6
Data and Analytical Tools for
Interregional Transportation
Planning and Decision Making
Countering the tendency to plan and program public investments in transportation on a mode-by-mode basis is a challenge. Mode-based funding exacerbates this challenge. Coordination in the interregional context is complicated further because of the large number of modes involved and because of the many federal, state, and local governments with infrastructure and operating responsibilities. Even though many policy goals—such as providing efficient service, relieving congestion, and protecting the environment—may be shared among these public entities, there may be few means of furthering them through actions coordinated at the corridor or interregional level.
The importance of a rational system of transportation planning and decision making has long been recognized in the urban context. Analysts have summarized the steps of urban transportation programming as consisting of (a) inventorying existing travel and activity patterns, (b) developing models of local transportation supply and demand relationships, (c) formulating options, (d) forecasting the effects of each option on travel, (e) evaluating each option on the basis of economic and other criteria, and (f) implementation (Button 2010, 397). Carrying out these steps requires an institutional arrangement for transportation planning and decision making under which consensus can be reached on goals and options and objective analyses can be undertaken. Although metropolitan planning organizations (MPOs) had existed in many urban regions for decades,1 their requirement as a condition for federal highway and transit aid expanded their use. They helped overcome inherent obstacles to
________________
1 For example, the New York Regional Planning Association was founded in 1922.
the following of these steps in metropolitan regions that can consist of numerous municipalities, counties, and transportation authorities.
The increase in the influence of MPOs has stimulated demand for improved travel data and analytical tools applicable to urban transportation.2 In addition to requiring the MPO process, the federal government supported the development of databases and tools.3 Thus, in the urban setting there is now an extensive base of research and practitioner guidance on the factors influencing travel behavior and demand, modeling and forecasting techniques, surveying and sampling methods, and project appraisal techniques such as benefit–cost analysis. The development of urban transportation planning tools has helped improve the analyses in support of decisions (Cambridge Systematics et al. 2012, 1). There are no similar coordinating mechanisms for transportation planning and decision making when interregional corridors cross multiple states; hence, there is less institutional demand for travel data, models, and evaluation techniques. In addition, in contrast to the metropolitan level, where public entities provide much of the transportation infrastructure and services, there is a substantial private-sector role in the interregional domain. Data collected and models developed by the private sector may be kept proprietary (Miller 2004).
In short, the data and analytical tools used for planning, evaluating, and developing transportation projects in the interregional domain are seldom derived from an ongoing transportation planning and priority-setting process. Analysis and evaluation are typically undertaken on an ad hoc basis; for example, a specific modal option for a corridor, such as the building of a high-speed rail line, might be assessed (Miller 2004). In this regard, the process is the antithesis of the multimodal, multioption, system-level approach used at the metropolitan level (Horowitz 2006). Indeed, the MPO approach has become more comprehensive over the years as urban transportation planners have assumed more project evaluation and programming responsibilities. In addition to planning highway and transit capacity expansions,
________________
2 Statewide and metropolitan transportation planning processes are governed by federal law (23 USC §§134–135). MPOs were first required by the Federal-Aid Highway Act of 1962 and became much more influential in metropolitan transportation planning and decision making after enactment of the 1991 Intermodal Surface Transportation Efficiency Act.
3 One example is the 20-year program of federal support for TRANSIM (Transportation Analysis Simulation System), which is an integrated set of tools developed to conduct regional transportation system analyses.
they now develop and assess options for managing travel demand and for achieving policy goals such as curbing greenhouse gas emissions; meeting air quality standards; and expanding community access to health care, jobs, education, and affordable housing.
In the remainder of this chapter, some of the analytical and data capabilities needed to inform transportation planning and decision making in the interregional context are described. As in the case of the MPOs, institutional mechanisms are necessary to create a demand for and to ensure regular use of these capabilities. Further consideration is given to that necessity in Chapter 7, which contains the study recommendations.
