Assessing Highway Tolling and Pricing Options and Impacts: Volume 2: Travel Demand Forecasting Tools (2012)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

11 2.1 Basics of T&R Forecasting A travel demand model predicts travel flows between origins and destination by time of day, mode, and route within each mode. In addition, these models produce the necessary infor- mation on toll facility patronage, as well as tolling and pricing impacts on all trips in the region. A travel demand model rep- resents a sequence of calculations structured by meaningful travel dimensions. There are two major approaches for struc- turing a demand model: the traditional trip-based (frequently referred to as 4-step) and an advanced tour-based (frequently referred to as activity-based) with numerous technical varia- tions. Traditional 4-step models were the foremost modeling technique used during the 1980s and 1990s for most MPOs in United States, and they constitute a majority of travel mod- els in practice even today. In recent years, however, advanced ABMs have been applied in practice, currently constituting the majority of newly-developed models for large MPOs (more than 10 such models in practice). The typical model structure and relevant travel dimensions modeled in each type of model are shown in Figure 2. A demand model represents a computerized travel simu- lation system where demand generation is integrated with network simulation, and equilibrium travel times and costs are sought. This equilibrium feature, which is technically implemented by means of feedbacks of travel time and cost to the demand generation stages, is essential for pricing studies. Pricing affects travel demand by shifting the equilibrium point to a solution where social/economic welfare is greater than without pricing. This is analytically consistent with the policy objectives of pricing, where pricing is often intended to affect travel choices in such a way that network capacity is utilized in a more optimal way. The following model features are specifically important for pricing projects: • Network simulation and associated route choice sensitiv- ity to tolls. Changes in route choice represent first-order response to pricing, and associated trade-offs between travel time savings and tolls are the cornerstone of toll facil- ity traffic forecasts. • Other first-order responses to pricing include mode choice and time-of-day choice. If pricing is applied and there is a reasonable transit alternative for the given trip, travelers may consider switching to the transit mode. In addition, travelers may consider switching to other time-of-day peri- ods to avoid paying tolls. The first order responses are char- acterized by the highest elasticity of substitution from an economic perspective and must be included in the model. ABMs have a more detailed structure of choices. In par- ticular, mode choice is modeled first for the entire tour and then for each trip on the tour, conditional upon the chosen entire-tour mode. Pricing impact on mode and time-of-day choice is modeled through the corresponding feedbacks. • There is a set of additional pricing impacts that can affect almost any travel dimension. For example, as a response to pricing, travelers may choose another destination for a trip or not implement the trip at all, substituting some other activity or linking the trip to another tour as a stop, etc. These impacts are generally considered second-order effects and are associated with a generally lower elasticity to pric- ing, although the accumulated effects over a longer period of time can still be significant and even affect residential location choices and land-use development (components of the land-use model). The second-order effects are mod- eled through additional feedbacks where the period-specific mode-choice Logsums are used to inform all upper- level models about the tolls and travel time savings for all affected modes. There are several reasons why pric- ing impacts on upper-level choices can be better modeled through mode choice Logsums rather than through direct feedback of auto travel times and tolls. First, mode choice Logsums combine all travel time and cost components in a single and theoretically sound measure of accessibility. Sec- ondly, highway pricing and associated congestion relief may C h a p t e r 2 State of the Practice in Forecasting Methods

