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.

6Background and application of the Method P a r t 1

7C h a P t e r 1 Introduction The topic of travel time reliability has been a significant focus in the transportation systems management and operations (TSM&O) community during recent years. With the end of the Strategic Highway Research Program (SHRP 2) Reliabil- ity research program in sight, agencies are working to figure out how to incorporate travel time reliability–related perfor- mance measures, analytical processes, and tools into their planning and programming processes. Travel time reliability describes the quality, consistency, timeliness, predictability, and dependability of travel. What is occurring today is a fun- damental 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 experiences of the Maryland State Highway Administration (SHA) and their ongoing efforts to include reliability into 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 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. Part 1 is organized as follows: • Chapter 1 provides a literature review of previous approaches to reliability valuation and focuses on whether or not, based on the existing literature, the use of 0.75 as a reliability ratio by SHA is defensible. • Chapter 2 describes the research approach. • Chapter 3 describes and presents the research findings and applications resulting from this project. • Chapter 4 provides conclusions and suggestions for future research. Previous approaches to reliability Valuation The literature review presented herein aims at using the results of various research studies conducted in the United States and elsewhere for both creating a benchmark for the data-driven approach and for reevaluating the current reliability ratio of 0.75 in use by SHA. First, various methods used in the litera- ture to determine the values of travel time (VOTT) reliability are summarized. Second, values of travel time reliability or reliability ratios (RRs) or ranges of ratios are summarized. Finally, putting the use of these research results into practice by local agencies is discussed and recommendations are made. In travel time reliability literature, two distinct approaches have been used to define travel time reliability for valuation purposes (Cambridge Systematics and ICF International, 2012): behavioral modeling approaches and an approach based on Real Options theory (a review of the literature on Real Options theory is included in Part 2: Description of the Method). With one exception, all studies in the reliability lit- erature used a behavioral approach in some form. The excep- tion, Evaluating Alternative Operations Strategies to Improve Travel Time Reliability, used an options-theoretic approach (SHRP 2 L11, 2012). The SHRP 2 L11 project was the first to use an options-theoretic approach for determining the value Background

8of travel time reliability by using speed and volume data as input. The options-theoretic approach introduced by the SHRP 2 L11 uses an analogy where premiums are set for an insurance policy on guaranteed speed levels. Specifically, the method calculates the dollar value of reliability by multiplying the certainty-equivalent penalty (measured in minutes-per- mile and obtained by applying the closed form Black-Scholes equation) by the value of time, thus it requires an estimation or adoption of VOTT as input. The SHRP 2 L11 study takes into account heterogeneity of the road users and different trip pur- poses by applying a separate value of time that corresponds to each user group. Use of an options-theoretic approach in transportation under SHRP 2 has led to significant discussion in the research arena by bringing a novel, data-driven approach to travel time reliability valuation. The discussions included some question- ing of the assumptions and methods used. The most signifi- cant question was with regard to the use of speed as a measure to set an insurance policy premium on guaranteed speed lev- els. The issue is, given speed is a measure that is not directly related to travel cost it cannot be discounted in the same way that financial analysts discount money. Another significant question relates to the assumption of the lognormal distribu- tion for speed variation; it does not address situations where speed/travel time is not distributed lognormally. Thus, the method used in SHRP 2 L11 is applicable only under a log- normal speed variation assumption. The research team con- ducting this project studied the questions resulting from the SHRP 2 L11 and attempted to clearly address these questions in its development of a new proposed data-driven methodology using an options-theoretic approach (see Part 2). Behavioral approaches followed two major paths: (1) statis- tical methods that directly estimate travel time distributions and variations, and (2) survey-based methods based on dis- aggregate data and discrete choice models. Among the two sta- tistical methods used to determine the VTTR, the first method uses a mean-variance approach which involves calculation of statistical measures to separate out the VOTT and VTTR. The second method is based on the schedule-delay concept, which focuses on the magnitude of the time encompassing both early and late arrivals in relation to a predetermined schedule. The mean-variance approach is easy to implement but has some theoretical drawbacks, since there is concern about dou- ble counting benefits. Double counting occurs if overall mean time is used to represent travel time (for the VOTT), since the mean time includes a portion of the variability component. The schedule-delay approach is conceptually more appealing, but it is more difficult to implement since it requires schedules of individual travelers and the distribution of their associated travel times. There are also methods that combine both mean- variance and schedule-delay methods, but they are more com- plicated to apply due to extensive data requirements that are not readily available. Survey-based methods, based on discrete choice models, typically use survey data in the form of stated preferences (SP) or revealed preferences (RP). Carrion and Levinson (2012) provides a comprehensive overview of the major behavioral approaches and evidence gathered over the years regarding the value of travel time reliability. Cirillo et al. (2014) provides a detailed review of behavioral approaches in the context of con- gestion pricing, including a systematic review of methodolo- gies, interpretations, findings and empirical applications on VOTT and VTTR estimations. After analyzing 14 congestion- pricing examples focusing on travel time reliability, they found that these two methods, survey-based and statistical, are the main research directions in the literature from a congestion- pricing context. Among the proposed survey-based methods, none of them were clearly superior to others. The analyses in the literature are often based on statistical methods and are based on the mean travel time and its variance while reliability is described using buffer indices and planning indices. How- ever, these studies usually involved complications due to the unknown theoretical distribution of travel time, which made comparisons of different studies impossible. For a meaningful universal comparison, the specific characteristics of the travel time distribution are needed. Much of the past research focuses on estimating VOTT rather than VTTR due to the complexity and difficulty of esti- mating VTTR (see Table 3 in Cirillo et al., 2014). As an alterna- tive, a typical approach is to use the reliability ratio (RR) (the ratio of VTTR divided by VOTT) as a convenient measure of travel time reliability for project evaluation purposes. An established RR along with knowledge of the VOTT simplifies the task of VTTR estimation. However, previous studies in the United States and elsewhere have shown that RR values vary significantly across different studies. Table 1.1 summa- rizes the average RR values and their ranges (minimum and maximum) found in previous studies. Note that the studies included in Table 1.1 are built on two previous studies: Carrion and Levinson (2012) and Cambridge Systematics and ICF International (2012). All of these studies in Table 1.1 used a survey-based behav- ioral approach, the majority of which are based on SP data or a combination of SP and RP data. There appears to be a lack of consistency in the values estimated, and average RR values vary significantly within and across studies from 0.1 to 2.51. The table shows that the most recent RR values, and 17 out of 25 average RR values, are higher than SHA’s current value of 0.75. It is worth noting that recent studies have used RP data. However, it should also be noted that RP and SP results are shown to differ significantly in the literature (Ghosh, 2001; Yan, 2002): RP estimates of VOTT and VTTR are almost double the median estimates of SP. Similarly, Shires and De Jong (2009) also showed that SP and joint SP and RP studies result in significantly lower VOTT savings. In addition to data

9sources (i.e., RP versus SP), these values show significant variation depending on the reliability measures used and modeling approach (e.g., heterogeneity, travel time unit, and choice dimensions considered). The most recent survey-based study to estimate social- economic values of travel time reliability was conducted by Significance et al. (2013) under the supervision of the KiM Netherlands Institute for Transport Policy Analysis for the Directorate-General of the Ministry of Infrastructure and the Environment. Previously, valuation of travel time reliability was determined based on the findings of an international expert meeting, organized by the Dutch Ministry of Public Works, Transport and Water Management. The Dutch values were last estimated in 1997 for passengers and in 2004 for freight trans- portation using major empirical research studies. The VOTT, VTTR, and RR values were updated annually in line with infla- tion and wage developments so that they could be used in benefit–cost analyses conducted for infrastructure projects. The Significance et al. (2013) study was the Netherlands’ first study to determine the social-economic values for travel time reliabil- ity based on empirical research (SP data). The data collection (SP) for passenger travel and transport was conducted in two steps: in the first survey, 240,000 partici- pants were recruited from the largest online panel (PanelClix) Table 1.1. Value of Reliability for Automobile Travel from Past Research No. Study Method Average RR Minimum Maximum Reliability Metric/Definition 1 Black and Towriss (1993) SP 0.55 — — Standard deviation 2 Senna (1994) SP 0.76 — — Standard deviation 3 Small et al. (1995) SP 2.30 1.31 3.29 Standard deviation 4 Koskenoja (1996) SP 0.75 0.33 1.08 Average schedule delay (late and early) 5 Small et al. (1999) SP 2.51 1.86 3.22 Standard deviation 6 Ghosh (2001) SP and RP 1.17 0.91 1.47 90–50 Percentile 7 Yan (2002) SP and RP 1.47 0.91 1.95 90–50 Percentile 8 Brownstone and Small (2005) SP and RP 1.18 — — 90–50 Percentile 9 Liu et al. (2004) RP 1.73 — — Median and the 80–50 percentile differences 10 Small et al. (2005) SP and RP 0.65 0.26 1.04 Ratio of standard deviation to mean 11 Tseng et al. (2005) SP 0.5 — — Scheduling approach; difference between early/ late arrival time and preferred arrival time 12 Bhat and Sardesai (2006) SP and RP 0.26 — — Scheduling approach; standard deviation 13 Hollander (2006) SP 0.10 — — Scheduling approach; mean-variance approach 14 Liu et al. (2007) RP 1.30 0.71 2.39 80–50 percentile 15 De Jong et al. (2007) SP 1.35 0.74 2.4 Standard deviation 16 Asensio and Matas (2008) SP 0.98 — — Scheduling approach; standard deviation 17 Borjesson (2009) SP 0.87 0.48 1.27 Ratio of sensitivity to standard deviation to sensitivity of the mean 18 Fosgerau and Karlström (2010) RP 1.0 — — Standard deviation 19 Tilahun and Levinson (2010) SP 0.89 — — Scheduling approach; difference between actual late arrival and usual travel time 20 Li et al. (2010) SP 0.70 0.08 1.59 Scheduling approach; standard deviation 21 Carrion and Levinson (2010) RP 0.91 0.47 1.20 90–50 percentile 22 Carrion and Levinson (2011) RP 0.91 0.69 1.12 Standard deviation 23 SHRP 2 C04 (2013a) RP 1.0 0.5 1.5 Standard deviation per unit distance 24 SHRP 2 L04 (2013b) RP 1.63 0.57 2.69 Standard deviation per unit distance 25 Significance et al. (2013) SP 0.6 0.4 1.1 Standard deviation Note: — = not reported; SP = stated preferences; RP = revealed preferences.

10 in the Netherlands, which led to 5,760 respondents. In the sec- ond survey, 1,430 respondents were recruited in the same manner as for the previous research study; namely, at petrol stations along the motorways, parking garages, train stations, tram and bus stops, airports (Schiphol and Eindhoven), and marinas (recreational navigation). For freight transport, face- to-face interviews were held with 812 respondents. The Significance et al. (2013) study determined VOTT, VTTR, and RR values both for passenger modes (including car, bus, tram, metro, train, airplane, and recreational naviga- tion) and freight modes (including road, rail, inland water- ways, sea, and air). The study is significant in the sense that the values of travel time for aviation (based on empirical research) and for recreational navigation were determined for the first time in the reliability literature. The new values are summa- rized in Table 1.2 (only passenger values are included in the table as other modes are not in the scope of this project). The Netherlands’ values in Table 1.2 are the result of the latest international work; however, other countries have also used either an estimate of their own or an adopted value for travel time reliability for benefit–cost analysis. The latest values estimated in the Netherlands and the values used by other countries are presented in Table 1.3. These values are compiled from various presentations from the International Meeting on Value of Travel Time Reliability and Cost-Benefit Analysis (15–16 October 2009, Vancouver, Canada). Table 1.3 also shows significant variation in RR values in different coun- tries as well. With the exception of the Netherlands’ updated values, they all are higher than SHA’s current 0.75 value, and even as high as 20 in France. The relatively low values of RR in the Netherlands is attributed to behavioral changes over time resulting from, for example, increased use of travel time by means of technological advances and methodological refine- ments in estimating these values. Given the significant variation in reliability ratios in the existing literature, the Maryland SHA and the research team chose an approach to estimate a new RR (or range of values) using available local travel time data. The proposed data- driven methodology using an options-theoretic approach developed under this project provides a VTTR for SHA based on readily available local travel time data. applying Vttr in Decision Making Prior to the SHRP 2 Reliability effort that started in 2009, no research existed for estimating reliability metrics based on the travel time distribution. These earlier works distinguished between recurring and nonrecurring delay (typically defined as incident delay), and then used nonrecurring delay as an indicator of reliability. Dowling developed a method for estimating recurring and nonrecurring delay for the California Department of Trans- portation (Caltrans) based on a probability tree to predict the expected number and duration of incidents (Dowling et al., 2004). The method is designed for application to a few selected facilities in a district and the results extrapolated to Table 1.2. Estimated VOTT, VTTR, and RR for Car Mode by Trip Purpose (in Euro/Hour per Person, Market Prices, Price Level 2010) Trip Purpose VOTT VTTR Reliability Ratio Home-to-work 9.25 3.75 0.4 Business 26.25 30.00 1.1 Other 7.50 4.75 0.6 Averagea 9.00 5.75 0.6 a Weighting is based on the division of the trip purposes in minutes traveled, derived from Onderzoek Verplaatsingen in Nederland (OViN) 2010. Table 1.3. RR Values for Cost–Benefit Analysis in Other Countries Country Reliability Ratio (RR) The Netherlands (Significance et al., 2013) 0.6 for auto and public transit (min 0.4, max 1.1) (old values 0.8–1.4 for personal auto and public transit, respectively) New Zealand (Taylor, 2009) 0.8 for personal autos Australia (Taylor, 2009) 1.3 for personal autos Sweden (Eliasson, 2009) 0.9 for all trip types Canada (Cambridge Systematics and ICF International, 2012) 1.0 for all trip types UK (Department for Transport, 2014) 0.8 for highways, 1.4 for transit France (Delache, 2009) 2 to 20 for auto, 6 for transit Japan (Fukuda, 2009) 0.966 for all trip types Note: The United Kingdom uses values estimated by the Netherlands, so these values may have been updated accordingly.

