Transportation-Air Quality Planning:
Current and Future Analysis Needs
by
Elizabeth Deakin
Greig Harvey
Deakin Harvey Skabardonis
with
David Reinke
Cambridge Systematics
for the
United States Environmental Protection Agency
Region IX
75 Hawthorn St
San Francisco, CA 94105
September 1992
Revised January 1993
-------
Page III
Table of Contents
Table of Content* fii
Preface v
CHAPTER 1: INTRODUCTION 1-1
CHAPTER 2: ADDRESSING CURRENT ANALYSIS NEEDS 2-1
2.1 Introduction 2-1
2.2 General Data and Modeling Needs 2-4
2.2.1 Introduction 2-4
2.2.2 Review of Basic Modeling Issues 2-4
2.2.3 A Basic Ensemble of Data and Models 2-6
2.2.4 Variations 2-9
2.2.5 Getting From Here to There 2-11
2.3 Specific Transportation-Related Air Quality Analysis Requirements 2-15
2.3.1 Emissions Inventories, Forecasts, and Tracking 2-15
2.3.2 VMT Baseline, Forecasts, and Tracking 2-16
2.3.3 TCM Assessment 2-18
2.3.3.1 Analysis of Section 108 Measures 2-18
2.3.3.2 Other TCM Analysis Issues 2-23
2.3.4 Conformity Assessment 2-32
2.3.4.1 RTP Conformity .. 2-34
2.3.4.2 TIP Conformity 2-35
2.3.4.3 Project-Level Conformity 2-36
2.4 Project-Level CO Analysis 2-38
2.4.1 Basic Issues 2-38
2.4.2 Alternate Analysis Methods 2-39
2.4.3 Particular Analysis Issues 2-41
CHAPTER 3: LONG-TERM IMPACTS OF TRANSPORTATION INVESTMENTS ... 3-1
3.1 Overview 3-1
3.2 What We Know 3-1
3.3 How to Address the Problem 3-8
CHAPTER 4: TRANSPORTATION PRICING: ANALYSIS AND IMPLEMENTATION
ISSUES 4-1
4.1 Background 4-1
4.2 Institutional Setting 4-2
4.3 Prototypical Pricing Concepts 4-4
4.4 Critical Analysis Issues 4-7
4.5 Pricing Analyses for the Bay Area 4-10
4.6 Analysis Approach for the South Coast 4-17
4.7 Implementation 4-17
Transportation-Air Quality Analysis Issuss January 1993
-------
Psgetv
WHMMHHMW
-------
P*ge v
Prefect
This document was prepared under Contract No. 68D90073 with the US Environmental
Protection Agency. Cambridge Systematica, inc. (CS) was the prime contractor and Deakin,
Harvey, Skabardonis, Inc. (DHS) was subcontractor for the work presented here.
Mark Brocker of EPA Region IX served as technical representative for this effort and provided
ongoing support and assistance. Douglas Eisin'ger and David Catkins of EPA Region IX
provided review comments on portions of the report
Elizabeth Deakln and Greig Harvey of DHS authored Chapters 1-5. David Reinke of CS*
Berkeley, CA, office authored Chapter 6. John Suhrbier of Cambridge Systematics, Inc., the
prime contractor, served as principal investigator for the wort and provided, assistance to the
authors at many stages of the work.
Materials on methods for addressing current analysis needs and on research on development,
in Chapters 2 and 5 of this document, also appear in a somewhat lengthier form in the Manual
of MPO Modeling Practices for Transportation and Air Quality Planning, prepared by DHS for
the National Association of Regional Councils under grants from EPA and the Federal
Highway Administration. The materials on surveys presented in Chapter 6 of this report will
be added, in edited form, to the NARC manual in Its next release. This sharing of resources
for the two projects has greatly benefitted both.
Notwithstanding EPA's financial and technical support, the views expressed in this report are
those of the authors and not necessarily those of the Environmental Protection Agency. The
authors are solely responsible for the material presented herein including any errors or
omissions.
Tnn§portatlorhAlr Qutllty Analysli Issubs
January 1983
/
-------
CHAPTER 1: INTRODUCTION
This document gathers together a series of discussion papers on analysis needs for
transportation-air quality planning, prepared during 1BB2 for the U.S. Environmental Protection
Agency, Region IX. Each of the chapters which follow covers a topic which EPA has
identified as important not only to air quality agency staff, but also to staffs at federal and
state transportation agencies and at metropolitan planning organizations in air quality
nonattainment areas.
The five following topics are addressed:
¦ Methods for addressing current analysis needs - an overview of -current analysis
practices and recommendations for incremental improvements (Chapter 2).
¦ Effects of added transportation eapadty - a discussion of potential short-term and
longer-term travel, land use, and growth impacts of new transportation investments
(Chapter 3).
¦ Analysis of transportation pricing - a review of the analytical issues that arise with
pricing strategies and a discussion of the implications for regional transportation
analysis methods (Chapter 4).
¦ Research and development needs - an overview of near-term and longer term
improvements that should be made to models used in practice (Chapter 5).
¦ Approaches for improving travel survey data • a discussion of the information than can
be obtained from travel surveys and a review of modem survey methods, with a
particular focus on panel surveys (Chapter 6).
The report is designed to serve as a reference for EPA staff as they assess regional analysis
capabilities and identify needed improvements for transportation-air quality planning. The
report also identifies a number of areas where research and development is needed in order
to improve the knowledge base for transportation and air quality planning, and to strengthen
federal, state, and local agency capabilities in transportation-air quality analysis.
Transportation-Air Quality Analysis Issuas
January 1993
-------
CHAPTER 2: ADDRESSING CURRENT ANALYSIS NEEDS
2.1 Introduction
The Clean Air Act Amendments of 1090 tat forth a number of analyais requirements for
transportation, including: 1) development of a baseline emissions inventory; 2) VMT tracking;
3) VMT forecasting; 4) long-range plan analysis; 5) TIP analysis; 8) TCM analysis; and 7)
project-level CO analysis. The statutory basis for these analysis needs was reviewed In
Chapter 2. In this chapter, each analysis need Is discussed in turn and issues to be
considered in fashioning an analysis response are set forth.
Two broad assumptions color the treatment of issues: 1) Analysis simply for procedural
completeness is not good public policy, and ultimately could undermine both the goals of the
Clean Air Act and Vie credibility of the analyst; and 2) resolution of some Issues Is hampered
by deficiencies in the state of Knowledge, and H is important to distinguish such deficiencies
from ones which could be overcome with straightforward investments of resources, It is
assumed that conformity analyses should be substantive, not simply "going through the
motions"; hence analysts should consider all factors which might reasonably be expected to
have a significant effect on the outcome, including factors whose evaluation may necessitate
improvements to data and models. On the other hand, certain difficulties In analysis posed
by Clean Air Act requirements (or other public policy concerns) may not be totally resolved
or resolvable; analysis has limits. The objective thus must be to interpret CAA requirements
in ways that are consistent with what models can do, as well as what MPOs can do.
The following sections go into some detail about analysis issues and approaches,
emphasizing practical ways to meet the spirit of the Clean Air Act However, the material
presented is intended to supplement, not substitute for, guidance on modeling and analysis
promulgated by FHWA, FTA, and EPA. Moreover, no attempt is made to dictate a uniform
"best practice". Instead, K is acknowledged that local conditions vary to a degree that requires
each region to chart Its own course through the analysis imperatives of the Clean Air Act
This will require a meeting of minds among local governments, regional planners, state
transportation officials, environmental and other interest groups, EPA, and FHWA/FTA on a
strategy for data collection, model development, and model application that yields needed
improvements at a rapid but prudent pace. Thus, the primary goal of the chapter is to clarify
the issues that must be resolved specifically for each region.
In the tailoring of a strategy for local conditions, a number of general principles will apply:
¦ Precision vs. Accuracy - The Clean Air Act now provides incentives to be both as
precise and as accurate as possible in emissions calculations. However, precision
must not be confused with accuracy. Considerable resources could be wasted
pursuing precise, but inaccurate, numbers. Sound data and models are a prerequisite
to producing accurate results, and in some cases sketch planning approaches
Tranaportatlon-Alr Quality Analysis /aauaa
January 1993
-------
Pag* 2-2
producing reasonable but no! very precis* estimates may be preferable to the exercise
of highly detailed models which have been poorly specified or which can be run only
on low quality data.
¦ Validity of Procedures - The Clean Air Act - through its facilitator) of citizen
enforcement, among other aspects • provides incentives for completeness and validity
in the analysis procedures used where the accuracy of analyses is at stake. In other
words, when there is agreement in the profession that a specific modeling procedure
or element could be important to the outcome of emissions analyses, it might be risky
for the MPO to ignore K; where an existing modeling procedure is hard to Justify
theoretically, Its continued use also may be risky.
¦ Model Improvements - There has been a pervasive underinvestment In regional
transportation models and data over the past 20 years. The large MPOs have not
been able to afford needed model improvements and emergent MPOs typically have
developed only rudimentary modeling capabilities. In general, model developments
have focused on a narrow set of project planning capabilities rather than the broad
spectrum of analysis issues raised by the Clean Air Act (and ISTEA). As a result, few
MPOs are prepared at this time to apply 'state of the practice" methods to the full
spectrum of analysts needs.
¦ Resource Constraints - The greatly expanded planning resource base provided by
ISTEA will help the MPOs to improve their data bases and models, but still will fall
short of what would be needed for immediate improvement of modeling capabilities
across the complete spectrum of issues. Improvements will have to occur
incrementally over a number of years.
¦ Timing of Model Improvements - Even if unlimited resources were available, some
types of model improvements would take years to complete. For example, good
survey data, Census data for corroboration, and well crafted networks are
prerequisites for model development Vet it typically takes two years to conceive,
design, test, administer, code, check, weight, and tabulate a good home Interview
survey. And, as all experienced modelers can attest, the prooess of model
development itself requires time and flexibility for experimentation with alternate
specifications and for identification and correction of model deficiencies. An MPO
beginning the survey process right now (1B92) likely would not be able to produce a
full set of new models until lata 1994 or 1995.
¦ Strategy for Model Improvements - Federal investment In transportation
infrastructure is too important politically and economically to be suspended while
planning capabilities are improved. On the other hand, such investment is too
important to the environment and social fabric of a region to ignore important analytical
deficiencies. This suggests a need for. 1) commitment to a strategy for making
critical model improvements as quickly as resources will allow, and 2} agreement on
reasonable assumptions about phenomena that the MPO cannot model, at least in the
near term (e.g., trip chaining, time of travel for each trip purpose.)
'ransportatlon-AIr Quality Analyal§ /sane# January 1993
-------
P*g» 2-3
¦ Six* of the Ar*a • The size of a metropolitan area Is an important determinant of the
resources available for transportation planning. However, the cost of developing good
planning tools is not necessarily correlated with size (network development is; survey
data collection and demand model development may not be). Moreover, the nature,
extent, and severity of an area's air quality problem is not necessarily correlated with
size.
¦ Nature of the Air Quality Problem * The specific pollutants for which a metropolitan
area is nonattainment should be taken into consideration in fashioning an analysis
capability. Also, the severity of the air pollution problem and the iocation(s) at which
it occurs should be taken into account1
¦ Regional Growth Dynamic - Rapidly growing and rapidly changing areas are more
likely to show land use effects of transportation investment than are areas where
overall growth is slight Slower-growing areas nevertheless may exhibit other
important land use changes which should be reflected in modeling efforts, e.g., shifts
of population and jobs to suburban areas.
¦ Mix of Modal Alternatives - Areas exhibiting a broad mix of modal alternatives
require a more complex modeling approach then areas dominated by auto travel. On
the other hand, various forms of ridesharing may be the chief travel alternatives in
euto-dominated areas, calling for a sophisticated approach to the modeling of auto
occupancy.
¦ Mix of Policy Alternatives - Areas considering more complex policy options • such
as land use and urban design interventions, or extensive or innovative transportation
pricing measures - will require more sophisticated analysis strategies than areas
relying on capital investments and operations changes to conventional highways and
transit
These complex - and often conflicting - fectors underscore the need to fashion a modeling and
analysis response that is matched to the needs and issues of the metropolitan area.
1 The area(s) required to be modeled lor afr quality purposes may not exadfy correspond to 9m area(t) the MPO
currently model*. Whan the nonattainment area It larger than the MPO boundaries, the MPO generally should tola
steps to enlarge its analysis area accordingly. In soma cases this may require special arrangements with neighboring
jurisdiction* When tha nonattainment area is smaller than the MPO modeling area, It usually wtn be possible to
adjust model runs andtot outputs to focus on the nonattainment area
Transportation-Air Quality Ana lya Is /seises January 1993
-------
Paga 2-4
2.2 General Data and Modeling Needs
2.2.1 Introduction
A number of modeling issues are common to many or all of the transportation planning
requirements of the Clean Air Act The purpose of this section is to introduce these modeling
issues. Subsequent sections deal with each type of analysis in detail.
The section begins with an enumeration of basic elements that should be present In every
transportation model, whether a region is large or small, fast-growing or alow growing, multi-
modal or auto dominated, severely polluted or moderately polluted. It then discusses the
principal variations in modeling requirements that arise from differences among regions. For
example, many of the small MPOs wilt need to make improvements simply to acquire basic
modeling skills, but may not require more extensive modeling investments. In comparison,
most of the large, well established MPOs already satisfy basic modeling standards, but will
require significant investments to address their more complex policy questions. Recognizing
that no region is in a position to fully address the Act's analysis requirements or to
immediately correct all deficiencies, the section ends with a discussion on setting priorities for
model improvements and devising interim analysis approaches to compensate for known
shortcomings in the models.
2.2.2 Review of Basic Modeling Issues
The transportation requirements of the' Clean Air Act pose significant challenge} for
transportation modeling and analysis. Based on their experiences with new conformity
determinations, on eariy drafts of EPA Clean Air Act guidance, and on efforts to date with
transportation control planning (both under the new federal legislation and under state law,
especially the California Clean Air Act), MPO staffers have reported broad concerns about the
degree of accuracy of transportation and emissions models in comparison with the reliance
being placed on them in transportation/air quality planning. They worry that reviews of the
standard four-step travel model system, focusing on the theoretical, econometric, and statis-
tical validity of the model hierarchies, component models, and data, have been sharply critical,
and that "accepted" practice may be open to successful legal challenge if it ignores key
phenomena or treats them in a way that is known to be inaccurate. 8ome of the specific
issues they have raised are as follows:
¦ Basic data such as household travel surveys reporting demographics, employment,
income, and trip-making; link volume and transit passenger counts; vehide
occupancies; parking prices; land use and employment data; emissions inventories;
and pollutant concentration data are frequently missing, dated, or too sparse to
support the detailed analyses being sought.
¦ The historical performance of large scale model systems has not been wed
documented, but several areas have had problems, reported in the literature, with fore-
casts of new transit ridership and predictions of other transit policy changes.
Transportation-Air Quality Analysis Itsuaa
January 1993
-------
Paga 2-5
Forecasts of highway volumes and speeds have been leu oflan publicly critiqued but
a! least in some key instances have reportedly been no more accurate.
¦ Some individual model components are more reliable than others. For example, mode
choice models often perform well, but the explanatory power of trip distribution models
is troubling.
a Many models use only highway travel time as an accessibility measure, aven when
transit accessibility is arguably very important
¦ Models often lack detail on household demographics which research Indicates as
being important for example, the number of children In the household, age and sex
of the traveler, ate. Some areas even lack data on household income (except, in the
aggregate, from Census data.)
a Network coverage varies considerably in terms of comprehensiveness (what
percentage of facilities are represented} and detail (how many categories of volume-
delay formulae are used to describe the facilities); many areas believe they lack
sufficient detail for some of the analyses they are expected to do.
¦ The accuracy of mode! estimates of Ink volumes and especially link speeds is often
poor.
¦ There are major gaps in knowledge about trip timing and trip chaining, and large, weak
assumptions must be made to handle these matters. For example, most areas
assume trip timing will remain the same in future years, and deal with trip chains only
in terms of the percent of trips that are non-home based.
•
a The ability to represent TCMs is mixed. Pedestrian and bicyde cosasures-andjmaayL
urban design options typically cannot be analyzed using the regional models. In some
areas, pricing measures also ire difficult to mode! because data sets lack detail on
parking and other auto operating costs, and/or models do not include price and
income variables (except in mode choice.)
¦ Regional models also are not well suited to assess many of the large changes in
transportation systems and operations being contemplated for the future, including:
o Intelligent Vehicle-Highway System (1VHS) technologies, i.e., 'smart highways"
o widespread road pricing
o fundamental change in land us« policy.
¦ Few models have been validated as integrated systems, although individual
component models are checked.
¦ Emissions calculations require inputs in far more detail than current transportation data
or models can produce, and therefore require numerous assumptions and extensive
post-processing.
Transportation-Air Quality Analysis laauaa
January 1993
-------
Pagt 2-6
¦ Documentation of models and applications generally ** not extensive or detailed
enough for outside reviewers.
While this list of common shortcomings may seem formidable, many MPOs have in fact made
significant progress in improving practice in a number of these areas. Thus the Ost can be
viewed more favorably as the set of improvements, many of them with good examples already
available, which many MPOs will want to implement over the next few years.
2.2.3 A Basle Ensemble of Data and Models
Every region affected by the Clean Air Act will need, at minimum, a good conventional
network-based travel modeling capability for creating grid cell emissions estimates and for
carrying out conformity analyses. The following features of such a model should be
considered basic.
¦ Zone System - The zone system should provide a spatial structure that 1) is
consistent with census data aggregates; 2) is consistent with the boundaries of major
political jurisdictions; 3) maximizes the betweervzone variation and minimizes the
wvthin-zone variation of key attributes; and 4} minimizes the proportion of vehicular
travel occurring wtthirvzones. in general, smaller zones (and larger numbers of zones)
will more easily satisfy these requirements. White the number of zones has often
been constrained by computer software and hardware limitations (e.g., oomputer time
for traffic assignment and matrix operations increases with the square of the number
of zones), modem RISC-based work stations have brought ample computing power
within reach of every MPO, including the smaller ones.
¦ Highway Network - The highway network should include all facilities in position to
cany significant interzonal traffic - limited access, arterial, and (when necessary)
collector. Enough categories should be, defined within each facility type to support a
representative range of volume-delay curves. Volume-delay parameters should be
selected to reflect actual rather than nominal performance, including free-flow speeds
in excess of 55 mph and effective lane capacities in excess of hypothetical values.
Critical or unusual facilities (such as major intersections and toll booths) should be
coded individually to ensure that delay is modeled correctly.
While the number of links in a network has been an issue in the past, modem work
stations with graphics-based topological editors make it relatively simple to manage
very large networks. The burden of database management can be eased (or shared)
by combining the link library for network modeling with the general purpose catalogue
of street characteristics now maintained in some regions (typically for pavement
management purposes).
ai Land Use - The basic data needed are population, average Income, average
household size, workers per household, employment by major category, and housing
stock; all should be available by zone for the most recent travel survey year, for the
current year, for the base year (ft different from the current year), and for the target
Tranaportatlon-AIr Quality Analyala laauaa
January 1993
-------
Paga 2-7
year(s). Whenever possible, additional demographic variables (age, sex,number of
children, household members with a disability, etc.) should be obtained on a
household or zonal average basis, along with additional land use data such as square
footage by land use category, measures of density, and measures of development
quality (price, rents, office dass, ate.), home ownership levels, tenure, ate.
m Trip Generation - At minimum, a simple (cross)-dasslfieation table ahould be
developed for each major trip purpose, ft is useful to distinguish at least five trip
purposes, including; home>based work, home-based shopping, home-based school,
home-based other, and non-home-based. The dimensions of the tables will vary with
purpose, and will depend somewhat on the range of demographic data available tor
each zone. For example, school trips are predicted more accurately by children per
household than by persons per household, but regional databases do not always
indude explicit variables for the number of children.
A basic travel model invariably addresses vehide trip generation rather than the larger
universe of person trip generation by all modes. While this is understandable as a
simplification of the process and as a short-term expedient, MPOs should be
cautioned that direct vehide trip generation leaves a model vulnerable to errors from
changes In land use charaderisties (especially residential and employment densities),
which are highly correlated with vehide trip making but not with person trip making,
if significant land use changes are contemplated in future studies, then K may be
sensible to initiate the model system on a more robust person trip basis rather than
retrofit the model later on to compensate for errors in vehide trip generation.
a Trip Distribution -Atrip distribution model should be developed for each trip purpose.
Transformed highway time (peak or off-peak) can be used as a simple impedance
factor, although areas wKh significant transit use should consider a more sophisticated
metric such as the logsum of the mode choice model, and areas contemplating pridng
strategies should consider adding -cost as well as time to the Impedance. Trip
generations (attractions and productions) can be used as simple attractiveness
measures.
a Peaking - Average AM and PM peaking fadors should be developed for each trip
purpose, with adjustments for interdistrid movements that are large enough to support
separate calculations.
a Trip Tables - Using the peaking fadors, trip distribution tables should be reconfigured
into total trip tables by time period (AM Peak, PM Peak, Off-Peak). If the trip
distribution tables represent person trips rather than vehide trips, this conversion also
requires knowledge of average vehicle occupancy by trip purpose, inter-regional and
off-model trips should be added to the tables to provide a complete pidure of traffic
in each period.
¦ Traffic Assignment - Constrained multi-path assignments should be earned out tor
peak conditions, and the resulting travel times should be recyded back to the trip
distribution models until a quasi-equilibrium is reached. In order to avoid mindless
Tranaportation~Alr Quality Analyala laauaa
January 1893
-------
Pagt 2-9
repetition, this step requires a careful Interpolation after a few Iterations tiava been
completed. -
In the model development phase, the trip distribution-traffic assignment loop should
he tested and adjusted ao that volumes and speeds are "replicated" lor a
comprehensive sample of count locations.
¦ Survey Data - Up-to-date home Interview survey data should he collected, to serve
as a resource for model development and to provide a direct source of Information on
mobile source emissions. As a general rule'of thumb a new survey every decade or
so is desirable. More effective sample design and use of special-purpose surveys
such as transit on-board surveys and employee commute surveys are a potential
source of data for more frequent updates (or could be used as an Interim data source
for some analyses), although this rarely has been successfully earned out In practice.
In addition, special surveys (or counts) will be needed to account.for off-model
(commercial and interregional) trips In the network.
a Validation Data - A comprehensive set of traffic counts (by vehicle type) and floating
car speed observations should be assembled as a basis for model validation and VMT
tracking. MPOs almost certainly will want larger and more rigorously gathered
samples than now collected for HPMS (even under the updated HFMS guidelines).
In addition, It may be desirable to collect data on a broader cross-section of the
highway system, Including local streets, and to do special studies to account for
interregional travel and Intm-zonal trips (VMT and average speed).
a Emissions Estimates/Inventory • The final step in an analysis for air quality planning
is the translation of link-specific travel estimates into emissions estimates, which are
used, with adjustments to account for "off-model" emissions, both to prepare emissions
inventories and to evaluate plans, programs, and projects. Typically this step involves
gathering data from "representative" links and using these data to estimate the
detailed link-level information that is sought as Input to the emissions models. These
inputs include: hourly data on traffic volume by vehicle class (light-duty auto, light-duty
truck, medium-duty truck, heavy-duty truck, motorcycle, and urban bus), fiiel type
(gasoline, diesel), and catalyst/non-catalyst: average link speed; and trip star! data.
Post-processors complete with default values for the representative links are
increasingly available for this step; their use should be preceded by a careful review
of their suitability. Note that for emissions inventories and attainment demonstrations,
It is necessary to account for all emissions, Including "off-mode!" emissions (generally
on local streets, collectors and arterials not represented In the model plus VMT
generated by any travel not represented in the model system.) Post-processors to add
these VMT and/or emissions also are available.
To summarize, the minimum prerequisites for a credible estimate of emissions Include: 1) a
complete network for the region, In sufficient detail to capture at least 85 percent of the
interzonal travel; 2) an ability to generate plausible vehicle trip tables based on current and
future land uses and travel options In the region; 3) software to assign the full spectrum of
vehicular traffic to the network; 4) sufficient field observations of traffic (average speeds,
TrantporUVon-AIr Quality Analyala laauaa
January 199S
-------
Paga 2-0
average daily volumes, average peaking factors for specific links that are-directly identifiable
in the network) to calibrate the traffic assignments for base year data; 5) software to calculate
•missions based on network flows and link speeds; S) software or procedures to account for
additional "off-moder transportation amissions; and 7) estimates of future land uses sufficient
to allow projections of future amissions.
2*2.4 Variations
The previous subsection describes travel modeling capabilities as they should exist in every
region affected by the Clean Air Act's transportation requirements, from the smallest to the
largest In many regions, notably the larger and mora complex urban areas, additional
features and capabilities are required to produce travel and emissions estimates with the
desired level of accuracy and flexibility. Among the phenomena to which these capabilities
must respond are:
m Implications of Modal Complexity • Additional networks, assignments, trip tables,
and peaking analyses are required to represent transit systems and, where there is
separate infrastructure, HOV systems (and their effects on the ridesharing mode).
Mode choice models must be present in these cases, as well. Where transit is a
aignificant contributor to interzonal accessibility (as It Is In many large U.S.
metropolitan areas, especially for the poor and for trips to the CBD), the accessibility
indicator used in trip distribution should Incorporate transit as wett as highway times.
As noted earlier, the simplest way to achieve such a complex hierarchical relationship
is to use the log of the denominator of a logH mode choice model as the accessibility
measure. This is both relatively simple computationally and recognized as the
theoretically correct approach.
a Implications of Land Use Complexity • Where land use densities differ among zones
by two or three orders of magnitude (e.g., as measured in persons per residential
acre), variations in vehide trip generation can be quite large. In these cases, It is
necessary to add land use variables to vehide trip generation equations (either by
adding a density dimension to each vehide trip attraction and production table, or by
dealing with person trip generation and splitting out the vehide trips in a later step).
Without such adjustments, there will be a tendency to over-predict vehide trips in
denser areas and under-predict trips tn less dense areas.
It is desirable to adjust the granularity of the zone system (i.e., select zone
boundaries) to provide for homogeneity of land use and population characteristics
within zones. Typically, this implies more fine-grained zones and (possibly) trip
generation and trip distribution modeling for several dasses of workers (e.g.,
manufacturing, service, other).
¦ Implications of Land Use Change - If a region is growing or undergoing significant
internal rearrangement of land uses, density attributes become even more important
For example, new development that is less dense than the regional average is likely
to produce a higher-tharvaverage number of vehide trips per unit, yet most existing '
vehide trip generation models do not allow land use characteristics to influence trip
Tranaportation-Alr Quality Anafyafa laauaa
January 1993
-------
#»«?• 2-10
generation for a given household category. Models to configured are likely to under*
predict VMT growth in the face of land use change.
A more controversial issue Is the degree of need for a land use allocation model
and/or a regional growth model capturing the effects of investment and aoeesslbfltty
on the spatial arrangement and amount of economic activity. As Chapter 3 makes
dear, the dominant style of analysis Involves the use of one or more fixed future
projections of land uses, treated as a "given" unaffected by transportation investments.
Vet most planners recognize that the location and level of infrastructure investment will
have at least some influence on the locus of economic activity. This implies a
potential effect on VMT and emissions forecasts, and on the outcome of conformity
assessments.
There is not a consensus among practitioners on the "right way" to model the land use
effects of infrastructure. However, workable models based on tested theories of
spstial interaction are available. Many of the largest MPOs have implemented or are
in the process of implementing such models simply to be in a position to offer an
informed opinion about the extent of land use effects from (and on) transportation
plans.
In one mode of operation, the land use models serve largely In a sensitivity testing
role. Analysts run the land use models to evaluate whether and to what extent the
land use pattern would shift in response to proposed transportation changes. If the
shift is significant, they then re-run the transportation models with the new land use
pattern to determine the resulting travel changes, or in some cases simply adjust the
core transportation model outputs by hand.
Looking at their conformity obligations, as well as their broader multimodal planning
responsibilities under the ISTEA, a number of larger MPOs likely will implement land
use models in the next few years no matter what (some are already doing so). As the
models are improved and their range of validity is established, K may become feasible
both to bring them more explicitly into the core modeling framework and to make them
available to smaller MPOs.
Variations In Travel Cost - Basic models typically assume that auto travel costs are
roughly proportional to highway travel times. In a region with few priced parking
spaces or toll facilities, this assumption is approximately correct by definition because
the out-of-pocket expense of driving an automobile is roughly proportional to distance.
However, where parking charges and tolls are present, travel choices will depend on
prices to the extent that price variation differs from time variation.
Where toll facilities exist or are contemplated, or parking prices exist or are expected,
the MPO must ensure that prices are adequately represented in transportation models.
For parking price, this implies a presence in the mode choice models and, to the
extent parking prices are a factor in destination attractiveness, a presence in trip
distribution. A linkage with trip distribution occurs automatically if the mode choice
logsum is used as the attractiveness variable.
TransporUtJon-AIr Quality Anafyala Is bum
January 1993
-------
Paye 2-11
For tolls, the modeling implications are somewhat more complex. Tolls not only have
the same mode choice and destination choice affects as parking prices, but usually
influence route choices as well. Given a suitable conversion factor (e.g., an average
value of time), tolls can be included in the generalized price (utility) of each affected
link or can be converted to a time equivalent and inserted as a link penalty. The
resulting traffic assignments will reflect the influence of tolls on route choices.
The appropriate treatment of tolls in mode choice becomes somewhat unclear in the
presence of demonstrable route choice affects. When tolls influence some but not all
- of the feasible routes between points, average time and cost for the mode choice
model must be calculated as weighted averages for tolled and non-tolled paths'
In general, price effects (hence value of time) have been shown dearty to diminish
with increasing income. Thus, wherever price appears in the model structure, its
influence should be inversely proportional to income. This can be accomplished in the
mode-choice models by dividing the price by some function of income. Such an
approach is not possible for route choice under commonly used traffic assignment
methods, and route choice results simply must be taken as representative of the
average traveler. (Because income is not among the available variables in route
choice (assignment) in the most commonly used software packages, it can be
incorporated only by contracting for special programming.)
a Peaking - Field studies show a significant variation in peaking characteristics with the
degree of congestion in a corridor, in a region like the San Francisco Bay Area, peak
hour flows will range from 11 percent of dally totals in uncongested corridors to 7
percent in a highly-congested corridor such as the Bay Bridge. Modelers in the Los
Angeles and New York areas report peak percentages as low as 6.5 percent. While
there are no accepted models of peaking available for general use, any region that
expects to experience a change in the level of congestion over time or among
alternatives should have a well documented procedure or simple model in place for
adjusting the peak hour percentages, for example, estimating the peak hour
percentages as a function of the 24 hour volume-to-capacity (V/C) ratios.
2.2.6 Getting From Here to There
As of this writing, no region's model exhibits all of the desired features, and no region is dose
enough to a "state-of-the-practice" modeling capability to have H all in place for the next round
of transportation-air quality analyses. Planning resources are increased significantly under
ISTEA, but not enough to "buy" all of the needed improvements immediately. Furthermore,
some of these improvements will require several years of data collection and model
development for complete realization. This suggests a pair of questions: 1) in what sequence
should data and model improvements be implemented? and 2) How can informed dedsions
2. A mora rigorous approach would involve representing the choice of rout* at tubcidary to tha choice of mode,
using a formal nested choice framework. Such a modal has not yet baan developed for practical use.