12 affect transit modes as well, such as buses in mixed traffic or using HOV lanes as well as park and ride options. 2.2 Travel Cost Representation in Demand and Network Models Highway pricings should first be properly incorporated in network assignments and skimming procedures through gen- eralized cost functions. Then, through generated travel time and cost skims, pricing will affect all other choice dimensions, specifically mode choice, time-of-day choice, trip/tour dis- tribution, and other upper level choices. In highway assign- ments, the generalized cost function is defined for each net- work link and further calculated for each origin-destination pair. It can be written in the following general way: G a T b Ck k k k k= × + × ( )Equation 1 where: k = vehicle types and auto occupancy classes, Tk = travel time, Ck = travel cost (only toll is normally included for assign- ment purposes), ak = coefficient for travel time, bk = coefficient for travel cost, and ak/bk = Value of Time (VOT). For highway tolling and pricing projects it is essential to sepa- rate vehicle types like private auto, light truck, heavy truck, taxi, etc., and auto occupancy classes like SOV, HOV/2, HOV/3, etc., in the traffic assignment for the following reasons: • Different vehicle types and occupancy classes may have very different VOTs. In this respect, some additional segmenta- tion by VOT (based on trip purpose and/or income group) is also recommended and will be discussed further below. • Toll rates might be differentiated by vehicle types and/or occupancy classes. A good example of this is an HOT-3 lane where vehicles with three or more passengers do not pay the toll, vehicles with two passengers pay half of the toll, and SOVs pay a full toll. • General prohibitions and eligibility rules can be applied for certain vehicle types on certain facilities [for example, trucks prohibited on expressways or truck-only toll (TOT) lanes] or auto occupancy classes (for example, HOT lanes). In order to satisfy all these conditions, traffic assignment should be implemented as a multi-class procedure (avail- able in all major transportation software packages) with 6–12 or even more trip tables depending on the model structure. While this is a certain complication, it is essential for proper modeling of all related choices. If different vehicle types and auto occupancy classes are mixed together (with some aver- age VOT), it is not only a source of bias in route choice, but since assignment procedures serve as the source for skimming LOS variables used in mode choice, time-of-day choice, and all other choices, the distortions in route choice will affect all these models as well. Tr ip -b as ed To ur -b as ed Tr ip ge ne ra ti on Tr ip di st ri bu ti on Tr ip mode ch oi ce Tr ip ti me of da y A ssi gnme nt (r ou te ch oi ce ) A ssi gnme nt ( rou te ch oi ce ) Da ily ac ti vi ty pa tte rn (t our ge ne ra ti on ) To ur pr im ar y de st in at io n To ur ti me of da y To ur mo de ch oi ce St op fr eq ue nc y St op lo ca ti on Tr ip mo de ch oi ce 1 s t o r d e r im pa ct s o f p r ic in g 1 s t o r d e r im pa ct s o f p r ic in g 2 n d o r d e r im pa ct s o f p r ic in g 2 n d o r d e r im pa ct s o f p r ic in g Figure 2. Typical structure of demand model.

13 Equation 1 corresponds to the general expression of high- way utility in its most common form. This expression consti- tutes a key component in all travel choice models. In the con- text of traffic assignment when choice is modeled between alternative routes, the travel time coefficient is normally set to 1.0. This arbitrary setting does not affect All-or-Nothing choice applied in the conventional Static User Equilibrium assignment (however, more advanced stochastic route choice models would be sensitive to this setting and should be cali- brated in a special way). With this simplification, the highway generalized cost function can be written in the following way: G T b C T VOT Ck k k k k k k= + × = + × 1 ( )Equation 2 While All-or-Nothing route choice embedded in the con- ventional assignment procedure is frequently applied in prac- tice to distinguish between free and tolled routes, it has been recognized that this is not an adequate tool in itself, since the traveler choice route is not a simple linear combination of time and cost. In particular, toll roads (or managed lanes) can represent a more attractive option because of enhanced reliability and other considerations that are not directly mea- sured by average time and cost. An explicit inclusion of travel time reliability in the highway generalized cost function rep- resents a technical challenge that will be discussed. A simpler, but still useful, approach that has been applied in many models in practice is to estimate an additional bias con- stant associated with priced facilities. This bias can be most effectively incorporated in a binary choice model frequently referred to as pre-route choice that is placed between mode choice and route choice. Technically, it can be included as the lower-level sub-nest in the mode choice nested structure. In addition, such models allow for probabilistic choice between free and toll options, helping to avoid the “lumpiness” of the All-or-Nothing assignment that yields unstable routes. With this enhancement, the highway generalized cost function can be written in the following way: G a T if C a T b C if C k k k free k k k k toll k k k = × = + × + × > , , 0 γ 0    ( )Equation 3 where gk represents the toll bias. Since the difference between utilities is all that matters in this choice framework, the expressions in Equation 3 can be rewritten in equivalent terms of