issues of transportation
pricing would require a separate study, a portion of our effort was devoted to exploring the impacts
of the various pricing strategies on different groups and interests. Indeed, the analysis procedures
described here were designed to produce as much information about the distributional
consequences of pricing as possible.
1 • ' ' ' I
The distribution of impact and equity can be thought of along many dimensions - income, class,
race, ethnicity, age, sex, and geography are among those commonly considered. For the illustrative
purposes of this section, however, we have chosen to focus our attention primarily on differences
by income level. We split the households of California into five equal household income groups,
and used the resulting quintile boundaries to categorize our findings throughout the analysis of
pricing policies. The five quintiles are:
Quintile Income Range (1994$)
1 <= $18,700
2 $18,701-$36,500
3 " $36,501-852,100
4 $52,101-$71,300
5 ' >=$71,301
B-48 U. S. Environmental Protection Agency
-------
Technical Methods far Analyzing Pricing Measures
' , to Reduce Transportation Emissions
Tables B19 and B20 present a distillation of quintile data based on the 1990 U.S. Census Public
Use Microdata Sample for California. It may be helpful to begin a discussion of equity by first
looking at some basic facts about the distribution of income in. California, as shown in these tables.
U. S. Environmental Protection Agency
B-49
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
table B.19: 1990 US Census California Statewide Summary
Public Use Microdata Sample
Share of Households in Each State Income Quintile
Region
Sacramento Region
San Diego Region
San Francisco Bay Area
South Coast
Balance of State
California Combined Total
Share by State Income Quintile
1
0.22
0.19
0.16 .
0.19
0.27
0.20
2
0:22 -
0.21
0.17
0.19
0.24
0.20
3
0.21
0.21
0.19
0.20
0.20
0.20
4
0.20
0.20
0.22
0.20
0.17
0.20
5
0,15
0.18
0.26
0.22
0.12
0.20
Total
Households
598405
885574
2242554
4560620
1744923
10032076
Share of Population in Each State Income Quintile
Region
San Diego Region
San Frandsco Bay Area
South Coast
Balance of State
California Combined Total
Share by State Income Quintile
1
0.17
0.15
0.11
0.15
0.21
0.15
2
0.21
0.20
0.15
0.18
0.24
0.19
3
0.22
0.22
0.19
0.20
0.22
0.20
4
0.23
0.23
0.25
0.22
0.19
0.22
5
0.17
0.21
0.31
0.25
0.14
0.23
Total
Population
1560521
2386031
5852335
13233643
4930037
279962567
Share of Autos in Each State Income Quintile
Region
Sacramento Region
San Diego Region
San Francisco Bay Area
South Coast
Balance of State
California Combined Total
Share by State Income Quintile
1
0.13
0.10
0.08
0.10
0.17
0.11
2
0.19
0.17
0.13
0.15
0.22
0.1 e
3
0.22
0.22
0.18.
0.19
0.23
0.20
4
0.25
0.25
0.26
0.24
0.22
0.24
5
0.21
0.26
0.36
0.31
0.18
0.29
Total
Autos
1080383
, 1577796
3941140
8077199
3163821
17840339
Share of Resident Workers in Each State Income Quintile
Region
Sacramento Region
San Diego Region
San Frandsco Bay Area
South Coast
Balance of State
California Combined Total
Share by State Income Quintile
1
0.09
0.08
o:os
0.08
0.11
0.08
2
0.18
0.17
0.12
0.15
0.21
0.16
3
0.23
0.22
0.18
0.20
0.24
0.21
4 5
0.27 0.23
0.26 0.26
0.27 0.37
0.26 0.32
0.25 0.19
0.26 0.30
Total
Workers
715029
1145517
2993791
6237629
2010851
13102817
B-50
U. S. Environmental'Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Table B.20: 1990 US Census California Statewide Summary
Public Use Microdata Sample
Autos per Worker in Each State Income Quintile
Region •
Sacramento Region
San Diego Region
San Francisco Bay Area
South Coast
Balance of State
California Combined Total
Autos per Worker by Income Quintile
1
2.16
1.75
1.89
1.75,
2.37
1.93
2
1.62
1.37
1.38
1.28.