Travel Demand Forecasting Models
There are three broad types of travel demand models. Aggregate models predict the number of trips taken in a geographic area on the basis of trip production and attraction factors. Disaggregate models estimate the probability with which a utility-maximizing individual (having certain quantifiable characteristics) will undertake a trip between a specific origin and destination by using a specific mode and a specific route. In highly disaggregate, activity-based models, travel behavior is analyzed in the broader context of participation in activities, often through the use of travel diaries.
Travel demand forecasting models used in the urban setting are typically hybrids of the three types. Most urban travel demand forecasts start with an aggregate model to estimate the total number of trips originating and ending in defined geographic zones or areas.4 Model inputs (or variables) used to explain trip production include household size, automobile ownership, and income. Trip attractions are chiefly workplaces and retail outlets but may include other household concentrations, schools, parks, airports, and education or recreational destinations. The models typically require data from a number of sources such as household surveys;
________________
4 This step refers to estimation of trips on the basis of “productions” (households are the most important source of production) and “attractions” (places of employment or retail establishments are obvious attractors).
census statistics; and bus, airline, and rail passenger counts and ticket sales. Additional models (which can be either aggregate or disaggregate) are applied to divide the total trips among specific origin–destination pairs (e.g., trip distribution or destination choice models),5 by mode (e.g., logit mode choice models), and by route (e.g., network assignment models). For example, the probability of choosing among modes and routes may be modeled as a function of the characteristics of individuals, trip purposes, and the relative costs of alternative modes. Activity-based models may be used to predict travel behavior from the perspective of the individual or households so that travel induced by a new mode or service or by other changes in the transportation system can be evaluated (see Pinjari and Bhat 2010).
Disaggregate and activity-based models are generally favored for more sophisticated evaluations of policy alternatives in more complex situations because they can better address traveler-specific trade-offs among alternatives having different characteristics with respect to time, price, and level of service. For example, when a metropolitan area with severe congestion is not meeting air quality standards, urban planners interested in assessing specific policy options such as variable tolling or bus rapid transit may desire a travel forecasting process that is sensitive to price and allows for the analysis of mode choice by time of day (Cambridge Systematics et al. 2012, 1–4). Such models require detailed information from representative samples of households to obtain statistically valid information on activities and preferences.
As in the metropolitan setting, the kinds of forecasting models that are most appropriate to the planning of interregional transportation are likely to depend on specific circumstances and the availability of travel data. Heavily traveled and modally diverse corridors such as the Northeast Corridor, where many alternatives involving complex conditions need to be assessed, may have characteristics that favor disaggregate models. Miller (2004, 94) suspects that earlier interregional demand
________________
5 The standard approach to estimating trip distribution is a gravity model, so named because the gravitational force between two bodies increases with the mass of each body and decreases with the distance between them. Trips are negatively affected by some measure of “impedance” or friction affecting the desirability of a trip between the two points, such as distance, travel time, cost of travel, or a combination of such factors.
forecasts have been developed by using aggregate models because they were constructed for a specific study with limited data, usually by proponents of a new investment in corridor infrastructure. The goal of these models is often to predict the number of trips in the corridor and the changes in mode shares in response to the investment rather than to assess a more varied and complex set of investment and policy alternatives.
As noted above, aggregate models are usually limited in their ability to portray or analyze the decisions of individuals, the effects of travel time valuations, and the utility of modal service attributes such as reliability. Miller (2004) is thus skeptical of the ability of aggregate models to account for a number of important demand factors, such as the effect of access and egress time and cost on airline and passenger rail mode shares (especially in relatively short interregional markets), and for details such as toll and fuel price levels, airline pricing strategies, and delays at airports due to security. Miller acknowledges that disaggregate models can present their own challenges. For example, data on relatively infrequent long-distance trips can be difficult to obtain, at least in comparison with data on the many regular trips made by households that inform urban transportation planning. The accurate portrayal of transportation system choice sets (i.e., level of service attributes, frequencies, prices) is also problematic for an entirely new system. These and other data-related issues are discussed below as they pertain to interregional travel.