11 obtain district totals. The recurrent and nonrecurrent delays for each sample facility are computed for three prototypical days (weekday, weekend, holiday) in each of the four seasons of the year (winter, spring, summer, fall). The delays com- puted for each prototypical day are factored to seasonal totals according to the number of days that each day represents of each season. The seasonal totals are then summed to obtain annual totals. The method requires geometric data, demand data, collision history, frequency of maintenance and con- struction activities, frequency of inclement weather days, and frequency of special events. Default parameters and distribu- tions are provided for use when local data are not available. The University of Florida developed a series of simple pre- dictive equations for total travel time based on binary com- binations of conditions (present/not present) for congestion, incidents, weather, and work zones (University of Florida, 2007). The analyst estimates the probability of each combina- tion occurring, and a weighted total travel time is computed. This method is currently being adapted for statewide use by the Florida Department of Transportation. The University of Maryland, as part of the ongoing Coor- dinated Highways Action Response Team (CHART) evalua- tions conducted for the Maryland SHA, developed a predictive equation model based on running experiments with micro- scopic simulation. Cambridge Systematics also developed a set of predictive equations for predicting recurring- and incident- related delay using a stochastic approach that varied both inci- dent characteristics and demand levels. This procedure was adopted for use by FHWA’s Highway Economic Requirements System (HERS) model. This same approach was also used by Cambridge Systematics to develop incident delay relationships for the Intelligent Trans- portation Systems Deployment Analysis System (IDAS) model (FHWA, 2014). In both the HERS and IDAS models, recurring and incident delay are assigned monetary values. Recurring delay is valued at a rate established by U.S. DOT (Transporta- tion Economics.org, 2010). Incident delay is valued at twice that rate, but the basis for the valuation is from older studies prior to 1999 (Cohen and Southworth, 1999). For the Integrated Corridor Management program, Cam- bridge Systematics developed a scenario-based approach for use with microscopic simulation models, for analysis at a cor- ridor level (Cambridge Systematics, 2008). The scenarios are primarily based on combinations of demand level and inci- dent characteristics. Empirical data are used to estimate the probability of each scenario occurring, and the results of each simulation are combined via weighting. The Second Strategic Highway Research Program (SHRP 2) was authorized by the U.S. Congress to address the nation’s most pressing needs related to the highway system: safety, renewal, reliability, and capacity. The SHRP 2 Reliability focus area has been the driver of research in this area since the onset of the program and is the main focus of the remainder of this literature review. SHRP 2 reliability research has focused mostly on reducing congestion through incident reduction, management, response, and mitigation by developing basic analytical techniques, design procedures, and institutional approaches to address the events that make travel times un- reliable (TRB, 2014). Among more than 25 research projects under the Reliability focus area, only a few of them address the estimation of value of travel time reliability. There are also few research projects under the Capacity focus area that also address estimating value of reliability. One of the most comprehensive SHRP 2 projects that address inclusion of reliability in travel demand models is under the capacity program SHRP 2 C04, Improving Our Understanding of How Highway Congestion and Pricing Affect Travel Demand (SHRP 2 C04, 2013a). The SHRP 2 C04 project aimed to synthesize past research on understanding and predicting changes in travelers’ behavioral response to changes in traffic congestion and travel price. Their synthesis is used to (statistically) test selected behavioral hypotheses on suitable data obtained from around the United States. In addition, the research provided guidelines for incorporating developed functions into existing travel demand and network simulation models. Although the C04 research is under the SHRP 2 Capacity Research Program, it involves building math- ematical models of highway user behavioral responses to travel time reliability in addition to congestion and pricing. The val- ues of travel time and travel time reliability are considered among the factors that affect traveler demand and route choice behavior. Other factors considered include demographic char- acteristics, car occupancy, situational variability, and an observed toll aversion bias. The SHRP 2 C04 study estimates various highway utility functions and finally suggests the use of the function given below: Time 1 Cost (STD ) 1 2 3 2U a a D a D b I O c De f[ ] ( ) ( ) = ∆ + × × + × + × + × × + × where D = alternative-specific “bias” constant for tolled facilities; a1 = basic travel time coefficient, ideally estimated as a random coefficient to capture unobserved user heterogeneity; Time = average travel time; D = travel distance; a2, a3 = coefficients reflecting the impact of travel distance on the perception of travel time; b = auto cost coefficient; Cost = monetary cost including tolls, parking, and fuel; I = (household) income of the traveler; O = vehicle occupancy;

12 e, f = coefficients reflecting the impact of income and occupancy on the perception of cost respectively; STD = day-to-day standard deviation of the travel time; and c = coefficients reflecting the impact of travel time (un)reliability. The SHRP 2 C04 team tested various functional forms for representing the reliability effect, including standard devia- tion in day-to-day time, the difference between the 90th and 50th percentile times, and the difference between the 80th and 50th percentile times. The measure that produced the most consistent results was the standard deviation in travel time divided by journey distance. Thus, the suggested main measure of travel time reliability is specified as the day-to-day standard deviation of the travel time by auto, divided by distance. This measure has some advantages: (1) avoids the problem of hav- ing correlation between travel time, travel cost, and any travel reliability measure including standard deviation or buffer time, and (2) a plausible behavioral interpretation that travelers may perceive travel time variability as a relative (qualitative) measure rather than absolute (quantitative) measure. This form of highway utility function used in the SHRP 2 C04 project report allows for deriving VOTT and VTTR as follows: VOTT ( ) 11 2 3 2( ) ( )= × + × + × × ×a b a D a D I Oe f VTTR ( ) = × ×c b I O D e f VOTT can be derived as a function of travel distance, income, and car occupancy for each travel segment. Similar to VOTT, VTTR also is a function of travel distance, income, and car occupancy for each travel segment unless a more detailed explicit segmentation is applied. Note that VTTR is inversely proportional to distance. However, as the travel distance increases, travel time variations dampen in a relative sense. Finally, the reliability ratio was calculated as a measure of the relative importance of reduction of (un)reliability versus average travel time savings as follows: RR VTTR VOTT 1 11 2 3 2( )= = × + × + × × c a a D a D D The SHRP 2 C04 project estimated VTTR and VOTT simul- taneously using real-world data from actual traveler choices (RP data). The study results suggest that improvements in travel time reliability are at least as important as improve- ments in average travel time. The reliability ratio for auto travel is estimated to be between 0.7 and 1.5 for various model specifications, and it is following an increasing trend based on the results from other research. These results are in line with previous research results, most of which are based on SP stud- ies from Europe. Typical values for auto travel are in the same range, while values for rail and transit can go up to 2.5. The results obtained from the SHRP 2 C04 project are significant in the sense that they reflect the actual choices of users, while SP based study results may vary significantly, as the previously described Carrion and Levinson (2012) review study presents, depending on how the reliability concept is presented to respondents in the hypothetical scenarios. The SHRP 2 C04 results indicate that the traveler’s value of travel time and value of travel time reliability changes by origin–destination (O-D) trip distance as well. Travelers value savings on average or typical travel time more highly for lon- ger trips than for short trips, except for very long commuting trips (over 40 mi). The value of reliability also shows a relative damping effect for longer trips. The SHRP 2 C04 study results indicate that incorporation of the reliability models into travel demand forecast models will need further research, particularly regarding collection of actual O-D level travel time variability data. Also, the net- work simulation models need to be extended to incorporate travel time reliability in route choice and to generate O-D travel time distributions (“reliability skims”) instead of aver- age travel times. Because the study found the variation of VOTT and VTTR by trip distance, using different VOTT and VTTR by different trip types will be necessary instead of assuming a constant for a wide range of short and long trips as is pertinent to most travel models currently. SHRP 2 projects such as L04, Incorporating Reliability Per- formance Measures in Operations and Planning Modeling Tools, and C10, Partnership to Develop an Integrated, Advanced Travel Demand Model and a Fine-Grained, Time- Sensitive Network, aimed at closing these gaps. The methods developed in C04 can be applied for corridor-level or facility- level forecasts while research is still ongoing on the modeling side (SHRP 2 L04, 2013b, and SHRP 2 C10, 2010a). SHRP 2 C04 also suggests that some simplified proxy measures of reli- ability, such as perceived highway travel time by congestion levels, can be applied to the existing traditional (static) model structures. The perceived travel time concept uses the notion that highway users driving in congested conditions might per- ceive the longer travel time as an additional delay or penalty on top of free-flow (or some expected) time (SHRP 2 C04, 2013a). It can be represented by segmenting travel time coefficients by congestion levels in the highway utility function. This would result in a larger disutility associated with congestion. The per- ceived travel time concept provides an operational proxy for a reliability measure where obtaining an explicit reliability mea- sure is not feasible or possible. Perceptions of travel time by congestion levels can be obtained by traditional network simu- lation models. The required level-of-service (LOS) skims can

13 be generated by static assignment methods, while advanced methods such as Dynamic Traffic Assignment (DTA) can be more beneficial, or rather necessary, as stated in Chapter 3 of the SHRP 2 C04 draft report: It is important to note that making this approach operational within the framework of regional travel models requires explicitly deriving these measures from simulation of travel time distributions, as well as adopting assumptions regarding the ways in which travelers acquire information about the uncertain situation they are about to experience. DTA and traffic microsimulation tools are crucial for the application of models that include explicit travel time variability, since static assignment can only predict average travel times. The methodology presented in the SHRP 2 C04 project is sound but requires extensive survey and modeling work. Even applying the suggested proxy approach with traditional models would require significant effort while not necessarily providing the desired accuracy in measuring reliability. Therefore, it is not easily applicable for many agencies due to the required level of data and modeling efforts. SHRP 2 C11, Development of Tools for Assessing Wider Economic Benefits of Transportation (SHRP 2 C11, 2010b), can be thought of as a simpler solution to the issues presented thus far. The C11 project aims to help planners in conducting impact assessment of transportation capacity projects on con- ditions that directly affect wider economic benefits. In this project, a value of travel time reliability is not estimated but a range of values of reliability ratio obtained from the literature are used to demonstrate calculation of the economic benefit of travel time reliability savings. The default reliability ratio used in the tool was 0.8 for personal travel, based on SHRP 2 L04 report (2013b), and 1.16 for commercial travel. In SHRP 2 C11, four tools are developed that provide mea- suring of impacts on travel time reliability, market access, and intermodal connectivity. These three metrics are incorpo- rated in an accounting system of economic benefit and eco- nomic impact analyses. The economic benefit and impact analysis tool is freely available as a Microsoft Excel spread- sheet. The advantage of the tool is the simplicity of data requirements that can easily be collected or obtained. The tool can also be used in conjunction with travel models, land use models, or economic models, if desired. The Puget Sound Regional Council (PSRC) incorporated reliability directly in their travel demand model, using prin- ciples established in the SHRP 2 C11 project. This essentially amounts to a shifting of the speed-flow curves to the left, to account for the extra “impedance” caused by unreliable travel (i.e., nonrecurring congestion sources). The tool developed in the SHRP 2 C11 project can readily be used by many agencies for conducting impact assessment of transportation capacity projects considering reliability of travel time as well. However, the C11 project does not provide a method or tool to estimate value of reliability but requires using a value obtained from either the literature or survey data. The SHRP 2 L05 project, Incorporating Reliability Perfor- mance Measures into Transportation Planning and Program- ming Processes (SHRP 2 L05, 2013c), looked at using previous research in transportation planning and programming pro- cesses by providing agencies guidance in incorporating reli- ability into their planning and programming processes. The project produced three reports: (1) a guide, (2) a technical reference, and (3) a final report to guide agencies on incor- porating reliability into their transportation planning and programming processes. This project also did not include estimating value of reliability but rather focused on (1) mea- suring and tracking reliability performance, (2) incorporat- ing reliability in policy statements, (3) evaluating reliability needs and deficiencies, and (4) using reliability performance measurement to inform investment decisions. These four main steps are explained in detail in the guide. The technical reference provided detailed descriptions of available analytic tools. The final report summarized all the research conducted, including validation of case studies. In these case studies, the L05 project team used a reliability ratio range of 0.9 and 1.25. The SHRP 2 L05 project team also developed a spreadsheet and variants, which were used to support calculations that were used in the case studies. Summary and Conclusions The value of reliability is disaggregate in nature and varies across individual travelers, by trip purpose, by trip distance, by trip time of day, by mode, and by many other possible fac- tors. Using a reliability ratio without establishing empirical values from locally collected data implies that the value of reliability is a function of the value of average travel time and assumes the same for all travelers, trip purposes, time of day, and so forth. This is a strong assumption, and the use of a single value makes it even stronger. However, establishing a value for travel time reliability or a reliability ratio with widely used methods (i.e., survey-based behavioral methods) is expensive and time-consuming due to extensive data col- lection requirements. Since these VTTRs and RRs are built on survey data, it is also difficult and costly to update them or generalize them because they likely are not transferable. Moreover, they are not perfect, either; in addition to data- related issues, they are vulnerable to modeling assumptions, simplifications, and errors. As discussed in this chapter, reliability ratios that are found in the literature are very different and subject to the specific characteristics of each study. Therefore, using a single VTTR or RR will likely be misleading. A methodology to establish values of reliability that are generally accepted and applicable

14 with relative ease has yet to be developed. Therefore, it is rec- ommended that a range of values be used in the absence of empirical data and sources to estimate them. Based on the literature, the dispersion among RR estimates from stated preference surveys is considerably larger than the RR estimates from revealed preference surveys. The latest revealed preference survey reports an average RR estimate of 0.91 (Carrion and Levinson, 2012), while the most recent stated preference survey (Significance et al., 2013) reports an average 0.60 RR estimate for all highway trip purposes. Compared with the recent revealed and stated preference survey-based estimates in the literature, SHA’s current RR value of 0.75 seems reasonable and may even be, to some extent, conservative. For instance, according to Concas and Kolpakov (2009), VTTR varies between 80% and 100% of VOTT in ordinary/everyday conditions (no major constraints). They also claim that VTTR can be up to three times the VOTT in instances where nonflexible arrival/departure constraints exist. Therefore, the adopted RR estimate in Maryland needed further detailed analysis based on local conditions and avail- able data. As noted previously, the proposed data-driven meth- odology using an options-theoretic approach developed under this project provides a VTTR for SHA based on readily avail- able local travel time data. Incorporating reliability into decision making requires data on existing travel time reliability and a measure of reliability, forecasting the reliability level after a project or policy is implemented (thus a method for predicting future reliability), and monetary values of reliability disaggregated at the appro- priate level of detail. Most these requirements, particularly forecasting future reliability, need further research. Besides, most of the existing research has been mainly focused on pas- senger transport, and research is needed for other modes, especially for areas with multimodal networks and significant freight corridors such as Maryland (International Transport Forum, 2012).

15 C h a p t e r 2 Through the adoption of various measurement and reporting methodologies and tools, Maryland State Highway Adminis- tration (SHA) has been able to quantify current mobility and reliability conditions and trends on its highways. This provides a basis for examining how those variables change with the evolving transportation environment, and to assess how the agency’s actions can efficiently impact the users of the state’s transportation system. This also gives the Maryland SHA the ability to develop better informed decisions regarding the use of its limited resources, identify critical transportation issues before they develop into more serious problems, and provide measurement of its success. Describe Sha’s established processes The Maryland SHA has a life-cycle benefit–cost analysis (BCA) process in place to identify and prioritize improvements. The research team held multiple meetings with SHA planning staff to document SHA’s baseline process. The baseline process was documented in the context of recent project evaluations per- formed by the agency so that the existing project prioritization and selection process could be used as a case study. It should be noted that while many planning and project programming processes exist within SHA, the research team focused on the existing congestion relief project selection process as this is where SHA is already applying both a reliability-based perfor- mance measure and value of travel time reliability. The research team paid special attention to note how the value of travel time is already established in the baseline approach. Identify and acquire Data Needed to perform research SHA has procured INRIX-based vehicle probe data sets for the entire state, which provide speed information at 15-, 5-, and 1-minute intervals. This data set augments the real-time freeway data SHA already receives from INRIX through the Regional Integrated Transportation Information System (RITIS). RITIS is an automated data sharing, dissemination, and archiving system housed at the University of Maryland’s Center for Advanced Transportation Technology (CATT) Laboratory. SHA uses this archived data along with other data for performance measurement, congestion, and reliability analysis of its transportation infrastructure. Using INRIX-based vehicle probe data, SHA has developed congestion and reliability-related measures [travel time index (TTI) and planning time index (PTI)] on all the freeways and expressways in Maryland. From a congestion standpoint, two major measures of highway performance are: (1) percent sys- tem congested during peak hours; and (2) percent of vehicle miles traveled (VMT) in congested conditions during peak hours. Vehicles traveling at 70% of free-flow speed (equivalent to TTI of 1.3) on a freeway are considered to be experiencing congestion. Level of congestion varies from light to moderate to severe. Similarly, depending on the PTI, segments of freeways are considered as highly unreliable, moderately un reliable, and reliable. Findings of these analyses have been summarized in recent reports on the status of mobility in the state of Maryland (Maryland SHA, 2012). Identify Method to Forecast Future travel time reliability Measures The Maryland Statewide Transportation Model (MSTM) is a long-term travel demand model developed by the National Center for Smart Growth Research and Education (NCSG) at the University of Maryland (National Center for Smart Growth and Parsons Brinckerhoff, 2011). This model covers transpor- tation and land use activities at three distinct layers: national, statewide, and Metropolitan Planning Organizations (MPOs). Figure 2.1 illustrates the four-step modeling approach under- taken by MSTM to model person (outlined with red dots) and Research Approach