1.66
1.40
3
1.48
1.33
- 1.30
1.24
1.47
1.32
4 5
1.38 1.36
1 1.33 1.36
1.25 1.27
1.23 1.28
1.37 1.42
1.27 1.30
Regional
Average
1.51
1.38
1.32
1.29
1.57
1.36
Work-at-Home Share in Each State Income Quintile
Region
Sacramento Region
San Diego Region
San Francisco Bay Area
South Coast
Balance of State
California Combined Total
Share by State Income Quintile
1 2 3 4 5
0.041 0.029 0.030 0.026 0.037
0.042 0.033 D.026 0.032 0.039
0.052 0.034 0.030 0.027 0.034
0.035 '0.023 0.022 0.022 0.032
0.044 0.038 0.032 0.031 0.047
0.041 -0.030 0.026 0.026 0.035
Regional
Average
0.031
0.034
0.032
0.027
0.037
0.03
Commute Time per Worker in Each State Income Quintile
Region
Sacramento Region
San Diego Region
San Francisco Bay Area
South Coast
Balance of State ,
California Combined Total
Minutes per Worker by Income Quintile
1 ' . 2 3 4 5
19.17 20.28 21.55 22.89 - 22.53
21.72, 21.84 22.28 23.23 23.20
23.24 23.64 25.35 26.20 26.37
25.65 25.30 25.90 26.81 27.28
18.00 18.70 19.45 20.47 20.26
22.84 23.04 24.03 25.20 25.83
Regional
Average
21.71
22.65
25.65
26.46
19.55
24.63
Drive Alone Share for Workers in Each State Income Quintile
Region
Sacramento Region
San Diego Region
San Francisco Bay Area
South Coast
Balance of State
California Combined Total
Drive Alone Share by Income Quintile
1
. 0.65
0.62
0:54
0.58
0.64
0.59
2
0.71
0.67
0.61
0.64
0.70
0.65
3
0.76
0.73
0.67
0.70
0.75
0.71
4
0.79
0.77
6.71
0.75
0.77
0.75
5
0.77
0.80
0.73
0.80
0.79
0.78
Regional
Average
0.75
0.74
0.69
0.72
0.74
0.72
U. S: Environmental Protection Agency
B-51
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
" : I ' '
By definition, each income quintile contains one fifth of the total number of households in the state.
But the distribution of household income within the state is uneven; there are notable differences
among regions. For example, the San Francisco Bay Area is relatively well off, with 48 percent of
its households in the top two quintiles and only 33 percent of its households in the bottom two
quintiles. In contrast, the small urban and non-metropolitan areas of the state have just 29 percent
of their households in the top two quintiles and 51 percent in the bottom two quintiles. While
housing prices and other cost-of-living factors may cloud the comparison somewhat, it seems clear
that the ability to pay higher transportation prices is not distributed evenly around the state, but is
higher in its metropolitan areas.
Other important points can be observed be examining the income quintile data. In particular:
• Population is not distributed evenly among the quintiles. Higher income households
tend to be larger, such that 23 percent of the population is in the highest quintile and
15 percent in the lowest quintile.
• Auto ownership increases with income. 53 percent of the vehicles for personal use in
California are owned by the top two quintiles, while only 27 percent are owned by
the bottom two quintiles. This suggests that policies which cause a general increase
in the cost of auto ownership may apply disproportionately to upper income groups.
• Households with workers tend to have higher incomes than those which do not. 56
percent of the workers statewide are in the top two quintiles, while only 24 percent
are in the bottom two quintiles.. This suggests that policies which cause a general
increase in the cost of commuting may apply disproportionately to upper income
groups.
• Autos per worker is consistently high in all income groups. Table 20 shows that
quintile 1 - the lowest income group - has the highest auto ownership per worker.
This counter-intuitive result is due to the large group of retirees falling into that
quintile. Removing the retirees from the data base produces a ratio of autos to
workers of 1.25:1 for each of the five quintiles. While this does not have direct
implications for pricing policy, it does suggest that access to an automobile for the
commute is widely distributed in California.
