Value of Travel Time Reliability in Transportation Decision Making: Proof of Concept—Maryland (2014)

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.

1Executive Summary The topic of travel time reliability has been a significant focus in the transportation systems man- agement and operations (TSM&O) community during recent years. With the end of the second Strategic Highway Research Program (SHRP 2) Reliability research program in sight, agencies are working to figure out how to incorporate travel time reliability–related performance measures, analytical processes, and tools into their planning and programming processes. Travel time reli- ability describes the quality, consistency, timeliness, predictability, and dependability of travel. What is occurring today is a fundamental shift from a past policy focus on average travel time to one that now focuses on variability of travel time. The specific problem that this research project addresses is to identify how an agency can include a value of travel time reliability (VTTR) in a benefit–cost analysis (BCA) when making congestion reduction–related project investment decisions. This project builds on the experi- ences of the Maryland State Highway Administration (SHA) and their ongoing efforts to include reliability in their planning and programming processes. In recent years, SHA has adopted a reliability performance measure and has included a VTTR in their BCA process when selecting congestion relief projects for implementation. The stated project objectives for this project were as follows: • Select and defend a value or range of values for travel time reliability for the Maryland State Highway Network. • Use the VTTR in the Maryland SHA project development process to prioritize operational and capital improvements and determine if (and how) the ranking of projects changes due to the addition of VTTR. • Report for the benefit of others the step-by-step process used to develop, justify, apply, and assess the use of VTTR in the Maryland SHA project evaluation and decision process. This research project is presented in two parts. Part 1: Background and Application of the Method provides the results of the project in four chapters: (1) Background, (2) Research Approach, (3) Findings and Applications, and (4) Conclusions and Suggested Research. Part 2: Description of the Method provides an in-depth treatment of the development and application of a travel-time data-driven methodology for estimating value of reliability, including the meth- odology’s assumptions, example application and calculations, and how it attempts to improve on a previous application of Real Options theory. The following sections provide a synopsis of how each objective was addressed, along with any related findings and products. The final section of the Executive Summary addresses conclusions and recommendations for further research.

2Select and Defend a Value or Range of Values SHA currently uses a VTTR in their existing life-cycle BCA for congestion relief projects. Follow- ing recent trends, particularly in European nations where reliability benefits are accounted for as a percent of congestion reduction–related savings, SHA adds 75% (known as the reliability ratio [RR]) of the congestion-related savings as reliability savings to overall project benefits. This research project demonstrates how this value can be defended by (1) a review of existing litera- ture and (2) a proposed data-driven methodology for determining a new value of reliability (or range of values) using mass quantities of local historical travel time data. Based on the results of (2), new localized values of the reliability ratio were calculated and input into the current life- cycle BCA methodology (as described in the next section). In the past, two distinct approaches have been used to define travel time reliability for valua- tion purposes, the first of which is based on behavioral modeling, which has been, by far, the most frequently used approach. Behavioral approaches followed two major paths: (1) statistical methods that directly estimate travel time distributions and variations and (2) survey-based methods based on disaggregate data and discrete choice models. A detailed literature search of these approaches is included in this report. Compared with the recent revealed and stated prefer- ence survey-based estimates in the literature, SHA’s current use of a reliability ratio of 0.75 seems reasonable and may even be, to some extent, conservative. The second approach is based on Real Options theory, which has been applied once under SHRP 2 L11: Evaluating Alternative Operations Strategies to Improve Travel Time Reliability. This SHRP 2 L35B research project made a concerted effort to improve on the L11 methodol- ogy by building off this previous work, while, at the same time, providing transparency in the newly developed methodology and clearly demonstrating how issues identified in the L11 approach have been addressed. There are three strong reasons for continuing to pursue this approach: (1) it is based on access to historical travel time data, which is becoming more acces- sible to agencies via contracts with third party probe data providers or the freely available Federal Highway Administration (FHWA)–sponsored National Performance Measures Research Data Set (NPMRDS); (2) because of access to ubiquitous archived travel time data, the methodology is readily implementable by agencies; and (3) it provides a different kind of “tool” for the travel time reliability valuation “tool kit,” in addition to the existing behavioral modeling approaches. Challenges remain, particularly in conveying a complex approach in a manner that is relatively intuitive and easy to understand. In an attempt to make the complex relatively simple, the proposed travel-time data-driven methodology for estimating value of reliability uses large quantities of historical travel time data, along with a value of typical/usual travel time (VOTT, also known as VOT) and produces an RR along with a value of travel time reliability (VTTR, or VOR). A brief summary of the steps involved in the methodology is as follows: First, the appropriate parameters of a stochastic pro- cess describing the evolution of travel time observations are calibrated. This stochastic process is used to predict the future distribution of travel time. Then, based on well-established results of other behavioral studies, the predicted late and early arrivals are transformed into equivalent monetary penalty values. The final step is to calculate the current certainty-equivalent expected value of future penalties, which results in the VTTR. Note that, conventionally, the RR is defined as the ratio of VTTR and VOTT. In providing a high-level explanation of how the methodology works, an analogy is used that is related to the purchase of an insurance premium that guards against the risk of being late. If travel- ers, based on experience, know that their morning commute to work takes 10 minutes on average, they might be willing to add 5 minutes to their trip time to avoid the risk of being late to work. This extra 5 minutes has a monetary value and represents the insurance premium that the traveler is will- ing to pay for this trip. The challenge is to determine this value (the extra 5 minutes in this example) using factors such as expected travel time, variations in historical travel time, tolerance of travel time variation, and how differences in expected travel time might impact the travelers’ experience.