Trangportttlon-AIr Quality Analy$ls /etties
January 1993
-------
Paga 2-12
be taken before a region's data and models are improved? Neither of these questions can
be answered In detail without reference to a particular region's context, but a number of
general observations are possible.
Prioritizing Improvements
Substantive model improvements should begin right away In each region. In most cases, the
Initial steps will Involve data collection and data analysis. With few exceptions, each region
will need the following Items:
¦ an up-to-date travel survey, to provide information for local model development The
survey should elicit key household information, including income, vehicle availability,
and household structure, and should include at least a one-day diary of vehicular and
non-vehicular trips in sufficient detail to support accurate geocoding. Sample size will
vary from region to region. After a period of economizing with small samples, many
regions have gone back to collecting the largest samples they can afford (e.g., 5000
or more observations, and as many as 10000-15000 observations in the largest
regions), both to ensure the maximum flexibility in later use and to have adequate
samples for subarea and corridor analyses. However, useful information can be
extracted from samples as small as 500, If that is all a region can afford. Also, a well-
designed supplementary survey of perhaps 500 observations could be used to obtain
missing data and then could be combined, via rigorous statistical procedures, with the
larger survey for analysis and forecasting purposes. (i.e., *n MPO could collect a
survey which includes data omitted from their larger survey • in some regions, this
would include data on incomes; prices paid, including parking price; and
characteristics of the vehicles owned.)
¦ an expanded set of traffic observations, to allow for accurate VMT and speed
monitoring and to provide data for model validation. Data about fleet mix by time of
day must be one of the observations made.
¦ an O/D survey at regional cordon lines, to provide estimates of internal-external traffic
origin-destination patterns.
¦ an expanded highway network, covering all freeway, arterial, and key collector
facilities. (A key collector facility is one which carries significant interzonal traffic, e.g.,
1000 trips/day.)
¦ a set of zonal demographic and economic data, updated to reflect the most recent
(1990) Federal Census results.
As soon as they are available, household survey data should be used to develop an improved
understanding of travel behavior in the region, and to make incremental improvements to
existing models. For example, an outdated (or "transferred") trip generation model might
immediately be replaced with a revised cross-classification table based on the new survey,
with a more sophisticated trip generation model developed as time and resources permit
Transportation-Air Quality Analysis Issuaa
January 1993
-------
Paga 2-13
Household survey data alto should be used to develop a deeper understanding of the origins
of mobile source emissions. With full geocoding and verification, and after base network
level-of-servioe Information is attached, the raw survey data can be used In a sample
enumeration framework to make accurate emissions estimates for a variety of trip categories.
Such an approach can draw on the survey for trip type, household, and vehicle data; on the
networks for trip speeds and distanoes; and on CPA's most recent MOBILE series model for
funning emissions and trip start emissions by trip type* In both San Diego and the Bay Area,
where extensive analyses of this type have been carried out numerous high-emitting trip
categories (such as high school and college trips) have been identified.
With an of the data in place, the MPO can undertake a (possibly) more far-reaching set of
model improvements, designed to achieve the level of complexity and sophistication deemed
appropriate for local conditions. For a small MPO with recently-acquired modeling
responsibilities, this may be a basic model that simply meets the core structural requirements
listed above (but with "home-grown" coefficients). For a large MPO with a sophisticated
model already in place, this would imply a refinement of models to address omitted or
problematic phenomena. For example, in highly-congested regions, a more formal
representation of the relationship between corridor-level peaking and congestion may be
sought in order to more accurately analyze future delays in a saturated network. Also, a
number of areas will want to improve their ability to represent transportation-land use
interactions in their models. For areas with complex transit choices, detailed nested models
might be developed (e.g., auto vs. transit; within auto, rideshare vs. drive alone; within
transit, rail vs. bus; within rail, light rail vs. heavy rail (or express vs. local); competing
eccess modes for each.) For areas with little transit but much ridesharing activity, the
rideshare options might be modeled in detail (e.g., rideshare vs. drive alone; within rideshare,
vanpool vs. carpool; within carpool, household members only vs. unrelated individuals;
access modes for each shared ride option.) In short, each MPO would decide what model
improvements are most important, given Ks existing models and the transportation issues and
options in the region, and would pursue improvements accordingly.
Interim Analysis Methods
As model improvements are being developed, the MPO may need to apply interim analysis
methods to accomplish certain of its responsibilities. Such methods need to be credible even
though they may be less detailed (accurate or precise) than the more sophisticated or more
formal methods to be used later.
The most basic and immediate need is for a credible current emissions inventory and a
projection of that inventory to the attainment year, as a basis for discussions over the level
of emissions reductions to be assigned to transportation sources. Most major MPOs are in
3 Currently, MOBILE outputs Induda tip dart •missions In tha ovaraB running amissions. Howavar, 1ha analyst can
produca trip start outputs by running tha modal tor a numbar of scanarios - zare cold start, zaro hot start (running
amissions only); 100% eoJd start, zaro hot start; 100% hot start, zare cold start - and comparing tha results to produca
astimstas of cold start and hot start amissions California's EMFAC program, in contrast, diradty providas cold start,
hot start, and running amissions astimatas.
TransporUtlon-AIr Quality Analysis lasuaa
January 1993
-------
P«0» 2-14
good position to produce at toast a first-cut 1990 Inventory, using HPMS data and/or their
available models; however, in some regions neither available data nor models are adequate
for this purpose, and the MPOs will have to catch up quickly.4 One approach would be to use
HPMS data and/or other historical traffic counts to establish baseline VMT and a trend in VMT
growth, and use the trend estimate to produce emissions forecasts. (MPOs In general may
wish to compare HPMS or other count-based estimates with estimates produced by their
travel models, and to make adjustments, as necessary, to reflect observed trends.)
if model improvements are required for other purposes (such as TCM analysis or conformity
assessment) but are not yet available, then K may be possible to "borrow" key models from
another region with adjustment as necessary to replicate local conditions. Alternately, an area
may choose to borrow an elasticity and either convert it to a parameter value or use H directly
in off-model applications.
Many areas may choose to apply sketch planning methods for TCM analyses, at least in initial
rounds. Sketch planning methods range from simple worksheets (available in various forms,
from hard copy to calculator-based to microcomputer versions), to more complex methods
such as the STEP model used in the Bay Area and the SRGP model used in Denver; a
number of these methods are documented In various EPA, DOT, and Dept. of Energy reports
(e.g., Sosslau, 1978; Harvey, 1978; Harvey, 1979; Atherton and Suhrtoier, 1978; among many
others). Caution should be exercised in using the simpler of these methods, since for some
TCMs they have been based on very limited data or experience.
Use of sketch planning methods may be a sensible strategy even for areas which have
adequate model systems. In the Bay Area, for example, the STEP model was used for much
of the TCM analysis (e.g., to explore policy options) and only the final proposed policies were
subjected to analysis using the full model system. (The full system also was used for
conformity runs.) This combined approach saved tens of thousands of dollars of computer
time.
4 VMT forecasting ^jidanc* (US Environmental Protadion Agency, 1992a) calls for MPOs to uu HPMS data as 9*
baseline and model-based growth facto* for projections, units* HPMS data art inadequate; In th« latter case* other
approach** may be aocaptad.
TnngportaHon'Alr Quality Antlytl* Ibmuea
January 1993
-------
Pag* 2-15
2.3 Specific Transportatlon-ftelated Air Quality Analysis Requirements
In addition to the basic data and modal improvements discussed in the preceding section, the
specific analysis needs result from requirements of the 1990 Clean Air Act Amendments.
These include emissions inventories, forecasts, and tracking; tracking and forecasting of VMT;
TCM evaluations; and conformity determinations. The analysis implications of each of the
major requirements Is discussed briefly in the following sections.
2.3.1' Emissions Inventories, Forecasts, and Tracking
Under the 1990 Amendments, emission* contributing to violations of the national ambient air
quality standards must be inventoried and tracked over time. Moreover, forecasts of
emissions will be used in determining the levels of reductions needed, and actual emissions
levels will be compared with forecasts in evaluating progress. Mobile source emissions will
need to be estimated at both the regional level and at the "grid cell" level, for air quality
modeling.
If aggregate regional totals were the only required emissions estimates, then a variety of
methods would be available, including:
¦ Totals expanded directly from traffic count data - An improved HPMS-type sample
(with added representation from local and collector streets) could be used to generate
a direct estimate of the regional emissions burden. Detailed volume and speed data
would form the basis for emissions calculations at each site, using MOBILE or EMFAC
emissions factors. The HPMS expansion factors then would be used to expand the
site-specific emissions calculations up to regional totals.
¦ Trip-based emissions summary~from a current travel survey - Travel diaries
generally provide a good picture of personal vehicle travel patterns. Given reliable
network-based speeds and distances, MOBILE or EMFAC factors can be used to
compute the emissions for each vehicle trip in the sample, and totat emissions can be
estimated by summing the emissions over alt trips. In some ways, this method is
likely to yield the most accurate estimates of emissions, because the MOBILE and
EMFAC factors are really trip-based (i.e., they derive from the Federal Test Procedure,
which was designed to represent a Typical" trip).
Even with a good survey of personal vehicular travel, additional effort is required to
determine the emissions from commercial and Inter-regional travel. As a practical
matter, H may be necessary to get at these off-survey elements through HPMS-type
traffic counts. (Some areas, e.g., Chicago, have done truck surveys.)
¦ Trip-based emissions summary from base travel model run • This method
resembles the sample enumeration approach described above, but draws the trip
sample from the model's trip tables. If recommended modeling practices are followed,
commercial and inter-regional trips will be present in the trip tables.
Trans portatlon-AIr Quality Anafyala laauaa
January 1993
-------
Paga 2-18
¦ Link-based emissions summary from bat* travel modal run - This method also
uses data from the regional model, but bases the calculation of emissions on final link
volumes and speeds. Emissions are determined for each link and time period, based
on the average link speed and volume by vehide type, and are then summed to yield
the regional totals. Experienoe in the Bay Area indicates that the trip-based method
and the link-based method will differ by no mora than 2 percent for CO and ROG
(VOC). The difference arises because of the greater speed variation among individual
links than among paths through the network. It is not at all dear that MOBILE and
EMFAC are as valid for the homogeneous link speeds assumed in transportation
models as they are for average path speeds*, but there is a tendency for MPOs to rely
on a link-based analysis because it more readily produces grid cell estimates of
•missions.
There really is only one reasonable method for produdng grid cell estimates of emissions,
which is to use the link flows and speeds in the manner described above.
2.3.2 VMT Baseline, Forecasts, and Tracking
Forecasts and estimates of actual VMT (along with certain other travel parameters) are
required under the 1990 Amendments as a way of monitoring compliance with transportation-
air quality requirements. As discussed In Chapter 2, ozone nonattainment areas classified
as Serious or worse must demonstrate that VMT, emissions, and congestion are consistent
with the assumptions in the 8IP, or SIP revisions will be triggered. Certain CO nonattainment
areas (dassified as Moderate or Serious with a design value over 12.7 ppm - currently nine
areas) have requirements for annual VMT forecasting and tracking; H estimated "actuar VMT
exceeds that forecasted, specific contingency measures must be implemented in these areas.
Thus, it is urgent that these areas develop the ability to produce reliable aggregate regional
"measurement" of VMT over time, with short- to medium-range forecasts of regional VMT
growth that are as accurate as possible.
EPA guidance on VMT forecasting and tracking for CO nonattainment areas with design
values above 12.7 ppm was issued in January 1992. While the guidance is not binding and
states (or regions) can depart from it, departures would have to be justified whereas simply
following the guidance would be accepted. The guidance specifies the use of ground counts
from FHWA's Highway Performance Monitoring System (HPMS) for tracking VMT, and
regional network models (or HPMS extrapolations) for forecasting VMT.
The use of HPMS for the baseline inventory and tracking and model-based VMT for
forecasting appears to reflect a compromise. On the one hand, several of the affected areas
lacked network models of suffident accuracy or detail to produce good estimates of VMT and
could not adequately improve the models in the amount of time available; therefore they were
looking for other ways to obtain VMT baseline information. The HPMS system was an
5 High «p«ads may b* m unption, tine* •mtnlont mod* for such would India* rtlrtvtty few
•ccefaratoni/cfocaleration*, «tc.
Tranaportatlon-AIr Quality Analysis Issuas
January 1993
-------
Paga 2-17
obvious option as it is already in usa in all 50 atates for VMT estimation, and is being
improved under FHWA direction. On the other hand, the HPMS "data base has recognized
limitations (for example, local streets are not counted - a limitation shared with regional
network models • and the area HPMS covers is usually smaller that the nonattalnment area,
with under-representation likely to be greatest in the urban fringe, the area where VMT growth
may be fastest.) Moreover H is especially problematic that the HPMS data for aeveral of the
states containing the CO nonattainment areas subject to the VMT forecasting and tracking
requirement - in particular, California, Washington, New York, and Connecticut - are currently
based on statewide sampling of grouped urbanized area data rather than regional samples
(though urban area samples are now required and wfll be available in the future). Also,
HPMS to not a forecasting mode!, so its use does not obviate the need for a forecasting
method. While the guidance recognizes these problems, calls for fix-ups to overcome aome
of them (e.g., more counts, methods for estimating local VMT and VMT outside the HPMS-
covered area, etc.), and allows alternative approaches under some conditions, considerable
uncertainty over the reliability of the baseline VMT estimates may be a troubling problem for
at least some of the areas.
Future VMT estimates are to be done by applying a model-based growth factor to the HPMS
1990 baseline estimates, or simply by extrapolating from HPMS samples. Thus errors in the
baseline estimate would be carried forward into the forecasts. Similarly, inaccuracies in the
expected growth rate (Itself a complex function of expected changes in employment,
population, household demographics and income, land use patterns, etc.) would produce
erroneous future estimates. Whatever the source of error, it could have serious
consequences, since overestimation of VMT would trigger more controls than actually are
needed, whereas an underestimate could complicate conformity findings, trigger a SIP
revision, and/or trigger the implementation of contingency measures*
Because of the uncertainties and risks, most areas will need to Invest in both improved counts
and improved network models. In the meantime, areas may wish to double-check their VMT
baseline estimates by comparing them to the results obtained via alternate methods. For
example, VMT estimates can be derived from any or a combination of the following:
¦ Regional gasoline sales
¦ Odometer readings
¦ A separate sample of traffic counts (different from HPMS)
¦ A household survey repeated at necessary intervals, supplemented by acommercial
VMT survey
¦ Model calculations (i.e., from regional travel models, with revision of key inputs based
on "measurements').
While each of these methods has limitations (see, e.g., Cambridge Systematica, Inc., 1891a;
6. The posslNltty of • (fscrepancy between forecasts and counts it Ikely to be quite high fltvan the uncertainty
attached to the many factor* which enter Into VMT trowth forecasts or counts themselves. However, the 6PA
guidance allows only a five percent tffference In forecasts and "actuaT, or in updated forecasts and the forecast relied
upon for the attainment demonstration, in comparison, a study by FHWA found that the variability of VMT
for 14 areas averaged *3.8% (fleet, 1M1).
Transportation-Air Quality Anafyals laauaa
January 1993
-------
P*g» 2-19
Smith and Ramadan, 1990), they may navarthalass (m useful as checks cm the
reasonableness of HPMS-basad estimates.
Alternative approaches also might be applied to forecasting VMT, including:
¦ Trend extrapolation
¦ Aggregate econometric models
¦ Regional travel model calculations.
Here, too, each approach has limitations, but can be useful as a way to check forecasts.
2.3.3 TCM Assessment
2.3.3.1 Analysis of Section 101 Measures
As noted earlier, Section 108 (f)(1) of the 1990 Clean Air Act Amendments requires EPA, In
consultation with DOT, to make information available on transportation control measures
including, but not limited to, the following sixteen Items:
1. Programs for Improved public transit
2. Restriction or construction of certain lanes or roads for use by buses or HOVs
3. Employer-based transportation management programs, including inoenthres
4. Trip reduction ordinances
5. Traffic flow improvement programs that achieve emissions reductions
6. Fringe and corridor parking facilities serving HOVs and transit
7. Programs to limit or restrict vehide use downtown or In other areas of amission
concentration, particularly during peaks
8. HOV / ridesharing service programs „
9. Time or place restrictions of road surfaces or areas to bikes and pedestrians
10. Bike storage, lanes, and other facilities, public and private
11. Programs to control extended vehicle idling
12. Programs to reduce extreme coid start emissions
13. Employer-sponsored programs to permit flexible work schedules
14. Localities' SOV trip reduction planning and development programs for special events
and major activity centers Including shopping canters
15. Pedestrian and non-motorized transport facility construction and reconstruction
16. Programs for voluntary removal of pre-1980 vehicles.
The EPA guidance documents address the relative effectiveness of various procedures and
methods, their potential effect on the transport system and the provision of transportation
services, and their energy, environmental, and economic impacts. Additional guidanoa
addresses:
¦ ways to reduce emissions during air pollution alerts;
¦ other measures to reduce public health impacts;
¦ information on the extent to which strategies to reduce one pollutant might lead to
Trana porta tlon-Air Quality Analyala laauaa
January 1993
-------
Pay# 2-19
increases hi another.
Although with specified exoeptions TCMs art not required, they nevertheless are likely to be
needed in most metropolitan areas lacing CO and ozone nonattainment problems, In order
to meet required milestones and to attain the standards by the deadlines. Hence the Section
108 Bst will be a starting point for MPO evaluations. o
Several observations are in order about the sixteen TCM categories Bsted in Section 108. For
one thing, some of the categories am vary broad, and wfthin a particular category a variety
of measures could be devised; lor example In a study on TCM implementation issues
conducted for EPA Just before the passage of the 1990 Amendments, over 70 specific
measures were reviewed (Eisinger, Deakin, at al., 1989). Each of these specific measures
could require a different analysis approach. As one example, employer-based incentives
could include reservation of dose-in parking lots for HOVs and visitors, with SOV drivers
having to walk two or three additional minutes for each trip. Or the employer could subsidize
transit and charge for parking, or establish shuttle services to transit Each of these options
would require a different set of analysis steps. Even more disparate are the analysis
approaches that would be suitable for freeway operations controls, downtown traffic signal
system coordination, and intersection redesign, all measures that fail under the improved
traffic flow category.
On the other hand, a number of the TCM categories overtap. For example, trip reduction
ordinances typically implement requirements for employer-based programs, which often (but
not always) include such items as employer-sponsored flexible work programs and rely on
local government commitments to improved transit, ridesharing assistance, etc. Similarly,
localities* programs to facilitate non-automotive travel and reduce the need for SOV travel as
part of transportation planning and development efforts are often implemented through trip
reduction ordinances or employer-based transportation management plans mandated as a
condition of development approval or building occupancy. Moreover some Hems on the list
provide alternate institutional frameworks rather than specific emissions-reducing measures;
the distinction appears in the implementation strategy specifics rather than in the way the
measures would affect travel behavior (or the differences are too subtle to be modeled: the
impact of a government-funded transit subsidy vs. an amployer-funded transit subsidy, e.g.)
In particular, TCMs #3, (employer-based transportation programs), #4 (trip reduction
ordinances), and #14 (trip reduction programs and ordinances as part of localities' planning
and development efforts) are ways to implement the more action-oriented TCMs.
These observations suggest the need to sort out TCMs into meaningful groups for analysis
purposes. First, note that the ability to model the various TCMs varies considerably. At least
six categories can be identified:
¦ Some of the measures are easily included in most currently available regional travel
models. For example, improvements in transit level of service (TCM #1) can be
readily modeled in most regional model systems.
¦ An additional set of measures could be modeled if additional data and variables were
incorporated into the model structure, though only some regions currently have models
TranaportatJon-Alr Quality Analyaia laauaa
January 1993
-------
Pag* 2-20
wtth the requisite capabilities. For example: the modeling of pricing strategies is
sometimes constrained by a lack of data on current prices and incomes, but available
models can and do accommodate these variables. Similarly, certain TCMs can be
analyzed by regional model systems if (but only if) special features are present in the
networks: an example would be HOV facilities (TCM #2) or other vehicle restrictions
on certain lanes or areas (TCMs #7 and 6). Bike and pedestrian network
improvements (TCMs f 10 and 15} could be represented in a mode choice model
which included these modes, or even could be treated via separate network coding,
although few areas deal with either mode so explicitly. In some cases "fix-ups" can
be devised to roughly incorporate the measures of question into the model, for
example, transfer of price coefficients estimated in another region or a special study,
or use of scalar variables to reflect the presence and quality of pedestrian
connections.
¦ Certain measures have been modeled in a few areas or in research projects, but
models of these measures are not in common use by MPOs. Examples include
models which incorporate variables to reflect the availability and nature of certain
features of ridesharing promotion (e.g., presence of an on-site coordinator as part of
employer-based transportation management programs - TCM #3, or of strong
rideshare matching outreach activities as part of a shared-ride services program • TCM
#8), and models which can reflect the availability of flexible work schedules (TCM
#13). For these measures spedal studies may be needed; or the analyst may simply
rely on inferences from available studies and data.
¦ Some measures are typically analyzed off-line on the basts of survey data, although
future modeling efforts incorporate their analysis measures into the regional model
system. For example, the decision to use park and ride (TCM #6) versus some other
form of access has frequently been modeled for transit (e.g., via a transit access mode
choice model}, but similar models have not been developed for ridesharing; in the
latter case most analyses estimate the impacts of parkmnd rideshare based on
empirical results at other applications.
¦ Some measures are readily modeled but not via the regional model system. These
include many traffic flow improvement measures such as traffic signal timing and
intersection redesign, as well as ramp metering and freeway weaving section
improvements (TCM #5). Special purpose models are best suited for these analyses.
MPOs either extrapolate from studies done on specific facilities, corridors, or areas of
the region, or commission special studies of the regional impacts of these TCMs.
¦ Finally, some measures are not readily modeled with transportation or traffic models
and data, but call for analyses using emissions and vehicle fleet information. The
measures addressing extended idling (#11}, cold starts (#12}, and pre-1880 vehicles
(#16) would fall in this category.
One way to group TCMs, then, is by the extent to which they can be analyzed with available
regional models, require model enhancements or fix-ups, rely on special studies or inferences
from available ones, call for analyses using transportation or traffic models other than the
Tranaportation-Alr Quality Anafyala /aauea
January 1993
-------
f*aga 2-21
regional ones, or call for non-traruiportation modeling and analysis.
The following listing suggests how various TCMs fit into each of these categories.
¦ Transit and highway Infrastructure Improvements (e.g.. TCMs 1 and 2; TCM 6
with respect to major facilities) - These infrastructure improvements usually can be
translated Mo changes in travel time, so a good regional travel model should be
capable of representing the effects on VMT, amissions, ate. Analysis of HOV lanes
generally will require the coding of specified links as HOV. {See Chapter 3.)
»
¦ Transportation aervlces and operations Improvements (e.g., TCM t, some
aspects of TCM 6) - This group of TCMs offers improvements in level of service,
reliability, flexibility, ate., mostly without adding new infrastructure. Some of the
strategies that would fall under this group Influence variables present in conventional
models: for example, increase in transit frequencies or discounts on fares. In such
cases it is possible to use a good regional model for TCM assessment. Other servioe
improvements provide new dimensions of service that do not translate well into
existing variables (e.g., improved raliebility due to guaranteed ride home programs)
or risk extending existing variables well beyond the range reflected in estimation data
sets (certain IVHS strategies might do this). In these cases H is necessary to make
"off-model" estimates and then integrate the results with the base outputs from the
regional model.
¦ Traffic flow improvements (TCM 8) - Improvements in traffic flow affecting facilities
in the networks generally can be modeled by edjusting travel times and costs to reflect
the new conditions. In many cases, analysts will use traffic operations models to
estimate the time savings (e.g., applying FREQ orTRAFLO to analyze ramp metering
or other freeway improvements; applying TRANSYT or PASSER to analyze signal
timing improvements, etc.) In some instances the analyst may prefer to work directly
with the outputs from these more detailed traffic operations models in estimating
emissions reductions, since the operations models provide detailed information on
changes in stops and delays (accelerations and decelerations).
For TCMs effecting facilities not included in the network "off-model* analyses must be
used. For example, this would be the approach for analyzing the impacts of signal
timing or intersection improvements on local streets.
a Programs Instituted at the work place (TCMs 3 and 13) - Employer-based
transportation management programs are a way of Implementing TCMs rather than
a TCM per se (although K is recognized that the very presence of an employer-based
program may support the use of travel alternatives.) Specific measures included in
an employer-based program could range from supplementary transit services to
parking pricing to transit pass subsidies to ridesharing marketing, and could extend in
some cases to employer-funded traffic signals, etc.
Measures that alter modal availability or change the times and costs of travel modes
generally can be represented in regional travel models, if the programs are ubiquitous
Transportation-Air Quality Analy*1* Issues
January 1693
-------
Paga 2-22
(or apply to all employers in particular zones, or apply to ad amployars of a particular
type of Job category represented in the model, e.g., all aervice workers). If the
programs are applied only to certain kinds of employers or employees, however, It
may be necessary to do extensive post-processing of results or to deal with the whole
measure "off-mode!" (e.g., by extrapolating from available evidence of effectiveness,
perhaps using a cross-classification approach to account for differences in geographic
area, employment size and type, etc.), and then integrate the results with base outputs
from the regional model. Such off-model approaches are also needed to aooount for
the effects of marketing and promotion, in-house coordinators, ate. (unless the MPO
happens to have one of the few models including a variable for these factors, or
chooses to implement such a model.)
« • ' •>
Flexible work schedules also must be analyzed "off-modef and then integrated with
base outputs from the regional model. Data on flextime programs indicate potential
effects on peaking (hence network characteristics experienced) and mode choloe.
(Here, too, a few regions have access to models, or at least heuristics,'for estimating
flextime effects.)
a Programs implemented or mandated by local governments (TCM 4,14, etc.) -
Like employer-based programs, local government programs can encompass a wide
variety of infrastructure Improvements, servioe Improvements, vehicle use restrictions,
pricing strategies, etc. The analysis of such approaches is according to the
substantive elements of the action rather than their implementation approach.
¦ Bicycle and pedestrian facilities and programs; auto restricted zones (TCMs 9,
10,16) • Since few regional models represent the bicycle and pedestrian modes in
their models, bike and pedestrian facilities and programs generally must be analyzed
"off-model", with the results used to adjust the regional model (e.g., vehicle trip
distribution tables would be adjusted downward, particularly for intrazonal trips and
trips between adjacent zones.) Most areas wfll refy on evidence from implementations
in their region or reported in the literature. In areas where these modes are
particularly important or are of key interest it may be useful to estimate special mode
choice models, add proxy variables to models, ate.
Pedestrian malls (auto-restricted zones) require somewhat more complex analyses,
depending on their size and the specifics of the restrictions proposed. Vehicle trip
productions and attractions could be modified, and traffic conditions on neaitoy portions
of the network could be altered. While several detailed studies involving modeling of
auto-restricted zones have been carried out, a serious proposal to implement auch a
measure on a significant scale would likely be the subject of a special study.
¦ Pricing measures - Pricing strategies win be analyzed in tome areas as possible
economic Incentive programs called for as contingencies under the 1990
Amendments. In other areas pricing may be considered as an alternative to the
command-and-control TCMs. Price, of course, also enters into mode choice and other
travel decisions in the form of fares, tolls, parking prices, vehicle operating costs, etc.
While the inclusion of price variables in regional travel models is highly variable from
TranaporUtJon-AIr Quality Analyala laauaa
January 1993
-------
Pagt 2-23
one region to another, a state of the art modal would hava price represented
ubiquitously in th« modal system, from traffic assignment through population and
employment allocation, with price responses appropriately tempered by Income, Given
such a model, the analysis of pricing measures should be about as straightforward as
the analysis of conventional infrastructure improvements. Without such a model, it will
be difficult to produce a oonvincing analysis of pricing strategies. Sinoe price and
income are absent from some models because the data are not available, alternate
approaches may have to be considered. One quick fix (recommended as a short-term
option only) is to convert prices to equivalent times for input to a conventional model,
.. using an income-variant value of time. A aeoond approach would be to transfer
model(s) from other regions with as much customization to local conditions as can be
supported by local data or studies. (See below for further discussion of pricing.)
¦ Vehicle controls (TCMs 11,12,11) - Measures aimed at controlling extended vehicle
idling, reducing the impact of extreme cold starts, or removing older vehicles (or high
emissions vehicles of any age) typically will be analyzed by referenda to motor vehicle
registry data (e.g., data on how many vehicles are pre-1980), special purpose studies
(e.g., patterns of use of vehicles of different ages, studies of idling in taxi lines), and/or
data on the effectiveness of technology applications (studies of emissions reductions
from pre-heated catalytic converters, Investigations of irt-use monitoring and
enforcement program effectiveness.) MFOs may leave these TCMs to air quality
agencies, or may work closely with them in conducting special studies, but relatively
few MPOs are likely to take the lead on their development and analysis. MPO data
and models may be of interest in studies of these TCMs if, for example, the number
of trips made by different categories of households is at issue, or If travel survey data
report which vehicle was used for each trip.
As this list indicates, a number of TCMs, particularly those that can be represented as ^
changes in vehicular travel times or travel costs, can be analyzed by typical regional travel 1
models; several more, especially detailed pricing options, could be analyzed as long as the
region has Invested in the requisite data collection and model development Several of the
more commonly used TCMs are best analyzed via traffic operations models rather than travel
demand models or network models, and a number call for special purpose studies or reliance
on evidence from previous implementation experiences rather than regional modeling. In
several instances "quick fixes" or simple indicators could be used as an interim measure,
while more detailed approaches are implemented. Overall, the choice of how to model a
particular TCM will depend on 1) the nature of the TCM itself, 2) the quality of available
models and data and their suitability for analyzing the TCM, 3) the importance of the TCM in
the region, and 4) the time and other resources available for model development or
refinement For further guidance, see the references to TCM guidance documents.
2.3.3.2 Other TCM Analysis Issues
Several other issues may need to be considered by TCM analysts. Of particular importance
are pricing strategies; TCMs* impacts on land use and development, and conversely, the
potential of land use and development measures as TCMs per se; and various TCMs*
Transportstion-Alr Quality Analyala laauaa
January 1993
-------
Pag* 2-24
implementabllity and its impact on the amissions reductions cradits that can ba da'tmed. Each
of these issues are discussed briefly in this section.
- Pricing
Economists have long argued that many functional problems in the transportation system
stem from inaccurate price signals. Diverse strategies for improving the pricing of
transportation have been proposed. Among them are toils; use of congestion pricing
techniques to allocate space on crowded urban highway facilities; market-based pricing of
parking and elimination of favorable tax treatment for aubsidized parking; and amissions fees
for vehicles operated in areas which have not attained air quality standards. Such strategies
increasingly find endorsement from other experts, and interest in transportation pricing has
been growing in recent years both as ways to generate revenues and as means of moderating
travel demand and reducing amissions.
In several parts of the country, new roads are being built as tollways in order to pay for the
needed facilities and deliver them faster than would be possible relying on public funds alone.
In some areas, tolls and other fees are increasingly seen as ways of generating revenues for
a variety of transportation projects including transit In addition, pricing has been proposed
as a way of dealing with congestion, air pollution, energy dependence, and other problems
associated with heavy use of the automobile.
