Chapter 3. Findings and Results

Findings from all four phases of the project are included in this section.

Phase I: Exploration Findings

Phase I involved exploring applied methods for estimating travel behavior change and benefits with a literature scan, understanding the possible gap typologies and how they might affect benefits estimation, and engagement with subject matter experts as well as practitioner focus groups to supplement published research information. Each of these three exploratory activities are summarized in this section. The full Phase I Report is available in Appendix A.

Scan of Applied Methods

The research team relied on existing research and input from technical experts and practitioners to inform the final guide. The effort began with a review of recent research and practice focused on estimating the impacts of closing gaps in active transportation networks. The review focused on research that could best be directly transferred, including research applicable to the US context, studies using revealed preference data (vs. stated preference or hypothetical behavior), and research with specific measures that could be applied to different contexts.

Overall, based on our scan of the literature, we can say:

• There are capable frameworks for estimating a range of benefits to active travel network gap closure, although most existing applications have considered entire systems or networks instead of individual projects or gaps.
• The vast majority of existing benefit estimation has taken behavior change as either given or calculated based on crude multipliers; exceptions have relied on regional travel model outputs or primary survey data collection.
• Applications have suggested health and travel cost/travel option benefits to users dominate overall benefits.
• Network gaps have been taken as given and not identified as part of the analysis process.

Regarding travel behavior change estimation, Table 3 provides a summary of the only applications we found that produced something like the objectives of this study: an estimate of the entire pathway from network gap closure through behavior change and economic benefit estimation.

Estimated benefits ranged widely, even though projects were broadly similar in measuring a single gap closure, usually by construction of an off-street path. There was remarkable consistency in the relative importance of each benefit category, though. Health and travel costs (including enhanced travel option benefits) were measured in all cases, typically accounting for a little more than half and a bit less than half the benefits, respectively. In the few cases where additional benefit categories were considered, they accounted for a much smaller share of benefits.

Understanding Gap Typologies

Gap typologies were also examined in the literature and interviews during this phase. Table 4 provides an initial scan of factors under consideration. Individual gap typologies could include permutations of each of these factors.

From the factors considered above, the research team selected the following three factors (in order of importance along with the primary question they help answer) as the most important:

• Barrier form: Why is it a gap?
• User type: Who is it a gap for?
• Land-use context: How does the surrounding environment impact the gap?

Our approach to gap typologies and benefit calculations in this study looks both at the gap types and the data availabilities. It is important to provide minimum viable options for agencies possessing only universally available data so they can calculate benefits for the variety of gap types. At the same time, it is important to demonstrate the additional capabilities for gap definition and benefit calculation when the non-universal data options are available. Table 5 consolidates the three primary factors of barrier form, user type, and land-use context and provides examples of possible case studies. Some gap types will only

Phase II: Product Development

The objective of this phase of the project was to document the proposed methods for quantifying active transportation gap closure benefits and calculate the benefits for the selected case study locations.

After the initial scan of the literature, the team conducted a comprehensive review of active transportation research that members of the team conducted for the American Association of State Highway and Transportation Officials’ Council on Active Transportation Research Roadmap (Dill et al. 2021) as well as four additional sources identified as valuable to the work. First is the Benefit Cost Analysis Guidance for Discretionary Grant Programs (BCA Guidance) from the United States Department of Transportation (USDOT) (2023a), which relies on a range of research sources to provide estimates of benefits of transportation improvements. Three other key sources are from researchers at the University of California, Davis; one documenting their efforts to create the Active Transportation Benefit – Cost Tool for Caltrans (Fitch et al. 2021) and two providing information to California on how to improve estimates of VMT reductions from new infrastructure (Volker et al. 2019a, 2019b). We supplemented these key sources with additional research, largely from peer-reviewed journals and technical reports, along with unpublished research conducted by team members.

The existing research identified several useful frameworks for estimating a range of benefits to active travel network gap closure, although most existing applications have considered entire systems or networks instead of individual projects or gaps. Most existing benefit estimation efforts have taken active travel behavior change as either given (e.g., assume that bicycling accounts for 5% of all trips) or calculated based on crude multipliers (e.g., triple the number of pedestrian trips); exceptions have relied on regional travel model outputs or primary survey data collection. The review also suggested that health and travel cost/travel option benefits to users make up a large share of overall benefits. The next section includes details of how we used specific research findings to inform quantitative estimates for travel behavior change in the guide.

