United SIH1*!S
Environmental Protection
Office of Transportation EPA420-B-05-018
and Air Quality October 2005
Procedures Manual for the
COMMUTER Model v2.0
-------
EPA420-B-05-018
October 2005
Procedures Manual for the
COMMUTER Model v2.0
Transportation and Regional Programs Division
Office of Transportation and Air Quality
U. S. Environmental Protection Agency
Prepared for EPA by
J. Ri chard Kuzmyak
Thomas R. Carlson
Robert G.Dulla
Sierra Research, Inc.
Stephen D. Decker
Christopher D. Porter
ErinE. Vaca
Cambridge Systematics, Inc.
NOTICE
This technical report does not necessarily represent final EPA decisions or positions.
It is intended to present technical analysis of issues using data thatC are currently available.
The purpose in the release of such reports is to facilitate the exchange of
technical information and to inform the public of technical developments which
may form the basis for a final EPA decision, position, or regulatory action.
-------
Table of Contents
page
1. Introduction 1
2. Overview of the COMMUTER Model
Analysis Procedure 4
What is the COMMUTER Model? 4
Conventional Analysis Methods 4
How the COMMUTER Model Approach is Different 7
Baseline Data Requirements 10
Selection of Strategies and Development of Scenarios 11
Computational Methods for Estimating Transportation Impacts 12
Emissions Calculations 20
Model Outputs 20
3. Assembling Required Background Information 22
Overview 22
Required Model Inputs 24
4. Estimating Travel Impacts of Employer Support Programs 34
Overview of Employer Support Strategies 34
Analytic Approach for Estimating Travel Impacts 36
5. Estimating Travel Impacts of
Alternative Work Schedules 45
Overview of Alternative Work Schedule Programs 45
Nature of Emissions Impacts 45
Analytic Approach for Estimating Travel Impacts 46
Multiple Options 54
6. Estimating Travel Impacts of Travel Time and Cost Changes 56
Overview of Travel Time and Travel Cost Strategies 56
Types of Time/Cost Strategies Covered by COMMUTER Model 59
Analytic Approach for Estimating Travel Impacts 61
Coefficients Used in Pivot-Point Logit Model Approach 66
Example Calculation of Change in Time or Cost through Logit Pivot-Point Model
69
7. Converting Travel Impacts to Emissions Reductions 73
Overview of Procedure 73
Changes in Total Trips 74
-i-
-------
Changes in Total VMT 76
Allocation Between Peak Period and Off-Peak 79
Emission Impacts Methodology 83
Appendix A
Definition of Modeling Terms 88
Appendix B
Fuel Parameters for Default Emissions Factors A-4
-11-
-------
List of Tables
page
1. Introduction 1
2. Overview of the COMMUTER Model
Analysis Procedure 4
What is the COMMUTER Model? 4
Conventional Analysis Methods 4
How the COMMUTER Model Approach is Different 7
Baseline Data Requirements 10
Selection of Strategies and Development of Scenarios 11
Computational Methods for Estimating Transportation Impacts 12
Emissions Calculations 20
Model Outputs 20
3. Assembling Required Background Information 22
Overview 22
Required Model Inputs 24
4. Estimating Travel Impacts of Employer Support Programs 34
Overview of Employer Support Strategies 34
Analytic Approach for Estimating Travel Impacts 36
5. Estimating Travel Impacts of
Alternative Work Schedules 45
Overview of Alternative Work Schedule Programs 45
Nature of Emissions Impacts 45
Analytic Approach for Estimating Travel Impacts 46
Multiple Options 54
6. Estimating Travel Impacts of Travel Time and Cost Changes 56
Overview of Travel Time and Travel Cost Strategies 56
Types of Time/Cost Strategies Covered by COMMUTER Model 59
Analytic Approach for Estimating Travel Impacts 61
Coefficients Used in Pivot-Point Logit Model Approach 66
Example Calculation of Change in Time or Cost through Logit Pivot-Point Model
69
7. Converting Travel Impacts to Emissions Reductions 73
Overview of Procedure 73
Changes in Total Trips 74
Changes in Total VMT 76
Allocation Between Peak Period and Off-Peak 79
Emission Impacts Methodology 83
-in-
-------
Appendix A
Definition of Modeling Terms .
Appendix B
Fuel Parameters for Default Emissions Factors A-4
Table 3-1 Default Values Provided in the COMMUTER Model 25
Table 3-2 Definition of Office and Non-Office Employment 28
Table 3-3 Default Shares for a Standard Regional Analysis 30
Table 4-1 Popular Employer Support Strategies to Encourage Specific Modes
of Travel 35
Table 4-2 Increase in Percent Using Model by Support Program Level 37
Table 4-3 Composition of Modal Support Strategy Programs 38
Table 5-1 Percent of Trips Shifted by Length of Peak Period 48
Table 6-2 Logit Mode-Choice Coefficients for Individual Urban Areas 67
Table 6-3 Default Coefficient Values 68
Table 7-1 Percent of Trips Shifted by Length of Peak Period 80
List of Figures
page
1. Introduction 1
2. Overview of the COMMUTER Model
Analysis Procedure 4
What is the COMMUTER Model? 4
Conventional Analysis Methods 4
How the COMMUTER Model Approach is Different 7
Baseline Data Requirements 10
Selection of Strategies and Development of Scenarios 11
Computational Methods for Estimating Transportation Impacts 12
Emissions Calculations 20
Model Outputs 20
3. Assembling Required Background Information 22
Overview 22
Required Model Inputs 24
4. Estimating Travel Impacts of Employer Support Programs 34
Overview of Employer Support Strategies 34
Analytic Approach for Estimating Travel Impacts 36
-IV-
-------
5. Estimating Travel Impacts of
Alternative Work Schedules 45
Overview of Alternative Work Schedule Programs 45
Nature of Emissions Impacts 45
Analytic Approach for Estimating Travel Impacts 46
Multiple Options 54
6. Estimating Travel Impacts of Travel Time and Cost Changes 56
Overview of Travel Time and Travel Cost Strategies 56
Types of Time/Cost Strategies Covered by COMMUTER Model 59
Analytic Approach for Estimating Travel Impacts 61
Coefficients Used in Pivot-Point Logit Model Approach 66
Example Calculation of Change in Time or Cost through Logit Pivot-Point Model
69
7. Converting Travel Impacts to Emissions Reductions 73
Overview of Procedure 73
Changes in Total Trips 74
Changes in Total VMT 76
Allocation Between Peak Period and Off-Peak 79
Emission Impacts Methodology 83
Appendix A
Definition of Modeling Terms -88-
Appendix B
Fuel Parameters for Default Emissions Factors A-4
Figure 1 Overview of Calculation Procedure 2
Figure 2-1 Traditional Transportation and Emissions Analysis Process 5
Figure 2-2 COMMUTER Estimating Procedure 9
Figure 2-3 Illustration of Share Adjustment Process 16
Figure 2-4 Logit Relationship 18
Figure 4-1 Illustration of Share Adjustment Process 44
Figure 6-1 Disutility of Mode m 63
Figure 6-2 Modal Share 65
Figure 7-1 Translation of Travel Demand Changes into Emissions Reductions 73
-v-
-------
1. Introduction
This manual describes a set of procedures that can be used to evaluate voluntary mobile
source emission reduction and commuter choice incentive programs for State
Implementation Plan (SIP) and employer-based commuter choice programs.* To
simplify their use, they have been implemented in a spreadsheet model, entitled
COMMUTER, whose operation is explained in a companion report entitled
"COMMUTER Model User Manual for Analysis of Voluntary Mobile Source Emission
Reduction and Commuter Choice Incentive Programs."
The procedures are intended to be simple to apply, with data requirements the minimum
necessary to evaluate alternative programs. At the same time, the analytical techniques
employed in these procedures are consistent with existing state-of-the-practice travel
modeling techniques widely employed by transportation professionals. The calculation
methodologies are largely based on those used in the Federal Highway Administration
Travel Demand Management Evaluation Model (FHWA TDM Evaluation Model),
developed by COMSIS Corporation in 1993. This model has been widely applied in the
evaluation of employer-based TDM programs for the purposes of calculating travel
impacts.
The analytical process described in this report is illustrated by the flowchart in
Figure 1-1. The basic sequence of steps in this process can be used to assess travel
impacts without use of the COMMUTER model. The sequence of procedural steps is
outlined below.
• First, adjust the baseline (existing) mode shares to reflect the impacts of
support and incentive programs such as rideshare matching and provision of
bicycle facilities. To make the adjustments, apply mode shift factors from
lookup tables that have been compiled from the literature and national
experience.
• Once the impact of support and incentive programs have been accounted for,
use travel modeling techniques to estimate the impacts of strategies related to
travel time or cost savings on mode choice. Support and incentive programs
include site-specific employer programs such as on-site transit pass sales,
rideshare matching, and guaranteed ride home programs. Those strategies that
can be measured by savings in travel times and costs include transit use
frequency, high occupancy lanes, and transit fare reductions.
Conformity regulations (40 CFR 93.110) require the use of the latest planning assumptions in conformity
analyses. Whenever the COMMUTER model is use for SIP or conformity purposes, the regulations require
that the latest planning assumptions be used for conformity determinations. Default data should not be
used.
-1-
-------
Figure 1
Overview of Calculation Procedure
Baseline Mode Shares
Adjust mode shares for
"soft" programs
Adjust mode shares for
time and cost incentives
Compute changes in
trips
Background
transportation data
Program
Information
Lookup tables
Coefficients
Compute changes in
VMT
Compute changes in
total emissions
• Convert the changes in mode shares to changes in motor vehicle trip ends. Use
the average trip length by mode to convert these changes in trip ends to
changes in vehicle miles traveled (VMT).
• Use predefined sets of emissions factors to estimate the emissions impacts.
The following sections discuss each step in the process in greater detail for both model-
and manually derived applications. The discussion covers definitions of terms,
-2-
-------
assumptions and analytical techniques, and coefficients and parameters that enter the
calculations.
It is highly recommended that analysts use locally derived values for the parameters
whenever possible to be consistent with local travel behavior characteristics in the
metropolitan area of analysis. In the absence of local data, the analyst may use default
values for program effectiveness that have been provided in this manual and built into the
spreadsheet model.
The following sections provide guidance in setting parameter values. Typically,
employers will want to use parameter values or values provided by their local
transportation planning agency, if available, before relying on the defaults listed in this
guidance.
###
-3-
-------
2. Overview of the COMMUTER Model
Analysis Procedure
What is the COMMUTER Model?
COMMUTER is a spreadsheet-based computer model that has been developed by EPA to
help areas obtain air quality credit from Best Workplaces for Commuters (BWC) and
other commuter choice-type programs. It is specifically intended to calculate the travel
and emissions effects that might result from implementation of voluntary employer
transportation management programs. The tool may be used by regional planning or air
agencies who are interested in pursuing this credit on a region-wide basis, or by
individual employers who may be seeking tax credit for implementing a voluntary
commuter choice incentive program.
Conventional transportation analysis tools normally used for regional planning and SIP
development are not well suited to this type of analysis task for several reasons, the most
important of which are the following:
• Traditional transportation analysis tools are unable to address many of the
voluntary commuter-choice measures, such as rideshare matching support or
alternate work schedules, in their structure; and
• Use of the traditional analysis tools is a complex and labor-intensive process.
This would limit analysis to only trained specialists, thereby excluding
employers and other interested groups from engaging in the process, and
dramatically cut down on the trial-and-error nature of determining what
particular combination of measures is the most effective and acceptable for a
given program.
COMMUTER offers a substantial savings in time and effort over the traditional approach
by employing some key shortcuts in the amount of data used and the number of micro-
level calculations that are performed. The result is a conscious but judicious tradeoff of
some accuracy for a significant increase in ease and flexibility for the user.
Conventional Analysis Methods
To understand the nature and operation of the COMMUTER model approach, it is
helpful to contrast it with the traditional approach. Figure 2-1 shows how a mobile
source emissions analysis would be done in normal practice. The process generally
consists of first estimating the transportation effects of a particular policy, program, or
improvement in the suite of models collectively known as the "four-step" process. This
results in a revised summary of vehicle trips, VMT, and operating speeds, which then
become the inputs to the emissions modeling procedure. Vehicle travel activity is mated
with emission production rates, typically through the latest version of EPA's MOBILE
-4-
-------
emissions model (EMFAC in California), which applies emissions factors that have been
specially adjusted to reflect local vehicle, regulatory, and climatalogical conditions.
Figure 2-1
Traditional Transportation and Emissions Analysis Process
"4-Step"
Transportation
Modeling
Process
Trip Generation
Trip Distribution
Mode Choice
Traffic Assignment
1
Trip
Tables
—
— '
Transportation
Emissions
Modeling
Process
Peak
Vehicle Trip
Volumes &
Speeds
24- Hour
Mobile Emissions
Factor Model
Mobile Source
Emissions
Inventory
Emissions Parameters:
• Vehicle Registrations
• Climatological Data
• Emissions Programs
-5-
-------
The four-step transportation analysis process is so named because of its separate
consideration of the elements of Trip Generation, Trip Distribution, Mode Choice and
Traffic Assignment.
• The Trip Generation step converts population, employment, income, vehicle
ownership, land use and other factors into an estimate of the number of trips
that will be made by households (productions) or attracted by economic and
personal activities (attractions).
• Trip Distribution then determines "where" households will travel to meet
their various needs, or conversely, "from where" travelers will come to serve
the various economic or personal attractions. The result is a set of tables (or
matrices) that depict trip flows between origins and destinations. The origins
and destinations in these "trip tables" correspond to specific geographic
"zones" defined by the planning agency. In a typical metropolitan area, there
is likely to be more than 1,000 such zones, meaning that their combined
function as both origins and destinations can result in several thousand origin-
destination pairs for which trip flows have been calculated, and upon which
analysis may be necessary. Also, trip tables are generally developed for
separate trip purposes, the most common of which are Home-Based Work,
Home-Based Non-Work, and trips that are not based from home, so-called
Non-Home-Based trips.
• Trip tables are then processed through a Mode Choice step, which predicts the
travel modes that will be used to make the various trips for each origin-
destination pair and for each trip purpose type (work, non-work, etc.). The rate
of use of particular modes (usually auto driver, auto passenger, and transit) is
based on which of these alternatives is available, the comparative service
offered by each alternative, and characteristics of the traveler. This
determination is usually made through a "mode split" model, and is calculated
separately for each origin-destination pair, resulting in a new (additional) set of
trip tables for private vehicle trips and transit person trips.
• Once vehicle trip volumes by origin-destination are known, the final step is to
distribute these trips onto the regional highway system in a process known as
Traffic Assignment. The computer is asked to overlay the regional trip tables
onto the system of freeways, arterials, collectors, and local streets and roads,
and to determine which combination of routes will be taken to complete the
trips most efficiently. To do this, the computer must have information on the
carrying capacity of each facility (on a segment-by-segment basis). Trips are
incrementally loaded onto the highway network to see what volumes and travel
speeds will result, and then reassigned to new paths in progressive iterations
until all travelers are realizing the most efficient (minimum travel time) path
for their particular trip given the overall volume of travelers. Major urban
areas will generally perform such a traffic assignment for an average 24-hour
day, and also for a case that represents peak period or peak hour conditions.
The peak period traffic assignment is crucial in identifying congestion
-6-
-------
conditions and in studying both work travel and competitive mode choice
situations.
The outputs of the transportation modeling process are then converted for use in the
MOBILE emissions model. MOBILE requires the user to supply information on the
local context in order to tailor the emissions rates appropriately to the individual area.
The information required includes composition of the regional vehicle fleet (by age and
size group) from vehicle registration and usage data; various climatological data,
including temperature and season; and existing emissions control programs in place, such
as reformulated fuels or inspection and maintenance programs. An important element in
adapting the MOBILE model and emissions factors to the region is also applying speed
correction factors to the respective emissions rates to reflect different rates of emissions
that occur at different operating speeds. The transportation inputs are used to make this
adjustment, which may be as simple as a single across-the-board adjustment for all travel,
or specific adjustments for VMT occurring on freeways vs. other facilities and under
peak or off-peak conditions.
How the COMMUTER Model Approach is Different
From the above, one can begin to see the value of having a simpler and more versatile
procedure for analyzing BWC and other commuter choice program strategies. While all
of the steps outlined above do not have to be repeated for every analysis using the
conventional models, it is still necessary to change the input data for each origin-
destination pair in the regional trip tables that will be affected by the policy or program
measure, re-estimate the modal split, and then redo the traffic assignment step in order to
get the new speed and VMT impacts required by MOBILE.
The COMMUTER model approach simplifies the quantification requirements for
workplace commuting programs by making selective simplifications, as well as
enhancements, to the conventional four-step model oriented process. The procedure is
heavily based on the Federal Highway Administration's Travel Demand Management
Evaluation Model (FWHA TDM Model), developed in 1993 for a similar purpose, to
facilitate analysis of TDM programs and strategies for congestion management and air
quality programs.* Prior to its acceptance by the FHWA for national distribution as a
planning tool, the TDM model underwent significant sensitivity testing, and has been
applied widely across the country by planning agencies, transportation agencies, and
employers or employer organizations. It was therefore viewed as a good base from
which to build the COMMUTER model.
The TDM Evaluation Model was developed by COMSIS Corporation in 1993 in conjunction with a
comprehensive program of research and development of reference and guidance tools by the Federal
Highway Administration and Federal Transit Administration. Additional information of potential value to
COMMUTER users may be found in the related reports from that research effort: Implementing Effective
Travel Demand Management Measures: Inventory of Measures and Synthesis of Experience, DOT-T-94-
02 (September 1993), Guidance Manual for Implementing Effective Employer-Based TDM Programs,
DOT-T-94-05 (November 1993), and Guidance Manual for Areawide Travel Demand Management
Programs, (October 1992).
-7-
-------
The decision to build a new modeling tool, rather than simply modifying and adopting
the FHWA TDM model, was based on the following factors:
• The FHWA TDM model was not designed to calculate emissions.
• The FHWA TDM model, while much simpler than the four-step models, was
still designed primarily to work with trip tables, and was also seen as more
complex than some members of the target audience would be able to work
with.
• The FHWA TDM model lacked a modern "Windows"-type user interface,
meaning the user would not be able to use a mouse or other convenience
procedures to ease application.
• It was seen as desirable to recheck and, as necessary, update or enhance the
calculation procedures or impact values coded into the TDM model.
The resultant COMMUTER model has the structure and general features pictured in
Figure 2-2. It is essentially a three-step procedure:
1. The user establishes a baseline by supplying essential information on local
conditions.
2. An analysis scenario is selected from among available options.
3. Impacts on vehicle trip making, VMT, and its distribution between peak and
off-peak travel periods are calculated and used to estimate the change in
emissions of volatile organic compounds (VOC), carbon monoxide (CO),
oxides of nitrogen (NOx), carbon dioxide (CO2), particulate matter (PM2.5),
and six air toxics.
Each of these steps is described briefly below, simply to provide an introduction to how
the overall technique operates. Subsequently, each remaining numbered section of this
Procedures Manual describes in some detail each of the individual steps and procedures.
The objective is to provide the analyst with a detailed explanation of how the model is
operating, the assumptions employed, and the data and methods that are used to calculate
the particular impact. Numeric examples are provided in the individual topic sections to
illustrate each module and its calculation procedure, supplemented in most cases by
sample applications across a range of starting conditions or values of the particular
measure to provide insight as to the measure's and the model's sensitivity. Uncertainties,
where important, are identified and described.
-------
Figure 2-2
COMMUTER Estimating Procedure
®
efine Scenario & Compute Travel Impacts
Q.
£
-------
illustrations of the respective model screens or functions, with direct instructions on what
data are needed and where and how they should be entered.
Baseline Data Requirements
The COMMUTER user is required to provide some basic information to the model in
order to establish a starting point, or baseline, from which to measure changes, as well as
to communicate some important local conditions that are used in performing the analysis.
This information consists of the items shown in box number © in Figure 2-2. The
required items are summarized briefly below.
• Area Size and Type: Qualifies the area as Large (over 2 million), Medium
(750,000 to 2 million), or Small (under 750,000), and whether the analysis is
being directed at the regional downtown core, a medium-density area or
activity center, or a low-density suburban area.
• Analysis Scope: Whether the analysis is being applied at the scale of an entire
region, for a specific employment site, or some variation in between.
• Employment Base: Since the analysis is being directed at work travel, it is
important to know the size and composition of the regional (or other
application scale) employment base. In particular, it is important to know the
number of employees working in "Office" as opposed to "Non-Office"
occupations, and also the proportion of this base that is expected to be affected
by the strategies under consideration. A discussion of the important distinction
between total and affected employment is presented in Section 3.
