Guidance for Calculating the Return on Investment in Transit State of Good Repair (2019)

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.

A-1 A P P E N D I X A Literature Review Summary This appendix summarizes a review of literature related to calculating the return on investment (ROI) in achieving or maintaining a state of good repair (SGR) for transit capital assets. The review is focused on factors that are unique to the research and/or extend beyond prior TCRP research. Specifically, the review details the existing literature on 1) the effects of transit quality of service on travel demand and 2) identifying and quantifying the economic and social impacts of transit investment. The following sections summarize the review approach and key findings from these two areas. An annotated bibliography of resources reviewed is provided in Appendix B. Effects of Transit Quality of Service on Travel Demand Approach Reliability is a key aspect of transportation services, whether by private or public modes, and there is a broad consensus in the research community that unreliability affects traveler decision making, transit ridership, and satisfaction. This literature review focused on four topics. First, theoretical aspects of modeling the effect of unreliability on travelers were considered—i.e., how unreliability has been incorporated into travel choice models. It was found that for benefit–cost analysis (BCA), two measures were potentially of interest: the reliability ratio and the reliability multiplier. Second, specific values for the reliability ratio and reliability multiplier were sought. Given the variability in results across studies and the need for generalizable results, review articles were selected in which the authors discussed the sources of variation and reported averages across larger numbers of studies. However, some individual papers were also included in the literature review if they discussed findings related to the interpretation and application of the reliability measures in BCA or in the public transportation context. Third, a small number of papers on ridership prediction were reviewed to evaluate whether there were any generalizable regression models that could be used in lieu of travel demand models, since the latter are not available to all agencies. This was found not to be the case. Instead, the recommended procedure is to use demand elasticities. Fourth, to derive specific elasticity values, a small set of secondary literature on elasticities was consulted.

A-2 Guidance for Calculating the Return on Investment in Transit State of Good Repair Key Findings Industry Practice Barron et al. (1) reports that only two metro systems worldwide track passenger impacts of unreliability, although the identity of the systems is not disclosed. Zhao et al. (2 ) describe an aggregate measure called “Excess Journey Time” that is used by Transport for London to track the difference between travel times under ideal conditions and observed travel times in congested or disrupted conditions. None of these metrics has been translated into a predictive model that could be used for BCA. However, a few European countries have recently adopted official Values of Reliability (VORs) for use in BCA, and some published studies have provided details on the approach used to derive the VOR. A recent example of such a study is Kouwenhoven et al. (3). In the United States, two transit-focused applications have been reported in the literature: Paterson and Vautin (4) included average transit delay times in a regional activity-based model, and the Los Angeles County Transportation Authority is reported to be using on-time performance in its regional travel demand model (5). Although reliability is only beginning to be incorporated in travel demand models, a number of studies have shown that passengers value reliability highly. A survey of transit users in San Francisco found that a large portion of users reported actively adapting their travel behavior to deal with unreliability on the network (6). In that survey, Carrel et al. further note that some passengers’ exclusive reliance on real-time information rather than on the timetable might contribute to perceptions of unreliability. Modeling Approaches To predict the effect of unreliability on transit passengers, one needs to be able to capture it in travel choice models. In general, there are three approaches to doing so. The following paragraphs describe each of the three approaches and the relationship of the modeling approaches to microeconomic theory. Scheduling Model. This model assumes that each traveler has a preferred arrival time at his or her destination and must choose an optimal departure time. The model is originally attributed to Gaver (7 ) and Vickrey (8) and was developed in its current form by Small (9). The premise of the model is that the traveler derives a marginal disutility from every minute of late arrival after the preferred arrival time (commonly called Schedule Delay Late, or SDL) and every minute of early arrival (Schedule Delay Early, or SDE), as well as an optional additional penalty from being late per se. Given a travel time distribution, the traveler trades off the disutility of being late against the disutility of being early and allocates an optimal safety margin. The utility equation is usually specified in a linear additive form as follows: (A-1) where td is the departure time,T is the travel time, and DL is a dummy variable for arriving late. This model specification is generally considered to have a behavioral underpinning that is consistent with microeconomic theory, but the major drawback is

