Skip to contentUnited States Department of Transportation - Federal Highway Administration FHWA Home
Research Home
Report
This report is an archived publication and may contain dated technical, contact, and link information
Publication Number: FHWA-RD-98-165
Date: July 1999

Guidebook on Methods to Estimate Non-Motorized Travel: Overview of Methods

3.0 Guide to Available Methods

 

3.1 Overview of Methods

This section describes eleven types of quantitative methods that can be used to forecast non-motorized travel demand or that otherwise support the prioritization and analysis of non-motorized projects. These methods are categorized according to four major purposes, as shown and described in table 3.1. Figure 3.1 illustrates how these four purposes relate to each other to support demand estimation. Following the overview, section 3.2 summarizes key characteristics of the methods. Section 3.2 also suggests appropriate methods according to specific purpose such as forecasting the number of new users of a bicycle/pedestrian trail.

Table 3.1 Categorization of Available Methods.

Purpose Method Description
Demand Estimation. Methods that can be used to derive quantitative estimates of demand.
     
  Comparison Studies Methods that predict non-motorized travel on a facility by comparing it to usage and to surrounding population and land use characteristics of other similar facilities.
  Aggregate Behavior Studies Methods that relate non-motorized travel in an area to its local population, land use, and other characteristics, usually through regression analysis.
  Sketch Plan Methods Methods that predict non-motorized travel on a facility or in an area based on simple calculations and rules of thumb about trip lengths, mode shares, and other aspects of travel behavior.
  Discrete Choice Models Models that predict an individual's travel decisions based on characteristics of the alternatives available to them.
  Regional Travel Models Models that predict total trips by trip purpose, mode, and origin/destination and distribute these trips across a network of transportation facilities, based on land use characteristics such as population and employment and on characteristics of the transportation network.

Table 3.1 Categorization of Available Methods (continued)

Purpose Method Description
Relative Demand Potential Methods that do not predict actual demand levels, but which can be used to assess potential demand for or relative levels of non-motorized travel.
  Market Analysis Methods that identify a likely or maximum number of bicycle or pedestrian trips that may be expected given an ideal network of facilities.
  Facility Demand Potential Methods that use local population and land use characteristics to prioritize projects based on their relative potential for use.
Supply Quality Analysis Methods that describe the quality of non-motorized facilities (supply) rather than the demand for such facilities. These may be useful for estimating demand if demand can be related to the quality of available facilities.
Bicycle and Pedestrian Compatibility Measures Measures that relate characteristics of a specific facility such as safety to its overall attractiveness for bicycling or walking.
  Environment Factors Measures of facility and environment characteristics at the area level that describe how attractive the area is to bicycling or walking.
Supporting Tools and Techniques Analytical methods to support demand forecasting.
  Geographic Information Systems Emerging information management tools, with graphic or pictorial display capabilities, that can be used in many ways to evaluate both potential demand and supply quality.
  Preference Surveys Survey techniques that can be used on their own to determine factors that influence demand, and that also serve as the foundation for quantitative forecasting methods such as discrete choice modeling.

 

Figure 3.1 Relationship of Methods Supporting Demand Estimation
Figure 3.1: Relationship of Methods Supporting Demand Estimation

 

For each of the 11 methods, a one-page summary is provided which includes an overview of the method, typical applications, advantages and disadvantages, and one or two real-world examples. Each summary also includes a quick reference guide, which provides a subjective rating of the method for five factors as described below. The ratings are provided only as a general assessment of the method's capabilities, and the quality of specific applications of each of these methods may vary. More detail on the specific ratings for each method is given in table 3.2, which follows the individual method overviews.

The five factors and criteria used to rate the factors are as follows:

  • Ease of Use - "Easy" if the method could be applied by a layperson with basic research and data analysis capabilities; "difficult" if the method requires extensive specialized training to understand and apply.

  • Data Requirements - "Minimal" if the method primarily uses existing data that can easily be collected and evaluated; "extensive" if it requires significant new data collection efforts.

  • Accuracy - "Low" if forecasts have not corresponded well to observations; "high" if forecasts have been found to closely reflect actual demand.

  • Sensitivity to Design Factors - "Low" if the method cannot assess the impacts of specific design factors on demand; "high" if the method can assess the impacts of multiple factors and the interactive effects of these factors.

  • Widely Used - "No" if only a few applications have been identified; "yes" if the method has been widely used in practice.

Finally, the overview page indicates whether the method can be used to predict demand at the facility level, area/regional level, or both. Facility-level methods predict the number of users of a specific facility such as a non-motorized trail, bicycle lane, or pedestrian bridge. Area-level methods predict total bicycle or pedestrian trips for an entire area such as a city, census tract, or other geographic area.

Section 2.0 of Supporting Documentation presents a more indepth, structured description of each method as well as specific variations and applications of the method. Section 3.0 contains bibliographic references for the real-world examples highlighted in this section. Section 4.0 identifies useful contacts, including individuals and organizations, in the area of non-motorized travel estimation.

