U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
202-366-4000
Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations
REPORT |
This report is an archived publication and may contain dated technical, contact, and link information |
|
Publication Number: FHWA-HRT-13-022 Date: August 2013 |
Publication Number: FHWA-HRT-13-022 Date: August 2013 |
Most travel behavior studies use surveys and travel diaries as data sources. There are only a few laboratory and field experiments that have been used for collecting data to capture travel behavior. The most widely used resources are national travel surveys that many countries collect. These surveys typically collect individual information (socioeconomic, demographic), household information (size, structure, relationships), vehicle information (age, make, model), and a diary of journeys on a given day (start and end locations, start and end times, mode of travel, accompaniment, purpose of travel). Recent and continuously collected national travel surveys are summarized in table 1. Armoguum et al. gave an overview of the different national travel surveys.(119)
Table 1. Recent and Continuous National Travel Surveys.
Country |
Source |
Years |
United States |
Federal Highway Administration |
2009, 2001, 1995 |
Germany |
German Mobility Panel |
1994–2011 |
Denmark |
Danish Transport Research Institute |
2006 |
Netherlands |
Institute for Road Safety Research |
1978–2011 |
United Kingdom |
United Kingdom Department for Transport |
2009 |
Switzerland |
Federal Department of Environment, Transport, Energy and Communications |
2010, 2005, 2000 |
Sweden |
Sika Institute |
2006 |
A number of papers accessed data from secondary sources such as the American Housing Survey 2001 or relied on specific surveys designed for the research. For example, Srinivasan and Rogers used a random stratified sample of 500 households in two suburbs of Chennai, India; Ampt et al. interviewed 102 participants in their survey of household travel behavior; and Rose and Ampt included only 46 participants in their study of car use reduction strategies in Sydney and Adelaide, Australia.(120–122) Thus, the number of participants in research projects varies considerably from less than 100 to many thousands in the case of national travel surveys. Methods of data collection also vary considerably, including face-to-face structured or semi-structured interviews, postal questionnaires, telephone surveys, and, more recently, online questionnaires.
Most studies of travel behavior use travel diaries, which include data from trips and the activity behavior that is collected with the trip. The most common practice with travel diaries is to collect data from a 1- or 2-day time span, which has a relatively low response burden and gives full information about trip frequencies, mode choice, and other decisions for aggregated models. Boarnet and Crane, Bowman and Ben-Akiva, Giuliano, Kunert and Follmer, and Newbold et al. all used 1-day travel diaries. (See references 38, 96, 34, 123, and 27.) A number of other projects used 2-day travel diaries.(23,124,125) However, to observe behavior changes and habitual travel and record data about how to influence and change travel behavior, a longer reporting period would be useful. Only a small number of research projects used 7-day or longer diaries, including Garvill et al., Kenyon, and Schlich and Axhausen.(126,127,41) In Schlich and Axhausen’s study of travel behavior in Karlsruhe and Halle (Saale), Germany, participants were required to keep a travel diary for a period of 6 weeks, recording all travel movements during that time. A total of 52,273 trips were recorded by the 361 participants.(41) Garvill et al. used the German National Travel Survey, a panel survey where participants are required to complete a travel diary several times during different periods.(126) Kenyon required 100 participants to complete a 7-day travel diary three times at 6‑month intervals in March 2004, October 2004, and March 2005.(127)
Not all surveys state the day of the week and season for the trips reported. Often, the day is only differentiated between weekdays and weekends/holidays. It is surprising that the season and day of the week are not discussed, as seasonality and associated weather has an influence on travel patterns, and travel patterns vary from day to day and week to week.
Larson and Poist provide an interesting overview of all papers using a postal questionnaire included in Transportation Journal between 1992 and 2003.(128) The authors report a total of 106,300 mailed questionnaires and that response rates have declined significantly since 1992.
To analyze likely behavior changes in response to changes in the transport system and other influence factors, RP data are often unavailable and SP data have to be used. As a result, significant literature has been developed around survey methods for estimating individuals’ behavior adaption in the absence of revealed system variation. These methods are widely used for developing optimal pricing strategies, forecasting responses to price changes, and modeling demand functions.
