Can information promote transportation-friendly location decisions? A simulation experiment

Journal of Transport Geography 19 (2011) 304–312

Contents lists available at ScienceDirect

Journal of Transport Geography journal homepage: www.elsevier.com/locate/jtrangeo

Can information promote transportation-friendly location decisions? A simulation experiment Daniel A. Rodríguez a,*, Jonathan Levine b, Asha Weinstein Agrawal c, Jumin Song b a

Department of City and Regional Planning, New East Hall Room 319, CB# 3140, University of North Carolina, Chapel Hill, NC 27599, United States 2208 B Art & Architecture Building, 2000 Bonisteel Boulevard, Urban and Regional Planning Program, Alfred Taubman School of Architecture & Urban Planning, University of Michigan, Ann Arbor, MI 48109-2069, United States c One Washington Square, Department of Urban and Regional Planning, San José State University, San Jose, CA 95192-0185, United States b

a r t i c l e

i n f o

Keywords: Transportation–land use connection Traveler information Location choice Experimental design

a b s t r a c t Where people live, work, shop, and recreate fundamentally determines their local travel options. Yet, information problems such as the cost of conducting comprehensive searches and cognitive load have been shown to limit decision-making. In the context of residential decision-making, information problems are likely to influence which locations get chosen. This study examines whether providing people seeking a rental home with map-based information about the transit and pedestrian accessibility of the available units might influence their residential location choices. More specifically, would some people make use of this information to select more accessible residences than they would have otherwise chosen? This proposition was tested through an experimental research design in a laboratory setting. Graduate student participants were asked to select their top choices of where to live after reviewing a database of residential properties custom-designed for this study. In order to assess the influence of accessibility information, we divided participants randomly into control and experimental groups, with the former receiving baseline information currently available and the latter receiving map-based supplemental property information on multi-modal accessibility to desirable destinations. The study results suggest that providing multi-modal accessibility information to people who are relocating will enhance the attractiveness of locations that support multiple travel modes. If this is true for broader populations, then planners and policy makers may be able to increase use of non-auto modes by providing multi-modal transportation information to people at the time when they are looking for a new residence. Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction In the last three decades, a strong core of policy research and innovation has sought to identify the most effective methods to encourage travelers to voluntarily reduce their driving and make more use of transit, walking, and biking. An emerging strand of this research has focused on whether providing the public with information about their travel options can lead to mode shifts away from driving. The most common approach has been to provide up-to-date (or even up-to-the-minute) information on transportation modes for individual trip-making decisions. For example, some transit systems have installed electronic message boards at bus stops to alert riders when the next bus will arrive. Another approach has been to provide travelers with one-time, personalized counseling on the travel alternatives available to them from their home. * Corresponding author. Tel.: +1 919 962 4763; fax: +1 919 962 5206. E-mail address: [email protected] (D.A. Rodríguez). 0966-6923/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.jtrangeo.2010.04.001

Despite the evident value of these short- and medium-term approaches to providing travelers with information, both suffer from an inherent weakness: once the traveler chooses a place of residence, the transportation die has largely been cast. Transportation decision-making involves both short-run and long-run decisions, with the choice of where to live being the most important of the long-run choices. An information policy that focuses on the daily modal choice alone may be overlooking valuable opportunities to affect travel behavior, because the choice of where to live partly determines the travel patterns of residents. The vast and growing literature associating characteristics of residential environments with residents’ travel choices supports the importance of location decisions for travel behavior (Ewing and Cervero, 2001; Saelens and Handy, 2008). Population and job density, land use diversity, distance to destinations, and street design have been related to mode choice, trip frequency, and distance traveled. One explanation for why location affects travel comes from rational choice theory. Depending on the characteristics of the