In contrast to the metropolitan context, no standard model structures or parameter values apply to interregional corridors. Default, or rule-of-thumb, coefficients compiled from empirical data are transferable among urban transportation settings when locally specific data are not available for model development.6 Similar standardization could prove valuable for modeling interregional travel demand, especially where regular data collection and analysis activities are not practical. Miller (2004) believes that the lack of entities responsible for ongoing evaluations of interregional transportation leads to a dependence on consultants whose
________________
6 Default coefficients are encouraged by the Federal Transit Administration in the New Starts process. Various reports (e.g., Martin and McGuckin 1998, Cambridge Systematics et al. 2012, Schiffer 2012) that compile typical ranges of coefficients and parameters are available to assist modelers, but practitioners are cautioned in the use of these values in the absence of good local validation data.
model specifications are often proprietary and not subject to continued review and improvement by users. In studying ridership estimates for the planned California High-Speed Rail System, the Government Accountability Office (GAO) concluded that “there are no industry standard or established criteria for developing or evaluating intercity passenger high-speed rail ridership forecasts” (GAO 2013, 23). GAO criticized the Federal Railroad Administration for not having established guidelines on acceptable approaches for developing reliable system ridership and revenue forecasts (GAO 2009).
In the United States, the modeling of interregional travel may be most advanced in states that enclose major interregional corridors, such as California. The most recent version (Version 2.0) of the California Statewide Travel Demand Model, released in 2014, integrates short- (under 100-mile) and long-distance (over 100-mile) personal travel models on the basis of more than 5,400 zones within the state and more than 50 external zones. The model’s long-distance component uses data from the 2012 California Household Travel Survey to consider five travel modes and five time periods of the day. It models long-distance travel choices in terms of whether to engage in such travel, trip purpose, party size, duration, destination, main mode, access mode, and egress mode. The model is sophisticated and includes feedback loops to take into account the impacts of traffic congestion on these choices (as well as short-distance travel choices). A consequence of this detail is that the California model is computationally intensive. Five iterations are typically required to reach equilibrium, and run times are several days.
To support state highway and transportation planning, the Federal Highway Administration (FHWA) has undertaken a comprehensive review of long-distance transportation demand models, including those used in 10 states and several from abroad (FHWA 2012). The report contained information on dozens of statewide models and found that many European countries have national transportation models that focus on interregional travel. FHWA found that the state models used many approaches, but most used aggregate models because of the ease of implementing them across the state and to aid in the comparison of results with those of models used by MPOs. The report indicates that most of the national models used by other countries for forecasting interregional travel are hybrids.
They contain logit choice submodels that describe mode choice and usually other choices such as access and egress modes, routings, and timing choices. FHWA described its review of the models as providing “foundational knowledge to support long-distance modeling.”
Evaluation and Representation of Uncertainty
All forecasting models are subject to sources of error and uncertainty. Errors are introduced during the data collection process as a result of sampling and survey methods. In addition, future conditions must be specified for a long time frame. For example, most travel forecasting models use the results of other forecasts, such as projections of population, households, and employment. Because these projections are usually developed independently, they can have variances and introduce uncertainties that go unrecognized by the travel demand forecaster. Miller (2004) points out that in the case of interregional markets, modelers may face the additional challenge of obtaining independent forecasts of demographic and economic conditions because, unlike individual states or metropolitan areas, the interregional market may not correspond to available forecasting sources.
All forecasts have uncertainties that need to be evaluated and recognized for informed decision making. Single-point forecasts are seldom realistic, since all of the inputs on which the forecast is based are unlikely to occur as projected. Many quantitative methodologies are available for associating a probable variance with each input factor (e.g., probability distributions for future conditions and modal parameters) and producing an expected error range for the final forecasts.7 Presenting model results with probability distributions, or an estimate of error, allows users to derive a point estimate (e.g., midpoint of the confidence interval) or to use a range defined by the confidence limits. In either case, the reliability of the model outputs will be clearer to users. Evaluation of the model structure and its estimates by independent peer reviewers may also be warranted, along with the development of reference cases based on the history of forecast outcomes for similar projects.