• Drive-alone share for commute .travel rises with income. The drive-alone share statewide
is about .59 in the lowest quintile and .78 in the highest quintile, with similar variation in
each region. Putting the mode shares (including the shared ride data not shown here)
together with the proportion of workers in each quintile, it becomes clear that only about
6 percent of the commute vehicles statewide will have drivers in the lowest quintile,
while about 35 percent will have drivers in the highest quintile.
• Commute time per worker rises with income. The average self-reported commute trip
time statewide is about 22.8 minutes for workers in the lowest quintile and 25.8 minutes
for workers in the highest quintile, with similar variation in each region. Because many
of the low income workers' miles are made by transit (or by foot) at speeds far below
auto speeds, even on congested networks, it is clear that higher income workers' trips
(J, S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
must be considerably longer (in VMT) than those of their lower income counterparts.
This illustrates a crucial point for pricing studies: higher income workers are the largest
contributors to work trip VMT, partly because high income jobs and high-end housing
are relatively sparsely distributed around each region.
• Both low and high income workers are more likely to work at home. About 3.5 percent
of workers in the highest quintile and 4.1 percent of workers in the lowest quintile listed
home as the primary place of work in 1990, compared to 3 percent of workers overall.
While these phenomena are not well understood, it is said that participation rates by
upper income households have been increasing in recent years. This may indicate that
upper income households have an important way to blunt the effect of large price
increases, namely by choosing to work at home some of the time.
.The PUMS data demonstrate one of the most important facts about equity of the current
transportation system. Truly poor people make relatively little use of the highway system as it
operates today and, consequently, would pay comparatively little under most transportation pricing
scenarios (in absolute terms, not necessarily as a share of income).
Equity Analyses Using PUMS and STEP
An unstated implication of the PUMS analysis is that the lower middle class - say, quintiles 2 and 3
- would sustain much of the impact of pricing policies. This hypothesis was explored in a range of
analyses using STEP, examples of which are shown in Tables B21 and B22. The STEP analysis
framework allows us to examine equity issues in detail because it utilizes specific demographic
information, at the individual household level, that can be associated directly with the effects of
each pricing policy.
U. S. Environmental Protection Agency
B-53
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Table B.21
Equity Implications of a VMT Fee in the Los Angeles Region -1991
VMT Fee
(cents/mile)
1
2
3
4
5
6
7
8
9
10
Absolute Change in Daily VMT by Income Quintile
Q1
-1.8
-3.4
-4.9
-6.2
-7.4
-8.5
-9.5
-10.5
-11.3
-12.0
Base VMT (millions) 25.5
Per Capita Daily VMT 11.7
Q2
-1.9
-3.7
-5.4
-7.0
-8.6
-10.1
-11.5
-12.9
-14.2
-15.4
45.0
17.3
Q3
-1.4
-2.8
-4.1
-5.5
-6.8
-8.1
-9.3
-10.5
-11.7
-12.9
54.8
19.1
Q4
-1.1
-2.2
-3.3
-4.4
-5.6
-6.7
-7.8
-8.9
-10.0
-11.1
71.9
22.0
Q5
-0.5
-0.9
-1.5
-2.0
-2.6
-3.2
-3.8
-4.5
-5.1
-5.8
92.8
25.8
Total
-6.6
-13.0
-19.2
-25.2
-31.0
-36.6
-42.0
-47.2
-52.3
-57.3
290.0
20.0
VMT Fee
(cents/mile)
1
2
3
4
5
6
7
8
9
10
Percent Change in Daily VMT by Income Quintile
Q1
-7.0%
-13.3%
-19.1%
-24.3%
-29.1%
-33.5%
-37.4%
-41.0%
-44.2%
-47.2%
Q2
-4.2%
-8.2%
-12.0%
-15.6%
-19.1%
-22.4%
-25.6%
-28.7%
-31.5%
-34.3%
Q3
-2.6%
-5.1%
-7.5%
-10.0%
-12.4%
-14.7%
-17.0%
-19.2%
-21.4%
-23.5%
Q4
-1.5%
-3.1%
-4.6%
-6.2%
-7.7%
-9.3%
-10.8%
-12.4%
-13.9%
-15.4%
Q5
-0.5%
-1.0%
-1.6%
-2.2%
-2.8%
-3.5%
-4.1%
-4.8%
-5.5%
-6.3%
Total
-2.3%
-4.5%
-6.6%
-8.7%
-10.7%
-12.6%
-14.5%
-16.3%
-18.0%
-19.7%
Note: Quintiles defined in terms of 1989 Census household incomes.