Demand Estimation: Aggregrate Behaviour Studies

Overview

The simplest form of demand forecasting, comparison studies compare usage levels before and after a change (such as a facility improvement), or compare travel levels across facilities with similar characteristics. The results of a comparison study can be used to predict the impacts on non-motorized travel of a similar improvement in another situation, assuming that all other influencing factors are roughly the same between the two situations.

 
Typical Applications

Before-and-after studies have been widely used in Europe to assess the mode choice impacts of programs to improve bicycle and pedestrian facilities. Some studies have focused on the change in mode split for an urban area as a whole, after a city-wide pro- gram of improvements. Others have focused on specific facilities, conducting user counts both before and after an improvement to the facility. Comparison studies have also been performed in the United States, using counts from existing trails to forecast the number of users on a new trail.

 
Advantages This method is simple to understand and relatively easy to apply.  
Disadvantages

Comparison studies only provide a rough estimate of demand for proposed facilities. Unless very carefully designed, comparison studies may not control for other factors unrelated to the facility improvement which may affect usage levels. It is often difficult to find truly comparable facilities. Because of possible differences in situations, trans- transferring results from one situation to another may lead to incorrect usage forecasts.

 

 

Central Massachusetts Rail Trail Bikeway
To estimate the potential usage of a proposed rail trail in Massachusetts, planning staff conducted bicycle counts on an existing trail which has characteristics similar to the proposed facility. These counts were then factored based on the ratio of total population within the corridors surrounding the two facilities to predict total trips on the proposed facility. Total volumes were distributed throughout the proposed corridor based on the population of communities along the corridor. An alternative method was also applied in which usage of the existing trails was factored by the ratio of bicycle commuting mode share in the two corridors, as determined from census data (Lewis and Kirk, 1997).
  Comparison of Trails in Australia
Wigan (1997) compared the characteristics of users and the surrounding population on two existing facilities in Australia. Trail users were surveyed regarding mode of access to the trail, access distance, and personal characteristics. Data on population in the surrounding area were also analyzed. The results indicate that the Lower Yarra trail attracted more users from a wider range of distances than the Lower Maribrynong, despite similar levels of surrounding population. The authors concluded that with better signage, improved linkages, and promotional efforts for the Lower Maribrynong facility, usage could be comparable to the Lower Yarra trail.

Demand Estimation: Aggregate Behaviour Studies

 

Overview Aggregate behavior studies involve the development of models to predict mode split and/or other travel behavior characteristics for an aggregate population, such as residents of a census tract or metropolitan area. Prediction is based on characteristics of the population and of the area. An example of an aggregate model is an equation to predict the percentage of trips taken by bicycle in individual census tracts in a metropolitan area, based on the average income of the tract and on the total length of bike- ways in the tract.

Typical Applications Aggregate behavior studies have been conducted in the United States and the United Kingdom, primarily utilizing census data and other readily available data sources to predict work-trip mode split at a tract, city, or metropolitan-area level.

Advantages Aggregate behavior models have isolated some factors that can be related to non- motorized travel and have developed quantitative relationships between these factors and modal split. Also, the results of these studies are potentially useful for the trip generation component of regional travel models which include non-motorized modes.

Disadvantages Aggregate behavior models have generally had low explanatory power and have not been successful at predicting mode splits when applied to other areas. Predicting behavior at an aggregate level suffers from a number of significant difficulties, including: (1) aggregate level data can mask significant variances within a population which affect behavior, e.g., the average income of a census tract may be much less important than the distribution of income; (2) the method ignores the impact of factors which are not readily available, such as attitudinal factors; (3) the primary data source on trips at a zonal/aggregate level is the census, which looks only at work trips; and (4) the available data generally do not include environmental variables which describe the overall quality of the area for bicycling or walking, the overall quality of alternative modes, etc.

Bicycle Journey-to-Work in the UK
Ashley and Banister (1989) used UK census and other data to (1) evaluate factors influencing cycling to work, (2) develop a model to predict the proportion of residents in a ward bicycling to work, and (3) test the model. The authors used regression analysis to test the effects of various factors on the proportion of ward residents cycling to work. Factors tested included personal characteristics, trip distance, avail- ability of cycling facilities, avail- ability of other modes, modes, traffic levels, and local climate and topography.
  Bicycle Mode Split in U.S. Cities
Nelson and Allen (1997) conducted a cross-sectional analysis of 18 U.S. cities to predict work trip bicycle mode split (from census data) based on weather, terrain, number of college students, and per capita miles of bikeway facilities. A positive association was found between the presence of bikeway facilities and bicycle work trip mode split.