• Starting Modal Split: The existing modal shares for work trips in the study
area, specifically percent of persons making work trips by auto drive alone,
auto carpool, vanpool, transit, walk, bicycle, and other .
• Vehicle Occupancy: Average number of persons traveling in a carpool unit or
a vanpool.
• Average Trip Length: Average trip length (distance), for all work trips
(including transit), and separately for vanpool, bicycle and walk.
• Peak Period Duration: Length of weekday peak travel period in hours.
• Peak/Off-Peak Work Trips: Percent of daily work trips that are made during
peak travel periods.
• Peak/Off-Peak VMT: Total daily VMT from work trips that occurs during
peak and off-peak travel periods.
• Peak/Off-Peak Travel Speeds: Average speeds on freeways and non-freeway
roads, during peak and off-peak travel periods.
-10-
-------
The need for this information is explained in greater detail in Section 3, and specific
instructions on where to obtain the information and how to use it in the model are spelled
out in the Users' Guide. Default values are provided in the model for many of the items,
although users are strongly urged to enter their own unique information. Some special
instructions apply to users of the COMMUTER model who are regional planning
agencies vs. employers and other less experienced in transportation planning methods
and data.
Selection of Strategies and Development of Scenarios
The COMMUTER model allows the user to select from and test a variety of strategies.
As shown in Figure 2-2, these include (as numbered in the diagram):
• Employer TDM Support Strategies: Non-monetary inducements to
encourage employees to use alternative modes rather than drive alone. These
include rideshare matching services, vanpool formation assistance, on-site
transit information and/or pass sales, transportation coordinators, guaranteed
ride home.
• Alternative Work Schedules: Arrangements such as flexible or staggered
work hours, compressed work weeks, and telecommuting.
Travel Time Improvements: On-site or adjacent area modifications to
improve access to work sites from transit, or by walking or biking. Also
includes preferential (close-in/reserved) parking for carpools or vanpools, and
improvements to transit service.
Travel Cost Changes: Measures such as imposition of parking fees,
differential rates or discounts for carpools or vanpools, transit fare subsidies, or
in specific modal incentives or disincentives to any or all modes.
Specific strategy options are discussed in Section 4 for Employer TDM Support, in
Section 5 for Alternative Work Schedules, and Section 6 for Travel Time Improvements
and Travel Cost Changes. With few limitations, the user may mix and match any number
or combination of these strategies within a given "scenario", and obviously can create
and test as many scenarios as desired. By so doing, the user will gain insight into which
strategies are the most effective for their particular situation, and will discover the
flexibility they may have to achieve their desired transportation management goal in
more than one way.
Computational Methods for Estimating Transportation Impacts
-11-
-------
The COMMUTER model escapes the technical and resource intensity of the traditional
four-step modeling approach by taking some calculated shortcuts. These shortcuts retain
the essence of the full-blown approach, but simplify the analysis effort by an order of
magnitude. Of course, because these shortcuts cut back on the many minute calculations
that are normally done in a trip-table-based analysis, a certain degree of uncertainty is
introduced by replacing a large number of individual computations with a single
"aggregate" calculation. However, given the modest increment of change typically
expected from most workplace commuter programs (and for which EPA is prepared to
grant credit), the COMMUTER model is seen as a reasonable compromise between ease
of use and accuracy.
Like the FHWA TDM model on which it is based, COMMUTER'S simplicity derives
from an "incremental" type of analysis procedure, often referred to by transportation
planners as "pivot-point" analysis. Rather than rerun the entire suite of transportation
models in the four-step process (or at least the mode choice and the assignment steps),
which entails making changes in the service variables and computing impacts for each
affected origin-destination pair in the regional trip tables, a pivot-point approach simply
extrapolates from the existing condition, or baseline. The degree of incremental
adjustment in this extrapolation depends on the current modal share balance, the types of
strategies that are being tested, to what modes and with what intensity they are being
applied, and what is known about the relationship, or sensitivity, between the particular
strategy (or service characteristics it affects) and demand for the mode. For modest
changes in mode shares, such as would be expected with many TDM-type strategies, this
incremental extrapolation is a convenient and acceptably accurate alternative to the more
rigorous analysis methods.
There are various ways this incremental adjustment/extrapolation can be done, some
much more analytically sophisticated than others.
• Elasticities: Applying a relationship derived from observing degree of change
in behavior in response to a change in an underlying variable.
• Share Adjustment Relational Factors: When formal elasticity relationships
have not or cannot be developed because of insufficient data or research on a
relationship, or because individual relationships cannot be isolated with
statistical confidence, factors may be developed that suggest the type of change
that would be expected in mode share based on a particular type of action
taking place, which may vary with the type of mode, setting, packaging
assumptions with other strategies, presence of a regulation, or type or size of
employer.
• Multimodal Travel Demand Models: Coefficients may be taken from
multimodal travel demand models, and applied either to an individual mode or
strategies or to multiple modes and strategies simultaneously.
The preferred approach among these options is the use of coefficient-based travel
demand methods. Elasticities can be very helpful as planning aids, since they provide
-12-
-------
some statistical basis for inferring a relationship between travel demand and a
transportation service or cost change. Where elasticities are limited is in being able to
take into account the interactive effects that occur when multiple actions are applied or
multiple modes are involved. For example, if a setting had two modes—auto and
transit—and two strategies were proposed—an auto parking charge along with a 10-
minute reduction in transit travel time, an elasticity method would not be able to project
the combined effect of these strategies on the subsequent demand for auto and transit. In
contrast, a coefficient-based multimodal approach has the ability to look at the combined
effects of multiple strategies and deal realistically with the cross-modal effects. The
relational factors approach is a fallback procedure when formal statistical relationships
have not yet been developed.
As with the TDM model, the COMMUTER model uses two of the above procedures for
calculating travel response to workplace commuting strategies:
• Logit Pivot-Point Model: A multimodal pivot-point model using coefficients
and computational procedures from accepted logit-based mode choice models;
and
• Look-Up Tables: Relational factors from empirical research, arrayed in look-
up tables where increments of change are associated with particular types of
programs, reflecting different application assumptions, levels of intensity, and
setting.
Figure 2-2 classifies the various strategies incorporated in the COMMUTER model into
four groups. The classification system mainly relates to how the strategies are analyzed
in the model. The first two—Employer TDM Support Programs (2), and Alternative
Work Schedules (3)—are analyzed using relational factors in look-up tables, with a
normalization procedure applied to the adjusted shares to ensure that changes are
proportionate across the available alternatives and do not allow final choices to exceed
100%. The strategies that involve changes to either travel time or cost, however,—
Travel Time Improvements (4) and Travel Cost Changes (5)—are analyzed through the
more rigorous logit pivot-point procedure. Each of these is described briefly below, and
then in more detail in the individual report section that deals with that strategy type.
Share Adjustments for Employer Support TDM Actions Using Look-Up Tables - While
virtually any attribute that plays a role in travel choice can be included in the structure of
a logit model, usually the models are limited to only factors that are related to travel time
or cost. Partly this is because there has not been great interest in trying to include such
diverse and specialized strategies as carpool matching programs and employer
transportation coordinators in regional travel models, and partly it is difficult to
incorporate these relationships within standard mode choice models for reasons of
available or appropriate data, or being able to demonstrate a functional relationship with
acceptable statistical confidence. Transportation analysts and economists have
commonly concluded that the biggest changes in travel behavior are those induced by
changes in the underlying economics among the alternatives, such as policies that try to
lessen the travel time disadvantage of using transit or carpool over driving alone, or
-13-
-------
seeking parity in the cost of travel by raising the cost of driving alone or discounting the
cost of traveling by alternative modes.*
As with the TDM Model, the COMMUTER model categorizes the various Employer
TDM Support Measures in group 2 into four discrete categories of effort, presented as
"levels." Each level constitutes a program of actions undertaken by the employer that is
intended to enhance the attractiveness of alternative modes and thereby increase the
likelihood that employees will use that mode over driving alone. Level 1 programs
reflect minimum effort by the employer, and hence produce the smallest change in
behavior, whereas Level 4 programs reflect the maximum effort and maximum travel
behavior impact. Specific named strategies describe what is assumed to be provided in
each level, with higher-level programs generally including everything that was offered in
the lower-level program, as well as some new measures or special enhancements.
There are individual support programs for each alternative mode except pedestrian:
Carpool, Vanpool, Transit, Bicycling. The support programs are also specific and
relevant to the individual mode, such as on-site pass sales being part of the Transit
support program. Invoking any of the Support Programs triggers a procedure that adds
an increment of mode share to the mode that is receiving the support, with the adjustment
increments being greater for higher levels of each program. The adjustment values are
stored in look-up tables for easy access and use by COMMUTER. A simple illustration
of how this analysis would be done in the COMMUTER model is provided in Figure 2-3.
When a scenario is designed that includes support measures for more than one mode at a
time (a common occurrence), the COMMUTER model must distribute the impact across
the modes. The model does this by adding the shares to the modes being targeted, and
then readjusting the shares of all modes—those targeted and those not—such that the
total for all modes once again equals 100%. This requires a proportionate adjustment
across all modes, which means that modes with the largest starting shares (such as drive
alone) when asked to forfeit shares will lose the greatest absolute number of users, while
small share modes (like vanpool) will lose the least. As part of the normalizing process,
even the targeted mode will have to forfeit some of its new share to ensure proportionate
redistribution.
See, for example Part III: Synthesis of Findings in the report: COMSIS Corporation, Implementing
Effective TDM Measures, for the Federal Highway Administration, DOT-T-94-02. Section 3 (September
1993). Also, in work done for the California Air Resources Board and the South Coast Air Quality
Management District by COMSIS in 1993 for the purpose of developing new plan development and plan
review software for Employer Trip Reduction Programs under Regulation XV, extensive analysis of Reg.
XV employer plan data from SCAQMD files coupled with new survey data from 43 California employers
failed to find significant impacts on travel arising from a variety of popular non-monetary support type
programs offered by employers.
-14-
-------
In Figure 2-3, a starting balance of modal shares is shown in the box at the upper left,
with a pending program of TDM Employer Support measures described on the right. The
example suggests that the employer (or group of employers) will implement Level 2
Carpool, Vanpool, and Transit support programs, and a Level 4 Bicycle support program.
The lookup tables show these options being selected for the test scenario: the Level 2
Carpool program contributes a 2% increase to the current 13% carpool mode share,
Level 2 Vanpool contributes a 1% increase to the existing 1% vanpool, Level 2 Transit
contributes 2% to the current 5% share, and Level 4 Bicycle contributes 1% to the
current
1% bike share. As shown in the Share Adjustment Process box in the lower half of the
figure, the Support Program share increments are added to the base shares, and a new
total computed. Obviously, the new total with the program increments added will exceed
the total for the baseline distribution; the baseline would normally total 100%, except
that a special treatment is provided for the Walk mode. Walk is the only mode for which
TDM support enhancements have not been developed, raising the concern that
walkers—like solo drivers—can only "lose" ground when support measures are applied
to the other alternatives, since those adjustments would otherwise come out of the Walk
share of 4%. To prevent this from happening until such time as a set of support factors
can be developed for Walk, the Walk mode is kept out of the mode share adjustment
process. Its 4% is shown in parentheses in the table, indicating this treatment.
With Walk out of the equation, the baseline mode shares total 96%, and the new mode
shares total 102%. An adjustment must be made to re-apportion the gains (or losses for
SOV) to account for the reality that (1) the total cannot exceed 100% (96% in this
example), and (2) that the modes will "compete" for the new shares. In other words, if an
improvement is made to Carpool as an option, it would be expected that it would attract
new users not only from solo drivers, but also in some measure from Transit and the
other modes if indeed Carpool had received a comparative advantage. And if both
Transit and Carpool received improvements, both would be likely to "steal" share from
SOV, but also from each other. The procedure to accomplish this redistribution in the
COMMUTER model for support programs is to "normalize" the gross adjusted shares.
In the example, the previous total of 96% is divided by the equivalent new total, or
102%, producing an adjustment factor of 0.941. Each share—except for Walk—is
subsequently multiplied by the adjustment factor to obtain the final revised share
expected from this scenario. As may be seen, the biggest adjustment is taken from Drive
Alone, which falls from 75% to 70.6%, while Carpool rises from 13% to 14.1%, Transit
rises from 5% to 6.6%, Vanpool and Bicycle go from 1% to 1.9%, and Other falls from
1% to 0.9%.
Section 4 provides greater detail on this procedure as it applies to Employer TDM
Support Programs. Specifically, the actual look-up reference tables used in the
calculations are provided, along with key related assumptions regarding characteristics of
the employment base, and special considerations that apply to using the model for an
individual employer vs. applying it at a regional or multi-employer level.
-15-
-------
Figure 2-3
Illustration of Share Adjustment Process
Starting Mode Shares:
Drive Alone:
Carpool:
Vanpool:
Transit:
Walk:
Bicycle:
Other:
75%
13%
1%
5%
4%
1%
1%
Carpool TDM Support Program
A Share
Program Level
1
+1.5
2
+2
3
+3
4
+5
Vanpool TDM Support Program
A Share
Program Level
1
+.5
2
+1
3
+1.5
4
+2
Transit TDM Support Program
A Share
Program Level
1
+1
2
+2
3
+4
4
+5
Bicycle TDM Support Program
A Share
Program Level
1
+.1
2
+.3
3
+.5
4
+1
Share Adjustment Process:
Adjustment
Drive Alone:
Carpool:
Vanpool:
Transit:
Walk:
Bicycle:
Other:
TOTAL
Base
75%
13%
1%
5%
(4%)
1%
1%
96%
A
+2%
+1%
+2%
+1%
Revised
75%
15%
2%
7%
(4%)
2%
1%
102%
Factor
.941
.941
.941
.941
1.0
.941
.941
Final
Shares
70.6%
14.1%
1.9%
6.6%
(4%)
1 .9%
0.9%
100.0%
Trip Adjustments for Alternative Work Hours Strategies - The procedure is virtually the
same as applied to Alternative Work Schedules, except that the changes that are induced
are solely to shift the given trip from the peak period to the off-peak period. No mode
shifts are calculated in conjunction with these strategies. The benefit in terms of
emissions, therefore, is not in the elimination of a trip but in its movement to a less-
congested travel period. With less congestion, vehicle speeds will be higher, which will
affect the emissions rates in the model. In most cases, as with VOCs, the speed change
results in fewer emissions, while in others, say with NOx, a higher speed may increase
emissions.
-16-
-------
As with the Employer TDM strategies, the COMMUTER contains a separate look-up
table for each strategy: Flexible Hours, Staggered Hours, 4/40 and 9/80 Compressed
Work Weeks, and Telecommuting. However, unlike the Employer TDM procedure
where the user selects a particular "level" of program package and intensity, with the
Alternative Work Hours strategies the user indicates only what percentage of employees
will be "eligible" for each program type. The procedure then uses this percentage to
determine how many trips are eligible for the given program, e.g., 4/40 work weeks, and
applies a factor that calculates how many of these trips will be shifted outside the peak.
The trips that are shifted from peak period to off-peak period are shifted at the current
mode split. In other words, if 75% of all trips in the peak period were Drive Alone, then
the model assumes that 75% of those shifted will be Drive Alone, with the remainder
comprised of Carpool, Transit, Bike, Walk, etc.
The procedure guards against an illogical number of employees participating in Work
Hours programs, to prevent double counting. However, users are able to test the
assumption that workers may be eligible for more than one Work Hours program, leading
to the result that the total eligible for Work Hours programs can exceed 100%. To
prevent against the number of people actually participating in these programs exceeding
this logical limit, the final calculated participation is determined by normalizing the
eligible percent to an assumed maximum participation of 100%. An example of this
procedure and illustration is provided in Section 5, along with presentation of the look-up
table values and underlying assumptions.
Analysis of Travel Time and Cost Strategies Using Logit Pivot-Point Procedure - A
certain class of travel demand models, known as "logit" models, provides relationships
that can be tapped for this complex analysis. These models are rooted in economic utility
theory, and assume that the attributes that comprise each travel option carry a certain
amount of utility, or value, to the customer. If the attributes for each alternative were
rolled up into a single measure of utility for each alternative, the theory holds that the
consumer will choose the alternative that provides the greatest overall utility. In the case
of travel mode choice, where the attributes of importance to the customer are the mode's
time and cost, the traveler would be expected to select the mode that minimizes his/her
time and cost. Higher values of time and cost are undesirable, and may be viewed as a
"disutility"; hence, the choice process becomes one of minimizing disutility, rather than
maximizing utility. This distinction is mainly a matter of semantics, in order to be
consistent with the terminology that will be used to explain the model.
Logit models do a remarkably good job of predicting what mode an individual traveler
will choose from a set of alternatives or, when applied in a regional context, what choices
an entire population will make on a proportional basis from an array of alternatives.
When applied for a population, the model estimates not a single preferred mode, but the
percentage of travelers who will choose each of the offered modes, resulting in an
estimate of modal split.
The mathematics behind the logit model may seem complicated, but the principal by
which it works is not too hard to grasp. The model computes the probability (a number
between 0% and 100%) that Mode^4 will be chosen when it is compared to its competing
-17-
-------
alternatives, Modes B, C, and D. It does this by first calculating the utility (disutility,
actually) of each alternative, or U, for a given trip. A typical disutility expression might
look like this:
U (Mode ,4) = a0 + ttj (In-Vehicle Travel Time) + a2 (Walk/Wait Time) + a3 (Cost)
The disutility of Mode A is a weighted sum of the travel time and cost incurred in using
Mode A for the given trip. The "weights" are provided by the model coefficients, which
are the oc0, a1? a2,and oc3 terms in the disutility equation above. The coefficients relate the
sensitivity of the traveler to the particular characteristic, i.e., signifying how important it
will be in the choice process. The values of the coefficients are determined through a
statistical estimating procedure (known as maximum likelihood estimation) applied to
data from local travel surveys. Once the coefficients have been determined, the
calculation of the probability of Mode A (or any of the other modes) being chosen is
computed as:
Utility Mode A
C
F (MOde A)- utilityModeA UtilityModeB UtilityModeC UtilityModeD
C T C T C T C
To use the logit model to analyze the effect of any policy or program, it is necessary to
change the disutility expression for each affected mode by relating the time or cost
change from the strategy policy to each individual travel unit in the analysis. If the travel
universe is an individual employer, the unit of analysis is the individual. If the universe
is a region or subarea, the unit of analysis would be each of the individual origin-
destination pairs that is influenced by the policy.
Because of the exponential form of the logit equation, the mode choice predictions from a
logit model produce an S-shaped curve, as illustrated in Figure 2-4.
This relationship is clearly non-linear, and as the figure implies, identical changes in the
Figure 2-4
Logit Relationship
100%
Probability of
Choosing
Mode a
0%
Dominant Share
Moderate Share
Limited Share
Disutility of Mode a
-18-
-------
travel characteristics of a given mode will not produce the same degree of change in
mode share in all cases, but will depend greatly on where the traveler is "starting" on the
S-shaped curve. In cases where the starting share is limited, say 1% to 3%, an
improvement of X minutes of travel time or Y dollars of travel cost would not produce as
large an increase in share as would occur if the mode were starting with a moderate
share, say 5% to 10%; a starting point in the dominant range, say 30% to 60%, would
produce the most substantial changes. Above a certain mode share the mode is so
dominant (so free of competition) that additional improvements in service or cost will
start resulting in modest or decreasing returns. The implications of this relationship are
important: if a program is to be sited in an area where existing transit or carpool use is
high, then strategies that provide additional advantages to those modes should result in
healthy increases. However, if a program is to be cited in an area where solo driving is
dominant and there is little or no transit or carpool use, significant enhancements and
incentives would be required to obtain the same shifts of travelers to transit or carpool.
While logit mode choice models are powerful and accurate, traditional application at an
individual origin-destination pair level would require significant data development effort
for every scenario tested. However, the tool can also be applied with reasonable
accuracy to the existing mode shares, by setting up the model to predict instead the
change in disutility that would result from the change in the individual attributes:
A U (Mode A) = A o^ (In-Vehicle Travel Time) + A a2 (Walk/Wait Time) + A a3 (Cost)
Instead of having to develop revised disutility expressions for every mode and every
analysis unit, the analysis can merely "pivot" from the existing share by knowing the
value of the coefficients and the degree of change that will be introduced by the
individual strategy or program. Moreover, in this equation, more than one service
variable can be changed at a time (e.g., walk time and cost), and more than one mode can
be affected at a time, and the choice probabilities for all modes may be calculated simply
by making the appropriate modifications to each mode's disutility expression. This
feature promises a realistic accounting for the competition among modes, and reflects the
gain/loss of different strategies applied in different measures to several modes.