________________
7Adler et al. (2014) provide a more detailed discussion of these methods.
The documented tendency for systematic biases in transportation forecasts is another consideration. In researching the accuracy of travel forecasts, analysts have found that many of them used to promote transportation projects (i.e., procure approval and funding) have produced substantial overestimates. Button et al. (2009) evaluated ridership forecasts for 47 urban transit projects completed from 1970 to 2005. They found that 34 overestimated ridership (by more than 50 percent in 18 cases). Bain (2009) evaluated more than 100 toll road projects worldwide and found that traffic volumes averaged 25 percent lower than originally forecast. Flyvbjerg et al. (2002) studied hundreds of transportation projects in many countries, including highways, urban and intercity rail projects, and bridges built over more than 50 years.8 They found that patronage is far more likely to be overestimated than underestimated and that costs are far more likely to be underestimated. Rail projects had the largest patronage overestimates and cost underestimates. In the 14 intercity high-speed and conventional rail projects studied by Flyvbjerg et al., project costs were underestimated by an average of 45 percent, and ridership overestimates were of the same magnitude. Flyvbjerg (2007) concluded that large cost underestimates combined with large ridership overestimates (and large standard deviations in both cases) result in a particularly high level of uncertainty and risk for rail projects in comparison with the other modes.
The frequency of high and low forecasts should be about equal if methods are unbiased. Researchers have characterized the empirical evidence of demand overestimates and cost underestimates as being indicative of “optimism bias” or even a tendency for strategic misrepresentation by proponents to gain project support (Flyvbjerg et al. 2005). To counter this phenomenon, Flyvbjerg and COWI (2004) have recommended that institutional checks and balances be introduced to develop independent forecasts. Two ways of doing this are by establishing peer reviews and by making empirically based risk assessments on the basis of data obtained from different projects (i.e., preparing probability distributions of the accuracy of project estimates). In the context of urban transportation planning, well-established model structures with known elasticities and interrelationships can be referenced and compared with those in the
________________
8 Fifty-eight of the total 258 transportation infrastructure projects studied were rail projects.
model used for a project. That capability is less clear for the modeling of interregional corridors.
Evaluations of the forecasting accuracy of models are critical before they are applied as standard tools for policy analysis, such as benefit–cost evaluation. Such evaluations would be highly misleading if they were based on faulty or misrepresented forecasts of demand, revenue, and costs.
Project Evaluation Methods
Transportation investments can have so many first- and second-order economic, social, and environmental impacts that evaluation criteria and methods must be diverse. Some impacts that need to be considered in terms of both their magnitude and their distribution by location and by social group are as follows [partial list developed by Goeller (1974)]:
- Transportation service impacts that occur to the users of systems, measured by trip volumes, door-to-door trip times and costs, changes in system congestion as traffic is diverted, and changes in traffic safety;
- Financial impacts on operators and the public sector, including the cost of building and operating vehicles and infrastructure in addition to the revenues from fares;
- Environmental and energy impacts such as increases or reductions in noise, air pollution, and greenhouse gas emissions, requiring consideration, among other factors, of whether the new or improved service reduces congestion on highways, on railways, or at airports and makes use of fuel-efficient technology;
- Economic impacts due to changes in employment and income in the region, including jobs added during construction, from system operation, and from multiplier effects; and
- Community impacts due to changes in activity patterns, property values, and tax bases and due to businesses and households displaced by system construction.