D.A. Rodríguez et al. / Journal of Transport Geography 19 (2011) 304–312

residential neighborhood, different travel modes will be more or less convenient. More convenient modes are those whose access, waiting, and travel times decrease as result of a relocation and therefore become a more attractive option for daily travel. A second explanation comes from viewing daily travel behavior as habits that tend to be environmentally conditioned and subject to formation at the moment of a change in one’s environment. In residential locations that support alternative transportation modes, environmental cues can encourage the development and trigger of habits related to non-auto travel. A third explanation is that people’s propensity for travel partly determines their choice of residential environment. Regardless of the causal mechanism(s), the fact that location partly determines travel options and decisions provides a long-term opportunity to change travel behavior by influencing upstream location choices. This study examines whether information about travel options delivered to individuals seeking a new home can alter their choice of residential location and thereby change their transportation environment. Conceptually, information about travel options could influence location behavior in at least two ways. First, it can reveal the existence of available locations, in a process that Chorus et al. (2006a) call alternative generation. Second, the integrated information can assist decision-makers in alternative assessment. That is, the integrated information would contribute to characterizing the alternatives already present in a decision-makers’ choice set, such as identifying travel modes at a particular location, the location’s accessibility to destinations, or its price and related attributes. Individuals would then use these characteristics in determining the desirability of particular locations. To test whether information can alter location decisions and travel options, we conducted a randomized simulation experiment in which students were asked to select their desired location from a database of rental properties. A control group was given conventional information about the characteristics of the properties, including attributes of the physical structure, its location and price. An experimental group was provided additional information about the property’s transit accessibility in an electronic map format. Subjects’ choices were assessed for differences in overall transit accessibility. The remainder of this paper is laid out as follows. In the next section of the paper we discuss the relevant literature on information and travel behavior. Section 3 details the methods used in the study, while Section 4 presents results. In the last section we conclude by suggesting the implications of the study for both policy makers and future research.

2. Relevant literature 2.1. Information and travel behavior Within the area of travel behavior and information, evidence that travelers actually change their behavior as a result of receiving new information is modest. In one study, Al-Deek et al. (1998) estimate that only 11% of drivers would vary their route choices in the presence of radio-provided information. For transit users, the situation appears to be even less promising. Hickman and Wilson (1995) question the benefits of real-time transit information on the basis of finding few changes in route choices, travel time, and trip reliability when such information is provided to transit users. Furthermore, transit users tend to make scant use of static information resources such as maps and schedules when they are available (Balcombe and Vance, 1996; Khattak and de Palma, 1997; Khattak et al., 1999). Non-regular users of transit appear to be even more impervious to efforts at transit information dissemination (Abdel-Aty, 2001; Chorus et al., 2006c).

305

A theme emerging from studies of information use in transportation is the importance of the habitual, satisficing, or otherwise less-than-optimizing behavior of travelers. While standard models of travel behavior assume that people rationally evaluate travel options for each trip, as a practical matter, a number of factors lead to a fair degree of inertia in travel behavior; travelers tend to rely on established behavioral patterns, and it is hard to budge people from these habits. Under ordinary circumstances, travelers make occasional ‘‘strategic” choices to change mode over the long term, but day-to-day modal choice tends to be ‘‘tactical” (i.e., derivative from their long-term behavior), and hence people are unlikely to change their patterns just because they are given new information (Reed and Levine, 1997). 2.2. Travel as habit Travel behavior has been characterized as a ‘‘habit” (Aarts et al., 1997; Matthies et al., 2002) an observation that has significant implications for an information-based strategy aimed at encouraging people to change their travel behavior. Habitual behavior is characterized by lack of awareness, in that people do not think about their actions; by efficiency, in that actions are carried out with little effort; and in some cases, by lack of control. Because habitual behaviors are taken without conscious thought, it is very hard to change them simply by providing information, since most people are unlikely to pay any attention. And the more habitual the behavior, the less the actor seeks or is even amenable to new information that might lead to altered behavior (Verplanken et al., 1997). On the surface, the habitual nature of most daily travel would seem to suggest that it is useless to try to encourage people to change their travel behavior habits by providing them with information; people who habitually rely on their cars will be relatively impervious to information on alternative travel options. This explains why short-run information–provision strategies aimed at influencing people’s daily choice of travel modes appears to have only limited effect. However, psychological research does offer some hope for information provision as an effective policy to promote behavioral change, at least under certain circumstances. Habits tend to be environmentally conditioned (Staats et al., 2004); that is, immediate surroundings send cues that trigger habitual behaviors. The observation that habits are largely environmentally triggered offers an insight into how information can stimulate behavioral change: information coming at the moment of a change in one’s environment is likely to be received better than at other times. The time when people move to a new home should thus be a moment when they are more susceptible to behavioral change in general, including information-induced behavioral change. For example, Wood et al. (2005) documented a change in students’ exercise habits associated with a shift to a different college campus. The environmental conditioning of habits has important transportation implications. In the short term, travel information may be particularly effective when habitual behaviors are disrupted, for example when an accident or construction disrupts a particular route (Chorus et al., 2006b). In the longer term, Heatherton and Nichols (1994) found that moving to a new location increased the likelihood that people successfully translated a desired shift in some aspect of their lives to an actual change. Bamberg, Rolle, and Weber (2003) examined the effects of offering a free bus ticket to people who had just moved, demonstrating that recipients responded with increased transit riding. The notion that people are more open to changing pre-established habits during times of change, when a number of rigid habits become temporarily looser, has clear implications for travel behavior: the time when people