But a presumption of political infeaslbflity has tended to make pricing strategies the exception
rather than the rule. Arguments challenging the practicality and fairness of transportation
pricing schemes include concerns that toll booths create delay and increase accidents, that
pricing might simply divert traffic to other "free" routes and cause problems there, that setting
congestion prices accurately would be near-impossible, that congestion prices would be so
high that only the rich could drive in many areas, and that emissions and congestion pricing
would be prone to fraud. While advances in technology could overcome some of the
objections raised (especially those concerning delays and cheating), other concerns would
remain.
Nevertheless, a number of metropolitan areas will consider pricing strategies as part of their
TCM analyses. For example, Los Angeles and the Bay Area have found that fees, tolls, and
the like would be necessary to meet state-mandated air quality standards. As proposed in
the Bay Area, the pricing approach would use congestion charges, smog charges, parking
fees, and gasoline taxes. These fall into two conceptual categories: charges that are firmly
rooted in the economics of transportation (i.e., "market-based") and fees that exploit a
convenient institutional framework for revenue collection (i.e., "fee-based"). With the
exception of toll increases, pricing proposals have not been adopted yet, but they have
received serious public airing and garnered substantial media support Similar proposals are
under consideration for Los Angeles.
Pricing Is also fikely to be considered because of the demonstration projects authorized by
ISTEA. At the time of this writing a number of MPOs are actively developing proposals for
the demonstration funds.
Tranaportatlon-AIr Quality Analysta laauaa
January 1993
-------
Paga 2-25
If market-based measures do receive consideration in Federal TCM planning, a number of
supportive actions may be necessary. In particular, suitable analysis tools will be needed;
currently a surprising number of MPOs lack data on such basic factors in travel choices as
household incomes and travel costs, and hence cannot adequately model any pricing policies.
A full-fledged pricing program would have tar-reaching effects on the pattern of mobility in a
region, by altering the perception of accessibility and the cost of auto ownership in a way that
depends on household income. The major questions for analysis are whether available travel
models can represent the role of price in accessibility in a systematic and comprehensive
way, and whether such models adequately capture socio-economic variation in the population.
The following is a brief review of critical modeling issues which arise in a pricing study.
One key issue is, What variables are critical? Clearly, the price of travel is central to the
analysis and must appear in appropriate places throughout a model system. Components of
price for a vehicle trip should indude parking costs, tolls, and perceived out-of-pocket auto
operating costs, the latter specified carefully to assure comparability with' less subjective
parking and toll costs. In addition, the annual cost of operation (both fixed and variable)
should appear as a determinant of auto ownership.
Household or personal Income is a second key variable that should mediate the effect of
price wherever it appears in a model system. Under best practice, household income typically
is included to make the coefficient of price inversely proportional to income.7 Unless income
appears jointly with price, the distributional consequences of pricing strategies cannot be
studied.
Congested travel time is a third key variable. Models should be capable of representing the
effects of predicted changes in demand on travel time, wfth as much time-of-day detail as
possible.
Behavioral responses to pricing strategies also need to be considered in building price
sensitivity into the model system. Under the conventional paradigm of travel behavior, price
could have noteworthy effects at several levels of the model hierarchy:
¦ route choice/traffic assignment - Tolls and congestion fees influenoe the
"impedance" of each route, which will produce changes in path assignments as fees
are differentially changed.
¦ time of travel - Fees that vary with congestion will induce some drivers with
scheduling flexibility to shift to less congested periods.
¦ mode choice • Price is a key determinant of modal competition for all types of travel.
7. This is don* by constructing • comport* variabl* that is som* function of poo* fend income.
Transportation-Air Quality Analysis Issuas
January 1993
-------
Paga 2-26
m trip distribution - Differential price increases will cause a spatial shift in the trip
distribution from any given location, and a general prioe increase will lead to shorter
trips overall.*
#
¦ trip generation - For non-work trips, a general price increase could reduce the
amount of discretionary trip-making. For work trips, a significant price increase (either
differential or general) could foster work-at-home policies, four-day work weeks, or
other reduced trip scenarios.
¦ auto ownership - By directiy or indirectly falsing the cost of auto ownership or
decreasing highway accessibility, price increases could reduce the incentive for
multiple auto ownership.
¦ residential and employment location - Significant price increases may cause lower
income working households to seek less expensive work places or residential
locations. Conversely, reductions in congestion may induce higher income households
to locate farther from their work plaoes.'
¦ residential and commercial construction - Pricing-induced changes in residential
demand or work force availability might shift the locus of regional growth, or perhaps
alter the overall rate of regional demographic and economic change.
It seems likely that some of the postulated phenomena are more important than others, but,
unfortunately, the literature does not provide much help in sorting out the first-order effects.
Depending on the specifics of the pricing policy and the time period in question, impacts could
be focused primarily on time of day, route choice, and mode choice, or could extend to the
shape and size of the region. Moreover the same hierarchy of effects could be posited for
other large or long-term changes in the transportation system, such as the cumulative effect
of gradually increasing congestion over a long period.
Finally, structural features of the models need to reflect pricing impacts in accordance with
theory. Key structural attributes of models involve linkages among elements of the behavioral
hierarchy, and degrees of disaggregation among places, people, times, and facilities:
¦ model linkages and feedback - The models have implicit linkages which should be
reflected in the results of a pricing analysis. Perhaps the most obvious is the
presence of time and price at many levels of the model hierarchy. If time and price
influence demand at many levels, and demand determines time and price (through
B. Th» behavioral process is quit* Afferent for work and non-work trip*, in tha non-work case, people have 0m
option to shift locations of dscretionary activities. In tfie work cast, peopla have fixed origins (residences) and
destination* (pieces of employment) In tha short run, but can change either or both in the long run. In effect, a work
trip distribution model is an attempt to represent long-run residential and employment location behavior, given
estimated numbers of dwaDings and work places.
9. Note that soma elements of this spatial response would be covered In a conventional modeling framework by work
trip distnbution. Caution is requirad in order to avoid double-counting these effects.
Tranaportatlon-AIr Qutlfty Analysis Issuaa
January 1993
-------
Paga 2-27
assigned link volumes and estimated levels of congestion), then It becomes necessary
to perform a recursive analysis that checks for consistency among the input and
output assumptions. In rigorous terms, models should be run to "equilibrium", but
since at present this often is computationally impractical, there must be at least a
systematic effort to achieve consistency between predicted times and costs and those
used in calculating the various elements of demand.
¦ disaggregation - In all transportation modeling there is a tradeoff between the detail
required for accurate representation of supply and demand and the resource
requirements of increasingly disaggregate analysis. For pricing studies, H is essential
to have a detailed representation of the income and vehicle fleet distributions, and
helpful to have as much specificity as possible about highway links (for micro-analysis
of congestion pricing), analysis zones (e.g., for exact representation of parking prices
and other land use-related measures), and times of day (for careful representation of
time shift effects). In a given analysis setting, the trick is to find the highest level of
aggregation that can support an evaluation of the "first-order" affects!
In short, full analysis of pricing strategies requires good data and sophisticated models.
- TCMs and Land Use and Development Patterns
The transportation measures listed in the 1B90 Amendments as potential TCMs range from
major capital investments such as HOV lanes and transit improvements, to operations-
oriented approaches such as traffic flow improvements and vehicle restricted areas, to minor
capital projects such as pedestrian and bike facilities, to vehicle technology options. However,
the Amendments note that TCMs are not limited to the Section 176 list Consequently at least
some metropolitan areas will consider additional measures, among them measures which use
urban design and land use planning as opportunities to reduce vehicle dependence.
Land use approaches have been part of the dialogue about emissions control since the 1970
Clean Air Act Amendments. They often are dismissed as impractical because of the
fragmented institutional setting of most land use diecisions in the U.S., and because of the
long implementation horizon. Yet recent debates about air quality and other aspects of the
urban environment have made much of the linkage between low density land uses and high
rates of per capita travel. Data from large cities worldwide show a consistent, strongly
negative correlation between residential density and measures of metropolitan average per
capita vehicular travel. (See, e.g., Kenworthy and Newman, 1990.) Environmental groups,
in particular, infer from the data that infrastructure investments will worsen per capita
emissions when they support development at the urban fringe (where the lowest density,
highest travel consumption districts are found) and will improve per capita emissions when
they create arrangements of land uses that require less vehicular travel (either by placing
compatible uses in close proximity or by linking activity centers and residential areas through
mass transit). While the studies on which these inferences are based have flaws and in any
event may be reporting correlations rather than causal relationships, the heightened interest
in the topic may put land use strategies on the TCM agenda in some metropolitan areas.
TranaportatJon-AIr Quality Analytls laauaa
January 1993
-------
Pag* 2-2$
Statutory authority differs in aach stata and metropolitan area, and available land use control
mechanisms also vary. In Minnesota and California, state law allows air districts to astablish
indirect source review (ISR) programs for oversight of land use and facility location decisions.
ISR in these states presumably could be used to elicit design features beneficial to air quality,
such as mixed uses at employment canters, high-quality pedestrian treatments, bicycle
facilities, and direct Inks to transit Ines. Alternatively, locally-originating policies and
programs could have the same effect. Such cities as San Diego, Portland, OR, Seattle, and
Boston have many of these policies already In place. In a few cases state planning acts or
regional planning laws may provide yet another'way for land use and transportation to be
more closely coordinated, though to date few areas have taken strong land use policy stances
in response to air quality concerns.
V
Whatever the implementation mechanism, the overall Intent Is simitar. Urban design and land
use approaches to transportation control are Intended to moderate travel demand and
influence mode choice by creating desirable urban development patterns In which lifestyles
less dependent on the automobile can flourish. Specific strategies at the small scale Include
transit-oriented development; doser attention to street layouts and street widths and provision
of sidewalks and bike facilities, to create improved environments for pedestrians and cydists
and greater ease of operation for transit and para transit; restraint in the amounts of parting
provided and location of parking to minimize conflicts with pedestrian flows; and site planning
for a balance of housing, jobs, and services, to reduce trip lengths and permit trip Hnking. At
the larger, community-wide or regional scale, the strategies include:
a Urban limit lines and urban development reserves.
¦ Mandatory consistency between local land use plans and local and regional
transportation plans.
¦ Requirements for the provision of adequate public facilities concurrent with
development, and/or attainment of minimum level-of-service standards.
¦ Minimum as well as maximum development densities and floor-area ratios to ensure
adequate development for transit to work.
¦ Incentives and bonuses for desired land uses and for developments that provide
desired transportation and land use amenities.
¦ Mandatory city, county, and regional batandng of Job growth with housing develop-
ment, priced and located to match the needs and incomes of the work force.
Advocates of these techniques believe they would produce both transportation and land use
benefits, by encouraging the efficient use of land, redudng infrastructure costs, and creating
lupportive environments for the operation and use of transit, and reducing trip lengths so that
Milking and cyding are feasible. Advantages to sotiety overall are thought to Indude
decreased requirements for travel, tower energy consumption, less air pollution; urban sprawl
would be reduced, sparing valued agricultural lands and other open space. Costs also would
iccrue. In some cases, development would likely spill beyond the urban boundaries into
Transportation-Air Quality Analyst* laauaa
January 1993
-------
Paga 2-29
unregulated, rum! town*, or perhaps would shift to other metropolitan regions. Empirical
evidence is accumulating, but remains too scattered and partial for affective use by policy-
makers; here is dearly an araa for future research and evaluation.
Currently, few metropolitan area model systems are wen equipped to Investigate either the
regional planning-Jeve! or the locaMevel transportation and land use policies. As rioted in
Section 4.3, only a few MPOs currently have formal land use allocation models, much less
models of the regional economy which would allow them to explore the impacts on
employment, population growth, land prioes, ate. as a function of public policy intervention*.
This, too, has been identified as an ansa where research is needed.
Modeling capabilities for urban design options are mixed. Occasionally, urban design changes
can be suitably represented in available models, e.g., strategies which would increase housing
in the vicinity of transit stations could be modeled lowering access times to transit However,
since few urban areas model walk or bike trips, many other urban design-level transportation
improvements (e.g., high quality walk access to shopping) would simply not be modeled, or
would require off-model "adjustments" to vehicle trip rates. Similarly, land use policies such
as purposeful mixing of retail and office uses typically would not be analyzed with regional
models because suitable land use variables are absent
Even when models including land use variables are available, they rarely are fine grained
enough to do a credible Job of evaluating such options as the affects of mixed use on trip
making or the impact on auto use of good pedestrian linkages to shopping areas. Some
areas do have models which can address these issues in approximate, heuristic ways (e.g.,
Montgomery Co., MD has a model which incorporates pedestrian and bike friendliness as a
scalar variable). But the performance and replicabiiity of such models have not been fully
tested. Other areas have done special studies of the issues (e.g., the recent Portland, OR
study of alternatives to the Western By-Pass, coordinated by 1000 Friends of Oregon, has
included modeling of a variety of land use options), and the Federal Highway Administration 's
is sponsoring a research project on urban design, demand management, and travel which
should produce additional case examples over the next few years. However, the most
common approach for analyzing land use planning and urban design strategies has been to
extrapolate from existing examples. For example, a study for Central New Jersey's MSM
Regional Council (Howard/Stein-Hudson Associates, at al., May 1991) applied findings from
Hooper's study of trip generation in suburban activity centers (Hooper, K.G., at al., 1989) to
adjust downward the estimated vehicle trip rates in mixed use developments, then estimated
impacts. Such approaches amount to welHnformed scenario testing, i.e., a particular level
of travel shift or reduction due to land use planning is inferred from experiences in comparable
situations, and the consequences of such shift or reduction is then analyzed.
Overall, then, capabilities of modeling of either regional or localized Sand use - transportation
policies are quite limited, and more research is needed.
The impact of land use strategies on travel demand is only one of several concerns about
land use - transportation interactions and their environmental consequences. For one thing,
not all transportation control measures are thought to support compact development and
pedestrian orientation. In particular, traffic flow improvements (including those created by
Transportatlon-AIr Quality Analyala taauaa
January 19 93
-------
Paga 2-30
HOV lanes which sort mufti-occupant vehicles from SOVs during peak periods, and act as
additional mixed flow lanes at other times) have come into question in tome areas.10 More
generally, large scale transportation investments are seen by many as instruments for the
shaping of metropolitan structure; for example, rail transit proponents express high hopes that
major capital investments in a new round of rail projects could redirect urban growth patterns
toward more compact, centered development, while highway proponents hope that major
roads will stimulate economic development The issue is whether transportation investments,
large or small, alleviate pollution problems by reducing congestion and smoothing out flows
(or, in the case of transit, shifting modes), or whether they ultimately lead to higher amissions
by stimulating development patterns and changes in travel behavior that would offset any
gains.
Theory says that transportation improvements (whether transit or highways) will tend,
simultaneously, to increase employment at benefitted sites and to decentralize workers'
housing. Conversely, worsening transportation services wilt favor decentralization of jobs but
support higher densities of housing. Empirical studies find that transportation availability and
quality are factors in location and development, but investments will do relatively little absent
other critical factors including appropriate land, labor, and capital. Environmentalists
sometimes argue that it is the shift in development potential that is of immediate concern, in
particular if development is induced to relocate from high-density areas where many trips
would be made by foot or transit to low-density areas heavily dependent on the auto.
Scenario testing exercises and a few modelling efforts using real data have explored this
issue sufficiently to report that such an effect could occur. But the magnitude of the effect
remains unclear, and controversy continues over when and to what degree a highway
improvement (or a rail transit line) will induce trips, shifl modes, and alter destination choices.
This is an area that is high priority for research.
While TCMs' impacts on land use and land use measures' potential as TCMs are likely to be
at issue in a number of nonattainment areas, -relatively few areas will be able address the
topic through detailed model-based analyses. A state-of-the-art model would have variables
accounting for the effects of density and mix of use on travel behavior, but probably would not
be able to address broader growth-related issues. In the absence of a state-of-the-art model,
off-model estimates of land use effects will have to be developed and then integrated into the
base model forecasts. In the longer run, research on this topic is much needed.
- TCM Implementation Analysis
Full analysis of a TCM cannot stop with the modeling of Its expected transportation and
emissions impacts. Implementation of a TCM requires:
¦ a full description of the measure and how H is intended to work
10. The 1900 Amendments themsetves contain the stipulation fiat Vw Section 100 guidance documents era to
address those traffic flow Improvement programs that achieve amissions reductions", with the dear Implication that
not all such programs would reduce emission*.
TrantporUVon-AIr Quality Analysis Issuas
January 1993
-------
Paga 2-31
a a schedule for oompfetinfl all steps of the analysis, for obtaining authority to proceed,
for securing funding, and for implementing the measure
¦ Identification or responsibility for planning, analyzing, programming, funding, and
implementing the measure
a analysis of the TCM and quantification of Ks amissions Impacts as well as
assessment (quantitative or qualitative) of its other social, •conomic, arid
environmental costs and benefits
¦ assurances of legal authortty to carry out assigned responsibilities on the part of
each identified responsible party involved in implementation and/or identification of
steps to be taken to obtain such authority and a schedule for completing such steps
¦ enforceable commitments of needed fiscal, personnel, and other .resources on the
part each identified responsible party involved in implementation and/or identification
of steps to be taken to obtain such commitments and a schedule for completing such
steps
a allocations of funds necessary to carry out each step of the planning and analysis
and/or a schedule for obtaining such commitments hi eases where funding is ¦
contingent on the results of an analysis, or on implementation of related TCMs or other .
projects
a a program for monitoring and feedback to assure the TCM is operating as intended *
or to identify problems and develop changes, as needed.
In many cases the analysis of legal, institutional, and related matters affecting implementation -<
will lead to adjustments of the more technical analyses. For example, for certain TCMs legal
authority to proceed may be lacking or uncertain, some parties may be unable or unwilling to
proceed, and/or funding may not be readily available. Pricing and parking management
strategies, vehicle fleet strategies, controls on idling and extreme cold starts, and employer-
based trip reduction programs are among those for which the ability to implement is not
always immediately available or complete. In such cases it is critical to develop a schedule
for obtaining the authorizations and commitments needed for the TCM to proceed, in order
to be able to take credit for estimated emissions reductions. Alternately, emissions reductions
estimates may have to be scaled back to reflect partial implementation or delayed
implementation. In addition, political aspects of implementation need to be recognized:
Certain measures may be legally feasible but sufficiently unpopular that rigorous
implementation Is unlikely, and this may need to be taken into account by adjusting estimates
of emissions reductions downward.
Tranaportatlon-AIr Quality Analyst* laaues
January 1993
-------
Paga 2-92
2.3.4 Conformity Assessment
Conformity provisions of the 1990 CM Amendments are considerably mom detailed and
exacting than the provisions of the predecessor 1977 Amendments. Specific directives for
making findings of conformity are now provided for plans, programs, and projects, and both
federal agencies and MPOs are explicitly given responsibilities. Conformity Is to be viewed
In reference to the overall purpose of eliminating or reducing the severity and number of
violations of the national ambient air quality standards (NAAQS) and attaining the standards
as expeditiously as possible. Activities must not 1) cause new violations, 2) increase the
severity or the number of violations, or 3) delay of attainment of standards or interim
milestones. Furthermore, the most recent estimates of population, employment, travel and
congestion levels must be used in assessing whether delays may occur.
Plans and programs can be found to be in conformance only If the emissions expected to
result from their implementation are consistent with the estimates of emissions and reductions
contained in the revised SIP, and the three findings listed above can be made.' Note that the
interim milestones provision brings the Reasonable Further Progress required emissions
reductions into the conformity determination. Programs furthermore must provide for timely
implementation of TCMs contained in the SIP, consistent with SIP schedules. Projects must
either come from a conforming plan and program, and have been described in terms of design
concept and scope in sufficient detail at the time of the conformity analysis for emissions to
have been determined; or must be subjected to an analysis which shows that emissions
together with those in conforming plans and programs will not cause an exceedance of
emission reduction projections and schedules. Until such time as an acceptable SIP is
available, conformity requires plans and programs to be consistent with most recent estimates
of mobile source emissions, provide for expeditious TCM implementation, and contribute to
annual emissions reductions; projects must come from a conforming plan and TIP and, in CO
nonattainmenl areas, must eliminate or reduce the severity and number of CO violations in
the area substantially affected by the project
Conformity analyses are heavily model-dependent, as evidenced by experiences to date with
conformity determinations in several metropolitan areas.11 Although data and models varied,
there was notable consistency in the results of the conformity analyses. Key findings are as
follows:
¦ Anticipated transportation investments were not found to alter either travel patterns or
emissions in a major way. Regional-level build/no build emissions differences were
modest - generally on the order of one percent or less.
11 Interim guidelines en conformity war* issued by EPA and DOT in June 1991. Fnal guidelines ware etjn under
review nit the time of this writing (October 1982). Metropolitan areas therefore are carryinfl out conformity via»y««s
under the interim guidelines. In adtftion, since the SIP revisions required under the 1690 CAA Amendments have
no! yet been submitted end approved, metropolitan areas are operating under their previously adopted SIP*.
Tranaportatlon-Alr Quality Analyst* laauaa
January 1993
-------
#»«0« 2-33
¦ The small differences ere due, in pert, to the modest overall changes being proposed .
in most TIPs, and in part to the offsetting effect of travel shifts induced by the
transportation investments.
ai Growth-inducing effects of transportation Investments, and their consequences for
travel and emissions, have been an issue in several urban areas, in general, where
models have been run with and without equilibration through trip distribution (to
capture effects of alternative levels of accessibility provided by transportation .
investments), the feedback effects were fOMnd to alter emissions estimates significantly
within the narrow range of improvement discussed above. In a prototypical case,
emissions benefits of the "build" scenario might be estimated at about 1 percent before
equilibration but at only 1/2 percent after taking the travel shifts into aocounL
¦ A basic change in the emissions vs. speed relation or in the method of calculating
emissions (e.g., to emphasize accelerations} might reverse the result in favor of the
no-build alternative in some areas.
a Disaggregation of regional emissions analyses of TIPs (which is done to prepare input
to air quality models which calculate ambient concentrations of various pollutants) can
reveal substantial variation in impacts at the subarea, corridor, or project level, with
some areas or corridors showing emissions reductions and others showing increases.
Since pollutant concentrations are affected not only by emissions but also by
temperature, wind speed and direction, topography, and other factors, these subarea
emissions changes do not necessarily indicate the direction of change in
concentrations. Additional analyses beyond the basic aggregate emissions
comparisons may be called for, however.
Thus, while the MPOs generally have been able to carry out acceptable conformity and TCM
assessments, they recognize that a number of issues are likely to arise as those outside the
traditional transportation planning community begin to look more closely at the inner workings
of transportation models and at their results.
The costs of conformity analyses have varied widely. Some MPOs have needed to develop
improved methods just to be able to produce an acceptable conformity analysis. They have
reported costs of as much as $250,000 and time requirements of four person-months to
complete the conformity analyses in the first year. In contrast, MPOs with ready-to-use
models have reported costs on the order of $30,000 and one person-month or less. Because
the higher estimates include costs of model updates and improvements that wilt be of more
general application, H probably is not fair to attribute all of these costs to conformity
requirements.12
Conformity assessments nevertheless will require a major effort for most MPOs. In the near-
term this involves bulld/no-build (action/no action) comparisons of the RTP and the TIP; over
12 Loa Angles reported conformity analysis tint* and dollar coats of tavara! timas tha "high* amounts reported hart,
but thair totals induded parboil arty Iar0« amounts for natworfc preparation and for rapaatad calibration runs, nattftar
of which ara onfnarfiy naadad to support conformity fencing*.
Tnn*portation-Alr Quality Analysis /sstie*
January 1993
-------
Paga 2-94
the longer run the conformity a*«#*»rT>errt will focu» primarily on making aure that the TIP
projects do not cium mobile source amissions to exceed levels assumed in the SIP. Project
conformity raises an additional set of issues, •specially in the interim period; the most
complex requirement will be to show that the project, when taken as a whole, will reduce or
eliminate the number and severity of violations of cartoon monoxide standards in the area
substantially affected by the project (The "area substantially affected by the project" includes
both the vicinity of the project in which receptors are located which could be affected by the
carbon monoxide amissions coming from vehicles using the completed project, and other
affected streets and arterials on which traffic could be expected to change significantly as a
result of the project} ...
In this section, analysis Issues raised by RTP and TIP conformity assessments are briefly
reviewed. The particular issues raised by project-level CO assessments are discussed in
Section 4.4.
2.3.4.1 RTP Conformity
The basic goat of RTP conformity analyses is to determine whether a region's adopted long*
range transportation plan is consistent with attainment and maintenance of national ambient
air quality standards. In atl eases the RTP conformity determination must be based on the
latest planning assumptions and emissions models, and must show timely implementation of
TCMs from the applicable SIP. In addition, the analysis must include all regionally significant
transportation facilities and operations expected to be in place by the target years of the
analysis (interim milestone years and attainment and horizon ysar(s}.)
In the interim period before a SIP revision is available, the conformity determination must be
based on a build-no build comparison (since a contribution to annual emissions reductions Is
required). The build-no build comparison generally will require that projects be specified in
more detail than has been the practice at the planning level in many areas, in order to be
analyzed in a meaningful way. For example, the location of new facilities will have to be
identified in sufficient detail for an accurate analysis of the impact on route choioe to be
carried out; the number of hours of operation and vehicle occupancy requirements for both
new and existing HOV lanes will need to be specified in order to determine their affects on
mode choice and on travel times on both HOV and other lanes. TCMs can be credited in the
plan analysis, but only If implementation is assured. In addition, plans will have to be more
grounded in fiscal reality than has been the practice (a requirement that ISTEA has Imposed,
In any case); projects which are not formally adopted or which lack funding sources which are
reasonably expected to be available cannot be considered as part of the plan.
The no-build scenario Is fashioned in a similar way, specifying the facilities and operations
that will be in place whether or not the plan proceeds. Once the two scenarios are adequately
specified, models can be run with each of them for the various analysis years and the
Tnn»porUtIon-Alr Quality Anatyala laauaa
January 1993
-------
Paga 2-35
•missions estimates can be compared.11 Interpolation generally will be adequate for interim
year analyses. If problems are identified plan revisions may be needed, e.g., delay of certain
projects, acceleration of others, addition of TCMs, etc.
Once a SIP revision Is approved, the test is whether the plan Is consistent with timely
attainment of the standards as well as with estimates of emissions and reductions, I.e., the
emissions budget, assumed in the SIP. Analysis is done for the plan (i.e.. the build scenario)
for the various analysis years and amissions estimates are compared to the emission budget
levels. Again, identified problems may necessitate revisions. Subsequent changes in the
transportation plan (or in underlying assumptions) may occasion an updated plan analysis to
determine conformity with SIP assumptions; aimilarty, a SIP revision which alters the
transportation emissions budget will generally trigger a redetermination of conformity.
A number of specific assumptions will need to be made in applying regional travel modeling
in either period. One area where there may be particular sensitivity concerns the assumptions
made about population and employment growth rates, and land use patterns. As noted
elsewhere in this report, most areas have treated these as exogenous inputs in their travel
models. Increasingly, however, pressures are mounting for areas to explicitly model the
impact of alternative transportation investments on urban growth and form. Nowhere is this
pressure likely to be greater than for the long-range plan analysis, where the possibility of land
use change is greatest. Large urban areas, fast-growing areas, and areas with particularly
troublesome air pollution problems are most likely to face demands for formal land use
modeling (at least reflecting impacts of transportation investments on trip distribution, and in
many cases going as far as a fully integrated land use -transportation model system).
However, other areas where land use policies are explicitly used to shape urban development
and/or land use policies are politically controversial may need to develop land use modeling
capabilities as well.
Developing capabilities to consider land use impacts will take time, whether the capabilities
needed are merely feedback of travel time changes to the trip distribution step or more
extensive feedbacks to trip generation, auto ownership, and location choices. Areas lacking
these capabilities may find it appropriate to describe, as part of the build scenario, how land
uses and activities are expected to change as a result of the transportation investments
contemplated, and to estimate the anticipated impact of such changes.
2.3.4.2 TIP Conformity
TIPs, like RTPs, must be analyzed using the most recent planning assumptions and
emissions models, and must show timely TCM implementation. TIPs also are subject to a
build-no build comparison in the interim period and a comparison to emission budgets after
SIP revision. During the build-no build comparison phase, new ISTEA provisions calling for
greater fiscal realism in plans and programs will also make a difference; past practices of
13 Recognizing the wide variation in the level of detail currently provided by regional transportation plant, the interim
conformity guidelines permit a qualitative analysis at the plan level. However many MP Of will use modeling.
Tranaportatlon-AIr Quality Analyala laauaa
January 1993
-------
Paga 2-36
showing extensive capital investments for which funding is unidentified will no longer be an
option, for example. In at least some areas, this may well make future plans look less bright
than they might have in the past.
As noted earlier, several MPOs that have conducted a thorough build-no build HP analysis
have found the difference in emissions (and other performance indicators) to be surprisingly
small - on the order of one percent. The statistical significance of such small differences is
often difficult to ascertain, especially In a complex multivariate, recursive model system. This
m turn could create problems for an MPO able to show only a small improvement in the build
scenario. For this reason, MPOs will want to move quickly to address the basic modeling
issues raised in forums such as the MTC lawsuit14, and/or will need to move past the interim
phase as quickly as possible.
In carrying out the TIP conformity analyses after a SIP revision is in place, emissions from all
projects and activities in the TIP taken as a whole must be estimated for the analysis years
(milestones, attainment) and compared to emissions budgets. Conformity requires emissions
to be no more than the budgeted amounts.
It should be noted that for areas classified as Moderate or higher, 1893 SIP revisions must
include provisions for a 15 percent reduction in VOCs from 1990 levels, after accounting for
growth; areas classified as Serious or higher must further show annual VOC reductions of at
least 3 percent a year until the attainment date." Hence, conformity with the SIP will require
that such reductions are demonstrated. While both plans and programs must be consistent
with the "necessary emissions reductions contained in the [SIP]*, CAA Section 176(c)(2)(A),
the TIP analysis is likely to be much more important to the required demonstration.
2.3.4.3 Project-Level Conformity
Projects which come from a conforming plan and program are subject to review only for CO
(and PM10 in areas where that is an issue), since ozone is a regional problem and project-
level ozone analysis is not meaningful.
14 Cttixartt for a Bettor Environment at al. v. Peter B.WUson at al., CM! No. 89-2044-TEH, and Siarra Club v.
Metropolitan Transportation Commission, at al., CM! No. C-69-20M-TEH (consolidated.) MTC was ordered to devise
a rigorous quantitative approach to assess the Impacts of proposed highway Investments. The plaintiffs' experts
harshly criticized the conventional fbur-atap modeling approiech as inadequate to the task and not in keeping with
accepted theory. MTC's response was to apply a modeling approach wtth detailed feedback affects explicitly
represented in the model system, and this approach was accepted by the court. See Harvey and Deakin (1982) for
a discussion of the case and Its implications for analysis.
15 A reduction of lets than 15 percent by 1696 can be approved by EPA only if fte plan includes stringent new
source review and RACT rules as required to1 Extreme areas, for "major sources" emitting 5 tons or more per year
of VOC, and further includes all technologically feasible measures for each source category. See CAA Section
l82fbX1KA). areas classified m Serious or higher aimBar provisions apply to plans which show VOC reductions
of less than 3 percent a year after 1996. Section 182{cX2XB).