In the COMMUTER model, the logit pivot-point procedure is used only to calculate the
impacts from the Travel Time Improvements or Travel Cost Changes strategies. The
application of logit pivot point to these strategies in the model is discussed in detail in
Section 6 of the Manual.
Sequencing Order of Calculations in the COMMUTER Model - The order in which the
COMMUTER model performs its calculations of travel changes is as follows:
1. It first calculates the changes due to Alternative Work Hours. This serves to
readjust the travel population baseline to determine how many trips will be
shifted to the off-peak, and how many will remain in the peak period and be
subjected to application and analysis of the mode-choice oriented strategies.
-19-
-------
2. Next, mode shares of the remaining peak trips are readjusted to reflect the
effects of the Employer TDM Support strategies.
3. All time and cost related strategies are tallied up and brought into the logit
pivot-point procedure, which is then applied to the revised mode share starting
point from step 2.
Emissions Calculations
Separate calculations are performed for each of the two activity parameters modeled:
VMT and trips. Emission reductions from each of these travel impacts are then added
together to produce combined emission reduction impacts from changes in travel due to
the TCMs evaluated. Each of these calculations is performed for both the peak and off-
peak periods.
First, reductions in VMT and vehicle trips due to the TCM strategies are computed by
subtracting existing activity levels from the "final" (after-TCM) activity levels.
Second, VMT-based emissions changes are calculated for peak and off-peak periods,
based on changes in VMT and on average regional speeds by period and facility type.
Emission factors (expressed in grams/mile) are provided with the model in 5-mph
increments and interpolated to represent emission factors at the actual speeds provided.
Third, start-based emissions changes are calculated for peak and off-peak periods, using
start emission factors (expressed in grams/start) and the change in number of starts.
Finally, the VMT- and trip-based emission reductions are summed together. Daily
reductions are also summed from peak and off-peak reductions.
Model Outputs
Key outputs include the following:
• Baseline and final mode shares for each mode, including percent of trips
eliminated;
• Percent of trips shifted from the peak to off-peak period;
• Baseline and final peak, off-peak and daily VMT;
• Baseline and final peak, off-peak and daily vehicle trips; and
• Changes in total daily emissions for each pollutant.
###
-20-
-------
3. Assembling Required Background Information
Overview
Types of Data Required - An analysis of commuter choice strategies using the
COMMUTER model requires certain basic information in order to establish the travel
baseline and to properly initialize some key parameters that are used in the analysis. This
section of the report lists each of the required items and explains why it is needed and
how it is important to the analysis. Items included in this list and discussed in this
section include the following:
• Metropolitan Area Size
• Application Setting Characteristics
• Affected Employment
• Mode Choice Model Coefficients
• Starting Mode Shares
• Average Trip Lengths
• Vehicle Occupancy
• Peak and Off-Peak Travel Characteristics
The discussion of input data requirements in this Procedures Manual focuses mainly on
confirming what information is required and optional ways to supply that information, in
relation to the methods being applied by the COMMUTER model. The accompanying
Users' Manual offers the procedural step-by-step guidance to the user on where to find
this information, how to format it properly, and how to enter it into the COMMUTER
Model.
Areawide vs. Individual User - The COMMUTER model has been designed for
application at either a regional (multi-employer) level or for an individual employment
site. While the types of strategies that might be evaluated for these two applications are
roughly the same, the input data requirements are somewhat different. Since an
application at a regional or areawide level requires an integration of results over a large
number of employers, it is necessary to have information that describes the size,
composition, and travel characteristics of that population. For an individual employer,
the regional composition characteristics are not relevant, and it becomes more important
to be able to describe in some detail how that particular set of employees is traveling.
Therefore, regional (or multi-employer) users of the model would be expected to provide
information that describes the corresponding employment base, and certain assumptions
about applicability of particular strategies to these employers, or the extent to which
these strategies are already in place. Much of this information will be derived from the
region's or local area's transportation planning process, which is the same information
that is used to prepare the SIP or the conformity analysis. Since regional COMMUTER
users would likely be the designated MPO, access to the appropriate information should
be fairly straightforward. For others who might conduct a COMMUTER analysis at
-21-
-------
either a regional or areawide level, such as departments of transportation, municipalities,
counties, employer organizations, or air agencies, it is expected that these groups would
need to work with the regional MPO to obtain this information and ensure its proper
interpretation.
Those users performing a COMMUTER analysis at an individual employment site will
not require most of the compositional or travel information needed for the areawide
analysis, but will require some fairly specific information on their individual case.
Primarily, the individual site users will need information on the travel characteristics of
the respective employees. This includes the following:
• Total employees working at the given site;
• Total traveling to the site on a given workday;
• Mode choice (percent by mode);
• Number traveling to the site during peak period;
• Number engaged in telework, compressed work weeks, or other alternative
work schedules; and
• Existence of staggered work hours or shifts, and percent of employees
accounted for in each.
This information is generally obtained through employee travel surveys, so the individual
site user will not need to be concerned with regional transportation data, models and
planning procedures. Guidelines are provided in the COMMUTER model and the Users'
Guide to help users determine what data they need, how to obtain it, and how to use it.
Default values are provided for relationships or data inputs that might not be readily
available. However, even though the COMMUTER model gives the user a considerable
degree of independence and flexibility, an individual site user is still strongly urged to
consult with the relevant local planning agency, for the purpose of answering questions
about assembly of input data, assumptions about background conditions, advice on
selection of program strategies to be tested, or interpretation of results. In particular,
employers should consult their planning agencies to obtain or confirm the model
coefficients that will be used in the analysis (see Section 6 for more discussion on these
coefficients).
Where differences occur in the data requirements or responsibilities between regional vs.
individual users, the Procedures Manual and the Users' Manual will point out these
differences and explain how each user should proceed.
Default Values - While most of the input data items that are requested by the
COMMUTER model should be relatively easy to obtain, the COMMUTER model also
provides default values for many of the items. These default values have been provided
mainly as a convenience to the user, to encourage use of the model by a broad audience
whose familiarity with transportation issues or access to certain data may be limited. The
defaults may also be of service in performing reasonableness checks of any user-supplied
data.
-22-
-------
While these default values are provided to help the user conduct an analysis without
undue difficulty or expense in obtaining local data, users are nevertheless urged to seek
out and utilize current local information wherever possible. This will increase the
validity of the resulting travel and emissions impacts, and help EPA reach a favorable
determination when reviewing the applicant's request for COMMUTER emissions
credits. Those measures for which default values are provided are listed in Table 3-1.
This listing shows the default values currently assigned to each measure, and the source
of the information used to develop the default value.
Users may wish to experiment with the COMMUTER using the default values in order to
develop familiarity with its operation and the types of impacts it projects for various
types of programs. However, programs submitted for air quality credit should give
evidence of due effort to utilize local data wherever possible.
Sources for Background Data - For regional or areawide application of the COMMUTER
model, most of the necessary background data should be available to or through the
responsible transportation planning agency in the area. This will normally be the
regional MPO, although county or municipal planning agencies or departments of
transportation may also be a resource depending on the extent of their planning activity.
These organizations should be capable of supplying information on regional employment,
modal split, VMT and travel speeds, such as are required by the model. For individual
site users, most of the background information will be obtained through an employee
travel survey.
Required Model Inputs
The remainder of this section is devoted to identifying each of the required data items
and providing information on typical sources. The availability of defaults for each item
is noted. In addition, a glossary of transportation planning terms is provided in Appendix
A to clarify the meaning of any terminology that the user may not be familiar with.
Metropolitan Area Population Size Category - The COMMUTER model distinguishes
among three size categories of metropolitan area when evaluating certain impacts:
• Large: Population of 2 million or more
• Medium: Population under 2 million but > 750,000.
• Small: Population under 750,000.
-23-
-------
Table 3-1
Default Values Provided in the COMMUTER Model
Data Item
Work Trip Mode
Shares
Average Commute
Trip Length (in
miles)
Average Vehicle
Occupancy (persons
per vehicle)
Duration of Peak
Period
Percent of Work
Trips Occurring in
Peak Period
Average Travel
Speeds (mph)
Affected
Employment
Suggested Default
Auto - Drive Alone 78.2%
Auto-Carpool 12.1%
Vanpool 0.5%
Transit 4.9%
Bicycle 0.4%
Walk 3.0%
Other 0.8%
Total 100.0%
Average Person-trip Length: 12.7
Average Trip Length - Carpool: 12.0
Average Trip Length - Vanpool: 20.4
Average Trip Length - Transit: 11.7
Average Trip Length -Bicycle: 2.9
Average Trip Length - Walk: 0.9
Average Carpool Occupancy: 2.25
Average Vanpool Occupancy : 7.19
3.0 hours
61.4%
33.8%in6-9A.Mpeak
27.6% in 4 - 7 P.M. peak
Area Freeway Freeway Arterial Arterial
Size Peak Off-Peak Peak Off-Peak
Large 40.2 49.022.6 25.1
Medium 46.8 54.723.2 25.5
Small 50.5 56.424.2 26.0
Distribution % only, user must supply total
Office occupations: 79.7%
Non-Office occupations: 20.3%
Source
2000 U.S. Census
2001 Nationwide
Household
Transportation
Survey (NHTS)
2000 Census (2-4
person carpool)
2001 NHTS (5+
person carpool)
1990 Nationwide
Personal
Transportation
Survey (NPTS)
Databook, Table 6-
32
2002 Highway
Performance
Monitoring System
(HPMS)
1997 Statistical
Abstract of U.S.,
Table No. 660
The primary use of this information is to select appropriate default coefficients for use in
the logit mode choice pivot point model, which is used for evaluating time or cost-based
strategies. However, since users will generally be expected to select and use the unique
coefficients that have been specifically developed for their own region, and used in the
officially sanctioned regional planning process, this information item is mainly for
-24-
-------
default situations only. This guidance on coefficient use is covered explicitly in Section 6.
Analysis Scope - This information item indicates to the COMMUTER model whether the
analysis is to be performed at a regional or site-specific level. This selection by the user
signals the modeling procedure to expect input data and specification of strategies to
correspond to a single employment site or a regional population of employers. This
affects what strategies are eligible, the data required of the user, and the calculation of the
results.
Analysis Area Type - The COMMUTER model has a "placeholder" for this input, but
does not currently use it in the travel impact calculations. If this input were active, it
would work as follows. For certain types of applications, the model would need to know
the "type" of area in which the program would be sited, with the options being:
• Central business district (CBD) or radial corridor;
• High density activity center or suburban CBD (other than the regional CBD);
or
• Suburban, low-density area.
This information is important because programs sited in these three different types of
areas are likely to have very different impacts in response to the same strategies, and as a
result, certain strategies may be more effective in one environment than another. In
regional CBDs or travel corridors into those areas, conditions generally favor transit and
carpool use, and so existing mode shares for transit and carpool are probably the highest
in the region. In contrast, suburban low-density areas typically show great reliance on
private vehicle travel, while suburban downtowns or major activity centers fall
somewhere in between. This distinction is important since the COMMUTER model's
pivot-point procedures calculate changes in behavior based on what was there before.
This was discussed briefly in Section 2 in the modeling discussion.
In most cases, supplying this information to the model will not be necessary for it to
correctly compute its results. This is because the travel input data will generally inform
the model of what the starting mode shares are. If the user is an individual site, the
employee travel survey data will contain the correct starting mode shares; if the analysis
is for the overall region, the work trip mode split from the regional transportation
planning process will provide the appropriate starting point. Where this item may
become important is if an analysis were being done for a subarea or municipality
involving multiple employers, where the total employment was known but the modal
shares were not. In this instance, the use of this parameter would instruct the model to
provide appropriate default mode shares for this type of environment as represented by
similar sites in a national sample.* This is discussed in more detail under "starting mode
shares" below.
2000 Census journey-to-work statistics.
-25-
-------
When representative default mode shares by area type can be developed in a future
version of the COMMUTER model, this input will be activated as described above.
Affected Employment Base - For regional (or any other areawide or multi-employer)
analyses of strategies, it is necessary to describe the employment base that will be subject
to the program. The COMMUTER model requires the following information:
• The total number of employees in the employment base to which the program
actions will be applied, referred to as affected employment; and
• The proportion in occupations defined as "Office" or "Non-Office" functions.
The total number of employees establishes the size of the traveling population under
study, and is used to calculate baseline vehicle trips, VMT, and emissions. This number
may constitute the entire regional employment population, or it may be only a portion of
that population based on locally established rules of eligibility. For example, it might be
assumed that employers below a certain size would not be expected (or encouraged) to
participate in voluntary commuter choice programs. For example, nationally more than
70% of all establishments have fewer than 10 employees, however, these employ only 12
percent of the workforce .* Therefore, some users may wish to restrict assumptions on
program coverage to only employers of 10 or more. Others may wish to restrict
applicability to only those workers who travel to a work site during peak travel hours (6
to 9 a.m., and 4 to 7 p.m.); nationally, only 61.4% of workers travel during these hours,
with 38.6% traveling at some other time.**
The calculation of impacts also takes into account differences between workers employed
in "Office" vs. "Non-Office" occupations. Generally, research on TDM program
effectiveness has shown that many strategies are more relevant and have greater impact
in "Office" occupations, consisting of clerical, professional and managerial fields, than in
"Non-Office," consisting of sales and manufacturing occupations. The FHWA TDM
model, from which this procedure was adopted, defined Office and Non-Office through
SIC (Standard Industry Classification) codes, as shown in Table 3-2. Of Course, SIC has
since been replaced by the North American Industry Classification System (NAICS),
however, the descriptions and breakdown in table 2-3 are still useable in setting
parameters for modeling.
* U.S. Bureau of Labor Statistics, Occupational Employment and Wages, May 2003; Bulletin 2567;
September 2004
" 1977 Statistical Abstract of the United States, Table 662.
-26-
-------
Table 3-2
Definition of Office and Non-Office Employment
Type
Office
Non-Office
SIC Code
40-49
50-59
50-59
60-69
70-89
90-99
10-14
15-17
24-39
Industry
Transportation & Public
Utilities
Wholesale Trade
Retail Trade
Finance, Insurance,
Real Estate
Services
Government
Total Office
Mining
Construction
Manufacturing
Total Non-Office
1996 Employment
6,316,000
6,587,000
21,597,000
6,977,000
34,359,000
19,461,000
95,296,000
570,000
5,407,000
18,282,000
24,259,000
Percent
5.3%
5.5%
18.1%
5.8%
28.9%
16.3%
79.7%
0.5%
4.5%
15.3%
20.3%
Source: 1997 Statistical Abstract of the United States, Table 660.
An extremely important issue with regarding this input is that it does not represent total
regional employment. The box below addresses this distinction.
"Affected Employment"
In addition to asking the regional user for information on overall employment size and
distribution, the COMMUTER model also takes into account what portion of this
population would be affected by a given strategy. It may be assumed that the entire
regional employment base is affected by the strategy, or only some portion, based on
minimum size criteria, location relative to the strategy, assumptions that the strategy is
already being offered, or other factors. This input is not part of the baseline data
requirement for the user, but rather is supplied on a strategy-by-strategy basis once the
user begins to design scenarios for testing. Hence, it is covered in more detail in
subsequent sections that deal with the respective strategy groups. However, it is
mentioned here because it represents another way in which the user can delimit the
employment base to which the strategies are applied, as an alternative to making these
restrictions in the baseline itself.
Existing Work-Trip Mode Shares - The pivot-point analysis techniques used by the
COMMUTER model to estimate travel changes rely on a starting set of mode shares for
the employment population being analyzed. If the application is for an individual site (or
-27-
-------
small number of sites), this mode share information comes directly out of the employee
travel survey data that the user must supply. If the application is for a region or similar
areawide/multi-employer setting, then the starting mode shares must be for the respective
population, and will probably be obtained from existing regional transportation planning
information. The regional MPO is the most likely source for this information, although
subarea analyses may find source data from county or municipal planning agencies,
and/or existing employer travel surveys.
The COMMUTER model calls for the percentage of daily work trips that are made by the
following modes:
• Auto - Drive Alone
• Auto - Carpool
• Vanpool
• Transit
• Bicycle
• Walk
• Other
Most major metropolitan areas will have information that can be used to calculate mode
shares for home-based work travel (HBW) for at least auto driver, auto passenger, and
transit. Vanpool is typically not broken out as a separate mode, nor is bicycle or walk,
and so assumptions will have to be made as to baseline shares of these modes. Auto
passenger may be used along with auto driver to determine carpool mode share (where
carpool is not reported separately), but typically this calculation cannot reveal the share
of multi-passenger trips that are vanpool. Where local information on mode shares for
particular modes is not available, defaults in the model have been provided, and the user
can opt to use the default shares for the absent modes and adjust the share values for the
local "known" modes proportionately.
The COMMUTER model also makes an adjustment internally to account for workers
who regularly work at home. This proportion averaged about 3.26% in the 2000 Census
Journey to Work. It nets out this percentage from the total employment base when it
calculates baseline work trips and VMT, and thereafter applies the strategies only to
those 96.74% of workers who travel.
The default shares provided in the model have been derived from the 2000 Census
Journey-to-Work files, and adjusted to reflect the work-at-home factor described above.
The default shares for a standard "regional" analysis developed in this manner are shown
in Table 3-3.
-28-
-------
Table 3-3
Default Shares for a Standard Regional Analysis
Mode
Auto-Drive Alone
Auto-Carpool
Vanpool
Transit
Bicycle
Walk
Other
Work at Home
Total
Total Excl. Work at Home
Reported 2000 Census
Commute Choice
75.70%
11.71%
0.48%
4.73%
0.38%
2.93%
0.81%
3.26%
100.00%
96.74%
Excluding Work
at Home
78.25%
12.11%
0.49%
4.89%
0.39%
3.03%
0.84%
100.00%
For individual site applications, shares for all modes of interest should be known directly
from the data obtained through employer surveys. However, the defaults can be used to
fill gaps as described above for the regional user.
Logit Model Coefficients - The COMMUTER model calculates changes in travel from
time- or cost-based strategies through a logit mode choice model procedure. The critical
elements in this procedure are the coefficients, which supply the "weights" that signify
how important each choice variable is to the traveler. The model employs coefficients
for In-Vehicle Travel Time, Out-of-Vehicle Travel Time (including Walk Time and
Transit Wait Time), Auto Parking Cost, and Transit Fare. These coefficients are used in
a metropolitan area's transportation planning model, and are developed uniquely for each
area from analysis of local survey data.
The COMMUTER model provides coefficient values for the user through a look-up
menu. The coefficient values for most U.S. metropolitan areas where the COMMUTER
model might be used are included in the menu. When setting up to use the model, the
user should designate those coefficients for their respective metropolitan area from the
menu, and the model will automatically bring them into play.
For users who do not find their area listed in the menu, there are two options. The
preferred option is to contact the responsible regional planning agency and obtain the
correct coefficients from the local model. The backup option is to make use of the
default coefficients that have been programmed into the model. A separate set of
coefficients is provided for Large, Medium and Small metropolitan areas. For users who
are able to obtain coefficients from the local planning agency, but do not have values for
-29-
-------
all of the required coefficients, it is recommended that the default coefficients be used
instead of the partial set of local coefficients. Substituting default coefficients for
missing values, thus creating a hybrid, is discouraged since the two sources of
coefficients may have been developed from very different assumptions.
Section 6 of this report presents further discussion on use of the logit coefficients, and a
listing of those metropolitan areas currently included in the COMMUTER model.
Average Trip Lengths - The model employs information on average trip lengths for all
modes when computing change in VMT. Since each travel mode has a different average
trip length, the shifting of a commute trip from drive alone (or any other mode) to another
mode will constitute a different VMT reduction depending on what modes are involved
in the shift. The model has internally coded default values as follows, which are based
on2001NHTSdata:
• Auto Drive Alone: 12.7 miles
• Auto-Carpool: 12.0 miles (2-4 person carpool)
• Vanpool: 20.4 miles (5+person carpool)
• Transit: 11.7 miles
• Bicycle: 2.9 miles
• Walk: 0.9 miles
• Other: 12.1 miles
Average Person Trip Length: 12.2 miles
(calculated from mode shares and modal trip lengths)
Average Vehicle Trip Length: 12.7 miles
(calculated from drive alone, carpool, and vanpool mode shares, trip lengths and
AVO)
The user is given the opportunity and is encouraged to supply local values for average
vanpool, bicycle and walk trip length, as well as for overall person trip length. These
data may be obtained through the regional planning agency or, for individual site
applications, from respective employee travel surveys.
Average Vehicle Occupancy - Average vehicle occupancy is also used in calculating
changes in VMT as well as vehicle trips. The model provides the following default
values for Carpool and Vanpool occupancy:
Average Carpool Occupancy: 2.25 (source: 2000 Census)
Average Vanpool Occupancy: 7.19 (source: 2001 NHTS)
The user is given the opportunity and is encouraged to supply local values for average
carpool and vanpool occupancy. These data may be obtained through the regional
-30-
-------
planning agency or, for individual site applications, from respective employee travel
surveys.