306

D.A. Rodríguez et al. / Journal of Transport Geography 19 (2011) 304–312

relocate to a new environment is a moment during which new travel habits can be formed. 2.3. Determinants of residential decision-making Since location partly determines people’s transportation options and travel decisions, a review of the determinants of residential location decisions is merited. Speare (1974) classified the factors influencing location decisions into four categories: individual, household, housing, and neighborhood. The latter includes characteristics of the neighborhood and the location and accessibility of the neighborhood relative to the metropolitan area. There is disagreement in the literature as to how much weight people place on accessibility when making residential location decisions. Controlled experiments relying on stated preferences and choices have confirmed the importance of factors belonging to Speare’s categorization (Molin and Timmermans, 2003, Zondag and Pieters, 2005; Molin et al., 1999). These studies have identified housing and intra-neighborhood characteristics such as socio-economic status and the quality of public services as more salient than location and accessibility factors. Similar conclusions have been reached by some studies relying on revealed preferences (Giuliano and Small, 1993; Sermons and Koppelman, 1998). An implication of these findings is that attempting to influence how people make residential location decisions through changes in actual or perceived accessibility will yield disappointing results. By contrast, other researchers have found that accessibility is an important factor in people’s decision of where to live. Using data from Calgary (CAN), Abraham and Hunt (1997) found that accessibility variables were the most important location factors influencing residential choices. In a study explicitly designed to examine the relative importance of accessibility, Levine (1998) estimated a nested logit model for the Minneapolis–St. Paul area and found that commuting time was more important than community characteristics such as school quality, taxes, and crime rate. Similar conclusions have been reached in other studies (Bhat and Guo, 2004; Levinson, 1998; Shen, 1998). The role of individual preferences for travel options in influencing residential decisions has received increased research attention. For example, it is likely that individuals with a preference for transit will select a neighborhood that is well-served by transit. The research has largely confirmed that, when possible, segments of the population do select locations partly based on their preferences for specific travel modes (Cao et al., 2006; Handy et al., 2006; Levine et al., 2005). As a result, providing a variety of urban forms and neighborhoods that support diverse modes of transportation would allow individuals to select a neighborhood that closely matches their travel preferences. Providing information about the availability of housing and its accessibility also would enhance people’s ability to self-sort based on their travel preferences. In summary, there is theoretical and empirical support for providing accessibility information about housing at the time when location decisions are being made. The extent to which such information is effective in steering individuals towards more accessible location and in changing travel behavior is an empirical question that remains to be explored. Building on the evidence presented, in this study we examine whether integrated accessibility and housing information presented in the context of a residential relocation will influence people to select residences with higher accessibility than the properties selected without the integrated information. We aim to answer three questions: (1) What is the effect of providing integrated information about housing characteristics and its multi-modal accessibility to destinations on residential choice selection? (2) Do the effects vary by population subgroups?