TrantportatJon-AIr Quality Anafyala taauaa
January 1993
-------
Paga 2-37
Occasionally an MPO may wish to proceed with a project which does not appear in the
conforming plan and/or program. In such cases an analysts must be done showing that
emissions from the project would be consistent with the emissions budgets in the applicable
SIP, after accounting for the TIP and plan projects. In general, this will require a model run
or runs with the project added, as well as Interpolation as needed for additional analysis years
(e.g., milestones, attainment year(s)), and hence is unlikely to be undertaken lightly by most
MPOs.
CO analyses also must be done for project-level conformity. Because CO analyses are
complex and may involve the MPO in carrying out or reviewing analyses of a type it has not
traditionally dealt with, the issues involved are addressed in some detail the following section.
TranaportatJon-AIr Quality Analysis laauaa
January 1993
-------
Paga 2-38
2.4 Projeet-Leve! CO Analysis
2.4.1 Basic Issuss
Project-level CO analyses present several complicated Issues for MPOs. Because of the
localized nature of many CO violations, project-level analyses are sensitive to specific design
and operations characteristics of the project, including not only size and location of the facility
per se but also such details about the facility as vertical and horizontal alignments, ramp or
intersection location, intersection controls Including signal parameters, and operating speeds.
In addition, CO concentrations are sensitive to meteorological conditions such as wind speed
and direction, mixing height, and temperature. As a result, project-level CO analyses calf for
the application of special-purpose models which account for these specifics.
A basic question that often arises is how to define a project so that the CO analysis is
meaningful and not misleading. In general, the project to be analyzed should be taken as a
whole even if it is phased or divided up for funding purposes. In addition, projects which
divert traffic either to or away from particular intersections or facilities appropriately call for an
analysis of all significantly affected routes/facilities as part of the "project" analysis. In this
way any potential for shifting a CO problem from one spot to another should be identified.
Most analysis measures produce outputs for each segment or link of the project, facilitating
location-specific assessment. Note that a change in operations or design could change the
results of the CO analysis, given the CO models' sensitivity to these factors."
In some instances it may be appropriate to group "projects" together in conducting localized
CO analyses. Certain projects may best be analyzed as part of a corridor or area study. For
example, a coordinated set of signal installations or other operations improvements along a
major arterial would fall into this category. Even if the various improvements would make
sense on their own accord and the proposal is to fund them separately, a corridor or area
analysis may be more efficient to carry out and more meaningful, since it is hard to isolate
project effects when a series of projects are interdependent and have cumulative impact.
Projects with intersecting or overlapping project areas or which are dependent on one another,
e.g., timed transfer centers and transit priority treatments, also might best be analyzed
together. Similarly, projects which would not be useful or needed except for other projects,
e.g., park and ride lots, light rail crossing barriers, also might be analyzed together (or with
due regard for their role enabling other projects or programs to proceed or function
successfully.) Defining the "project" in this way would help internalize mitigation.
Certain projects are exempt from the conformity requirements under the interim guidelines.
These projects include such Hems as planning efforts, safety projects, land acquisition,
program administration, and research programs and projects. Final guidelines may alter the
16. Not* that a significant change In the dttign concept or acope could frigger a ri«w TIP analytic. For project-fevet
CO analyse*, mora modest design and operations details may be important to the results.
Tranaportation-AIr Quality Analysis laauaa
January 1993
-------
Pag* 2-39
project list and their treatment somewhat.
Exemptions for other projects, including some projects of the TCM variety, would probably not
be permitted; however, such projects might be first subjected to a quick response (screening)
analysis, with a detailed analysis only If a potential problem Is identified. The screening
process might ask a series of questions, with a "no" answer resulting in a conclusion that no
further analysis is needed. For example, if auto use is not involved or is trivial, e.g.,
pedestrian projects, then no further analysis would be done; if the project involves auto use
but does not increase the number of cold starts and reduces VMT, no further analysis would
be needed, ate.
Some MPOs have developed procedures which call for an initial screening (review of potential
impacts) of all projects, including those on the interim guidelines' exempt list This approach
was developed in response to a concern raised by some participants in the planning process,
who worried that in some situations "exempt list" projects might create CO problems. The
initial screening is seen as a simple, low cost way to either alleviate a concern or assure that
analyses are carried out as needed.
2.4.2 Alternate Analysis (Methods
At least six methods are in widespread use for modeling CO from transportation projects.
They are CALINE4, CAL3QHC, TEXIN2, GIM, IMM, and VOL9MOB4. A number of other
methods are available, however, and additional ones are reported to be under development
(including FHWA and EPA sponsored software.) In addition, several studies of methods for
project-level CO analysis are underway or planned, including a model evaluation being done
for EPA and a major NCHRP study to be undertaken in FY *83.
In general, CO project-leve! methods are used in conjunction with traffic volume estimates
produced separately. Traffic performance estimates may be produced by a separate modeling
effort as well (e.g., applying HCM), though for some methods, e.g. CAL3QHC, TEXIN2, and
CALINE4 this step is built into the software used to analyze CO. Usually this is simplified, eg,
CAL3QHC uses a reduced form of HCM and CALINE4 uses a modal emissions approach
which omits certain factors which would be considered in an intersection traffic analysis.
Similarly, emissions estimates may be separately produced by applying the appropriate
version of MOBILE (in California, EMFAC) or may be built into the CO analysis method (the
case for TEXIN and IMM).
Several of the models have borrowed extensively from the CAUNE model, which was first
developed by the California Department of Transportation in the early 1970s. Hence
substantial similarities are found among the models. Differences are due to 1) the number
and type of refinements added, such as ability to model street canyons; 2) the treatment of
traffic performance in or out of the model and the level of detail with which traffic is modeled
(e.g., whether approach lanes are fully represented, signal timing is considered, etc.); and 3)
whether emissions analyses are integrated directly into the model, or separate runs with
MOBILE or EMFAC are required.
Transportation'Air Quality Analysis lasuas
January 1993
-------
Paga 2-40
The general approach is to assume Gaussian dispersion of pollutants, as adjusted by such
factors as wind speed, wind direction/angle, stability, temperature, surface roughness,
elevation, etc. This treatment is an abstraction (simplification) of a complex set of
phenomena, but simplified though they may be, the models are quite data hungry. They
require very specific project descriptions (centertine location, elevation, width/number of lanes,
traffic controls, etc.). substantial traffic data (volumes, speed limits, slgnalization. etc.), and
for some models, estimates of anticipated traffic performance (expected average speeds,
acceleration and deceleration times, idle time, stops and delays, etc.) In addition, in order to
use the emissions factor models (or model components) estimates of the vehicle fleet mix are
needed. Hot/cold weighting, Inspection/maintenance program parameters, and other inputs
also are required.
Most models which have extensive requirements for input data and assumptions do provide
default values to reduce the burden on the user. Generally, the defaults are worst-case
assumptions (very low wind speeds, etc.) The level of analytical skills required .to apply the
models is largely a matter of whether default or other pre-spedfied assumptions and inputs
are to be used, or whether the analyst will exercise discretion. The latter case calls for
training or study in traffic engineering as welt as air quality modeling (though such education
and training are certainly helpful even If defaults are used.) Even well trained users report
difficulties in applying some of the models, however, in part because documentation is not
very extensive.
Most of the models now run on desktop or work station computers as well as on mainframes,
so computer resources are not a major problem. However, some of the desktop versions are
notably slow and some have reduced set of features (i.e., features were removed to enable
quick transfer from the mainframe to the PC environment) Also, for some of the models the
programming is not particularly user-friendly (e.g., it is easy to lose input data files or to
accidentally abort a run.)
Studies currently being done for EPA suggest that most of the models perform reasonably
well under worst-case scenarios. CALINE4, CAL3QHC, and TEXIN2 appear to be the better
sf the available programs, based on statistical tests of model results vs. measured results.
The preliminary reports suggest that CAL3QHC and TEXIN are somewhat more accurate than
DA LINE for isolated intersection analysis, though statistical differences are not large.
However, direct comparisons among the models are not a straightforward matter because of
he different ways traffic performance is handled. Differences probably stem from the traffic
steps rather than the emissions calculations and dispersion analysis components, which are
lighly similar (all are based on the CALINE routine).
Overall, model results are highly sensitive to the project description and assumptions about
:onditions. Early results from the testing studies suggest that all of the available models tend
o under-estimate emissions in comparison to field measurements when "actual conditions"
lata are used. This probably is due to the difficulty In measuring "actual conditions". For
ixample. such items as wind speed and direction can and do change over a typical
nonitonng period: "actual conditions" modeling is in fact modeling "average* or "typical"
anditions for the period. In addition, recent evidence suggests that estimated emissions
ates from the vehicle fleet may not be highly accurate. Moreover, the vehicle fleet in use at
'ranaportatlon-Alr Quality Analyala laauaa
January 1993
-------
Paga 2-41
a particular location is generally assumed to be the same as for the fleet for the region (or
state!); yet observation suggests that the mix of vehicles in use at any particular location can
and often does differ from the average.
CO analyses underscore some of the difficulties that arise in transportation-air quality
planning. Emissions modeling can never be more accurate than the data on which H is
based, and CO analyses depend on transportation or traffic measurements and models which
are widely understood by transportation professionals to be of varying accuracy. For example,
traffic counts on major facilities are more likely to be available, and accurate, than on minor
ones.
Emissions estimates prepared for future years reflect the uncertainties of the baseline
transportation and emissions data, plus uncertainties stemming from the estimates of future
conditions. Forecasts of future traffic volumes are dependent on a number of factors ranging
from anticipated increases or decreases in vehicle use per capita and mode choices, to
expectations for growth of the economy and population. Forecasts of future Vehicle fleets and
emissions depend on expectations for technological change, fleet turnover rates, and
assumptions about the driving patterns (driving cycles) for which vehicles will be used, among
other things.
Given these uncertainties K is important to exercise reasoned judgement in selecting and
applying CO models. It is not reasonable to insist on higher accuracy in CO modeling than
is feasible for the input data.
2.4.3 Particular Analysis Issues
A number of particular issues arise in CO analyses. They are reviewed briefly in this section.
Identifying Receptors In the Area Substantially Affected by a Project
A receptor location is the point at which pollutant concentrations are monitored or estimated.
The general rule is to locate or analyze receptors at a reasonable sample of sites where
people might realistically be exposed to high pollutant concentrations for a number of hours
corresponding to the ambient standard in question. EPA guidance suggests that reasonable
receptor sites would residences, hospitals, rest homes, schools, playgrounds, and the
entrances and air intakes to other buildings. On the other hand, while CO might build up in
a tunnel no one would be exposed for very long inside K, and so it would not be an
appropriate spot for a CO analysis.
For many projects a great many receptors could be identified in the general vicinity of the
project. A strategy for carrying out the analysis would be to identify the receptor location(s)
likely to measure the highest ambient concentrations. In case of doubt about what the worst
case receptor might be, the analyst would be well advised to examine all high-risk receptors.
In general, receptors more than 300 meters from a facility are unlikely to be significantly
impacted (though some experts have recommended a screening-level examination of all
Tmnaportation-Alr Quality Analyals laauaa
January 1993
-------
Paga 2-42
"sensitive receptors" - schools, nursing homes, convalescent homes, etc. • within 1 km.)
Analysts should be look especially carefully at areas where pollution is likely to build up (e.g.,
street canyons; major intersections; other areas previously identified as over or near the
standards).
The location of "future receptors" is an issue in modeling CO concentrations for future years.
One concern is what to do about areas where receptor development is permitted under local
land use plans and zoning, is anticipated as of the target year (i.e., is forecast to be in place
by the regional land use model or assumed to be in place in future year land use input data),
but is not yet approved. One view on this is that the project proponent must consider such
development to be extant for the purposes of future year analyses. Another view is that the
project sponsor need consider the same traffic assumptions as would result from the
development, but would not be held responsible for future CO impacts (and mitigation) on
such not-yet-approved future receptors. The latter approach would appear to be consistent
with environmental impact analysis regulations, under which only developments already
approved at the time of the sponsor's analysis must be considered "rear. This would have
the effect of shifting the burden of dealing with CO concerns to sponsors of future projects on
the affected now-unbuilt parcels.
Estimating Current and Future Year Levels of CO
Analysts frequently have doubts about whether to use data from permanent monitors to
estimate current levels of CO for a particular project area, or to collect new data specifically
for the project area. In many urban areas, there is concern that permanent monitors are too
few and far between to simply rely on the closest one (especially in cases where hills or
valleys intervene.) in some areas data from permanent monitors have been supplemented
by data from mobile monitoring, project-specific monitoring, and other special studies to form
a more extensive data base than the permanent monitors alone would provide, and air
agencies permit this data base to be used. In other areas, project-level monitoring is
frequently done.
It is not a simple matter to collect good project-specific CO data. Because K is extremely
difficult to "translate" data collected, say, during the summer to estimate winter month
readings, it generally is necessary to carry out the data collection at the time of year when
violations are most likely (winter). EPA generally asks for a minimum of four months of good
data collected during winter months in order to be assured of the statistical validity of the
monitoring findings. State and local agency requirements may differ, some California air
districts, for example, seek at least one full year (and preferably two or three years) of
monitoring data. Given the time and costs involved, M seems realistic to expect original data
collection only for larger projects, and then only if available data are insufficient or
problematic.
Once a current level of CO has been estimated, estimated CO levels for future years, with
and without the project, must be produced. This is often a difficult step. The rollback method,
which treats future concentrations as proportional to current ones in the same ratio as future
emissions are to current emissions, is frequently used. However, there are known problems
Tranaportatlon-Alr Quality Analyala laauaa
January 1993
-------
Pag* 2-43
with th® rollback method, which has been shown riot to predict well. One limitation is that the
rollback method treats meteorology as constant. (Other reasons for poor predictions may be
that data underlying emissions estimates, including estimates of future VMT and emissions
factors, have not been especially accurate.)
Air district staff could greatly assist in this step of the analysis process by 1) mapping areas
in which data from each monitor could, fn their view, be used appropriately; and/or 2)
providing grid cell emissions estimates or isopleths of CO levels, as well as forecasts of future
year levels for the no-build alternative.
Emission* Factors
CO models require use of a set of emissions factors for the vehicle fleet, both for current
estimates of project emissions and for forecasting future impacts. For most-models analysts
also must make a number of adjustments to the vehicle fleet data to reflect project-specific
conditions. In particular, the share of heavy duty trucks is a key assumption for CO analyses.
In addition, the fleet may vary by time of day; the percent cold starts can be expected to vary
as a function of project type; and corrections may need to be made to the emissions factors
to account for altitude. Data on these matters usually must be collected for the project,
although some states provide guidance on these choices.
Shifting CO Violations
A very difficult issue is what to do if the project shifts the location of a CO violation. One view
has been that this situation should be treated as a "new" violation, which would prevent a
positive conformity finding. An alternative view is to consider the net impact of the project,
and apply a "no net increase* test. This would permit a conformity finding if the new location
would be no more severe a violation than the eliminated one. A third option is to require a
showing that the number of violations has been reduced (not increased) and/or the severity
lessened, i.e., to show a net improvement in air quality, taking into consideration the overall
objective of the CAA in reducing or eliminating the number and severity of violations. The
issue was not resolved in the interim conformity guidelines and at the time of this writing this
issue is still under debate among the drafters of the final conformity regulations.
Dispersion Modeling Assumptions
A large number of assumptions need to be made in running dispersion models, and H can
require a significant effort to determine what are reasonable worst-case assumptions for each
project. In general, however, the models are relatively insensitive to variables such as surface
roughness and quite sensitive to wind speed, wind direction, stability class, and temperature.
It generally is recommended that worst-case assumptions be used for screening purposes.
If a project is marginal or problematic using the worst-case assumptions, then the analyst
would develop the data base for project-specific worst-case conditions and conduct further
Tranaportatlon-AIr Qualify Analysis lasums
January 1993
-------
Page 2-44
analyses.
Mitigation
Mitigation of project-specific CO violations may be difficult because of their localized nature.
Nevertheless, mitigation may sometimes be a useful option, i.e., mitigation measures might
be added to a project in order to reduce CO estimates to acceptable levels. (When mitigation
elements are already present or expected due to SIP provisions or other planned actions, their
effects should be accounted for in estimating the "baseline" emissions for the project area.)
What measures are appropriate as mitigation will depend in targe part on the specifics of the
project. Some areas are known to be prone to CO buildups, hence projects located in those
areas are more likely to be problematic than projects located elsewhere. When projects are
located in violation areas or areas prone to CO buildups, alignments that reduce, the exposure
of the population to CO should be sought whenever possible.
For some kinds of projects specific traffic mitigation measures are appropriate. For example,
traffic flow improvements (intersection redesign, signal retiming, grade separation) could
reduce or eliminate CO problems related to queuing. Restrictions on idling or use of dean
vehicles might be considered if a CO problem arises with regard to a terminal, e.g., a bus
timed transfer facility. Time of day restrictions might be suitable mitigations in some cases.
Tougher ridesharing requirements for use of an HOV lane might be imposed in some
circumstances.
Mitigation measures must be in addition to those already accounted for in the analysis
"baseline." If mitigation measures are TCMs, H will be necessary to show that the effort is In
addition to that already committed in the TCM or contingency portions of the SIP.
Tranaportation-AIr Quality Anatyala taauaa
January 1993
-------
CHAPTER 3: LONG-TERM IMPACTS OF TRANSPORTATION INVEST-
MENTS
3.1 Overview
The impact of addino transportation capacity has been a recurring focus of research and
policy debate. It is widely agreed that, for both highways and transit, capacity increases can
increase speeds and smooth traffic flows, and in to doing shorten travel times and reduce
vehicle operating costs while serving more people. On the other hand, both theory and
observation suggest that capacity increases can alter travel choices in ways that could offset
(and potentially more than offset) some of the time and cost improvements thai a static
analysis would predict. From a broader perspective, transportation enhancements may
redirect and restructure the spatial distribution of development opportunities, and in so doing
alter the patterns of land use and development, the opportunities for economic growth, and
the quantities, characteristics, and locations of local and regional population and employment
increases. These structural changes in turn may affect travel patterns in ways that may alter
the transport impacts of the capacity investments.
While this complex set of potential effects Is generally recognized, the likelihood and relative
magnitude of the various impacts are not well understood or agreed upon. Moreover,
practical methods for estimating these impacts are only now beginning to come into
widespread use. Thus, questions range from what we know, to what we can estimate, to
what we should do to improve our understanding of the issues.
3.2 What We Know
Previous Impact Analysis Results
A large literature examines the impacts of transportation investments on travel choices and
development patterns. This has been the subject of a number of literature reviews, e.g.,
Deakin (1991); Garrison and Deakin (1991); Giuliano (1988).
Much of the "impacts" literature is retrospective (i.e., It looks at the impact of facilities without
a "before" or "control* case). Some of H reports the findings of reasonably well designed
investigations of rail investments; the BART Impact Program (summarized in Webber, 1976)
and David Boyce et al.'s (1972) analysis of the Linderrwold rail line between Philadelphia and
the South New Jersey suburbs are examples of credible post-project evaluation work.
Numerous post-implementation studies of highways' impact also appear in the literature, but
these are of uneven quality. Well-constructed investigations include the work done by
Garrison and Berry in the 1950s and '60s, as well as some of the later studies of the impacts
of highways and beltways (e.g., Payne-Maxie Consultants and Blayney-Dyett Associates,
1980).
TnnMportaiJon-AIr Quality Analysla laauaa
January 1993
-------
Page 3-2
Based on these studies, the general conclusion is that « at least under current levels of
deployment and investment - transportation infrastructure has an indeterminate but probably
minor effect en overall growth rates, and has only modest impact on development patterns.
The limitation inherent In the bulk of the studies reported In this literature is that conditions
rarely permit an independent determination of what would have occurred without the
investment in question (i.e., the comparisons are before-after, but not with-wtthout). In other
words, there is no control sufficient to contrast with the experiment Indeed, H may not be
possible to find a control, since the underlying conditions of each project can differ in so many
ways.
In the BART example, ten years' prior knowledge of the exact location of each BART station
appeared to have had little effect on the subregional location of activities, particularly at
suburban locations: except in San Francisco, development patterns around BART were little
different from long-range trends for the region as a whole. However, major investments in
highways during this same period had created many nodes where enhanced transportation
capacity could facilitate development The BART studies did not explicitly account for these
competing nodes and their potential impact in spreading development to many locations. In
addition, local government planning restrictions actively prevented proposed development and
redevelopment at high densities around several of the BART stations, a factor noted but not
sufficiently examined in the BART Impact Study work.
When controlled case analyses are difficult, more abstract modeling may offer insights, and
another part of the literature involves theoretical and simulation model results. (See, e.g.,
Lerman, 1975; Anas, 1985; Harvey, 1990; and reviews by Giuliano, 1988, and Berechman
and Small, 1987.) This body of work suggests that the accessibility afforded by new capacity
is one of a number of factors that influence location decisions. Only in cases providing an
order of magnitude increase in accessibility - such as would result from a plantation road
which literally opens up new territory - is new capacity likely to be the chief factor affecting
location. In already developed areas, the increase in accessibility afforded by the kinds of
capacity increases typically considered is much more likely to operate at the margin. Other
factors, including land and building availability and prices, infrastructure availability and cost,
local services and amenities, access to appropriate labor and other inputs (for businesses),
and access to suitable schools or other social services (for households) are equally or more
important than the marginal transportation differences usually observed.
Nothing in the literature, however, supports a sweeping conclusion that whole transportation
programs would have little effect on development patterns. In other words, significantly
different infrastructure proposals (e.g., massive freeway expansion versus massive transit
expansion versus large-scale pricing to reduce congestion) may lead to significantly different
land use outcomes, especially over long periods of time, even if large individual projects do
not seem to. Moreover, work suggesting that the aggregate level of transportation investment
is positively associated with economic growth rates (e.g., Aschauer, 1988) implies that targe
scale regional, interregional, and national impacts may be present (but hard to directly
attribute to specific projects.)
Government action to coordinate land uses with infrastructure provision also may shape
growth patterns. For example, recent statutes that tie development approvals to the
Tranapoitatlon-AIr Quality Analyalt laauaa
January 1993
-------
Paga 3-3
availability of adequate transportation capacity (reviewed in articles by Deakin, 1888, 1989)
likely will strengthen the perception of causal links between transport investments and
development patterns. In addition, some developers are now pursuing strategies to
coordinate transportation and land development. Transit station area plans and larger-scale
transit- and pedestriarvoriented "neotradrtionar town plans - tee, e.g., Duany and Plater-
Zyberk, 1988, and the articles in Kelbaugh (ed.), 1990 - ane implicit attempts to achieve a
particular land use-infrastructure linkage.
Effects on System Performance
The impacts of added capacity on system performance depend on the specific issues under
consideration. If the question is: "How do point-to-point travel times change when new
capacity is added to a highway system", the answer is fairly well understood. If there is no
significant effect on the number, timing, and destinations of trips, new travel times can be
calculated for the affected facilities and new routes can be determined for the fixed pattern
of trips. Added capacity would reduce travel times and improve speeds.
If there is an effect on the underlying tripmaking pattern, however, the calculation of new
travel times (and speeds) must account for the changes. In general, one would expect new
capacity to increase accessibility of served points, and to make the benefitted mode(s) of
travel more attractive relative to those not improved. The increased accessibility would
support an increase in trip-making and generally would increase the envelope of reasonable
destination choices, whereas the change in modal competitiveness should change modal
choice. In addition, improved capacity may lead some people to shift the time of their travel
to periods that formerly were too congested for them, but otherwise would be preferred, in
the longer run, changes in location of workplaces and of residences may result from
alterations in accessibility.
The magnitude of these shifts depends on the level of capacity improvement in question, in
comparison to the baseline. Overall, the impact will be to offset at least some of the
transportation system benefits of increased capacity, although it should be kept in mind that
this is not the same as total social benefit. (Total social benefit could increase if more people
are able to engage in a desired activity, even if overall travel times, speeds, etc. do not
improve.)
The determinants of system performance outlined above are reasonably well understood,
although estimating the impacts is not necessarily an easy task. Most of the uncertainty
surrounds the travel demand response. Having said this, K should be acknowledged that
traffic flow phenomena and network complexities are not perfectly understood. The
relationship between traffic volume and speed, for example, has been represented as a
smooth polynomial function, but is increasingly understood as a more complex "chaotic"
phenomenon. To what extent this observation is valid, and how important it might be in
practical transportation analysis, remain to be determined.
The issue of network complexity has more intriguing implications for transportation planning.
There has emerged in the research literature a concern that certain network additions in
Transportation-Air Quality Analysis Issuas
January 1993
-------
P*ga 3-4
already developed systems may actually worsen overall system performance (a phenomenon
referred to as Braess* Paradox), For example, research has shown that Manhattan's street
grid might function more efficiently at current volumes If certain thoroughfares were dosed to
through traffic. This result stems from the route choice behavior of travelers: as individuals
act to minimize their own travel times (or times and costs), the resulting patterns of congestion
generally will be more severe than If everyone were assigned to a route by tome sort of
centralized control. This is true because some individuals may find certain routes shorter
even though their use of the route causes delay for a far larger group of travelers.
in concept, it should be possible to capture this effect with existing traffic assignment
software. However, the practice of transportation network modelling is not, in general, earned
out at a level of complexity or sophistication which would support such an analysis, in part,
this is a reflection of the models in use: For example, some smaller urban areas still use all-
or-nothing route choice algorithms without capacity constraints; a somewhat larger number
of urban areas use incremental assignment approximations of equilibrium flows, in addition,
there is little motivation for transportation modeling staff to actively seek out potential negative
impacts of this sort when evaluating proposed capacity additions. Hence, for the most part,
the exploration of paradoxical effects of system expansion remains a matter for research.
Effects on Travel Behavior
Key travel behavior variables, in the short run, can be thought of as reflecting a series of
decisions on: whether or not to make a trip or set of trips, whether trips should be combined
(linked) or made separately, the time(s) of day to travel, the mode(s) to use, the route(s) to
choose, and for all but the work trip, the destination(s) to choose. For some of these trips,
which car to use and whether to travel with others also are at issue, in the longer run, work
trip destination, auto ownership levels, and housing location are additional choices that
travelers make.
Suppose that substantial levei-of-service improvements are made in a major corridor.
Travelers already using the corridor might
¦ change their routes to make use of the newty improved facilities
¦ change to more desirable trip times that previously entailed congested conditions
¦ switch to the improved travel mode
¦ decouple trips that previously were linked, because K is now easier to make multiple
trips in the corridor
¦ make additional trips now that it is easier to do so.
Others not now using the corridor might:
¦ change the destination for a given discretionary trip (such as clothes shopping),
because locations in or near the comdor have now become more attractive
¦ decide to make additional discretionary trips, because particularly desirable locations
are now made more accessible
¦ decide to own more automobiles to support the increased trip-making (or to own fewer
Tranaportatlon-AIr Quality Analyst* laauaa
January 1993
-------
Paga 3-5
automobiles, If the improvements are to an alternative to the auto aueh as transit)
¦ switch to a workplace whose accessibility characteristics are affected by the corridor .
¦ move to a new residence whose accessibility characteristics are affected by the
corridor.
Developers and commercial establishments might
¦ construct additional commercial and office space in or near the corridor, in response
to market pressure or in the belief that additional infrastructure capacity will attract
more potential customers to the area
¦ construct new homes on land affected by the corridor, on speculation or In response
to market pressure
¦ move significant new economic activity into the corridor, occupying either vacant or
newly-constructed space.
Finally, at an inter-regional scale:
¦ substantial construction expenditures may themselves attract additional population
(e.g., construction workers who become permanent residents) or economic activity
(e.g., construction management or supply firms that remain in the area)
¦ new employers may be attracted to the region by improved accessibility (which leads
to lower costs and higher productivity)
m new residents may be attracted by the availability of additional jobs..
The latter two categories are property considered development effects (discussed in more
detail below) rather than travel behavior effects, but their influence on travel demand could
be profound.
A modest amount is known about these effects as individual phenomena (especially about the
more immediate travel effects). In particular, work mode choice has been extensively studied
using sophisticated statistical techniques and carefully collected data. This has led to an
appreciation of: 1) the critical role played by time and cost; 2) the different perceptions of
time depending on qualitative aspects (such as wait time versus In-vehicle time); 3) the role
of income in restricting travel choices; 4) the interdependence of mode choice and trip
scheduling, and 5) the importance of personality, life cycle, intra-household dynamics, land
use characteristics, and institutionally-based incentives.
The research community has addressed many aspects of travel behavior, including: 1) the
relationship between congestion and work trip timing; 2) route choice behavior as a function
of time, cost, and quality of information about alternatives; 3) the role of accessibility In
selecting trip destinations and residential locations; 4) the effect of auto accessibility and local
land use characteristics on number of vehicle trips and auto ownership; and 5) the locational
responses of large employers to congestion. With a few exceptions, this work Is not mature
enough to find its way into everyday practice, although some regional model systems are
remarkably advanced (e.g., the San Francisco Bay Area models, the Portland OR models).
A relatively short-range additional research investment could produce implementable models
in several of these categories.
Transportation-Air Quality Analyala laauaa
January 1093
-------
Paga 3-6
In particular, travel behavior is well understood to be a function of income and socio-economic
factors as well as transportation characteristics and land use patterns; but socio-economic and
attitudinat factors are oflen poorly represented in travel models. For example, age, sex, race,
health, disability, education, job title, dress requirements of the job, need to transport work
material, social status, political outlook, previous experiences, and before- end after-work
responsibilities all probably affect travel behavior, though they are hardly ever modeled
explicitly. (The impact of these factors can be envisioned quite clearly If one considers who
bicycles to work; the impacts may be less obvious for other modes and travel choices, but
nonetheless are present.)
When variables such as these are not explicitly represented in models, their impact is
implicitly treated as fixed, and therefore captured in model coefficients. One issue is how
stable these variables actually are, particularly If basic factor inputs should change. For
example, the changing role of women may influence their travel choices far beyond what
would be predicted using standard time, cost, and income variables. Also, elderly people who
grew up driving automobiles may behave considerably differently from elderly people who
grew up before Worid War II.
Effects on the Environment
The impacts new highway capacity will have on auto emissions is currently a key concern.
Transportation specialists have long understood that low speed, stop and go travel conditions
produce high levels of carbon monoxide and hydrocarbon emissions. For a given level of
traffic, capacity increases which smooth out flows and/or increase speeds can reduce
emissions.
There are some important limitations. Emissions of oxides of nitrogen increase with speed,
and in those areas where NO, is an issue this means that speed improvements are not wholly
beneficial. Also, with current vehicle technology CO and hydrocarbon emissions increase at
high speeds (over 55 mph), along with energy consumption, noise, and possibly accident
potential. In the many areas where increased capacity facilitates midday and evening travel
at such high speeds (whether legal or illegal), this must be taken into consideration.
Nevertheless, in many applications speed increases will have an overall beneficial impact on
vehicular emissions, and also on fuel consumption and vehicle wear and tear costs, for a
specified traffic volume.
The above observations are for a specified traffic volume X. As discussed earlier,
increasingly at issue (especially for highway projects) is whether the estimate of X has
properly accounted for changes in route choice, changes in time of day of travel, and induced
vehicular travel (mode shifts, destination shifts, increased trip generation). The time of day
and induced travel issues are particularly controversial given the forecasting methodologies
currently used in practice, which - with some exceptions - are relatively arbitrary (especially
for time of day) and/or simplistic (for destination choice and trip generation.)