Peak Period Duration - The model calculates emissions reductions based on changes in
travel during both the peak and off-peak periods. Typically travel conditions are less
congested in the off-peak period, resulting in higher average travel speeds and lower rates
of emissions.
While the actual timing of the peak travel period may differ from area to area, the typical
length of the peak period (either a.m. or p.m.) as recorded in the 1990 NPTS survey is
three hours. This number is coded as a default into the model, but the user has the option
to use local data and is encouraged to do so. Again, these data are obtained from the
local planning agency.
Work Trips Occurring in Peak Period - While the peak period is when most workers are
assumed to commute to jobs, nationally only 61.4%* of all work trips occur in the a.m. or
p.m. peak travel periods. The remainder, 38.6%, take place in the off-peak hours, and
occur under less congested off-peak travel conditions. The COMMUTER model uses
this relationship to calculate commute trip emissions in both peak and off-peak periods,
where emissions "rates" are different. The user is encouraged to supply the local value
for this relationship, which may be obtained from the local planning agency.
Peak and Off-Peak Travel Speeds - This information is essential in calculating emissions,
since the emissions factors are established through a relationship with average speed (this
relationship is discussed in more detail in Section 7).
Other Background Information - In addition to the information items presented above, the
user will probably need to obtain some other information in the course of doing an
analysis. These items are briefly mentioned here, and discussed in more detail in the
respective technical section that follows. The items include the following:
• Pre-existing offerings of the various employer TDM support strategies in the
region, the level of the strategy or program offered, and the extent of the
program's offering across the employment base (Section 4).
• Pre-existing offerings of and employee participation in alternative work
schedule programs, including compressed work weeks, telecommuting, and
flexible or staggered work hours (Section 5).
• The number (or percentage) and characteristics of employers who will be
affected by or who will be expected to participate in a given program scenario
(Sections 4, 5, and 6).
1995 Nationwide Personal Transportation Survey.
-31-
-------
• Peak and off-peak travel speeds (Section 7).
• Percentage distribution of the regional vehicle class by EPA size class
(Section 7).
• Information on current emissions requirements, standards, and programs
(Section 7).
###
-32-
-------
4. Estimating Travel Impacts of Employer Support
Programs
Overview of Employer Support Strategies
One way in which employers can try to persuade employees to consider traveling by
alternative modes, rather than driving alone, is to provide various types of support that
make it easier and more attractive to use those modes. These support programs typically
consist of measures that heighten awareness of the availability of other modes, provide
information on their service or use, or generally make it easier and more attractive to the
employee to consider their use.
These employer support programs generally do not include measurable time or cost
incentives or disincentives. Rather, they serve to provide an improved set of conditions
for the employee to use an alternative, and provide incentives that are tangible and
important, but not necessarily quantifiable by the employee. They may best be seen as
"catalysts" in supporting a change in behavior away from driving alone and providing an
accommodating environment in which those changes can take place, but by themselves
may not be capable of engineering major changes in employee travel.* Nevertheless,
employers find these types of strategies attractive because they present little economic
risk, and generally earn employee gratitude.
Many of the strategies in this category of programs are specific to the needs of the
particular mode. For example, Table 4-1 lists typical support strategies that might be
offered to encourage use of the respective mode.
See: Implementing Effective Travel Demand Management Measures, Federal Highway Administration,
September 1993; or Characteristics of Effective TDM Programs, TCRP Project B-4 Final Report, COMSIS
Corporation, May 1996.
-33-
-------
Table 4-1
Popular Employer Support Strategies
to Encourage Specific Modes of Travel
Mode
Carpool
Vanpool
Transit
Bike
Walk
Strategies
Ride matching (individual employer or tie in to regional
matching programs)
Preferential (reserved and/or close-in) parking
Ride matching (individual employer or tie in to regional
matching programs)
Preferential (reserved and/or close-in) parking
Use of company fueling or maintenance facilities
Employer capital cost or insurance underwriting
Startup subsidies
On-site transit information booths
On-site transit pass sales
Sidewalks or shelter
Bike lockers or racks
Shower and changing facilities
Sidewalks or other barrier removal
Showers and changing facilities
In addition to the mode-specific types of strategies listed in Table 4-1, there are actions
employers can take that are almost universal in their applicability across all of the
alternative modes. Examples of these strategies include the following:
• Employee Transportation Coordinators. Persons who are trained to provide
information or advice to employees regarding use of any alternative mode, in
terms of where to go for information, company policy and benefits, etc.
• Guaranteed Ride Home. Provision to get an employee home by alternative
means if it necessary to work late or in event of a personal emergency, and the
employee did not drive on that day.
• Flexible Work Hours. A formal or informal policy that allows employees
some flexibility over the official office hours in order to meet the schedule of
the chosen alternative mode.
• Promotions. Marketing and other methods to increase awareness of a given
mode or employer incentive, or to provide prizes or awards for meeting some
usage challenge.
-34-
-------
Analytic Approach for Estimating Travel Impacts
Because most support-type strategies do not translate into changes in the time or cost of
travel, it is difficult to calculate their impacts on travel in the same way as with time and
cost-based strategies (Section 6). The latter are readily translated into terms that can be
fed into a standard mode choice model, whose coefficients almost universally apply only
to time- or cost-based criteria. While a number of attempts have been made to
incorporate these support measures in traditional model structures, or to develop
representative coefficient or elasticity relationships on their link with travel behavior,
these efforts have not been particularly successful in developing valid mathematical
relationships.*
For this reason, travel impacts related to support programs are estimated by the
COMMUTER model in the same way they were calculated in the FHWA TDM model:
through relational look-up tables. This approach associates the offering of a particular
employer support measure with an incremental change in the mode share of the mode to
which the program is applied. This is done separately for each of the designated
alternative modes: Carpool, Vanpool, Transit, and Bicycle (the model currently does not
incorporate measures for Walk). Table 4-2 lists the changes in mode share that are
credited by the COMMUTER model for each mode and its respective program level.
The mode share adjustment values for support programs shown in Table 4-2 are taken
from the FHWA TDM model. The authors of the TDM model obtained these adjustment
values through a synthesis of empirical results from a large number of case studies. The
TDM model researchers determined that three factors were particularly important in
predicting the level of impact from a program of support actions:
• Level of Program: The significance of the program being offered in terms of
the support measures themselves, the number and type of such measures
combined into the program, and the level of investment reflected.
• Type of Employer: A difference in effect was attributed to whether the
measures were being applied in an "Office" or a "Non-Office" employment
environment.
* A Survey and Analysis of Employee Response to Employer-Sponsored Trip Reduction Incentive Programs,
Final Report and Technical Appendix A, COMSIS Corp. for California Air Resources Board (contract
#A932-187) and the South Coast Air Quality Management District (August 1993). This study attempted to
develop statistical relationships for a wide range of employer support incentive strategies, drawing upon
both extensive Reg. XV employer plan data from SCAQMD files coupled with new survey data from 43
California employers. Logit model estimation was used to determine whether statistically valid
relationships existed between these measures and travel behavior. Results were a mixed success, with the
time and cost-related strategies demonstrating the strongest relationships (and reasonable consistency with
conventional models), while most of the non-monetary support strategies exhibited weak or implausible
statistical relationships with demand. Those support strategies showing the strongest relationships were
guaranteed ride home (all modes), transportation coordinators with rideshare matching (carpool and
vanpool), use of company-provided vehicles (carpool and vanpool), transit information center and on-site
pass sales (transit), and bike racks with lockers/showers (bike and walk).
-35-
-------
Table 4-2
Increase in Percent Using Model by Support Program Level
Program Level
Program Type of Workplace 1 234
Carpool
Vanpool
Transit
Bicycle
Office
Non-Office
Office
Non-Office
Office
Non-Office
Office
Non-Office
0.40%
0.20%
0.40%
0.20%
0.20%
0.20%
0.20%
0.10%
1.00%
0.40%
1.00%
0.40%
0.50%
0.50%
0.50%
0.25%
2.00%
1.40%
2.00%
1.40%
1.50%
1.50%
1.50%
0.75%
4.00%
2.00%
4.00%
2.00%
2.00%
2.00%
2.00%
1.00%
Sources: FHWA TDM Model, 1993 (carpool, vanpool, and transit); Cambridge Systematics, Inc.
for COMMUTER version 1.0, 1999 (bicycle). Carpool and vanpool office levels 3 and 4 updated
2005 by Eric Schreffler. The Level 3 and Level 4 carpool and vanpool support impacts have
been reduced to reflect current professional opinion that support measures are less effective than
direct financial incentives and disincentives to commuters, all else being equal.
• Level of Employer Participation: On an areawide basis, the portion of the
employment population that can be expected to take part in the program,
and/or to offer a program of a given level
Program Level - For simplicity, employer support programs have been categorized into
four different levels, ranging from Level 1, reflecting the minimum level of effort, to
Level 4, representing the maximum. Each level is associated with a particular program
or "package" of support actions. The identification of what measures have been assumed
to make up each level is shown in Table 4-3. If an employer (or group of employers) is
presumed to adopt one of these programs, i.e., a given level for a particular mode, it is
assumed that the shown measures will be implemented by the employer.
-36-
-------
Table 4-3
Composition of Modal Support Strategy Programs
Mode
Carpool
Vanpool
Transit
Bicycle
"u
s
1
2
o
J
4
1
2
3
4
1
2
3
4
1
2
3
4
Strategies Included in Program
Carpool information activities (tied in with areawide matching)
Quarter-time transportation coordinator
All the above, PLUS:
In-house carpool matching service and/or personalized carpool candidate get-togethers
All the above, PLUS:
Preferential parking (reserved, indoor, and/or close-in)
Flexible work schedule policy to accommodate carpool schedules.
Half-time transportation coordinator
All the above, PLUS:
Full-time transportation coordinator
Vanpool information activities (tied in with areawide vanpool matching and/or third
party vanpool programs)
Quarter-time transportation coordinator
All the above, PLUS:
In-house vanpool matching services and/or personalized vanpool candidate get-
togethers
Non-monetary vanpool development assistance
Policy of flexible work schedules to accommodate vanpool schedule
All the above, PLUS:
Vanpool development and operating assistance, including financial assistance such
vanpool purchase loan guarantees, consolidate purchase of insurance, and a startup
subsidy.
Supporting services such as van washing and fueling
Half-time transportation coordinator
as
All the above, PLUS:
Major financial assistance for development and operations, such as employer purchase
of vans with favorable leaseback, continuing subsidy, free maintenance, free insurance.
Full-time transportation coordinator
Transit information center
Quarter-time transportation coordinator
All the above, PLUS:
Policy of work hours flexibility to accommodate transit schedules/delays
All the above, PLUS:
On -site transit pass sales
Half-time transportation coordinator
All the above, PLUS:
Guaranteed ride home
Full-time transportation coordinator
Provision of on-site bicycle parking (racks or lockers)
All the above, PLUS:
Shower and change facilities
All the above, PLUS:
Provision of secure bicycle parking (storage lockers or indoor storage)
Development of local bike-friendly infrastructure
All the above, PLUS: Workplace information and promotional activities
-37-
-------
Of course, not all employers would be expected to implement the same modal programs
at the same levels uniformly across an entire region. As a result, the model allows for the
user to assume and test different levels of participation across the employment
population. Also, it would be unrealistic to assume that all employers would be "starting
from scratch" in terms of existing programs. Many employers will already have been
doing one or more of these actions, and the analysis allows for identifying these pre-
existing efforts and netting out their effects in order to avoid double counting of benefits.
Type of Employment - The research that developed the FHWA TDM model also
concluded that the impact of support programs varied according to the type of workplace,
in which it distinguished between Office and Non-Office employment. This shows up in
travel responses for Office employment that are somewhat greater than the same strategy
applied to Non-Office employment. The explanation for this is that Office employees are
more likely to travel on a routine daily schedule to given work sites, while Non-Office
workers (those in construction, mining, manufacturing, and utility occupations) are more
likely to be traveling at non-peak times, or working shifts, or traveling to different
locations depending on where the current job is.
Employer Participation - If the user is conducting an analysis at a regional or multi-
employer level, it is necessary to account for the type of participation that is expected
from the sample of employers. The user will be asked to specify the percentage of
employers who will be expected to participate in each modal support program by level of
the program, as well as the percentage that would not participate at all.
So, for example, if a scenario were created that envisioned employer support programs at
level 2 for vanpool and transit, level 4 for carpool and level 3 for bicycle, the
communication of this information to the COMMUTER model would resemble the
following:
Percent of Employers Participating by Level of Program
Mode
Carpool
Vanpool
Transit
Bicycle
None
Level 1
Level 2
100%
100%
Level 3
100%
Level 4
100%
Total
100%
100%
100%
100%
More typically, what might happen when a scenario is drawn up is that there would be
much more variation in the types of participation expected, since programs would be
expected to emphasize different modes based on, for example, employer location, or not
include certain groups, for example, employers with fewer than 50 employees. To
-38-
-------
illustrate this point, the participation information communicated to the COMMUTER
model might instead look like this:*
Percent of Employers Participating by Level of Program
Mode
Carpool
Vanpool
Transit
Bicycle
None
38%
38%
38%
38%
Level 1
42%
12%
Level 2
20%
50%
Level 3
12%
62%
Level 4
50%
Total
100%
100%
100%
100%
This distribution suggests that 38% of all employers in the base would not be expected to
implement any of the programs, and that of the remaining set, full participation at the
design level (nominal levels shown in previous table) is probably unlikely, with some
estimated percentage implementing programs that are probably at a level lower than the
level intended. When this distribution is related to the model, the computation of the
change in carpool mode share associated with this participation pattern would be as
follows:
Carpool Mode Share Increase:
38% @ Level "0" = 0.38 x 0.0%
12% @ Level 3 = 0.12 x 2.0%
50% @ Level 4 = 0.50 x 4.0%
Net increase in carpool share = 0.0% + 0.24% + 2.0% = 2.24%
So with this example, the model would add 2.24% to the existing carpool share for this
scenario, and similar calculations would be done for the other modes. This simple
illustration assumes two conditions, however, that are not likely to occur in practice:
• The employment base is comprised entirely of Office employment; and
• Each of these programs is being implemented on a zero base, i.e., there are no
pre-existing program efforts of this type in place.
Some employer support programs may apply combinations of strategies that are different from those
described in Table 4-2. In these cases, the analyst may either choose the program level that most closely
compares to the actual program being implemented, or split the impact among programs of two or more
different levels. For example, if a carpool program includes all the elements of a Level 3 program except
for flexible work schedules, the user may want to approximate this program by assigning half of the
participation to Level 2 and half to Level 3.
-39-
-------
With regard to employment type, the calculation would be expanded to account for the
different level of impact projected in non-Office situations, as shown in Table 4-2. Thus
if the baseline employment were comprised of, say 80% Office and 20% Non-Office, the
calculation above would be revised as follows, yielding 2.03% as the Carpool share
adjustment instead of 2.24%:
Carpool Mode Share Increase:
38% @ Level "0" = 0.38 x 0.0%
12% @ Level 3 = 0.12 x [(2.0% x 80%) + (1.4% x 20%)]
50% @ Level 4 = 0.50 x [(4.0% x 80%) + (2.0% x 20%)]
Net increase in carpool share = 0.0% + 0.23% + 1.8% = 2.03%
With regard to accounting for existing program efforts, this is done easily in a table
equivalent to the one above, but instead specifying existing participation rates by level.
An example is shown below.
Percent of Employers by Existing Program
Mode
Carpool
Vanpool
Transit
Bicycle
None
50%
95%
50%
95%
Level 1
50%
5%
40%
5%
Level 2
10%
Level 3
Level 4
Total
100%
100%
100%
100%
Given this information, the COMMUTER model will compute the net impact from the
new scenario of actions by subtracting out the portion of the effect that may be attributed
to existing efforts. Using the Carpool example again, the share that may be attributed to
existing efforts is calculated as follows:
Carpool Mode Share Increment Due to Existing Efforts:
50% @ Level "0" = 0.50 x 0.0%
50% @ Level 1 = 0.50 x [(0.4% x 80%) + (0.2% x 20%)]
Credit for Existing Program = 0.0% + 0.36% = 0.36%
-40-
-------
The net increase that may then be attributed to the Carpool support program of the
scenario, after crediting for existing efforts, is 2.03% - 0.36%, or 1.67%.
Site vs. Areawide - Users of the COMMUTER model who are evaluating employer
support programs at an individual site level will find slight differences in how the model
treats a single employment site vs. the more challenging simulation of these programs
applied across a diverse employer population, as discussed above. The basic analytics
are the same, however, employing the same mode share adjustment procedures and
affording the same level of impact. The only difference is that the user does not have to
specify the distribution of the programs across the employment base.
Model users are asked at the early stage of model application whether the analysis will be
for an individual site or areawide. The model then automatically directs them to the
individual site screen, where they are asked to enter the following information:
Program Focus
Carpool
Vanpool
Transit
Bicycle
Existing Level
New Level
As with the multi-employer case, the individual site user first enters the "Existing Level"
that best corresponds to the type of program that is being offered now (from 1 to 4, with
the default being Level 0), and then selects the New Level that best represents the
program that will be offered. This is done individually for each mode. The model then
determines the change for each mode under the New Level, and nets out the credit for the
Existing Level.
The COMMUTER model also offers an alternative procedure that give the user even
more flexibility to experiment with the potential impact of support programs. The
following alternative table is presented in the model:
Program Focus
Carpool
Vanpool
Transit
Bicycle
Existing
Share
13.2%
0.5 %
5.3 %
0.4 %
Increase
%
%
%
%
New
Share
13.2%
0.5 %
5.3 %
0.4 %
-41-
-------
To use this procedure, the user enters the percentage increase that is expected for each
listed mode through the given set of program actions, and the model automatically makes
the adjustment from the Existing Share to the New Share. This type of procedure is
useful only if the user has previously compiled company survey data on how employees'
mode shares would change if a support program were implemented. (Note that the
existing shares in this input table come from the baseline mode share inputs.)
Absence of Walk Mode - While employers can make an effort to encourage those
employees who have the opportunity to walk to work, the specific incentive measures are
perhaps less tangible for pedestrian travel. Neither the individual site nor the areawide
applications for support programs in the COMMUTER model provide for increasing
walk mode share; however, the model does preserve the share that is already there, by
freezing the starting share of walk trips and not including it in the normalizing process
(described below) that balances the share adjustments across all the other modes. Walk
mode share can be increased through improvements in walk access time to the site, as
described in Section 6.
Normalizing Across Modes - Once the mode shares of each commute mode have been
adjusted upward by the increment associated with the respective support program, the
sum of all modal shares will almost always exceed 100%. In the logit mode split model
procedure (Section 6), the total shares are constrained to 100% by virtue of the way the
equation works, i.e., by dividing up the "pie" in proportion to the respective "utilities" of
each mode. In this more simplistic procedure, however, where increments are added to
each mode without accounting for "where" (what other mode) those new riders come
from, some readjustment must be made to bring the sum of all shares back to 100%. This
is done through a process called normalizing, where each mode's share is divided by the
new total, proportionately reducing each share so that the total once again is 100%. This
is equivalent to assuming that the increase in each mode share affected by a program will
come from the other modes in proportion to their starting mode shares.
The important exception to the procedure of proportioning the adjustments across modes
is that Walk is not included in the adjustment process. The normalizing process simply
freezes the Walk share at its existing level, and it neither gains nor loses as a result of the
support program impacts.
A simple example shown as Figure 4-1 illustrates how the normalizing process works
(this example was also used earlier in Section 2). Initial shares of carpool, vanpool,
transit and bicycle are increased through individual support programs. Carpool is
increased from 13% to 15%, Vanpool from 1% to 2%, Transit from 5% to 7%, and
Bicycle from 1% to 2%. The problem, though, is that these various programs add 6% to
the targeted modes, but no mode has had to give up any users. To correct this, the 6% are
"found" from among all of the other modes, including the ones that have been targeted
(except Walk, which remains excluded from this process). This amounts to multiplying
each share by the ratio of the new total to the old total, or 96%/102%, or 0.941. This
means that Carpool's increase from 13% to 15% is dropped back to 14.1%, Vanpool's
and Bicycle's 2% dropped back to 1.9%, and Transit's 7% dropped back to 6.6%.
-42-
-------
However, in this process Drive Alone drops from 75% to 70.6%, losing the greatest share
because its share was initially the largest.