(3) If an effect is identified, what are the price consequences of selecting more accessible locations? In other words, will people select properties that have higher rents in exchange for obtaining a home with higher non-auto accessibility? 3. Methods This section of the paper describes the simulation experiment used to test the research questions described above.1 The experiment was conducted with 236 University of Michigan (UM) graduate students, who were asked to use a specially crafted housing search database to choose the rental properties they most preferred. Half the participants received information about the transit and pedestrian accessibility of each property in a map format and half did not, allowing us to compare the housing preferences of the two groups. 3.1. Study area The Ann Arbor metropolitan area, located in southeastern Michigan, offers various transportation options for the UM campus community. On-campus parking for students is very limited, with the only other option for drivers being satellite lots requiring students to take a shuttle bus to the campus. Bus transportation is a convenient travel option for UM students. All UM faculty, staff and students receive free, unlimited access to AATA, making bus transportation economically attractive. In addition, the Ann Arbor Transportation Authority (AATA), which operates the area’s public transportation system, provides relatively extensive bus service throughout the whole metro area. Constrained and expensive parking combined with free bus service lead many students to choose public transit for their commute to school. AATA’s routes cover most of the metro area, including downtown Ann Arbor, major shopping malls, and the four University of Michigan campuses: Central, North, Medical, and South Campuses. Central Campus, the main campus, is located close to downtown Ann Arbor, and there is a bus stop within a quarter mile of almost any location in Central campus and the downtown area (Fig. 1). As for North Campus, it is located about 2 miles away from Central Campus. The Medical Campus, which is also removed from the downtown and Central Campus, includes the university hospitals and relevant academic facilities. The South Campus includes only athletic facilities, and since no regular academic activities occur there, it was excluded from this study. 3.2. Research design In a computer laboratory setting, we asked the 236 UM graduate students to select their top five choices of where to live after searching a database of residential properties that was custom-designed for this study.2 The simulation used data from actual rental properties listed on the UM’s off-campus housing database. Of the 9324 residential units in the university’s database, we selected 286 units that provided adequate variation with respect to accessibility, rent, number of bedrooms, furnished versus unfurnished status, availability of off-street parking, and availability of in-building laundry facilities (Table 1). 1 For a more complete description of the study methodology, see Levine et al. (2006). Can consumer information tighten the transportation/land-use link? A simulation experiment. San Jose, CA, Mineta Transportation Institute. 2 We used a power analysis to identify the minimum required sample size that would at least 80% of the time detect a 0.25-mile difference between the experiment and the control groups in the distance from the selected rental unit to a bus stop. The power analysis suggested a sample size of 230 participants (115 in the experimental group and 115 in the control group). The type-I error probability was set to 0.05, and the variance was assumed to be the same for each group.

307

D.A. Rodríguez et al. / Journal of Transport Geography 19 (2011) 304–312

Dexter village Barton Hills village

City of Ann Arbor

North Campus Medical Campus Downtown Central Campus South (Athletic) Campus

City of Ypsilanti

City of Saline

2

0

2

4 Miles

Fig. 1. The four University of Michigan campuses within the Ann Arbor metropolitan area.

Table 1 Characteristics of rental units in the housing database.

a

Variable

Mean

SD

Min.

Max.

Obs.

Monthly rent ($) Bedroomsa (number) Furnished (1 = yes, 0 = no) Pets allowed (1 = yes, 0 = no) Parking (1 = on premises, 0 = no parking or on street) Laundry (1 = yes, on premises 0 = no) Smoking (1 = yes, 0 = no) Wheelchair accessible (1 = yes, 0 = no) Unit type (1 = apartment, 2 = house, and 3 = room) Distance to closest transit stop (miles) Distance to central campus (miles) Distance to north campus (miles) Distance to medical campus (miles) Distance to downtown Ann Arbor (miles)

880.46 1.59 0.23 0.37 0.78 0.83 0.34 0.05 1.67 0.47 2.09 3.33 2.34 2.08

637.74 1.22 0.42 0.48 0.42 0.38 0.48 0.21 0.80 0.53 2.19 2.00 2.11 2.11

250 0.5 0 0 0 0 0 0 1 0.01 0.20 0.62 0.20 0.17

5800 6 1 1 1 1 1 1 3 4.19 14.50 15.04 14.80 15.58

286 286 286 279 286 286 286 263 286 286 286 286 286 286

Studio apartments and single rooms were recoded as 0.5.

To determine the influence of integrated accessibility and housing information, we followed a classic experimental research design by dividing study participants randomly into two groups. The ‘‘control” group received information about each rental unit, in table form, that only included the attributes currently standard in most private and university housing databases, such as price, the number of bedrooms, and the availability of off-street parking. In separate laboratory sessions, the ‘‘experimental” group searched for housing with a database that included all the information available to the control group, along with additional information about how far the unit is from a transit stop, the transit service frequency at that stop, whether travel to campus by bus would require a transfer, and the distance to the part of the university campus that the student visits most often. Data on housing opportunities were presented to the experimental group in map form, with each property classified according to its accessibility to the individual’s UM campus.3 After choosing their desired rental properties, all participants filled out a survey that asked them about their current travel 3 Excellent accessibility: walking distance from campus (<0.5 mile) or bus route (<0.33 mile) with zero transfer and high bus frequency (<15 min headways); high accessibility: walking distance to bus route (0.33 mile) with zero transfer, high bus frequency (<15 min headways), and walking distance to campus P 0.5 mile; medium accessibility: walking distance to bus route with one transfer, regardless of frequency, and walking distance to campus P 0.5 mile; low accessibility: none of the above.