relative travel time savings, where the free route generalized cost is set to zero as the refer- ence point: G if C a T T b C if C k k k k k toll k free k k k = = + × −( ) + × > 0 0 γ , 0   ( )Equation 4 Equation 4 constitutes the essence of many models applied for T&R forecasting in practice. It also has many possible technical modifications. One such modification (which was adopted for many pricing studies in Texas and Colorado) represents a non-linear transformation of the following form (WSA 2001; CSI 2005; Vollmer 2001): G if C a T T b C k k k k k free k toll k k = = + × + −( ) + × ( ) 0 0 1γ ln 2 0, ( ) if Ck >   Equation 5 The model form in Equation 5, however, still only corre- sponds to route choice, and should be further generalized to include other relevant choice dimensions (possible traveler responses) like mode choice, time-of-day choice, destina- tion choice, and others. The corresponding generalization to incorporate mode choice is done by the inclusion of the generalized highway cost, as part of the mode choice utility for highway modes in the following form: U a T b C Sm p m p m p m m p m vm p v v = + × + × + ∑γ λ ( )Equation 6 where: m = set of modes including auto occupancy classes, p = travel purpose and other possible segments, v = person, household, and zonal variables, Tm = travel time by mode, Cm = travel cost by mode, Sv = values of the person, household, and zonal variables, gpm = mode-specific constant for each purpose/segment, apm = coefficient for travel time by mode and purpose/ segment, bpm = coefficient for travel cost by mode and purpose/ segment, apm/b p m = VOT, and lpvm = coefficients for person, household, and zonal vari- ables for each mode by purpose. The most frequently used person, household, and zonal variables in 4-step models include income, car ownership, and urban density. In research works and ABM, the set of explanatory variables and possible dimensions for segmenta- tion have been significantly extended and will be discussed in more detail in Chapter 5. Travel time and cost variables in themselves include many components. In particular, for auto modes, travel time can include parking search and parking time as well as additional time for picking-up and dropping- off passengers (for HOV) while travel cost can include toll, parking cost, and vehicle operating cost (fuel and some frac- tion of maintenance cost that depends on the mileage). Examples of mode choice models incorporating pre-route choice developed for T&R studies in Montreal and San Francisco

14 are shown in Figure 3 and Figure 4. These models are dis- cussed in detail in Appendix A (Section A.1.2). Mode utility functions that include travel time savings and additional cost associated with highway pricing (Equa- tion 6) represent the basis for the most theoretically consis- tent formation of the impedance functions used for destina- tion choice (trip distribution) and/or time-of-day choice by using Logsums of the lower-level choices as components of the utility functions in the upper-level choices. However, in addition to using mode choice log-sums, there is a simplified option available (and frequently used in practice) to employ the highway generalized cost itself (Equation 1) in the utility function of destination choice or time-of-day choice. This simplified option, however, is not recommended unless tran- sit shares are very low. The details of these models depend on how the destination choice, mode choice, and time-of-day choice are sequenced in either the 4-step or ABM. We will illustrate the basic prin- ciples following the typical model structures shown in Fig- ure 2. The time-of-day choice utility can be formed using mode choice Logsums in the following way: V U St p mt p m vt p v v = × ( )   +∑ ∑µ λln exp (Equation 7) where: t = time of day periods (TOD), 0 < m ≤ 1 = scaling coefficient that should be in the unit interval, and lpvt = coefficients for person, household, and zonal variables for each TOD. In aggregate 4-step model systems, TOD choice models normally operate with broad 3- or 4-hour peak periods, and longer off-peak periods. This might require additional peak spreading or peak-hour factoring sub-model. In disaggre- gate AB model systems, TOD choice models operate with a finer temporal resolution of 1 hour or even less (Vovsha and Bradley 2005). In addition to mode choice Logsums, such person, household, and zonal variables as income and density (especially at the destination end) prove to be significant. The TOD choice utility is sensitive to tolls and associated travel time savings through the mode choice utilities included in the Logsum calculation. The destination choice utility (or trip distribution imped- ance functions) can be formed using a Logsum over all TOD periods. While it is possible to calculate this Logsum, which Figure 3. Montreal mode choice model—nested structure incorporating free vs. toll route choice. Figure 4. San-Francisco mode choice model—nested structure incorporating free vs. toll route choice.