Figure 4-1
Illustration of Share Adjustment Process
Starting Mode Shares:
Drive Alone:
Carpool:
Vanpool:
Transit:
Walk:
Bicycle:
Other:
75%
13%
1%
5%
4%
1%
1%
Carpool TDM Support Program
A Share
Program Level
1
+1.5
2
+2
3
+3
4
+5
Vanpool TDM Support Program
A Share
Program Level
1
+.5
2
+1
3
+1.5
4
+2
Transit TDM Support Program
A Share
Program Level
1
+1
2
+2
3
+4
4
+5
Bicycle TDM Support Program
A Share
Program Level
1
+.1
2
+.3
3
+.5
i
IP
Share Adjustment Process:
Adjustment
Drive Alone:
Carpool:
Vanpool:
Transit:
Walk:
Bicycle:
Other:
TOTAL
Base
75%
13%
1%
5%
(4%)
1%
1%
96%
A
+2%
+1%
+2%
+1%
Revised
75%
15%
2%
7%
(4%)
2%
1%
102%
Factor
.941
.941
.941
.941
1.0
.941
.941
Final
Shares
70.6%
14.1%
1.9%
6.6%
(4%)
1 .9%
0.9%
100.0%
###
-43-
-------
5. Estimating Travel Impacts of
Alternative Work Schedules
Overview of Alternative Work Schedule Programs
Rather than trying to induce a change in mode of travel, this class of commuter choice
strategies is aimed at reducing the frequency of the commute trip (i.e., the number of trips
made by the employee to the site per week), or shifting the timing of the trip outside of
the peak hour or period. These are broadly referred to as Alternative Work Schedules or
Arrangements, and consist of the following popular options:
• Flexible Work Hours: A relaxation in the official daily hours of business
allows employees the flexibility to adjust their personal work schedules to
either come early/leave early, or come late/leave late in order to avoid the most
congested portion of daily commute periods.
• Staggered Work Hours: A more formalized version of the above, where the
employer sets one or more starting/ending times within a small time increment
of each other, so that all employees are not arriving/departing at the same time.
For example, one half of the workforce may arrive at 8:00 a.m., and the other
half at 8:30 a.m.
• Compressed Work Weeks: Rather than working a standard 5-day/8-hour-
per-day work week, some employers will allow employees to work a longer
work day, usually either 9 or 10 hours, and build credit in order to be exempt
from traveling to the site on a particular day. The most common versions of
compressed work week arrangements are 4/40, where the employee works four
10-hour days and then takes the fifth day off, or 9/80, where the employee
works nine 9-hour days and takes the tenth day off.
• Telecommuting: Also sometimes referred to as "telework", this arrangement
allows employees to work off-site usually one or more days per week, being in
communication with the worksite via telephone or computer modem
connection.
Typically an employer will offer only one of these programs to its employees, since, to a
certain extent, taking part in one nullifies the ability to participate in the other. However,
there may be occasions when more than one of the programs may be offered, perhaps to a
different segment of the employer's work force. The COMMUTER model estimates the
impact of each of these Alternative Work Schedule programs, and permits offering more
than one program at a time with some practical constraints imposed to prevent unrealistic
cases of dual eligibility.
Nature of Emissions Impacts
-44-
-------
Whereas employer commute choice programs that induce changes in travel mode lead to
reductions in vehicle trips and VMT and thus fewer emissions, Alternative Work
Schedule programs reduce emissions by eliminating trips or pushing them to a less
congested time period. A brief description of how each of the listed programs would be
expected to reduce emissions is provided below.
• Flexible Work Hours: Emissions reductions from flexible work hours are
determined entirely by the number of vehicle trips shifted from the peak period
to the non-peak period. Emissions reductions occur because congestion is less
and speeds are higher in the off-peak. An important assumption with flexible
work hours is that shifting work hours does not take away the incentive to use
transit or ride in a carpool or vanpool, since driving alone may be more
attractive outside the peak period.*
• Staggered Work Hours: Same as flexible work hours, emissions reductions
from staggered hours are dependent entirely on the number of vehicle trips
shifted out of the peak period. The same important assumption as to whether
employees have the same incentive to use transit or rideshare applies as with
flexible work hours.
• Compressed Work Weeks: This strategy reduces emissions by eliminating
vehicle trips to the work site. 4/40 programs eliminate one out of five trips per
week, while 9/80 programs eliminate one trip out of every ten, or 0.5 per week.
A critical assumption in claiming emissions credits from these trips shifted is
that equal or greater travel does not occur on the individual's day off.
• Telecommuting: This strategy also reduces emissions by eliminating vehicle
trips to a work site. As with compressed work weeks, the critical assumption is
that travel on the telecommute day is negligible because the employee is home
working.
Analytic Approach for Estimating Travel Impacts
Three factors enter into the calculation of the travel impacts from Alternative Work
Schedules:
1. Which strategy or strategies are being offered?
2. What type of occupation is the target employment engaged in, i.e., is it suitable
for the particular type of alternative work arrangement? The COMMUTER
Flexible work hours as presented here is not to be confused with the flexible work schedule policy
described in Chapter 4 as a way of encouraging alternative mode use by granting leniency to employees
who need to adjust their hours to meet the schedule of a carpool/vanpool or transit.
-45-
-------
model calculates Alternative Work Schedule travel impacts only for Office
employment.
3. What percentage of employees are made eligible for (or are expected to
participate in) the programs by employers?
A fourth factor becomes important in specific relation to emissions impacts, namely, the
length of the peak period as a factor in determining the number of trips that would
actually be shifted outside the peak.
Like the Commuter Choice Support Strategies (Chapter 4), Alternative Work Schedule
programs are also analyzed using factors supplied in lookup tables. The user enters
preferences as to which strategies will be offered, and indicates the percentage of
employees who will be eligible for (or, alternatively, would be expected to participate in)
each. The COMMUTER model then makes a calculation that either removes vehicle
trips from the total or shifts those trips to the off-peak period. Research reviews
identified no elasticities or coefficients that could be used for estimating the response of
travelers to these strategies. As a result, simple factor relationships taken from empirical
research, in combination with some simple mathematical identities, are used to calculate
the travel impacts for these strategies.
The calculation of the travel impact for each of the Alternative Work Hours programs is
explained below.
Flexible Work Hours - Using relationships developed for the FHWA TDM model, the
COMMUTER model assumes that in employment situations where employees are made
eligible for Flexible Work Hours options, 22% will in fact take advantage of that option
and participate in a program that will shift their hours of travel.* Alternatively, users can
specify directly what percent are expected to participate, and in effect replace the 22%
default with their own estimate; there should, however, be some explanation as to what
information justified this assumption.
The user is also asked to supply information on the existing percent of employees eligible
for (or already participating in) Flexible Work Hours. As with the Employer Support
Programs (Chapter 4), the net new percentage of eligible (or participating) employees is
used to estimate the impact.
Exactly how many of the employees who elect to participate in Flex Hours actually shift
their time of travel such that the new trip falls outside the peak depends on the length of
the peak period. If the peak period is short (e.g., two hours in duration), then the
likelihood that a trip shifted by flexible work hours will fall outside the peak (and hence
become associated with lower emissions rates) is fairly high, whereas with a long peak
This percentage was derived from data compiled in a 1980 study "Behavioral Impacts of Flexible Work
Schedules."
-46-
-------
period, say 4 or more hours, only a relatively small percentage would be expected to fall
outside the peak. Table 5-1 shows the shift percentages as related to length of peak
period that are coded into the COMMUTER model.
Table 5-1
Percent of Trips Shifted by Length of Peak Period
Length of Peak Period (hrs)
2.0
2.5
3.0
3.5
4.0
4.5
5.0
Percent of Trips Shifted
28.7
19.2
13.9
10.6
8.5
7.1
6.0
Source: Estimating the Effect of Alternative Work Schedules on Travel Activity
and Emissions, FHWA TDM Evaluation Model, 1993.
The user supplies the local information on length of peak period. This item appears on
the screen for Alternative Work Hours strategies as "Percent of Trips Shifted from Peak
Period." The user may use the default value of 3 hours if local data are not easily
available. If local data are available on the percentage of trips shifted from the peak,
they should be used in place of the defaults given in Table 5-1.
It is important to note that not all of the trips being shifted under Flexible Work Hours
are vehicle trips. Employee trips are shifted at the same modal split as exists in the
overall population; hence, if only 80% of employees drove alone, and 10% rode in 2-
person carpools, then only 85% (80% + 10%/2) of the shifted trips would be assumed to
be vehicle trips.
The calculation of the travel impact from flexible work hours is then as follows:
-47-
-------
Number Daily Vehicle Trips Shifted,
Peak to Off-peak =
Total Affected Employment
X
Percent of Employment that is Office
X
NET Percent of Employees Eligible
X
22% Participate in Program
X
Percent of Trips Shifted
(Based on Length of Peak Period)
X
Current Private Vehicle Mode Share
The following example illustrates the effect of Flexible Work Hours:
Total Vehicle Trips Shifted from Peak =
(100,000 employees) X (75% Office) X (30% - 10% eligible) X (22%) X (13.9%
shifted)3'"P6* X (.85 vehicle trips per employee) = 390 daily vehicle trips shifted
Alternatively, the user could specify the percent of employees that would be expected to
participate in the program if offered (less the percent already participating). In the above
example, if the "Eligible" percentages were in fact already "Participating," the
calculation would change as follows:
(100,000 employees) X (75% Office) X (30% - 10% participation) X (22%) X
(13.9% shifted)3hrpeak X (.85 vehicle trips per employee) = 1,772 daily vehicle
trips shifted
This is a very different outcome, 1,772 vs. 390 trips, than when the participation is
entered directly. For this reason, the user should be aware of the difference in
significance between using the Percent Eligible vs. the Percent Participating option in the
model.
Staggered Work Hours - The travel impacts from Staggered Work Hours are determined
in exactly the same manner as Flexible Work Hours above. The user indicates [either]
the percentage who will be Eligible, or who will be expected to Participate, with the same
caveats as discussed above. The model then calculates the number of trips that will be
shifted, and the number that will be assumed to fall outside the peak and qualify for
favorable emissions treatment.
-48-
-------
Compressed Work Weeks - Calculation of travel impacts from Compressed Work Weeks
follows a very similar procedure to the one above for Flexible and Staggered Work
Hours. The principal difference is that employees who opt to participate in Compressed
Work Weeks are assumed to make fewer trips to the worksite, rather than the same
number of trips but shifted out of the peak.
To estimate the effects from Compressed Work Weeks, the user must indicate:
• The percent of employees who will be eligible for the new program, or who
will be expected to participate;
• The percent who are already eligible for (or participating in) a similar
compressed work hours program; and
• Whether the program type will be a 4/40 or 9/80 arrangement (percent by
each).
The COMMUTER model then calculates the number of daily vehicle trips that would be
eliminated from the Compressed Work Week programs, as a function of the total number
of Office (only) employees, the net new percent eligible or participating, the current
modal share, and the effective daily trip reduction attributable to the type of program
(4/40 or 9/80).
As with the Flexible or Staggered work hours programs, if the user specifies the percent
or employees Eligible, the model will assume that 22% of these eligible employees will
participate in the program. If given directly the percent who will participate, it will use
100% of this estimate.
For each employee who participates in a compressed work week program, the following
trip reduction will be estimated:
4/40 Work Week: 1 trip per week or 0.20 trips per day.
9/80 Work Week: 1A trips per week, or 0.10 trips per day.
Again, as with the work hours strategies, the elimination of trips is assumed to be
proportional across all current modes, i.e., they are not all vehicle trips being eliminated.
-49-
-------
The calculation of the travel impact from Compressed Work Weeks is then as follows:
Number Daily Vehicle Trips Eliminated =
Total Affected Employment
X
Percent of Employment that is Office
X
NET Percent of Employees Eligible
X
22% Participate in Program
X
Daily Person Trips Eliminated
(0.20 for 4/40, 0.10 for 9/80)
X
Current Private Vehicle Mode Share
The following example illustrates the effect of Compressed Work Weeks (based on
eligibility):
Total Vehicle Trips Eliminated =
(100,000 employees) X (75% Office) X (30% - 10% eligible) X (22%) X
(0.20 trips)4740 WeekX(.85 vehicle trips per employee) = 561 daily vehicle trips
removed
Alternatively, the user could specify the percent of employees that would be expected to
participate in the program if offered (less the percent already participating). In the above
example, if the "Eligible" percentages were in fact already "Participating," the
calculation would change as follows:
(100,000 employees) X (75% Office) X (30% - 10% participation) X (22%) X
(0.20 trips)4740 Week X (.85 vehicle trips per employee) = 2,550 daily vehicle trips
removed
This is a very different outcome, 2,550 vs. 561 trips, compared to when the participation
is entered directly. For this reason, the user should be aware of the difference in
significance between using the Percent Eligible vs. the Percent Participating option in the
model.
The user should also apply caution in interpreting the trip reduction number, given
uncertainty as to the travel behavior of the employee on the "day off."
-50-
-------
Telecommuting - Calculating the travel impacts from Telecommuting is slightly different
from both prior classes of strategies, but similar to the Compressed Work Week in that
impacts are registered as trips eliminated as opposed to trips shifted from peak to off-
peak.
To estimate the effects from Telecommuting, the user must indicate the following:
• The percent of employees who will be eligible for the new program, or who
will be expected to participate;
• The percent who are already eligible for (or participating in) a similar
Telecommuting program; and
• The average number of days per week that employees will telecommute, i.e.,
work at home, if different from the 1.5 day default value coded into the model.
The COMMUTER model then calculates the number of daily vehicle trips that would be
eliminated due to Telecommuting as a function of the total number of Office (only)
employees, the net new percent eligible or participating, the current modal share, and the
average number of days per week that the employee will work at home under the
telecommute arrangement.
As with the Flexible/Staggered work hours or Compressed Work Week programs, the
user may either specify the percent of employees who will be eligible for a Telecommute
program, or the exact percentage that will participate. If the user opts for the "Eligible"
approach, the model will assume that 10% of the eligible employees will participate in
the program. If the percent who will participate is provided directly, the model will use
100% of that estimate.
The percent of eligible office commuters who would be assumed to telecommute, if given
the option, has been reduced from 32 percent in COMMUTER version 1.0 to 10 percent
in COMMUTER version 2.0. This change reflects more recent survey findings and
analyses that suggest about one in ten workers who are in a job that is suited to working
from home, have an employer who would allow telecommuting, and who want to begin
telecommuting will do so.* This percent shift of eligible workers can be modified within
the COMMUTER Model if the user has local empirical data that suggests employees will
shift at a greater level. This might be due a more aggressive program that encourages a
greater proportion of the workforce to adopt telecommuting, or if the user is modeling a
single worksite where the proportion of targeted telecommuters is known and is higher
than 10 percent.
See: Joanne H. Pratt and Associates, Telework America 1999 Survey and Key Findings, prepared for
International Telework Association and Council, 1999; LDA Consulting, 2003 State of the Commute
Report, prepared for the Metropolitan Washington Council of Governments, June 2004; Mokhtarian, P. and
Salomon, I., "Modeling the Choice of Telecommuting 2: A Case of the Preferred Impossible Alternative,"
Environment and Planning A 28, 1996, pp. 1859-1876.
-51-
-------
For each employee who participates in a Telecommute program, the following trip
reduction will be estimated:
1.5 days per week work at home / 5 days per week = 0.3 trips per day eliminated.
(user has option of changing the 1.5 day default)
Again, as with the other strategies, the eliminated trips are assumed to be proportional
across all current modes, i.e., they are not all vehicle trips being eliminated.
The calculation of the travel impact from Telecommuting is then as follows:
Number Daily Vehicle Trips Eliminated =
Total Affected Employment
X
Percent of Employment that is Office
X
NET Percent of Employees Eligible
X
10% Participate in Program
X
Daily Person Trips Eliminated
(0.3 based on 1.5 days)
X
Current Private Vehicle Mode Share
The following example illustrates the effect of Telecommuting (based on eligibility):
Total Vehicle Trips Eliminated =
(100,000 employees) X (75% Office) X (30% - 10% eligible) X (10%) X
(0.3 trips)1'5 days/week X (.85 vehicle trips per employee) = 383 daily vehicle trips
removed
Alternatively, the user could specify the percent of employees that would be expected to
participate in the program if offered (less the percent already participating). In the above
example, if the "Eligible" percentages were in fact already "Participating," the
calculation would change as follows:
-52-
-------
(100,000 employees) X (75% Office) X (30% - 10% participation.) X (32%) X
(0.3 trips)1'5 days/week X (.85 vehicle trips per employee) = 3,825 daily vehicle trips
removed
This is a very different outcome, 3,825 vs. 383 trips, compared to that obtained when the
participation is entered directly. For this reason, the user should be aware of the
significant difference between using the Percent Eligible vs. the Percent Participating
option in the model. The user should also apply caution in interpreting the trip reduction
number, given uncertainty as to the travel behavior of the employee on the "day off."
Research suggests that some of the travel benefits of telecommuting will be offset by
additional trips made from home for other purposes.
Multiple Options
The COMMUTER model permits the user to examine a program scenario with more than
one of the alternative work hours strategies at one time. If the analysis is being done at a
regional scale over many employers, it is practical to assume that more than one option
would be tested in the same scenario. Generally, the user is advised to limit the total
eligibility—or participation—in all programs to 100% of the affected employment base.
When the model is asked to estimate the effects of multiple programs in the same
scenario, it simply calculates the program effect for each program, e.g., Flexible Work
Hours, individually, and then simply adds the results to a grand total of "number of trips
shifted to off-peak" and "number of daily trips eliminated." Unlike the Employer
Support Strategies of Chapter 4, however, because the user is advised to limit total
participation to 100%, there is no concern about one program competing for share with
the other and necessitating a tradeoff and normalizing process.
Even an individual employer using the model might opt to make more than one
Alternative Work Schedules strategy available, and so long as the total eligibility or
participation does not exceed 100% of all employees, no reconciliation of shares across
programs is necessary in the model.
However, one could envision the scenario where an employee might be eligible for more
than one program option, or in fact be eligible for any of the options simultaneously. In
this instance, the employee might choose to participate in more than one program, limited
only by the interference that might occur between one program and another. For
example, an employee "eligible" for Compressed Work Weeks, Flexible Work Hours,
and Telecommuting might find it difficult in practice to be able to take advantage of more
than one of these options. If the employee is on a 4/40 work week, for instance, it is
difficult to see how they could also telecommute, or participate in a Flexible Work Hours
program. Nevertheless, to accommodate this real possibility, the model allows the user
to specify more than 100% eligibility or participation across all programs, in essence
letting the employee "decide" which program or combination of programs to take part in.
To deal with this in the model calculations, each program's share is first computed
individually, as before, assuming that the employee has only one option available.
-53-
-------
However, once all the individual choices are computed, the sum of the eligibility or
participation rates is computed, and if in excess of 100%, the computed effects of each
program is normalized back to an effective participation of 100%. This gives a very
liberal opportunity to the user to test broad strategies, while still adhering to basic
reasonableness criteria regarding use of more than one alternative at a time.
###
-54-
-------
6. Estimating Travel Impacts of Travel Time and Cost
Changes
Overview of Travel Time and Travel Cost Strategies
Perhaps more than any other type of strategy, those measures that change the time or cost
of making a trip seem to have the most significant impact on travel behavior.* As a
simple example, if a traveler had two alternatives for making a trip, A and B, which the
traveler saw as basically equal options, reducing the travel time or the travel cost
associated with alternative B would make B more attractive, and the traveler more likely
to choose that alternative. Exactly how much more attractive it would be depends on
how big the relative changes are, which components of time or cost are changed (since
they are weighted differently by the traveler), and other intangibles such as the purpose
of the trip, the occupation and income of the traveler, and so forth.
Travel time and cost are also the most measurable elements of transportation choice.
Whereas it may be difficult to precisely measure the value of, for example, a guaranteed
ride home program as an employer incentive to encourage transit or ridesharing,
incentives that involve changes to travel time or cost are quite measurable. As a result,
statistical analysis of travel behavior shows a strong and consistent relationship between
travel choice and the level of time and cost associated with the trip.
The economic framework upon which travel demand analysis and forecasting is based
suggests that travelers view travel as a means of accomplishing a need, and consider
several factors when deciding whether and how to make a trip:
• Trip Purpose: Depending on the type of activity to be satisfied, the need to
make a trip may be less flexible for some purposes than others. Travel to
work, for example, is generally less negotiable than making a social or
recreational trip.
• Trip Frequency: In addition to deciding whether to make the trip, the traveler
also has the choice, depending on the purpose of the trip, of how often (daily,
less than daily, etc.) the trip will be made.
• Destination: When the traveler is not committed to a single site for satisfying
the activity, as is generally the case with the commute trip, more than one
location may be acceptable.