behavior patterns, socio-demographic characteristics, and desired features in housing. For the latter, participants rated the importance of each of 17 housing characteristics (assuming that the rent or purchase price were held the same for all properties). Participants rated the characteristics using a scale of one (unimportant) to four (very important). Collectively, their responses can be interpreted as preferences guiding residential locational decisions. We used electronic mail to recruit the 236 study participants. To avoid bias, graduate students in architecture, urban design, and urban planning were excluded, since these students are more likely to be sensitive to issues of transit use than the general population. We also aimed to have a distribution of participants that resembled the population distribution by UM campus type (20% Medical, 30% North, and 50% Central). As an incentive to encourage participation, participants received $20 in cash. 3.3. Outcomes and hypotheses Relative to individuals in the control group, for the first research question we hypothesized that individuals exposed to the integrated housing and accessibility information provided in a map format (the experimental group) would choose more accessible rental units—units that were closer to transit lines, closer to transit lines with a high level of transit service, closer to transit lines serving a variety of destinations, and closer to destinations. To test this

308

D.A. Rodríguez et al. / Journal of Transport Geography 19 (2011) 304–312

# of routes

ANOVA

Non-auto accessibility % of properties in locations with low, medium, high and excellent accessibility

%

Ordered logistic

we hypothesized that there would be a stronger reaction to the accessibility information among people who do not have strong a priori commitments to driving, transit, or walking and therefore may be more amenable to mode shifts. By contrast, people determined to walk would be expected to seek residences close to their destinations with or without an information system; groups of regular bus riders already know where the routes are and need no information system to help guide them in their residential choices. If transportation ‘‘fence-sitters” can be identified, these are likely to be the individuals most amenable to behavioral change through information provision. For the third research question, about the potential price consequences of selecting more accessible locations, we examined the rental prices of properties selected by the control and experimental groups. We expected more accessible units to be more expensive, all else held equal, as some transportation literature has suggested. By steering participants towards more accessible locations, the experiment group members could select properties with higher rents.

Pedestrian accessibility to destinations Average network distance to closest destinationb Average sum of network distance to four main destinations

Miles Miles

ANOVA ANOVA

3.4. Statistical analyses

Table 2 Accessibility measures for properties selected by participants. Outcome Access to transit Average network distance to transit Average crow-fly distance to transit Average network distance to high-quality transit (to home campus)a Average network distance to high-quality transit (to central campus)a % of preferred properties within 1/4 mile of transit route Total buses per hour within 1/4 mile of properties selected Transit destination diversity Sum of # of routes within 1/4 mile of properties selected

Units

Statistical test

Miles Miles Miles

ANOVA ANOVA ANOVA

Miles

ANOVA

%

Ordered logistic ANOVA

# of buses/h

a High quality transit is defined as a bus route with no-transfer service to the destination of interest—central campus or one’s home campus—and a service frequency of 15 min or better. b Closest destination refers to the closest of the following four main destinations: Central campus, North campus, Medical campus and downtown Ann Arbor.

hypothesis with the simulation data, we defined 10 different measures of accessibility for each property in the simulation database, grouping these measures into four categories (Table 2): (1) Access to transit: This measure relates to the ease and time needed to catch a bus. (2) Transit destination diversity: For this measure we calculated the total number of transit routes passing within a quarter-mile of the unit, to measure how easily residents can reach a variety of destinations by transit (Dimitriou, 1992). (3) Non-auto accessibility to campus: For this holistic measure, we looked at the ease of travel by transit, walking, and cycling, calculating the measure in terms of both proximity to campus and also transit service quality.4 (4) Pedestrian accessibility to key destinations: This measure reflects how easily residents can walk (or bicycle) to campus or downtown Ann Arbor. We also suspected that the accessibility information would impact participants in different ways, depending on their sociodemographic characteristics, current travel behavior, prior exposure to transit, and perceptions of transit convenience. Sociodemographic characteristics examined included age, sex, income, and current housing expenditures. Current travel behaviors were measured using the proportion of trips by different travel modes (Table 3), and prior transit exposure and convenience perceptions were measured with agreement questions using a Likert-like response scale (Table 3). Therefore, for our second research question

4 Units with excellent non-auto accessibility are within 0.5 miles from campus or within 0.33 miles of a bus route with zero transfers and a service frequency lower than or equal to 15 min. Units with high non-auto accessibility are within 0.33 miles of a bus route with zero transfers and a service frequency higher than 15 min. Units with medium non-auto accessibility are within 0.33 miles to a bus route requiring one transfer, regardless of frequency. Units with low non-auto accessibility are all the rest.