Demand Estimation: Aggregate Behaviour Studies

Overview Sketch plan methods can be defined as a series of simple calculations to estimate the number of bicyclists or pedestrians using a facility. These methods generally rely on data that already exist or can be collected with relative ease (such as census and land use data), and can be combined with behavioral assumptions derived from other studies. Sketch plan methods vary widely in their specific approaches and in their level of sophistication.

Typical Applications A variety of pedestrian sketch-plan methods have been developed to estimate pedestrian volumes under existing and future conditions in a pedestrian activity area, such as a central business district or shopping center. These methods generally use pedestrian counts and regression analysis to predict pedestrian volumes as a function of adjacent land uses and/or indicators of transportation trip generation (parking capacity, transit volumes, traffic movements, etc.) Alternatively, data on surrounding population and employment may be combined with assumed trip generation and pedestrian mode shares to estimate levels of pedestrian traffic. At least one bicycle sketch plan method has also been applied to predict usage of a new bicycle lane in Seattle. This method relies on census data and simple travel survey data to estimate the travel impact of the project.

Advantages Sketch plan methods tend to be relatively simple to understand and to apply. If the methods and data are selected carefully, they may give reasonable estimates of the number of users of a proposed facility. These methods are best for developing rough estimates for planning purposes and for comparing potential usage levels among facilities or areas to prioritize actions.

Disadvantages Sketch plan methods tend to rely on limited local data and on general assumptions about behavior. Therefore, they can be imprecise and may not account well for specific local conditions such as characteristics of the facility, network, surrounding population, destinations, or competing modes of travel. In addition, methods and assumptions developed for specific applications may not always be relevant to applications in other geographical areas.

Estimating Pedestrian Corridor Activity
Matlick (1996) describes a method to determine the level of pedestrian activity in 0.8 km buffer areas in specific corridors. A variety of sources was used to estimate activity within the corridor: population, mode split, and trip characteristics from census and National Personal Transportation Survey data; land use data from local data bases; and estimates of school and transit trips.
  Estimating Peak Pedestrians per Hour
Ercolano (1997) describes a method that determines site, corridor, and subarea pedestrian per hour volumes using local vehicle per hour turning movements and mode share census data (at a minimum). Other features of this method include the ability to estimate sidewalk and intersection trips and the ability to adjust trips based on completeness of pedestrian infrastructure and climatic conditions.

Demand Estimation: Aggregate Behaviour Studies

Overview A discrete choice model predicts a decision (choice of mode, choice of route, etc.) made by an individual as a function of any number of variables, including factors that describe a facility improvement or policy change. The model can be applied across a population to estimate the total number of people who change their behavior in response to an action. The model can also be used to derive elasticities, i.e., the percent change in bicycle or pedestrian travel in response to a given change in any particular variable.

Typical Applications Discrete choice models are widely used by regional travel modelers to predict auto vs. transit mode choice. Mode choice models have also been developed that include bicycling and walking as options; a model was recently developed in Chicago to predict the impacts of pedestrian and bicycle improvements on transit access mode (see sidebar). Discrete route choice models have also been developed for bicyclists which model bicyclists preference for various facility design features when selecting a route.

Advantages Discrete choice models based on local survey data are the most accurate tool available for predicting travel behavior impacts. These models can be a powerful tool for isolating and quantifying the effects of specific factors, both personal and environmental, on travel behavior. They can also be used to examine the interaction of each factor with other factors, e.g., whether age has an impact on the type of facility preferred.

Disadvantages Development of a discrete choice model generally requires the collection of extensive survey data and requires expertise in discrete choice modeling techniques. Also, since the number of factors (facility design, personal, etc.) which can be considered in any particular modeling exercise is limited, it is not possible to identify or control for all factors which may influence behavior. Furthermore, a model developed for a specific situation may not be applicable to other situations if important factors not considered in the model differ between the two situations.

Transit Access Mode Choice in Chicago
The Chicago Regional Transit Authority recently developed a set of discrete choice models to predict the impacts on transit access mode of bicycle and pedestrian improvements to rail station areas in Chicago. Surveys to deter- mine existing commuters mode choice, station access distance, and other characteristics were used in conjunction with visual simulation surveys to estimate whether people would shift to non-motorized access modes as a result of various improvements. Bicycle improvements tested included removal of debris, provision of parking, slowing of traffic, and development of curb lanes, paths, and bicycle routes. Pedestrian improvements tested included sidewalks, recreation paths, slowing of traffic, and various improvements to intersection crossings (Wilbur Smith Associates, 1997).
Demand Estimation: Aggregate Behaviour Studies

Overview Regional travel models, commonly referred to as four-step travel demand models, use existing and future land use conditions and transportation network characteristics, in conjunction with models of human behavior, to predict future travel patterns. These models are described in more detail in section 2.4 of this overview and section 2.8 of the supporting documentation.