* See for example: Federal Highway Administration. Implementing Effective Travel Demand Management
Measures. (1993), and Transit Cooperative Research Program. Cost Effectiveness of TDM Strategies.
TCRP Project B-4 (1995).
-55-
-------
• Travel Mode: The traveler may have more than one travel mode to choose
from, e.g., driving, walking, taking transit, depending on the coordinates of the
trip and what modes are available.
• Time of Day: Again, depending on the purpose of the trip and the flexibility of
the schedule, the traveler may adjust the hour of the day that the trip is made in
order, for example, to avoid traffic congestion that will make the trip take
longer.
In the case of travel made for the purpose of work, many of the traveler's options are
limited: generally the trip must be made on a daily basis, to a fixed destination, and at a
given time of day. The Alternative Work Schedule strategies presented in the previous
chapter have introduced some flexibility into this set of choices, allowing some travelers
some discretion over how often they travel from home to the worksite, or at what times of
day.
Generally, however, the principal choice to the commuter is in what mode of travel to
use. Except for those persons whose households do not own vehicles, the most
commonly used alternative is driving alone. Principal alternatives that represent choices
to driving alone are carpooling, riding transit, or joining a vanpool. Persons whose
residence is reasonably close to the work place may also have the choice of walking or
biking.
When evaluating these modal choices, the consumer is believed to look at how long it
will take and how much it will cost to travel from door-to-door, as depicted in the figure
below.
In other words, the comparison among the options is based on how much time and effort
is involved in getting from the home to the mode of travel (access time), traveling in that
mode to the destination (in-vehicle time), and then getting from the point of debarking to
the final destination (access time). For most travelers, driving a personal vehicle
Access
Time
In-Vehicle
Time
Access
Time
represents the most convenient solution to this comparison. The auto user walks to the
parked vehicle (usually in a driveway or garage), drives directly to the workplace, and
parks within fairly short walking distance of the workplace. For the great majority of
American workers, the auto user is also likely to find ample free or subsidized parking at
the worksite.
-56-
-------
In contrast to the option of driving alone, which is virtually "door-to-door", each other
mode involves some type of compromise. The carpool or vanpool user must give up
some flexibility of arrival and departure time to accommodate the schedule of the other
riders, and often spend a bit more time traveling just to connect up with the other pool
users (access or in-vehicle time). The transit user must first get to the transit line, either
by walking or driving, then wait for the transit vehicle to arrive (wait time), and then
walk from the transit stop at the end of the trip to the final destination. In some instances,
it is also necessary to use more than one transit service, involving a transfer and potential
wait (transfer time). Persons who would walk or bike to work face not only limitations
on "reasonable distance" (generally 1 mile for walk, about 2 miles for bicycle), and
exposure to weather and traffic, but also physical barriers that make access to the site
difficult and add time.
What makes transit, ridesharing, biking, or walking attractive to some travelers, other
than matters of pure personal preference or constraint, are time or cost advantages of one
mode over the other. If users of carpools, vanpools, or transit can gain a travel time
advantage over the single-occupant vehicle by having a separate facility that avoids
congestion, that can alter the balance. Changes at the destination end that reduce the time
and difficulty of reaching the site are also important. For transit users this might mean a
stop closer to the building or a more direct path of access to the building. For carpool or
vanpool users this might mean close-in preferential parking. For bikers or pedestrians, it
might mean improved access to the site that reduces the time and distance.
Travel cost can also be a factor that tips the balance toward one of the alternative modes.
The typical SOV user usually only recognizes the cost of operation of the vehicle. Tolls
and parking fees are clearly noted, but are seldom in place, so the SOV traveler sees their
mode as very cost-effective. Transit users, on the other hand, must pay a fare that is no
longer a "nominal" charge in most urban systems. Vanpool users usually pay a monthly
or weekly charge, and carpoolers split their costs to where it is less than driving alone,
but if the trip does not involve parking, tolls or other out-of-pocket fees, the savings may
not be significant. Obviously, bikers and pedestrians have the cheapest alternative,
trading their time and effort for avoidance of monetary cost.
Actions that can make transit, ridesharing, or biking/walking more attractive to potential
users from a cost point of view would entail either increasing the cost of driving alone or
decreasing the cost of using the alternative. Fare subsidies are the most common cost
incentive for transit users. Employers have also been known to subsidize or underwrite
the costs of vanpooling, and there have been instances of employers providing
"equipment" subsidies to encourage biking or walking.
Perhaps the most significant incentive to the decision to drive or not drive, however, is
the cost of parking. Since most employers offer free or subsidized parking to employees,
this substantial benefit is a major subsidy to employees who drive. Where employers
have imposed charges on parking, major differences in drive alone mode share can be
-57-
-------
observed.* There are many variations on this theme of using parking cost as a
transportation demand management incentive, including the following:
• Combining parking fees for SOVs with free or subsidized parking for carpools
or vanpools.
• Offering a "Cash-Out" payment in the amount of the employer's cost to
provide parking to the employee as an incentive to not drive alone.
• Taking the revenues from employee parking fees and distributing them back to
employees as targeted or general subsidies to support use of other modes.
Types of Time/Cost Strategies Covered by COMMUTER Model
The COMMUTER model is fairly flexible in its ability to look at a variety of travel time
or travel cost incentive strategies. These strategies can be targeted at individual modes,
and numerous strategies can be included within the same scenario. The model will deal
with the interactions among the modes under the influence of the selected package of
strategies through a multinomial logit modeling procedure, described in the next section.
Site Access Improvements - The user can test strategies that change the current time
required to access the employment site once it is reached by the respective mode. This
can be done for any or all of the modes at the same time, with different assumptions for
each mode. Testing a negative value of change in access time (e.g., -two minutes)
constitutes an incentive since it reduces travel time, while testing a positive value
represents a disincentive for that particular mode.
Changes in walk access time for carpools, vanpools, or drive alone can be realized
through parking supply management techniques, such as preferential close-in parking for
pools or on-site vs. off-site parking privileges. Changes in access time for transit, biking,
or walking can be realized through improved site access design, affording a more direct
and safe connection to the transit stop or the local community or sidewalk/street network.
Transit Service Improvements - Most employers are not likely to initiate strategies that
will change the travel time associated with transit service, other than the improvements to
on-site access described above. However, users examining regional applications of the
COMMUTER model may envision and wish to test the consequences of improving the
travel time associated with improved transit service.
Transit service can provide shorter door-to-door travel time as a result of the following
changes:
* See again: FHWA Implementing Effective Travel Demand Management Measures (1993) and TCRP B-
4: Cost-Effectiveness of TDM Strategies (1995).
-58-
-------
• Increased frequency of service (i.e., shorter headways between vehicles),
which allows reduction in wait time or transfer time.
• More direct service, which could result in both reduced in-vehicle travel time
as well as elimination of transfer time.
• Faster service, which could result from introduction of express type service,
more direct service, or operation on an exclusive facility/right-of-way that is
free of highway traffic congestion.
COMMUTER model users can incorporate improved transit service in their scenarios by
making changes in either Frequency of Service (which is analyzed as "wait time") or
Faster Service (which is analyzed as "in vehicle time"). Improvements in ease of
accessing transit—even if they occur off-site—can be evaluated through the Access Time
feature described above.
Changes in Parking Costs - The model allows users to examine the effectiveness of
imposing parking fees. These may either be changes in fees that are currently in place, or
introduction of fees where parking is currently free. Because the user provides the model
with information on how parking cost would change for each relevant mode—Drive
Alone, Carpool, and Vanpool, it is possible to assume a different fee for each mode.
Through the use of positive (+) and negative (-) entries, it is also possible to treat the
change in parking fee as a surcharge (increase) or as a reduction or subsidy. For
example, if a site currently imposed a $1.00 per day charge for all vehicle parking, it
would be possible to test a $1.00 daily increase (+$1.00) on SOV users while keeping the
fee for carpools at the present $1.00 a day and making parking for vanpools free by
entering -$1.00.
When the model calculates the impact of the parking cost, or any other cost strategy, it
assumes that the cost is divided evenly among the vehicle's occupants. Thus, for SOVs,
a $1 daily charge increases the cost for the SOV traveler by $1, while an increase in $1
for a 2-person carpool would increase the cost for the carpool user by only $.50, and for a
10-person vanpool by only $0.10.
Fare Cost - The cost element that is most relevant to transit use is the level of fare paid.
The COMMUTER model allows the user to test the effect of reducing transit fares by
entering the daily change (i.e., per round trip) in cost to the user. Entering a negative (-)
value constitutes a subsidy (or cost reduction), while entering a positive (+) value
constitutes a fare increase. Under certain circumstances a user might want to examine
the effect on employee travel if a general fare increase was proposed by the local transit
agency.
Other Financial Incentives or Disincentives - In addition to the specific parking and
transit fare measures described above, employers may also utilize pricing strategies in
other ways to either encourage alternative mode use or discourage SOV use. For
-59-
-------
example, some employers have imposed a cost for parking on site and then used the
revenues to provide general subsidies to some or all of the alternative modes, including
walk and bicycle. To permit analysis of this type of measure, the COMMUTER model
has a place where the user can enter and test daily round trip cost changes—either
increases or decreases—for any or all of the available modes. These strategies can be
tested simultaneously with changes in parking pricing or transit fares, although the user
should not apply the same strategy in both places in the same scenario.
Analytic Approach for Estimating Travel Impacts
Logit Mode Choice Model - The impact of travel time or cost strategies on travel
behavior are determined in the COMMUTER model through use of an analytical
technique known as a logit mode choice model. Models of this type are considered to be
the state of the practice in transportation planning and policy studies. Most major
metropolitan areas have developed these models from local data and use them routinely
within their regional planning process. They are used by metropolitan planning agencies
for developing regional plans, analyzing the impact of major investment alternatives, or
responding to regulatory requirements (such as conformity analysis).
Logit mode choice models are particularly appropriate for examining traveler response to
change in travel time or travel cost. Their structure explicitly includes the various
elements of time and cost that were discussed in the previous section. The value of these
elements to the traveler is reflected through coefficients that have been developed through
a statistical estimation process applied to extensive regional travel data. The structure of
the model also expressly incorporates the very practical consideration of determining the
effect of having several strategies operating at the same time—possibly differently on the
individual modes—and having the modes compete with each other based on their relative
cost and time advantages.
A logit mode choice model relates the utility or "attractiveness" of a travel mode to the
consumer in relation to the utility of the alternatives. The traveler selects the mode that
maximizes his/her utility, i.e., which provides the greatest total value. The model does
this by calculating the statistical probability that each mode will be chosen, and the
individual is presumed to choose the mode with the highest probability. When this
process is performed for a travel population, rather than an individual traveler, the
probabilities become modal shares. For example, if the model computes probabilities of
0.758 for drive alone, 0.133 for transit, and 0.109 for carpool, the individual would be
expected to drive alone, since drive alone commands the highest probability. However, if
these probabilities were obtained as the result of applying the model to a population of
travelers, the probabilities would signify mode shares for that group, i.e., 75.8% would
drive alone, 13.3% would use transit, and 10.9% would ride in carpools.
In the logit model, the share of each available mode depends on the existence and relative
attractiveness of other available travel modes. The probability structure ensures that the
total of the shares of all modes will not exceed 100%.
-60-
-------
The probability of selecting a given travel mode m is calculated through the following
equation:
P(/n) =
eum
where
p(m) is the share of mode m;
e is the exponential function (e to the power x, where e = 2.7183);
um is the utility equation for mode m; and
UJ...U; represents the utility of other available modes 1 through i.
The utility of each alternative mode is usually a linear function that incorporates the
important choice attributes of the mode, such as travel time and cost, coupled with
weights, or coefficients, that represent the relative importance of the attributes to the
traveler. Logit mode choice model coefficients are typically calculated from household
travel survey data using a statistical curve-fitting technique known as maximum
likelihood estimation. A typical utility function might look like the following:
U(auto) = c+ a(time) + 6(cost)
Where:
U(auto) is the utility of the auto mode;
time and cost represent travel time and cost for the trip using auto;
c is a constant representing the utility of the mode that is not captured by travel
time or cost; and
a and b are coefficients applied to time and cost, respectively.
This equation produces an S-shaped curve, as shown in Figure 6-1. The graph shows the
percentage use of mode m as a function of the disutility of mode m* A distinguishing
* While the traveler choosing among alternatives attempts to "maximize his utility" when picking his
preferred alternative, in the case of travel mode choice where the key attributes are time and cost, the best
(continued...)
-61-
-------
feature of this curve is that its slope is steepest in the middle, and this is where small
changes in utility (or disutility) have the greatest effect on mode share. At the high and
low end of the curve, the curve flattens out, indicating that even relatively large changes
in disutility have little impact on the mode's share at that point.
The Pivot-Point Approach - In the COMMUTER model, just as in the FHWA TDM
Model, the logit procedure is applied in what is known as apivotpoint approach. In a
traditional application of a logit mode choice model in a regional planning study, it
would be necessary to relate to the model detailed information on the time, cost and other
relevant factors for each alternative, and for each traveler or origin destination pair that
make up the analysis. This must be done for each alternative or policy analyzed,
requiring a significant amount of time and effort.
The pivot point approach is a less data-intensive method for applying logit mode choice
models. Rather than gathering information on the time and cost components for each
mode for each trip, the analyst has only to describe how much travel time or cost
components will change as a result of the test strategies. In the pivot point approach, the
logit model is not used to estimate a mode share, but to estimate the change in a known
mode share, given a change in transportation system characteristics. In this case, the
procedure starts with a set of mode shares and an assumed change in travel time or travel
cost for each mode. The same logit coefficients described above can then be used to
estimate what the new mode shares would be. The mathematical expression for this,
which can be derived from the basic logit model equation, is as follows:
Figure 6-1
100%
90%
80%
70%
£!
Jl 60%
OT
£ 50%
01
O 40%
30%
20% '
10% '
0%
Disutility of Mode m
(...continued)
alternative is the one that offers the least time and cost. In this situation, because higher values of time and
cost are seen as disutility, the consumer's selection of the best alternative is instead a process of
"minimizing his/her disutility."
-62-
-------
where:
p '(m) = original share of mode m
p(m) = the new share of mode m
AU(m) = the change in disdisutility of mode m = a (Travel Time) + b (Travel
Cost)
Because this equation does not contain any reference to the original (or background)
values of travel time or Travel Cost, it is not necessary for the user to know those values
or any other conditions about the original trip. All of that information is "contained" in
the disutility value for the original mode share. Thus, the user only needs to inform the
model of the starting mode shares, and the procedure will "pivot" off the point described
by the existing shares and adjust those shares in relation to the magnitude of time or cost
changes introduced by the scenario and the mathematical relationships in the logit
equation.
As indicated in the preceding section, mode shares predicted by a logit model follow an
S-shaped curve relationship between disutility for a mode and the probability of its
choice, or, as shown in Figure 6-2, the Modal Share.
As Figure 6-2 illustrates, a given mode m has a starting share of 10%. A strategy or set
of strategies is applied that reduces the disutility of using mode m by a particular amount,
and this change improves the attractiveness of mode m sufficient to increase its share to
30%. The reader can readily see that if the starting share for mode m were different from
10%, then the same change in disutility would produce a very different outcome in
resulting share. If the starting share were higher, the predicted increase would be greater;
if it were smaller than 10%, the predicted change in share would be less than in the
example.
-63-
-------
Figure 6-2
Modal Share
Disutility of Mode m
"Seeding" of a Zero Share Mode - One limiting characteristic of the pivot point
approach occurs when one of the modes being targeted for change has an initial share of
"zero." This is the case if, for example, transit or vanpool were to be introduced under a
commuter choice program, but there was no current history of use with that mode.
Because the pivot point approach requires some initial share value to pivot from, any
time or cost changes introduced through a set of strategies would have nothing to pivot
from, and the model would predict no change.
There are two basic ways of dealing with this situation:
1. A nominal starting share can be assumed, such as 0.1%. While the changes
induced in this share through time or cost incentives will be small, they will at
least register the mode as part of the alternative choice set.
2. The user can first employ one of the Employer Support Strategies (Chapter 4),
which serves to arithmetically add an incremental share to what is already
there. If the starting share is 0% and the user specifies a program that
increases share by 2%, then the starting share for the zero mode becomes 2%
for subsequent application of time or cost strategies when the procedure
reaches the logit pivot-point procedure at the end of the computation process.
If the user does not employ option (2) above, the COMMUTER model has been set up to
enter a default share of 0.1% for modes that enter the logit pivot point process with zero
starting share and that have time or cost strategies targeted at them.
-64-
-------
Coefficients Used in Pivot-Point Logit Model Approach
The weights used in the utility expression of the logit model are coefficients, derived
from statistical analysis of relationships embedded in travel survey data. Generally, these
coefficients are developed from an estimation process using local survey data. Many
metropolitan areas conduct new urban travel surveys approximately once a decade for the
purpose of updating and upgrading previously-developed models. Some areas do not
conduct travel surveys and develop new models from scratch because these are very
resource intensive and mathematically challenging exercises. Rather they will "borrow"
models and coefficients from other areas believed to be similar in character, and
"calibrate" the models to their own situation using modest local information. Areas that
have their own existing travel demand models that were developed from very dated
information may opt to simply update their coefficients from modest new information
and "calibrate" those coefficients to where they replicate local behavior when tested.
The logit pivot point procedure in the COMMUTER model incorporates coefficients for
the following time or cost elements:
Coefficient
In- Vehicle Travel Time
(IVTT)
Out-of- Vehicle Time
(OVTT), Walk Time
Out-of- Vehicle Time
(OVTT), Wait Time
Auto Out-of-Pocket Cost
Transit Cost
Strategies Applied to:
Faster Transit Service
Walk Access Time for:
- Auto drive alone
- Carpool
- Vanpool
- Transit
- Bicycle
- Pedestrian
Transit Wait (or Transfer) Time
Parking Cost for:
- Drive Alone
- Carpool
- Vanpool
Transit Fare
COMMUTER model users are asked to indicate what coefficient values they will use in
their analysis. The model provides for two alternatives:
• Site Specific Coefficients: All users are encouraged to obtain and use the
coefficients that have been specifically developed for their respective area by
-65-
-------
the regional planning agency, or MPO. Coefficients have been compiled for a
large number of U.S. metropolitan areas, and have been incorporated into the
COMMUTER model. Users may search the menu for their respective areas,
and if it is included in the list, simply click on the choice to load those
coefficients into the model. Table 6-2 lists the sites that are included in the
COMMUTER model at this time, and the respective values of their
coefficients.
Table 6-2
Logit Mode-Choice Coefficients for Individual Urban Areas
Location
Albuquerque
Atlanta
Baltimore
Boston
Chicago
Cleveland
Columbus
Dallas
Denver
Detroit
Houston
Los Angeles
Milwaukee
New York
Philadelphia
Phoenix
Portland
Reno
Sacramento
Year
1992
2002
1993
1991
1990
1994
1999
1996
1997
1996
1985
1996
1991
1996
1986
1991
1994
1991
2001
In-Vehicle
Travel Time
(min)
All Modes
-0.0209
-0.0256
-0.0300
-0.0314
-0.0282
-0.0178
-0.0213
-0.0544
-0.0180
-0.0512
-0.0220
-0.0450
-0.0157
-0.0113
-0.0391
-0.0167
-0.0394
-0.0275
-0.0250
Out-of-Vehicle
Travel Time (min)
Walk
Time
-0.0219
-0.0639
-0.0750
-0.0330
-0.0440
-0.0444
-0.0640
-0.0640
-0.0540
-0.0186
-0.0568
-0.1073
-0.0412
-0.0380
-0.0316
-0.0206
-0.0646
-0.0550
-0.0380
Transit
Wait
-0.0978
-0.0256
-0.0750
-0.0550
-0.0960
-0.0378
-0.0465
-0.0640
-0.0180
-0.0186
-0.0568
-0.0423
-0.0412
-0.0554
-0.0511
-0.0304
-0.0397
-0.0550
-0.0380
Out-of-Pocket Travel
Cost (cents)
Auto
Parking
-0.0031
-0.0031
-0.0043
-0.0173
-0.0021
-0.0034
-0.0016
-0.0056
-0.0014
-0.0041
-0.0154
-0.0025
-0.0045
-0.0004
-0.0026
-0.0053
-0.0135
-0.0167
-0.0025
Transit
Fare
-0.0031
-0.0013
-0.0043
-0.0083
-0.0008
-0.0024
-0.0016
-0.0055
-0.0012
-0.0041
-0.0061
-0.0025
-0.0045
-0.0004
-0.0012
-0.0053
-0.0135
-0.0067
-0.0025
-66-
-------
Location
San Diego
San Francisco
San Juan
Santa Cruz
Seattle
Tucson
Washington D.C.