16 truck travel (outlined with blue dots). In the MSTM frame- work, it is possible to incorporate travel time variability mea- sures into the utility of mode and route choice alternatives between each origin–destination (O-D) pair. MSTM is a func- tional travel demand model currently used for a number of practices at SHA. In addition, NCSG researchers are currently enhancing MSTM by incorporating a Dynamic Traffic Assignment (DTA) capability. Chapter 3 identifies how MSTM was used to explore the impact of incorporating the value of travel time reliability into long-term project prioritization and selection decisions. Calculate a Local Value of travel time reliability Ultimately, the research team chose to focus on building off previous work on an options-theoretic approach to determine VOTT and VTTR analytically. The reliability ratio (RR) is a convenient way of estimating VTTR for project evaluation purposes. While Chapter 3 provides an overview of the pro- posed travel-time data-driven methodology for estimating value of reliability, Part 2 of the report provides an in-depth treatment of the methodology’s development, its assumptions, example application, and calculations, as well as how it tries to improve on the previous application of Real Options theory. The proposed method is data driven and requires access to fine granularity and long-term archived travel time data. This method is based on the analogy of an insurance policy designed to cover travelers against the negative impacts of unexpected variations in travel time. The proposed method has been designed to provide maximum flexibility for valuing travel time reliability based on existing local information and expe- riences. A review of the previous attempt to apply Real Options concepts to the problem of travel time reliability valuation is provided. Reasons as to why the previous attempts have received a cautious review are explained. Also, Part 2 sets out to unravel some of the less clear aspects of the previ- ous work by venturing further into the nuts and bolts of the approach and clearly identifying the distinctions between the proposed method and the earlier effort. Part 2 also includes a brief background on classical utility theory and its application in travel time reliability valuation. Strengths and limitations of utility-based estimation methods are discussed. A travel time insurance analogy is adopted to illustrate the different aspects of the proposed approach. Setting a premium on the proposed travel time insurance is presented and discussed in the context of options-theoretic valuation and asset pricing. Examples are provided throughout the technical report to facilitate the discussions and to demonstrate applica- tion of the concepts. Applications of the proposed methodology using a year’s worth of travel time data in the state of Maryland are reported. Analysis performed on the results of this applica- tion are presented and models to relate the travel time reliability ratio and average travel time (as well as 95th percentile travel time and average travel time) are calibrated. Finally, Part 2 includes two appendices. Appendix D pro- vides a brief review of stochastic processes, and in particular, the geometric Brownian motion (GBM) process including its properties and relationships with random walks. Appendix E presents more details about the application of the proposed Figure 2.1. Overview of MSTM—Phase 3. (II  internal to internal; EI  external to internal; IE  internal to external; EE  external to external.)

17 methodology to the ten directional corridor cases in Maryland and their various results. Incorporate Value of travel time reliability into project evaluation 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 life-cycle BCA base- line approach that was documented as previously noted focused on congestion relief projects prioritized for the Baltimore Beltway. Sensitivity of the baseline prioritization results to changes in VTTR was investigated. The VTTR that made pairs of projects comparable or resulted in a re-prioritization of projects was identified. Brief Sha Management on Methods Used to Select and Defend Local Value of Vttr and Impacts of application to existing Decision processes The Maryland SHA Office of Planning and Preliminary Engineering leadership and 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 Preliminary Engineering. A presen- tation was prepared and made to upper management within SHA to gauge their reaction to the findings of this research. The presentation used during this meeting is included in Appendix C, and the results of the meeting are presented in Chapter 3.

18 C h a p t e r 3 Overview of process Used to apply Value of travel time reliability in Maryland The high-level steps used to incorporate value of travel time reliability (VTTR) into the Maryland State Highway Admin- istration (SHA) project evaluation and decision process were as follows: Step 1: Document Existing Project Selection Process This step involved documenting the existing life-cycle benefit– cost analysis (BCA) process for which VTTR was being used in consideration of prioritizing congestion relief projects for implementation. Step 2: Define Trips/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 DOTs across the country are already using vehicle probe–based travel time data. Step 4: Calculate Reliability Ratio/Value of Reliability The research team used the travel-time data-driven method- ology for estimating value of reliability developed as part of this project for calculating a local reliability ratio and value of reliability. The methodology used is explained in this chapter as well as in Part 2. 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 replacing 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 Maryland Statewide Trans- portation Model (MSTM), a long-term travel demand model. The results are presented in this chapter, along with details of the process used. Step 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. The presentation used during this meeting is included in Appendix C and the results of the meet- ing are presented at the end of this chapter. Description of established processes The Maryland SHA project investment decision-making pro- cess is performed within an elaborate and complex frame- work that has been established over many years and involves Findings and Applications

19 the Maryland Department of Transportation (MDOT), local jurisdictions, and Metropolitan Planning Organizations (MPOs). Within the last 2–3 years, SHA, through the adop- tion of various measurement and reporting methodologies and tools, has been able to quantify current mobility conditions and trends on its highways including reliability performance measures. This provides a basis for examining how mobility conditions change with the evolving transportation environ- ment, and to assess how the agency’s actions can efficiently impact the users of the state’s transportation system. What follows in this section is a description of SHA’s short- term congestion relief project prioritization and decision- making process, because it is this process in which SHA currently uses a value of travel time reliability. However, this short-term congestion relief process falls within a much larger decision-making framework and there are many other specific programming decision processes within this framework. For an overview of the larger Maryland Department of Trans- portation’s investment decision process followed by SHA’s high-level investment decision process, the reader is referred to Appendix A. Appendix A also includes some detail regard- ing other specific programming decision processes internal to SHA. Description of reliability in Congestion relief project Decision Making From a reliability perspective, SHA has made significant inroads with incorporating reliability within short-term improvement studies to identify priority congestion relief projects. The four- step process that will be described (see Figure 3.1) for making investment decisions in these congestion relief projects is rela- tively new and incorporates an adopted reliability measure, value of time, and value of travel time reliability. Step 1: Diagnose Problems Including Highly Unreliable Segments/Corridors In 2012, the Maryland SHA published its first of what has become an annual mobility report. This annual mobility report is an important document that helps SHA decision makers identify problematic state roadways where short-term conges- tion relief project investments should be made and reliability is a key component of the problem diagnosis. Significantly, reli- ability is becoming ingrained in SHA transportation policy as evidenced by this excerpt from the foreword of the 2013 Mary- land State Highway Mobility Report as written by the SHA Administrator, Melinda B. Peters (Mahapatra et al., 2013): In addition to safety and congestion, transportation system reli- ability is another key factor to providing our customers with a good travel experience. The congestion and reliability measures reported in the Maryland Mobility Report include travel time index (TTI) and planning time index (PTI). For PTI, 95th percentile reliable travel time is the selected measure that is calculated for SHA roadways and reported. These measures, TTI and PTI, were selected because they are easily computed from speed data and are relatively easy to communicate to a broad range of audi- ences. Speed data comes from a private company providing both real-time and historic traffic speed data collected from an estimated 100 million vehicles nationwide, including com- mercial vehicle fleets. Note that this is the same data source that is used in the travel-time data-driven methodology for estimating value of reliability developed as part of this project for calculating a local reliability ratio and value of reliability. For the purposes of reporting PTI, the Maryland SHA has categorized the reliability-based value of PTI as follows: • Reliable (PTI < 1.5) • Moderately Unreliable (1.5 < PTI < 2.5) • Highly Unreliable (PTI > 2.5) This categorization was closely coordinated with the Wash- ington and Baltimore Metropolitan Planning Organizations (MPOs) to ensure regional consistency in definition and report- ing. Analysis and reporting of congestion and reliability mea- sures is done by (1) entire state network, (2) major geographic regions, and (3) regionally significant corridors in the morning and evening peak hours. In addition, the Maryland SHA reports Figure 3.1. Four-step process described for making investment decisions.

20 on the extent of reliability by reporting, for example, the per- cent of peak hour vehicle miles traveled (VMT) experiencing unreliable (PTI > 1.5) conditions. Figure 3.2 shows an example of how SHA reports on reliability with a map focused on the Baltimore-Washington region. In this region, 19% of the morn- ing peak hour VMT experiences unreliable conditions. The executive summary of both the 2012 and 2013 Mary- land SHA Mobility Reports provides a summary of the top five most unreliable segments, as measured by PTI, in the morn- ing and evening peak hours. In 2012, three of the top 10 most unreliable segments were on the Baltimore Beltway (I-695) as were three of the top 10 most congested segments. Based on these findings, the Baltimore Beltway was targeted for identify- ing and prioritizing congestion relief projects. Step 2: Identify Congestion Relief Alternatives and Prioritize Using Benefit–Cost Analysis In this step, ongoing studies and projects already in the plan- ning or design phase are identified for the targeted facility. In 2012, using the Baltimore Beltway (I-695) as an example, there were 10–15 projects in various stages of planning and design. In an effort to refine targeting of problem locations, input is gathered through field observations as well as input from regional planning personnel, Office of Highway Design, Dis- trict personnel, Office of Construction, Office of Traffic and Safety, and the Office of Planning and Preliminary Engineer- ing. A traffic simulation model (VISSIM) is used to support the project sequencing evaluation process for improvements to the roadway and to summarize the results of the analysis and provide prioritization for projects. Proposed projects are low-cost solutions that exclude any major roadway improve- ments, such as bridge widening or anything requiring major right-of-way acquisition. Proposed projects also take into account any projects that already are in the planning and design phases. The I-695 study area included the southwest, northwest, and northeast segments as shown in Figure 3.3. Data used in the VISSIM analysis include morning and evening peak hour volumes (including ramp turning movements), the number Figure 3.2. SHA reports on reliability with a map focused on the Baltimore-Washington region.

21 of lanes that service the volume, percent trucks, and speeds based on vehicle probe data. The VISSIM models are calibrated using the following crite- ria for each roadway segment along I-695 between interchanges as follows: • Traffic volumes must be within 10% of the input volume. • Auto speeds must be ±5 mph of the vehicle probe data speed. • Auto travel times must be within 10% of the vehicle probe data travel time. In order to calibrate the model, adjustments are made to driver behavior including lane change parameters, headways, and desired speed decisions. Models are run five times and data are averaged for the combined runs. The majority of proposed improvements for I-695 pro- vide auxiliary lanes between interchanges or the extension of acceleration lanes. Proposed improvements are run through a benefit–cost analysis. The speeds and travel times from VISSIM are compared between existing conditions and proposed improvements. In the I-695 study, the following assumptions were made in calculating benefits and costs, respectively: • Benefits 44 Three hours of both the AM and PM peaks are considered 44 250 working days per year 44 20 year time horizon 44 10% trucks 44 Auto congestion cost: $25.68/hour (2010) 44 Truck congestion cost: $66.08/hour (2010) 44 1.2 average vehicle occupancy 44 Fuel cost estimated to be 10% of delay savings 44 75% of delay savings as reliability savings (0.75 reliability ratio) 44 Safety benefits made using crash modification factors and year 2011 crash data • Costs 44 Major quantities and unit pricing developed in accor- dance with 2010 SHA Highway Construction Cost Esti- mating Manual. Figure 3.3. The I-695 study area included the southwest, northwest, and northeast segments.

22 44 Bridge widening that is necessary (outside of restrip- ing alternatives) is by separate preceding contract unless otherwise noted. 44 Retaining walls and concrete traffic barrier are assumed to stay within right-of-way within developed or known environmentally sensitive areas. 44 An 800-foot length of grinding/resurfacing is assumed on each end of each alternative for maintenance of traf- fic (MOT) traffic shifts. 44 Pavement section consists of 2-in. surface course, 15-in. base course, and 2 courses of 6-in. graded aggregate base. 44 Ground mount signing and pavement markings estimated on cost-per-mile basis. 44 Utility relocation estimated at 8% of neat construction cost. 44 A 35% contingency and 15.3% overhead factor was applied to each alternative estimate. SHA Business Plan objectives require at least a 5% reduc- tion in delay due to the implementation of its congestion relief projects. Note that SHA’s BCA process uses travel time savings (both savings in average and reliability) as part of the benefits calculation. Average travel time savings are calculated using traffic volume affected, average travel time improvement, and value of travel time (VOTT). Following recent trends of other transportation agency practices, particularly in Europe (where reliability benefits are accounted for as a percent of congestion reduction–related savings), SHA includes 75% of the congestion-related savings as reliability-related savings to project benefits. Note that the literature search performed on relevant national and international studies as well as the options-theoretic analysis on Maryland travel time data point to the fact that the current value (75%) is well within the range of viable values for the state of Maryland. Table 3.1 summa- rizes the latest basic parameters used in SHA’s BCA process to estimate monetary value of travel time savings, travel time reliability savings, and fuel cost savings. In the baseline approach, the value of travel time for auto- mobile passengers, truck drivers, and freight cargo are declared. These values are based on a series of studies that are primarily sponsored under SHA’s CHART program to evaluate eco- nomic value of its incident management initiatives (Chang, 2011). In the Chang 2011 study of CHART incident management program benefits, which reports on 2011 values, the passen- ger unit value of time is based on U.S. Census Bureau data (IndexMundi 2013; U.S. DOT 2013). A truck driver’s value of time is based on information from the Bureau of Labor Sta- tistics, the U.S. DOT, and FHWA’s Highway Economic Requirements System (HERS) (FHWA, 2013). Similarly, the cargo value of time is based on a study by the Texas Transpor- tation Institute, a study by Levinson and Smalkoski (2003), and a study by De Jong (2000). Step 3: Congestion Relief Project Selection The output of the previous step provides a list of potential con- gestion relief projects along with their associated benefit–cost ratios. Table 3.2 below is an example of a subset of improve- ment projects that were developed for I-695 (note that a total of 16 projects were identified in the study area). Recommenda- tions for project selection are made by the study analysis team. Final selection of projects is made by SHA leadership after meetings are held with various stakeholders, such as MDOT, the MPO, FHWA, and the district offices. Ultimately, projects selected are based on both quantitative and qualitative input as well as available budget. They are then programmed and moved forward into the design phase. Step 4: Post-Congestion Relief Project Implementation Assessment After completion of the project, an impact assessment on con- gestion and reliability resulting from implementation is made. This is usually done four to six months after the project has opened in order to allow traffic to adjust to the new patterns. Maryland uses congestion and reliability measures in project- specific impact assessment as well as in annual corridor assess- ments made as part of their Mobility report development and reporting process. An example of assessing a major capacity improvement project, Maryland Route 200, which is commonly known as the Intercounty Connector (ICC), was analyzed to determine its postconstruction impacts on congestion and reliability. Maryland Route 200 is a six-lane electronic toll facil- ity connecting Interstates 270 and 95 in the Washington, D.C. metropolitan area. The analysis found that although the metro- politan area generally experienced better traffic conditions in Table 3.1. Parameters Used by SHA in Project Benefit Estimation (2012 Values) Saving Type Parameter Unit Categories SHA Value Travel time VOTT $/h Passenger 29.82 Truck driver 20.21 Cargo 45.40 Travel time reliability VTTR $/h Passenger 22.36 Truck driver 15.16 Cargo 34.05 Fuel cost na $/gal Gasoline 3.69 Diesel 3.97 Note: na = not applicable.