Literature Review Summary A-3 that estimating it requires the modeler to know the preferred arrival time of each traveler. In practice, researchers often discard the additional lateness penalty, DL, to simplify model estimation. An estimated scheduling model allows the researcher to derive two important marginal rates of substitution: The value of SDL ($/minute of late arrival) and the value of SDE ($/minute of early arrival). Centrality-Dispersion Approach. This is often called the mean–variance approach. This formulation, which is attributed to Jackson and Jucker (10), includes a term for the expected travel time and a term for the travel time variability (and hence, unreliability) in the utility equation. Most frequently, it is expressed as follows: (A-2) where is the mean of the travel time and is the standard deviation. As with the scheduling model, it is assumed that the traveler is aware of the travel time distribution. After estimating the coefficients, and using the formulation with a mean and a standard deviation, marginal rates of substitution can be calculated to obtain three frequently reported measures: The value of time (VOT, in $ per minute of expected travel time), the value of reliability (VOR, in $/minute of standard deviation), and the reliability ratio (RR = VOR/VOT). The reliability ratio describes how many minutes of expected travel time travelers are willing to trade off for a minute of standard deviation. This model is appealing due to its simplicity and because it does not require knowledge of the preferred arrival time. A further advantage is that the dispersion term cannot only capture the disutility of being late or early but also the general nuisance of travel time uncertainty. Some researchers have reported success with other measures of centrality and dispersion, such as the difference between the median and the 80th or 90th percentiles of travel time—e.g., Brownstone and Small (11). Mean-Lateness Approach. This was introduced as a performance metric for low- frequency, scheduled passenger rail services in various European countries. In this approach, the utility equation contains the scheduled travel time and the mean lateness at the arrival station. The formulation was expanded by Batley and Ibáñez (12) to include mean lateness at the boarding station, resulting in the following model: (A-3) where is the scheduled travel time, the mean lateness at arrival, and is the mean lateness at boarding. This model assumes that the scheduled departure and arrival times are known by the traveler and that lateness with respect to the schedule is evident to the operator and to the traveler. From the estimated coefficients, it is possible to derive the so-called reliability multiplier, which is the ratio of the value of a minute of lateness over the value of a minute of scheduled travel time.

A-4 Guidance for Calculating the Return on Investment in Transit State of Good Repair Since this model has seen limited use outside of the UK, Sweden, and Greece, there are fewer published reliability multipliers in academic literature. Furthermore, some concerns have been raised regarding the validity of this approach by Börjessen et.al. (13), who found nonlinearities in passengers’ valuation of delay times. In an unrelated project in Los Angeles, Vovsha et al. (5) made similar observations. Relationship to Microeconomic Theory. An important contribution to the state of the art was by Fosgerau and Karlström (14). They showed that under certain assumptions, the centrality-dispersion model can be derived from the scheduling model and therefore that the centrality-dispersion model is also consistent with microeconomic theory. In illustrating this result, Small (15) presents two different decompositions of the expected (generalized) user cost. The first one includes the standard deviation of the travel time as a standalone reliability measure. This decomposition, which mirrors a common formulation of the centrality-dispersion approach, is empirically convenient if the shape of the travel time distribution is unknown. However, it becomes clear from the derivation that the VOR, which is multiplied with the standard deviation to obtain the generalized cost of unreliability, will depend on the shape of the travel time distribution. This means that calculated VOR will not be comparable between studies unless the travel times follow the same standardized distribution. The second decomposition shifts the variable describing the shape of the distribution to the reliability measure (i.e., it is multiplied with the standard deviation), thus turning it into an unweighted average of all percentiles of the travel time distribution beyond a certain threshold. The threshold depends on the relative disutility of arriving late or early; for a conventional choice of parameters, roughly based on empirical evidence, it is the 80th percentile. This measure is interpreted as the “mean lateness in standardized travel time,” and with this decomposition, the VOR is purely a behavioral quantity that should be portable between studies with similar populations. Note that in this case, the mean is subtracted from the travel times and that despite the similarity, this measure is not the same as the difference between the median and the 80th percentile used in older studies (11). To the best of our knowledge, this reliability measure has not yet been operationalized in any large-scale study, and there is an ongoing discussion in the research community about the assumptions underlying the derivation (13). Many of the empirical studies to date have been conducted on highway travel, where the shape of the travel time distribution is often approximately normal or lognormal and where travelers can choose their departure time freely. On the other hand, transit travel time distributions can be rather unique to a system or context and can depend on service intervals and transfer times, as discussed in depth by Bates et al. (16). This presents a modeling difficulty that has yet to be fully addressed. Another issue is that, in reality, travelers’ choices are a function of perceived travel time distributions rather than objectively measurable distributions, and that the former are often distorted by perceptive biases. In a similar vein, stated preference surveys