Although longer time studies might have information about system changes, there are often limitations to RP data. For example, observation of choices might not occur and changes take time for adoption. These limitations could be overcome with real-life controlled experiments; however, opportunities for such experiments have been limited. The Federal Highway Administration provides some opportunities to design and collect data from such field experiments through the integrated environment at the Saxton Transportation Operations Laboratory, which has a data resources test bed, a concept and analysis test bed, and a cooperative vehicle-highway test bed. If such experimental environments are not available, SP surveys provide an approximation by asking questions about hypothetical situations. The design and configuration of such questions is not trivial and has been the subject of research in recent years. Bliemer and Rose proposed an approach to generate an efficient experimental design that minimizes standard errors in estimating the parameters from the utility function that underlies travel behavior decisions.(129)
Because SP data may not include the history of individuals, which is needed to assess the effect of habits, models estimated with SP data often overestimate the impact of changes in the system on behavior adaption. Questions included in SP surveys such as car use and trip frequency can give hints about the history so that models can capture inertia. Mixed RP/SP data include more likely inertia indicators, as they are constructed based on previously chosen alternatives and real situations are used to construct the hypothetical SP survey response.(130) Since such data is vulnerable to serial correlation, Cantillo et al. proposed discrete choice models with both inertia and serial correlation with mixed RP/SP data. They found that inertia and serial correlation in mixed RP/SP data are significant.(131)
In 1997, Kitamura and others pointed to the need for more extensive data and improved methodologies for understanding travel behavior.(132) The case for collection of such data has strengthened in the last 15 years as others have called for more comprehensive transportation modeling and planning contexts.(58) Beyond trips and travel networks, analysts need data to understand “why, with whom, where, and when activities are engaged in and how activity engagement is related to the spatial and institutional organization of an urban area.”(132) If travel is assumed to be a derived demand, then sociodemographics and attitudes are thought to be the primary inducements for observed travel. When travel for its own sake is evaluated, attitudes and perceptions become even more important.(133) From this perspective, travel control measures and travel demand measures affect urban quality of life more than facility expansion.(133) Thus, any effort to understand the impact of demand-side interactions must examine human time use. This requires improved data collection and methodologies that are able to evaluate both induced and suppressed travel as well as the implications of this travel on perceived quality of life.(132) Time-use data are particularly useful for understanding location substitution of activities (e.g., telemobility alternatives) and activity and departure timing. Jara-Diaz noted that as the utility of travel depends on activities as a result of time and monetary budget assignments, then travel must be understood in the context of human activities and the nature and perception of time use.(134) Thus, meaningful models require linking data, methods, and knowledge from sociology, psychology, and economics.(134)
Kwan’s review of time-use research, time-geographic research, and studies on human activity-travel patterns in space-time discusses the integrated nature of time, space, and information technology (IT).(135) Transportation research has explored the implications of IT on time use and travel behavior, noting that telemobility alternatives may be complementary, substitutive, or synergistic to traditional travel behaviors.(136) Furthermore, mobile devices allow for en-route or continuous adjustment of travel plans. A significant amount research on the dynamics of route choice behavior is based on laboratory-like experiments that repeatedly ask the participants to respond to hypothetical route choices. Because of the lack of detailed, disaggregate spatial and temporal data, much research has focused on two dimensions at a time (e.g., location and time use, time use and IT, or location and IT).(137,138) Though the availability of such data represented a significant challenge in the past, the current challenge lies in the design of postprocessing algorithms to enable analysis of the massive quantities of data available from smartphones and increased Internet use. Methods for analysis of complex space-time data are also needed.(135)
IT presents an opportunity to relax the time-space constraints often imposed in travel behavior studies. The ability to mingle work and non-work activities and locations using mobile devices blurs the distinctions between home and work and between public and private.(139) Kwan suggested that these time-space constraints will not disappear altogether, since IT accessibility and quality of service is often constrained by location.(140) Future research should examine how social and geographical contexts shape the impact of IT on specific social groups and, in particular, urban areas.(135) Furthermore, analysts should examine interactions among household members within social groups and evaluate within-household variations of IT use. IT provides additional data sources such as social networking sites that can be used for a variety of detailed destination attributes and trip purposes that were previously unavailable.
Experimental methods have an increasing role to play in the study of complex activity and travel behavior dynamics, especially as information and communication technologies increase the realm of spatiotemporal opportunities available for individual and household activity engagement. In a synthesis of experimental economics approaches to travel behavior, Mahmassani identified the following situations for which laboratory experiments may be needed in the study of the relevant behavior and system properties:(141)
·
Complex dynamics and collective
effects are essential aspects of the system under consideration, making
joint measurement in the real world considerably complicated
or costly.
· Situations or policies of interest are not available in the real world (e.g., new technologies) or are mutually inconsistent in the same system.
· Control for extraneous factors is desired.
· Understanding of dynamics and learning processes is of concern.
Experiments that entail varying degrees of sophistication in context design, task design, and delivery environment and that contain different scales of experimentation in terms of number of participants and environmental perturbation are common in many disciplines concerned with the study of human systems. Transportation planning professionals and travel behavior-activity researchers have been slow to adopt experimental methods in research or practice (with the exception of stated response methods and full-scale operational tests). However, from modest beginnings in the early 1980s, there appears to be growing interest in experimental methods for the study of human behavior in transportation decision situations. The following reasons can be surmised for this phenomenon:(141)
· Growing interest in experimental economics as an approach for the study of economic systems.
· Related development in complexity science and its application to human, economic, and sociotechnical systems.
· Advancement in computing capabilities and networked environments, especially the Web, and interest in large-scale collective phenomena in networks.
· Continued development of travel behavior as a focus of interdisciplinary research, with entry of professionals from varying disciplinary backgrounds.
· Increased sophistication in methods, theories, and intellectual constructs in travel and activity behavior research.
· Significance of policy questions and concerns that require better understanding of behavioral dynamics and multiagent interactions (e.g., environmental sustainability, vehicle use, and congestion mitigation).