15 t would represent the most consistent impedance measure, it is computationally very intensive to do so since it should be implemented for each origin-destination pair. The usual practical approach taken with a 4-step model (as well as adopted for some ABMs) is to use representative TOD peri- ods for each travel purpose to economize on calculations. For example, for trips/tours to work, a combination of AM period (for the journey to work) and PM period (for the jour- ney home) is normally applied; while for non-work trips, the midday (off-peak) period is assumed. Multiple representative periods can be applied or a weighted linear combination of LOS variables between several periods can be used if necessary. In any case, the destination choice utility (and impedance function as part of it) can be generalized in the following way: W U Aod p od m t p p m d p = × ( )   + ( )( )∑η ln exp ln (, , Equation 8) where: o, d = origin and destination zones, 0 < h ≤ 1 = scaling coefficient that should be in the unit interval, t(p) = representative TOD period for each purpose, and Apd = destination zone attraction (size variable) for each purpose. The size variables represent destination zone attractions for each purpose. The most frequently used attraction size variables are total employment for work purpose, enroll- ment for school purpose, and retail employment for non- work purposes. Many ABMs include more complicated size variables that mix several employment and population vari- ables and can be segmented by urban type and density. Size variables are not added to the impedance function in doubly- constrained gravity models of trip distribution since they are applied directly as constraints on the destination side. The destination choice utility is sensitive to tolls and associated travel time savings through mode choice utilities included in the Logsum calculation. By using the destination choice utilities sensitive to high- way pricing and travel time savings, zonal accessibility indi- ces can be calculated and used as an explanatory variable for trip generation, activity pattern, car ownership, and land-use development models. Accessibility indices essentially represent mode/destina- tion choice Logsums calculated by trip purpose in the fol- lowing way: Z Wo p d od p = ( )  ∑ln exp ( )Equation 9 If Equation 9 is directly applied in combination with Equa- tion 8 it may result in very intensive calculations. For this reason, in most model systems, the destination choice utili- ties used in accessibility calculations are simplified in such a way that they can be pre-calculated based on a limited num- ber of origin-destination skims and for a limited number of modes, travel purposes, and population segments. Even with these simplifications, accessibility measures represent useful explanatory variables sensitive to highway pricing and travel time savings. Further extensions of the formulas for highway utilities that include travel time reliability measures are discussed in Appendix A (Section A.3). 2.3 Models Included in the Synthesis For this research, documentation was obtained and ana- lyzed in detail. Table 2 shows the list of transportation mod- els and their applications for highway pricing studies. This review has revealed that there is a great variety of travel models and analytic approaches currently applied in practice by different agencies. In order to constructively ana- lyze and synthesize them, we have developed a template that includes their most important features. Each model has been analyzed in this format based on the available model docu- mentation. This approach makes it possible to meaningfully compare different models, and also helps to identify their commonalities, as well as gaps in particular model structures. The following main model features were included in the tem- plate (Table 3). The details of the selected transportation models applied for pricing studies in the template format are presented in Appendix A (Section A.1). The models are grouped into the following two major classes: 1 = trip-based 4-step models, and 2 = tour-based ABMs. 2.4 Conclusions from the Review of Existing Models The most important findings and conclusions are summa- rized below: • There is a great deal of variation in approaches. In most cases, the model applied for the highway pricing project was essentially a modification of the existing regional model available for the study. Thus, limitations and deficiencies of the existing regional model were inevitably adopted for the study. There was not a single practical case uncovered yet of a regional model specially designed and developed for highway pricing studies or at least having these specific requirements in mind.

16 • In most cases, only route itinerary (assignment) and binary route type choice (toll versus non-toll) models were employed for comparison and evaluation of pricing alternatives. This achieves reasonable results under the assumption that pricing would not affect mode choice, time-of-day choice, trip distribution, and trip generation. While this simplification might be somewhat acceptable for intercity highways, it is more difficult to defend for most of the metropolitan/urban facilities. • Pricing effects on trip distribution have been incorpo- rated by using mode choice Logsums as the measure of accessibility in destination choice or gravity-type distribu- tion models. The use of mode choice Logsums in doubly- constrained gravity models needs to be tested extensively; unlike destination choice frameworks, where appropriate elasticities with respect to cost are expected when reason- able Logsum parameters are used, it is not clear that gravity models behave