Literature Review Summary A-5 have been the predominant form of data collection due to a number of practical considerations. However, researchers have found it difficult to present concepts of reliability and probabilities of various travel times to survey participants in a generally understandable way and consistent way. Li et al. (17 ) identify differences in presentation as an important source of variation in empirical results. Passenger Impacts of Unreliability Unreliability can cause a shift in mean travel times (e.g., due to slow zones) and contribute to the day-to-day variance of travel times (e.g., due to unpredictable vehicle or infrastructure failures). This imposes two types of costs on passengers. First, there is a direct cost of delays. Passengers who are on board a vehicle that fails or who are in a vehicle that is stuck behind another failed vehicle or track segment experience additional on-board travel time. Passengers who are waiting for a vehicle that is delayed due to a failure experience additional wait time. There is evidence that direct delay times are perceived differently depending on whether they are experienced on board a vehicle or waiting at a stop (18). In a study with users of the San Francisco transit system, Carrel et al. (18, 19) found that the marginal contribution of a minute of in-vehicle delay to passenger dissatisfaction and ridership losses was higher than the marginal contribution of an additional minute of wait time at the origin stop. Paterson and Vautin (4) name the in- vehicle delays “type 1 delays” and out-of-vehicle delays “type 2 delays.” Second, there is a routine cost of departing earlier to allow a safety margin, given the uncertainty of day-to-day travel times. A further impact of unreliability is increased crowding in some (or all) vehicles, either due to fewer vehicles being available or to larger variations in headways between vehicles. Van Oort et al. (20) note that unreliability can also affect public perception of the transit system in the general population, especially if incidents are publicized. Benefit–Cost Analysis Although the scheduling model is considered most accurate from a microeconomic and behavioral point of view, the data requirements (in particular, the reliance on a preferred arrival time) make it impractical for large-scale BCA. The mean-lateness approach, although it has previously been used in BCA in the UK, requires further research due to the concerns raised by Börjessen and Eliasson (21). The centrality-dispersion model, on the other hand, has been applied in BCA in a variety of settings, and the result by Fosgerau and Karlström (14) lends theoretical credibility to this approach. The Netherlands, Sweden, and Norway have defined an official reliability ratio for BCA, which in turn has led to the widespread use of a formulation with the mean and standard deviation. In fact, Kouwenhoven et al. (3) argue that the mean and standard deviation are currently the only realistically feasible

A-6 Guidance for Calculating the Return on Investment in Transit State of Good Repair measures for large-scale BCA. An advantage of using a formulation with the mean and the standard deviation of travel time is that by choosing this approach, modelers can draw on a large body of published VOR from the United States, Europe, and Australia. For public transportation projects, the centrality-dispersion approach is particularly suitable for non-scheduled services with high travel time variability (13), and Van Oort (22) describes a successful application to urban light rail in the Netherlands. A generalizable procedure for including transit reliability in a BCA framework is described by Van Oort (22) and summarized below, with slight modifications. Current literature does not provide an indication of how transfer times should be treated, and no separate reliability ratios have been published for wait times at the boarding stop. Therefore, the wait times at the boarding stop will have to be converted into equivalent in-vehicle travel times, and the analysis is limited to unlinked trips on individual routes. The first step is to derive the wait time and in-vehicle travel time distribution for every for every origin-destination pair on every route, which yields the means and variances of the distributions. Optionally, a weighting factor can be applied to wait times to convert them to equivalent in-vehicle travel time. The variances of in-vehicle travel time and wait time are assumed to be independent, which is admittedly a simplification, but allows them to be added while omitting a covariance term. How the travel time distributions are computed depends on the data available to the transit agency: If the locations and probabilities of failure of individual assets are known, it is recommended to calculate delays at a route level. If they are unknown, average values will need to be computed for all routes, similar to the approach taken by Paterson and Vautin (4). After taking the square root of the variance of the generalized travel cost to yield the standard deviation, it is then multiplied by a reliability ratio, and the additional travel time component is added to the mean travel time. This results in a generalized travel time between every origin-destination pair, as follows: (A-4) Where: TT - Total (generalized) travel time , - Mean and variance of in-vehicle travel time , - Mean and variance of wait time - Optional weighting factor for wait time - Reliability Ratio The second step is to use the generalized travel time as input to a ridership prediction model. This can either be done directly, or TT can be multiplied by a VOT coefficient to obtain a generalized cost. If possible, the generalized origin-destination travel times should be weighted by origin-destination flows on each route. If a regional travel demand model is available through the local Metropolitan Planning Organization, it is recommended to calculate the ridership impacts with the travel demand model since it most accurately captures the availability and travel times of alternative modes. If no