23 2012 (after) than before (2010), the area in the vicinity of the ICC experienced greater magnitude improvements than did the region overall, by a margin of 3–4 percentage points, which is an indication of the ICC net effect (Pu et al., 2013). The analysis looked at the spatial extent of congestion, intensity of conges- tion, and reliability of travel both before and after the ICC in the morning and afternoon peak hours. Travel time reliability in the ICC study area, as measured by the 95th percentile travel time– based PTI, improved significantly after the ICC was constructed. As Figure 3.4 shows, in the AM peak hour, the ICC study area average PTI was 2.11 in 2010, and decreased to 1.85 in 2012, an 11% drop. In the PM peak hour, the PTI went from 2.04 in 2010 to 1.82 in 2012, an 11% drop. Referring back to the Baltimore Beltway (I-695) example, the congestion relief projects selected to move forward to design have not yet been constructed. The Maryland SHA does, how- ever, continue to monitor I-695 congestion and reliability per- formance overall as shown in the most recent mobility report (2013). The Baltimore Beltway is one of many regionally significant corridors that is measured annually in terms of con- gestion and reliability performance (see Figure 3.5). proposed travel-time Data-Driven Methodology for estimating Value of reliability/reliability ratio As described in the previous section, SHA is using planning time index (PTI) to measure travel time reliability on high- way facilities. The Maryland SHA has also adopted a 0.75 reliability ratio (RR) to measure the economic benefits of improvements in travel time reliability when conducting benefit–cost analysis of congestion relief projects. Conven- tionally, RR is defined as the ratio of value of travel time reli- ability (VTTR) and value of travel time (VOTT). This value was adopted by the Maryland SHA based on a comprehen- sive literature search of existing national and international resources as well as existing federal recommendations for a Table 3.2. Subset of Improvement Projects That Were Developed for I-695 Location Project Description Total Savings ($, 103) Construction Cost ($, 103) O&M Cost ($, 103) Total Cost ($, 103) Benefit/Cost (%) I-695 outer loop: US 40 (Baltimore National Pike) Interchange Extend outer loop auxiliary lane prior to interchange to connect to deceleration lane to eastbound US 40. Widen I-695 outer loop to provide exclusive decel- eration lane for westbound US 40. Total project length is 2,200 ft. $32,894 $5,000 $500 $5,500 598 I-695 inner loop: MD 147 ( Harford Road) Inter- change Remove eastbound I-695 to northbound (NB) MD 147 (Harford Road) ramp and replace with Signalized Spur off of eastbound I-695 to southbound (SB) MD 147 (Harford Road) ramp. $9,117 $2,368 $237 $2,605 350 I-695 inner loop: MD 26 (Liberty Road) to I-795 (Northwest Expressway) Provide 3 through lanes and 2 auxiliary lanes from eastbound MD 26 ramp to inner loop I-695 continuing to existing auxiliary lanes for northbound I-795 ramp. Project will require restriping and constructing new pavement and placement of a retaining wall. Total project length is 2,750 ft. $30,702 $9,900 $990 $10,890 282 I-695 inner loop: MD 542 (Loch Raven Boulevard) to MD 41 (Perring Parkway) Extend MD 542 (Loch Raven Boulevard) northbound to I-695 inner loop accel- eration lane to bridge over East Joppa Road. Project includes milling and overlay for restriping and widening of I-695. Total project length is 3,000 ft. $17,801 $5,900 $590 $6,490 274 I-695 outer loop: I-83 (Jones Falls Expressway) to Stevenson Road Provide additional through lane from I-83 (Jones Falls Expressway) ramp to outer loop I-695 continuing to Stevenson Road off ramp. Project will involve mill and overlay to facilitate restriping exist- ing pavement. Total project length is approximately 2 miles. $15,177.26 $5,400 $540 $5,940 256

24 RR value. One of the objectives of this research was to develop a methodology to defend this number or provide a basis for changing it based on local data. A travel-time data-driven methodology is proposed for estimating a reliability ratio (RR) and ultimately a value of travel time reliability (VTTR). The methodology has been implemented in MATLAB to automate the process (the MATLAB code is provided in Appendix B). An entire year’s worth of archived probe-based travel time data was used to estimate the local RR and VTTR values on five different cor- ridors in Maryland. What follows is an overview of the proposed methodology to value travel time reliability which is based on Real Options theory. A detailed in-depth treatment of the methodology’s development, its assumptions, example application, and calcu- lations, as well as how it differs from and builds on the previous application of Real Options theory, is provided in Part 2. The proposed method is based on an analogy of a travel time insurance policy. The method requires historical travel time data over an extended period of time as input and per- forms the necessary analysis to identify the nature and size of travel time variations that are experienced by travelers. Once the stochastic nature of variations in travel time is identified, it can be used to build a projected probability den- sity function of travel time realizations over an extended period given prevailing infrastructure and traffic conditions. Figure 3.4. ICC study area average PTI for AM and PM peak hours. (Before  2010; after  2012.) Figure 3.5. The Baltimore Beltway is measured annually in terms of congestion and reliability performance.

25 Travelers are assumed to incur penalties associated with arriving earlier or later than their planned arrival times at their destination. In the proposed method, these penalties are defined as a fixed portion of the amount of time by which the traveler is early or late relative to their planned arrival time. The estimated penalties are used to evaluate the certainty- equivalent insurance policy that will offer the traveler equal coverage against expected future penalties. Note that in characterizing the valuation method the fol- lowing questions need to be answered: 1. How can travel time evolutions over time be modeled? 2. How can a penalty/reward (payoff) of early/late arrivals at the destination be determined? 3. What is the guaranteed level of travel time? 4. What is the duration of time for which the travel time insurance policy is issued? 5. How do the future payoffs get valued at the outset of the trip? Figure 3.6 illustrates the above-mentioned components of an options-theoretic valuation method. Note that this is a generic graphic. The methodology is fully described by speci- fying each component of the method in Part 2. In essence, the following set of responses to the corresponding set of ques- tions above provides a high-level description of the proposed methodology: 1. Travel time series can be characterized as geometric Brownian motion (GBM) with drift stochastic process; hence, given the process parameters, future travel time probability distributions can be specified. 2. Penalty is simply defined as an asymmetric bilinear func- tion of the amount of time by which the traveler is late or early at the destination. 3. Expected travel time is taken as the guaranteed travel time level. 4. Travel time insurance policy is issued for the longest trip time possible under recurrent congestion scenarios (95th percentile travel time is used for this purpose). 5. A certainty-equivalent payoff valuation strategy is adopted. This payoff valuation method takes advantage of the GBM assumption for the travel time process to greatly simplify the insurance valuation process. The results of applying the methodology indicate that SHA’s use of the current RR of 0.75 is conservative for commute trips. According to U.S. Census Bureau statistics, average commute trips in Maryland during the 5-year period (2006–2010) have been approximately 31 minutes long (IndexMundi 2013; U.S. DOT 2013). However, the corresponding RR value (0.87) is believed to be at the upper range of values for travel time reli- ability. Further analysis was conducted to justify any decision to increase the current value of travel time reliability. Maryland Statewide Transportation Model (MSTM) long- term demand and travel time estimates are used in aggregating the results for all origin–destination (O-D) pairs in the state. Based on MSTM, for all current trip purposes, an average reli- ability ratio value of 0.52 is obtained. This value is expected to increase to 0.55 over the next 15 years until 2030. Similarly, the current average reliability ratio for commute trips in Maryland is estimated to be 0.68 and would remain relatively unchanged until 2030. However, it should be noted that in comparison with U.S. Census Bureau estimates, MSTM travel times are on Figure 3.6. Various components of a travel time insurance pricing method. Time Travel Time Guaranteed Travel Time Policy Duration ETA Penalty/ Claim Travel Time PDF Time 1 2 3 4

26 average about 6 minutes smaller. Note that due to bias in self- reporting, Census Bureau estimates tend to be an overestimate. At the same time, it may be argued that MSTM travel times are underestimates caused by spatial aggregations in zone defini- tions as well as the use of long-term performance functions. In summary, it can be concluded that, during peak hours in congested urban areas, the average reliability ratio ranges between 0.68 and 0.87 derived from MSTM and Census Bureau travel times, respectively. In nonurban areas and at off-peak hours, the average reliability ratio can be taken as 0.52. Therefore, it seems the current value (0.75) is reasonable when the reliability of commute travel times during peak hours in congested urban areas is concerned. Incorporating results into Short-term prioritization and project Selection In order to incorporate the findings of this study into the short-term prioritization and project selection process at the Maryland SHA, improvement projects on I-695 (Baltimore Beltway) were selected as a case study. All proposed congestion relief projects are low-cost solutions that exclude any major roadway improvements, such as bridge widening and major right-of-way acquisition. Projects were analyzed using VISSIM. The resultant travel time and reliability savings as well as cor- responding project costs are used to rank each project. This study includes I-695 between MD 43 in White Marsh and I-95 in Arbutus and will be expanded in the future to include the remainder of the Beltway, which includes the entire east side. The I-695 study area includes the entire Bal- timore Beltway in Baltimore County, Anne Arundel County, and Baltimore City. For analysis purposes, I-695 was divided into the following segments as shown in Figures 3.3 and 3.7: • Northeast—from I-83 (Harrisburg Expressway) to MD 43 (White Marsh Boulevard) • Northwest—from I-70 to I-83 (Harrisburg Expressway) • Southwest—from I-95 (Arbutus) to I-70 Existing AM and PM peak hour volumes were developed for the study area using information provided by the High- way Information Services Division (HISD) website as well as the O-D study conducted for the I-695 inner loop weave from Figure 3.7. Baltimore Beltway (I-695) study area.

27 northbound MD 41 (Perring Parkway) to MD 43 eastbound. The volumes include the turning movements at ramp ter- mini. The truck percentage throughout the study area varies between 5% and 12% for both peak hours. The models were created using VISSIM 5.3-09 for both the AM and PM peak hours. In order to minimize the effort in the calibration process, signalized intersections were excluded from the models. Calibration criteria for each roadway seg- ment along I-695 between interchanges are as follows: • Traffic Volumes must be within 10% of the input volume. • Auto Speeds ±5 MPH of the INRIX speed. • Auto Travel Times must be within 10% of the INRIX travel time. All models were calibrated within the targeted ranges. In order to calibrate the model, adjustments were made to driver behavior including lane change parameters, headways, and desired speed decisions. Most modifications were made at heavy merge and weave areas. Seeding times varied between 15 minutes and 1 hour depending on the congestion level of the roadway. Models were run five times and data was aver- aged for the combined runs. Improvement projects Several improvements were proposed for the I-695 corridor. These improvements do not include any bridge widening other than those bridge widening projects that are already funded for construction. Most improvements provide auxil- iary lanes between interchanges or the extension of accelera- tion lanes. Tables 3.3 through 3.5 provide a complete list of proposed improvements in each quadrant of the Beltway. The resultant speeds and travel times obtained from VISSIM models were compared between the existing condi- tions and proposed improvements. Benefit-to-cost comparison Table 3.3. Proposed Improvement Projects in the Southwest Quadrant of the Baltimore Beltway Project Code Location Improvement Description SW1 I-695 outer loop: MD 144 (Edmonson Avenue) on ramp continuing to MD 372 (Wilkens Avenue) Provide additional through lane from on ramp at Edmonson Avenue to end of accel- eration lane from Edmonson Avenue. Project includes widening and restriping of I-695 outer loop and removal and placement of retaining wall. Total project length is 2,500 ft. SW2 I-695 inner loop: US 40 (Baltimore National Pike) Interchange Extend inner loop auxiliary lane prior to interchange to connect to deceleration lane to westbound US 40. Widen I-695 inner loop to provide exclusive deceleration lane for eastbound US 40. Includes construction of retaining wall. Total project length is 2,200 ft. SW3 I-695 outer loop: US 40 (Baltimore National Pike) Interchange Extend outer loop auxiliary lane prior to interchange to connect to deceleration lane to eastbound US 40. Widen I-695 outer loop to provide exclusive deceleration lane for westbound US 40. Total project length is 2,200 ft. SW4 I-695 inner loop: I-70/MD 122 ( Security Boulevard) to Windsor Mill Road Extend I-70 WB to I-695 NB acceleration lane by 500 ft. Extend MD 122 to I-695 NB acceleration lane by 1,250 ft. Project will require restriping of I-695, widening to accommodate acceleration lane and construction of a retaining wall. Table 3.4. Proposed Improvement Projects in the Northwest Quadrant of the Baltimore Beltway Project Code Location Improvement Description NW1 I-695 inner loop: MD 26 (Liberty Road) to I-795 (Northwest Expressway) Provide 3 through lanes and 2 auxiliary lanes from eastbound MD 26 ramp to inner loop I-695 continuing to existing auxiliary lanes for northbound I-795 ramp. Project will require restriping and constructing new pavement and placement of a retaining wall. Total project length is 2,750 ft. NW2 I-695 outer loop: I-795 (Northwest Expressway) to MD 26 (Liberty Road) Provide auxiliary lane from I-795 (Northwest Expressway) Ramp to outer loop I-695 continuing to MD 26 (Liberty Road) off ramp. Project will include restriping, widen- ing, construction of retaining wall, and placement of W-beam traffic barrier. Total project length is 3,800 ft. NW3 I-695 outer loop: I-83 (Jones Falls Expressway) to Stevenson Road Provide additional through lane from I-83 (Jones Falls Expressway) ramp to outer loop I-695 continuing to Stevenson Road off ramp. Project will involve mill and overlay to facilitate restriping existing pavement. Total project length is approximately 2 miles. NW4 I-695 inner loop: Stevenson Road to I-83 (Jones Falls Expressway) Provide additional through lane from Stevenson Road to auxiliary lane for southbound I-83 (Jones Falls Expressway). Project will involve mill and overlay to facilitate the restriping for the additional lane. Total project length is approximately 7,900 ft.

28 Table 3.5. Proposed Improvement Projects in the Northeast Quadrant of the Baltimore Beltway Project Code Location Improvement Description NE1 I-695 inner loop: MD 139 (Charles Street) to MD 146 (Dulaney Valley Road) Provide auxiliary lane from West Road exit to northbound MD 146 (Dulaney Valley Road) exit. Project includes widening for 500-ft deceleration lane at West Road exit. Project will also require milling and overlay for restriping and construction of retaining wall. Total project length is 5,200 ft. NE2 I-695 inner loop: MD 146 (Dulaney Valley Road) to Providence Road Provide auxiliary lane from MD 146 (Dulaney Valley Road) northbound off ramp to Providence Road underpass. Includes mill and overlay for restriping, I-695 inner loop widening, and placement of W-beam traffic barrier. Total project length is 6,300 ft. NE3 I-695 outer loop: MD 542 (Loch Raven Boulevard) to Providence Road Provide additional through lane from on ramp MD 542 (Loch Raven Boulevard) to outer loop I-695 continuing to Providence Road off ramp. Includes mill and overlay for restriping, I-695 outer loop widening, and placement of noise barrier and W-beam traffic barrier. Total project length is 3,700 ft. NE4 I-695 outer loop: Providence Road to MD 146 (Dulaney Valley Road) Provide auxiliary lane from Providence Road to Dulaney Valley Road off ramp. Includes mill and overlay for restriping, I-695 outer loop widening, and placement of noise barrier and W-beam traffic barrier. Total project length is 5,200 ft. NE5 I-695 inner loop: MD 542 (Loch Raven Boulevard) to MD 41 (Perring Parkway) Extend MD 542 (Loch Raven Boulevard) northbound to I-695 inner loop acceleration lane to bridge over East Joppa Road. Project includes milling and overlay for restriping and widening of I-695. Total project length is 3,000 ft. NE6 I-695 inner loop: MD 41 (Perring Parkway) to MD 147 (Harford Road) Provide auxiliary lane from MD 41 (Perring Parkway) northbound ramp to inner loop I-695 continuing to and terminating at off ramp at MD 147 (Harford Road) southbound. Total project length is 3,900 ft. NE7 I-695 outer loop: MD 147 (Harford Road) to MD 41 (Perring Parkway) Provide auxiliary lane from southbound MD 147 (Harford Road) ramp to outer loop of I-695 continuing to off ramp for northbound MD 41 (Perring Parkway). Total project length is 3,900 ft. NE8 I-695 inner loop: MD 147 (Harford Road) Interchange Remove eastbound I-695 to northbound MD 147 (Harford Road) ramp and replace with signalized spur off of eastbound I-695 to southbound MD 147 (Harford Road) ramp. was developed, and the results are shown in Table 3.6. The following assumptions were made in the development of user savings under the current process: • Three hours of AM peak and three hours of PM peak considered • 250 working days per year • 20 years • Assume 10% trucks • Auto congestion cost: $25.68/hour • Truck congestion cost/hour: $66.08/hour • Assume 1.2 average vehicle occupancy • Fuel cost savings is assumed to be 10% of delay savings • Assume 75% of delay savings as reliability savings (non- recurrent savings) • Safety benefit using crash modification factors and year 2011 crash data Major quantity estimates have been developed for each primary and long-term auxiliary lane alternatives using the following nine assumptions: 1. Measurements have been taken from base mapping, when available. When such base mapping was not available, measurements and cut heights were estimated using Google Map. Significant embankment and retaining wall heights within fill conditions were visually estimated by field visits as necessary. 2. Major quantities and unit pricing were developed in accor- dance with the 2010 SHA Highway Construction Cost Estimating Manual as practical. Major quantities percent- ages were supplied for Categories 1, 3, and 7 for the appro- priate pavement type (restriping or pavement widening). 3. Bridge widening that is necessary (outside of restriping alternatives) is by separate preceding contract unless other wise noted. 4. Estimates are for construction costs only. Retaining walls and concrete traffic barrier are assumed as noted to stay within right-of-way within developed or known environ- mentally sensitive areas, as well as to avoid impacts to noise walls. Right-of-way costs for environmental mitigation may be significant and should be estimated separately during preliminary design. 5. Except where otherwise noted, an 800-foot length of grinding/resurfacing is assumed on each end of each alter- native for MOT traffic shifts. In remaining instances, MOT shifts on entire lengths of approach curves were assumed as noted.