Benefit–cost analysis is an established method for evaluating such impacts. It is required for large transportation projects undertaken with public funds in the European Union (EU).9 The EU guidelines for benefit–
________________
9 Benefit–cost analysis is required for projects of €50 million more (European Commission 2008).
cost analysis for large transportation investments require an investigation of a project’s net impact on economic welfare in comparison with alternative actions or scenarios (business as usual, do minimum, do something, and do something else). The evaluation is to be undertaken by considering whether an investment yields incremental social benefits that exceed incremental costs. The EU guidelines explain how to conduct such evaluations for specific types of transportation investments, including highway, airport, and high-speed rail projects (European Commission 2008, 82).
The EU has developed Europe-wide models, such as TRANS-TOOLS,10 for use in forecasting for such appraisals. Many European countries have their own national transport models and routinely apply benefit–cost analysis to projects requiring government investment. Great Britain, where benefit–cost analysis has been undertaken from a welfare-maximizing perspective according to standardized methods for many years, is an example. All projects requiring government funding, including all highway projects and virtually all rail infrastructure projects, are appraised in accordance with the online guidance provided in WebTAG (Web-Based Transport Analysis Guidance).11
Benefit–cost analysis is also required in the United States for most federally funded projects.12 However, as GAO (2013) points out, there are multiple federal guidelines for valuing public benefits, and none is designated for use in analyzing high-speed rail projects. For example, high-speed rail service that reduces congestion on highways or at airports and makes use of fuel-efficient technology may provide environmental benefits (i.e., reduced pollution and greenhouse gas emissions), but there are no standards for valuing these effects. Similarly, the appraisal of new infrastructure to serve interregional travel demand requires assumptions about traveler valuations of time. Traditionally, the time spent traveling has been regarded as a penalty or disutility, with leisure and business time valued at percentages of the wage rate. Assumptions about travel time can
________________
10 Tools for Transport Forecasting and Scenario Testing. http://energy.jrc.ec.europa.eu/transtools/TT_model.html.
11https://www.gov.uk/guidance/transport-analysis-guidance-webtag.
12 Executive Order 12893 states that expected benefits and costs should be quantified and monetized to the maximum extent practicable when federal infrastructure investments in transportation, water resources, energy, and environmental protection are evaluated.
be particularly important for benefit–cost evaluations—for example, by giving priority to transportation investments that save the more highly valued time of business travelers. As communication technologies change to allow for more productive use of time spent in travel, new guidelines on time valuation assumptions and methods may be warranted.
Other intangible effects, such as economic development benefits, can be difficult to estimate, and methods for evaluating them have not been standardized. There are no standardized ways for assessing and weighting distributional or equity impacts, knowledge of which may be important along with a project’s total net cost or benefit.
GAO (2009) reviewed existing plans for 16 intercity rail passenger investments in the United States. The agency found that most of the project sponsors cited a variety of public benefits of the projects, such as congestion relief or emissions reductions, but a formal benefit–cost analysis was carried out in only four cases. Of the four analyses, none compared the proposed project with alternative modal investments, such as airport or highway expansion, although GAO (2009, 27) noted that “the proposed high speed rail line between Los Angeles, California, and San Francisco, California, has created a rough comparison of high speed rail investment with stated investment needs on the highway and air modes.” A benefit–cost assessment of California’s high-speed rail plan by Brand et al. (2001) found positive benefits when both user and nonuser benefits were considered.
The Federal Transit Administration (FTA) has had to confront the difficulty of monetizing all public benefits and costs in assessing urban transit projects proposed for “New Starts” grants. To do so, the agency has shifted the emphasis away from quantifying total net benefits and net costs to measures aimed at showing the distribution of benefits and costs according to defined socioeconomic accounts or categories, including mobility improvements, environmental benefits, operating efficiencies, land use effects, and economic development.13 Proposers assign and support qualitative values, such as high, medium, and low, for each impact category, including one category that addresses user benefits on the basis of travel time savings to users of the regional transit system. For each
________________
13http://www.fta.dot.gov/documents/FY12_Evaluation_Process%281%29.pdf.
impact category, FTA has predetermined weights that are used to rank projects. The Federal Railroad Administration has consulted with FTA to develop more standardized evaluation criteria for intercity passenger rail grants.