Environmental effects of added capacity, transit or highway, are not limited to air quality. In
many areas a key issue is the impact of right of way taking Itself - the amount and nature of
Tranaportatlon-AIr Quality Analyst* faauaa
January 1993
-------
Paga 3-7
the land consumed in building or expanding the facility.. The implications of the capacity
expansion on the natural and built environments also are major concerns. Increasing traffic
through residential areas, imposing a harsh noise and vibration impact on abutting structures. ^
loss of valued farmlands, destruction of plant and wildlife habitat, and disruption of parklands
and wild areas are among the issues that are raised. Longer term impacts include growth
inducement and its effects, positive and negative - reduced viability of farming in areas made
available for housing development by their increased accessibility, for example.
Effects on Development
It is well accepted that additional transportation capacity will alter the development potential
of the benefitted sites and thus change the competitive advantage of different parts of a
region. Common sense, anecdotal evidence and a limited amount of formal empirical work
Indicates that, all else being equal, developers prefer areas with good accessibility to ones
that are congested or capacity constrained. Particularly in areas where developers may be
asked to help finance capacity deficits, they tend to seek out locations that already have
adequate capacity to accommodate their projects. But transportation capacity is but one
factor in the development equation; developers (and the banks that lend to them) need
reassurances that there will be demand for their products, and consider the full range of
factors that affect residential and business location choice. Consequently, they often find
congested, capacity constrained sites more desirable than ones with good transportation
access but few other attractions.
Along similar lines, residents value the accessibility provided by good transportation facilities
and services, but transportation is only one factor in housing location choices. Consumers
look for housing in their price range, with the number of bedrooms and other features
(including yard space) they need and want, taking into consideration a bundle of other
attributes including crime rates (especially a consideration for women), local government
services, neighborhood status, school quality (for families with children), as well as
transportation. Moreover, transportation access is not a simple function of capacity (since
travel to shops, etc. often is carried out off-peak when capacity is not an issue.)
Businesses also value accessibility, but in combination with a number of other factors such
as the availability, quality and cost of labor, land, and public utilities, state and local
government regulatory environment, tax policies and tax rates, access to specialized services
in banking and law, and a host of industry-specific considerations. Many businesses cluster ,
in specific locales in order to take advantage of agglomeration economies. For some
businesses, areas that offer particular amenities to executives and top professionals keep and
attract businesses despite high costs (dollar and other). Here, too, transportation is hardly
ever a dominant consideration or a simple factor in the location calculus.
Overall, then, added transportation capacity can be expected to shift development patterns
and potentials by altering the pattern of accessibility within the region. The magnitude of such
shifts will depend on the nature and magnitude of the capacity increase and accessibility
change relative to existing capacity and accessibility, and on the presence or absence of other
factors that support or deter development
Tranaportatlon-Alr Quality Analysis laauaa
January 1993
-------
Pag* 3-9
The preceding discussion has treated development Impacts as a within-region allocation
question; I.e., the issue considered is where expected population and employment growth will
be located. There is increasing interest, however, in whether and under what circumstances
investments in transportation can make a significant difference in the overall level of regional
growth. Much of this interest has been stimulated by work by Aschauer, who has argued that
aggregate levels of investment in infrastructure are reflected in the rate of economic growth
and development. This work has led to an ongoing controversy involving questions of
causality versus correlation and the direction of causality that may exist. These remain
unsettled research questions.
Finally, there is a question of competition between metropolitan regions and whether
transportation investments may affect the outcomes. Here, too, there is relatively little work
to support or refute competing claims. A few observations may help frame the issues, though
not resolve them. If there are significant differences among regions in the overall level of
transportation service available, transportation improvements potentially could, affect overall
rates of economic and population growth. Viewed from this scale, differences in within-region
transportation quality and availability are likely to be small, in part because federal funding for
transportation has tended to diminish state to state and region to region differences in
expenditures on transportation infrastructure. The transportation differences that do exist are
likely to be less critical than other differences affecting business and household location (wage
rates, taxes, services, cultural opportunities, climate ) Moreover the Importance of
transportation investments will surely be highly sector-specific, reflecting the rote of
transportation in the production of the goods and services in each economic sector.
More important to the inter-regional competition issue art the intercity and international
transportation connections - airports, seaports, rail, intercity highways. Both the quality of the
transportation facilities and the region's location (central/peripheral, close/far) with respect to
major national and international markets and trading partners will influence the region's
competitive position economically. Changing transportation technologies have altered these
patterns (consider the impact of railroads in the 19th century, airports and containerization of
freight in the 20th), and will undoubtedly continue to shift the competitive edge • integrated
multimodal port-road-rail double stack container connectivity is doing this right now, and IVHS
technologies may do so in the coming years. Again, this is a topic catling for considerable
research.
3.3 How to Address the Problem
Better understanding of the travel effects of added transportation capacity will require
advances In the theoretical understanding of the issues posed, research and development on
forecasting methods and models which better reflect agreed-upon relationships and
interactions, and empirical studies to test theory and models against evidence. Moreover
models and empirical studies are a means of testing theory. Some of the areas where
research and development efforts might begin are discussed in this section.
Tranaporiatlon-Alr Quality Analyst* la»ua$
January 1993
-------
Paga 3-9
Forecasting Models:
A project currently is underway to specify agreed-upon acceptable practices for regional
modeling, to provide an initial set of guidance documents on modeling, and to encourage
further research and development efforts where identified uncertainties persist. The project
is being carried out under the auspices of the National Association of Regional Councils, using
funding from EPA and FHWA. The initial workshop of this project, attended by some 80
representatives of MPOs, state DOTs, federal and local agencies, and environmental groups,
as well as consultants and academics, revealed that the transportation-growth issues are of
concern nearly across the board - in areas with severe air quality problems, and in areas with
clean air but concerns over other impacts of growth, in the largest MPOs, in mid-sized regions
where concerns over maintaining the quality of life are key, and in small regions where loss
of farmlands and growth in vacation homes is at issue. The desire to upgrade data bases and
forecasting capabilities was widely shared.
Some improvements to operational models were generally fett to be feasible and necessary
in the next few years. They include:
¦ expansion of highway networks to incorporate all arterials and significant collectors
¦ improved equations for calculating link speed as a function of volume
¦ improved representation of ridesharing in networks and mode choice models
¦ incorporation of a somewhat broader concept of accessibility into trip distribution
¦ routine recursion to ensure consistency among travel times computed in trip
assignment and travel times assumed in trip distribution and mode choice
¦ use of transportation-land use models in areas where growth stimulus is a high profile
issue.
Other improvements would require additional research and development Among these are:
¦ explicit modeling of time of travel and peaking
¦ inclusion of highway and transit accessibility in trip generation and auto ownership
¦ representation of walk trips throughout the model hierarchy
¦ full-scale modeling of transportation-land use interactions.
In each case, progress has been made, but more work is needed.
Improved models should be useful in helping to clarify the possible impacts of alternate
choices, and to support "What if..." analyses. Models are tools, however, and there is a
difference between using models to inform choices and relying on model output to determine
decisions. Model results must be put into perspective and tempered by common sense and
knowledge of factors not modelled. On the other hand, it is not unreasonable to expect
planners and analysts base their decisions on data and analyses rather than rely solely on
judgment and assertions of expertise, and furthermore to expect the analyses to comport with
accepted theory and practice, or be improved so that they do so. Until additional modeling
research and development is funded adequately, modelers - and decision-makers - may find
themselves in an increasingly difficult position.
Tnnsportatlon-AIr Quality Anafytla faauaa
January 1993
-------
Paga 3'10
Empirical Studies:
While modeling can provide important Insights into relationships among transportation, land
use, and development, many decision-maker* would be more satisfied with on-the-ground
evidence. Can the theoretically hypothesized relationships discussed in this paper be
observed in the real world? For many of the hypothesized changes resulting from travel
investments (time of day of travel, route choice, mode choioe, destination choice) the answer
is surely yes, and indeed recent technological advances such as smart cards or satellite
tracking might support allow a relatively straightforward (if initially costly) data collection effort-
Using more mundane data sources, it is possible to infer many of these changes through
systematic data analyses, and indeed much more work of this sort should be encouraged.
Observation of longer-term changes, e.g., the housing location choices and auto ownership
impacts of congestion, or the growth patterning effects of transportation investment, is more
difficult. A modest body of work has been carried out to explore location and auto ownership
choice processes, using data of both the revealed preference (longitudinal surveys, time
series surveys, surveys on moving behavior) and the stated preference (attitudinal surveys,
focus group, experimental-situationa!) varieties. Much of the work to date has been for PhD
dissertations (or is proprietary: some of the efforts on moving behavior and housing type
choice, for example) and has not been carried further or generalized beyond the cases
examined. Here too, much more work could be profitably done.
Unfortunately, the methods proposed as alternatives to modelling run into serious difficulties.
Before-after studies, for example, are frequently subject to problems of interpretation because
of changes in uncontrolled variables (e.g., if the economy is strong before and weak after,
how much might the observed changes in travel be dampened due to the downturn? if a local
government promotes growth before and restricts H after, how can that effect be taken into
account?). Even defining what's "before" can be problematic (e.g., real estate prices may
increase in anticipation of a transportation investment). In addition, case study data may be
subject to time-slice effects, e.g., the factors they measure may be due to the timing of the
data collection rather than a more basic underlying phenomenon. There is evidence that this
is much of the reason for the low trip generation rates observed in some suburban office
complexes: the trip rates reflect low occupant per sq. ft. ratios, which in turn are a reflection
of office overbuilding and resulting low rents.
Despite these difficulties, however, additional empirical work, if carefully designed and
adequately funded, could add to the understanding of these complex phenomena. Perhaps
the biggest risk in this regard is that underfunded, rushed, and poorly conceived work in this
area will undermine interest and support for the serious study of the topic.
Institutional, Financial, and Social Considerations:
Institutional, financial, and social issues also strongly color transportation-land use policies
and decisions and deserve research attention. Among the topics which arise in this regard
are the following:
Tranaportation-AIr Quality Analyaia laauaa
January 1983
-------
#>*£• 3-11
m the capacity (financial and technic*f) of current institutions to cany out the analyses
implied by the debates over the modeling of transportation-growth effects
¦ effective uses of models in decision-making, recognizing the uncertainties Inherent in
modeling and forecasting and the widespread view that elected officials, riot models
and forecasts, should determine the land use plans to which transportation plans
respond
a the role of citizen participation in planning and decision-making and the concern that
complex modeling is a means by which professionals attempt to exert and legitimize
their control over transportation and growth decisions
¦ the changed set of opportunities and challenges posed by the greatly increased
flexibility posed by the Intermodal Surface Transportation Efficiency Act
Each of these topics would be an appropriate one for investigation and research.
REFERENCES AND BIBLIOGRAPHY
Alcaly, W. G., "Transportation and Urban Land Values: A Review of the Theoretical
Literature", Land Economics. V. 52,1876.
Alonso, William, Location and Land Use: Toward a General Theory of Land Rents. Joint
Center for Urban Studies Publication Series. Cambridge, Mass: Harvard University
Press, 1964.
Anas, A., "Modeling the Dynamic Evolution of Land Use in Response to Transportation
Improvement Policies", in G.R.M. Jansen, P. Nijkamp, and C. Ruijgrok (eds.),
Transportation and Mobility In an Era of Transition. North-Holland, Amsterdam, 1985.
Aschauer, D., "Government Spending and the Falling Rate of Profit", Federal Reserve Bank
of Chicago, 1988.
Baerwald, Thomas, "Land Use Change in Suburban Clusters and Corridors", Transportation
Research Record 861. 1982.
Bajpai, Jitendra, "Forecasting the Basic inputs to Transportation Planning at the Zonal Level",
NCHRP Report 328. Washington, D C., June 1990.
Bay Vision 2020 Commission, "Bay Vision 2020 Report", San Francisco, 1991.
Berechman, Joseph, and Kenneth Small, "Modelling Land Use and Transportation: An
Interpretive Review for Growth Areas", Institute of transportation Studies, University of
California at Irvine,1987.
Berry, Brian, Geography of Market Centers and Retail Distribution. Prentice-Hall, Englewood
Cliffs, NJ, 1967.
Boyee, David E. and others, impact of Rapid Transit on Suburban Residential Property
Values and Land Development-prepared for U.S. DOT, Office of the Secretary, University
of Pennsylvania, Philadelphia, November 1972. NTIS PB 220 693.
Burgess, E.W., "The Growth of the City", in M. Steward (ed.), The Citv: Problems of Planning.
Penguin Books, Baltimore, 1972.
Cervero, Robert, America's Urban Centers: A Study of the Land Use Transportation Link.
Unwin Hyman, Boston, 1989.
Cervero, Robert, "Light Rail Transit and Urban Development", JAPA. Spring 1984.
Deakin, E. "The Politics of Exactions," in R. Alterman (ed.), Private Supply of Public Services.
Tranaportatlon-Alr Quality Analyala laauaa
January 1983
-------
Page 4-2
plans. But this poses • pair of difficult questions: 1) is there enough evidence of effectiveness
in reducing emissions to warrant regulatory consideration of pricing measures as
transportation controls? and 2) in light of historical resistance to tolls and fees, is there any
chance that public opinion will support effective transportation pricing in the near future?
This Chapter provides an introduction to the major issues by describing the way in which
pricing has been handled as an air quality measure in California. The discussion draws from
a rich literature on transportation pricing as well as from three pricing analyses conducted for
California cities. These were:
¦ State Transportation Control Measure (STCM) Plan development for the Bay Area'
¦ A transportation pricing study for the South Coast Air Basin4
¦ Design of a congestion toll pricing demonstration for the San Francisco Bay Bridge*
The discussion also draws from recent studies on transportation modeling and air quality
carried out by the authors, including:
¦ A review of recent developments in transportation-air quality planning for the Federal
Highway Administration*
¦ An effort to define the state of the art in regional transportation modeling for the
National Association of Regional Councils7.
4.2 Institutional Setting
Transportation pricing concepts have been debated in the two largest metropolitan regions
of California, as well as by the California Air Resources Board (CARB), the California Energy
Commission (CEC), and committee staff of the California Legislature. Within metropolitan
areas, organizations as diverse as large businesses, environmental groups, MPOs, air
districts, and even advocates for the disadvantaged have sought to place transportation
pricing concepts on the policy agenda. While these groups do not necessarily advocate
3. Harvey, Greig anc Elizabeth Deafen. (1891) "Estimated Emissions Reductions from 9m Bay Ana
Transportation Control Manure Plan." Working paper prepared for the Bay Area Air Quality Management District
June.
4. Environmental Defense Fund, (1891) "Transportation Efficiency. Tedding Southern California's Air Pollution
and Congestion * Paper prepared for the Regional Institute of Southern California. March.
5. Harvey, Greig. (1881) "The Suitability of Bay Area Toll Bridges for a Congestion Pricing Experiment" Paper
prepared for the University of California Transportation Center Conference on Congestion Pricing, San Diego,
February 28 - March 1.
6. Harvey, Greig W„ and Elizabeth A. DeaWn (1982) "Transportation and Air Quality" Searching for Solutions:
A Policy Discussion Series, Number 5. Office of Policy Development, Federal Highway Administration,
Washington. August
7. see Harvey, Greig and Elizabeth DeaWn, (1881) "Toward Improved Regional Transportation Modeling
Practice." National Association of Regional Councils, Washington, DC. December, and Harvey, Greig and
Elizabeth Deatan. (1902) Manual of MPO Modeling Practice. National Association of Regional Councils,
Washington, DC. October.
Transportation-Air Quality Analyalt Isauaa
January 1993
-------
Pag* 4-3
implementation, they seem to agree that now is the time to debate the form of a broad pricing
strategy and to explore how such a strategy might be made acceptable to the general public.
A number of developments have converged to make pricing an Issue of debate in urban
transportation policy circles:
¦ More stringent ozone standards • The California one hour ozone standard is .09
ppm with no exceedences (versus the Federal standard of .12 ppm with not more than
three exceedences in three years). The California standard is based on a reading of
the epidemiological literature,* and there does not appear to be much sentiment at this
time for relaxing It. Even in a region that is relatively dean by the Federal standard,
such as the San Francisco Bay Area, the tighter State standard is extremely difficult
to meet without significant changes in automotive technology or extensive travel
reduction.
¦ Explicit treatment of transportation emissions under the new Clean Air Acts - For
example, some regions are required to plan for a 15 percent reduction in reactive
organic (ROG) emissions by 1896, beyond what would be achieved through
preexisting tailpipe controls. With hesitation about additional tailpipe controls, initial
plans may have to rely only on TCMs.
¦ Plan and program conformity under the Federal Clean Air Act Amendments -
Regional transportation plans (RTFs) may not be able to contribute enough to a
showing of attainment without a pricing option for reducing VMT. Without a showing
of attainment on schedule, and without a showing of specific progress at three year
intervals, H may be impossible to obtain the necessary plan and program approvals.
¦ Experience with "command and control" transportation measures -Trip reduction
programs based on voluntary, advertising-induced mode shifts by commuters have
had modest effect. Regulation 15 (the South Coast mandatory employer trip reduction
measure) has been effective mostly when employers instituted parking fees. There
is some sentiment to be forthright about parking fees if that's what really works.
¦ The ubiquity of congestion - Policy-makers have concluded that neither funding nor
public will is present in sufficient quantity to support "building our way out of
congestion.* This has led the business community (among others) to search for other
means of reducing congestion.
¦ Advances In technology - Automatic vehicle identification (AVI) has become reliable
enough to support large-scale monitoring of the vehicle fleet Given an identifier on
each vehicle, all congested points on the freeway and arterial system could be
monitored and priced (dynamically, if so desired).
¦ Fairness of the existing funding stream - The gasoline tax and other fees
proportional to use provide less than half of all transportation revenues in California,
while local sales taxes now account for over 25 percent* Increasingly, the sentiment
8. Recent press reports of lung lesions among Los Angel#* children, and of Increased asthmatic aensHMty from
chronic exposure to low levels of ozone, give some sense of the direction of this Weraturt,
8. California Office of the Governor. Governor's Budget Summary 1SI0-41.
Transportation-Air Quality Analysis Issuas
January 1993
-------
Paga 4-4
is heart! that explicit pricing might be fairer than the current system, especially if funds
were directed at the transportation needs of the low income community. Some groups
for whom equity is an overriding concern have supported this position.
¦ Evolution of the antHax movement - Taxes remain a political anathema, but the
willingness of voters to impose new taxes for specific, desired projects has become
apparent. It is possible that a large restructuring and expansion of the transportation
funding stream would be feasible if accompanied by dear, iron-dad expenditure
commitments.
4.3 Prototypical Pricing Concepts
Broad pricing strategies for the Bay Area and the South Coast are outlined in Tables 1 and
2. These fall Into two categories; "Market-based" and what might be called fee-based."
Market-based charges are ones that can be can be justified in a (more or less) pretise way
by the internal or external costs of transportation. The Bay Area Economic Forum, a business
coalition based in San Francisco, introduced the "market-based" terminology to the Bay Area
debate with a 1989 proposal for pricing as a substitute for "command and control" emissions
reduction measures. Two of their suggestions were adopted by Bay Area air quality planners:
¦ Congestion fees - Localized tolls in congested corridors throughout the region,
employing AVI technology. Under the Economic Forum proposal, revenues would be
reinvested in new infrastructure (transit or highway) until the marginal cost of a
capacity increment matched the congestion charge. Bay Area planners retreated from
such a rigid approach, assuming that some portion of the revenues might have to be
bartered for political support They substituted an arbitrary highway level-of-service
criterion (LOS D/E) in place of the Forum's marginal cost criterion. EDF proposed a
similar measure for the South Coast
¦ Emissions fees • Annual fee based on an estimate of vehicle emissions in the
previous year. At the time of registration, the vehicle odometer would be read and
tailpipe emissions would be measured. Coupled with information about the
characteristic performance of each make and model, these data would be used to
develop an estimate of annual emissions. Health and damage costs per unit of
emissions then would be used to set the annual fee. Vehicle fleet data suggest that
fees for the existing fleet might vary between $5 and 11000, with the average at S12S.
Institutional fees are ones arising from a convenient administrative framework for revenue
collection. In the Bay Area and the South Coast, analyses made it dear that market-based
measures alone could not achieve the state emissions reduction mandate. Planners fell back
on two other pricing strategies with known administrative requirements:
¦ Employee parking fees - The intent of this proposal was to achieve a minimum
employee parking charge of $3.00 per day, payable monthly, with the bulk of the
revenues recyded as added transit and ridesharing incentives. The $3.00 level was
loosely selected to represent the "opportunity cost" of land dedicated to parking in a
typical suburban location.
Tranaportatlon-AIr Qualify Analyal« laauaa
January 1993
-------
Table 1: Policies for th« Bay Area Pricing Study
Strategy
Description
Percent Change From 1M7 MoMIe Source Baseline
VWT
Tt*a
Feel
noo
CO
CO,
nes^e"^"^e cer*eee*
(•mi Pftelnf [hul e*.
•Mtkt OICI - X>M(M
se.ie per mm
Am m ¦ -* ** - " »* * Ua-JKmI.,. /AIM -
mi rawnpv vfnRW WPfmORVIn lufivinv
vwuM top u**4 to into# to# toMwvy spm!
- -» ¦* -* - ¦¦ ¦» ^ * a - « - * < ja r\a|
tm.
•u
•U
•ej
4J
.7M
•XJ
•MM
¦ nes^enwrtse ®^w^®eyee
I Parking Chart* el
] MM r*f Day
/Ul VPBf%9t9 to to# IVfploCI W##to #WJ>##to^t©# # Willi*
mum $3.00 (Iftf) pm 4&f ehfp# tor »w»Mn m
••to, pfcJrvp, #f ftn st to# MMlcpitocMi,
-u
-1J
•«
•M
A A
•tj
-IS
1 p>»»n Paifttof Maty*
1 «netfBy*tofto
tow# tostopwi#*
¦41
44
-«.r
-4J
4.1
4J
¦i.T
9 Bate* WftftfHn
9 Pee |a*er*§e ttiS per
I ««kicli|
£##• #wwto Imp jp#to towy#^f #t to# tow ttt i#pto*
#fwto®iofMi tof ##g§i vdite®# ^^#fto#dl town IN# #^*
#f toB|>^# •mtoatom).
•ej
«e.t
•ua
¦4 J
•4J
41
.1M
1 SimSm Tai tocreeie
1 kyS3.ee
Ftoi vwvto top p#M d to# jNHfflp*
-a.t
.fJB
-«a.s
•rj
'TM
-T J
•*$M
I Net tor lawtet-Seeea
1 atrateflee
-14 J
-i»t
-47.T
41.T
•MJ
.«7J
¦47.7
Definition of column (loading*: VMT » auto and private transit vehtele-rnliee travel**; trip* » aeto vehicle trip*; Nel ~ gallons of
fuel consumed; ftOO * amlaalonaI of reactive or§anlcs; CO • emissions of carbota monoxide; NO, » emlaalona of oxides
of nitro«en; CO, » amissions of carbon dioxide.
Each value represents the midpoint of the estimated range of effect Numbers In brackets Indicate variation above and below tfie
midpoint, based on sensitivity tests of hey parameters related to pricing (swell as the travel cost coefficients). Accuracy of
the estimates will depend as well on other uncertainties that are Inherent In any travel forecasting exercise, such as In
regional and sttbreglonal growth projectlona and In assumptions about future Infrastructure Investmsnts.
-------
Table 1: Policial for th« South Coast Pricing Study
Paicant Chang* Pram 2010 Mot) II* Sourca Batata*
Strategy
Daacription
W»f
TfV»
r«M
DM
CO
MO,
CO,
Pilthig pant if Mnttt
ami ¦ «w»r»i» m.is pm
mm
M#% Antomik Vfiitti MwMRbp(vi (AVfJ wowW
to HMtitoln t»t«l rn mnk» (Lost 0€.
*»
!~.»!
41
!~«!
41
1* •»
41
|#J|
•«.l
|*10|
44
(~I*
41
(~•I
Fw%»>t owee*mj»
Pmtmt
AS w^hMS to ftQ wa^®^ mmU s
1J 00 (1M1| pm Hj *wt» Iw »"** 1 m mm, fhMm.
-11
(~21
•U
l*-H
•IT
l*H
•t.f
!~*
41
IM
•tj
l*J|
4.r
l*-1
x f»iu»l cmni ml
¦ WHNHHHI
0£ hpQP 4PP^*4P®WHP^ In
job ^o^pj jwp^ pmprtppwi^w^ 99 9/tf^f90^^
¦M
l»«l
¦4J
|»4
-»•
l*JI
44
|*J|
41
!~«!
44
!~«!
41
(~J)
ltM|
-------
Paga 4-7
a Gasoline taxis - A simple increase in the pump price of gasoline by $2.00 per gallon.
The $2.00 level was loosely selected to reflect the average cost of a state-
administered automobile insurance program (but no explicit recommendation for a
state takeover was made).
Many other pricing approaches would be possible, but these were ehosen as representative
for the purposes of analysis and public discussion.
4.4 Critical Analysis Issues
A full-fledged pricing program could have far-reaching effects on the pattern of mobility in a
region, by altering the perception of accessibility and the cost of auto ownership in a way that
depends on household income. The major questions for analysis are whether available travel
models can represent the role of price in accessibility in a systematic and comprehensive
way, and whether such models adequately capture socio-economic variation-in the population.
These are not simple questions. They raise issues not only of representation and structure,
but of the fundamental behavioral validity of the dominant modeling paradigm. The following
is a partial list of critical modeling issues which arise in the consideration of pricing.
Key Variables
Clearly, the price of travel Is central to the analysis and must appear in appropriate places
throughout a model system (structural issues are discussed below). Components of price for
a vehicle trip should include parking costs, tolls, and perceived out-of-pocket auto operating
costs, the latter specified carefully to assure comparability with less subjective parking and
toll costs. In addition, the annual cost of operation (both fixed and variable) should appear
as a determinant of auto ownership.
Household or personal Income is a second key variable that should mediate the effect of
price wherever K appears in a model system. Under best practice, household income typically
is included to make the coefficient of price Inversely proportional to income,10 Unless income
appears jointly with price, the distributional consequences of pricing strategies cannot be
studied.
Congested travel time is a third key variable. Models should be capable of representing the
effects of predicted changes in demand on travel time, with as much time-of-day detail as
possible.
Behavioral Effects
Under the conventional paradigm of travel behavior, price could have noteworthy effects at
several levels of the model hierarchy;
10. This is dona by constructing • composite variable that it soma function of pfice and incoma.
TranaportatJon-AIr Quality Analyala laauaa
January 1993
-------
Pag* 4-9
¦ route choice/traffic assignment - Tolls and congestion fees Influence the
"impedance" of each route, which will produce changes in path assignments as fees
are differentially changed.
¦ time of travel - Fees that vary with congestion will induce some drivers with
scheduling flexibility to shift to less congested periods.
¦ mode choice - Price is a key determinant of modal competition for all types of travel.
¦ trip distribution - Differential price increases will cause a spatial shift in the trip
distribution from any given location, and a general price increase will lead to shorter
trips overall.11
¦ trip generation • For non-work trips, a general price increase could reduce the
amount of discretionary tripmaking. For work trips, a significant price increase (either
differential or general) could foster work-at-home policies, four-day work weeks, or
other reduced trip scenarios.
¦ auto ownership - By directly or indirectly raising the cost of auto -ownership or
decreasing highway accessibility, price increases could reduce the incentive for
multiple auto ownership.
¦ residential and employment location - Significant price changes may cause lower
income working households to seek less expensive workplaces or residential locations.
Conversely, reductions in congestion may induce higher income households to locate
farther from their workplaces."
¦ residential and commercial construction - pricing-induced changes in residential
demand or workforce availability might shift the locus of regional growth, or perhaps
alter the overall rate of regional demographic and economic change.
The same hierarchy of effects could be postulated for other large changes in the
transportation system, such as the cumulative effect of gradually increasing congestion over
a long period.
This hierarchy is of course hypothetical, and tied to a specific paradigm of travel behavior.
It seems likely that some of the postulated phenomena are more important than others, but,
unfortunately, the literature does not provide much help in sorting out the first-order effects.
In the intense debate over such a large change in transportation policy, there is pressure to
address each potential effect in a meaningful way.
11. The behavioral process is quite cfifferent for work and non-work tips. In the non-work case, people have Ihe
option to shift locations of liscretionary activities. In the work cast, people heva fbted origins (residences) and
destinations (pfacas of employment) In the short run, but can change either or both in the long run. In affect, a
work trip distribution model is an attempt to represent long-run residential and employment location behavior,
given fixed numbers of dwellings and wori^eoes.
12. Note that some elements of this spatial response would be covered in a conventional modeling framework by
work trip (Sstribution. Cautf on is required In order to avoid double-counting the** effects.
Transportation-Air Quality Analysis Issuas
January 1993
-------
Pag* 4-9
Structural Features
Key structural attributes of models involve linkages among elements of the behavioral
hierarchy, and degrees of disaggregation among plaoes, people, times, and facilities:
¦ model linkages and feedback - The models have implicit linkages which should be
reflected in the results of a pricing analysis. Perhaps the most obvious is the
presence of time and price at many levels of the model hierarchy. If time and price
influence demand at many levels, and demand determines time and price (through
assigned link volumes and estimated levels of congestion), then It becomes necessary
to perform a recursive analysis that checks for consistency among the input and
output assumptions. In rigorous terms, models should be run to "equilibrium", but
since this often is computationally impractical, there must be at least a systematic
effort to achieve consistency between predicted times and costs and those used in
calculating the various elements of demand.
¦ disaggregation « In all transportation modeling there is a tradeoff between the detail
required for accurate representation of supply and demand and the resource
requirements of increasingly disaggregate analysis. For pricing studies, H is essential
to have a detailed representation of the income and vehicle fleet distributions, and
helpful to have as much specificity as possible about highway links (for micro-analysis
of congestion pricing), analysis zones (e.g., for exact representation of parking prices
and other land use-related measures), and times of day (for careful representation of
time shift effects). In a given analysis setting, the trick is to find the highest level of
aggregation that can support an evaluation of the "first-order* effects.
Philosophical Issues"
Pricing and other recent issues in transportation planning have raised fundamental questions
regarding the state of knowledge about travel behavior. These include:
¦ Is the basic paradigm valid? Conventional transportation modeling is drawn in an
eclectic way from the methods of economics and quantitative geography. It has
served adequately - many would say well - in studies of incremental change to the
existing transportation system (such as the alternatives analyses and corridor studies
carried out by many MPOs). But there has not really been a thorough test of the
underlying concepts, including the assumption of a sequential hierarchy in decision-
making, the validity of linear utilities and unrestricted choice sets, the division of
tripmaking into a relatively small number of supposedly homogeneous trip purposes,
and so on. If these assumptions are deeply flawed, then It Is unlikely that existing
models will perform reliably in addressing any of the large questions implied by pricing
13. This section reflects • aynthesit of comments received by tie euthor over the two year* elnoe pricing
emerged on the Bey Area policy agenda.