Typical Applications Traditionally, regional travel models have been oriented toward predicting trips by automobile and transit. However, a number of models in the United States, Canada, and Europe have recently been modified to estimate non-motorized mode splits based on ratings of the pedestrian friendliness or bicycle friendliness of individual zones. Some models have also been modified to include bicycle and/or pedestrian facility networks and to predict the route choice impacts of improving or adding facilities. Models have also been developed specifically for bicycle or pedestrian travel. For example, in the 1970s pedestrian demand models were developed for various commercial business districts in the United States. These models related pedestrian trips to land uses at a block level and assigned trips between blocks based on characteristics of the pedestrian network.

Advantages Regional travel models have been developed for all major urban areas in the United States. The regional travel model structure provides an integrated framework for analyzing travelers choices between modes. Given sufficient data collection and enhancements to the model structure, regional travel models could serve as a powerful tool for analyzing bicycle and pedestrian travel. Regional travel models can also serve as a source of data, such as total trips generated in an area, which are useful for other bicycle or pedestrian modeling or sketch-planning efforts.

Disadvantages The current generation of regional travel models was developed at a spatial scale appropriate for automobile rather than bicycle or pedestrian travel. Also, incorporation of non-motorized modes may require significant data collection to create a zone-level "environment factor" or develop a network of bicycle and pedestrian facilities. Current regional travel models also do not consider trips made for the sole purpose of recreation. Finally, the development and modification of travel models require considerable expertise and the use of specialized software packages.

Edmonton Transport Analysis Model (Canada)
The Edmonton Transport Analysis Model recently developed for the Edmonton, Canada region includes both walk and bicycle as separate modes and also includes bicycle network characteristics in determining mode choice. Links in the network model can be coded in three ways: bicycle path, bicycle lane, or mixed traffic. Bicycle travel time on each link is adjusted by a factor representing the relative onerousness of bicycling by facility type. These factors are derived from a hypothetical choice survey of bicyclists in which bicyclists are asked to choose between different routes based on distance, facility type, and other factors (Hunt, Brownlee, and Doblanko, 1997).
Demand Estimation: Aggregate Behaviour Studies

Overview This is a general approach which estimates the maximum potential number of trips by bicycle or walking in an area, based on (1) current trip length distributions, usually by trip purpose; (2) rules of thumb on the maximum percentage of bicycling or walking trips by trip distance and purpose; and/or (3) the percentage of the population likely to switch to bicycling or walking, based on identifying a target market of bicyclists or walkers according to commute distance, demographic characteristics, etc. An ideal network of facilities is assumed, i.e., this method estimates how many trips might take place if the quality of facilities was not an issue.

Typical Applications Market analysis is a relatively common approach that can be applied in many different ways, with varying levels of detail. Some studies have taken aggregate data on trip lengths by purpose for an area and applied a rule of thumb about the maximum bicycle or walk trip length, in conjunction with a best guess as to the likely mode share diversion, to estimate the potential bicycle or walk mode share. Others have focused on defining the demographic characteristics of people most likely to walk or bicycle, and subsequently using demographic information for an area, in conjunction with trip length distributions, to obtain an overall maximum potential mode split under ideal conditions.

Advantages Market analysis methods generally define an "upper bound" on the number of trips by cycling or walking and may therefore give municipalities a target to shoot for in developing plans to improve facilities city-wide. This type of analysis can also be helpful in identifying areas of greatest potential demand, as an aid to prioritizing projects.

Disadvantages Market analysis methods are intended only to achieve rough estimates of the maximum number of trips that could be diverted to bicycling or walking. The methods are not useful for estimating changes in demand in response to an improvement, and they shed little light on factors affecting the decision to walk or bicycle.

Market for Bicycle Commuting in the San Francisco Bay Area
Deakin (1985) defined a demographic target group for Bay Area commuter bicycling, based on data from the Bay Area Travel Survey, a review of the literature, and interviews with local and state officials. Her market was defined as employed full-time, under 40 years old, travels less than 11.3 km one-way to work, drives alone during the peak period, and owns a bike suitable for commuting. She used these criteria to estimate a reasonable upper bound on the size of the potential bicycle commuter market.
Demand Estimation: Aggregate Behaviour Studies

Overview Measures of potential demand have been developed for both bicycle and pedestrian facilities for the purpose of prioritizing facility improvements according to areas of highest potential demand. Demand potential is measured based on characteristics and levels of the surrounding population, trip generators, as well as other environmental factors such as topography and the quality of connecting facilities.

Typical Applications Measures for both bicycle and pedestrian facility demand potential have been developed and applied to prioritize improvements (see sidebar).

Advantages Measures of potential demand can be a useful aid to prioritizing locations for improvements, particularly when applied in conjunction with measures of supply or facility quality to identify areas of both high potential demand and significant deficiencies. In addition, these measures can frequently be constructed from readily available data sources such as the census and local land use data bases.