29 Table 3.6. Improvement Projects Benefit–Cost Analysis Under Current Value of Reliability (RR  0.75) Project Code Vehicle Minutes Saved Peak Period Savings (h, 103) Auto Cost Savings ($, 103) Freight Cost Savings ($, 103) Delay Cost Savings ($, 103) Fuel Cost Savings ($, 103) Reliability Savings ($, 103) Safety Savings ($, 103) Total Savings ($, 103) Construction Cost ($, 103) O&M Cost ($, 103) Total Cost ($, 103) Benefit/ Cost (%) Rank SW1 AM PEAK 1,542 386 10,692 2,547 13,239 1,324 9,929 989 27,164 16,500 1,650 18,150 150 10 PM PEAK 106 27 735 175 910 91 683 SW2 AM PEAK 663 166 4,597 1,095 5,692 569 4,269 3,408 14,558 10,900 1,090 11,990 121 12 PM PEAK 39 10 270 64 335 33 251 SW3 AM PEAK 352 88 2,441 582 3,022 302 2,267 26,398 32,894 5,000 500 5,500 598 1 PM PEAK 57 14 395 94 489 49 367 SW4 AM PEAK 0 0 0 0 0 0 0 4,397 26,665 13,300 1,330 14,630 182 6 PM PEAK 1,402 351 9,721 2,316 12,037 1,204 9,028 NW1 AM PEAK 62 16 430 102 532 53 399 26,779 30,702 9,900 990 10,890 282 3 PM PEAK 185 46 1,283 306 1,588 159 1,191 NW2 AM PEAK 457 114 3,169 755 3,924 392 2,943 2,252 10,416 11,300 1,130 12,430 84 15 PM PEAK 57 14 395 94 489 49 367 NW3 AM PEAK 447 112 3,099 738 3,838 384 2,878 1,597 15,177 5,400 540 5,940 256 5 PM PEAK 408 102 2,829 674 3,503 350 2,627 NW4 AM PEAK 106 27 735 175 910 91 683 4,540 6,922 5,700 570 6,270 110 14 PM PEAK 44 11 305 73 378 38 283 NE1 AM PEAK 114 29 790 188 979 98 734 1,573 27,717 16,100 1,610 17,710 157 8 PM PEAK 1,532 383 10,622 2,531 13,153 1,315 9,865 NE2 AM PEAK 4 1 28 7 34 3 26 989 4,403 8,000 800 8,800 50 16 PM PEAK 211 53 1,463 349 1,812 181 1,359 NE3 AM PEAK 494 124 3,425 816 4,241 424 3,181 798 8,644 6,300 630 6,930 125 11 PM PEAK 0 0 0 0 0 0 0 NE4 AM PEAK 486 122 3,370 803 4,173 417 3,129 858 14,200 7,500 750 8,250 172 7 PM PEAK 354 89 2,454 585 3,039 304 2,279 NE5 AM PEAK 6 2 42 10 52 5 39 1,347 17,801 5,900 590 6,490 274 4 PM PEAK 1,030 258 7,142 1,702 8,843 884 6,632 NE6 AM PEAK 107 11 309 74 383 38 287 1,049 14,860 8,800 880 9,680 154 9 PM PEAK 1,980 206 5,720 1,363 7,083 708 5,312 NE7 AM PEAK 1,225 128 3,539 843 4,382 438 3,287 2,228 10,937 8,800 880 9,680 113 13 PM PEAK 91 9 263 63 326 33 244 NE8 AM PEAK 155 16 448 107 554 55 416 83 9,117 2,368 237 2,605 350 2 PM PEAK 1,210 126 3,496 833 4,329 433 3,246

30 6. The assumed pavement section consists of 2-in. surface course, 15 in. of base course, and two courses of 6-in. graded aggregate base. 7. Ground mount signing and pavement markings were esti- mated by cost-per-mile estimates. With the exception of the restriping alternatives and replacement of sign structures, roadway lighting and intelligent transportation systems (ITS) were estimated separately as noted within each estimate. 8. Utility relocation costs were estimated at 8% of the neat construction cost. 9. A 35% contingency and overhead factor of 15.3% was applied to each alternative estimate. Table 3.6 presents a detailed description of various cost and savings estimates associated with each improvement. Benefit–cost analysis and resulting priority rankings for each improvement project under the existing reliability ratio sce- nario (0.75) are also reported. Table 3.7 summarizes the sensitivity of project rankings to the reliability ratio scenario when RR values are varied between zero and 1.2 at 0.05 increments. In other words, Table 3.7 indi- cates how increasing relative value of travel time reliability sav- ings as an index of travel time (delay) savings has contributed to the ranking of different projects. Figure 3.8 exhibits the same sensitivity analysis findings as in Table 3.7. Note that Figure 3.8 facilitates the visual inspection of changes in the rankings of a given project when RR values are varied. Figure 3.8 illustrates the changes in project rankings when RR is varied from 0 to 1.2. It should be noted that in this analy- sis, the top ranked project (SW3) has a high benefit–cost ratio, approximately 600%. As a result, SW3 is not challenged by any other project as the RR is varied. Among the top five projects, the project ranked second goes progressively down in ranking when the RR increases to 0.35, 0.85, and 1.05. Projects ranked six through nine are stable throughout this range. The project originally (when RR = 0) ranked 10 also dropped in the rank- ings as the RR is progressively increased to 0.1, 0.4, and 0.75. From this graph it can be seen that the majority of changes hap- pen in the 0.35–0.45 range. At higher RR values (larger than 0.7) the switch between projects is few and far between. Figure 3.9 demonstrates the effect of budget constraints on project selection under different RR scenarios. It should be noted that at budgets less than $31,425,000, the top five proj- ects (SW3, NW1, NE8, NE5, and NW3) compete for funding. SW3, with a total cost of $5,500,000, is always the first choice. When RR varies between 0.65 and 0.90, NE8, with a total cost of $2,605,000, is always the second choice. In this range, when RR is less than 0.85, NW1, with a total price tag of $10,890,000, is the third choice. However, at RR levels larger than 0.85, NE5, with a smaller total cost of $6,490,000, will be the third choice. Throughout this range, NW3 is the fifth choice for funding at a total price tag of $5,940,000. So, it can be concluded that at low budget levels the choice of RR can be crucial in prioritizing and selecting projects as is evident in the switch between more expensive NW1 and cheaper NE5. In this case increasing RR to 0.85 has caused NE5 (which is relatively more advantageous in terms of reliability) to obtain higher priority over NW1. Delving a bit deeper into the details of projects NW1 and NE5 shows that quantitative analysis of improvement costs and savings depends on various project-specific factors includ- ing existing and projected volumes, safety-related statistics, adopted mitigation factors, and the number and configuration of existing lanes, among other things. Therefore, among low- budget-type improvements considered on I-695 (which are typically of a similar nature), rankings are mainly influenced by relative improvements in delay and travel time reliability, as well as traffic demand levels and presence and frequency of severe incidents at each location. 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 con- stant factor of 0.75 was applied to the reported VOTT savings. The reader should note that this is an approximation and effectively reflects the implicit assumption that all 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 analysis (and perhaps in similar practical decision-making scenarios), this exemplifies the versatility of the proposed reliability valuation method. Also, note that, in this analysis, each improvement is evalu- ated independent of the other proposed improvements. In practice, the interactions between nearby improvements should be taken into account. In the next section the results of incorporating the pro- posed VTTR estimation method into long-term prioritization and project selection are presented. In this case, disaggregate O-D information is used to estimate VOTT and VTTR sav- ings. Also, in this application interactions between different projects under the framework of a long-term regional trans- portation planning model are taken into account. Incorporating results into Long-term prioritization and project Selection In order to incorporate the findings of this study into the long- term prioritization and project selection process at the Mary- land SHA, a postprocessing module was developed for the Maryland Statewide Transportation Model (MSTM). These efforts illustrated that the travel time reliability valuation pro- cess and its corresponding savings estimation can be easily inte- grated into any regional travel demand model. However, note that these results should only be regarded as a proof of concept.

31 Table 3.7. Sensitivity Analysis on Improvement Project Rankings with Various Reliability Ratios Project Code 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 1.2 SW1 11 11 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 SW2 12 12 12 12 12 12 12 12 11 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 SW3 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 SW4 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 NW1 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3 3 4 4 4 4 5 5 5 5 NW2 15 15 15 15 15 15 15 15 15 15 15 15 15 15 15 15 15 15 15 15 15 15 15 15 15 NW3 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 4 4 4 4 NW4 10 10 11 11 11 11 11 11 13 13 13 13 13 13 14 14 14 14 14 14 14 14 14 14 14 NE1 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 NE2 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 NE3 13 13 13 13 13 13 13 13 12 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 NE4 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 NE5 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 3 3 3 3 3 3 3 3 NE6 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 NE7 14 14 14 14 14 14 14 14 14 14 14 14 14 14 13 13 13 13 13 13 13 13 13 13 13 NE8 3 3 3 3 3 3 3 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2

32 Figure 3.8. Improvement project rankings under various reliability ratios.

33 Figure 3.9. Impacts of different reliability ratio and budget levels on selected improvement projects. 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 1.2 0 10000 20000 30000 40000 50000 60000 70000 80000 90000 100000 110000 120000 130000 140000 150000 160000 Reliability Ratio B u d g e t ( $ x 1 0 0 0 ) SW1 SW2 SW3 SW4 NW1 NW2 NW3 NW4 NE1 NE2 NE3 NE4 NE5 NE6 NE7 NE8

34 Future research directions should include integration of a calibrated reliability ratio model into travel behavior models. One of the integral findings of SHRP 2 L35B is the data- driven empirical model to compute reliability ratio (RR). Previ- ously, RR has been defined as the ratio of VTTR to VOTT; however for the purposes of this long-term prioritization analy- sis, it can also be defined as the ratio of the system benefits from travel reliability enhancements to the system benefits from travel time savings. This ratio, in theory, should differ based on transportation facility type, level of congestion, vehicle fleet composition, time of day, trip purpose, etc. The proposed empirical formula for RR was used to compute travel time savings and travel time reliability savings for four scenarios: 1. Base year—build; 2. Base year—no build; 3. Future year—build; and 4. Future year—no build. The base and future years are 2010 and 2030 respectively. The base case—no build scenario represents the network conditions prior to construction of the Intercounty Connec- tor (ICC). The base case—build scenario represents the land use for year 2010 and current network with the ICC. The future year—no build scenario includes future-year land use along with the base-year network. The future year—build scenario consists of land use forecasts for the year 2030 with all proposed projects as currently contained in the Maryland SHA’s Constrained Long Range Plan (CLRP). A step-by-step process of the methodology used is shown in Figure 3.10. The first step was to prepare the necessary input files to run MSTM. Input files for four scenarios were then cre- ated. The next step was to complete the model run and sum- marize the results. In preparing the model summary, a congested skim matrix was developed to represent congested travel times for each O-D pair. Similarly, corresponding trip matrices were obtained. After summarizing model results for each scenario, reliability ratios for each O-D pair were obtained. Disaggregate travel time savings and travel time reliability savings for all O-D pairs were computed for the base-year and future-year scenar- ios. In the comparison, average travel times by O-D pair and by time of day, both before and after system enhancements, were Figure 3.10. Step-by-step process for incorporating SHRP 2 L35B travel time reliability into MSTM.

35 captured. System benefits were estimated based on the resulting improved travel time reliability at the O-D level. The base-year comparison shows benefits resulting from the ICC, and the future-year comparison shows benefits resulting from projects included in CLRP. The findings of this analysis are summarized at varying geographic levels: statewide, county, zone and corridor. Both travel time savings and travel time reliability savings were computed at these geographic levels. The analysis was con- ducted for the AM peak period only and by considering all the trips as a medium income group. However, the results can be summarized for other peak periods and by considering the other five income classes included in the MSTM. Statewide Findings Statewide findings were estimated by taking travel time improvements for all O-D pairs when multiplied by corre- sponding trips. The findings suggest that both the base and future-year scenarios result in savings when compared with their no-build counterparts. Future-year savings are higher than the base year as expected. At the statewide level, travel time reliability savings are approximately 92% of travel time savings for the base year. Table 3.8 shows statewide travel time and travel time reliability savings during peak hours (includ- ing AM and PM peak) for a whole year. County Level Findings Travel time savings and travel time reliability savings are plot- ted at the county level for base (Figure 3.11) and future years (Figure 3.12). County level savings are shown for a typical day in the AM peak period. In the base year, Montgomery and Prince George’s counties received higher savings. The major- ity of these savings can be attributed to the opening of the ICC in the base year under the build scenario. In the future-year scenario, Anne Arundel and Baltimore counties received higher savings as a result of CLRP project implementation in these counties. Transportation Analysis Zone Level Findings Transportation Analysis Zone (TAZ) level findings are shown in Figures 3.13 through 3.16. Base-year findings suggest that zones near the ICC enjoyed higher savings in terms of travel time and travel time reliability values. Future-year findings suggest that the savings are spread over major urban and sub- urban areas. Figures 3.13 and 3.15 represent travel time savings in min- utes for zones in the following three categories: • Less than 1 minute; • Between 1 and 5 minutes; and • More than 5 minutes. Figures 3.14 and 3.16 represent travel time reliability value savings in dollars for zones in the following three categories: • Less than $0.25; • Between $0.25 and $1; and • More than $1. Table 3.8. Statewide Peak Hour Savings for a Year Year Total Savings Travel Time (min) Travel Time ($) Base Year Travel Time 449,915,060 104,965,240 Travel Time Reliability 416,446,020 97,157,160 Future Year Travel Time 1,812,587,810 422,876,590 Travel Time Reliability 1,837,341,380 428,651,620 Figure 3.11. County level savings comparing base year—build with base year—no build. Qu ee n An ne ’s Wo rc es ter Ba ltim or e Cit y 0 20,000 Mo ntg om er y Pri nc e Ge or ge’ s An ne Ar un del Fre der ick Ba ltim or e Ho w ar d Ha rfo rd Ca rr oll Ch ar les St. Ma ry’ s Ca lve rt Wi co m ico Ta lbo t Wa shi ng ton Ce cil Al leg an y Do rc hes ter Ca ro lin e So m er se t Ke nt Ga rr ett 40,000 60,000 80,000 100,000 120,000 140,000 Travel Time Savings ($) Travel Time Reliability Savings ($)

36 Do rc hes ter Qu ee n An ne ’s Fre der ick Pri nc e Ge or ge’ s 0 20,000 40,000 60,000 80,000 100,000 120,000 140,000 160,000 180,000 Travel Time Savings ($) Travel Time Reliability Savings ($) Wo rc es ter Ba ltim or e Cit y Mo ntg om er y An ne Ar un del Ba ltim or e Ho w ar d Ha rfo rd Ca rr oll Ch ar les St. Ma ry’ s Ca lve rt Wi co m ico Ta lbo t Wa shi ng tonCe cil Al leg an y Ca ro lin e So m er se t Ke nt Ga rr ett Figure 3.12. County level savings comparing future year—build with future year—no build. Figure 3.13. Travel time saving per trip comparing base year— build with base year—no build. Figure 3.14. Travel time reliability saving per trip comparing base year— build with base year—no build.