Literature Review Summary A-7 travel demand model is available, standard factors for elasticity of demand with respect to travel times or generalized costs can be applied. An intermediate option is to use a regression model to predict changes in demand (23, 24). This is only recommended if a fully calibrated regression model for the respective urban area is available, and even then, one should be cautious: As Iseki and Ali point out, omitted variable biases are relatively prevalent and difficult to account for in regression-based ridership prediction studies (24). When predicting changes in ridership, it is important to consider that the full impacts of increased unreliability are often only observable after a time lag, as users adjust their travel behavior and auto ownership. Long-run elasticities of demand are often 2–3 times as large as short-run elasticities, as reported by Wardman (25) and Litman (26). Preston et al. (27 ) cite a study on passenger rail in the UK that showed a lag of approximately 6 months before performance affected revenue. Value of Reliability Due to variations in study protocols, data sources, and model specifications, the literature has not been able to converge on generally accepted VORs. Furthermore, as shown by Small (15) and Börjessen et al. (13), different types of travel time distributions and other modeling assumptions limit the transferability of published results. A recent review study by Burris, et al. (28) reviews the literature on VOT and VOR and describes an effort to measure these for travelers on a freeway corridor in Texas utilizing transponder data. In a review of European evidence, Wardman et al. (29) find that values of time derived from Revealed Preference (RP) data significantly exceed values derived from SP data, by a factor of 1.38 for non-business and 1.46 for business trips. While the VOR is not analyzed separately by Wardman et al., it is reasonable to expect differences as well. Furthermore, geographical variations are possible, but at present it is not known whether there are systematic differences in the VOR between countries or regions. To determine a value of the reliability ratio for BCA, four broad review articles can be considered, where authors have tried to understand the differences and commonalities between published studies on the reliability ratio. The first review is by Carrion and Levinson (30): The authors compile a list of published reliability ratios ranging from 0.1 to 3.3. Although a meta-model developed by the authors yields inconclusive results, their work indicates a ballpark figure for the reliability ratio to be around 1 across all modes. It should be noted that the majority of this evidence is from car travel. The second review is by Li et al. (17 ), who suggest a reliability ratio between 0.8 and 1.4 for car travel and between 1.4 and 2.1 for public transportation, citing earlier work. They argue that a higher reliability ratio for public transportation is plausible since the consequences of unreliability are more severe for transit users due to missed transfers, fixed service intervals, and crowding.