3In practice, the methodology involves the complex application of Real Options theory. Addi- tional detail about the methodology is included in Chapter 3. However, for a detailed in-depth treatment of the methodology’s development, its assumptions, example application and calcula- tions, as well as how it improves on the previous application of Real Options theory, refer to Part 2. The methodology uses large quantities of historic travel time data for a trip to (1) calcu- late the future distribution of travel time and (2) using this future time distribution, apply a recursive process to estimate the present value of reliability. The proposed data-driven methodology was implemented in Maryland and used to estimate a local value for RR and ultimately a travel time VTTR. The methodology was implemented using MATLAB to automate the process (note that the MATLAB code is provided in Appendix B). A year’s worth of archived probe-based travel time data was used to estimate the local RR and VTTR values on five different corridors in Maryland. Results of the data-driven methodology application indicate that the currently used RR of 0.75 is within the calculated range of values for commute trips (0.68 to 0.87). Use VTTR in the Maryland SHA Project Development Process As noted previously, the Maryland SHA has an existing short-term project development process that is focused on congestion relief projects. The details of this process are included in this report, but the high-level steps include 1. Diagnosis. This involves identification of the most unreliable segments of the highway system. SHA uses the planning time index (PTI) (95th percentile travel time) as the reliability perfor- mance measure. 2. Analysis. SHA uses an existing 20-year life-cycle BCA analysis for project prioritization. SHA adds 75% of the congestion-related savings as reliability savings to overall project benefits as the value of travel time reliability. 3. Selection. Based on this prioritized list, SHA works with various stakeholders to select proj- ects to program for design and construction. 4. Assessment. Postconstruction reliability improvements are assessed using the planning time index. Given that the data-driven methodology estimated a range of RR values that could be used to calculate reliability-based savings, a sensitivity analysis was conducted to determine the impact of a range of RR values on congestion relief project selection. This was accomplished by selecting a case study to document how congestion relief projects were prioritized on the Baltimore Beltway (I-695) in 2012. This short-term project improvement selection process focuses on low- cost solutions that exclude major roadway improvements, such as bridge widening and or anything requiring major right-of-way acquisition. A range of reliability ratios was applied to the BCA pro- cess used on the Baltimore Beltway to determine how congestion relief project prioritization might change based on changes in VTTR. It was determined that at low budget levels, the choice of RR can be an important factor in project prioritization. Note that the analysis results obtained from these short-term improvement projects are based on aggregate travel time savings. Therefore, to estimate the VTTR benefits, a constant factor of 0.75 was applied to the reported value of travel time (VOTT) savings. The reader should note that this is an approximation and effectively reflects the implicit assumption that all origin– destination (O-D) pairs affected by the proposed improvements have the same travel times and volumes in before/after scenarios. The research team acknowledges this significant assumption; however, in the absence of detailed O-D information for short-term improvement project analy- sis (and perhaps in similar practical decision-making scenarios), this exemplifies the versatility of the proposed reliability valuation method.