Disadvantages Measures of potential demand only indicate relative levels of demand between areas, rather than predict the actual number of users of a facility. They do not indicate the extent to which usage is likely to increase as the result of a particular improvement, and they do not indicate which improvements to a specific facility or area should be given the highest priority. Also, the factors used in constructing the index may or may not be good indicators of the true potential demand for the facility.

Latent Demand Score
A Latent Demand Score (LDS) technique has been developed to estimate the latent or potential demand for bicycle travel assuming the existence of a bicycle facility. Trips are estimated based on the size and proximity of population and activity centers to the proposed facility, using Geographic Information System (GIS) analysis tools. The LDS has been applied in a number of cities with the purpose of prioritizing existing bicycle facility improvements or new bicycle facility improvements or new bicycle facilities. (Landis, 1996). The LDS may be combined with bicycle level of service measures.
  Pedestrian Potential Index
A Pedestrian Potential Index has been developed and applied in Oregon to prioritize locations for pedestrian improvements. The index uses three main factors: (1) proximity factors that refer to pedestrian generators such as schools, transit or neighborhood shopping; (2) environmental factors such as mixed use and street connectivity; and (3) policy factors that identify certain areas as critical for pedestrians. The index has been applied in conjunction with a Deficiency Index to identify areas with both high potential demand and significant deficiencies. (City of Portland, 1997).
Demand Estimation: Aggregate Behaviour Studies

Overview A variety of compatibility measures have been developed to indicate the suitability of a particular facility for bicycle or pedestrian travel. These measures have been given names such as "Level of Service," "Stress Level," "Compatibility Index," and "Interaction Hazard Score." The measures combine factors such as motor vehicle traffic volume and speeds, lane or sidewalk width, pavement quality, and pedestrian amenities into an index of overall suitability for travel. The measures can be used alone or in conjunction with measures of potential demand to prioritize facilities for improvements.

Typical Applications Compatibility measures have been used in a number of cities to rank facilities for purposes of prioritizing projects. For example, Orange County, NC, has applied the Bicycle Stress Level index to determine the suitability of their planned bicycle routes. Level-of- service measures have also been applied in conjunction with the Latent Demand Score to prioritize projects in various urban areas in Florida. Oregon has developed a Deficiency Index which it uses in conjunction with potential demand indicators to rank and prioritize pedestrian facilities.

Advantages Compatibility measures can serve as a useful means of prioritizing facilities for improvement as well as determining which improvements will be most beneficial. Compatibility measures may also become a key component of non-motorized travel demand forecasting, if relationships can be developed between the indices and individuals' likelihood of making a bicycling or walking trip.

Disadvantages Existing indices primarily rate individual segments rather than describing the overall compatibility of a route. They cannot account for the effects of intersections and other discontinuities, and they do not sufficiently describe the overall compatibility of a route made up of different segments with different ratings. Also, the indices may not include all relevant factors (or may require significant data collection to do so), and they may not properly reflect perceptions if not validated through surveys. In addition, they do not predict the actual number of trips on the segment.

Bicycle Compatibility Index
The Federal Highway Administration has recently developed a bicycle compatibility index (BCI) to describe the compatibility of a facility for cycling (FHWA, 1998). The BCI uses a formula based on traffic volume, speed, lane width, and other indicators of bicyclist stress to rank a road segment for compatibility on a scale of 1 to 6, which is then equated to a level-of-service (LOS) rating. Qualitative adjustment factors were developed to consider instances of high volumes of trucks or buses, right-turning vehicles, and vehicles turning into and out of driveways. The index was developed using a video survey methodology which asked participants to rate their comfort level on various videotaped facilities.
Demand Estimation: Aggregate Behaviour Studies

Overview Pedestrian and bicycle environment factors describe the friendliness of an area (such as a city block, census tract, or traffic analysis zone) for walking and/or bicycling. The factors are quantitative and may be a composite of a number of quantitative descriptors and subjective factors. Examples of factors considered include lane or sidewalk width, street continuity, topography, and the aesthetic quality of the environment.

Typical Applications Pedestrian and bicycle environment factors have been developed primarily for use in regional travel models. A pedestrian environment factor has been developed and applied to the regional travel model in Portland, OR and modified versions have been applied in Sacramento, CA and Washington, DC. Montgomery County, MD, has developed a different pedestrian/bicycle environment factor for use in its travel model. A transit friendliness factor describing the quality of pedestrian access to transit has been developed in Washington State.

Advantages Considerable research has been performed recently on factors that make areas inviting to pedestrians, and much of this knowledge has been incorporated in the current generation of environment factors. The factors have been found to enhance the performance of travel models in Portland, OR and Montgomery County, MD particularly for predicting vehicle trips from an area. These factors may also be useful in prioritizing areas for improvements, based on the relative ratings of individual areas.

Disadvantages Environment factors are frequently based on subjective ratings and their performance at predicting actual variations in travel behavior has not yet been widely validated. Also, separate bicycle environment factors have not been developed; the ability of these or of combined pedestrian/bicycle factors to predict bicycle trip activity has not yet been tested. In addition, environment factors require considerable field data collection to develop for a specific area.