A-8 Guidance for Calculating the Return on Investment in Transit State of Good Repair This assessment is echoed by Bates et al. (16 ), who also write that, in addition to direct impacts on travel times, there is some evidence that travelers dislike failure to adhere to the schedule. Li et al. (17 ) also observe that travelers with flexible arrival times tend to place a lower value on reliability, and that the VOR for business trips is generally higher than for other trip purposes. Another review conducted by Wardman et al. (29) and focused only on Europe, names an average reliability ratio of 0.64 across all modes. Typically the estimates provided in the reviews are based on SP data. In general, non-commute trips are often found to have higher reliability ratios than commute trips, which is primarily due to the lower VOT for non-commute trips. Given the large variability in previous results, it is difficult to conclusively recommend a reliability ratio for use in BCA. Therefore, we recommend running the analysis with three different values of the reliability ratio: 1, 1.5, and 2. The value of 1.5 is chosen as the most plausible reliability ratio for public transportation, based on the analysis by Li et al. (17 ). 1 and 2 represent lower-end and higher-end estimates of the reliability ratio. Compared to the results of the review studies summarized above, these values have been slightly inflated to reflect the higher valuations expected from RP data (29, 30). As mentioned above, no separate reliability ratios have been published for wait times. Economic and Social Impacts of Transportation Investments Approach In addition to the review of literature on the effects of transit quality of service on travel demand, the research team also reviewed publications/reports related to the economic assessment of transit SGR. While the individual articles vary in their technical focus, collectively the articles trace the connection between the condition of a public transit asset and economic outcomes. This chain of linkages is as follows: 1. A transit asset’s SGR describes its physical condition, which may range from well-maintained (assets are functioning at their ideal capacity within their design life) through various degrees of physical deterioration that correspond to compromised performance. While different transit agencies may define SGR in different terms, from an economic perspective the key relationship is between the physical condition of the asset and its operating performance. 2. The performance of the asset impacts the generation of user benefits (safety and travel time costs). These are the direct economic outcomes of a transit asset operating at peak or compromised levels of performance. 3. Travelers’ decisions to select transit for their travel is based on the relative attractiveness between transit and other competing modes for that trip. As transit is perceived to be less safe, more time consuming, or less reliable than other modes, travelers will shift increasingly to other modes or not make the trip.

Literature Review Summary A-9 4. As the travelers shift from transit to auto, travelers and the local economy may experience secondary economic impacts in terms of higher travel costs per trip, greater peak-hour congestion on roads, and greater emissions. 5. While points 1–4 outline an economic efficiency perspective, there are equity considerations as well. For low-income or transit-dependent riders, a reduction in system performance plays a disproportionate impact on their household budgets and access to employment, shopping, health services, and entertainment. The increase in costs associated with shifting away from transit to another mode represents a larger share of a low-income household’s disposable income than a household with a median or higher income. Unreliable transportation imposes other costs in terms of day-care late fees or missed appointments, which again erode the households’ disposable income. Materials reviewed include studies that suggest types of economic effects or methods, quantify the ROI of bringing a transit system to SGR, and provide economic estimates. Results from specific case studies might be used as benchmarks in deriving the economic benefits of transit SGR investments. Key Findings Several papers reviewed suggested the types of economic and social impacts of transportation investments. Weisbrod et al. shows that investment in public transportation, including SGR investment, has wide economic benefits (31). Specifically this report highlights the impacts to wages and business income. This report also summarizes productivity impacts as public transportation improves mobility and spending impacts from increased investment in capital and operations programs. APTA published an updated of the report in 2014 (32). SGR investment has social impacts as well as economic impacts. One goal of public transportation may be to equally distribute the benefits and burdens of public transportation projects, thereby improving equity. State of Good Repair and Transit Equity (33) suggests that equity analysis can help prioritize and allocate SGR strategies and investments. This paper provides a methodology for conducting service equity analysis that can be used in conjunction with other economic based benefit/cost analyses. In addition to the positive impacts of SGR investment, a 2011 American Society of Civil Engineers report analyzes the impact of failing to invest adequately in SGR (34). Specifically, this report highlights the economic implications of surface transportation infrastructure deficiencies on the U.S. economy. It provides a picture of worsening conditions if SGR investments are not prioritized in the coming years, including values for vehicle operating costs, travel time delays, safety costs, and environmental costs. Further economic and social impacts of transportation projects on their surrounding communities are provided in NCHRP Report 456 (35). This guidebook includes sections on analyzing the impacts of projects on accessibility and community cohesion, for