37 Figure 3.15. Travel time saving per trip comparing future year—build with future year—no build. Figure 3.16. Travel time reliability saving per trip comparing future year— build with future year—no build. Corridor-Level Findings To illustrate the performance of MSTM in evaluation of sav- ings at the corridor level, a regionally significant corridor on the northwest side of the Capital Beltway was considered. Travel time and travel time reliability savings were deter- mined for the I-270 corridor (Figure 3.17). Table 3.9 shows that, for the I-270 corridor, travel time savings are achieved for both the base and future years when compared with their respective no-build scenarios. Overall, these results indicate that reliability measures pro- posed in this study can be integrated into MSTM. For this pur- pose, four scenarios were considered: base case—no build, base case—build, future year—no build and future year—build. Travel time and travel time reliability savings were shown at the statewide, county, TAZ, and corridor levels. Based on the analy- sis results presented, savings in travel time reliability appear to be significant at all geographic aggregation levels. results of presentation 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

38 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 NB base year NB future year SB base year SB future year Travel time savings(min)/Traveler Travel time reliability savings(min)/Traveler Figure 3.17. Travel time and travel time reliability savings in minutes per traveler on I-270. Table 3.9. I-270 Travel Time and Reliability Results for Four Different Scenarios Scenario I-270 Travel Time (min) I-270 RR I-270 Travel Time Savings (min/ traveler) I-270 Travel Time Reliability Savings (min/ traveler) NB SB NB SB NB SB NB SB Base Case—No Build 20.2 23.8 0.74 0.79 1.6 2.0 1.7 2.2 Base Case—Build 18.6 21.8 0.71 0.77 Future Case—No Build 21.6 25.7 0.76 0.82 1.8 2.0 1.9 2.2 Future Case—Build 19.8 23.7 0.73 0.79 the findings of this research. The entire presentation used during this meeting is included in Appendix C. What follows is a summary of some of the key points presented and the feedback obtained. The research team’s overall approach to presenting the SHRP 2 L35B project results to SHA management was to (1) explain the travel-time data-driven methodology devel- oped at a high level and NOT get into specific details of its technical development and implementation; and (2) focus on the results of the methodology and its application to both short-term and long-term decision-making processes. A few slides from Appendix C are included here for ready reference in describing the presentation. The slide in Figure 3.18 was used to explain the underlying analogy for the travel-time data-driven methodology. If a traveler, based on experience, knows 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; toler- ance of travel time variation; and how differences in expected travel time might impact the traveler’s experience. The following slide in Figure 3.19 was used to provide a high-level explanation of how, essentially, the travel-time data-driven methodology works. In an attempt to make the complex relatively simple: the pro- posed travel-time data-driven methodology for estimating

39 Figure 3.18. Explanation of the underlying analogy for the travel-time data-driven methodology. Figure 3.19. Explanation of how the travel-time data-driven methodology works.

40 value of reliability uses large quantities of historical travel time data based on probe data, along with a value of typical/usual travel time (VOTT), and produces a RR along with a value of reliability (VTTR). Discussion of the “calculations” was cursory and an attempt was not made to discuss any technical details. It was mentioned, however, that SHRP 2 would be enlisting out- side technical expert reviewers to review the entire methodol- ogy developed, its assumptions, and application calculations. The following slide in Figure 3.20 was used to explain the output of the travel-time data-driven methodology results. Based on the results obtained from application of the pro- posed travel-time data-driven methodology, it can be con- cluded that, during peak hours in congested urban areas, the average reliability ratio 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 reliability ratio can be taken as 0.52. Therefore, it seems the current value (0.75) is reasonable when reliability of commute travel times during peak hours in congested urban areas is considered. The slide in Figure 3.21 was used to demonstrate the impact of including a value of reliability (using sensitivity to RR) in SHA’s congestion relief project life-cycle BCA selec- tion process (as explained earlier in this chapter). The slide shows how project rankings are impacted for the top 6 highest ranked projects. If for example, SHA was decid- ing on priority congestion relief projects with a budget of $15M, not taking into account a value of reliability would likely result in selection of projects ranked 1 (cost is $5.5M) and 2 (cost is $10.9M). Both of these projects involve construc- tion of auxiliary lane extensions; however, project 2 requires construction of a retaining wall, which adds significantly to the cost of the project. Using SHA’s current RR of 0.75 results in the project previously ranked 3 jumping into the second ranked slot. This project costs considerably less at $2.6M and involves removing a ramp on the inner loop of I-695 and replacing it with a signal. Finally, if SHA selected a RR value of 0.85 (which is the top of the range of values obtained using the travel-time data-driven methodology), the project previously ranked 4 (cost is $6.5M) jumps into the third ranking. Ultimately, this might mean SHA would choose to do three projects instead of two if the budget was $15M. SHA was also presented with slides showing the travel time reliability savings at various geographic levels based on con- struction of the ICC (explained earlier in this chapter). In terms of conclusions, based on the results of this project, the research team expressed the opinion that SHA’s current RR of 0.75 is a good, and defensible, estimate based on the literature as well as the proposed travel-time data-driven methodology. That said, while the travel-time data-driven methodology shows significant promise, it does require a rigorous valida- tion of hypotheses underlying the methodological develop- ments as well as validation of application results (see suggested further research). Figure 3.20. The output of the travel-time data-driven methodology results.

41 The overall response from SHA, including management, was positive. Interestingly, and perhaps not surprisingly, SHA management did want to learn more about the techni- cal details regarding the travel-time data-driven methodol- ogy developed. There was also an interesting discussion, led by SHA management, that perhaps our collective goal should not be focused on “fixing congestion” as that is not Figure 3.21. Demonstration of the impact of including a value of reliability in SHA’s congestion relief project life-cycle BCA selection process. necessarily feasible in today’s world of financial constraint and other competing issues. Perhaps a better goal is to work toward making the system more reliable; however, the key will be communicating system reliability benefits in a way that is ultimately useful to decision makers. So the goal becomes improving reliability rather than eliminating congestion.

42 Conclusions and Suggested Research Overall Findings An overall conclusion from this research suggests that agencies who do not account for VTTR in their BCA processes might be undervaluing project benefits resulting from improvements to trip reliability. Valuation tools and techniques, both existing and newly developed as a result of this research, along with a signifi- cant body of literature, provide a basis for incorporating VTTR in an agency’s BCA process. While this research project focused on Maryland State Highway as a case study, the information (literature review, data-driven methodology, and application examples) has the potential to help agencies looking to incorpo- rate VTTR in their investment decision processes. Compared with the recent revealed and stated preference survey-based estimates in the literature, the current RR ratio value of 0.75 used by SHA seems reasonable. Based on the development 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 reasonable when the reliability of commute travel times dur- ing peak hours in congested urban areas is considered. 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 both conges- tion (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 planning time index, is not currently included. The VTTR estimates obtained from this research could be used to bridge the gap in reporting costs of unreliability in the annual mobility report. As noted above, this part of the report can help agencies incorporate VTTR into their investment decision processes. Every effort has been made to fully document the data-driven valuation methodology developed under this research to facil- itate its transferability to agencies beyond the Maryland 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 a software tool (or build into an existing performance-measure calculation and report- ing tool) that can process the historical travel time data and estimate RR/VTTR using the methodology developed (this is expanded on at the end of the next section). In addition to this suggestion of follow-on work to facilitate the practical applica- tion of the results of this research, a number of ideas for future research to build on and enhance the data-driven methodology developed are included in the next section. Suggested Future Research In future research, rigorous validation of hypotheses under- lying the methodological developments as well as validation of application results should take the highest priority. The assumptions made regarding travel times following a certain stochastic process (GBM with drift) in this study should be further investigated. It is particularly important to identify a set of stochastic processes with theoretical properties that are consistent with empirical travel time distributions. Note that the proposed valuation method can easily be modified to take into account any other stochastic process to model the projec- tion of travel time distribution over time and into the future. The stochastic volatility family of models (in which GBM is a member), and in particular, the Generalized Auto-Regressive Conditional Heteroskedasticity (GARCH) family of models, are deemed to be potential candidates for this purpose. C h a p t e R 4

43 The other assumptions regarding the payoff function used in the proposed method needs further validation based on local data. Survey-based measurements of penalties (or rewards) associated with arriving earlier or later than expected can be used as a comparison with the assumed bilinear form of the payoff function and its parameters. In jurisdictions where such survey-based measurements and models are readily available, it is recommended that the VTTR and RR estimates that can be obtained from the proposed data-driven method used in this study be validated against their survey-based counterparts. The payoff function also includes the same valuations for all trip purposes at all times of the day. Research should be conducted on the impact of changes in these factors on the payoff function. In applications regarding future scenario demand levels, aver- age travel times and travel time variability measures are inevi- tably estimated using some type of model. These available modeling techniques may vary widely by local jurisdiction in terms of their complexity and accuracy. In this study, micro- simulation and four-step modeling techniques were used for short-term and long-term evaluation of the impact of improve- ments on travel time reliability savings, respectively. Other traf- fic analysis techniques, as simple as sketch planning or as complicated as Dynamic Traffic Assignment (DTA), may be used in practice for this purpose. Given data availability, it is highly recommended that the effectiveness and accuracy of these modeling tools in recreating the needed measures of travel time variation and reliability be further investigated in real- world cases. Interactions between trip characteristics, traveler decision making, and travel time reliability valuation should be further investigated. Trip purpose (commute versus noncommute), mode (auto versus freight), facility type (freeway versus arte- rial), income level, trip distance, geography (urban versus rural), geometry (number of travel lanes) and presence of alternatives (e.g., mode, route, trip time) are among the fac- tors that conceivably have a direct impact on the value of travel time reliability. In the context of the proposed method developed in this study, the impact of these factors on VTTR estimation can be traced through their impact on the VOTT estimate, travel time variability (model specific parameters), and terminal payoff function characterizations. Different methods can be potentially used to aggregate travel time data. The respective impact of these aggregation methods on travel time variability and reliability valuation could be significant. In this study an instantaneous travel time aggregation method is used to estimate path travel times based on link travel times. It is conceivable that more elaborate path travel time estimation methods (e.g., trajectory construction- based models), will result in more accurate travel time esti- mates for long distance trips. Also, in this study 1-minute travel times are used. At this level, travel time data provides a very high level of resolution that essentially captures much of the variation in travel time experienced by users. However, it is possible that other jurisdictions may not have access to data at this resolution level, or they may decide to perform some tem- poral aggregation to avoid higher computational costs. It is recommended that in the future, the sensitivity of VTTR esti- mates to the accuracy and granularity of path travel time esti- mates be investigated. From a practical perspective, it is important that both spa- tial and temporal transferability of VTTR and RR models and estimates be investigated. The result would inform decisions as to how often the analysis needs to be repeated considering recent data, and whether or not similar (maybe nearby) juris- dictions need to perform the analysis using their respective local data. One potential outcome of such a study could be a set of recommended VTTR and RR values that can be used by local jurisdictions where access to accurate speed data and other resources needed to perform the proposed data-driven analysis is limited. Finally, a logical next step would be to develop a software tool to process the historical travel time data and to auto- mate the estimation of VTTR and RR values. The software tool should provide the opportunity to perform hypothesis testing and to calibrate appropriate stochastic process param- eters. This tool will also facilitate the sensitivity analysis through enabling seamless variation of different assump- tions regarding the time series process, payoff function spec- ifications, and estimation parameters. Additionally, the tool should provide the capability to perform sensitivity analy- sis on all assumptions that go into the project benefits quantification.

44 References: Part 1 Asensio, J., and Matas, A. 2008. Commuters’ Valuation of Travel Time Variability. Transportation Research Part E: Logistics and Transporta- tion Review, Vol. 44, 1074–1085. Bhat, C., and Sardesai, R. 2006. The Impact of Stop-Making and Travel Time Reliability on Commute Mode Choice. Transportation Research Part B: Methodological, Vol. 40, 709–730. Black, I., and Towriss, J. 1993. Demand Effects of Travel Time Reliability. Technical Report. Centre for Logistics and Transportation, Cran- field Institute of Technology, United Kingdom. Borjesson, M. 2009. Modelling the Preference for Scheduled and Un- expected Delays. Journal of Choice Modelling, 2(1), 29–50. Brownstone, D., and K. A. Small. 2005. Valuing Time and Reliability: Assessing the Evidence from Road Pricing Demonstrations. Trans- portation Research Part A: Policy and Practice, Vol. 39, pp. 279–293. Cambridge Systematics. 2008. Integrated Corridor Management Analy- sis, Modeling, and Simulation Experimental Plan for the Test Corri- dor. Report No. FHWA-JPO-08-035EDL 14415. Cambridge Systematics and ICF International. April 26, 2012. Synthesis Report & Workshop Working Paper. SHRP 2 Workshop on the Value of Travel Time Reliability. Carrion, C., and Levinson, D. 2010. Value of Reliability: High Occu- pancy Toll Lanes, General Purpose Lanes, and Arterials. In Confer- ence Proceedings of 4th International Symposium on Transportation Network Reliability, Minneapolis, Minn. Carrion, C., and Levinson, D. 2011. A Model of Bridge Choice Across the Mississippi River in Minneapolis. In Network Reliability in Prac- tice: Selected Papers From the Fourth International Symposium on Transportation Network Reliability (D. Levinson, H. Liu, and M. Bell, eds.), Springer, Chapter 8, pp. 115–129. Carrion, C. and Levinson, D. 2012. Value of Travel Time Reliability: A Review of Current Evidence. Transportation Research Part A: Policy and Practice, Vol. 46, pp. 720–741. Chang, G.-L. 2011. Performance Evaluation of CHART: An Incident Management Program, Report prepared for Maryland State High- way Administration, University of Maryland. Cirillo, C., Liu, Y., and Serulle, U. January 2014. Travel Time Reliability of Road Transportation Systems: Methods and Case Studies. White Paper. Cohen, H., and Southworth, F. 1999. On the Measurement and Valuation of Travel Time Variability Due to Incidents on Freeways. http://ntl.bts .gov/lib/7000/7500/7598/2cohen.pdf Concas, S., and Kolpakov, A. January 2009. Synthesis of Research on Value of Time and Value of Reliability. Report No. BD 549-37. De Jong, G. 2000. Value of Freight Travel Time Savings. In Handbook of Transport Modeling (Hensher D. A. and K. J. Button, eds.) Elsevier. De Jong, G., Tseng, Y., Kouwenhoven, M., Verhoef, E., and Bates, J. 2007. The Value of Travel Time and Travel Time Reliability. Technical Report. The Netherlands Ministry of Transport, Public Works and Water Management. Delache, X. 2009. Cost Benefits Analysis Applied to Reliability Policies in France. Presented at International Meeting on Value of Travel Time Reliability and Cost-Benefit Analysis, Vancouver, B.C., Canada. Department for Transport, Transport Appraisal and Strategic Modelling Division. 2014. Transport Analysis Guidance (TAG) UNIT A1.3 User and Provider Impacts. https://www.gov.uk/government/uploads/ system/uploads/attachment_data/file/313222/webtag-tag-unit-a1- 3-user-and-provider-impacts.pdf Dowling, R., Skabardonis, A., Carroll, M., and Wang, Z. 2004. Methodol- ogy for Measuring Recurrent and Non-Recurrent Traffic Congestion. Presented at the 83rd Annual Meeting of the Transportation Research Board, Washington, D.C. Eliasson, J. 2009. Reliability in Swedish CBA—Current Practice and What Needs to be Done. Presented at the International Meeting on Value of Travel Time Reliability and Cost-Benefit Analysis, Vancouver, B.C., Canada. FHWA. 2013. Highway Economic Requirements System. http://www .fhwa.dot.gov/infrastructure/asstmgmt/hersindex.cfm FHWA. 2014. Intelligent Transportation System Deployment Analysis Sys- tem (IDAS). http://www.fhwa.dot.gov/research/deployment/idas.cfm Fosgerau, M., and Karlström, A. 2010. The Value of Reliability. Trans- portation Research Part B: Methodological, Vol. 44, No. 1, pp. 38–49. Fukuda, D. 2009. Reliability Valuation in Japan. Presented at Inter- national Meeting on Value of Travel Time Reliability and Cost- Benefit Analysis, Vancouver, B.C., Canada. Ghosh, A. 2001. Valuing Time and Reliability: Commuters’ Mode Choice from a Real Time Congestion Pricing Experiment. PhD dis- sertation. University of California, Irvine. Hollander, Y. 2006. Direct Versus Indirect Models for the Effects of Unreliability. Transportation Research Part A: Policy and Practice, Vol. 40, 699–711. IndexMundi. http://www.indexmundi.com/facts/united-states/quick- facts/maryland/average-commute-time#map. Accessed 2013. International Transport Forum, Joint Transport Research Centre of the OECD and the ITF. 2010. Surface Transport Networks: Improving Reliability and Levels of Service, http://www.internationaltransport forum.org/pub/pdf/10Reliability.pdf