Portland, OR, Pedestrian Environment Factor
Portland's Pedestrian Environment Factor (PEF), developed for use in its regional travel model, includes four elements: sidewalk availability, ease of street crossing, connectivity of street/sidewalk system, and terrain. Each traffic analysis zone is ranked for each element on a scale of zero to three, with higher numbers representing higher quality pedestrian environments, so the overall PEF can range from 0 to 12 (1,000 Friends of Oregon, 1992 - 1997).
Demand Estimation: Aggregate Behaviour Studies

Overview Geographic Information Systems (GIS) relate environmental and population data in a spatial framework, using location points, lines (commonly roadway links and corridors), corridors), and polygons (surface areas and analysis zones). GIS are employed as a mechanism for the physical inventory of transportation facilities; as a planning tool to relate available environmental, personal transportation and household characteristics data; as a spatial analysis tool for calculating distances and areas; as a network performance monitor; and as a vehicle for the graphic display of data and analysis in a geographic context.

Typical Applications GIS have been used in non-motorized planning to inventory and evaluate facilities such as roads and sidewalks; establish spatial relationships between roadway network links, features such as activity centers, and area population characteristics; compare and display current conditions with projected travel and conditions; assess total network performance and identify optimal routes; produce printed maps; and develop network measures (e.g. street density and connectivity) and land use measures (e.g., mix of residential, office, and retail) which can be related to the likelihood of walking or bicycling.

Advantages GIS can greatly increase the ease of analyzing data relevant to non-motorized travel forecasting. For example, a corridor surrounding a facility can be defined and the characteristics of the population within the corridor easily identified. GIS allows development of spatial measures and analysis of data relationships which might otherwise be prohibitively time-consuming or impossible. The display capabilities of GIS are also valuable for conveying information to policymakers and the public.

Disadvantages GIS require considerable user skill as well as specialized software to develop, although future developments will make them more accessible to laypersons. Also, since GIS can only manage and analyze data, the data must still be collected through other means.

Warwick, RI, Bicycle Network Study
A Bicycle Network Study in Warwick, RI, was assisted by GIS methods. Trip generation estimates were calculated as a function of employment, school enrollment, and total population for traffic analysis zones adjacent to the bicycle network. Composite trip generation scores were then attributed to network segments within the areas of influence of trip generators. The results of this analysis were compared to the existing designated bicycle route network. Alternative route designations were suggested where an undesignated roadway link's potential scored higher than a parallel or adjacent designated route (Beltz and Burgess, 1997).
Demand Estimation: Aggregate Behaviour Studies

Overview Using survey research techniques, preference surveys (also known as stated preference surveys) focus on the choices that people would make given discrete alternatives. Respondents are asked to express an attitude or make a choice as to how they would act under certain conditions. Two basic types of preference surveys exist. Attitudinal surveys ask respondents directly how they would respond to various actions (e.g., would they bicycle if bike lanes were available), or ask them to rate their preferences for various improvements. Hypothetical choice surveys require respondents to make choices between hypothetical alternatives with varying attributes, and survey results are then used to develop models of behavior.

Typical Applications Attitudinal surveys have been widely used to estimate the potential impacts of bicycle and pedestrian improvements and to determine relative preferences for such improvements. Hypothetical choice surveys are generally used to develop discrete choice models and to estimate the relative importance of each attribute (time, cost, presence of bike lanes, etc.) in common terms.

Advantages Attitudinal surveys are relatively easy to design and implement. They can also be good tools for evaluating relative preferences and for estimating the maximum possible response to an action. Hypothetical choice surveys, if carefully designed, can be used to develop relatively accurate models of behavior and to give quantitative information on the relative importance which people place on various factors.

Disadvantages Attitudinal surveys often significantly overestimate the response to a bicycle or pedestrian improvement, since people tend to be more likely to state that they will change their behavior than to actually do so (Goldsmith, 1992). Therefore, they are not well-suited for predicting actual shifts in travel demand. While hypothetical choice surveys overcome many of the limitations of attitudinal surveys, they must be designed carefully and require considerable time and expertise to implement. Both types of preference surveys suffer from the further drawback that people may not have any real-world experience with the choices they are asked to make, and may therefore be unable to indicate their preferences or actions with accuracy.

Transit Access Mode Choice in Chicago
The Chicago Regional Transit Authority (RTA) surveyed transit and auto users to determine reasons why they did not currently walk or bicycle to a transit station. (These surveys were also used to develop models of individual behavior, as described under Discrete Choice Models.) Respondents were asked to identify specific reasons for not bicycling or walking, such as lack of secure parking, dangerous traffic conditions, or inadequate sidewalks or path- ways. Two different survey methods were employed: an intercept survey in which respondents were asked directly to rate factors, and an interactive video survey in which respondents were asked to make tradeoffs between vari-

3.2 Key Characteristics and Uses of Each Method

This section summarizes key characteristics of the methods and suggests appropriate methods according to specific purpose such as forecasting the number of new users of a bicycle/pedestrian trail. More specifically, table 3.2 summarizes key characteristics of each of the 11 methods, providing more detail on the factors (e.g., ease of use and data requirements) rated in the quick reference guide for each method.