Behavior Change Technical Details

To support behavior-change estimation, we developed a range of general factors that blended various existing tools—particularly the Caltrans Tool, which provided the initial tables and estimates—and adapted research findings. We labeled these “Caltrans” and “Research” estimates. Because evidence for responses to many pedestrian and bicycle infrastructure types has been sparse to date, we tried to make use of all available evidence, rather than pick a single “best” source. The wide range of supporting studies and methodologies made construction of confidence intervals or precision estimates impossible. Users are encouraged to test the sensitivity of their analyses to different input assumptions and to consider results at different response levels (low, medium, high) to better understand behavior-change uncertainty. Appendix A: Behavior Change Technical Details in the final guide documents the construction of mode and route shift percentage change estimates (low-capacity behavior change) and network impedance factors (medium-capacity behavior change).

Percentage Change Estimates

We relied on five primary research sources to augment the existing Caltrans estimates for bicycle behavior change:

1. Broach et al., 2012 - revealed preference (RP) bicyclist route choice model
2. Broach and Dill, 2016 - RP mode choice model, including walking and biking
3. McNeil et al., 2015 - stated preference (SP) study of bike lane treatments

4. Broach and Dill, 2015 - RP pedestrian route choice model
5. Sevtsuk et al., 2021 - RP pedestrian route choice model

From the RP studies, we calculated elasticities from the published models that roughly corresponded to the various combinations of facility type and degree of improvement over existing conditions (low, medium, high). Choice elasticities were calculated as arc elasticities from the multinomial logit (MNL) models presented using the usual formula:

where, E_P,i|dX is the elasticity of the probability (P) of choosing alternative i given a change in attribute X, and B_X,i is the MNL parameter estimate of attribute X for alternative i. With certain exceptions as noted in subsequent tables and text that follows, general assumptions were:

• Set X equal to 0.5 (the midpoint of the arc between not having the facility and having it).
• Set initial probability of selection P at 0.5 in all cases (equivalent to assuming, e.g., current option captures half the potential bike volume or half the relevant trips) for convenience and to provide generally conservative estimates (since setting low selection probabilities inflates the elasticity calculation).
• Assumed no detour required to use new facility.
• Structured Basic and Advanced methods such that each step was roughly equivalent to a single “step” in quality (e.g., for an off-street path, “low” would reference an on-street bike lane, “medium” moderate traffic minor arterial without a bike lane, and high a major arterial without bike lane).
• Where separate parameters were provided for commute and non-commute cycling, weight by approximate share of bike trips in most recent NHTS (2017) for each purpose, rounded to 20% commute and 80% non-commute.
• Research-based mode shift estimates did not account for new trips. We applied the Caltrans Tool’s estimate of new demand to inflate the calculated elasticities (1.3x mode shift for bike, 1.1x mode shift for walk). E.g., if calculated mode shift for bike was 50%, we used 50% * 1.3 = 65% as the research-based estimate.

Resulting elasticities under different scenarios were used directly as the research-based estimates for percent change (e.g., a calculated elasticity of 0.3 was interpreted as an expected 30% increase in cycling at a location). Where available, route and mode shift elasticities were added together to produce a total expected percent change. Table 6 and Table 7 describe the sources and calculation results for estimated impacts of each gap-closing facility type. Where both Caltrans and research estimates were available, results were averaged to produce a final percentage change estimate.

The Augmented Method for percent change estimation (Appendix C in the NCHRP Research Report 1149) requires an additional assumption that original research relating network quality to shifts in bicycling will apply similarly to shifts in walking from pedestrian network improvement. The research team consulted data from original research to explore whether pedestrian sensitivities to route quality potentially should be less than this elasticity, based on findings that people walking are less willing to shift routes modes (Broach and Dill 2012; Broach and Dill 2015; and Broach and Dill 2016). Instead, we found it could be justified to be higher, in relative terms. To remain conservative, we use an assumed elasticity of 2.37 for both modes.