Transportatlon-AIr Quality Analysis Issuat January 1993
-------
Paga 4-10
strategies.14
¦ It there ¦ direct tradeoff between time end cost? The placement of time and cost
variables In a linear utility equation Implies a direct compensatory relationship between
the two, i.e., travelers must be perfectly willing to substitute time for money, and wee
versa. Researchers have found evidence that cognitive assessment of time and cost
is perhaps sequential or related in some other less direct way. Any such finding would
profoundly influence the assessment of pricing and other congestion reduction
strategies.
¦ Is H acceptable to extrapolate beyond the range of experience with time and
cost? With all of the linear approximations, and absent some example of time of day
pricing in the estimation data, a full pricing strategy would be likely to extend beyond
the valid range of price coefficients. While indisputable, this concern Is mora valid In
some areas than in others. For example, Bay Area estimation data sets exhibit large
variations in parking costs, include tolls for an extensive subset of trips, and contain
detailed information about the vehicle characteristics that influence ownership costs.
The resulting models arguably should be valid for a range of pricing strategies, though
not for time of day measures. Other regions with less existing price variability would
not be able to use their models with as much confidence.
¦ Do we give adequate consideration to confidence intervals and other statistical
properties of travel models? t-statistics routinely are reported for model estimation,
but error calculations rarely appear with model forecasts. While it might be desirable
to provide confidence intervals, there are many good reasons why this is not done.
Chief among these is the realization that exogenous variables such as fuel price, rate
of regional population growth, and real household income are the greatest sourcas of
uncertainty. It is not uncommon to see alternate forecasts for scenarios of these key
variables, which is a way of expressing subjective confidence intervals.
While these are important concerns, It would be a mistake to overstate the conceptual
difficulties of urban transportation modeling. Research is ciearty needed on these and other
matters, but existing "best practice" models can be used as long as the greatest potential
inaccuracies are acknowledged and accounted for.
4.5 Pricing Analyses for the Bay Area
The previous section identifies a formidable set of considerations for analyzing pricing
strategies. One issue alone « the presence of price at all appropriate points in the model
hierarchy - rules out most of the operational travel demand models available to MPOs. The
14. One example often raised is the Zahavl time budget worfcofftt late 1870s. Zahavl found a r*m*ktf>le
consistency in the average deity time spent traveling among in dividual all over the developed wotld, wid inferred
the existence of a time constraint (or "time budget") analogous to the budget constraint of microeconomic theory.
His in real work with tie concept was promising, but, sadly, he died before carrying out a complete assessment
end, aside from one issue of the journal Transportation Research from the earty 1980s, no one has picked up this
strain of research. The Implication of Zahavfs theory Is that travel time Improvements due to new infrastructure
mey simply facilitate more travel rather than produce an absolute reduction in vehicle-hours of ftwel.
TranaportaHon-AIr Quality Anatyala laauaa
January 1993
-------
Paga 4-11
price of travel generally does riot appear outside of the mode choice equation in conventional
model systems.
The San Francisco Bay Araa Is fortunate both in having what is arguably the most
sophisticated operational travel demand model in the U.S. (at least in terms of structure and
specification), and in having served as the test case for a number of advanced travel demand
modeling research efforts. The pricing analyses for the Bay Araa and the South Coast drew
heavily on this existing model base.
Of primary interest was the travel demand mod*! operated by MTC. This model covers a
standard set of trip purposes (work, shopping, social/recreational, and non-home-based) but
does so in a manner that accounts for the affects of accessibility through much of the
behavioral hierarchy. Since K can be argued that the structural characteristics of the MTC
model were essential to carrying out a credible analysis, K is helpful to describe that model
in some detail:
The MTC model begins with standard network representations of the regional
highway and transit systems. Given interzonal trip tables by purpose,19 and
knowledge of peaking characteristics for each trip purpose, network assignment
algorithms are used to create peak and off-peak level-of-service tables by mode (auto,
transit, NOV)- Capacity-constrained assignments are performed for the peak periods.
Computed level-of-service data are then fed into a hierarchy of probability and
demand equations for each trip purpose. For non-work trips, the hierarchy begins with
a joint model of mode and destination choice, in which the zone-to-zone accessibility
is defined in terms of both travel times and travel costs. Trip frequency then is
computed using a non-linear equation with variables drawn from personal and home
zone characteristics, including the natural log of the denominator of the
destination/mode choice model.1' This "iogsum" variable captures the effect of overall
accessibility from the home zone on trip generation.
Work trips are modeled in a similar way, with mode choice nested under
destination choice rather than treated jointly. The "Iogsum" from the mode choice
model essentially serves as the accessibility measure in destination choice. Work trip
generation is treated exogenously to the travel model (as a fundamental output of the
land use modeling process).
Given a full set of travel calculations, auto ownership probabilities are computed
as functions of income, local land use characteristics, and relative accessibility by
transit versus auto. Accessibility is inferred from the denominator of the shopping and
work destination choice models, by dividing the composite transit utilities by the
15. Note that trip tables are both an Input and an output of the modeling exercise. The process begins with a
'guess" about what the fins) tip table will look like {usually based on a factored version of a previously-calculated
trip table). This initial set of tables is used to produce initial level-of-service estimates, which In turn are used to
calculate new trip tables. Then, new tnp tables are used to produce new level-of-service estimates, and the
demand estimation process is repeated. An Iterative procedure Is required to achieve consistency between level-
of-service assumptions and estimated demand.
16. For a log'rt discrete choice model, the log of tie denominator is an exact measure of the expected utility from
the choice depicted by the model.
Tranaportatlon-AIr Quality Analyala laauaa
January 1993
-------
Paga 4-12
composite auto utilities in each case (work trip accessibility is included only for
households with workers).
The four travel model levels - mode choice, destination choice, trip frequency, and
auto ownership - yield new trip tables, which can be integrated and assigned to the
networks to generate new estimates of level-of-sarvice. Technically, it ts necessary
to iterate the whole procedure until the travel demand estimates and tovel-of-service
estimates converge to an apparent equilibrium. Since repetition of the entire
sequence is expensive and time consuming, MTC exercises some discretion in
deciding how far up the hierarchy to check for equilibrium.17 More importantly, MTC
finds it beneficial to invest resources in generating the initial "guess" at a trip table, to
that first estimates of level-of-service are as good as possible.
A regional land use model, POUS, is operated by the Association of Bay Area
Governments (ABAG). It uses highway levei-of-servioe data in allocating new
development among the cities and towns of the region. In some applications, MTC
provides ABAG with revised levehrf-service tables in order to verify that land use
allocations would not be affected by changes in highway accessibility.1* Experience
to date has not shown much sensitivity of POLIS to typical level-of-service
adjustments, but future studies could require iteration of POUS with the MTC model
system.1*
The MTC travel model was developed in the mid-to-late 1970s,*1 using data from
a 1665 regional home interview survey and from more focused surveys conducted
before and after the opening of BART. Some adjustments were made based on a
1981 regional home interview survey, and MTC contemplates a major update based
on their recently-collected (1990) survey. The model is described here as originally
delivered by the consultants.
While an evaluation of the accuracy and validity of the MTC model was beyond the
scope of the earlier work, it is appropriate to comment briefly on these matters. The
strength of the model lies in its normative treatment of travel demand, i.e., as a
hierarchical arrangement of interrelated travel choices. From the model
documentation, it is dear that some of the linkages are significant while others may
17. For example, a routine transit alternatives analysis coveting • small area might be balanced only through
mode choice, whereas the annual TIP analysis covering all new projects over 5 years would be balanced through
auto ownership.
18. Transit accessibility could be added ts the POUS framework, but tie resources to do so turn not been
made available.
19. POUS is an optimization model showing a significant effect of highway accessibility on location decisions but
also reflecting the constraints on development imposed by existing zoning maps. \M»en POUS is run without
releasing zoning constraints, the location el effects of accessibility changes are quite muted, if land use
constraints were released, the locations! effects would be much greater.
20. Model development was carried out by Cambridge Systematica with support from researchers at the
University of California, Berkeley.
Tranaportatlon-AIr Quality Analyst laauaa
January 1993
-------
Pag* 4-13
be less so.21
Much less clear is the superiority of the MTC specification and structure in
comparison with other possible realizations of hierarchical, accessibility-dependent
travel choice. In particular, advances in nested choice modeling have been made in
the last decade that are not reflected in MTC's model because they were not well
developed at the time.33 Also, model development resources did not allow a complete
exploration of alternate specifications. MTC can be expected to update its models in
due course to address these matters.
With two exceptions, the MTC travel model appears to offer a reasonable first-
order treatment of the key structural effects of time and cost - a statement thai cannot
be made about any other operational travel model. However, the exceptions are
important and instructive about the state of travel demand theory. The MTC mode!,
like all other regional travel demand models, ignores trip chaining (the traveler's ability
to save time and cost by making two or more stops before returning home) and treats
time of travel in an ad hoc way. A behavioral basis for time of travel choice is a
particularly important omission, since a major effect of congestion pricing would be to
encourage workers and employers to shift work start times.
The Bay Area and South Coast pricing studies dealt with time of travel through a
procedure which relates the degree of peak spreading in a corridor to the level of
congestion (indicated by the ratio of peak to off-peak travel time). Congestion prices
were converted to equivalent increases in peak time, using the value of time for each
traveler, and then translated into expected increases in the width of the peak. This
dearly is a simplified approach. Travel demand modelers with an interest in pricing
or in congestion management have now recognized the need for formal models of
peak spreading (or time-of-trave! choice), and have placed a high priority on
incorporating such capabilities in the next round of model improvements.
Two options were available for practical use of the MTC model: standard operation in the
UTPS large-scale modeling framework (MTCFCAST); and non-standard operation in a
microsimulation framework (STEP).
MTCFCAST reflects an adaptation of the "pure" MTC model to the constraints and
practicalities of large-scale modeling. There are detailed highway and transit networks, and
demand estimates are carried out for a system of about 700 regional zones (average
population -9000). Many elements of the demand models are treated in a simplified way. For
example, socio-economic data are input as zonal averages rather than as ranges that would
reflect the variability of the estimation data. Also, some models have been re-estimated
21. For example, In standard statistical terms, the Influence of non-work trip accessibility on trip generation and
auto ownership is supported by coefficients with the correct signs and ^statistics In the range [2.3| to |5.2| {for
sample suet of about 1300).
22. These developments might affect the precise form of accessibility in the auto ownership model, for example.
Transportation-Air Quality Analysis lasuat
January 1983
-------
P*g» 4-14
without accessibility feedback, in the interest of computational tractability.2® It would be fair
to characterize MTCFCAST as a model that achieves great detail in representing
infrastructure by reducing its detail in representing demand. (This is also true of other MPOs*
models.)
STEP reflects an application of the "pure" MTC model at the household level. It works with
a sample of households from the region, as large and as up-to-date as possible.34
Base level-of-servioe data are supplied exogenously for each zone* along with standard land
use and socio-economic Information. (This data ensemble has exactly the tame form as a
model estimation data set) STEP reads through the household sample, adding level-of-
service and land use data to each household record as necessary, and calculates alt of the
household's travel probabilities. Full model specifications are used, and the sampling
framework preserves the richness of the underlying distribution of population characteristics.1*
Household totals are expanded to represent the population as a whole, and summed in
various regional and subregional categories.
Any change in the population or in the transportation system that 1) can be represented in
terms of the variables of the MTC model and 2} can be associated with a specific geographic
area or grouping of households, Is possible to test with STEP by reprocessing the household
sample using the new data values. What STEP lacks is a detailed network representation
and traffic assignment component, meaning that changes in travel time resulting from changes
in demand must be calculated in another way. A simple routine for estimating changes in
level-of-service has been incorporated, but H is intended only as an approximation.37 It thus
would be fair to characterize STEP as a model that achieves great detail in representing
demand by reducing Ks detail in representing infrastructure.
23. Specifically, non-work ftp generation was de-coupled from damnation and mod* choice, not from lade of
statistical significance but from Ks ralativa unimportance to investment policy (as an off-peak phenomenon that
does not impinge on capacity requirements). Standard cross-dassification models ware substituted. The auto
ownership linkages to destination and mode choice were not altered.
24. New modets have been estimated for use with STEP, and these new models perform better statistically than
the MTCFCAST models. However, for consistency the MTCFCAST models were used in STEP for these
analyses.
25. Intenona! ievet-af-eervice tables typically would be obtained from large-scale model output and used as the
STEP base case.
26. For example, households in the lowest income decOa are addressed In STEP but loat In the higher zonal
average Incomes of MTCFCAST.
27. The simplified level of service model uses peak and off-peak travel times and base case demand estimates
to "calibrate" a Bureau of Public Roads (BPR) - type supply function for appropriate spatial groupings of trips (e.g.
trips In broadly-defined corridors). For each change in demand, the calibrated fandion can be used to compute a
new "equilibrium* in the corridor. This technique Is documented in:
Harvey, Grelg. (1983) Methodology for Incorporating Transportation System Effects Into Regional
Transportation Energy Demand Forecasts. Report prepared for tie California Energy Commisaion under
Contract No. 400-82-023. September.
Transportation-Air Quality Analystt lasuaa
January 1993
-------
Paga 4-15
A comparison between the available large-scale and microsimulaiion approaches suggests
that neither is ideal for the analysis of transportation pricing policy. The large-scale model is
too expensive to run in an exploratory mode, lacks some potentially important feedback loops,
and is incapable of capturing the distributional effects of pricing. The microsimulation model
does not allow pricing of individual links and supports disaggregation of outputs only where
sample size is adequate. On balance, H was decided to use the microsimulaiion model
(STEP) for pricing studies because of the importance accorded to equity considerations in
early Bay Area debates about pricing. The microsimulation methodology fits well with the
regional scope of policies addressed here* However, a more facility-specific approach to
pricing, such as the toll road proposals for Southern California, could be studied only with a
correctly-specified large-scale model.
Using STEP, each of the prototypical pricing concepts was studied in some detail. The
analysis results are summarized in Table 1, and key assumptions are described briefly here:
¦ Congestion Fees - Congestion was expressed in terms of the .ratio of peak to
uncongested travel times. Values of this ratio reflecting various levet-of-servica (LOS)
ratings were selected. For each LOS rating, STEP was run repeatedly to determine
the average per mile price that would have to be charged in each corridor to reduce
highway travel (and increase speed) enough to achieve the required travel time ratio.
All models through auto ownership and work trip distribution were exercised. An
additional correction was introduced for the peak spreading that would be induced by
high peak-period prices.38 No effort was made to account for residential and
employment allocation effects.
¦ Emissions fees - The annual smog-based registration fee was represented in two
ways. First, an age-based fee was calculated as a proxy for the true smog fee.®0
CARB data on average emissions and VMT by model year were obtained for a 1997
fleet. An informal survey of air pollution experts was conducted to estimate the
28. And It has ¦ secondary value, as wall. Microaimulation fundemerrtalty '* a survey tabulation technique
employing sophisticated data transforms. In the Bay Area studies, much use was made of the ready-at-hand
survey tabulation capability implicit in STEP. For example, Bay Ana surveys contain detailed vahide data, so
that exact usage patterns by model year or vahide type can be determined. Using STEP, these were correlated
with personal and household cheracteristics to yield useful information about low-Jncome dependence on old
vehicles.
29. Absent any information about the effect of prica on peak spreadng, we looked at the evidence on time
responses in the STEP household sample. Start times were determined for all workers beginning their Jobs
between 5 am and 11 am. The mean and standard deviation of worti start time were calculated for ranges of the
peak-to-eff-peek time ratio {1.0-1.0®; 1.1-1.18, ate.) The mean start time did not vary eignlficantly with the
peeing ratio (approximately 7:55 am in all cases). The standard deviation did not vary for values of the peaking
ratio up to 1.6. Above 1.8, the standard deviation rose dramatically (from about .8 to 1.1 hours). The obvious
interpretation is that peak spreading has two components: as congestion builds, workers change trip start times
but do not change work start times; when congestion reaches a critical point (apparently related to the presence
of LOS E/F conditions over much of the commute), workers also shift work start times.
The same process was assumed to occur with peak cod The price of the auto trip was converted to
equivalent time using the mode choice model's value of time. True peak spreading was introduced when 9m
combined time/cost measure exhibited a peek ratio greater than 1.6.
30. The work of Stedman and others has shown that vehicle age is an imperfect predictor of amissions rate.
Tranaportatlon-Alr Quality Analyala laaua*
January 1993
-------
Page 4-16
average annual cost of mobile source pollution in the Bay Area. A value of $500
million was selected for use in the analysis.'1 The fleet and pollution cost data were
translated into per mile estimates of pollution costs by model year. These costs then
were added to the annual auto ownership costs of each household according to the
makeup of the existing household fleet and STEP was run with the full set of models.
Because no possibility of substitution was allowed, a modest drop in the size of the
vehicle fleet was predicted, with the burden falling almost entirely on tow income
households. Limited reductions in emissions occurred, however, because most
households simply retained their older cars. A vehide type choioe model would be
required for developing a more accurate estimate of this shift
A second scenario was tested by simply assuming the emissions fee would result
in some portion of the vehicle fleet older than eight years being replaced by eight year
old vehicles fl.e., perfect substitution would occur). The STEP travel demand models
were not exercised in this case (i.e., STEP was used only for survey tabulation with
emissions calculations).
The results reported in Table 1 were for the first analysis scenario. It'Es quite likely
that this approach understates the true potential of an emissions fee, however. Even
a modest substitution under the second scenario produces higher estimated emissions
reductions.
¦ Parking fees - All worker drive-alone parking fees were set to a minimum of S3.00 per
day,*3 and worker transit fares in impacted districts were set to zero. STEP was run
with the full set of models to estimate the combined effect of parking and transit
incentives.
¦ Gasoline taxes • The effect of a gasoline tax was studied in a very simple way. The
auto operating cost per mile was increased under varying assumptions about improved
fleet mileage, and STEP was run with the full set of models. Figures reported in Table
1 assume fleet improvement to 40 mph, reflecting a substantial demand pull for more
efficient vehicles. If the gasoline tax really covered insurance, then it might not be so
easy to reduce the fee by purchasing a.more efficient vehide.*1 Under this scenario,
the effects of a gasoline tax might be substantially higher.
How well do these analyses stand up against the criteria outlined above? Simply put, the
STEP approach addresses: 1) the need for appropriate price, income, and time variables; 2)
the need for structural completeness and feedback through auto ownership; and 3) the
advantages of disaggregation. On the other hand, STEP: 1) does not address residential and
31. Estimate* of the total cost of air pollution in the region variad between $1 billion and $4 billion. The J500
million figure was eelected as a conservative estimate of what could be attributed to mobile aouroas.
32. Existing fees higher than S3.00 were not changed, and no fee was assumed for HOV users
33. Because aggregate insurance costs would be relatively independent of (or might even rise with) fuel
efficiency, the per mile tax would have to stay about the same in order to recover full insurance coats.
Tranaportatlon-AIr Quality Analyala laauaa
January 1903
-------
Paga 4-17
employment allocation14 or absolute regional growth; and 2) treats time of day in a very
simplified way. In addition, STEP does not resolve the philosophical concerns enumerated
here; it is unequivocally a model in the geographic/economic tradition of travel demand
analysis, albeit in a novel implementation framework. However, given the state of knowledge
and available models, K would be difficult to carry out a more thorough analysis of pricing
without integrating the full MTC demand models into a fully-detailed network model. Such an
effort was well beyond the scope of the original Bay Area and South Coast studies.*9
4.6 Analysis Approach for the South Coast
The South Coast analysis strategy dosely resembles the Bay Area approach. It relies on a
computer package resembling STEP, but does to in the absence a rich a tradition of data
collection and model development SCAG's regional demand model could not satisfy enough
of the criteria for pricing analysis to support independent development of a microsimulation
program for the South Coast. New model development was not feasible. In the end, it was
decided to adapt the MTC model to the South Coast by recalibrating its constant terms using
the 1876 SCAG home interview survey. The same survey then was reweighted to reflect
future conditions a closely as possible, and used (with other SCAG data) in constructing a
microsimulation program for the South Coast"
Transferring a model from another region Is not as bad as K sounds. Time and cost
coefficients tend to be reasonably stable from region-to-region, whereas constant terms do
not. And the approach used in the EDF study for the South Coast involved recalibrating
constant terms precisely to capture local conditions. Nevertheless, it would always be
preferable to use a "home-grown" model, if only for political reasons. As new model
development tends to produce more MTC-like models in other parts of the country, versions
of STEP should be crafted for each region.
4.7 Implementation
The Bay Area and South Coast studies also involved much discussion of the conditions that
might foster implementation of pricing.
The credibility of data and analysis was often cited as essential to building a coalition of
diverse interests.
34. Work with the POUS modal sugsersts that allocation effects would be femtted without tome relaxation of land
use constraints. Of course, many pricing advocates expect that congestion prices would produce poWcal
pressure for such a change.
35 A hybrid approach In which trip tables from MTCFCAST are adjusted according to Interim resutts from STEP
has been employed for TIP conformity studies in the Bay Area. This method requires further refinement, but
holds promise for pricing studies where fine-grained network characteristics are of critical importance.
36. Given this provenance for the South Coast model. It seems appropriate to add geographic transferability of
model coefficients to the list of epistamological tsaue*.
Transportation-Air Quality Analysis laauaa
January 1993
-------
Paga 4-18
The need for detailed information on "winners and losers" was also highlighted. While the
identification of winners and losers will occur partly through political channels, thera also is
a critical role for a model such as STEP, which has the capability to produce detailed
information about the distributional features of transportation policies. It is helpful to have a
solid means of corroborating or refuting claims of unfair treatment when they emerge through
the political process.
It became dear in both areas that a major portion of the stream of funds generated by pricing
would have to be dedicated to compensating the losers". This is important both because It
implies another cycle of modeling after the compensation structure Is in place, and because
an economically rigorous determination of the efficiency of a pricing scheme will depend
heavily on the uses to which revenues are put
A final strategic issue concerns the appropriate way to get started toward a comprehensive
pricing system. The assumption has been that pricing could start In a few places - the Bay
Bridge, or a new toll road in Orange County, for example - and build upon an initial positive
experience. Yet K seems obvious thai a few facilities dropped into an otherwise unpriced
network will behave differently than a more widely applied scheme - local distortions in route
choice might be quite significant, and the public might be led to the wrong conclusions. This
could argue for a different kind of gradualism In Implementation, with much earlier
implementation of more widespread congestion fees, Initially set at very low levels and only
gradually increased toward optimal levels.
TransportatJon-AIr Quality Analysis lasuas
January 1993
-------
CHAPTER 8: AN OVERVIEW OF RESEARCH AND DEVELOPMENT
NEEDS
Improvements in modeling practice will be needed in the near term, and much can be
accomplished by upgrading data bases and implementing available methods. Advances in
software and hardware should greatly aid the effort Such improvements will, however,
require a commitment of staff resources and funding for data and analyses. In addition,
applied research and development will be necessary to address remaining limitations and
shortcomings of modeling practice, and basic questions should be raised about longer term
research needs, the role of modeling, the possibilities for alternative paradigms, and the need
for institutional change. All of these matters are discussed in this chapter.
8.1 Selecting a Strategy for Model Improvements
Currently, the quality of models in practical use varies significantly, in the short term, efforts
might be directed toward bringing all metropolitan areas' models, and the data that support
them, up to acceptable levels. Improvements should be selected and prioritized based on the_
current capabilities and most pressing needs in each area, which will vary depending on
current and anticipated travel conditions, policy options of greatest concern, non-attainment
status, and resource availability.
As 9 startingj>oint. each urban area should be encouraged to maintain a network-based travel
forecast modal system which incorporates key phenomena in a model structure that is in
keeping with theoretical considerations and empirical evidence. This implies, for example,
that models at minimum should feed back travel times resulting Jrom the traffic assignment
step to the mode choice and trip distribution (and possibly, to the trip generation) steps, and
should be run to an approximate equilibrium. Model systems which omit such feedback loops
should be upgraded.
In addition, individual models should be upgraded, where necessary, to incorporate key
variables that are widely agreed to be strong determinants of travel behavior and that are
needed to analyze key policy options. For example, common shortcomings of models in
current use include: (1) no trip generation variables beyond auto ownership and income (e.g.,
household composition: workers per household); (2) inadequate representation of trip
attractions; (3) trip distribution models which omit transit and walking accessibility (needed in
areas where transit and walk modes are important); (4) lack of peaking information on trips
by type and market segment; (5) simplistic representation of socioeconomic variables affecting
Transportation-Air Quality Analyst* laauaa
January 1993
-------
Paga S-2
Improvements to address these shortcomings would be in order.
Among the variables that some areas have omitted from their models, and should add as
soon as possible, are: (1) hpmahoid in coma (a key variable that should appear wherever
cost appears); f2) marking shames and auto operating ^osts (without which analyses of
parking pricing strategies, congestion pricing, tod roadsTetcTCan only be done off-line); and
(3) the number of workers in the household (a key variable In ridesharing estimation).
For many areas, better models of land use allocation or residential and employment location
choice also would be appropriate. Hers, one of the~difficu!iies Is the political sensitivity of land
use forecasts. Local land use plans are rarely tempered by economic analyses of regional
and intra-regional development competitiveness; regional land use forecasts may challenge
both overall growth daims and individual localities1 hopes for a large share of the growth.
Many areas instead use politically-negotiated land use forecasts which are assumed not to
change regardless of infrastructure investments. Such practices may be politically pragmatic,
but they are not necessarily theoretically or empirically defensible. Indeed, in some areas
these practices are being challenged by outside groups who view them as ways to perpetuate
the status quo in transportation investment (and land development) policy. Overall, land use
modeling is likely to be a ticklish problem for many areas, requiring careful work with elected
officials and interest groups as well as on the models themselves.
Although it would be desirable for every urban area to quickly develop state-of-the-art
modeling practices, the reality Is that there will be wide variation in practioes among areas at
the start and undoubtedly for years to come, even if an aggressive program of improvements
is undertaken.3 Moreover, variation in practice is desirable, to match activities with context-
specific needs. For example, as discussed in previous chapters, practice might differ with:
(1) required transportation and Clean Air Act analysis activities, which vary with area size,
pollution types, and pollution severity; (2) the magnitude and location of anticipated growth
in the region; (3) travel characteristics (e.g., transit share); and (4) other related policy issues
of importance to the community (e.g., location of employment growth, housing affordabllity).
Determination of what analysis practices are appropriate for an urban area might be done by
agreement among interested parties. For example, before beginning transportation-air quality
analyses, agencies might negotiate the analysis approach with EPA, DOT, and perhaps other
concerned parties such as environmental groups. Agreements reached on the scope and
complexity of travel forecasting to be achieved over a specific time frame would be
documented in a strategic plan or a work program for improving data and analysis tools, in-
cluding a reasonable, negotiated schedule for implementing improvements. The agreement
might specify, for instance, that current practice Is the best that can be achieved over the next
six months, but within the next 18 months a specific set of improvements will be implemented,
2 The state of the art ateo wff change as research add* to the toiowledge base, anafysfe techniques are Improved,
and advances art mad* in computer hardware and software. In addition, changes In vehicle technology and in
transportation systems potentially could fransfomi fransportaton-air quality planning and modeling For example,
alternate fuels and electric vehicles would cfrastieally change emissions modeling. Intelligent vehicle highway systems
(IVHS) could transform network specification, route choice, time of day of fravel, and potentially many other factors.
The introduction of sophisticated congestion pacing also could require advances in tie state of the art of modeling
Tranaportation-AIr Quality Analyala laauaa
January 1993
-------
P*Q9 6-3
and over the longer term new data and models wilt be developed.
6.2 Data Need*
Models are only as good as the data on which they are based, and better data are a pressing
need In most urban areas. Part of the model enhancement effort therefore must be to
develop and maintain high quality data bases.
Data on current land uses and land use regulations, as well as land market information,
should be updated regularly. More specific and detailed data on economic and demographic
characteristics and changes would be useful in preparing population and Job forecasts,
improved network representation would in turn improve travel forecasts and impact analyses;
networks and their underlying data bases should represent all facilities down to arterials (and
in many instances, major collectors). Travel surveys should be done perhaps once a decade,
via household surveys of adequate size to support the detailed analyses contemplated. Data
from special purpose studies or from a smaller panel could be used to track changes and
provide interim updates.
Monitoring data will be particularly important in carrying out the growth tracking requirements
of recent legislation. As discussed in earlier chapters, iherlBSO CAA Amendments set forth
a number of requirements for data and analyses to which MPOs will need to respond over the
next few years. These include: (1) development of emissions inventories and forecasts; (2)
VMT, speed, vehicle emissions, and congestion checks; (3) TCM analyses and
implementation; and (4) determination of conformity of transportation plans, programs, and
projects with the State implementation Plan for air quality.
The monitoring requirements of the CAA imply the availability of detailed information on both
the highway network itself and network operation (link volumes, vehicle mixes, speeds by time
of day, locations of high acceleration, locations of parked vehicles, trip start information).
Many areas will need to update and enrich this information. Among other Hems, ipeed
validation and congestion assessments should be done through a robust sample of floating
car studies; VMT should be checked by comparing model outputs to traffic counts.
Currently, the accuracy required for VMT forecasts and comparisons to be meaningful for air
quality planning purposes is not achieved in practice. The data base for monitoring VMT is
inadequate in most areas, with counts off the Interstate system and in new growth areas
particularly problematic. A monitoring program that provides a reliable time series of data
collected in a consistent way is needed. The HPMS data base has been used for needs
studies but in many urban areas it contains too few samples to accurately monitor VMT for
the purposes called for in the new legislation. Improvements to the HPMS data base,
including more and better traffic counts, are being planned and should be an important step
toward better monitoring capabilities. Most urban areas wilt want to compare the findings of
these counts with model outputs and other sources of information, if available, since large
discrepancies could prove to be problematic.
Transportation-Air Quality Analyala laauaa
January 1993
-------
Paga 5-4
As a second example, many areas have assumed speeds not exc*edinQ toga! speed fern its
(posted, or 55 mph on urban Interstate*) tven on off-peak networks; In most areas this would
fail a "reality check". (The Impact on emissions estimates is uncertain; while high freeway
speeds generally result in increased emissions, higher arterial speeds could reduce emissions
estimates.)
In addition, many high growth areas significantly underestimated population growth in the 70s
and '80s. Such inaccuracies if repeated in the future would pose major problems for
conformity findings and SIP attainment demonstrations.
Better data will take 2-3 years to collect, and even more time « as much as 5-10 years - will
fee needed where time series data or panel data are required. Also, data collection activities,
and in particular surveys, are very expensive; the case for them will need to be clearly
articulated.
8.3 Other MPO Resource Needs
In addition to good data, many MPOs will need additional binding and staff in order to carry
out enlarged responsibilities for data collection, analysis and forecasting, in many areas this
will mean overcoming some significant resource constraints, although new funding for
planning available through the Intermodal Surface Transportation Efficiency Act (ISTEA) may
ease this problem.
One issue is that models are neither user friendly from a computer applications perspective
nor simple from an analytical perspective. Because the computer skills needed to run
transportation models are extensive, many areas have assigned these tasks to programmers
and data processing staff, many of whom have had little training in transportation systems
analysis. Consequently these staff are not well prepared to deal with questions and concerns
about the theoretical and empirical validity of the models. Training for the computer staff to
improve their understanding of the models they are running would be helpful, but should not
be considered an adequate substitute for an expert modeling staff. Many MPOs may need
to add staff positions to handle CAA (and ISTEA) analysis obligations, and to supplement staff
with consultant contracts.