Tables 3.3 through 3.6 are intended as a guide for practitioners who need to choose the most appropriate method for a specific situation. Each table lists a specific purpose for which non-motorized demand forecasting methods may be applied and suggests which methods are most appropriate for that purpose. Generally the methods are ordered from simpler to more complex in Tables 3.3 - 3.5. For each of these methods, the table describes the specific way in which the method would be applied and identifies major advantages and disadvantages of using the method for the given purpose. These purposes include:

Table 3.3 - estimating the number of users of a new facility;

Table 3.4 - estimating the number of new bicycle or pedestrian trips area-wide, as a result of facility or network improvements;

Table 3.5 - prioritizing design features for a specific facility; and

Table 3.6 - prioritizing facilities for improvement.

roadway without a sidewalk with pedestrians walking along it
Figure 3.2: If Sidewalks Were Built Here,
How Many People Would Use Them?
How Far Up on the Priority List is This Project?

 

Table 3.2 - Key Characteristics of Available Methods.

Method Ease of Use Data Requirements Accuracy Sensitivity to Design Factors Where Used
Demand Estimation
Comparison Studies
Simple to understand and relatively easy to apply Requires facility user counts; data on surrounding population and land uses are optional May provide rough estimates of demand if truly comparable case studies can be found. Accuracy has not been formally tested. Relatively low; requires identification of comparable facilities within a comparable environment Massachusetts; Netherlands; Germany; Australia
Aggregate Behavior
Studies
Requires simple statistical analysis skills Varies; can use existing sources such as census and local land use data bases Models have generally had low explanatory power and have not been transferable Low, since detailed information on facilities has generally not been collected UK; Berkeley, CA
Sketch Plan Methods
Methods are relatively simple to apply Varies; can use existing sources such as census and local land use data bases Varies by method; some methods may give reasonable estimates others have not been formally tested Low; rely on general assumptions Seattle, WA (bicycle); New York City, NY; Plattsburgh, NY; Milwaukee, WI; Toronto and Montreal, Canada (pedestrian)
Discrete Choice Models
Knowledge of statistical analysis and specialized survey and modeling techniques is required Usually requires survey data collection specific to situation being analyzed Can be relatively accurate in predicting impacts of specific actions High, although only limited number of factors can be considered at once Wisconsin; California, Chicago, IL; Raleigh, NC
Regional Travel Models
Requires established capabilities for travel demand modeling May require additional data collection on bicycle and pedestrian travel patterns and/or facility characteristics Including bikes/peds has improved performance of some models at predicting auto and transit trips Potentially high; limited by data availability and tradeoff information Portland, OR; Montgomery County, MD; Sacramento and San Francisco, CA; Edmonton, Canada; Leicester, UK; Netherlands
Relative Demand Potential
Method Ease of Use Data Requirements Accuracy Sensitivity to Design Factors Where Used
Market Analysis Methods are relatively simple to apply Data required on trip length distributions (from travel survey or regional travel model); other population data may be needed Untested. Methods are designed to predict an upper bound under ideal conditions Low; assumes ideal network of facilities San Francisco, CA; Chicago, IL; Bend, OR; Minneapolis, MN; Europe
Facility Demand Potential Methods are relatively simple to apply Data required on local population and land use, some methods require trip distributions by length and purpose Attempts to apply Latent Demand Score in practice have had mixed results Low; assumes ideal network of facilities (Pedestrian Potential Index); Florida; Birmingham, AL; Philadelphia, PA; Portland, OR (pedestrian)
Supply Quality Analysis          
Bicycle and Pedestrian
Compatibility Measures
Methods vary but are generally relatively simple to apply Requires data on facility characteristics; some may exist, others may need to be collected, depending on method Has not been tested with respect to forecasting demand High; factors included depend on specific index Orange County, NC; Gainesville, FL; Buffalo, NY; Ames, IA
Environment Factors Relatively simple to apply; may require judgment in developing ratings Generally requires field data collection on facility/ environmental characteristics Have improved performance of some regional travel models at predicting auto, transit trips High; factors included depend on specific index Portland, OR; Montgomery County, MD; Sacramento, CA
Supporting Tools and Techniques
Geographic Information
Systems
Generally requires specialized knowledge of GIS analysis techniques GIS can manage and analyze a wide variety of data based on availability and needs Has potential to improve accuracy of forecasting methods Potential to store information on a variety of facility design factors Portland, OR; Seattle, WA; Buffalo, NY; Warwick, RI; Orange County, CA; Fort Collins, CO; Buffalo, NY; Ames, IA
Preference Surveys Requires knowledge of survey research techniques; may require specialized survey design and analysis skills Requires survey data collection Performance depends on quality of survey design and implementation Variety of design factors can be considered in survey Widespread

 

Table 3.3 - Methods for Estimating the Number of Users of a New Facility.