A-10 Guidance for Calculating the Return on Investment in Transit State of Good Repair example. These are important aspects that may be considered when justifying and prioritizing SGR investments. Several other papers reviewed provided guidance on calculating various economic benefits and costs and the ROI of SGR investment. The primary resource the research team focused on in the review was the New Zealand Transport Agency’s Economic Evaluation Manual (EEM) (36). This document is the industry’s standard for the economic evaluation of land transport activities for New Zealand. The EEM sets out economic evaluation procedures and values used in calculating benefits and costs. Of particular interest to the current project is the guidance this document provides on justifying SGR investments. Detailed methodologies for calculating costs and benefits from this document will be useful in ultimately providing guidance to transit agencies on calculating the ROI of SGR investments. Finally, case study papers were included in the literature review. Two case studies focus on Chicago’s transit system and estimate the expected ROI for several capital improvement scenarios outlined in the Regional Transportation Authority’s (RTA) Strategic Plan (37 ) and in (38). These may be useful in developing the guidance for this project as the team seeks to build upon the existing literature and contribute innovative methods for calculating the ROI of SGR investments. References 1. Barron, A., P. Melo, J. Cohen, and R. Anderson. “Passenger-Focused Management Approach to Measurement of Train Delay Impacts.” Transportation Research Record: Journal of the Transportation Research Board, No. 2351, 2013, pp. 46–53. 2. Zhao, J., M. Frumin, N. Wilson, and Z. Zhao. “Unified Estimator for Excess Journey Time Under Heterogeneous Passenger Incidence Behavior Using Smartcard Data.” Transportation Research Part C: Emerging Technologies, 34, 2013, pp. 70–88. 3. Kouwenhoven, M., G. C. de Jong, P. Koster, V. A. van den Berg, E. T. Verhoef, J. Bates, and P. M. Warffemius. “New Values of Time and Reliability in Passenger Transport in the Netherlands.” Research in Transportation Economics, 47, 2014, pp. 37–49. 4. Paterson, L., and D. Vautin. “Evaluating the Regional Benefit/Cost Ratio for Transit State of Good Repair Investments.” Journal of Public Transportation, 18(3). 2015. 5. Vovsha, P., G. S. O. Marcelo, and D. William. “Statistical Analysis of Transit User Preferences Including In-Vehicle Crowding and Service Reliability.” Transportation Research Board 2014 Annual Meeting Compendium of Papers. 2014. 6. Carrel, A., A. Halvorsen, and J. Walker. “Passengers' Perception of and Behavioral Adaptation to Unreliability in Public Transportation.” Transportation Research Record, No. 2351, 2013, pp. 153–162.

Literature Review Summary A-11 7. Gaver, D.P. “Headstart Strategies for Combating Congestion.” Transportation Science 2, 1968, pp. 172–181. 8. Vickrey, W. S. “Congestion Theory and Transport Investment.” The American Economic Review, 59(2), 1969, pp. 251–260. 9. Small, K. A. “The Scheduling of Consumer Activities: Work Trips.” The American Economic Review, 72(3), 1982, pp. 467–479. 10. Jackson, W. B., and J. V. Jucker. “An Empirical Study of Travel Time Variability and Travel Choice Behavior.” Transportation Science, 16(4), 1982, pp. 460-475. 11. Brownstone, D., and K. A. Small. “Valuing Time and Reliability: Assessing the Evidence from Road Pricing Demonstrations.” Transportation Research Part A: Policy and Practice, 39(4), 2005, pp. 279–293. 12. Batley, R., and N. Ibáñez. “Randomness in Preferences, Outcomes and Tastes, an Application to Journey Time Risk.” International Choice Modelling Conference, Yorkshire, UK. 2009. 13. Börjesson, M., J. Eliasson, and J. P. Franklin. “Valuations of Travel Time Variability in Scheduling Versus Mean–Variance Models.” Transportation Research Part B: Methodological, 46(7), 2012, pp. 855–873. 14. Fosgerau, M., and A. Karlström, A. “The Value of Reliability.” Transportation Research Part B: Methodological, 44(1), 2010, pp. 38–49. 15. Small, K. A. “Valuation of Travel Time.” Economics of Transportation, 1(1), 2012, pp. 2–14. 16. Bates, J., J. Polak, P. Jones, and A. Cook. “The Valuation of Reliability for Personal Travel.” Transportation Research Part E: Logistics and Transportation Review, 37(2), 2001, pp. 191–229. 17. Li, Z., D. A. Hensher, and J. M. Rose. “Willingness to Pay for Travel Time Reliability in Passenger Transport: A Review and Some New Empirical Evidence.” Transportation Research Part E: Logistics and Transportation Review, 46(3), 2010, pp. 384–403. 18. Carrel, A., R. G. Mishalani, R. Sengupta, and J. L. Walker. “In Pursuit of the Happy Transit Rider: Dissecting Satisfaction Using Daily Surveys and Tracking Data.” Journal of Intelligent Transportation Systems, 20(4), 2016, pp. 345–362. 19. Carrel, A., and J. L. Walker. “Understanding Future Mode Choice Intentions of Transit Riders as a Function of Past Experiences with Travel Quality.” European Journal of Transport and Infrastructure Research, in print. 2017. 20. Van Oort, N., T. Brands, E. de Romph, J. A. Flores, and B. Eng. “Unreliability Effects in Public Transport Modelling.” International Journal of Transportation, 3(1), 2015, pp. 113–130. 21. Börjesson, M., and J. Eliasson. “On the Use of ‘Average Delay’ as a Measure of Train Reliability.” Transportation Research Part A: Policy and Practice, 45(3), 2011, pp. 171–184. 22. Van Oort, N. “Incorporating Enhanced Service Reliability of Public Transport in Cost-Benefit Analyses.” Public Transport, 8(1), 2016, pp. 143–160. 23. Chakrabarti, S. “The Demand for Reliable Transit Service: New Evidence Using Stop Level Data from the Los Angeles Metro Bus System.” Journal of Transport Geography, 48, 2015, pp. 154–164.