Year
1995
1990
1990
1990
1990
2000
1994
In-Vehicle
Travel Time
(min)
All Modes
-0.0250
-0.0333
-0.0366
-0.0163
-0.0176
-0.0178
-0.0300
Out-of-Vehicle
Travel Time (min)
Walk
Time
-0.0500
-0.0931
-0.0717
-0.0325
-0.0206
-0.0400
-0.0750
Transit
Wait
-0.0250
-0.0523
-0.0752
-0.0325
-0.0155
-0.0200
-0.0750
Out-of-Pocket Travel
Cost (cents)
Auto
Parking
-0.0069
-0.0021
-0.0066
-0.0045
-0.0024
-0.0018
-0.0043
Transit
Fare
-0.0025
-0.0021
-0.0066
-0.0036
-0.0024
-0.0018
-0.0043
Source: Coefficients obtained from a review of current practice by Cambridge Systematics, Inc. (2005).
• Default Coefficients: For users who do not find their city in the menu, a set of
default coefficients has been developed through a synthesis of the cities that
are included.
These coefficients are listed in Table 6-3.
Table 6-3
Default Coefficient Values
All Cities
In-Vehicle
Travel Time
(min)
All Modes
-0.0253
Out-of-Vehicle
Travel Time (min)
Walk Time
-0.0473
Transit
Wait
-0.0466
Out-of-Pocket
Travel Cost (cents)
Auto-
Parking
-.0.0056
Transit- Fare
-0.0040
Source: Cambridge Systematics, Inc., 2005.
Model users reflecting upon the values of the various coefficients will reach the
following conclusion as to which strategies are likely to have the greatest impact. In
general, cost strategies will have a greater impact (per percentage change in existing
conditions) than travel cost. Auto parking costs are known to have major impact on
travel choice, and this is reflected in the coefficient that is higher than any other in the
group. Auto parking cost has almost twice the impact on mode choice as a change in
transit fare. Among travel time elements, strategies that reduce the travel time that
occurs "outside" the vehicle, i.e., where the user is either walking or waiting to connect
-67-
-------
with the mode, are generally valued more than changes that occur to the travel time once
the user is "inside" the vehicle. This is because walking and waiting require more active
effort from the traveler, whereas once aboard and underway, the traveler is generally
seated, comfortable, and able to use the time for some other purpose. These relationships
can be a guide as to where attention should be placed when designing an effective
commuter choice program.
Example Calculation of Change in Time or Cost through Logit Pivot-Point
Model
The following is a simple example to illustrate how the logit pivot-point analysis
procedure would evaluate a typical program that involves strategies that change travel
time and cost. This example is built on the following assumptions:
• Area type and size: This example involves a "large" metropolitan area, and
we are conducting a "regional level" analysis. Because this is a hypothetical
area, the specific model coefficients for the area have not been included in the
model's menu. As a result, we will use hypothetical user-supplied coefficients
for this area. The values for these coefficients are as follows:
IVTT (all modes) = - .0281 min
OVTT (walk, all modes) = - 0.0521 min
OVTT (wait time, transit) = - 0.0584 min
Cost (auto parking) = - 0.0094 cents
Cost (transit fare) = - 0.0065 cents
• Current Mode Shares: The following local modal shares are assumed:
Drive Alone: 75.4%
Carpool: 13.2%
Vanpool: 0.5%
Transit: 5.3%
Bicycle: 0.4%
Walk: 4.0%
Other: 1.1%
• Strategies: The following time and cost related strategies will be included in
the test scenario:
Access Time Changes:
Preferential parking: Drive alone, +2 min.; Carpool, -1 min.; Vanpool, -1 min.
Site Access improvements: Transit: -2 min.; Bike: -2 min.: Walk: -2 min.
Transit Service Improvements:
-68-
-------
More Frequent Service (Less Wait Time): Transit, -2.5 minutes
Faster Service (Less Travel Time): Transit, -5.0 min.
Travel Cost Changes:
Parking Charges: Drive Alone, +$1; Carpool, $0; Vanpool, -$2
Transit Subsidy: Transit, -$1 per day
The first step taken by the model will be to recompute the change in utilities for each
mode [-AU(m)], as follows:
Drive Alone:
-AUDA = - 0.0521 (AOVTTWalkDA) - 0.0094 (AParking CostDA)
= - 0.0521 x (2 min) - 0.0094 x ($l)x 100 cents/$ = -1.0442
Carpool:
-AUCP = - 0.0521 (AOVTTWalkCP) - 0.0094 (AParking CostCP)
= - 0.0521 x (-1 min) - 0.0094 x ($0) x 100 cents/$ = 0.0521
Vanpool:
-AUVP = - 0.0521 (AOVTTWalkw) - 0.0094 (AParking CostVP)
= - 0.0521 x (-1 min) - 0.0094 x ($-2) x 100 cents/$ = 1.9321
Transit:
-AlV = - 0.0281 (AlVTTTR/2) - 0.0521 (AOVTTWalk^) -
0.0584 (AOVTTWaiV2) - 0.0065 (AFare CostTR)
= - 0.0281 x (.5 min/2) - 0.0521 x (-2 min) -
0.0584 x (-2.5 min/2) - 0.0065 x ($-1) x 100 cents/$ = 0.8975
(Note that the changes in transit service frequency and time are divided by two
to reflect two-way commute trips being applied to daily coefficients.)
Bicycle:
-AUBK = - 0.0521 (AOVTTWalkBK)
= -0.0521 x (-2 min) = 0.1042
Walk:
= - 0.0521 (AOVTTWalk^)
= -0.0521 x (-2 min) = 0.1042
-69-
-------
Next, revised mode shares are computed using equation (1):
P'DA = [PDA x exp (-AUDA)] / {[(exp (-AUDA) - 1) x PDA] + 1}
= [0.754 x e -1-0442] / {[(g-1-0442 - 1) x 0.754] + 1}
= 0.2655/0.4460 = 0.5953 = 59.53%
P'CP = [Per x exp (-AUCP)] / {[(exp (-AUCP) - 1) x Pcp] + 1}
= [0.132 x e +o.o52i] / {[(e+o.°52i . i) x 0.132] + 1}
= 0.1393/0.8862 = 0.1572 = 15.72%
P'w = [PVP x exp (-AUW)] / {[(exp (-AUVP) - 1) x Pw] + 1}
= [0.005 x e +1-9321] / {[(e +1-9321 . ^ x o 005] + ^
= 0.0345/0.9952 = 0.0347 = 3.47%
PV = [PTR x exp (-AUTR)] / {[(exp (-AUTO) - 1) x PTR] + 1}
= [0.053 X e +0.8975] / {[(g +0.8975 _ ^ x Q Q53] + 1}
= 0.1308/0.9537 = 0.1371 = 13.71%
P'BK = [PBK x exp (-AUBK)] / {[(exp (-AUBK) - 1) x PBK] + 1}
= [0.004 X e +0.1042] / | [(g +0.1042 _ ^ x Q QQ4] + ! j
= 0.0044/0.9960 = 0.0045 = 0.45%
P'WK = [PWK x exp (-AIV)] / {[(exp (-AIV) - 1) x PWK] + 1}
= [0.040 X e +0.1042] / | [(g +0.1042 _ ^ x Q Q4Q] + j j
= 0.0446/0.9616 = 0.0464 = 4.64%
(Note that the revised share for the other (e.g., telecommute) mode is still 1.1% since
it is unaffected by these specific strategies.)
Final mode shares are then determined by normalizing the revised shares (to total 100%):
-70-
-------
Mode
Drive Alone
Carpool
Vanpool
Transit
Bicycle
Walk
Other
Total
Revised
59.67%
15.72%
3.47%
13.71%
0.45%
4.64%
1.10%
98.61%
Normalization
0.5953 -
0.1572-
0.0347 -
0.1371-
0.0045 -
0.0464 -
0.0110-
-0.9861
-0.9861
-0.9861
-0.9861
-0.9861
-0.9861
-0.9861
Final
= 60.4%
= 15.9%
3.5%
= 13.9%
0.5%
4.7%
1.1%
100.0%
###
-71-
-------
7. Converting Travel Impacts to Emissions Reductions
Overview of Procedure
Once the basic travel impacts for a given set of strategies have been calculated as
described in Chapters 4, 5, and 6, the COMMUTER model then proceeds to estimate the
corresponding change in emissions. This procedure is illustrated in Figure 7-1.
Figure 7-1
Translation of Travel Demand Changes into
Emissions Reductions
Change in
Vehicle
Trips
Calculate
Change in
Peak Period
VMT
Calculate
Change in
Peak Period
Emissions
Alternative
Work
Schedules
Shift of Trips
From Peak
to Off-Peak
Calculate
Change in
Off-Peak
Period VMT
Calculate
Change in
Off-Peak
Period
Emissions
-72-
-------
The changes in travel result from application of Employer Support Programs, Travel
Time or Cost Strategies, and Alternative Work Schedules. As pictured in Figure 7-1, the
first of these two strategy groups produces changes in overall trip making (i.e., number of
vehicle trips), while the Alternative Work Schedules strategies produce changes in the
time that the trips occur, namely in the Peak or Off-Peak.
The change in modal split induced by the Employer Support Programs and the Time/Cost
Strategies is mated with information about average modal trip lengths, and used to
calculate the change in vehicle miles of travel (VMT). In combination with the
information on the proportion of trips shifted from peak to off-peak, the COMMUTER
model then calculates the change in VMT in both peak and off-peak periods.
Using the change in VMT for the two time periods, the COMMUTER model then applies
per-mile vehicle emissions rates (or factors) to these VMT estimates to produce an
estimate of the change in emissions due to the particular analysis scenario. Emissions
changes are calculated for the following pollutants:
• Volatile organic compounds (VOC);
• Carbon monoxide (CO);
• Oxides of nitrogen (NOx)
• Carbon dioxide (CO2);
• Particulate matter (PM2.5), reported for particles smaller than 2.5 microns
(PM2.5); and
• Six air toxics — acetaldehyde, acrolein, benzene, 1, 3-Butadiene,
formaldehyde, and methyl tertiary butyl ether (MTBE).
The procedures and assumptions associated with each of these steps are presented in the
sections below.
Changes in Total Trips
Once final mode shares have been calculated based on application of the lookup table or
logit pivot-point model techniques, these mode shares are used to estimate changes in
vehicle-trips and VMT. The model calculates the percentage of persons who will travel
by each mode. For example, assume that the affected travel population consists of 5,000
employees, and the starting mode shares are as follows:
-73-
-------
Drive Alone: 75.4%
Carpool: 13.2%
Vanpool: 0.5%
Transit: 5.3%
Bicycle: 0.4%
Walk: 4.0%
Other: 1.1%
100%
Then 5,000 x 0.754 = 3,770 people would drive alone, 660 would travel by carpool, 25
would travel by vanpool, etc. To determine the number of vehicle trips represented by
this population's modal distribution, it is necessary to divide each vehicular mode by its
passenger occupancy. Only the drive alone, carpool, and vanpool modes are assumed to
generate vehicle trips in this computation; transit, bicycle, walking and other are not
associated with making of a private motor vehicle trip. Passenger occupancy for drive
alone is, of course, 1.0. Values of 2.2 persons are assumed for carpools, and 6.0 persons
for vanpools. The user has the option of changing these assumptions if better information
is available locally.
The vehicle trips generated by the example population in relation to the starting mode
shares would be as follows:
Drive Alone: 3,770 + 1.0 persons = 3,770 vehicle trips
Carpool: 660 + 2.2 persons = 300 vehicle trips
Vanpool: 25 + 6.0 persons = 4.2 vehicle trips
for a total of 4,074 vehicle trips being generated by this population of 5,000 employees.
Whether these are regarded as only one-way trips or daily round trips depends on how
the analyst has defined them in the database. In the COMMUTER model these would be
daily round trips, meaning that it would be necessary to multiply 4,074 vehicle round
trips by 2, to yield 8,148 daily vehicle trips.
The same process is then applied to the new mode shares to arrive at the change in
vehicle trips produced by the test scenario. To illustrate, assume that the test scenario
brought about the following new pattern of mode shares:
-74-
-------
Drive Alone: 72%
Carpool: 15%
Vanpool 2%
Transit: 6%
Bicycle: 1.5%
Walk: 7%
Other: 0.5%
100%
The new vehicle trip total generated by this population would be calculated as follows:
5,000 persons [ (0.72)/(1.0)DA + (0.15)/(2.2)CP + (0.02)/(6.0)VP] x 2
or 7,916 vehicle daily trips
So the tested scenario in this case would reduce vehicle trips by 8,148 - 7,916 = 232
trips.
Changes in Total VMT
To convert changes in trips to changes in VMT, the COMMUTER model applies
information on average trip length for each of the modes to the modal split for both
existing and new conditions. An average trip length is specified in the model for each
mode.
The example assumes the following user-specified trip lengths:
Drive Alone:
Carpool:
Vanpool:
Transit:
Bicycle:
Walk:
Other:
Average
11. 85 miles
12.21 miles
17. 70 miles
11. 42 miles
1.80 miles
1.00 miles
11. 42 miles
11. 42 miles
The COMMUTER model would calculate baseline VMT for the 5,000 employee
example used above as follows:
Total VMT = [ (3,770 DA trips) x (11.85 miles) + (300 CP trips) x (12.21 miles)
+ (4.2 VP trips) x (17.70 miles) ] x 2 trips per day
= 96,824 daily VMT
-75-
-------
When a test scenario has been applied that changes mode split, a slightly different
procedure is used to calculate the new VMT. The reason for this is the assumption that,
when travelers are taken from the drive alone mode and shifted to a new mode, they also
assume the trip length of that new mode. Hence the trips diverted from drive alone are
assumed to reduce the cumulative VMT for drive alone by the amount of the new mode's
trip length. Assume, for example, that an auto driver shifted to vanpool. The average
trip length for a vanpool trip is 17.7 miles, compared to 11.85 for the drive alone. So the
shifted trip in this case would deduct 11.85 daily drive alone VMT, and add 17.7 miles
divided by 6.0 occupants, or 2.95 miles after switching to vanpool, for a net change of
-8.9 miles.
To effect this debit and credit accounting for VMT based on which modes are
exchanging riders, the following rules guide the analysis:
1. The overall average person-trip length for all individuals making a work-trip
remains the same (i.e., no one changes the location of their workplace or
residence as a result of the implemented programs).
2. Average person trip-lengths for all modes except drive-alone remain fixed
before and after the program. This means, for example, that those who shift
from driving alone to bicycling have shorter-than-average trip lengths, while
those who shift from driving alone to vanpooling have longer-than-average
trip lengths.
3. Trips eliminated due to telecommuting or compressed work weeks are
assumed to be the same length as the overall average person-trip, so that
elimination of these trips does not change the average person or motor-vehicle
trip length. (This conservative approach assumes that the average trip lengths
will remain the same, even with a reduction in the number of automobiles
using the transportation system as a result of increased telecommuting or
compressed work weeks.)
4. The average person-trip length for drive-alone commuters is then re-computed
so that the overall average person-trip length is the same as it was before
implementation of the programs.
An illustration of how this calculation would be performed is shown in the following
table using the previous 5,000 employee example.
-76-
-------
Mode
D. Alone
Carpool
Vanpool
Transit
Bike
Walk
Other
Total
Starting
Share
75.4%
13.2%
0.5%
5.3%
0.4%
4.0%
1.1%
100%
Average Trip
Length
11.85
12.21
17.70
11.42
1.80
1.00
11.42
11.42
Daily
VMT
89,349
7,326
149
0
0
0
0
96,824
New Mode
Share
72%
15%
2%
6%
1.5%
7%
0.5%
100%
New Avg. Trip
Length
11.66
unchanged
unchanged
unchanged
unchanged
unchanged
unchanged
unchanged
New Daily
VMT
83,952
8,325
590
0
0
0
0
92,867
The baseline mode shares combined with the baseline average trip lengths produce
96,824 total daily VMT. Under the test scenario, the mode shares are changed as shown
in column 5. To calculate the associated new VMT, a new average trip length must be
calculated for drive alone. It is assumed that all other modes retain their original average
trip length and that overall average trip length remains at 11.42 miles, so that any
changes in individual trip length are assumed to occur in the drive alone mode. To
calculate the adjusted drive alone trip length, the new shares in column 5 are used to
weight the original trip lengths (for all modes except drive alone), and then the new drive
alone trip length that preserves the original 11.42 mile overall trip length is calculated by
solving the following equation for ATLDA:
(0.72 x ATLDA) + (0.15 x 12.21) + (0.02 x 17.7) + (0.06 x 11.42) + (0.015 x 1.8) +
(0.07 x 1.0) + (0.005 x 11.42) = 11.42
Solving for ATLDA yields 11.66 miles for the revised drive alone trip length. Using these
new shares and average trip lengths, the daily VMT resulting from the new scenario is
calculated as 92,867.
As an additional input to the emission calculations, average motor-vehicle trip lengths
(existing and new) must be also be computed. The average motor-vehicle trip length
(ATLMotorVehide ) is simply the average of trip lengths for the drive-alone, carpool, and
vanpool modes, weighted by the mode shares for each mode. For the example above,
these values would be calculated as follows:
Baseline ATLMotorVehicle = (75.4% x 11.85) + (13.2% x 12.21) + (0.5% x 17.70)
= 10.64 miles
Post-Scenario ATLMotorVehicle = (72% x 11.66) + (15% x 12.21) + (2.0% x 17.70)
= 10.58 miles
-77-
-------
Allocation Between Peak Period and Off-Peak
Once the changes in vehicle trips and VMT due to the tested scenario have been
computed, the next step before calculating the emissions impacts is to reallocate the trips
and VMT between the peak and off-peak periods. This adjustment would be necessary
only if the scenario had also included Alternative Work Schedule strategies that affected
the timing of trips. Specifically, this would be the result of applying Flexible Work
Hours or Staggered Work Hours strategies.
The COMMUTER model begins with an initial allocation of trips between the peak and
off-peak periods. This is done by first computing the number of vehicle trips and VMT
for the "affected employment" using the methods above, and then using information
provided by the user on the percentage of daily commute trips that occur in the peak
period. This factor, appearing on the Local Data screen of the model as Percent of Work
Trips in Peak Period, is supplied with a default computed from national data of 61.7%.*
The user is encouraged to supply local information for this value if available.
When Flexible Work Hours or Staggered Work Hours strategies are applied in a program
scenario, the COMMUTER model estimates the percentage of trips that would shift the
time of travel from the peak to the off-peak periods (process discussed in detail in
Chapter 5). As part of this calculation, the COMMUTER model also takes into account
the length of the peak period. For areas where the peak period lasts the "standard" three
hours, the model predicts that 13.9% of the eligible employee trips will actually be
shifted "outside" the three-hour peak period. If the peak period is shorter, say two hours,
a higher percentage (28.7%) of trips will be shifted, whereas if the peak is unusually
long, say four hours, then only 8.5% will be shifted. The user is asked to supply the
model with the length of the local peak period, if it is different from the 3.0 hour default
that is provided in the model. The information in Table 5-1 is repeated below in Table 7-
1 to illustrate the relationship between length of peak period and percent of trips shifted
that is used in the model.
1995 Nationwide Personal Transportation Survey, Federal Highway Administration.
-78-
-------
Table 7-1
Percent of Trips Shifted by Length of Peak Period
Length of Peak Period (hrs)
2.0
2.5
3.0
3.5
4.0
4.5
5.0
Percent of Trips Shifted
28.7
19.2
13.9
10.6
8.5
7.1
6.0
Source: Estimating the Effect of Alternative Work Schedules on Travel
Activity and Emissions, FHWA TDM Evaluation Model, 1993
The sample illustration is used below to provide an example of how the COMMUTER
model would perform an allocation of trips between peak and off-peak based on
Alternative Work Schedule programs and local information about the duration of the
peak period and the current share of commute trips occurring in the peak. The
calculation process is described below.
The model first determines the distribution of baseline trips that occur in the peak and
off-peak periods. The user has selected the default of 61.7% to represent the percentage
of these trips that are occurring in the peak and off-peak periods:
Peak-Period Trips = 5,000 person trips x 0.617 = 3,085
Off-Peak Trips = 5,000 person trips x 0.383 = 1,915
(Note that these are daily round trips; if the individual home-to-work and return
work trips are to be separated, the basis would increase to 10,000 one-way trips.
Care should be taken when calculating VMT that the basis is consistent with
whether the user is assuming daily or one-way VMT and emissions.)