Data are required for constructing travel demand forecasting models and for applying and validating them. Collecting such data for interregional trips is more challenging than for local trips because of their relative infrequency and high potential for mode transfers.
The necessary data can be considered in both macro and micro terms. Macro, or aggregate-level, data are required for forecasting total travel demand in an interregional market. Such data may include household income, employment status, number and size of households, automobile availability, and other socioeconomic attributes of the population. In forecasting interregional travel demand, the relevant study area boundaries or populations of interest must be identified. They may not correspond to the official statistics routinely collected by public agencies such as the Census Bureau. Transportation planning organizations such as MPOs can obtain such data over time for their respective metropolitan regions, but this may not be the case for an interregional market that lacks such a coordinating body.
The term “microdata” refers to individuals’ characteristics and behavior. These data are usually obtained from surveys of individuals or households. As noted in Chapter 2, the value of the 1990s-era American Travel Survey database in supporting detailed and region-specific evaluations of travel demand is highly questionable because of its age. Such data must often be collected specifically for a proposed project. If the project involves a type of service that already exists, revealed preference data—that is, data based on the observation of travel behavior—may be used. The collection of revealed preference data can be challenging for interregional markets because (a) standard household surveys may not capture this infrequent travel and (b) private carriers who serve these markets (e.g., buses and airlines) may be reluctant to provide information such as traffic counts or ticket sales (Miller 2004).
When a new mode of transportation is being considered, model developers may need to use stated preference surveys. Representative trip makers are asked to make hypothetical choices between the proposed mode and existing or proposed alternatives across a range of scenarios. California is proceeding with its plan to build a high-speed passenger railway in reliance on ridership forecasts that are based largely on assessments of existing airline traffic and analyst efforts to develop realistic choice-set scenarios for stated preference surveys (Corey, Canapary, and Galanis Research 2005). These scenarios require many assumptions about the future attributes of the new service as well as those of modal alternatives, such as wait times, ease of access and egress, schedule frequencies, fares, and travel and transfer times (Cambridge Systematics 2006). Considerable attention must be given both to sample selection and size and to the design of these surveys and their scenarios to avoid inadvertently biasing respondent choices. The risk of bias may be greater when stated preference surveys and experiments are conducted by project proponents as opposed to ongoing planning entities that have no predisposition to a particular outcome or solution and that may be more inclined to have their survey instruments made public.
The Transportation Research Board committee that produced How We Travel: A Sustainable National Program for Travel Data (TRB 2011) recognized the need for up-to-date and representative data on interregional travel behavior. In its report, the committee urged the U.S. Department of Transportation to establish a National Travel Data Program, a key component of which would be a national program for passenger data collection and analysis. This proposal may be compared with the British National Travel Survey, which is continuous and surveys around 20,000 individuals in 8,000 households each year.14 The passenger data survey would collect information on how, why, when, and where people travel and on factors affecting personal travel such as car availability, driver’s license holding, and access to key services. It would provide detailed information on travel behavior and the ability to track how such behavior changes over time. By aggregating data from adjacent
________________
14https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/243957/nts2012-01.pdf.
years, geographical detail could be provided with reasonable accuracy, and—as in the British survey—the data could be used both for monitoring and for key inputs to statewide models and a national travel model.15
Finally, other data needed for modeling and evaluating interregional travel and transportation options are detailed descriptions of the existing system’s capacity, speeds, service levels, cost, and traffic congestion for the line-haul and local highway and transit networks. In particular, the attributes of the available set of transportation alternatives and all their characteristics, or choice sets, need to be well described for the corridor or region. As noted earlier, access and egress availability must be represented correctly in models and in stated preference surveys, particularly for shorter-haul interregional trips, whose beginning and ending phases can account for a significant portion of total travel time. The competitive advantage of upgrading rail and air travel speeds may be reduced or nullified if terminals have poor transit access or are located in areas that are difficult to reach because of highway congestion (Miller 2004). Similarly, peak and nonpeak representations of highway and transit service levels are needed to establish the realistic choices available to travelers in deciding to use a common carrier service or to drive.