A-12 Guidance for Calculating the Return on Investment in Transit State of Good Repair 24. Iseki, H., and R. Ali. “Fixed-Effects Panel Data Analysis of Gasoline Prices, Fare, Service Supply, and Service Frequency on Transit Ridership in 10 U.S. Urbanized Areas.” Transportation Research Record: Journal of the Transportation Research Board, No. 2537, 2015, pp. 71–80. 25. Wardman, M. “Review and Meta-Analysis of UK Time Elasticities of Travel Demand.” Transportation, 39(3), 2012, pp. 465–490. 26. Litman, T. Understanding Transport Demands and Elasticities: How Prices and Other Factors Affect Travel Behavior. Victoria Transport Policy Institute. 2013. 27. Preston, J., G. Wall, R. Batley, J. Ibáñez, and J. Shires. “Impact of Delays on Passenger Train Services: Evidence from Great Britain.” Transportation Research Record: Journal of the Transportation Research Board, No. 2117, 2009, pp. 14–23. 28. Buris, M., C. Spiegelman, A. Abir, and S. Lee. Travelers’ Value of Time and Reliability as Measured on Katy Freeway. Texas A&M Transportation Institute Report PRC 15-37 F. 2016. 29. Wardman, M., V. P. K. Chintakayala, and G. de Jong. “Values of Travel Time in Europe: Review and Meta-Analysis.” Transportation Research Part A: Policy and Practice, 94, 2016, pp. 93–111. 30. Carrion, C., and D. Levinson. “Value of Travel Time Reliability: A Review of Current Evidence.” Transportation Research Part A: Policy and Practice, 46(4), 2012, pp. 720–741. 31. Weisbrod, W., and A. Reno. Economic Impact of Public Transit Investment. Technical report prepared for APTA through TCRP Project J-11, Task 7. 2009. 32. APTA. Economic Impact of Public Transportation: 2014 Update. 2014. 33. Amber Ontiveros Associates, LLC, and Econsult Solutions. State of Good Repair and Transit Equity. 2015. 34. American Society of Civil Engineers. Failure to Act: The Economic Impact of Current Investment Trends in Surface Transportation Infrastructure. 2011. 35. Forkenbrock, D., and G. Weisbrod. NCHRP Report 456: Guidebook for Assessing the Social and Economic Effects of Transportation Projects. TRB, National Research Council, Washington, D.C., 2001. 36. New Zealand Transport Agency. Economic Evaluation Manual. 2016. 37. Chicago Metropolis 2020. Time is Money: the Economic Benefits of Transit Investment. 2007. 38. Cambridge Systematics, Inc., with Vlecides-Shroeder Associates, Inc., B. Welsh, and Ernest Sawyer Enterprises, Inc. Investment in Public Transportation: The Economic Impacts of the RTA System on Regional and State Economies. 1995.