Shift Proportion Estimates

The Caltrans Tool includes suggested shift change proportions attributable to different sources for both ped and bike projects: route, shift from vehicle modes, shift from other modes, new travel. We adjusted these proportions in two ways from the original:

• Where no significant route shift effect expected, re-allocate among remaining shift categories total 100%. This avoids underestimating the impact of projects with primarily mode shift potential (e.g., a new bridge without existing alternatives).
• Where research estimates were available for both route and mode shift:
  • Calculate the implied route shift allocation as route shift / (total shift) and the remaining categories as 100% - route shift.
  • Allocate among remaining based on Caltrans shift proportions.
  • Average these new estimates with those of other facility types to arrive at a general average shift allocation.

Impedance Adjustment Factors

Level of traffic stress (LTS) provides a method for rating segments (and, more recently, intersections) into discrete categories thought to correspond to perceived stress or network quality by different cycling subpopulations. The discrete categories lead to deterministic, all-or-nothing networks that do not map easily to conventional network impedances (time or perceived cost of travel), creating issues for traditional network analysis. To create a more flexible network definition that better aligns with the low capacity, percentage change option, impedance adjustment factors were created by mapping existing RP route choice research models to LTS categories.

LTS categories were mapped to research model variables based on a mix of overlapping definitions along with some qualitative judgment about the intended meaning of a given LTS category; e.g., “comfortable for people of all ages and abilities.” There are currently a limited range of facility attributes that have been modeled, and LTS definitions include multiple factors, making one-to-one matching infeasible. Users are encouraged to consider their unique local context, and to seek out updated research that might update the mappings or impedances provided here.

Once a mapping from LTS to research models was established, impedances were established based on the marginal rate of substitution (MRS) between route distance and a given segment or intersection improvement. Distance could then be converted to travel time based on assumed base speeds for a given travel mode. The resulting MRS-based distance/time values represent the combined effect of delay and perceived cost relative to a reference case. The reference case is typically an off-street or very low volume and low speed facility with adequate bicycle or pedestrian infrastructure. An example impedance example is provided:

• Context: biking on an off-street path vs. an on-street, conventional striped bike lane;
• Relevant research model MRS: 0.24 (i.e., cyclist willing to ride 24% farther to use the better facility);
• Impedance factor: 1.24 (i.e., 1 mile on a striped bike lane is equivalent to 1.24 miles on an off-street path); note that it does not matter whether this is applied to travel speed/time or distance, since travel speed/time is fixed for each mode.

Table 8 and Table 9 provide the specific mappings used to generate the impedance values from various research model MRS estimates. The mappings should be used as a guide to customize factors as needed to suit a specific project context. For example, a protected bike lane might use the LTS 1 value (1.0) or an

Health Benefits Estimation - Technical Details

NCHRP Research Report 1149 provides instructions on quantifying the health benefits of increased physical activity related to behavior shifts from closing gaps in the active transportation network. Both HEAT (developed by the World Health Organization, WHO) and the Integrated Transportation and Health Impact Model (ITHIM) are tools that can estimate the value of improved community health through increases in walking and bicycling. The underlying calculations for these tools rely on established research linking regular physical activity at various intensity levels (metabolic equivalent of task, or METs)—including walking, bicycling, and, more recently, micromobility—to changes in health outcomes measured as risk of premature death (mortality) or living with chronic disease (morbidity). While capturing an important component of total health impacts from increased active travel, many additional pathways exist linking transportation shifts to health outcomes (e.g., safety improvements leading to reduced injuries and fatalities).

There are three levels of methods included in NCHRP Research Report 1149:

• Basic Method based on FHWA’s BCA Guidance for Discretionary Grant Programs, see the Final Guide Step 3, Health Benefits section.
• Augmented Method based on HEAT’s active travel module, see the Final Guide, Appendix L.
• Advanced Applications based on ITHIM’s physical activity impact estimation module, see the Final Guide, Appendix M.

Each option follows the same basic principle: changes in walking and biking affect health outcomes by changing the likelihood of disease mortality or morbidity. The dose-response relationships are defined by the guidance or tool. As capacity increases, more localized and disaggregated inputs and outputs are possible.