Intergovernmental and multimodal planning and programing procedures have guided the development of metropolitan transportation systems for decades. They have helped improve the capacity to analyze urban travel demand and to evaluate transportation initiatives with regard to policy goals. In comparison, the planning and evaluation of transportation projects to accommodate interregional travel take place largely outside the multimodal context. These activities are often pursued by proponents of new modes and services who lack a broader policy perspective and who have limited incentive and ability to assess alternatives.
________________
15 Another model is the French National Travel Survey, which is conducted every 10 years. The survey consists of a computer-assisted personal interview, a 7-day travel diary, a more detailed followup questionnaire about long-distance trips taken during the past 3 months, and more thorough monitoring (via GPS) of travel by a small (750- to 1,100-person) subsample of respondents. See Roux and Armoogum 2011.
The experience with metropolitan planning indicates that an ongoing structure for multimodal and multijurisdictional decision making would help improve the capacity to prioritize policy goals, forecast travel demand, collect relevant data, and formulate and evaluate alternatives. The large number of active MPOs has prompted the development and refinement of standard methods for forecasting travel demand, assessing policy and investment options, and collecting data. The federal government, which mandates the MPO process, has provided leadership and resources to aid in these planning, analysis, and data collection efforts. Examples of the modeling, evaluation, and data capabilities required to inform transportation planning from an interregional perspective are given in this chapter. The project-specific, ad hoc approach to planning in this sector has not furthered these capabilities, whose development is hindered by a lack of ongoing institutional demand. The experience at the metropolitan level suggests that the creation of new institutions responsible for planning at the interregional level is essential to stimulating that demand.
REFERENCES
Abbreviations
| FHWA | Federal Highway Administration |
| GAO | Government Accountability Office |
| TRB | Transportation Research Board |
Adler, T., M. Doherty, J. Klodzinski, and R. Tillman. 2014. Methods for Quantitative Risk Analysis for Travel Demand Model Forecasts. In Transportation Research Record: Journal of the Transportation Research Board, No. 2429, Transportation Research Board of the National Academies, Washington, D.C., pp. 1–7.
Bain, R. 2009. Error and Optimism Bias in Toll Road Traffic Forecasts. Transportation, Vol. 36, No. 5, pp. 469–482.
Brand, D., M. R. Kiefer, T. E. Parody, and S. R. Mehndiratta. 2001. Application of Benefit–Cost Analysis to the Proposed California High-Speed Rail System. In Transportation Research Record: Journal of the Transportation Research Board, No. 1742, Transportation Research Board, National Research Council, Washington, D.C., pp. 9–16.
Button, K. 2010. Transport Economics, 3rd ed. Edward Elgar, Cheltenham, United Kingdom.
Button, K., M. Hardy, S. Doh, J. Yuan, and X. Zhou. 2009. Transit Forecasting Accuracy: Ridership Forecasts and Capital Cost Estimates. Transportation and Economic Development
Center, George Mason University, Fairfax, Va. http://ntl.bts.gov/lib/31000/31300/31361/Transit_Forecasting.pdf.
Cambridge Systematics, Inc. 2006. Bay Area/California High-Speed Rail Ridership and Revenue Forecasting Study: Levels-of-Service Assumptions and Forecast Alternatives: Final Report. http://www.hsr.ca.gov/docs/about/ridership/ridership_revenue_model_LevelService0806.pdf.
Cambridge Systematics, Inc., Vanasse Hangen Brustlin, Inc., Gallop Corporation, C. R. Bhat, Shapiro Transportation Consulting, LLC, and Martin/Alexiou/Bryson, PLLC. 2012. NCHRP Report 716: Travel Demand Forecasting: Parameters and Techniques. Transportation Research Board of the National Academies, Washington, D.C.