4In addition to short-term congestion relief project selection, the research team looked at the impact of incorporating a value of travel time reliability into long-term project prioritization and selection. This was accomplished using the Maryland Statewide Transportation Model (MSTM), a long-term travel demand model. In this case, disaggregate O-D information is used to estimate VOTT and VTTR savings. The results presented in this report should only be regarded as a proof of concept, as development of the base-year and future-year travel demand models is still in progress. However, this research demonstrates that incorporating travel time reliability valuation into a regional travel demand model can be relatively easy. Report the Step-by-Step Process Used by the Maryland SHA The high-level steps used to incorporate VTTR into the Maryland SHA project evaluation and decision process were as follows. Step 1: Document Existing Project Selection Process This step involved documenting the existing life-cycle BCA process for which VTTR was being used in consideration of prioritizing congestion relief projects for implementation. Step 2: Define Trips and Corridors to Be Analyzed This step involved selecting the routes and corridors connecting major O-D pairs for which a local value of reliability is desired. The selection should be done in conjunction with Step 3 to ensure that the required historical travel time data are available. Step 3: Acquire Data to Be Used for Analysis The Maryland SHA has access to link-based historical travel time data based on vehicle probes [both INRIX and the National Performance Measures Research Data Set (NPMRDS)] for all highways and major arterials. Many departments of transportation across the country are already using vehicle probe–based travel time data. Step 4: Calculate RR/VTTR The research team used the travel-time data-driven methodology for estimating value of reliabil- ity developed as part of this project for calculating a local reliability ratio and value of reliability. The methodology used is explained in Chapter 3 as well as in Part 2: Description of the Method. The MATLAB code used to automate this process is included in Appendix B. Step 5: Incorporate RR into the Existing Short-Term Congestion Relief Project Selection Process The local VTTR calculated using the travel-time data-driven methodology for estimating RR/VTTR was used to replace the current value in the baseline approach. The impact of replac- ing the RR currently used with a range of RRs was analyzed using projects selected in the past as a case study. Step 6: Incorporate RR into Long-Term Project Selection Process This was accomplished using the MSTM, a long-term travel demand model. The results are presented in Chapter 3 of this report, along with details of the process used.

5Step 7: Present to SHA Management Maryland SHA stakeholders were briefed on project progress throughout the conduct of the research project. The research team was led by a member of SHA’s Office of Planning and Pre- liminary Engineering. A presentation was prepared and made to upper management within SHA to gauge their reaction to the findings of this research. This presentation is summarized at the end of Chapter 3, and the presentation slides used are included in Appendix C. Conclusions and Recommendations An overall conclusion from this research suggests that agencies that do not account for VTTR in their BCA processes are undervaluing project benefits resulting from improvements to trip reli- ability. Valuation tools and techniques, both existing and newly developed as a result of this research, along with a significant body of literature, provide a basis for incorporating VTTR into an agency’s BCA process. While this research project focused on Maryland State Highway as a case study, the information (literature, data-driven methodology, application examples) documented in this report could help agencies looking to incorporate VTTR into their investment decision processes. Compared with the recent revealed and stated preference survey-based estimates in the litera- ture, the current RR ratio value of 0.75 used by SHA seems reasonable. Based on the develop- ment and application of the data-driven approach to reliability valuation methodology developed under this research, it can be concluded that, in Maryland, during peak hours in congested urban areas, the average RR ranges between 0.68 and 0.87, derived from MSTM and Census Bureau travel times, respectively (IndexMundi 2013; U.S. DOT 2013). In nonurban areas and at off-peak hours, the average RR can be taken as 0.52. Therefore, it seems the current value of 0.75 is reason- able when the reliability of commute travel times during peak hours in congested urban areas is considered. Note, however, that while this value appears reasonable based on the application of the newly developed data-driven reliability valuation methodology, the results obtained under this research do not necessarily validate this value because the data-driven valuation methodol- ogy itself must be validated. Future research identified in Chapter 4 of this report will facilitate methodology validation. The reader is also cautioned that this ratio can differ based on trans- portation facility type, mode, level of congestion, vehicle fleet composition, time of day, trip purpose, and so forth. Estimates of the value of reliability may be modified when these factors are taken into consideration. Given that the Maryland SHA is able to account for the benefit of project-related travel time reliability improvements, a potential next step is to incorporate the results of this project into a future iteration of the Maryland State Highway Mobility Report in the form of costs due to unreliability. Currently, the report includes performance measures based on congestion (travel time index) and reliability (planning time index). While the statewide cost of congestion is reported, an estimate of the additional cost users incur as a result of a lack of reliability in travel times, and as measured and reported using the planning time index, is not currently included. The VTTR estimates obtained from this research can now be used to bridge the gap in reporting costs of unreliability in the annual Mobility report. As noted above, Part 1 of this report can help agencies incorporate VTTR in their investment decision processes. Every effort has been made to fully document the data-driven valuation meth- odology developed under this research to facilitate its transferability to agencies beyond the Mary- land SHA. However, doing so at this time would likely require teaming with a university or consultant. A logical next step that would facilitate transferability among agencies, and overall ease of implementation, would be to develop (or build into an existing performance-measure calcula- tion and reporting tool) a software tool that can process the historical travel time data and estimate RR/VTTR using the methodology developed. In addition to this suggestion of follow-on work to facilitate the practical application of the results of this research, ideas for future research to build on and enhance the developed data-driven methodology are included in Chapter 4.