Assuming the same baseline mode shares as previously, this would produce the following
peak and off-peak vehicle trip and VMT allocation:
-79-
-------
Mode
D. Alone
Carpool
Vanpool
Transit
Bike
Walk
Other
Total
Starting
Share
75.4%
13.2%
0.5%
5.3%
0.4%
4.0%
1.1%
100%
Daily
Vehicle
Trips
7,540
600
8.4
0
0
0
0
8,148
Daily
VMT
89,349
7,326
149
0
0
0
0
96,824
Daily Peak
Vehicle
Trips
4,652
370
5.2
0
0
0
0
5,027
Daily
Peak
VMT
55,128
4,520
92
0
0
0
0
59,740
Daily Off-
Peak
Vehicle
Trips
2,888
230
3.2
0
0
0
0
3,121
Daily Off-
Peak VMT
34,220
2,805
57
0
0
0
0
37,083
When the hypothetical scenario is then applied that changes the mode shares, the table
above changes to the following:
Mode
D. Alone
Carpool
Vanpool
Transit
Bike
Walk
Other
Total
Revised
Mode
Shares
72%
15%
2%
6%
1.5%
7%
0.5%
100%
Daily
Vehicle
Trips
7,200
681
16.7
0
0
0
0
7,898
Daily
VMT
83,952
8,325
590
0
0
0
0
92,867
Daily Peak
Vehicle
Trips
4,442
420
10.3
0
0
0
0
4,872
Daily
Peak
VMT
55,128
4,520
92
0
0
0
0
57,299
Daily Off-
Peak
Vehicle
Trips
2,758
261
6.4
0
0
0
0
3,026
Daily Off-
Peak VMT
34,220
2,805
57
0
0
0
0
35,568
Now if the test scenario applies a Flexible Work Hours program strategy that assumes
that 30% of all employees will be eligible for the program, the following impacts would
be calculated (and assuming zero existing eligibility):
First, the model calculates the number of employees who would be affected as:
5,000 employees x 61.7% in peak x 80% office x 30% eligible
-80-
-------
= 740 employee trips eligible for shifting
Next, the model determines how many trips may be shifted outside the peak
period based on the length of the peak period. In this example, it is assumed that
the peak period has a 3.0 hour duration. Hence the model would assume that
13.9% of the eligible trips would actually shift to the off-peak. Coupled with the
default participation rate of 22%, the example yields:
740 person trips x 13.9% x 22% = 23 person trips shifted from peak to off-
peak
(23 x 2 = 46 per day)
Those person trips would be presumed to shift from peak to off-peak in the same
proportion as the modal shares for all affected employees as engineered by the
test scenario. In other words, the person trips shifted would be at the rate of 72%
drive alone, 15% carpool, 2% vanpool, 6% transit, 1.5% bicycle, 7% walk, and
0.5% other. Thus, for each person trip shifted, the model in this case would shift:
7,898/10,000 = 0.7898 vehicle trips
and
92,867/10,000 = 9.29 VMT
So if the Flexible Work Hours strategy tested would shift 23 daily person trips,
the model would remove 23 x (2 directions) x 0.7898 = 36 vehicle trips from the
peak period(s) and move them to the off-peak, and similarly shift 23 x 2 x 9.29 =
427 daily VMT from the peak to the off peak.
Daily Vehicle Trips, Peak
Daily VMT, Peak
Daily Vehicle Trips, Off-Peak
Daily VMT, Off-Peak
Pre-
Scenario
4,872
57,299
3,026
35,568
Change
-36
-427
+36
+427
Post
Scenario
4,836
56,872
3,062
35,995
This would be the new distribution of vehicle trips and VMT by time of day that would
be taken forward into the emissions analysis.
-81-
-------
Emission Impacts Methodology
The COMMUTER model uses the estimates of travel changes described above as the
means for computing corresponding emission reductions.
In a full analysis of travel and emissions impacts such as would occur in the regional
planning process when demonstrating SIP conformity, the detailed computation of travel
changes and their effect on overall regional trips, VMT, and travel speeds would be
processed through the accepted emissions model. In most areas, the MOBILE6
emissions model would be used; in California, the EMFAC equivalent model would be
used.
As with the traditional transportation analysis procedure that the COMMUTER model
simplifies and replaces for this purpose, so too is the MOBILE emissions model a fairly
time and data intensive analysis process. As a result, a simplified process has also been
developed and incorporated within the COMMUTER model to be consistent in detail and
accuracy with the transportation methods being employed, and maintaining the desired
application ease for the user.
The computed changes in travel activity from baseline levels are coupled with
MOBILE6-based emission factors to generate daily emission reduction estimates for
VOCs, NOx, CO, CO2, PM2 5, and air toxics. These reductions are computed as tons per
day or pounds per day changes in each respective pollutant, and are computed and
reported separately for peak and off-peak travel situations. This section explains the
basic emissions modeling methodology that is used in the COMMUTER model.
However, the complexity of the calculations in this process makes it impractical to
provide easy-to-use, manual procedures for generating emission reduction estimates.
Beyond this summary, the user is referred to the COMMUTER model itself to generate
these estimates.
MOBILE6.2 Emission Factors - To eliminate the need for the user to supply emission
factors representing local fleet conditions in the analysis area of interest, an extensive
series of MOBILE6.2 emission factors were generated to represent a variety of local
conditions and loaded into the COMMUTER model. A total of 1,152 separate sets of
emission factors were generated representing combinations of the following fleet
parameters:
• Calendar Year of Analysis (6) - 2007, 2009, 2010, 2013, 2019, or 2021;
• Season (2) -summer or winter;
Climate type (3) - mild, moderate, or severe;
• Inspection and maintenance (I/M) program type (4) -none, basic, enhanced, or
on-board diagnostics (OBD) only;
• Fuel type (2) - conventional gasoline or reformulated gasoline (RFG); and
• Petroleum Administration for Defense District (PADD) (5) - 1, 2, 3, 4, or 5.
-82-
-------
The "season" and "climate type" parameters are used in combination to determine the
temperature ranges specified for the MOBILE runs. These ranges are as follows:
Season
Summer
Summer
Summer
Winter
Winter
Winter
Climate Type
Mild
Moderate
Severe
Mild
Moderate
Severe
Min Temp, (deg F)
55
65
75
52
32
12
Max Temp, (deg F)
75
85
95
72
52
32
The fuel type and PADD parameters are used to establish a variety of fuel parameters.
Weighted average values are developed for each PADD based on refinery-level data.
The fuel parameter values used in the COMMUTER model default runs are shown in the
table attached as Appendix B. Emission factors were not generated for every possible
combination of parameters since, for example, RFG is not produced in some PADDs.
For each combination of the above parameters, separate emission factors were generated
for the following:
• Vehicle type (10) - Nine individual classes of light-duty vehicles, as well as
transit buses (diesel);
• Facility type (4) - Freeway, arterial, local road, and freeway ramp;
• Emissions type (2) - Starts (grams/start) and running (grams/mile); and
• Speed categories (13) - 5 mph increments from 5 to 65 mph.
In MOBILE6 database output, CO and NOx exhaust start emissions are labeled as
emissions type 2 and exhaust running emissions are labeled as emissions type 1. For
VOC and air toxics, the eight emissions types, representing both exhaust and evaporative
emissions, were allocated between starts and VMT (running) emissions in the following
manner:
-83-
-------
Source
Exhaust
Evaporative
Emissions Type
1
2
O
4
5
6
7
8
Description
Start
Running
Hot Soak
Diurnal
Resting
Run Loss
Crankcase
Refueling
Assigned To:
Start
VMT
Start
Neither
Neither
VMT
VMT
VMT
Diurnal and resting evaporative emissions are more or less independent of vehicle
activity, and therefore would not be reduced by reducing either vehicle starts or distance
traveled. Therefore, they were not assigned to either component and were removed from
the analysis.
Particulate matter is reported as an aggregate of the seven solid components of
particulates computed by MOBILE6.2. These include sulfates (SO4), organic carbon in
diesel exhaust (OCARBON), elemental carbon in diesel exhaust (ECARBON), total
carbon in gasoline exhaust (GASPM), lead, brake wear, and tire wear. Sulfur dioxide
(SO2) and ammonia (NH3), as gaseous components of exhaust emissions, undergo
secondary atmospheric reactions which make it difficult to directly relate these emissions
to PM quantities. Therefore, these components are not reported.
For CO2 and PM, the MOBILE6.2 model does not compute start emissions separately
and does not vary emissions based on speed. Therefore, only a single running emission
factor is reported for each vehicle type and facility type combination, and changes in
these emissions are based on changes in VMT only. For the other pollutants analyzed,
start-based emissions do not vary by speed so only a single start emission factor is
reported for each vehicle type and facility type combination.
Emission Reduction Calculation - Separate calculations were performed for each of the
two activity parameters modeled: VMT and trips. Emission reductions from each of
these travel impacts were then added together to produce combined emission reduction
impacts from changes in travel due to the TCMs evaluated. Each of these calculations,
described below, was performed for both the peak and off-peak periods.
First, reductions in VMT and vehicle trips due to the TCM strategies are computed by
subtracting existing activity levels from the "final" (after-TCM) activity levels. Note that
-84-
-------
under a regional analysis, the existing activity levels used in the COMMUTER model are
not the same as the baseline activity for the entire region. The existing activity in the
model is based on affected employment, not total population.
Second, the regional speed inputs were used to linearly interpolate the running emission
factors by speed increment to represent these specific speeds. Interpolated emission
factors were developed for each vehicle type and for freeway and arterial facility types,
for both peak and off-peak speeds. (Emission factors for local roads and freeway ramps
are calculated at a set speed in MOBILE and do not vary.)
Third, a weighted average running emission factor was developed for each pollutant for
the peak and off-peak periods. This factor is the weighting of individual factors by
vehicle and facility type, based on VMT mix and facility type mix provided by the user
or specified from COMMUTER default values.
Fourth, VMT reductions were then applied to the weighted average running emissions
factors to compute daily emission reductions due to changes in VMT.
Fifth, the change in vehicle trips (existing vs. final) was multiplied by the start emission
factors to compute daily emission reductions due to changes in vehicle trips (starts).
Finally, the emission reductions from changes in VMT and changes in trips were added
to obtain overall daily reductions in emissions.
Transit Bus Emissions - Some TCM strategies may include additional transit service
(e.g., increased transit service frequency). If the user chooses to analyze transit service
improvements, emissions increases from this additional service are estimated and
subtracted from the emissions decreases from reduced personal vehicle travel. Emission
factors for diesel transit buses were developed and incorporated in the model in the same
manner as for light-duty vehicles. The user is prompted to enter the additional daily
VMT from transit vehicles, and may also enter an assumed average speed for these
vehicles or use the default speed provided (14.8 mph). Speed-based emissions are then
calculated as described above, although speeds are not assumed to vary for peak vs.
off-peak periods. Start emissions for buses are not calculated as they are not relevant to
this analysis. In scenarios where transit service is not changed in the inputs, overall
transit service, and hence, overall bus emissions, is assumed to be unchanged by any
additional commuters shifted to transit in the mode share calculations.
User-Input Emissions Factors
The user may select the option of importing locally-developed emission factors from
their own MOBILE6.2 output file. If this option is selected, COMMUTER runs a
subroutine to process a MOBILE6 database-format output file into a file format suitable
for import into the COMMUTER model. The procedure for calculating emission
reductions is identical to the procedure applied to the default emissions scenarios
included with the COMMUTER model.
-85-
-------
###
-86-
-------
Appendix A
Definition of Modeling Terms
-------
Definition of Terms
Sources for Definition of Terms:
1. Manual of Regional Transportation Modeling Practice, 1993.
2. TDM Evaluation Model Documentation, 1993.
Affected Employment - Affected employment considers the amount of employees that
will be impacted by a particular commuter choice support program in a given analysis
area for both office and non-office categories. In the analysis presented in this Manual,
the effectiveness of some programs is assumed to differ for these two employment
categories. Total employment affected is also needed to calculate changes in total trips
and vehicle-miles of travel (VMT).
Destination - The geographic area in which any trip terminates.
Elasticity - In a causal relationship, the elasticity of i with respect to j is the percent
change in variable i with respect to the percent change in variable].
Functional Classification - The classification of urban and rural roadways by function.
Roadways at the top of the hierarchy (freeways) serve intercity and other long distance
traffic movements, roadways at the bottom (local roads) provide access to land.
Home to Work Trip - A person or vehicle trip that begins either at home or work
(including one trip end at work and the other at home).
HPMS - Highway Performance Monitoring System, a federally-mandated database
consisting of a representative sample of highway roadways by functional classification.
Level of Program Effort - Research has shown that the impact of employer support and
incentive programs and strategies vary according to the level of effort expended on the
program. The level of program effort is a particularly important dimension as it defines
what is actually meant by, for example, a carpool support program. This is an important
distinction in the analysis presented in this Manual. The program level will vary from 0
to 4, 0 representing no program and 4 representing a program of maximum effort. These
program levels are defined separately for carpooling, vanpooling, transit, and bicycle
commuting. Shown below is an example of the program level definitions for a carpool
support program.
• Level 0 - no program.
• Level 1 - includes carpool information activities (tied in with area-wide
matching) and a quarter time transportation coordinator.
• Level 2 - includes an in-house carpool matching service and/or personalized
carpool candidate get-togethers (including information activities) and a
quarter time transportation coordinator.
A-l
-------
• Level 3 - includes in-house carpool matching and information services plus
preferential (reserved, inside, and/or especially convenient) parking for
carpools, a policy of flexible work schedules to accommodate carpools, and a
half time transportation coordinator.
• Level 4 - includes in-house carpool matching and information services plus
preferential parking for carpools, flexible work schedules, and a full time
transportation coordinator.
Logit - A choice model formulation based on the principle that individuals maximize
utility in choosing among available alternatives. The logit formulation involves
specifying a utility function for each individual, with a deterministic component (that is,
one which depends on characteristics of the individual and of the alternatives) and a
stochastic disturbance (or error term). The logit model, follows the assumption that the
error terms are independent and share the same probability distribution. Most
transportation demand model systems use a logit formulation for mode choice.
Mode Share - Mode shares represent the percentage of person and vehicle travel by type
of transport. Typically, mode shares are measured by drive alone (single occupant
vehicles), high occupancy vehicles (carpooling, vanpooling, and other ridesharing
activity), transit (bus, rail), walking, and bicycling modes of transport.
Off-Peak Period - Occurring during periods of relatively low traffic, not during a peak
period. Trip making in this period typically represents mid-day, non-commute conditions
occurring between 9 a.m. and 3 p.m.
Origin - The geographic location at which a person or vehicle trip begins.
Peak Hour - The hour during which the maximum traffic occurs within the defined peak
period(s). The peak hour during which traffic is highest varies by roadway functional
classification and place to place (i.e., origin-to-destination).
Peak Period - Whether categorized by purpose or by geographic area, person or vehicle
trips occur at different rates at different times of the day. A graph of trips by time of day
typically reveals one or more peaks. These peaks play a key role in conventional travel
demand analysis, which focuses on maximum infrastructure need in each corridor. The
dominant weekday peaks are in the morning (AM Peak) and the late afternoon (PM
Peak), obviously related to the timing of home to work (commute) trips. A peak can be
characterized by its maximum trip rate (in trips per unit of time) or by duration over
which some threshold trip rate is maintained. The portions of the peak before and after
the peak hour are called shoulders of the peak.
Peak Spreading - Lengthening of the peak period, usually accompanied by a flattening
of the peak period(s).
A-2
-------
Person Trip - The movement of a person from an origin to a destination, as opposed to
the vehicle trip associated with the same origin-to-destination movement. A carpool
carrying three people from origin-to-destination has made one vehicle trip, its occupants
together have made three person trips.
Soft Programs - "Soft Programs" represent employer support and incentive programs
such as site-specific employer programs. These strategies include on-site transit pass
sales, rideshare matching, and guaranteed ride home programs.
Trip End- The number of vehicle or person trips ending at a given origin or destination
location.
Trip Frequency - The number of trips per unit of time.
Utility - In transportation modeling, the value (positive or negative) of a particular
option, usually estimated as a function of the travel option's characteristics as well as
traveler or population characteristics.
Vehicle Miles of Travel (VMT) - The total usage of vehicles in miles traveling on the
transportation system for a given condition - metropolitan area, corridor, facility types
(freeways, arterials), or modes of transportation.
Vehicle Occupancy - For a specific group of travelers, e.g., morning peak workers, this
represents the ratio of person trips to vehicle trips. It is often used as a criterion in
judging the success of trip reduction programs.
Vehicle Trip - An origin-to-destination journey by a single vehicle, as opposed to a
person trip, the origin-to-destination j ourney of an occupant of the vehicle. A bus
carrying 40 people from an origin to a destination makes one vehicle trip, while its
occupants make a total of 40 person trips.
A-3
-------
Appendix B
Fuel Parameters for Default Emissions Factors
-------
tfatcn volume
^Itur
Aromatics
Uletms
uenzene
uxygen
ivietnanoi
IVllUJt
J^tnanoi
&IK&
lame
1 -uutanol
Kvp
ezuu
ejuu
ApiGravitj
1003 RFC Summer
PADD 1 WEIGHTED AVERAGE
PADD 2 WEIGHTED AVERAGE
PADD 3 WEIGHTED AVERAGE
4,635,357,866
1,546,419,377
5,648,131,887
97
116
131
21.16
19.88
18.24
12.81
3.73
10.91
0.61
0.82
0.55
2.12
3.54
2.28
0.01
0.00
0.01
10.77
0.00
9.56
0.00
9.42
1.56
0.02
0.00
0.00
0.98
0.00
0.77
0.03
0.00
0.03
6.80
6.85
6.85
47.02
45.36
49.09
83.96
85.13
84.29
57.8(
58. 6(
59.4?
1003 RFC Winter
PADD 1 WEIGHTED AVERAGE
PADD 2 WEIGHTED AVERAGE
PADD 3 WEIGHTED AVERAGE
7,856,210,556
819,908,912
7,500,187,347
128
210
200
20.51
16.43
18.71
12.57
7.37
9.50
0.67
0.82
0.62
1.92
2.00
2.31
0.01
0.00
0.01
8.95
0.00
8.20
0.73
5.24
2.68
0.00
0.00
0.00
0.17
0.00
0.54
0.02
0.00
0.03
10.48
13.07
11.29
55.49
60.66
55.98
86.00
86.52
84.08
61.7^
63. 6i
62. i:
1003 CG Summer
PADD 1 WEIGHTED AVERAGE
PADD 2 WEIGHTED AVERAGE
PADD 3 WEIGHTED AVERAGE
PADD 4 WEIGHTED AVERAGE
PADD 5 WEIGHTED AVERAGE
3,651,741,910
11,324,147,175
21,336,476,945
1,908,376,956
3,027,379,030
233
351
306
250
203
26.87
28.70
26.84
27.68
31.18
13.23
9.64
13.15
10.71
7.36
1.00
1.39
0.95
1.71
1.34
0.86
0.33
0.17
0.00
0.32
0.00
0.00
0.00
0.00
0.00
2.62
0.00
0.83
0.00
1.60
1.01
1.09
0.00
0.00
0.00
0.10
0.00
0.00
0.00
0.01
0.22
0.00
0.11
0.00
0.00
0.01
0.00
0.00
0.00
0.00
8.38
8.47
8.16
8.32
7.84
47.72
45.85
44.45
45.81
44.48
82.84
80.81
79.38
84.28
81.96
58. i;
58.0!
57.8:
59.2'
57.3^
1003 CG Winter
PADD 1 WEIGHTED AVERAGE
PADD 2 WEIGHTED AVERAGE
PADD 3 WEIGHTED AVERAGE
PADD 4 WEIGHTED AVERAGE
PADD 5 WEIGHTED AVERAGE
4,920,137,802
12,848,436,621
23,415,419,586
1,975,690,694
3,653,470,712
220
267
269
202
140
23.94
25.14
24.54
24.40
27.99
14.67
9.08
12.23
10.76
8.09
0.90
1.27
0.92
1.53
1.51
0.40
0.43
0.14
0.02
0.07
0.00
0.00
0.00
0.00
0.00
0.84
0.00
0.70
0.00
0.06
0.81
1.30
0.01
0.05
0.22
0.01
0.00
0.01
0.00
0.01
0.24
0.00
0.05
0.00
0.00
0.00
0.00
0.00
0.00
0.00
11.05
13.07
11.72
11.68
10.61
51.40
51.50
49.31
51.06
48.56
84.24
83.62
81.47
86.39
84.60
61.6
ei.s;
61. o;
62. 5(
60.5?
These numbers include refineries, terminals, imports, etc.
Some refineries were excluded from the averaging (shown in italics), based on their fuel properties implying that their batches were not substantially 'normal' gasoline.
The volumes shown here reflect the total volume included in the averaging.
The averaging algorithm omits from the average for each fuel property any refineries who have a blank value for that fuel property (i.e. blanks are not included as zeros).
B-l
------- |