Synthesis of Traveler Choice Research: Improving Modeling Accuracy for Better Transportation Decisionmaking

Travel Behavior Studies Review

This section reviews studies of different levels of traveler decisionmaking as outlined in the conceptual framework, organizing travel behavior knowledge by decision horizon. At the within-day and day-to-day levels, route and departure time choices have been the primary focus. Travelers’ experiences from day to day influence their future decisions, and the line between these daily choices and a traveler’s behavioral pattern quickly blurs. For example, a traveler may eliminate public transportation from his or her choice set after a bad experience, even if the utility is otherwise perceived as quite high. Since mode choice is subject to available modes, it tends to be modeled and studied as a behavioral pattern in the time horizon of weeks or months. Finally, lifestyle and mobility choices reflect the self- or otherwise-imposed constraints to which travelers are subjected (and choose) over longer time frames. Much work has been done within each area to understand how various factors influence these choices, but there is less understanding of the mechanisms that work to define these travel habits, patterns, and long-term constraints. These mechanisms and how they relate to different levels of traveler decisionmaking are discussed last.

Day-to-Day and Within-Day Behavior Changes

Jan et al. found that travelers habitually follow the same route for the same trip, but route variations increase with longer travel distances.⁽⁵⁾ The dominant factors for route choice are travel time and distance.^(5–7) Significant research effort has been focused on the effects of route choice behavior under traffic information systems, the dynamic aspect of route choice behavior, and the relationships among route choice, departure time, and trip-chaining decisions.^(8–10)

Traveler information influences route choice substantially. Abdel-Aty et al. studied route changes in Los Angeles, CA.⁽⁷⁾ Only a small share of the respondents (15 percent) reported using more than one route on their commute. Of that 15 percent, 34 percent said they changed routes after actually seeing traffic conditions. Higher incomes and education levels predicted more route changes, perhaps reflecting schedule flexibility and arrival-time expectations for such workers.

Mahmassani and Herman performed a survey of commuters in Austin, TX, and yielded a binary logit model that relates route switching propensity to four types of factors: geographic and network condition variables, workplace characteristics, individual attributes, and use of information (radio traffic reports).⁽¹¹⁾ They found that variables describing the characteristics of the commute itself had a dominant effect relative to workplace rules or individual characteristics. Information in the form of radio traffic reports also appeared to have a strong impact. Regular listeners to traffic information were more likely to switch routes. The only sociodemographic attribute significant in the model was age.

In a similar experiment, Avineri and Prashker examined the impact of information on traveler learning, differentiated by travelers’ risk aversion.^(12,13) The results suggest that when information about travel times is provided, travelers do not always choose the route with the least expected time. Giving static information to users increases traveler heterogeneity; in this case, individuals learned more quickly to prefer either routes with less travel time or routes with less variability in travel time. When examined at an aggregate level, this combination could be seen as a “non-learning effect” or no change. Furthermore, higher variation in travel times is associated with lower sensitivity to travel time differences. Avineri and Prashker found in some cases that “increasing travel time variability of a less attractive route could increase its perceived attractiveness.”⁽¹²⁾ This underscores the need for better models of learning and reinforced habits as an alternative to utility maximization.

Beyond these dimensions, only a couple of studies have addressed destination adjustment in response to real-time information for discretionary travel (shopping).⁽¹⁴⁾ The remainder of this section discusses the effects of other network and non-network factors that have been explored
in more depth.

Effect of Tolling and Other Costs on Mobility Decisions

Travel cost as part of demand management is a powerful tool to influence travel behavior. Hensher and King examined the influence of parking costs in the central business district, a park-and-ride facility with public transit access, and the related mode choice as well as destination choice (including the alternative to forgo the trip) in Sydney, Australia.⁽¹⁵⁾ Each of the participants was required to consider six alternatives in a stated preference questionnaire. In 97 percent of the responses, cost was the most significant factor in location and mode choices. Similar results were found by Handy et al., who studied whether Americans drive by choice or necessity.⁽¹⁶⁾ The study found that most drivers chose the car because of the costs and a lack of alternatives. However, studies of this nature in Europe may reflect stricter land use norms that have led to denser, more compact urban form and increased use of public transit but that also decrease costs for public transit operators and increase the cost of parking.