Safety Benefits Estimation - Technical Details

Existing methods are well established (e.g., 2022 BCA Guidance), though they have their limitations. The 2022 BCA Guidance method involves applying crash modification factors (CMFs) to a baseline number of crashes calculated from 3-7 years’ worth of crash history. Per the official guidance, crash history should be as localized to the project site as possible, not derived from areawide totals. The estimated number of crashes reduced is then monetized using standard published values.

There are three levels of safety methods included in the NCHRP Research Report 1149:

• Basic Method uses crash history or crash estimates obtained via multipliers, see the Final Guide Step 3, Safety Benefits section.
• Augmented Method uses crash history or crash estimates derived from local crash profiles, see the Final Guide, Appendix E.
• Advanced Applications uses Safety Performance Functions (SPFs) to estimate crash frequencies based on existing roadway information, see the Final Guide, Appendix F.

General limitations of existing methods for active transportation safety benefits include:

• Limited to treatments that already have a CMF – many newer ped/bike treatments that might be used in gap closure may lack CMFs.
• Can be challenging to choose which CMF to use if multiple exist – based on crash type, crash severity, modes involved, evidence quality, etc.

• Gap closures may entail several different treatments, each with its own CMF. Methodology for combining CMFs is still fuzzy, although the generally accepted practice is to multiply them.
• Using only crash history is inherently reactive. Note that USDOT BCA Guidance explicitly encourages this: “The baseline data should be closely aligned with the expected impact area of the project improvements, rather than reflecting outcomes over a much larger corridor or region.”
• Only using crash history can be problematic in multiple types of locations – where conditions are so poor that people avoid walking and bicycling, resulting in low exposure and few crashes, and where the gap closure involves forging an entirely new network connection, so there is literally no crash history available for the location.
• Only using crash history leads to a possible underestimation of safety benefits, since we multiply crash modification factors by a low number of existing crashes (due to low exposure).
• That said, one could argue that the safety benefit is expressed through increase modal usage and related downstream benefits, and that estimating a benefit from hypothetical crashes avoided is “double dipping.” This may be why BCA guidance specifically wants localized crash history instead of areawide estimates.
• Monetizing safety benefit requires crash reduction estimate to be for specific severity levels, which may or may not be possible based on treatment(s) in question and available CMFs.

Emissions Benefits Estimation - Technical Details

Emissions benefits of gap closure projects are driven by mode shifts, and specifically, shifts from driving to active travel induced by the gap closure, which result in reductions in VMT. Mode shifts could be induced by any type of gap closure. Emissions benefits are most likely to be realized when the closure mainly serves utilitarian trips (e.g., commuting, shopping) that might be substituting for car trips, rather than recreational trips that are generally not substituting for driving.

There are three levels of methods included in NCHRP Research Report 1149:

• Basic Method based on estimate of mode shift due to the gap closure, national average trip lengths, and national emission rates per vehicle-mile, see Final Guide Step 3, Emissions Benefits section.
• Augmented Method based on local or regional data rather than national data, see Final Guide, Appendix H.
• Advanced Applications Method based on using a model for estimating emissions, see Final Guide, Appendix I.

Enhanced Amenities Benefits Estimation - Technical Details

Enhanced amenities benefits measure a reduction in generalized travel cost due to improvements in one or more travel mode options. Generalized travel costs typically include out-of-pocket expenditures and time costs but may also capture quality improvements experienced by the user, such as reduced crowding on transit or the installation of lower-stress active transportation facilities like protected bike lanes or off-street paths.

There are three levels of methods included in NCHRP Research Report 1149:

• Basic Method based on USDOT BCA Guidance (USDOT 2023a), see Final Guide Step 3, Enhanced Amenities section.

• Augmented Method is also based on USDOT BCA Guidance (USDOT 2023a), see Final Guide, Appendix J.
• Advanced Applications based on using a regional travel demand model, see Final Guide, Appendix K.

Local Economic Benefits Estimation - Technical Details

There have been many studies of the economic impacts of active transportation facilities, examining impacts of 1) increased spending by active transportation tourists; 2) increased business sales or spending based on mode or destination shifts of local travelers; and 3) impacts of facilities on property values.