Another issue has to do with the need to "translate" models and analyses for non-expert
decision-makers and reviewers. Models can be difficult to explain to elected officials whose
decisions they are intended to support, but increasingly these officials want to know how
reliable the model forecasts are and how they were developed. Environmental groups and
community groups also seek a dear understanding of the models and their strengths and
weaknesses. Providing this information will necessitate resources specifically for this purpose.
Lack of documentation and inadequate documentation could be a barrier. In past years
documentation of data, models, and analyses has been under-funded and has fallen by the
wayside when deadlines are tight However, lack of documentation makes it more difficult to
repeat analyses or monitor trends, and may be a serious problem if analyses are challenged.
Greater expenditures in this area could have an important payoff.
Tranaportatlon-Alr Quality Anatyala laauaa
January 1993
-------
Paga 5-5
Seven! issues are likely to arise concerning documentation. First, documentation for some
software is provided only to registered owners. However, other groups may wish to examine
the software in detail. This could create difficulties for the agency, and possibly necessitate
the preparation of extensive "model reviews" to explain the details of the programs to
outsiders.
A related issue is that, in some cases, documentation is of the "how to run the mode!" variety
rather than "how the models work". Someone may need to prepare a description of the latter.
Documentation and explanation of TCM analyses are particularly difficult when a variety of
methods in addition to the regional modeling system are used to estimate impacts, then folded
back into the overall analysis. Nevertheless, having a dear explanation of the analysis may
prove to be key in building support for TCMs.
6.4 Hardware and Software Advances: An Opportunity
Hardware and software advances should speed up and lower the cost of many of the analysis
steps, and at the same time should support far more sophisticated analyses than have been
possible in the past On the software and data base management side, for example, TIGER
(Topological^ Integrated Geographic Encoding and Referencing) files - digitized block
boundaries or segments - are now available from the Census Bureau and permit block-level
data (of whatever variety is coded, Census and other) to be aggregated to any zone system
desired through the use of a GIS system. Some areas have already begun to use GIS to
encode such additional information as tax parcel data, structures data, zoning, land use, slope
and soils, environmental conditions, sidewalk and bike facility inventories, conditions of
approval including traffic mitigation requirements, crime rates, and many other factors, in
addition to the Census data on housing, trade, employment, and the fike. These flexible,
extensive, integrated data bases and data management tools would support advanced
modeling, but only if the advantages of the advances are recognized and seized. (This may
require R&D sponsored by federal agencies or consortia of MFOs.)
Computer hardware also has greatly increased modeling capabilities, with desktop models
now superior to the mainframes of a decade ago and work station versions capable of running
most models quickly and efficiently. Unfortunately, the programming of many irvuse models
and model frameworks does not always take full advantage of these advances, and hence
loses some of the benefits of hardware gains. Re-programming may be an important option.
Hardware and software for data collection and analysis is a third area where significant strides
have been made and more are expected in short order. Automatic data collection via
roadside markers or roadbed counters, traffic signal detectors, speed sensors, and ticketing
databases are but some of the available methods that have yet to be fully exploited in most
areas.
TnnaportatJon'AIr Quality Analyala laauaa
January 1993
-------
Paga S-0
6.6 Research Priorities
8.6.1 Applied Research and Development
Significant improvements in practice could be accomplished through more widespread
implementation of advanced methods that already are in use in some areas. However, more
basic improvements will require applied research and development Such R&D might be
supported with federal funding or might be undertaken by MPOs with funds drawn from other
(perhaps local or foundation) sources.
Possible R&D topics include the following:
¦ Review and Enhancement ef Location and Land Use Models: Identify key
variables and relationships which must be present in land use models, and their
relative contribution to the quality of estimates. Assess the sensitivity of model results
to variations in major input variables, as well as the relative accuracy of each model
component. Evaluate whether simple land use models that are easy to apply and
require low levels of input data (in contrast to available procedures that are highly
complex and data Intensive) could perform acceptably. Examine how local plans and
economic base models might be integrated into a realistic land use forecasting pro-
cess. Assess the leads and lags which occur in transportation-land use relationships,
and how they might be taken into account Explore how to handle growth distribution
in areas where new construction is not the key or sole issue, i.e., shifts in the location
of activity in existing buildings, or which buildings will fill up first in areas where office
space supply will exceed demand for some time.
¦ Zoning Controls and Urban Design: Identify and rank small scale urban design
options that could have important impacts on travel patterns, and develop methods for
analyzing these measures. Assess the role of zoning controls in land use forecasts
and in transportation policy implementation, and develop methods for analyzing such
policies as zoning for higher density around transit stations, or the use of mixed use,
high and medium density zoning to create less auto-dependent communities.
¦ Data: Identify and evaluate data sources and surrogates which could reduce high
data costs for model development, forecasting, and monitoring. Assess ways to make
better use of available data such as en-board transit surveys and traffic signal data
bases. Assess panel data applications and evaluate whether the near-term results
justify the added expenses.
¦ Networfc Models: Develop better methods for estimating travel times as a function
of network congestion, accounting for speed changes resulting from shifts in route
choice, time of travel, and growth impacts. Evaluate methods for accounting for the
impacts of traffic incidents (non-recurring congestion). Identify ways to achieve more
detail and comprehensiveness in network specifications and/or roadway classification
schemes, and to assure consistency between, or adjust for inconsistencies in, network
specifications for periodic conformity determinations.
Tranaportat/on-Afr Quality Analyala laauaa
January 1993
-------
Paga 5-7
m Socioeconomic and Lifestyle Issues: Assess the importance as travel determinants
of such variables as age, sex, race, ethnicity, occupation, and household structure.
Assess the role of auto ownership as a variable in models, considering both the areas
where the number of vehicles equals or exceeds the number of drivers and the experi-
ence of very dense areas (e.g., Manhattan) where the correlation between rising auto
ownership and trip making is weak.
¦ Trip Distribution Models: Assess the reasons that K-factors historically have been
so important in calibrating trip distribution models, and evaluate approaches that might
improve model fit wfth less dependence on these factors. Assess the performance
and treatability of formal nested models through destination choice (e.g., with the
expected mode choice utilities forming the accessibility variables).
¦ Walk Mode: Evaluate methods to represent walking as a travel mode and to model
strategies that alter pedestrian facilities and amenities.
¦ Time-of-Oay: Develop behavioral models of the choice of time of travel.
¦ TCM Effectiveness: Assess the accuracy of methods which apply findings from other
urban areas, sometimes from a very limited number of studies, to estimate the
effectiveness of TCMs.
a Supplemental Analyses: Assess the role that can be played by "extra modef
analysis toots for TCMs, such as spreadsheet tools based on empirical evidence of
effectiveness. This is especially an issue for commercially available analysts tools
which have not been fully evaluated by a disinterested third party (other than clients
and sponsors, e.g.), and for which detailed documentation is not available to the
general public.
m Vehicle Emissions Factors: Develop more accurate vehicle emissions estimation
techniques, particularly to account for speeds above 55 mph and to reflect actual
driver behavior (frequent accelerations, etc.)
¦ Software Improvements: Develop improved software, including both a more support-
ive software environment for transportation - land use modeling and a more powerful
"post-processor" for emissions calculations.
¦ Model Precision and Accuracy: Assess the current precision and accuracy of data
and models, identify sources of uncertainty, and evaluate how these conditions may
change in light of CAA and ISTEA requirements. Develop approaches which could
improve precision and accuracy and reduce uncertainty.
5.8.2 Basic Research Needs
In the longer run, research to provide a deeper understanding of the regional transportation-
land use system is needed. Here, the controversy over the impacts of infrastructure
TrantporUUon-Alr Quality Analysts laauaa
January 1993
-------
Pag* M
development will be utad to illustrate the issues and lay the basis for observations of a more
general nature. "
As noted earlier, debate over the land development impacts of transportation investments,
•specially in highways, has re-emerged in a number of uitoan areas. One side argues that
transportation investments are major growth shapers; the other side argues that transportation
investments art of limited impact except in special, unusual circumstances. How might
research cast more light cm the issues?
Some have proposed a "quick fix" research strategy which attempts to measure the actual
impacts of a sample of infrastructure improvements, through measurements of changes in
traffic flow and similar aggregate measures of system use and/or case analyses of land use
shifts. The problem is that the changes are broader than that, reflecting a web of interactions
among mode choice, route choice, time of day of travel, destination choice, location choioe,
and overall growth of the region. Such complexity cannot be revealed by simple observation;
more fundamental analysis of travel behavior and location decision-making is needed.
Moreover, many of the phenomena at issue occur gradually over long periods, and thus be-
come difficult to separate from other trends. Hence, case studies of new infrastructure outside
the context of high-quality longitudinal data sets are unlikely to yield definitive results.
Several large metropolitan areas - notably the Bay Area and Seattle - have initiated and
begun to use the data from panels, but their efforts are hampered by a lack of adequate
funds. One reason that funds have been difficult to come by is that H is hard to show a dear,
Immediate payoff from the substantial investment needed. Ironically, then, research on the
fundamental social and behavioral effects of infrastructure is stymied by the lack of data from
carefully maintained long-term panels in a number of urban settings, and such panels are
stymied in pari because their results will be long in coming.
Research is also limited by the constraints of available models. Most operational models
were developed under conditions of. 1) scarce computational resources, requiring the most
parsimonious problem definitions possible; 2} a limited pool of professionals with the ability
to reliably apply sophisticated models; and 3) a focus on a specific set of applications for
which a particular set of simplifications seemed appropriate. Conditions have changed, the
models have not, and in most cases the state-of-the-art does not satisfy the need for improve-
ment
Although much can be done to improve model performance within the conventional "four-step'"
paradigm, a more fundamental examination of the issues Is in order. For some time the travel
behavior research community has recognized a need to rethink the basic paradigm of travel
demand analysis in light of three decades of advances in the cognitive sciences, in econom-
ics, and in computational capabilities. The emerging theory might be described as activity
participation in the face of time and monetary constraints. The implications for modeling are
substantial: for example, models might focus on activities, with travel consumption as a by-
product This creates pressure for research on virtually every element of travel behavior.
If an understanding of the urban activity system is the goal, researchers and research
iponsors must acknowledge the inherent complexity of the problem, which could be compared
Tranaportatlon-Alr Quality Analysta /saves
January 1993
-------
Pag* 6-9
wtth research on global warming or human cognition. Other discipline* facing Inherently
complex problems have developed a research style that emphasizes a detailed understanding
of specific, isolatable phenomena, together with computer simulation of feedback and similar
complex Interactions. In these disciplines, alternate theories of system structure are tasted
by evaluating the performance of their respective simulation models.
Similar research styles will be needed in transportation to explore, develop, and implement
analysis approaches reflecting fundamental changes in knowledge and method. Research
sponsors and their user-clients will need to tolerate work that may not have any immediate
applications - tome of which will not prove out - If basic advanots are to be made. The
emphasis on pragmatic investigations producing quick answers to pressing issues is under*
standable, but H should not be the only kind of research. At least some funding should be
directed toward a broader disciplinary scope and a more basic, deeper, set of questions for
transportation research.
One issue is whether current institutions are capable of supporting activities which may
challenge established beliefs and ways of doing things. Research sponsorship is one matter;
put In broader terms, the issue may well be whether current institutions permit a search for
improved mobility along many dimensions. Provisions of the ISTEA challenge urban areas
to begin such a search. Some institutional arrangements and assignments of responsibility
may be better suited to the task than others, and this too would be a valuable topic for
investigation.
A decision-making paradigm that is more informed than simple "fair-share* distribution of
public capital, yet is less dependent on deterministic "knowledge of the future" than current
rational planning approaches, would be another area for attention. Modeling assumes an
ability to forecast the future that may not be realistic or necessary. Scenario testing
approaches suggest an alternate use of modeling as a means of exploring policy implications;
it gives explicit recognition to the "if-then" character of the models, darifies the assumptions
on which they rest, and provides opportunities for the introduction of qualitative information
into forecasts. Control theory suggests another direction: data from monitoring could be used
to make adjustments in operation and to identify needed improvements, perhaps selecting
from a set of responses previously agreed upon in contingency plans. A broader look at such
options might uncover new directions for transportation planning, policy, and institutions.
In the debate over the development impacts of transportation, K may be the case that both
sides are right. Within the limited domain of current land use regulations, current pricing
practices, current technology, and current financial resources, many (or most) congestion-
relieving highway investments may well improve system performance. But a different social
optimum may exist when the current constraints are relaxed. (Imagine the sort of "bubbly"
functional surface one might expect of a non-linear, multivariate, mixed-behavior, time-
dependent system.) Moreover, present levels of public expenditures on transportation are
insignificant in comparison with the aggregate of private expenditures, and K may be
unrealistic to expect the public sector to have a strong influence on patterns of mobility under
such circumstances. But ways to exert broad influence over private decisions are well known
and available for use, if the public will to do so is present A wider consideration of options
might identify new approaches and open up new opportunities for advancement
Tranaportatlon-AIr Quality Analyala /aauea
January 1983
-------
Paga 5-10
t.C Putting Modeling Into Perspective
As the preceding section Indicates, transportation and its interactions with land use and the
environment are highly complex phenomena for which substantial additional research, both
basic and applied, would be appropriate. Research into these matters should reveal ways to
improve models and analyses and their utility in decision-making. Nevertheless, K must be
kept In mind that models are tools; they need to be interpreted with car* and not expected
to "make decisions." Moreover expectations for models must be tempered by practical
realities including time and cost considerations. Thus a few words of caution are in order
here.
The review of current modeling practices and their strengths and weaknesses raises questions
about the requirements for modeling promulgated by federal and state transportation
agencies, especially as these requirements are combined with those implied in the Clean Air
Act. For example, current transportation planning regulations vary planning and analysis
requirements with population of the metropolitan area. More detailed and demanding require-
ments apply to the larger urban areas. From an air quality perspective, however, the size of
the metropolitan area is not necessarily a good indicator of the severity of the pollution prob-
lem^} or of the complexity of the issues faced in air quality (or transportation) planning. Thus,
small and medium-sized metropolitan areas might need to develop better planning and analy-
sis capabilities than otherwise would be expected, In order to respond to air quality planning
needs - or to the transportation and land use challenges of the region.
There are concerns that the technical and financial capabilities to support extensive data
collection, model development, and model application are largely lacking among the smaller
metropolitan areas, whereas large urban areas have greater resources to carry out
sophisticated monitoring and analysis efforts. Clearly, there are exceptions In both directions.
For example, data and modets may be relatively up-to-date and sound in the urban areas with
recent experience in transit alternatives analysis, regardless of area size (although some
would challenge this claim). Conversely, some of the larger urban areas have not paid
attention to modeling for a number of years and their practices may actually have declined
in quality and sophistication. Size, in short, may be only one indicator of modeling capacity,
and a rough one at that
Growth rates have not been considered as a factor in setting modeling requirements, but they
may be an important indicator of needs. Very fast growth areas may need to develop
sophisticated data collection, monitoring, and analysis capabilities, regardless of their pollution
or urban size classification, in order to permit them to track changes In travel patterns, trip
making, and VMT more accurately. Conversely, less complicated methods might suffice in
low growth areas, although even there, intraregional shifts from city to suburb and from
downtown to outlying commercial areas are often large and important
Overall, despite the difficulties, H seems reasonable to recommend that urban areas large and
small should be encouraged to improve their data bases and enhance their modeling capa-
bilities, but common sense must be exercised in setting expectations. Smaller areas may not
be able to afford locally-based research and extensive methodological innovation, but their
Tranaportaiion-AIr Quality Analyala laauaa
January 1993
-------
Paga 5-ff
travel demand models nevertheless should be good examples of the application of state-of-
the-practice models (i.e., trip generation, trip distribution, mode choice, and traffic assignment,
with feedback through trip distribution.) Their land use forecasts should reflect key data on
housing and employment trends and forecasts. Their network models should be checked
against ground counts. Areas where fast growth Is occurring or where air pollution problems
are severe may need to enhance specific models to address those Issues.
Where tn-house resources are lacking, urban areas typically have turned to consultants for
model development and calibration, with local staff taking over the applications In some cases
and working with the consultants on applications In others, Developments In a few states
suggest an alternate approach: cooperative agreements in which the state provides hardware,
software, standardized model structure, and technical support; regional agencies and local
governments provide data (sometimes, however, with state funding); and local universities
provide training for staff and ongoing technical assistance. Florida and Texas provide
examples of successful state-regional programs of this sort
Transportation modeling regulations and needs are but one part of a broader set of issues
concerning data and analysis requirements with which planning agencies must contend,
however. A second set of issues stems from the transportation - air quality planning and
analysis requirements set forth in the 1090 Clean Air Act Amendments. As discussed at
some length in earlier chapters, this highly complex legislation sets different deadlines for
attainment of national ambient air quality standards (NAAQS) depending on the pollutant (CO
or ozone) and the severity of the violation, resulting in a patchwork of due dates and target
analysis years both among metropolitan areas and, probably more seriously, within particular
areas. Moreover, increasingly stringent requirements for planning, monitoring, and control
apply to each classification. All but areas classified Marginal must reduce VOCs
(hydrocarbons) by at least 15 percent within six years. Although only those areas classified
Severe or worse are required to identify and adopt TCMs, Serious non-attainment areas must
adopt TCMs if emissions prove to be underestimated in the SIP, and even Moderate areas
may need TCMs to meet the standards by-the deadlines. The differing requirements match
mandated actions to problems and reflect the greater difficulty of achieving the air standards
in the more severely polluted areas, and hence are a pragmatic response; but they also
create a much more complex and varied set of requirements than previously applied.
Conformity provisions also are significantly expanded in the 1890 Amendments, and require
growth rates to be taken into account Furthermore, changes In VMT must be monitored,
reported, and taken into account in SIPs and transportation control plans. This may raise
particular difficulties in fast growing areas.
Overall, the complexity and comprehensiveness of the requirements would appear to require
extensive data collection, monitoring, and modeling.
From a modeling perspective, two issues arise. One is that the different dean air milestones
•nd deadlines for attainment become target years for transportation planning and analysis,
but these years do not necessarily coincide with available transportation data, forecasts, or
planning horizons for the region. This has meant that analysts must extrapolate or interpolate
their transportation forecasts. Since transportation plans and programs rarely are precise
TranaportatJon-AIr Quality Anaiyaia laauaa
January 19$$
-------
Paga §-12
about the implementation year for particular projects and policies (arid such precision, tf
imposed, would probably no! b« a ecu rata, especially for actions to be Implemented some
years In the future), this atap introduces numerous assumptions and approximations. "Tbe
results are than treated as "givens*, however, In calculating estimated amissions.
A second and perhaps larger issue concerns the requirements for amissions and air quality
modeling and the way those models interface wfth transportation models. In particular,
emissions and air quality models require as input hourly volumes by link, plus speed and fleet
mix estimates. But transportation models produce much less spedfic output Henca, post-
processing of the transportation output must be ttone, and again rests upon numerous
assumptions. Moreover many transportation models constrain freeway speeds to the legal
limits (i.e., 55 MPH), and hence Introduce inaccuracies wfth potential repercussions throughout
the travel forecasts as watl as In the amissions estimates.
Overall, the current precision and accuracy of data and models hardly seem to be in keeping
with the expectations for them implicit in air quality planning and modeling requirements.
While improvements in method could partly narrow the gap, a more fundamental assessment
of the uses and limitations of transportation forecasts might be in order.
Despite these concerns, the possibilities for immense Improvements are many. Today, there
is reawakened interest in models and their performance, new mandates for analysis,
legislative changes that open up important opportunities for institutional development,
advances in a variety of disciplines which could be brought to bear on transportation prob-
lems, and funding to support both short-term and longer-term research and development
Both improved planning, modeling and analysis practices and a richer understanding of
underlying phenomena should be the sought-after results.
TranaportatJon-AIr Quality Anafyaia laaua*
January 1993
-------
CHAPTER 6: IMPROVED TRAVEL SURVEY DATA
6.1. INTRODUCTION
Regional travel models have been the primary analysis tool for transportation planning and
for inputs to air quality planning. These models are based on data from large-scale
household travel surveys, which are expensive to collect. Transportation and air quality
analysis capabilities therefore depend ultimately on the quality of household travel survey
data.
Because household travel survey data are expensive to collect, many regions have not had
the resources to collect new data, and have had to rely on out-of-date data sets. It has also
become evident to transportation practitioners that current transportation and air quality
analysis needs have severely tested the capabilities of existing travel models; new modeling
practices may be required, which will in turn require collecting new household travel survey
data. In the meantime, existing household travel survey data contain information that can
be used to extend analysis capabilities beyond that provided by regional travel models.
New types of surveys are being developed that will enable much more sophisticated types
of analyses than have been possible with existing data sets and models. This paper
presents a review of current household travel survey practice, some examples of additional
uses of household travel survey data, and a look ahead at new directions in household travel
surveys.
Existing regional travel models are based on household travel data sets, many of which are
20 or more years old. There are concerns that older household travel survey databases may
not accurately represent travel behavior characteristics of the current population. Many
economic demographic changes have taken place over the last 20 years that call into
question the validity of using travel models based on old data sets. These include the
following:
• Real Income has not grown — and has declined for some groups — during the
1980s.
• Family size and structure have changed In particular, the proportion of households
with more than one worker has increased significantly in the last 20 yews,
corresponding to the Increase in labor force participation rate for women. Single-
person households have also increased in proportion to the general population.
• Travel behavior characteristics have been assumed to remain constant over time (trip
generation rates, trip length frequency distribution, etc.). But this is an untested
assumption that Is questionable given the changes In travel supply In the past 20
years (e.g., increased congestion, Increasing implementation of HOV facilities, trip
reduction programs).
Transportation-Air Quality Analysis Issue*
January 1993
-------
Pag9 6-2
There are further needs for new household travel data collection efforts. Existing regional
travel models were originally developed In the 1950s and 1960s to answer the major policy
questions of the time: where additional road capacity should be provided. Current
transportation policy issues — e.g., new rail starts, congestion management, air quality
management — have stretched the capabilities of existing travel models. The 1990 Clean
Air Act Amendments (CAAA) will place even more stringent requirements for transportation
and air quality analysis on regional planning agencies. There Is a broad consensus among
regional transportation planning agencies that current analysis practices are inadequate to
meet the new needs for transportation and air quality planning In the 1990s and for the next
century.1 ft is therefore likely that regional agencies will need to collect new household
travel data to support the development of new transportation and air quality analysis
practices.
Their high cost has deterred many regions from carrying out new household travel surveys.
Hence, these regions continue to rely on old, and possibly out-of-date, data sets as the
basis of their transportation and air quality analyses. Many regional travel databases are
therefore inadequate to support current analyses, let alone to provide data for developing
new types of travel models.
Furthermore, the transportation planning profession typically regards household travel survey
data only as inputs to the development of a four-step model. By restricting the use these
data to this purpose, transportation planners Ignore a significant amount of useful
Information that Is present In the data Hence, there Is an additional concern that travel data
are not being fully used.
f Experience from the past 20 years suggests that household travel surveys can be conducted
much more efficiently than they were In the 1950s and 1960s. Advances In survey
techniques have also led to a growing body of practice on the collection of longitudinal data
on travel behavior, which offer the potential for development of better behavioral travel
models.
This paper presents an overview of the collection and use of household travel survey data.
The following section reviews household travel survey data collection experience, focusing
on recent surveys in the San Francisco Bay Area; the discussion includes a summary of
lessons learned from survey experience.2 Section 3 discusses the use of household travel
survey data beyond that of estimating traditional travel demand models. Section 4 looks at
new advances in household travel survey data collection, with an emphasis on longitudinal
data
6.2. REVIEW OF HOUSEHOLD TRAVEL SURVEY DATA COLLECTION EXPERIENCE
1. Greig Harvey and Elizabeth Deakin, "Toward Improved Regional Transportation Modeling Practice,*, working
paper prepared (or National Association of Regional Councils, Washington, DC, 1992.
2. These surveys wera selected because they were successful (good response rates, good quality data sets,
etc.), and there is thorough documentation on the design and conduct of these surveys.
Transportation-Air Quality Analysis Ittuas
January 1993
-------
Paga 6-3
6.2.1 Early Surveys
Household travel survey data were originally collected through home interview surveys, as
part of the original large-scale regional transportation studies in the 1950s and 1960s (e.g.,
Chicago Area Transportation Study, Detroit Study, Bay Area Transportation Study). These
surveys typically had large sample sizes and high cost; for example, the 1965 survey for the
Bay Area Transportation Study covered 30,000 households at a cost of over $217 per
household in current dollars.®
These home Interview surveys set the tone for alt household travel surveys slnoe then. The
surveys gathered information on characteristics of the household (e.g., number of persons,
dwelling unit type, number of vehicles, income), persons In the household (e.g., age, sex,
workplace location), and trips made by each person in the household on a designated
travel day* (e.g., origin and destination times, activities, and locations; travel mode).
Transportation planners used the survey data almost exclusively for estimation of large-scale
regional travel demand models, but tended to ignore other potential uses of the data
These surveys were the first of their kind, and provided the first experience with collecting
household travel data. But they suffered from a number of disadvantages, chief of which
was their large cost. Another disadvantage was their susceptibility to Interviewer bias.
Because Interviewers were reluctant to go Into 'rough* areas, minorities and low-income
persons were undenrepresented in home interview surveys.
The high cost of the home Interview survey method made regional agencies reluctant to
undertake further data collection to update the data sets. As a result, in many regions, these
original data sets have remained the only source of household travel data for the past 15 -
25 years.
6.2.2 Recent Surveys
As the original household travel survey data sets grew older, regional planning agencies
became increasingly concerned that the data were becoming out of date. But the high cost
of conducting a home Interview survey deterred follow-up data collection efforts.
The increasing age of the data sets — hence, the need to collect additional data — and the
high cost of home interview surveys led some regional planning agencies to consider
gathering household travel data by means of telephone surveys. Telephone surveys have
a significant advantage over home interview surveys because their cost was significantly
lower; furthermore, because interviewers would not have to personally visit households,
Interviewer bias resulting from reluctance to visit "rough* areas was not a factor. There are,
however, some potential drawbacks:
• The sampling frame excludes a priori those households without telephones.
3. Charles L Purvis, The San Francisco Bay Area Household Pan#! Survey: A Response to Claan Air and
Mobility Initiatives,* paper presented at the Firat U.S. Conference on Panels for Transportation Planning, Lake
Arrowhead, California, October 1&92
Transportation-Air Quality Analyslt fssues
January 19 93
-------
Paga 6-4
• Unless precautions are taken, the sample may be biased against households with
unlisted numbers.
• K the sample Is not carefully drawn to reduoe the chance of geographic bias, the
geographic distribution of the sample may not match that of the population as a
whole.
• A reliable mechanism must be established for collecting trip data from persons In the
household. Simply asking persons to recollect their travel on a specified day will
result In missing trips, especially nonwork trips.
These potential drawbacks can be overcome by proper design of the survey. Chief among
the key features of a successful telephone survey are the following:
• Use of a valid technique for drawing a random sample of residential telephone
numbers.
• Use of a two-stage Interview wfth diaries for recording travel. During the Initial
contact, information Is secured from the household on characteristics of the
household and persons In the household; at this time the travel day is established
and a follow-up Interview to collect travel information is scheduled. The household
is then mailed a set of diaries for recording trips. In the second Interview, each
person in the household is interviewed about his trips on the travel day; alternatively
the diaries can simply be mailed in.
In 1i80, Caltrans conducted a travel -survey in the San Francisco Bay Area of 2,000
households using a two-stage telephone survey. The survey sample was randomly drawn
from a reverse-address telephone book, with controls on the sample to avoid geographic
bias. The survey was judged by Caltrans to be successful because it was carried out at a
much lower cost than a traditional home interview survey and the response rate was
sufficiently high: over 50%. Hence, the survey demonstrated the validity of a telephone
survey for gathering household travel data
At the same time, the Metropolitan Transportation Commission (MTC) was planning a new
household survey to update Its 1965 database. The survey design and questionnaire format
were based on those of the 1980 Caltrans survey. The survey collected data from 7,200
households at a cost of about $70 per completed interview *
The main difference between the MTC and the Caltrans surveys was that the MTC survey
used directory-based random-digit dialing to obtain the sample; at the end of the survey,
approximately 10 numbers were drawn from the reverse telephone directory to Improve the
4. Crain & Associate#. Thm 1981 Bay Ana Trtvl Sunny, report prepared for the Metropolitan Transportation
Commission, Berkeley, California, July 1981. See alio Marilyn M. Reynold*, S yd well M. Flynn, and David B.
Reinke, The 1901 San Francisco Bay Area Travel Survey,* Transportation Rese&rch Record 1220,
Transportation Research Board, National Research Council, Washington, D.C., 1982, pp. 51-58. The oost is In
1990 dollars.
Transportation-Air Quality Analysis Issuas
January 19 93
-------
Page 6-5
geographic coverage of the sample.8 As will be discussed below, this difference had
important implications for the validity of the sample. A two-stage interview process was
used as follows:
• The initial contact with the household was used to explain the purpose of the survey,
obtain the cooperation of the household, and gather Information about the household
and Its occupants (see below for a list of the specific information obtained). At this
time a specific travel day was designated for the household. Household occupants
were then sent a set of travel diaries on which to record their trips and Instructions
for filling the diaries.
• The second contact was scheduled for one or two days after the travel day. Each
household member of about 10 years of age and over was interviewed to get
information on his or her trips for the travel day; travel information was gathered by
as many follow-up callbacks as necessary to contact every household member. In
some instances, households were contacted during the coding process if trip origin
or destination locations could not be coded as originally given to the interviewers.
The survey was carried out over a three-month period in the Spring 1981. Interviews were
completed at 7,200 households at a cost of about $70 per household. The overall response
rate (number of completed Interviews divided by number of households contacted) was 70
per cent.
The success of this survey led MTC to plan a new survey coincident with the 1990 Census.
As carried out, this survey had three subsamples:
• Single-day trip data were collected from 9,000 households.
e Multi-day trip data were collected from 1,200 households.
• An additional, special-purpose sample of 1,100 households was funded by the San
Francisco Bay Area Rapid Transit District (BART); this sample was divided roughly
5. In directory-based random dialing uMd for the 1981 Bay Area Travel Survey, the following procedure was
followed.
1. Select residential telephone number* at random from the telephone book.
2. For each number selected, add one to the number.
3. tf the number it a working residential telephone number, atlempt to contact the household at least 8
times at different times of the day. different days of the week, and weekdays and weekends.
4. II the household is contacted, attempt to recruit the household for the Interview. If the household
agrees to participate, conduct the interview.
5. If a successful interview cannot be conducted at the number (nonworking number, nonresidential
number, cannot establish contact, household refuses to participate, etc.), add one to the number and
go back to Step 3.
Transportatlon-Alr Quality Analysis Issues
January 1993
-------
P*g» 6-6
evenly between BART users and households considered as part of BARFs "latenf
market (households who could potentially use BART, but didn't).8
As an incentive to participate, households who were asked to maintain multi-day trip diaries
were given an Incentive payment of $10.
The questionnaire for this survey was based on those used In the earlier Bay Area surveys,
but questions were added to gain additional information on household and job location
dynamics, and on other factors that could influence travel behavior. Among these were the
following:
• Households whose length of residence at the current address was less than 5 years
were asked their previous city of residence.