Congestion pricing of roadways presents a valuable opportunity to rationalize road networks by helping ensure that travelers pay for the delay costs they impose on others. A study of Seattle, WA, travelers with Global Positioning System (GPS) vehicle units estimated that variable network pricing (to reflect the congestion impacts of different demand levels over space and time) would reduce regional vehicle miles traveled (VMT) by 12 percent and total travel time by 7 percent with a 6:1 benefit-cost ratio.⁽¹⁷⁾ Using GPS tolling meters, the study followed participants to establish a baseline tolling routine. Participants were then given a monetary travel budget sufficient to cover the cost of their routine for the duration of the study period, creating an incentive to reduce certain forms of travel to save or make money. This policy approach is very similar to Kockelman and students’ credit-based congestion pricing policy proposal. However, VMT results differed in their network simulations of the Austin and Dallas-Ft. Worth regions of Texas, where marginal social-cost pricing of freeways for all links by time of day was consistently estimated to result in VMT savings of less than 10 percent.^(18,19)

Saleh and Farrell investigated the influence of congestion pricing on the “peak-spreading” of departure time choice.⁽²⁰⁾ Results suggest that non-work activities and work schedule flexibility impact departure time choice for the trip to work. Furthermore, respondents were less willing to pay a toll to depart earlier than usual.

In a similar vein, Transit Cooperative Research Program (TCRP) Report 95 discusses a number of elements that influence a traveler’s decision to use a high-occupancy vehicle (HOV) lane.⁽²¹⁾ The report concludes that because so many urban, facility, and vehicle characteristics interact with one another it is difficult to delineate the effect of HOV lanes on travelers. However, the success of HOV lanes—both in terms of drivers served and benefits to the road network—is attributed to combinations of the following attributes:

·        Urbanized population of 1.5 million or more.

·        Orientation, preferably radial, to a city center, “focusing on major employment centers with preferably more than 100,000 jobs.”

·        Geographic barriers.

·        Congestion in general purpose lanes.

·        Realistic potential for 25–30 buses per hour.

·        Peak-hour travel time savings of preferably 1 min/mi or more or at least 5 min of total travel time.⁽²¹⁾

Walk Quality on Day-to-Day Travel Behavior and Patterns

Beyond information and pricing, the quality of the urban environment can influence route and activity timing decisions. Cervero and Kockelman examined many features of urban form that may reduce auto dependence.⁽²²⁾ Their gravity-based accessibility measure for access to commercial jobs was found to have an elasticity of -0.27, suggesting neighborhood retail shops and pedestrian-oriented design are more significant than residential densities in mode selection. Integrating aspects of pedestrian-oriented design such as four-way intersections and vertical mixing of land uses may result in significant VMT reductions. For example, a 10 percent increase in the number of four-way intersections in a neighborhood was associated with an average reduction of 5.19 percent of person miles traveled for non-work trips. A doubling of land use mix or variety is associated with a roughly 11 percent increase in modes other than single-occupancy vehicle (SOV) for non-work travel.

In addition to urban density, mixed land use and high-quality pedestrian-oriented urban design increase the use of public transit and non-motorized transport modes.⁽²³⁾ Naess and Jensen found that, in general, car use increases with increasing distance from the city center.^(24,25) This could also be an indicator of self-selection or endogeneity (a topic discussed in more depth in later sections). Similarly, Cervero studied the impact of compact, mixed-use, pedestrian-friendly design on mode choice.⁽²⁶⁾ The study quantified density and diversity and estimated the influence of each on mode choice. The influences were significant but modest. Surprisingly, the most important influence factor for mode choice was the sidewalk ratio. In well-developed pedestrian areas, commuters were more likely to use public transit or join carpooling initiatives.

Information, pricing, and urban form influence day-to-day and within-day behaviors, but they are understood and applied over time such that they also influence travel patterns. These and other influences are discussed in the context of habits in the following sections.