These studies have used a variety of methods, most commonly visitor surveys; time-series analysis comparing business activity before and after an improvement and/or in improved vs. non-improved areas; and hedonic analysis of property values. These studies have not found consistent impacts across regions or projects; instead, the research and findings tend to be context-specific and not often generalizable. Studies have not found consistent impacts across regions or projects. As a result, for the Basic Method we do not present multipliers that can be used universally; instead, we recommend only discussing the local economic benefits qualitatively.

There are three levels of methods included in NCHRP Research Report 1149:

• Basic Method based on summarizing findings from other studies to qualitatively assess the potential benefits.
• Augmented Method based on applying factors from other studies to develop rough quantitative estimates of visitor spending, while describing other impacts qualitatively, see Final Guide, Appendix N.
• Advanced Applications involve original data collection that will provide an assessment specific to the project and its context, see Final Guide, Appendix O.

Identified Challenges

The research phases revealed several challenges that are difficult to directly address through the methods presented in the guide:

• Differences in evidence for modal behavior change. The existing research on behavior change is much more developed for conventional bicycling. Research on walk, e-bike, and micromobility behavior-change estimates is sparse, and research on responses to Americans with Disabilities Act accessibility improvements is even more limited.
• Uncertainty. The estimates in this guide are just that—estimates. There is an unknown level of uncertainty or error associated with every estimate that users should consider when applying the methods and conveying results. Users should check for updates in documents referenced, such as the USDOT BCA Guidance. Users can also conduct both conservative and optimistic evaluations using lower and higher estimates of behavior change or other benefits estimates. Step 4 of the Final Guide provides some ideas for communicating uncertainty in analysis results.
• Destination shifts. The guide’s process for estimating behavior change includes the generation of new trips, trips shifted from other modes, and trips shifted from other routes. In general the guide does not account for the potential that people will choose different destination locations as a result of the gap closure. For example, a person who would normally drive to a grocery store several miles away might bike to a closer store after a key gap is closed. Alternatively, someone

There were some areas of substantive technical feedback, including:

• Reviewers felt our approach to limiting benefit calculations to the most primary effects was generally the correct approach; however, they felt it would be helpful to users to make this more explicit, especially in cases where a user might only apply a selection of the methods.
• There was consensus that, despite their known limitations, CMFs have no viable substitute.
• While recognizing the limitations of conveying uncertainty in a static document, reviewers noted some ways we might better convey the uncertainty underlying methods or specific parameter values, and how users should consider that uncertainty when analyzing or presenting results. There was agreement that the most effective option was to raise awareness of uncertainty throughout the guide, and to provide concrete examples of impacts on results where possible.
• The definition of areas influenced by a gap closure is a key input to many benefit calculation methods. Commenters agreed that this was a difficult problem with no easy solution in practice. Reviewers further noted that influence areas in safety analysis posed a particularly difficult problem.
• In the Phase II Report, we considered the following behavioral shifts as part of benefit calculation: shift from driving modes, shift from non-driving modes, route shift, and induced travel (new trips). We did not explicitly consider shifts in destination. Commenters noted that shifts in destination for the same trip—for example, replacing a longer driving trip with a shorter walk or bike trip—could affect benefit calculations. Reviewers noted that the errors could go both ways, leading to under- or over-estimation of benefits. Overall, it was agreed that destination choice impacts here were interesting but likely marginal.
• Reviewers voiced some reservations around reliance on LTS as a primary measure of network change. The consensus was that LTS has become a de facto standard and, therefore, useful as a starting point.

Additional details on the technical vetting results can be found in Appendix B: Phase III Report, which will be published separately on the project webpage. Appendix B: Phase III Report can be found on www.nationalacademies.org/publications by searching for NCHRP Web-Only Document 426: Developing a Guide for Estimating Benefits of Closing Gaps in Active Transportation Networks and looking under “Additional materials.”

Usability Testing Results

We invited eight transportation planning practitioners to review various sections of the guide and provide feedback to shape the guide’s look and feel. The selected reviewers represent regional, state, and federal transportation agencies.

Reviewers felt that the methods presented in the Phase II Report were useful but somewhat dense and text-heavy. They thought the level of detail is appropriate for practitioners using the guide but could be made more accessible to a wider audience by adding blank space, including additional visuals, highlighting key concepts with callout boxes, and using internal document links to make cross-referencing content from other parts of the document easier.