• Persons who had worked at their work location for less than 5 years were asked for
the city in which their previous workplace was located.
• Workers were asked about how much flexibility they had in their work schedules.
This survey experienced a somewhat lower response rate (about 60%) than the earlier
surveys. The cost of the survey was about $70 per household for single-day trip diaries, and
about $120 per household (including Incentive) for multi-day trip diaries.
6.2.3 Lessons from Recent Surveys
Recent experience in the collection and use of household travel survey data has provided
some lessons on what constitutes good practice for household travel surveys. These are
discussed in the following areas: information to be gathered, sampling, response rates, and
survey conduct.
Information from Survey
The main aim of a household travel survey has, up until now, been the collection of data for
use in estimating travel demand models; Table 1 lists a typical set of data from a household
travel survey that would be used to develop a standard set of models in the four-step"
process. But this list represents only a part of the data contained in a typical household
travel survey. As discussed in the next section, there are other potential uses of these data
that may over time be more important than modeling. Moreover, the survey should gather
enough information about the household, its members, and Its travel, so that future model
development options are not Inadvertently foreclosed.
6. The BART sample was derived from art on-board survey of transbay riders during the period of the Bay
Bridge cloture after tha 1989 Loma Priata earthquake. The sample was subsequently divided into BART uaar*
and nonusers baaed on the response to tha question on how frequently they had used BART before the
earthquake. So-called latent* rider* ware assumed to be those who were riding BART at tha time of the
survey, had ridden BART less than once per week before the Bay Bridge was closed.
Transportation-Air Quality Analyst$ Itsuat
January 1993
-------
f>«0* 6-7
1.
Household data
• household size
• dwelling unit type
• number of vehicles
• Income
2.
Person data
v%
• employment status
.
• workplace location
3.
Trip data
• beginning and ending times
• beginning and ending locations (coded to traffic analysis zone or Census
tract)
• trip purpose
• travel mode
1 - Household Travel Survey Data Typically Used in Travel Model Estimation
Information gathered from a.household travel survey falls into three categories:
• Household
•• Persons in the household
• Trips by persons in the household on the, given travel day(s)
The following tables present lists of the information that was gathered during the 1990 MTC
and BART travel surveys. Table 2 lists information on the household; Table 3 lists
information on household occupants; Table 4 lists Information on trips by household
occupants.
Note that the information that was obtained goes beyond that needed for development of
existing state-of-the-practice travel models. Information on previous household or workplace
location was sought to obtain better Information on residence and job movement patterns
within the San Francisco Bay Area. The question on flexibility of work schedules was asked
to gain information on the sizes of markets for transportation demand management
measures such as fiextime and ridesharing. Questions on parking were intended to obtain
not only parking cost but also the extent to which subsidized parking exists for work trips.
Transportation-Air Quality Analysis Issue*
January 19 93
-------
Page 6-8
1.
Number of persons
• total number
• number under 5 years of age
2.
Location of residence (street address or nearest intersection)
3.
Length of residence at current location (H under 5 years, place of previous
residence)
4.
Dwelling unit information
• type (e.g., single-family house, duplex, multi-family, mobile home, etc.)
• owned or rented
• (optional) purchase price
5.
Vehicles in operating condition
• number of cars, vans, trucks
• number of motorcycles
• number of mopeds
• number of bicycles
• list of cars, vans, trucks
— make
— model
— model year
— fuel type
6.
Income
2 • Household Data
Activity questions should, ideally, be asked for both the trip origin and the destination. This
aids checking the continuity of an individual's travel for the day It is also necessary In
instances where a day's travel begins at a place other than home/
7. In the 1901 Bay Area Travel Survey, activity questions were asked for both origin and destination. In the
1990, only a single question was asked lor each trip: trip purpose.
Transportation-Air Quality Analysis Issu**
January 1993
-------
Pmgo B-9
1.
Age
2.
Sex
3.
Relation to head of household
4.
Driver's license?
5.
Handicapped? (H so, what condition?)
6.
Occupation (employed full time, emptoyed part time, student full time, student
part time, retired, unemployed, homemaker, etc.). If more than one
occupation (e.g., student and employed, or more than one job), record
information for all occupations.
7.
If employed, occupation type (e.g., manager, clerical, operative, etc.;
corresponding to Census categories)
S.
Workplace or school location
9.
Business type at workplace
10.
How long at workplace location (if less than 5 years at workplace location,
previous workplace location)
11.
If employed, flexibility in work times? (e.g., how many minutes early or late can
person be to work, leave work)
3 • Person Data
Survey Design and Sampling
Conducting a telephone survey is considerably less expensive than a home interview survey.
But it introduces several biases of its own:
• Obviously, persons without a regular telephone number are excluded from the
sample. This includes households without telephones and homeless persons.
• Persons living in institutionalized residences (nursing homes, residence hotels, etc.)
are excluded from the sample.
• There is a greater likelihood of establishing contact with larger households.
Transportation-Mr Quality Analysts /ssues
January 1993
-------
Page 6-10
t. Beginning and ending trip times
2. Origin and destination activities (e.g., home, work, school). Distinguish
between different types of activities that would affect travel behavior (e.g.,
comparison shopping, convenience shopping, and grocery shopping should
be treated as separate types of activities).
3. Trip origin and destination locations; street address, building, or nearest
intersection. These are typically coded to a census tract or block group or,
ideally, point coded in a GIS database.
4. Travel mode
5. Car trips:
• identify specific vehicle in household from list
• number of persons In vehicle
• parking information;
— parking type (e.g., street free, street meter, lot free, lot meter, employee
lot, subsidized parking, etc.)
— parking cost
6. Transit trips;
• fare paid (or pass type)
• number of transfers
• waiting time for transfers
A - Trip Data
• It may be more difficult to Interview non-English speaking persons, either through
lack of capability on the part of the survey team, or through Inability of the interviewer
to identify which language Is being spoken.
It may be argued that the first two sources of bias may not matter for the purposes of travel
modeling. The persons who are excluded from the sample very likely account for a small
percentage of the population. And they tend to be persons who travel less often.
Nevertheless, it should be made explicit when reporting household survey results what
groups are excluded from the sample, and what the likely effect is on Inferences drawn from
the sample.
Because the proportion of single-person households has been Increasing, survey bias
against smaller households Is becoming more significant as a problem. This source of bias
can be reduced by making as many attempts as possible to contact a household at a
working residential telephone number. The 1981 and 1990 Bay Area Travel Survey designs
Transportation-AJr Quality Analysis /tsuec
January 1933
-------
Page 6-11
called for a maximum of 8 attempts to contact persons at a working residential number.
Each number was tried at different times of day, different days of the week, and on
weekdays and weekends.
The method of drawing the sample can introduce biases of its own. As discussed above,
the 1981 Bay Area Travel Survey used directory-based random-digit dialing to draw the
sample, with a small supplementary sample from the reverse telephone directory. The 1990
Bay Area Travel Survey sample was drawn from a commercially purchased list of residential
telephone numbers. Both of these methods provided adequate geographic coverage of the
Bay Area; the socioeconomic characteristics of the households in the sample closely
matched the Census data
At the conclusion of the 1981 survey, an analysis was conducted to test the validity of the
1980 Caltrans survey sample, which was drawn using the reverse telephone directory. Each
household in the 1981 survey was coded according to whether or not it could have been
included in the 1980 Caltrans sample; i.e., whether or not the household appeared in the
reverse telephone directory, The analysis showed that the Caltrans sample contained biases
against households that had resided at the current address for less than one year,
households in multi-family units, low-Income households, and households without cars.
Reverse directories also do not contain unlisted numbers, which are more likely to be held
by households with lower incomes and shorter lengths of residence In the same place."
Hence, drawing a sample of telephone numbers from a reverse telephone directory appears
to be an inferior method when compared to a sample drawn either by random-digit dialing
or from a purchased sample of telephone numbers. Furthermore, when using random-digit
dialing, directory-based random digit dialing is more efficient than random-digit dialing within
prefixes. Nationally, random-digit dialing within prefixes requires calling on average about
3.8 different numbers per working residential number contacted.9 The 1981 Bay Area Travel
Survey, using directory-based random-digit dialing, averaged fewer than 2 numbers per
working residential number reached.10
Response Rates
Nonresponse to a household travel survey can significantly bias estimates obtained from the
sample. In particular, it has been found that households that are less mobile are less likely
to respond to travel surveys. Because travel survey nonresponse Is correlated with
endogenous variables (especially travel frequency), there is no objective way to estimate or
correct for nonresponse bias In a travel survey. A high response rate is therefore an
B. Clyde R Rich, Is Random Digit Dialing Really Necessary,' Journal of Marketing Research Vol XIV (August
1877), pp, 300-305.
B. Telephone companies ueually assign number* in "blocks* of 1000 (e.g., 464-6XXX). Hence, random-digit
dialing within prefixes can result in repeated attempts within blocks where few numbers have been assigned.
Nationwide, random-digit dialing will average attempting to call 3.6 numbers per working residential household
reached. See Gerald J. Giasser and Gale D. Metzger, 'National Estimate of Nonlisted Telephone Households
and Their Characteristics.* Journal of Marketing Rasmrch v. XII (August 1B75}, pp. 395-461.
10. Crain & Associates, Inc., *1981 Bay Area Travel Survey.*
Transporiatlon-AIr Quality Analysis /tsues
January 1993
-------
Page $-12
essential goal of travel survey design and implementation. Experience from various types
of transportation surveys suggests the following broad guidelines on response rates:
• 70% and above • Nonresponse bias is likely to be low.
• 50% to 70% ~ Nonresponse bias may present somewhat of a problem, especially for
nonwork trips and for trips made by households whose members make fewer trips
than average for the population (e.g., low-income, retired),
• Below 50% - Nonwork travel is likely to be seriously underrepresented. Trip rates
estimated from surveys with high nonresponse could be significantly overestimated.
The 1981 Bay Area Travel Survey achieved an overall response rate of 70% from initial
contact to completed interview. It has become Increasingly difficult to obtain high response
rates for a number of reasons, including the following:
• Persons are becoming more reluctant to respond to surveys. This is partly due to
feelings of increasing Invasion of privacy over the telephone by solicitors and by
commercial surveys.
• Telephone answering machines are becoming more common. This makes it more
difficult to establish contact with households that use them for screening calls.
• Persons with unlisted numbers may be quite suspicious of getting calls from
strangers. A frequently encountered response from persons with unlisted numbers
is, *How did you get my number?"
• A household travel survey goes into considerable detail on the socioeconomic
characteristics of the household and the travel behavior of its occupants. Over the
years, many persons have become increasingly reluctant to let a 'big brother*
government agency know all about them, or to know exactly what they have been
doing during the travel day(s); or, there may be a fear that the information may be
misused by unscrupulous persons who would want to know when they are away
from home.
These were particular problems for the 1990 Bay Area Travel Survey; the response rate was
under 60%. In particular, respondents who cooperated with the first part of the interview on
characteristics of the household would refuse to give out their address so that cards could
be sent to them to record their trips. This was especially a problem with single-woman
households.
Nevertheless, a high response rate remains the main goal of survey design and conduct
Several methods have been successfully used to increase the response rate of a survey and
to minimize nonresponse bias, Including the following:
• A publicity campaign before and during the survey, consisting primarily of public
service announcements In the media Also helpful are press releases by public
Transporiatlon-AJr Quality Analyst* /ssues
January 1993
-------
Paga 6-13
officials to explain the purpose of the survey and to encourage persons who are
contacted by the survey to respond.
• An envelope and cover letter included with the travel diary that bear the letterhead
of the sponsoring agency.
• A clew Introduction to the survey on initial contact wHh the household. The caller
should identify the sponsoring agency and explain the purpose of the survey. This
implies that interviewers should be trained from the beginning to understand the
purpose of the survey and the reason for asking each question, so that they can
explain it to persons they contact. Persons who are contacted and who question the
purpose of the survey are usually satisfied with the answer This survey Is being
used to provide information to transportation planning agencies in the region so that
they can plan transportation facilities to serve you better.* This 'up front* approach
is In sharp contrast to marketing surveys, where persons who are contacted are told
neither who is collecting the information nor the purpose of the survey. Persons are
more likely to respond when they know what the information will be used lor,
especially If this is to be used for public purposes.
• Respondents with unlisted numbers who say, "... how did you get my number?"
should be told that their number was reached by chance because telephone
numbers were dialed at random. This explanation is usually sufficient.
e Prompt callbacks to a household no more than two days after the designated travel
day(s). These and any further necessary callbacks should be scheduled to make it
as likely as possible that someone will be on hand to answer the telephone.
e Respondents should be repeatedly assured that the information that is gathered will
be kept confidential and will be used only In the aggregate for statistical purposes.
To preserve confidentiality for persons who refuse to give their home address,
Interviewers should seek to obtain the Intersection nearest the residence; trip cards
can be sent to the person's place of employment
• In some cases, an Incentive may be effective in increasing the response rate. In the
1990 MTC travel survey, respondents who were asked to keep multi-day trip diaries
were paid $10 for completing the Interview. In the Puget Sound Transportation
Panel (see below), Incentive payments of $2 and $10 were tested, and found to be
about equally effective In encouraging responses.11 In the Los Angeles regional
household travel survey in 1991, each trip diary mailed to respondents contained a
$1 bill. Experience with various types of surveys shows that the presence of my
Incentive, no matter how small, will encourage a greater response rate.
11. Eaine Murakami and Cy Ullbwg, 'Current Statue of frit Puget Sound Transportation Panel," paper
presented at the Firet U.S. Conference on Panel! for Transportation Planning, Lake Arrowhead, California.
October 1992.
Transportation-Air Quality Analysis Issuas
January 1993
-------
Page 6-14
Obtaining Travel Dam
Obtaining travel data Is the most difficult and expensive part of a household travel survey.
It is crucial that all trips are recorded If the survey sample is to property represent the mix
of travel in the region. Work trip data are typically reported with more accuracy than all
other types of trips. The problem for the survey is to obtain information on trips that are less
likely to be recorded: mainly nonwork and walk trips (e.g., a trip wtth walk as the only
mode, a walk trip from a parking lot to the workplace}."
There are two alternatives for obtaining trip information from a household travel survey
conducted by telephone: sending out trip diaries to be filled out and mailed In, or
conducting a second telephone interview of the household. The first method is less
expensive, but there Is the danger respondents may forget to record some trips; moreover,
If detailed Information beyond origin, destination, mode, purpose, and times are sought for
each trip (e.g., type of parking, vehicle used, number of persons in car), the trip diary
becomes more lengthy, which tends to discourage complete responses. The second
method can be effective for gaining information on trips that would otherwise be omitted,
especially If interviewers we instructed to prompt persons to remember all trips.13 An
important part of the second method is to send households cards on which to record the
important parts of each trip (locations, times, modes), so that the written information serves
as a reminder during the callback interview.
Recently, a third alternative has been developed and used to collect trip information.
Households are sent diaries on which to record their activities (e.g., home, work, shopping),
and how and when they traveled between them, on a designated day or days. These
activity-based diaries have been found to be easy to fill out because persons are more likely
to be aware of their activities than their trips. This method was used in the 1991 travel
survey of 16,000 households for the Southern California Association of Governments.
Nevertheless, it is still necessary to include safeguards to ensure that all legs of a trip,
especially walk modes, are recorded.
Quality Control
The issue of quality control goes beyond simply developing a good survey design and
ensuring that interviewers are doing their job well. Once the data are collected, It Is
necessary to edit, code, and enter them onto a computer medium. Careful control over this
process will reduce the chance of errors.
12. Trip information from household travel survey# does not Indude Information on vehicle movements that art
not considered to be trips: e.g.. moving a car from the driveway to a street parking space. If trip data are used
to estimate emissions (see Section 3), they would underestimate emissions depending on to what percentage
of cold starts are accounted for by these types of vehicle movements.
13. In the 1981 and 1990 Bay Area Travel Surveys, interviewers were instructed to periodically prompt
respondents during the trip interview with the question; *... and did you stop anywhere else along the wayV
Transportation-Air Quality Analysis Issues
January 1993
-------
Paga 6-15
The following are especially important elements of a good quality control program for a
household travel survey:
• Survey data should be coded as soon as possible after the interview is completed.
The most frequent problem encountered in coding travel surveys arises from
recorded origin or destination locations that cannot be coded (e.g., an address that
does not exist, or a location given as an intersection of two streets that are parallel),
in such cases, it is necessary to call back the respondent to correct the error.
• Survey information should be cross-checked for consistency. For example, if an
activity at a location is given as "home*, the location of that activity in the trip data
should agree with the residence location recorded in the household data section;
similarly, a location where the activity is "work* should agree with the workplace
location recorded under data for occupants of the household.
• For each person's travel, tip starting and ending times should follow in sequential
order; i.e., the ending time of each trip should be greater than the starting time, and
the starting time of a trip should be greater than the ending time of the previous trip.
• Activities should follow in logical sequence for each person's trips when travel data
are recorded as trips. For example, if the destination activity of a trip is recorded as
"work*, the origin activity of the next trip should also be "work".
The use of computer-assisted telephone interviewing (CATI) can automate several aspects
of quality control as well as reduce survey and coding costs. Use of this technique involves
developing a questionnaire that appears on a computer screen. Skip parterns can be
programmed in so that responses to questions determine which questions will be asked
next; for example, if a trips is recorded as being made on transit, only transit-related
questions such as fare and transfer information appear, and automobile-related questions
such as parking cost are skipped.
Survey Period and Scheduling
Interviews for household travel surveys typically require one to several months to complete.
If the survey Is to represent a typical* travel day, it is usually desirable to schedule the
survey for the fall or the spring, when school is In session. To minimize the effects of
variations In weather or traffic conditions, the survey should span a sufficiently long period
so that most travel interviews cover days that are as close to typical as possible. Holiday
periods should be avoided; hence, spring is usually a better period than fail in which to
conduct a household travel survey because of the lower incidence of holidays.14
6.3. ALTERNATIVE USES OF HOUSEHOLD TRAVEL SURVEY DATA
14. This can, howevar, prasani a consistency problem when using the results of household trave* survey* for air
quality analysis. Air quality modeling typically focuses on January (for CO) and on the summer months for
ozone. Both timss ara considered "nontypical* for othe* transportation planning purposas.
Transportation-Air Quality Analysis Issue*
January 1993
-------
Page 6-16
As discussed above, planning agencies view the main purpose of household travel surveys
as providing data for developing regional travel models. These surveys typically collect
other data that are not needed for travel model development, but could be used to address
transportation and air quality analysis issues that are not possible to analyze by using typical
travel demand models. This section discusses two examples: vehicle operating mode
information and market-based analysis of transportation control measures (TCMs).
6.3.1 Vehicle Operating Mode
An important issue in air quality planning is estimating the operating mode of passenger
vehicles, particularly the percentages by trip type of vehicle trips that begin In the cold-start
mode. In practice, these percentages are derived from assumptions on characteristics of
each trip type; e.g., almost all trips from home to work begin in the cold-start mode.
A household travel survey can provide significant Information on operating modes because
start and end times are recorded for each trip; hence, for auto driver trips, the time between
trips can be used to infer what percentage of automobile trips begin In the cold-start mode.
The trip diary could therefore be regarded as a sample of auto driver trips that could be
analyzed to estimate the percentage of trips by purpose that begin in the cold-start mode.'*
If the trip information also contains information on the particular vehicle used, as In the 1981
and 1990 MTC travel surveys, It is possible to develop even more detailed information. The
household travel survey data and the regional network can be used in conjunction to
develop a database that effectively acts as a sample of tripmaklng by purpose. Developing
emissions estimates from the data could be carried out in the following steps;
• Create a database for each vehicle in the trip file, containing the following information
for each automobile trip:
— running time and mileage driven on the previous trip
— amount of time the vehicle was idle before beginning the trip
— running time and mileage driven on the current trip
— vehicle type
• Using an appropriate algorithm, determine whether each trip began In the hot start
or the cold start mode.
• As a further refinement, estimate emissions for each vehicle trip based on vehicle
type and start mode.
15By exienaion, K would b« poaalble to define on a or more intermediate mode* bafwaen oold-start and hot-
start based on the ratting time between tripe, and to estimate tha numbar of tripa that begin in each of aavaral
operating modas.
16. If, in future air quality anaiysea. It became common jxactica to define one or modes between cold start and
hot start, It would be possible to use these data to determine in which mode (extreme or intermedials) each trip
began.
Transportation-Air Quality Analysis Issue*
January 1993
-------
Page $-17
• Run statistics on the sample to derive estimates of the following for trips by type in
the region:"
— percentage of trips by type that begin In cold start mode
— average emissions per mile
6.3.2 Market-Baaed Analyses of TCMs
Existing travel models can be used to estimate the effects of TCMs insofar as the TCMs
affect travel supply (e.g., travel time and cost). But other TCMs cannot be represented
within the existing travel modeling framework; e.g., ridesharing promotion programs, transit
marketing, introduction of an employer-based transit pass marketing program. Hence,
existing travel models are not well suited to analyzing the effects of TCMs that do not directly
affect travel times and costs.
A valid alternative to developing Information on the relative effectiveness of TCMs is to
estimate the markets for each TCM, then to estimate the potential market penetration of the
TCM. This can provide at least a bound on the maximum effect of each measure.
Consider an example of measures designed to encourage greater transit use for work trips.
To better consider the likely effectiveness of these measures, the travel market could be
segmented into several groups, for example:
• Two or more working adults and children of preschool age.
• Single adult and children of preschool age.
• One or more working adults who use their cars during the day for work-related travel.
• No children, or teenage children only.
• Two adults, two or more vehicles, and only one working adult.
• One or more nonworking adults.
The market penetration of a transit-related TCM Is likely to be lower for the first three groups
than the last three. Using information from the household travel survey, It is possible to
estimate the percentages of households with workers, and therefore the percentages of work
trips, that fall into each of these categories.
6.3.3 Exploratory Analyses
17. These data would not cover vehicles such as those that are part of a company fleet. Bui information on
operating modes for these vehicles could be estimated from cample surveys of tripmaking using fleet vehicles
Transportatlon-AJr Quality Analysis Issuts
January 1993
-------
Page 6-18
Household travel survey data offer the possibility of conducting exploratory analyses of travel
behavior that can indicate subsequent refinements
6.4. NEW DIRECTIONS IN HOUSEHOLD TRAVEL SURVEYS
Transportation researchers are becoming increasingly aware that traditional household travel
survey data present significant problems for analyzing and forecasting travel behavior. In
particular, as they are collected and used, household travel survey data represent a cross-
sectional sample of household travel behavior*, yet they are used to Infer longitudinal
changes in travel in response to changes in travel supply. Furthermore, over time, Individual
households go through changes that affect their travel behavior. There is growing
awareness that to accurately look at these changes and to Infer behavioral responses to
changes in the transportation system, It is necessary to develop longitudinal information on
household behavior. This can be done by asking retrospective questions on household
travel behavior. A more reliable method is to make repeated observations on the travel
behavior of a set of individual households over time; this is known as a panel survey.
Panel surveys have been used extensively In medical research since the 1950s, for example,
in long-range studies of heart disease, where observations on the same individuals are
conducted at frequent intervals. Their use in travel behavior studies is relatively recent,
although the technique was advocated as early as 1966.1# A 1963 paper discussed a
number of ways in which repeated measurement from a panel of households would Improve
our understanding of travel behavior.'* As summarized by Duncan ef. a/., these include
the following:20
e Describing and analyzing changes in travel behavior In response to changing prices
or the availability of public transportation.
• Analyzing the sequencing of joint decisions about the place of residence, place of
work, and home-work trips.
• Understanding the changes in energy consumption In response to changes in energy
prices.
• Forecasting car ownership and driving licenses.
18. William L Garrison and Rjchard D. Worrall, Monitoring Urban Travel: Final Report of Project 2-B. Estimation
and Evaluation of Diverted and Generated (Induced) Traffic, unpublished draft report prepared for National
Cooperative Highway Research Program, Highway Research Board, Washington, D.C., 1866.
19. B. Baanders and K. Slootman, "A Panel for longitudinal Research Into Travel Behavior,* In S. Carpenter and
P. Jones (ed« ), Recant Advances In Travel Demand Anaiys/s, Hants., England, Gower, 1963. Crted in G.J.
Duncan el. aJ., "Panel Studies In Research on Economic Behavior," Transportation Research A, Vol. 21A, No.
4/5,1967, pp. 249-263.
20. G.J. Duncan at. a!., "Panel Studies*
Trans porta tlon-AJr Quality Analysis Issues
January 1993
-------
Pag# 6*19
• Estimating the price elasticity of public transportation by measuring behavior before
and after price changes.
Panel surveys are often the onfy realistic method for collecting longitudinal data. Asking
respondents to remember past behavior can frequently lead to misleading results. For
example, an on-board transit survey in Sacramento, followed up by telephone survey data
on the same Individuals, showed that in many cases the same Individual gave varying
responses to the same question In both surveys on frequency of transit use*
The most extensive panel survey to date is the Dutch Mobility Panel, which began with a
sample of over 5,000 households. The Initial Intent of the survey was to monitor changes
in travel behavior in response to transit fare changes. The survey consisted of week-long
travel diaries administered at six-month Intervals. Since the first wave of the survey In 1984,
a total of ten additional waves were conducted at six-month intervals through 1989."
In the U.S. the most extensive panel to date is the Puget Sound Transportation Panel, which
consists of two waves of surveys from 1989 to 1990." The sample covered 1,700
households in the first wave, and 1,800 households in the second wave; trip diaries covered
two days. The survey was designed to overs ample bus and carpool commuters.
In what was to have been the most ambitious panel survey in the U.S., MTC enlisted 9,600
households out of 10,900 surveyed during the 1990 household travel survey; this was in
addition to the 1,100 households that comprised the BART portion of the survey, which were
intended to be the foundation of a BART user/nonuser panel.24 MTC has tried to obtain
funding for subsequent waves of surveys of 3,000 households in this group, but has been
unable to do so. As time goes on, there is an increasing danger that MTC will not be able
to maintain this panel.
Panel surveys provide information on the dynamics of change that is simply not available
from cross-sectional surveys. A longitudinal model of tripmaking on five travel modes was
estimated from the first three waves of the Dutch mobility panel; the model, which contained
lagged variables to capture time effects, showed significant "inertia* effects (households
tending to exhibit the same behavior despite changes in the transportation system), but also
21. Respondents to the both surveys were asked: "How many day* did you use transit the pat! week?*
Individual responses to both questions showed a high degree of variance. See David Reinke, Trtrmft Pass
Distribution Methods In Sacramento, California, report prepared (or Transportation Systems Center, U.S.
Department ot Transportation, May 1065.
22. Henk Meurs and Geert Ridder, 'Attrition and Response Effects In the Dutch Mobilty Panel.* paper
presented at the First U.S. Conference on Panels for Transportation Planning, Lake Arrowhead, California,
October 1992.
23. Murakami and Ullberg. 'Puget Sound Transportation Panel.* See also E. Murakami and W. T. Watterson,
•The Puget Sound Transportation Panel After Two Waves,' Transportation, Vol 19, 1992. pp. 141-158.
24. Purvis. "The San Francisco Bay Area Household Panel Survey.'
Transportatlon-AIr Quality Analyst* Issues
January 1993
-------
Paga 6*22
We haw learned from past experience. We can now do household travel surveys that are
much better and less expensive than the earliest home Interview travel surveys. Household
travel surveys from now on will rely mainly on telephone interviews for reasons of coverage
and economy.
Design and Implementation of a good household travel survey requires considerable
care. Accurate and inaccurate data look remarkably alike. There Is no sure way to
distinguish between them unless one (mows how they were collected. An improperly
designed or drawn sample can result In surveying a group of households that does not
accurately represent the region. Nonresponse bias in a household travel survey can lead
to significant overestimates of trip rates because persons who travel less often are less likely
to respond to travel surveys. Collecting a// travel data requires care in collecting trip
information. A well-conducted survey requires painstaking work by survey managers,
interviewers, and data processing staff. If a survey cannot be done well, it is better not to
do it than to gather inaccurate information on which important policy decisions may be
based.
If household travel survey data are used only for developing travel models, they are not
being fully used. Existing travel models are not very good tools for analyzing policies such
as most TCMs, or for estimating vehicle operating modes. But household travel survey data
contain a information that can significantly inform these analyses.
Panel surveys represent a promising new method for understanding the dynamics of
changes In population characteristics and travel behavior. We are still learning how to
best conduct panel surveys, and how to develop models of travel behavior based on these
data. But longitudinal observations hove yielded important insights into the dynamics of
change. It is not necessary to develop working travel models for panel survey data to
provide information. Simple data tabulations of panel survey data can provide significant
information by themselves, for example, In analyzing a transit fare change by looking at
before-and-after data.
Where do we go from here? New policy initiatives will require new analysis procedures,
which will require new data Funding for traditional household travel surveys remains
scarce, although new federal initiatives on funding for travel model research may lead to new
data collection efforts. State and regional agencies seldom have the wherewithal — or the
will — to fund surveys on their own. Yet the need for new data remains. The following are
several actions that could be taken.
Develop national guidelines for household travel survey data collection. The past 20
years have given us a body of experience that should be distilled and disseminated to all
U.S. transportation planning agencies. Similar to an existing project on Identifying best
travel modeling practices currently being carried out for the National Association of Regional
Councils, a manual of best household travel survey practices should be developed and
implemented as a set of guidelines, especially If federal or state funding is sought for
surveys.
Transportatlon'Alr Quality Analysis Ittuas
January 19 93
-------
Page 6-23
Make evaluation — and aurveylng at part of evaluation — an Integral part of funding
for major transportation projecta. Evaluation of major transportation projects can provide
important information for future policy analysis, yet this Is seldom done In practice." A
major component of any evaluation would be to analyze the effects on travel behavior.
Travel behavior changes would be best assessed by a panel survey that include
observations before and at severed times after implementation of the project so that, at least,
short-term and Intermediate-term effects of the project could be assessed. The surveys
would Include control groups to allow for cross-sectional comparisons between market
segments (i.e., those who would be affected by the project and a control group of persons
from similar socioeconomic groups who are outside the area affected by the project). For
transportation projects such as new rail starts, costing on the order of $100 million to $1
billion, it Is asking very little to set aside 1% to 2% of the project funding for evaluation,
including surveys. This would also provide the regional planning agency with an up-to-date
data set for model development and other policy analysis.
Fund existing panel surveys. The Puget Sound Transportation Panel remains the oldest,
and oniy, ongoing regional household travel survey panel in the U.S. Two waves have
already been conducted, and a third wave is scheduled for 1992. Experience with this
survey will guide future panel survey efforts. The 1990 MTC travel survey contains the
largest known group of respondents who have been Identified for future contact for a new
wave of a panel survey. This sample could potentially yield a panel of such a large size that
strong consideration should be given to continuing funding this survey as a major
demonstration of collection and use of panel survey data in the U.S. Given provisions in
ISTEA that call for looking at congestion pricing, and given that the San Francisco Bay Area
is one of the areas under consideration for implementation, the MTC sample — and the
BART subsample — would be extremely valuable for evaluating the effects of congestion
pricing.
33. Exceptions to this statement include the BART impact tludies ot the 1870s and the current or planned
studies of BART, Metro (Washington, DC), and MARTA (Atlanta) that are being funded by FTA.
Transportation-Air Quality Analysis Issues
January 1993
------- |