Corey, Canapary, and Galanis Research. 2005. High Speed Rail Study: Survey Documentation. Cambridge Systematics and the Metropolitan Transportation Commission. http://www.hsr.ca.gov/docs/about/ridership/ridership_revenue_model_survey_0505.pdf.
European Commission. 2008. Guide to Cost–Benefit Analysis of Investment Projects. http://ec.europa.eu/regional_policy/sources/docgener/guides/cost/guide2008_en.pdf.
FHWA. 2012. Foundational Knowledge to Support a Long-Distance Passenger Travel Demand Modeling Framework: Review of Experience. DTFH61-10-R-00036.
Flyvbjerg, B. 2007. Cost Overruns and Demand Shortfalls in Urban Rail and Other Infrastructure. Transportation Planning and Technology, Vol. 30, No. 1, pp. 9–30.
Flyvbjerg, B., and COWI. 2004. Procedures for Dealing with Optimism Bias in Transport Planning: Guidance Document. Department for Transport, London.
Flyvbjerg, B., M. S. Holm, and S. Buhl. 2002. Underestimating Costs in Public Works Projects: Error or Lie? Journal of the American Planning Association, Vol. 68, No. 3, pp. 279–295.
Flyvbjerg, B., M. K. S. Holm, and S. L. Buhl. 2005. How (In)accurate Are Demand Forecasts in Public Works Projects? The Case of Transportation. Journal of the American Planning Association, Vol. 71, No. 2, pp. 131–146.
GAO. 2009. High Speed Passenger Rail: Future Development Will Depend on Addressing Financial and Other Challenges and Establishing a Clear Federal Role. GAO-09-317, March. http://www.gao.gov/new.items/d09317.pdf.
GAO. 2013. California High-Speed Passenger Rail: Project Estimates Could Be Improved to Better Inform Future Decisions. GAO-13-304. http://www.gao.gov/assets/660/653401.pdf.
Goeller, B. F. 1974. System Impact Assessment: A More Comprehensive Approach to Public Policy Decisions (unpublished). Cited by E. S. Quade, Analysis for Public Decisions, North Holland, New York, 1982.
Horowitz, A. 2006. NCHRP Synthesis 358: Statewide Travel Forecasting Models. Transportation Research Board of the National Academies, Washington, D.C.
Martin, W. A., and N. A. McGuckin. 1998. NCHRP Report 365: Travel Estimation Techniques for Urban Planning. Transportation Research Board, National Research Council, Washington, D.C.
Miller, E. J. 2004. The Trouble with Intercity Travel Demand Models. In Transportation Research Record: Journal of the Transportation Research Board, No. 1895, Transportation Research Board of the National Academies, Washington, D.C., pp. 94–101.
Pinjari, A. R., and C. R. Bhat. 2010. Activity-Based Travel Demand Analysis. In A Handbook of Transport Economics (A. de Palma, R. Lindsey, E. Quintet, and R. Vickerman, eds.), Edward Elgar, Cheltenham, United Kingdom, pp. 213–248.
Roux, S., and J. Armoogum. 2011. Calibration Strategies to Correct Nonresponse in a National Travel Survey. In Transportation Research Record: Journal of the Transportation Research Board, No. 2246, Transportation Research Board of the National Academies, Washington, D.C., pp. 1–7.
Schiffer, R. G. 2012. NCHRP Report 735: Long-Distance and Rural Transferable Parameters for Statewide Travel Forecasting Models. Transportation Research Board of the National Academies, Washington, D.C.
TRB. 2011. Special Report 304: How We Travel: A Sustainable National Program for Travel Data. National Research Council of the National Academies, Washington, D.C. http://onlinepubs.trb.org/onlinepubs/sr/sr304.pdf.