There was also consensus that the guide should focus on low-capacity methods and include more detailed methods in an appendix while retaining enough information about the medium- and high-capacity methods so that people understand what sort of analysis is available to them, why it exists, and how it should be used.

Overall, they found the low-capacity methods to be presented clearly, with an appropriate level of complexity for each capacity level. However, the medium-capacity methods were perceived as less accessible to the average reader. Reviewers suggested we add explicit notes to clarify that the high-capacity methods mentioned are provided as examples, and individuals capable of executing such methods should be able to develop their own analysis techniques. Furthermore, they recommended that the guide include a more explicit discussion of both analysis and data limitations.

Reviewers generally felt that terminology was explained well within the document but said the guide could benefit from defining terms more than once.

Reviewers made several recommendations about visuals, which will be addressed during guide development. First, the final visuals should be original vector graphics rather than screen captures from external sources. The report, which is currently text-heavy, would benefit from the inclusion of additional graphics.

The case studies were well received and contributed to the overall understanding of the analysis methods. Reviewers thought the length of the case studies was generally appropriate and struck an appropriate balance between the amount of information presented and the reader’s engagement time. Reviewers did suggest adding photos of the roadway and a small amount of narrative describing existing conditions and proposed gap closure methods to help users contextualize each gap.

Additional details on the usability testing results can be found in Appendix B of the Phase III report.

Phase IV: Implementation

The final guide is a resource on best practices for estimating the benefits of closing gaps in active transportation networks. It has been designed for use by practitioners, including at state DOTs, as well as regional and local governments. It provides guidance on identifying gaps, estimating the benefits of closing gaps, and communicating the results. Based on the reviews in Phase III, and summarized in Appendix B: Phase III Report, major changes were incorporated into the final guide.

First, and most importantly, the entire guide has been streamlined to only include the low-capacity methods for estimating benefits. The medium- and high- capacity methods were moved to appendixes, for reference by organizations with greater capacity. Then nomenclature for different capacity methods has also changed from low-, medium-, and high-capacity methods to Basic, Augmented and Advanced Application Methods for additional clarity in the guide.

Other additions and changes include:

• New introductory narrative;
• Walk-through calculation examples;
• Editing for brevity and clarity, and readability throughout the document;
• Critical concepts, such as LTS, are highlighted and presented early enough in the document to provide context for subsequent analysis;
• New section on how to communicate the results of quantifying benefits;
• Simplifying the process for estimating benefits, as well as identifying gap typology;
• New visuals and graphics to explain critical concepts and each method of estimating benefits;
• Summaries of case studies were highlighted in callout boxes for more visual interest.

Overall, NCHRP Research Report 1149 has been organized in the following fashion. The guide has been built to allow any user to quickly understand the needed data, assumptions, and limitations as well as final outcomes of each type of benefit estimation. It is intended to be used sequentially, starting with Step 1 and continuing through Step 4. The Steps in Action section provides walk-throughs of the methods with example calculations, and the last section of the final guide contains references.

Introduction - This section overviews the purpose of the guide, defines key terms, describes the research behind the guide and describes how to use the guide.

Step 1. Understanding the Gap Closure - This section describes identifying the gap for analysis, determining the gap type, describing the gap closure, and thinking ahead to communicating the results of the analysis.

Step 2. Estimating Behavioral Change - This section involves choosing an appropriate methodology and assembling data to quantify the change in active travel due to the gap closure. This includes understanding the impacts on route shift, mode shift, and trip generation.

Step 3. Estimating Benefits - This section involves choosing specific benefits that apply to the project and quantifying the benefits that can be expected from the gap closure. We describe each type of benefit and then provide an overview of the methods to choose, data needed, and detailed methods for analysis.

Step 4. Communicating the Results - This section helps users of this guide properly communicate the estimated benefits, including considerations for different audiences.

Steps in Action (Analysis Walk-Throughs) - This section provides two examples of fictional gap closure projects and outlines how benefits would be estimated for these projects.

References – This section lists the many references used in the creation of this guide and supporting materials.