· Technological advances in information and communication technologies that enable improved simulation/gaming environments, delivery platforms, and multiplayer interactions.
In the mid to late 1980s, Mahmassani and Herman conducted a series of three experiments involving actual commuters in a simulated congested traffic corridor.(11) Those experiments were conducted before the widespread availability of personal computers and the Internet and entailed overcoming significant logistical challenges. The participants were all actual daily commuters who responded to traffic conditions with their selection of a particular time to depart or route to use in a commuting corridor. The experiments provided a basis for articulating a theory of departure time and route switching decision mechanisms in repeated decision situations such as work commuting. The experiment results were indirectly validated with 2-week diary surveys of commuters in Austin and Dallas, TX.(142,143) An important methodological question is the extent to which behavioral findings from laboratory experiments are indeed representative of actual behavior in real traffic systems. The main conclusion from the comparative analyses was that behavioral mechanisms developed on the basis of laboratory experiments provided a good explanation of observed behavior, with essentially similar model specifications and correct signs but different coefficient magnitudes.(143)
An extensive set of experiments was conducted by Mahmassani and colleagues to investigate user dynamics under real-time information of varying types. In contrast to earlier experiments, which addressed only the day-to-day dynamics of user decisions, the ATIS investigation addressed both real-time and day-to-day dynamics. As such, these experiments required a special purpose simulator that allowed real-time interaction between respondents and the traffic system. The interactive simulator provided ATIS information that was consistent with the traffic conditions on the network. The prevailing traffic conditions, in turn, were the result of collective decisions of individuals on the network, whose interactions in traffic were modeled using a dynamic traffic simulation model. Thus, the simulator ensured mutual consistency between user behavior, experienced traffic network conditions, and real-time information.(144)
Three sets of experiments were performed over a 3-year period: (1) en-route path choice and day-to-day departure decisions under a given overall congestion level, (2) effect of congestion and experimental exposure sequence, and (3) effect of information type, quality, and feedback to users on user decision processes.(145,146) An overview was presented by Mahmassani and Srinivasan.(147)
In the past decade, interest has grown in the potential role of experimental economics approaches in travel behavior research. Methodologically, the following guidelines are generally followed in experimental economics:(141)
· Use real monetary payoffs to incentivize subjects; in other words, the payoffs should be designed so as to induce the same behavioral response as the experienced consequences in a natural context.
· Publish complete experimental instructions.
· Do not use deception. There is considerable debate regarding this matter in the field; experimental evidence suggests that deception (false consequences to deny participants monetary payoffs) leads to unreliable responses and loss of goodwill.
· Avoid introducing specific, concrete context (i.e., keep the decision context stylized and generic and hence transferable and generalizable).
The precepts of experimental economics differ from prevailing practice in transportation and travel behavior research because the latter have generally sought to elicit responses to the actual attributes that influence choices in the real world, rather than some monetary surrogate that may be of questionable realism.
Selten et al. conducted laboratory experiments of a highly stylized day-to-day route choice game with two route alternatives (a main road and a side road) and two experimental treatments corresponding to feedback about one’s own travel time and feedback about the travel times of the alternative route in addition to one’s own route.(148) Each experiment consisted of 18 players at a time (equilibrium consisted of 12 players on the main road and 6 on the side road). Methodologically, the payoffs increased according to a simple linear formula with decreasing travel time, itself related linearly to volume. The researchers ran 200 iterations, considered a long time in experimental economics, but they still encountered large fluctuations. The results seemed to converge toward equilibrium but not perfectly, as fluctuations persisted under both treatments (fluctuations appeared to be smaller under the full-information treatment). Both direct and contrarian response modes could be identified among the players, with direct players changing routes after a bad payoff and contrarians changing routes after a good payoff.
An important element in experimental economics that is of considerable relevance to travel behavior dynamics is the role of learning and judgment in repeated decision situations (e.g., day-to-day adjustment). Psychological studies have examined some of these questions through experiments on individual subjects but have typically ignored the effect of other decisionmakers and different information environments. Information availability plays an important role in determining which theories are feasible in different environments. Economists have investigated learning behavior both experimentally and theoretically but on a macroscopic scale, studying how simple information adjustment rules drive equilibrium processes in games under different information environments.(149)
More recently, opportunities offered by online gaming environments have emerged as a promising approach for studying individual activity and travel choices.(150) Activity and travel behavior in virtual environments is of interest for the following reasons:
· It is a manifestation of human behavior in a domain that is occupying a greater share of the time and resources of a growing segment of the population and increasingly cutting across social, demographic, and economic lines.
· Virtual world engagement is integrally linked to physical world behavior and, as such, becomes essential in studying and predicting behavior in the latter.
· It is likely to provide insight into activity and travel behavior in the physical world and to help identify fundamental mechanisms underlying such behaviors.
· It may eventually provide a laboratory for observing behavior under controlled experimental conditions (e.g., in response to contemplated policies).
However, designing games to address the questions of interest while retaining the players’ engagement remains a challenge.