Method Specific Application Major Advantages or Drawbacks
Comparison Study
Look at usage on comparable facility May be difficult to find truly comparable situation
Sketch-Plan Method
Look at local population, trip generators, non-motorized work trip percentages for area around facility to estimate potential trips Easy way to get a rough estimate of potential usage; however, difficult to consider factors such as non-work trips, whether facility serves local travel patterns, existence of supporting facilities/network, etc.
Preference Survey
(Attitudinal)
Survey local residents and commuters as to whether they would use the facility Will give relative indication of interest, but will generally overstate actual likelihood of using facility
Preference Survey
(Hypothetical Choice)
and Discrete Choice
Model
Conduct survey of whether people would use facility under various scenarios; develop behavior model to predict usage A carefully-designed hypothetical choice survey may be the most accurate method but is also resource-intensive
Regional Travel Model
Modify existing regional travel model to include new facility Requires travel model which already includes bicycling/walking networks; will not capture recreational travel

 

Table 3.4 - Methods for Estimating the Number of New Bicycle or Pedestrian Trips Area-wide as a Result of Facility or Network Improvements.

Method Specific Application Major Advantages or Drawbacks
Preference Survey
(Attitudinal)
Survey residents to ask if they would choose to walk or bicycle given improvements Survey results tend to overstate willingness to change mode of travel
Aggregate Behavior
Study
Develop relationship between levels of non-motorized trip-making and overall facility/network characteristics, based on data from other cities/areas Requires data on many cities or areas which includes indicators of non-motorized trip making as well as information on existing facilities/
networks comparable to the improvements being considered locally
Preference Survey
(Hypothetical Choice)
and Discrete Choice
Model
Conduct survey of whether people would bicycle or walk under various city-wide improvement scenarios; develop behavior model to predict usage A carefully-designed hypothetical choice survey may be relatively accurate but is also resource-intensive
Regional Travel Model
Modify pedestrian/bicycle environment factors or network links in regional travel model Requires travel model which already includes bicycling/walking environment factors and/or networks, and that these networks include facility characteristics that are desired to be improved; models must also be based on data relating behavior responses to design improvements

 

Table 3.5 - Methods for Prioritizing Design Features for a Specific Facility.

Method Specific Application Major Advantages or Drawbacks
Supply Quality Analysis
Compare improvements in quality rating as a result of various design improvements Good for identifying facility deficiencies and most effective improvements, but using this technique alone does not predict benefits in terms of new users
Preference Survey
(Attitudinal)
Ask local residents, employees, bicyclists, pedestrians, etc., which design improvements are highest priority Responses may vary depending on population surveyed; for example, just surveying existing users will not indicate number of new users attracted to facility as a result of improvements
Preference Survey
(Hypothetical Choice)
and Discrete Choice
Model
Conduct survey to determine relative-preference for facility improvements, and build model to determine likely number of new users Determining who to survey can be a problem; however, can actually predict benefits of each improvement based on change in usage as well as benefits to existing users
Regional Travel Model
Modify facility travel times to reflect proposed new facilities or design improvements, to determine travel-time equivalent benefits to existing users and number of new users Considers most types and origins/destinations of trips. However, requires that the travel network is coded with the bicycle or pedestrian facility design features to be analyzed, and that the valuation of travel time by bicycle or foot has been related to these design features.

 

Table 3.6 - Methods for Prioritizing Facilities for Improvement.

Method Specific Application Major Advantages or Drawbacks
Supply Quality Analysis
Rate facilities based on existing bicycle or pedestrian compatibility, environment factors, or deficiency indicators; prioritize according to ratings Does not look at existing or potential demand/usage on facilities
Preference Survey
(Attitudinal)
Ask local residents, employees, bicyclists/
pedestrians, etc., which are highest priority facilities to improve
Responses may vary depending on population surveyed; for example, just surveying existing users will not indicate number of new users attracted to facility; need to survey population of potential users as well
Facility Demand
Potential
Look at potential demand for facility based on surrounding population, land uses, etc., and prioritize according to highest potential Serves as a good basis for prioritization assuming that measures of potential demand are proportional to actual future demand across projects.
Combination of Facility
Demand Potential and
Supply Quality
Analysis
Rate facilities both on potential demand and existing quality; prioritize facilities with highest potential and lowest quality Combines strengths of both methods; however, still does not indicate actual number of new users

 

FHWA-RD-98-165

Previous | Table of Contents | Next

ResearchFHWA
FHWA
United States Department of Transportation - Federal Highway Administration