appropriately to changes in LOS variables, such as the introduction of tolls. • In some cases there is an inconsistency between the travel times and costs used for the trip distribution and mode choice models, in that the travel times reflect priced condi- tions, while the toll cost itself does not enter the impedance function. This is the case when travel times are fed back from a generalized cost assignment into a distribution model that is a function of travel times only. • There are a growing number of applications where mode and/or occupancy choices were included. In several cases, mode, occupancy, and binary pre-route choices were com- bined in one multi-level nested logit choice model structure. • In a few cases utility functions in multinomial or nested logit mode choice models are miss-specified. Undesirable City / Area Agency developed the model Pricing study Corridor and Sample-Enumeration models: New York, NY PANYNJ Congestion Management for New York – New Jersey Crossings Minneapolis-St. Paul, MN MNDOT, FHWA’s STEAM model Pricing Study Washington, D.C. FHWA’s SMITE-ML model Value Pricing Study of the Capital Beltway in Northern Virginia Regional Trip-Based 4-Step models: Alameda County, CA MTC I-580/I-680 Corridor Value Pricing Study Atlanta, GA ARC Managed, HOV, and Truck Toll Lanes Study Austin, TX CTRMA Central Texas Turnpike Project Dallas – FW, TX NCTCOG Regional Value Pricing Corridor Evaluation & Feasibility Study Denver, CO DRCOG Northwest Parkway Traffic & Revenue Study Denver, CO DRCOG I-25/SR-36 Value Express Lane Feasibility Study Colorado Tolling Enterprise DRCOG, etc Preliminary Statewide Traffic & Revenue Study Houston – Galveston, TX HGAC Road Pricing Study (QuickRide System), I-10 Katy MIS Montgomery County, MD Road Pricing Study Oakland, Bay Area (MTC) MTC HOT Lanes Study Orange County, CA OCTA SR-91 Value-Priced Express Lanes Orlando / Tampa Bay, FL FDOT Turnpike Enterprise San Diego, CA SANDAG I-15 FasTrak and SR-125 South Tollway Salt Lake City, UT UDOT Mountain View Corridor Pricing Study Sonoma County, CA SCTA US-101 Variable Pricing HOV/HOT Lane Study Phoenix, AZ MAG Managed Lanes Study Pittsburgh, PA PENNDOT HOV & HOT Lanes Study Sacramento, CA SACOG Managed Lanes Study Twin Cities, St. Paul, MN MC I-394 HOT Lanes Washington, D.C. MWCOG Managed Lanes Study, HOT Lane in Northern Virginia Seattle, WA WSDOT/PSRC SR-520 Toll/HOV Feasibility Study Toronto, ON MTO Highway 407 Traffic & Revenue Study Regional Activity-Based Tour-Based models: Montreal, QC MTQ A-25 and A-30 Traffic & Revenue Study New York, NY NYMTC Manhattan Area Pricing Study San Francisco, CA SFCTA Congestion Pricing Feasibility Study Portland, OR METRO Traffic Relief Options Study Table 2. Forecasting models applied for highway pricing projects.

17 Major model feature Detailed feature / sub-model Possible characteristics Spatial scale Regional Corridor / sub-area Facility Coverage of time periods Regular weekday AM peak & shoulders PM peak & shoulders Midday Other off-peak (night, early) Daily traffic Annualization factors Weekend traffic (assumptions) Holidays Seasonal variation Demand model structure Sample enumeration Aggregate trip-based 4-step Microsimualtion activity-based Network simulation tool Static user equilibrium assignment Dynamic traffic assignment / microsimulation Representation of priced highway facilities Link tolls & toll equivalents in generalized cost / time functions Toll plazas / payment types / delays Modeled pricing impacts (traveler responses), sub- model structure, form of utility function, incorporation of pricing Trip-level decisions Route itinerary in highway network Principal decision to take a toll route vs. non-toll route (pre-route choice) Tour & trip-level decisions Auto occupancy Mode choice Time-of-day choice (including peak spreading effects) Destination choice (spatial distribution) Day-level decisions Trip/tour/activity frequency Activity re-sequencing as result of time-of-day shifts Mid-term mobility decisions and household / person attributes Transponder acquisition Transit pass acquisition Long-term parking arrangement Free parking eligibility at workplace/school Household car ownership Long-term location choices Residential location and dwelling type Usual workplace location Firm / businesses location Willingness to pay / VOT and user segmentation Vehicle classes (in the demand model and network simulation) Auto Commercial vehicle / light truck Heavy truck Taxi Vehicle occupancy categories (in the demand model and network simulation) SOV HOV/2 HOV/3+ Trip purpose segmentation (in the demand model) Work Work/business-related School University Shopping Escorting children Other household maintenance Discretionary / leisure / sport Trip to airport / rail station / port for long-distance travel Intercity business travel Intercity non-business travel Table 3. Template for transportation model analysis. (continued on next page)