VMT is vehicle-miles traveled in millions per day. Sales tax relief,
Improved transit, and other potential expenditures to mitigate
impacts on lower income households are not reflected here.
B-54
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Table B.22
Equity Implications of Congestion Pricing in the
San Francisco. Bay Region -1991
Average
Peak Fee
(cents/mile)
1
2
3
4
5
6
7
8
, ; 9
10
Absolute Change in Daily VMT by Income Quintile
Q1
-0.2
-0.3
-0,4
-0."5
-0.6
-0.7
-0.8
-0.8
-0.9
-0.9
Base VMT (millions) 7.2
Per Capita Daily VMT 10.0
Q2
' -0.2
-0.3
-0.5
-0.6
-0.7
-0.8
-0.9
-1.0
-1.1
-1.1
14.0
15.3
Q3
-0:1
-0.2
-0.3
-0.4
-0.5
-0.6
, -0.7
-0.7
-0.8
-0.9
19.6
16.8
• Q4
-0.1
-0.1
-0.2
-0.3
-0.4
-0.4
- -0.5
-0.6
-0.7
-0.8
30.3
19.5
Q5
0.0
0.1
0.1
0.1
0.2
0.2
0.3
0.3
0.3
0.3
44.0
22.6
Total
-0.6
-1.1
-1.6
-2.1
-2.5
-2.9
-3.3
-3.6
-3.9
-4.2
115.0
18.3
Average
Peak Fee
. (cents/mile)
1
2
3
4
5
6
7
8
9
10
Percent Change in Daily VMT by Income Quintile
Q1
-2.2%
. -4.2%
-6.0%
-7.5%
-8.8%
-10.0%
-11.0%
-11.8%
-12.4%
-12.9%
Q2
-1.2%
-2.3%
-3.3%
-4.2%
-5.0%
-5.7%
-6.4%
-7.0%
-7.5%
-8.0%
Q3
-0.6%
-1.1%
'-1.7%
-2.2%
-2.6%
-3.0%
-3.4%
-3.8%
-4.2%
-4.5%
Q4
;-0.2%
-0.5%
-0.7%
-1.0%
-1.2%
-1.5%
-1.8%
-2.0%
. -2.3%
-2.6%
Q5
0.1%
0.2%
0.3%
0.4%
0.5%
0.5%
: 0.6%
0.6%
0.6%
0.6%
, Total
-0.5%
-1.0%
-1.4%
-1.8%
-2.2%
-2.5%
. -2.9%
-3.2%
-3.4%
-3.7%
Note: Quintiles defined in terms of 1989 Census household incomes.
VMT is vehicle-miles traveled in millions per day! Sales tax relief,
improved transit, and other potential expenditures to mitigate
impacts on lower income households are not reflected here.
U. S. Environmental Protection Agency
B-55
-------
Technical Mefhods for Analyzing Pricing Measures
to Reduce Transportation Emissions
Table B2I presents results for VMT fees in the Los Angeles region at levels ranging between 1
cent and 10 cents per mile. The STEP analysis shows that daily VMT is skewed heavily toward the
upper income quintiles - the highest income quintile accounts for about one-third of total VMT,
while the lowest quintile accounts for less than 10 percent. Nevertheless, the absolute drop in VMT
resulting from a VMT fee is largest in quintile 2 (the second lowest income level) and smallest in
quintile 5 (the highest income level). The absolute drop in VMT is of the same basic magnitude in
each of the first four quintiles, and the percentage drop is progressively larger the lower the income.
level. (Percentages are shown in the second part of the table).