45 International Transport Forum. 2012. Policy Brief: Making Reliability Part of Transport Policy, April. KiM Netherlands Institute for Transport Policy Analysis. November 2013. The Social Value of Shorter and More Reliable Travel Times. Report. Ministry of Infrastructure and the Environment. Koskenoja, P. 1996. The Effect of Unreliable Commuting Time on Com- muter Preferences. PhD dissertation. University of California, Irvine. Levinson, D., and Smalkoski, B. 2003. Value of Time for Commercial Vehicle Operators in Minnesota. TRB International Symposium on Road Pricing, Key Biscayne, Fla. Li, Z., Hensher, D., and Rose, J. 2010. Willingness to Pay for Travel Time Reliability in Passenger Transport: A Review and Some New Empiri- cal Evidence. Transportation Research Part E: Logistics and Transpor- tation Review, Vol. 46, 384–403. Liu, H., Recker, W., and Chen, A. 2004. Uncovering the Contribution of Travel Time Reliability to Dynamic Route Choice Using Real-Time Loop Data. Transportation Research Part A: Policy and Practice, Vol. 38, 435–453. Liu, H., He, X., and Recker, W. 2007. Estimation of the Time-Dependency of Values of Travel Time and Its Reliability from Loop Detector Data. Transportation Research Part B: Methodological, Vol. 41, 448–461. Mahapatra, S., Wolniak, M., Sadabadi, K. F., Beckett, E., and Jacobs, T. September 2013. Maryland State Highway 2013 Mobility Report— 2nd Edition. Maryland Department of Transportation, Baltimore. Maryland Department of Transportation. 2009. 2009 Maryland Transpor- tation Plan. http://www.mdot.maryland.gov/Office_of_Planning_ and_Capital_Programming/Plans_Programs_Reports/Reports/ MTP/2009MTP.pdf Maryland Department of Transportation, State Highway Administra- tion. 2011. Fiscal Year 2011 Annual Report. http://sha.md.gov/OC/ Annual_FY11_Report.pdf Maryland State Highway Administration. June 2011. CHART Planning & Programming Process, Office of CHART and ITS Development. Maryland State Highway Administration. June 2012. Maryland State Highway Mobility Report. Maryland Department of Transportation, Baltimore. National Center for Smart Growth and Parsons Brinckerhoff. June 2011. Maryland Statewide Transportation Model User’s Guide: Draft Final Report. Report for Maryland State Highway Administration. Pu, W., Mahapatra, S., Tadayon, M., and Simpson, D. N. August 2013. Identifying Net Effect of Highway Project in Before-and-After Evalua- tion. Paper submitted for presentation at the 93rd Annual Meeting of the Transportation Research Board, Washington, D.C. Senna, L. 1994. The Influence of Travel Time Variability on the Value of Time. Transportation, Vol. 21, 203–228. Shires, J., and De Jong, G. 2009. An International Meta-Analysis of Value of Travel Time Savings. Evaluation and Program Planning, Vol. 32, 315–325. SHRP 2. 2010a. SHRP 2 Draft Reports for Projects C10A and C10B, Part- nership to Develop an Integrated, Advanced Travel Demand Model and a Fine-Grained, Time-Sensitive Network, Transportation Research Board of the National Academies, Washington, D.C. SHRP 2. 2010b. SHRP 2 Draft Report C11: Development of Tools for Assessing Wider Economic Benefits of Transportation, Transportation Research Board of the National Academies, Washington, D.C. SHRP 2. 2012. SHRP 2 Draft Report L11: Evaluating Alternative Opera- tions Strategies to Improve Travel Time Reliability. Transportation Research Board of the National Academies, Washington, D.C. SHRP 2. 2013a. SHRP 2 Draft Report C04: Improving Our Understanding of How Highway Congestion and Pricing Affect Travel Demand, Transpor- tation Research Board of the National Academies, Washington, D.C. SHRP 2. 2013b. SHRP 2 Draft Report L04: Incorporating Reliability Per- formance Measures in Operations and Planning Modeling Tools. Transportation Research Board of the National Academies, Wash- ington, D.C. SHRP 2. 2013c. SHRP 2 Draft Report L05: Incorporating Reliability Performance Measures into the Transportation Planning and Pro- gramming Processes. Transportation Research Board of the National Academies, Washington, D.C. http://onlinepubs.trb.org/onlinepubs/ shrp2/SHRP2prepubL05Guide.pdf Significance, VU University Amsterdam and John Bates Services in col- laboration with TNO, NEA, TNS NIPO and PanelClix. November 2013. Values of Time and Reliability in Passenger and Freight Trans- port in The Netherlands. Report for the Ministry of Infrastructure and the Environment. Project 08064, Revised Final Version. Small, K., Noland, R., and Koskenoja, P. 1995. Socio-Economic Attributes and Impacts of Travel Reliability: A Stated Preference Approach. Cali- fornia PATH Research Report No. UCB-ITS-PRR-95-36. Small, K., Noland, R., Chu, X., and Lewis, D. 1999. Valuation of Travel- Time Savings and Predictability in Congested Conditions for Highway User-Cost Estimation. National Cooperative Highway Research Pro- gram Report No. 431. Small, K., Winston, C., and Yan, J. 2005. Uncovering the Distribution of Motorists’ Preferences for Travel Time and Reliability. Econometrica, Vol. 73, 1367–1382. Taylor, Michael A. P. 2009. Reliability and Benefit–Cost Analysis in Aus- tralia and New Zealand. Presented at the International Meeting on Value of Travel Time Reliability and Cost-Benefit Analysis, Vancouver, B.C., Canada. Tilahun, N., and Levinson, D. 2010. A Moment of Time: Reliability in Route Choice Using Stated Preference. Journal of Intelligent Trans- portation Systems, Vol. 14, 179–187. TransportationEconomics.org 2010. http://bca.transportationeco nomics.org/benefits/travel-time/categories-of-travel-time Transportation Research Board of the National Academies. October 15–16, 2009. International Meeting on Value of Travel Time Reliabil- ity and Cost-Benefit Analysis. Jointly organized by SHRP 2 of the Transportation Research Board, Washington, D.C., and the Joint Transport Research Centre, Vancouver, B.C., Canada. Transportation Research Board of the National Academies. 2014. http:// www.trb.org/StrategicHighwayResearchProgram2SHRP2/General .aspx Tseng, Y., Ubbels, B., and Verhoef, E. 2005. “Value of Time, Schedule Delay and Reliability. Estimation Results of a Stated Choice Experi- ment among Dutch Commuters Facing Congestion,” European Regional Science Association Congress. University of Florida, Transportation Research Center. April 2007. Travel Time Reliability and Truck Level of Service on the Strategic Intermodal System. Final report to Florida Department of Transportation. U.S. Department of Transportation, Research and Innovative Technol- ogy Administration (RITA). State Transportation Statistics. Wash- ington, D.C., 2013. http://www.rita.dot.gov/bts/sites/rita.dot.gov .bts/files/publications/state_transportation_statistics/state_ transportation_statistics_2013/index.html/chapter4/table4_1 Yan, J. 2002. Heterogeneity in Motorists’ Preferences for Travel Time and Time Reliability: Empirical Finding from Multiple Survey Data Sets and its Policy Implications. PhD thesis, University of California, Irvine. Yusufzyanova, D., and Zhang, L. January 2011. Multi-Modal and Multi- Jurisdictional Transportation Investment Decision Making: The Case of the Washington DC–Baltimore Region. Paper submitted to TRB for publication.

46 Overview of Maryland Department of Transportation Planning The Maryland Department of Transportation (MDOT) is one of the state’s largest agencies, with nearly 9,000 employ­ ees committed to delivering a balanced and sustainable multi­ modal transportation system for all Maryland’s residents and businesses. As a truly multimodal transportation agency, MDOT is responsible for coordinating statewide transporta­ tion planning activities across all methods of transportation, including highways, tunnels, bridges, railways, rail transit, buses, ports, airports, bike paths, sidewalks, and trails, as well as driver services. MDOT provides oversight of, and coordi­ nates with, five administrations that have unique functional responsibilities for the transportation facilities and services in Maryland as shown in Figure A.1. State Report on Transportation Each year MDOT publishes the State Report on Transporta­ tion (SRT). The SRT contains three important documents: the Maryland Transportation Plan (MTP), the Consolidated Transportation Program (CTP), and the annual Attainment Report (AR) on Transportation System Performance. Fig­ ure A.2 gives a visual example of how it is compiled. Maryland Transportation Plan The MTP is a 20­year vision for transportation in Maryland. It outlines the state’s transportation policies and priorities and helps guide statewide investment decisions across all methods of transportation. The MTP is one component of the annual State Report on Transportation, which also includes the CTP and the AR. The CTP is Maryland’s six­year capital budget for transportation projects. The annual AR tracks MDOT’s progress toward attaining the goals and objectives of the MTP using outcome­oriented performance measures. The current MTP was completed in 2009 (MDOT, 2009). The five stated goals of the current MTP include 1. Quality of Service—enhances users’ access to and positive experience with all MDOT transportation services. 2. Safety and Security—provide transportation assets that maximize personal safety and security in all situations. 3. System Preservation and Performance—protect Mary­ land’s investment in its transportation system through strategies to preserve existing assets and maximize the effi­ cient use of resources and infrastructure. 4. Environmental Stewardship—develops transportation policies and initiatives that protect the natural, commu­ nity, and historic resources of the state and that encourage development areas that are best able to support growth. 5. Connectivity to Daily Life—supports continued economic growth in the state through strategic investments in a bal­ anced, multimodal transportation system. The goal of improving quality of service basically reflects improvements in accessibility and mobility. This should include reduction in travel time or delay, or increase in travel time reliability for non­motorized travelers, private vehicle users, transit users, and freight/commercial users. Figure A.3 presents the current MTP milestones. Over time, changes to Maryland’s population, economy, and environment will result in far­reaching effects on the transportation system. The picture of transportation in Maryland in 20 years may look quite different from today’s picture. Though not a comprehensive list of the challenges that MDOT will face in the coming years, the following criti­ cal issues are some of the most important issues that will shape the decisions made by MDOT, its modal administra­ tions, and the Maryland Transportation Authority (MDTA). The MTP provides a path to help MDOT address these chal­ lenges in the future. They are • Transportation and the economy • Freight demand and infrastructure capacity • Planning for development A P P e n d i x A

47 Figure A.1. Maryland Department of Transportation with its modal administrations. Figure A.2. Components of the State Report on Transportation. Figure A.3. Current Maryland Transportation Plan milestones. • Transportation and the environment • Transportation needs outpacing funding resources • Transportation­related fatalities and injuries Maryland department of Transportation Budget Allocation MDOT has a somewhat unusual system for funding trans­ portation projects. The state’s Transportation Trust Fund (TTF) is a unified pot of money that provides MDOT the flexibility to fund high­priority projects across the state regardless of transportation modes (Yusufzyanova et al., 2011). Local roads in Maryland are controlled and main­ tained by cities and counties. Also, MDOT provides Mary­ land’s entire share of funding for the regional transit system in the D.C. area known as the Washington Metropolitan Area Transit Authority (WMATA). Figure A.4 illustrates MDOT’s TTF allocation between jurisdictions and modes in the state. TTF is first divided into separate funds to meet dif­ ferent transportation needs categories (e.g., maintenance, capital programming) and then allocated to different modal agencies, where it is then subject to the investment process of the modal agencies.

48 Overview of SHA investment decision-Making Process The Maryland State Highway Administration receives high­ way transportation funds from MDOT, and works with Metro­ politan Planning Organizations (MPOs) and local jurisdictions to allocate funds to meet highway preservation and capital programming needs. In the last two decades, system preser­ vation projects have received a higher and higher share of SHA’s transportation funds due to aging infrastructure, and this trend is likely to continue in the future. The Admin­ istration identifies system maintenance and preservation needs through an internal technical evaluation process, and has created a large number of funding categories for different preservation and maintenance needs. For instance, SHA per­ forms technical evaluation of pavement and bridge con ditions every year, and has set the goal of keeping 84% of pavements under “acceptable conditions.” While pavement and bridge maintenance consumes the majority of SHA’s system preserva­ tion budget, there are also 24 smaller funding categories dedi­ cated to specific needs including drainage, traffic signs, and community improvement. For capital improvement and sys­ tem expansion projects, SHA coordinates with six MPOs and local jurisdictions (through a priority­letter process discussed below). The SHA transportation investment process centers on MPO­level transportation improvement programs (TIPs) and the statewide Consolidated Transportation Program (CTP). TIPs represent projects within the boundary of each MPO, and SHA provides technical assistance with those proj­ ects on request. TIPs consist of projects funded by federal money and matching state/local contributions. The CTP is a six­year program that is financially constrained by the Mary­ land Transportation Trust Fund. Figure A.5 shows the time­ line for the CTP development process. There is a financially unconstrained predecessor to the CTP, often referred to as the 20­year state Highway Needs Inventory (HNI). The HNI is a technical document (based on performance/condition moni­ toring and travel demand forecasts) that identifies all required highway improvements as well as safety and structural prob­ lems on the existing highway facilities. Usually, only “serious” projects from the HNI undergo detailed engineering planning Figure A.4. MDOT’s TTF allocation between jurisdictions and modes in the state. (Given percentages are FY 2009 budget allocation (may vary year to year).)