18 specifications include toll utilities that are a function of both the toll alternative travel time and travel time sav- ings with respect to the free alternative. This type of speci- fication may result in counter-intuitive results when the LOS attributes on either the toll or the free routes change. Another potentially problematic specification is the use of thresholds, such as making the toll alternative available only if it meets a pre-defined minimum time savings goal. The nesting coefficients on these models sometimes result in models with unreasonably high elasticities to toll or time differences when the toll diversion is examined at the root level of the model (where they are comparable with the elasticity of route type binary choice models). • There is no consensus on whether road pricing costs should be shared among vehicle occupants, and if so how. Most models either assume that the full toll cost is either borne by all occupants, or that it is equally shared among the occupants. Some models differentiate between cost sharing for HBW trips and cost sharing for other pur- poses. Sharing road pricing costs among vehicle occupants makes carpools less cost-sensitive, an assumption that may be acceptable for work trips, but is questionable for other purposes, where the majority of carpools are among members of the same household and often times include minors. • In some models willingness-to-pay differences between cash-payment users and ETC users are explicitly made (by specifying different values of time or different toll con- stants). Other models simply use the average toll cost per transaction. • In some regional model systems that were specifically modified for congestion pricing projects, peak-spreading models were applied. Conventional 4-step models are normally based on time-of-day (peak) factors that are not sensitive to relative congestion levels at different periods of the day. Thus, 4-step models require a post-model peak- spreading sub-model that is difficult to incorporate in the overall equilibrium framework. Activity-based tour-based models can offer a better framework where peak-spreading effects are captured by time-of-day choice sub-model. • Peak-spreading or time-of-day models are sensitive to dif- ferences in travel times by time of day, but not to differ- ences in toll costs by time of day. This may be simply a result of the limited number of localities where road pric- ing costs vary by time of day combined with observed data insufficient to estimate appropriate model parameters. • Very few models to date have incorporated all trip and tour-level dimensions in a consistent way, and there have not yet been any practical examples of the incorporation of pricing impacts on the day-level, mid-term, and long-term choices, even with the activity-based models now that have recently come into use. • Almost all models, including activity-based tour-based models are characterized by a significant discrepancy between the user segmentation by (VOT) in the demand model compared to network simulation. While at the demand modeling stage, segmentation normally includes several trip purposes, income groups, car occupancy, and time-of-day periods; network simulations are character- ized by a limited segmentation. Traffic assignments are Major model feature Detailed feature / sub-model Possible characteristics Time of day (in the demand model and network simulation) AM period PM period Midday period Night period Household / person characteristics (in the demand model and network simulation) Household income group Person work status Gender Demand-network equilibrium Application flowchart with feedbacks Feedback implementation Number of iterations Averaging rules Convergence criteria / statistics Surveys and other data sources for model estimation / calibration / validation Household travel survey Size / sample, year, structure / questionnaire Survey of existing toll road users Size / sample, year, structure / questionnaire Stated Preference survey Size / sample, year, structure / questionnaire Model validation for the base year General validation targets / reported measures of fit AADT Traffic counts by time of day / vehicle type Project-specific calibration Traffic counts Travel time / speed data Table 3. (Continued).

19 implemented by periods of the day and for multiple vehicle classes that typically include vehicle type and occupancy. However, trip purposes and income groups are blended together before assignment, creating strong aggregation biases with respect to VOT. • There are also discrepancies in the cost functions used to build best paths between the network simulations used to build travel time and cost matrices for the demand mod- els, and the network simulations used to assign trips to the highway network. Best paths for the demand model may be built on the basis of travel time only, while the assignment is performed on the basis of generalized cost, or vice-versa. • In almost all modeling efforts where pre-route choice (toll versus non-toll) was involved, a problem of inconsistency between the generated trip tables for toll-users and their assignment onto the highway network was reported. This leakage of toll users in the network simulation can be sig- nificant and constitutes a non-trivial analytical problem that requires special modeling efforts to resolve. • Most models attempt to equilibrate supply and demand by feeding back travel times and cost from the assignment step to the trip distribution or mode choice steps. In most cases feedback is executed for a fixed number of iterations, so convergence is not necessarily guaranteed. This may be problematic when forecasting under conditions of high population growth, where congestion effects may be far more pronounced than for the calibration year. • Most models break down the network simulation into four broad time periods, typically AM Peak (2 to 4 hours long), Midday, PM Peak (2 to 4 hours long) and Night, and are therefore able to compute LOS differences by time of day only at this level of aggregation. Only one of the regional models performs the network simulation at a finer time of day disaggregation. • With few exceptions, network simulations are validated to 24 hour traffic volumes, even when highway assignments by time periods are available. It is not clear that the models are adequately reproducing LOS attributes and demand for the different times of the day used in the simulation.