Table B22 presents results for congestion prices ranging from one cent to ten cents per mile, on
average, for the San Francisco Bay Area. Here we find that absolute VMT decreases are roughly
the same among the lowest four quintiles, while VMT for the highest quintile actually rises (as one
would expect for high-value-of-time travelers).
Comparable analyses for parking fees in the San Diego region and fuel taxes in the Sacramento
region (details not shown here) yielded similar results: The largest VMT decreases per capita are
concentrated in the four lowest income categories. While the VMT drop should not be read as a
pure decline in mobility, since some trips are still made by other modes, shortened, etc., it does
show how the mobility impacts would be distributed in the absence of efforts to improve modal
alternatives for impacted individuals.
A
Another way to think about equity is in terms of the per capita daily payment by each quintile.
Based on Table B21, the quintile total payments for a 5 cent VMT fee in the Los Angeles region
would be:
Quintile
' 1
2"'":
3
. 4' '
5
Daily Payment (million
, 0.9
1.8
2.4
3,3 •
4.5
Out of a daily total of $12.9 million, 35 percent is paid by the top quintile and 61 percent is paid by
the top two quintiles. Similarly, only about six percent of current fuel taxes are paid by members
of the lowest income quintile and 10 percent by the second quintile. Thus, while the
travel/mobility impact falls disproportionately on the lower income quintiles, the financial burden
falls squarely on the upper income quintiles.
It is harder to say how VMT fees and vehicle emissions fees would affect different income groups;
we can estimate impacts on trip making and location clroice, and can forecast auto ownership levels
by income group, but we have no direct evidence on how the various groups would change the type
and age of the vehicles they own in response to new fees (our analyses on vehicle type and age
changes were based on assumptions provided to the models rather than computed outputs of the
models). Nevertheless, available data do provide some insights into equity impacts. Using data
collected by Caltrans as part of a statewide travel survey, we find that about 55 percent of the
vehicles over eight years old are owned by the top three income quintiles, mostly as second, third,
or even fourth or fifth cars. The remaining 45 percent of the older cars are owned by the two-fifths
B-56
U. S. Environmental Protection Agency
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
of the households with low or moderate incomes. To the extent that vehicle registration fees fall
most heavily on these older vehicles, they also would fall somewhat more heavily than proportional
on low and moderate income households.
\
Implications
As with any change in tax policy, the distributional consequences of the proposed change should be
carefully examined: The distribution of the burden of the proposed tax among income groups
should be compared with the distributional consequences of tax alternatives, including the current
tax system. If adverse equity consequences are deemed to be significant, policies to ameliorate
those burdens should be examined, including the use of tax revenues to benefit those who might be
disproportionately affected by the tax change (such as increased funding for public transportation
services, use of revenues to provide tax exemptions (life-line rates) for low-income users, or use of
tax revenues to replace existing taxes that disproportionately burden low-income taxpayers).
U. S. Environmental Protection Agency
B-57
-------
Technical Methods for Analyzing Pricing Measures
to Reduce Transportation Emissions
References
1.
2.
3.
4.
5.
6.
i •
Greig Harvey, STEP: Short-Range Transportation Evaluation Program, prepared for the
Metropolitan Transportation Commission, Oakland, California, 1978.
Greig Harvey "Methodology for Incorporating Transportation System Effects into
Regional Transportation Energy Demand Forecasts," report prepared for the California
Energy Commission under Contract No. 400-82-023, September 1983.
I
D. Pickrell, "What Gasoline Taxes Can - and Can't - Do," paper presented at the Graduate
School of Architecture and Urban Planning, UCLA, Los Angeles, March 16, 1993.
; •' • ' f '
D. Pickrell, "VMT and Vehicle Fuel Economy Elasticities," Memorandum, Volpe
National Transportation Systems Center, Cambridge, Massachusetts, February 14, 1995.
G. Orcutt, Policy Evaluation through Discrete Microsimulation, Second Edition, Brookings
Institute, Washington, D.C., 1976.
J.D. Wardrop, "Some Theoretical Aspects of Road Traffic Research," proceedings of the
Institute of Civil Engineers 2(1), 1952, pp. 325-378.
B-58
U. S. Environmental Protection Agency
-------