49 phases and cost estimation procedures. The HNI lists only major capital improvement projects (i.e., no system preserva­ tion projects), and is the main source of candidate projects for the SHA transportation investment process. Another source of candidate projects is the priority letters submitted to SHA by individual counties in Maryland. Priority letters represent each county’s internal ranking of projects based on local needs and local inputs. All candidate projects for capital improvement from HNI and county priority letters are evaluated by SHA planners based on three main investment criteria: safety, congestion mitigation, and support for economic development, though there is no formal quantitative evaluation procedure. Prior­ ity letters should detail how each priority project supports the goals of the Maryland Transportation Plan and are consistent with the county’s land use plan goals. MDOT provides a two­page project questionnaire that summarizes all the needed information about each project (note that the questionnaire specifically mentions travel time reliabil­ ity as an objective under the goal of improved quality of service). NEPA (National Environmental Policy Act) and political considerations also play a role in this prioritization process, though the actual influence of these two factors can only be analyzed on a project­by­project basis. Although there is a formal procedure for SHA to discuss project prioritization with counties each fall (known as the “fall county tour” dur­ ing which MDOT and SHA engineers and planners visit each county and hold public meetings; there are also meetings between SHA and local jurisdiction representatives before the tour), it is possible that a county may not get any high­ priority projects for the county funded by SHA. If a project proposed by a county meets all SHA requirements but does not receive enough federal or state funding to be included into the CTP, the county may “come to the table” and share the cost with SHA. Typically, only the counties with high lev­ els of economic development (e.g., Montgomery and How­ ard counties) participate financially as project sponsors. After needs­based analysis and negotiations with counties are com­ pleted, SHA submits the draft CTP each year to the MDOT secretary, which may be revised and then submitted to the Maryland state legislature for possible further revisions and budget approval. Revisions to CTP at these later stages often originate from political influences and changes in budgetary situations. The complete high­level process for the SHA investment process, including interactions with counties and MPOs, is shown in Figure A.6. Figure A.5. Timeline for CTP development process.

50 Maryland State Highway Administration Budget Allocation—example from FY2011 SHA’s annual expenditure can be divided into two distinct areas with each area further breaking down into three main categories: • Capital ($738.3M) 44 Construction ($634.3M) 44 County and Municipality ($98.3M) 44 IT Development ($5.6M) • Operating ($409.7M) 44 Maintenance ($236.7M) 44 County and Municipality ($157.5M) 44 Highway Safety ($15.5M) The numbers within parenthesis indicate SHA’s expendi­ tures in each category during FY2011 as reported in the Mary­ land SHA Annual Report (MDOT SHA, 2011). The pair of pie charts and the table in Figure A.7 further illustrates how SHA use of funding for capital and operating projects has been apportioned among various programs. CHART (Operations & Management) Planning and Programming Process After several years of experience in deploying intelligent trans­ portation system (ITS) technology, the Maryland SHA has established a process within its CHART program for planning, programming, designing, building, operating, and maintain­ ing ITS to provide benefits to its customers (Maryland SHA, June 2011). What follows is a high­level description of the plan­ ning and programming portion of the CHART program’s deployment process. Planning is the initial step within the CHART ITS proj­ ect process. Once an operational need is established for a par­ ticular CHART project, it is first planned using inputs from all relative users and stakeholders, and then the appropriate Figure A.6. High-level process for SHA investment process, which includes interactions with counties and with MPOs. Funding is ensured Local jurisdictions and MPOs need identification Fall Tour, Final Consolidated Transportation Plan 6 year program MDOT allocates Transportation Trust Fund between Different Modes and Jurisdictions Minor Projects Current year + Next Year P U B L IC IN P U T Highway Needs Inventory Local jurisdictions prepare Priority Letters SHA gets funding from MDOT Major Projects Current year + Next 5 Years SHA Business Plan SHA Performance Measures MDOT Policies SHA Call for Projects SHA Technical Analysis Draft CTP SHA evaluates Priority Letters, Some projects are excluded SHA Environmental Analysis

51 funding is programmed to carry out the project. Once plan­ ning and programming efforts have been conducted, the project then (typically) enters into the design phase. Follow­ ing the final design acceptance, the project is then constructed or deployed, and acceptance testing is performed on the final deployment. Eventually, the deployed assets will be operated and maintained for a number of years until their life expec­ tancy is met. As can be seen in Figure A.8, the overall CHART deployment process is cyclic. When the life expectancies of deployed assets are met, there comes a need for replacement assets to be deployed through a new project. The CHART Board of Directors also oversees the entire life cycle of each ITS project. This is a brief description of the high­level steps within CHART’s project planning and programming process. 1. This step in the CHART project planning and program­ ming process involves gathering information from various inputs that are both internal and external to the CHART program. One of the CHART program’s primary objec­ tives is to coordinate with other offices/agencies/partners in order to effectively operate Maryland roadways. As such, CHART has an established place within several forums and processes that involve planning/interaction with other agencies (e.g., bordering/regional states, local and county agencies, other state modal transportation agencies, pub­ lic safety agencies, emergency and medical operational agencies, among others), as well as other offices within the Maryland SHA. Like the CHART program, these part­ ner agencies also have planning processes and documented initiatives, many of which identify resources that CHART will be responsible for deploying/providing. CHART’s planning efforts, therefore, also need to account for var­ ious CHART resources allocated to support other agency initiatives. 2. Once projects are identified in the initial phase of the CHART planning and programming process, official doc­ umentation of these projects is initiated through the high­ level summary process prior to being entered into the MDOT CTP. 3. The Maryland SHA and the Office of CHART, being part of MDOT, are responsible for contributing its portion of the six­year capital investment program within the CTP. As such, the Office of CHART’s contribution to the MDOT CTP includes project titles and cost estimates to be pro­ grammed over the next six years. This includes budget Figure A.7. SHA funding breakdown in FY2011.

52 projections for each project in yearly increments. CHART updates its projects and budgets every year for submittal to the MDOT CTP, showing the latest CHART capital investment six­year projection. 4. The CHART deployment plan presents and describes capital improvement projects that the Maryland SHA’s Office of CHART is responsible for within the six­year MDOT CTP. Updated on an annual basis, the primary purpose behind the CHART deployment plan is to docu­ ment detailed information on CHART projects to receive funding for the next six years through the CTP. As a result, the CHART deployment plan directly coincides with the CHART projects for the MDOT CTP document within the CHART project planning and programming process. 5. This step involves detailed project descriptions and ITS architectures and systems engineering (SE) analysis. This level of documentation takes place once projects are docu­ mented in the MDOT CTP and the CHART deployment plan. The detailed project description and ITS architecture/ SE analysis phase is required to be carried out prior to a project going through the preliminary engineering phase (if applicable), and eventually entered into the Federal and MDOT project setup phase. 6. Once the needed project information is documented in the detailed project descriptions and/or SE analysis/ project­level ITS architecture, the project can enter the preliminary design phase where all needed details about the deployment are gathered prior to beginning the final project design. It should be noted that not all Office of CHART projects require preliminary engineering services. An example could be a situation in which specific equip­ ment will simply be procured through the project, and therefore engineering design services are not needed. In general, the most common types of projects that require preliminary engineering services are those in which ITS field devices are being deployed in new locations. 7. When preliminary engineering services are carried out and documented, the project needs to be set up in the Fed­ eral and/or MDOT project tracking systems, which track budget, payments, scheduling, and so forth. As discussed above, those projects that do not require project­level ITS architecture, SE analysis, or preliminary engineering services may be entered directly into the project setup phase. 8. Once the project is set up in the U.S.DOT/FHWA and/or MDOT project system, it can then move forward with design and deployment services. As such, the Office of CHART typically does not conduct design services for many of the projects it initiates through its planning pro­ cess, and therefore, a design request is submitted by CHART to the Office of Traffic and Safety (OOTS) in order to officially move project design and construction management services to OOTS. This step also moves the Figure A.8. Overall CHART deployment process.

53 planning and programming process into project design and deployment. Other example of a SHA Programming Process: Crash Prevention, Safety and Spot improvement, and intersection Capacity improvement Maryland has a number of additional internal project identi­ fication and programming processes, and following are three specific programs (Crash Prevention, Safety and Spot Improve­ ment, and Intersection Capacity Improvement) that follow the same general process flow, but with differing criteria for rating candidate projects. It should be noted that while these projects could have travel time reliability impacts, reliability­ based criteria or considerations are not included as part of the candidate project ratings. The current general process (which has been abbreviated) involves the following steps: 1. The SHA district offices identify a need, conduct a traffic study, and forward study to the Office of Traffic and Safety (OOTS). 2. If OOTS approves study, concept funding may be obtained through the Office of the Chief Engineer/Administrator to complete a concept development study (project impact report). 3. If OOTS approves the concept development study, a request is made for preliminary engineering funding through the Office of the Chief Engineer/Administrator. 4. Design is conducted; a plans, specifications, and estimates (PS&E) package is developed; and OOTS completes a benefit–cost analysis along with a completed rating and ranking form (criteria specific to each program is identi­ fied below). 5. Funding is requested through the Office of the Chief Engineer/Administrator and, if approved, is added to the CTP. 6. District moves forward with project and project is eventu­ ally constructed. Following is a summary of rating criteria used in evaluat­ ing projects for selection under each program as mentioned in Step 4. These criteria are used as part of a candidate project rating form that determines an overall project rating based on weighted scores within associated weighted categories. Crash Prevention Program Candidate projects are given a rating based on the categories of Safety (30%), Impacts (40%), Support/Difficulty (20%), and Congestion/Operations (10%). The percentages are the weights given to each category. The Safety category criteria include • Whether or not the improvement is on a list of previous Safety Improvement Candidate Locations • Accident experience • Police reported safety concern • Conflicts observed/reported • To what extent project improvement will address problem The Impacts category criteria include • Right­of­way and property • Historical/archaeological • Structures • Environmental (wetlands, floodplains, critical areas) • Utilities • Storm water management and drainage • Signals and lighting The Support/Difficulty category criteria include • Degree of support (from “Overwhelming Opposition” to “Overwhelming Support”) • Difficulty and associated cost (from “Difficult/Expensive [>$1M]” to “Easy/Cheap [<$500K]”) The Congestion/Operations category criterion includes • Percent change in v/c ratio in AM and PM peak hours for existing and proposed conditions Safety and Spot improvement Program Candidate projects are given a rating based on the categories of Safety (60%), Congestion/Operations (30%), and Support/ Opportunity (10%). The percentages are the weights given to each category. The Safety category criteria include • Relative position with regard to list of Safety Improvement Candidate Locations • Accident experience • To what extent project improvement will address problem The Congestion/Operations category criteria include • Need based on level of service (delay/capacity problems) • To what extent project improvement will address problem The Support/Opportunity criteria include • Degree of support (from “Overwhelming Opposition” to “Overwhelming Support”)

54 • Benefit/Cost/Difficulty (from “Expensive/Difficult” to “Cheap/Easy”) Safety and Spot improvement Program Candidate projects are given a rating based on the categories of Congestion/Operations (80%), Safety (10%), and Support/ Opportunity (10%). The percentages are the weights give to each category. The Congestion/Operations category criteria include the percentage change in the following measures of effectiveness in the AM and PM peak hours for existing conditions versus conditions after improvement: • Intersection delay • 95th percentile queue • Level of service (Highway Capacity Manual) • Volume to capacity (v/c) ratio The Safety category criteria include • Relative position with regard to High Accident Location (HAL) list • Non­HAL accident experience • To what extent project improvement will address problem The Support/Opportunity criteria include • Degree of support (from “Overwhelming Opposition” to “Overwhelming Support”) • Difficulty/Cost (from “Very Difficult/$2.5–$4M” to “Easy/<$300K”)

55 MATLAB Code A p p e n d i x B GBM Calibration and Hypothesis Testing Function: function [tt_mean,alpha,sigma,h]=gbm_calibrate(time,tt,period,corridor_name,segment_na me,L,fig_handle,axis_handle) if isempty(time) tt_mean=nan; alpha=nan; sigma=nan; h=nan; return end idx=isfinite(tt); time=time(idx); tt=tt(idx); A=[time tt]; A=sortrows(A,1); time=A(:,1); tt=A(:,2); tt_mean=nanmean(tt); [~,~,D,H,MN,S]=datevec(diff(time)); dt=D.*1440+H.*60+MN+S./60; sqrt_dt=sqrt(dt); log_inst_interest=diff(log(tt)); sigma=nanstd(log_inst_interest)/nanmean(sqrt_dt); alpha=nanmean(log_inst_interest)/nanmean(dt)+sigma^2/2;

56 % Black-Scholes formula % Input: % alpha: long-term trend (%) % sigma: instantaneous variation (%) % tau_initial: initial travel time % tau_guaranty: guarnateed travel time % optlength: option length (time) % evaltime: time at which option is to be evaluated (time) % tol: tolerance level (%) % type: 'call' or 'put' option % Output: % X: option value tleft=optlength-evaltime; d1=(log(tau_initial/tau_guaranty)+(tol+.5*sigma^2)*tleft)/(sigma*sqrt(tleft)) ; d2=d1-sigma*sqrt(tleft); if strcmpi(type,'CALL') X=tau_initial*normcdf(d1,0,1)-tau_guaranty*exp(- tol*tleft)*normcdf(d2,0,1); elseif strcmpi(type,'PUT') X=-tau_initial*normcdf(-d1,0,1)+tau_guaranty*exp(-tol*tleft)*normcdf(- d2,0,1); end end %check for log-normal distribution Y = log_inst_interest; [h,p] = chi2gof(Y,'cdf',{@normcdf,alpha,sigma},'nparams',2); if ~isempty(axis_handle) figure(fig_handle); subplot(axis_handle); histfit(Y,max(1,round(sqrt(size(Y,1)))),'normal'); if h==0 str1='CANNOT'; str2='IS'; else str1='CAN'; str2='IS NOT'; end title({ ['ALPHA: ' num2str(alpha) ' SIGMA: ' num2str(sigma)];... % ['NULL HYPOTHESIS ' str1 ' BE REJECTED (@ 5% SIGNIFICANCE LEVEL)'];... ['TRAVEL TIME ' str2 ' LOG-NORMALLY DISTRIBUTED'] }); xlabel('TRAVEL TIME LOGARITHM (LOG-MINUTE)'); ylabel('FREQUENCY'); end Black-Scholes Option Valuation Function: function X=BS(alpha,sigma,tau_initial,tau_guaranty,optlength,evaltime,tol,type)

57 p_prime=1-p; % forward binary tree development tree=nan(n+1,n+1); tree(1,1)=tau_initial; for i=1:n %time steps for j=1:i %travel time states tree(j,i+1)=tree(j,i)*u; end tree(j+1,i+1)=tree(j,i)*d; end % assigning binary probabilites to each node in the tree prob=nan(n+1,n+1); prob(1,1)=1; for i=1:n prob(1:i+1,i+1)=binopdf(i:-1:0,i,p)'; end % backward option valuation option=nan(n+1,n+1); option(:,n+1)=late_penalty*max(tree(:,n+1)- tau_guaranty,0)+early_penalty*max(tau_guaranty-tree(:,n+1),0); for i=1:n for j=1:n+1-i option(j,n+1-i)=(option(j,n+2-i)*p+option(j+1,n+2- i)*p_prime)/(1+tol*delta_t); end end C=option(1,1); % plot(tree(:,end),prob(:,end)); hold all; % xlim([0 1000]); % ylim([0 1]); end Binary Tree Option Valuation Function: function C=BinT(alpha,sigma,tau_initial,tau_guaranty,optlength,tol,n,late_penalty,earl y_penalty) % Binary Tree Option Valuation % Input: % alpha: long-term trend (%) % sigma: instantaneous variation (%) % tau_initial: initial travel time % tau_guaranty: guarnateed travel time % optlength: option length (time) % tol: tolerance level (%) % n: number of steps (whole number, positive) % late_penalty: portion of VOT traveler will be penalized for % arriving late (unitless) % early_penalty: portion of VOT traveler will be penalized for % arriving early (unitless) % Output: % C: travel time option value delta_t=optlength/n; delta_h=sigma*sqrt(delta_t); u=exp(delta_h); d=exp(-delta_h); q=.5*(1+(alpha/sigma-sigma/2)*sqrt(delta_t)); q_prime=1-q; % risk neutral probability p=(1-d+tol*delta_t)/(u-d);

58 Presentation to Maryland State Highway Administration A p p e n d i x C

59 L35B Local Methods for Modeling, economic evaluation, Justification, and Use of the Value of Travel Time Reliability in Transportation decision Making

60

61

62

63

64

65

66

67

68

69

70

71

72