To determine the statistical significance of continuous outcome variables between the experimental and the control groups (question 1) and whether subgroups within the experimental group might have been more or less susceptible to acting on the information provided (question 2), we used analysis of variance (ANOVA). For proportion-type variables, such as the percentage of units selected within one-quarter mile of a bus stop, we used ordered logistic regression with the proportion as the dependent variable and a dummy variable identifying group membership as the independent variable. The test-statistic of the dummy variable coefficient reveals whether statistically significant differences between the two groups exist. Finally, to examine whether participants selected properties with higher rents in exchange for obtaining a home with higher non-auto accessibility, we compared median rent prices using a Wilcoxon rank-sum non-parametric test. 4. Results and implications 4.1. Characteristics of participants Table 3 summarizes participants’ responses to our survey. The average age of participants was 26 years, 56% were female, and all reported some yearly household income (average $22,700). The average current monthly rent or mortgage payment was $695 a month, with some participants reporting no payments and others reporting up to $5000 per month. We combined the monthly payments for housing with the income data to calculate a measure of housing expenses (monthly housing expenditures as a percentage of monthly income) and found that participants spent an average of 60% of their household income in rent or mortgage payments. Although this percentage is high compared to that of the general population, it is not unexpected to find that students spend a high percentage of their monthly income on housing. Furthermore, since our focus is on graduate students, it is likely that some students have very little income but are tapping into personal savings or external support to pay for housing while in graduate school. With respect to travel patterns, respondents reported that more than 67% of all their trips to or within the university were by bicycle, walking, or transit. By contrast, participants reportedly made 63% of all other trips during the same time period by car, either as the driver or passenger. This trend for participants to use non-auto modes for campus trips, but cars for most other trips, is

309

D.A. Rodríguez et al. / Journal of Transport Geography 19 (2011) 304–312 Table 3 Summary statistics of survey responses. Description

Mean

SD

Min.

Max.

Obs.

Respondents’ socio-demographic characteristics Age (years) Sex (1 = female) Midpoint of yearly income range ($1000s) Current monthly rent or mortgage ($1000s) Affordability (% of monthly income for housing)

25.7 0.56 22.7 0.70 60.1

4.0 0.5 17.1 0.50 65.2

20 0 5 0 0

52 1 75 5 576

235 236 225 232 222

Current travel patterns % of UM trips by UM transit % of UM trips by non-motorized modes % of UM trips by AATA % of non-UM trips by non-motorized modes % of non-UM trips by AATA % of non-UM trips by car % of all trips (UM & non-UM) by car

17.0 35.2 15.2 26.9 6.9 63.1 42.6

22.3 34.6 25.3 33.1 19.9 38.1 28.3

0 0 0 0 0 0 0

100 100 100 100 100 100 100

236 236 236 225 225 236 236

Respondents’ locational preferences Closeness to destinations among top three reasons to choose a residence (1 = yes) Transit nearby among the top three reasons to choose a residence (1 = yes)

0.58 0.22

0.50 0.42

0 0

1 1

236 236

Respondents’ exposure to transit and perceptions of its convenience Know how to use bus (or know how to find out) (1: strongly disagree; 5: strongly agree) Transit is inconvenient (1: strongly disagree; 5: strongly agree) Was there a period in life when used transit regularly (1 = yes)?

4.46 2.68 0.86

0.81 1.18 0.34

1 1 0

5 5 1

235 236 236

Respondents’ perception of how easy it was to use the search tool Ease of using tool (1: very difficult; 5: very easy)

4.13

0.81

2

5

236

consistent with the urban spatial structure of Ann Arbor, where parking is ample and cheap at most destinations except in the town’s center and at the university. Summary statistics for the 17 survey questions about unit and building preference show significant variation in the criteria that participants deem important. The average rating for quiet units, safety, on-site parking, and size was relatively high, while the average ratings for school quality, building amenities (e.g., an on-site pool or gym), and property type (condo, single family, or townhouse) were the lowest. When asked to select their first, second, and third priorities from among the list of 17 attributes, 58% selected proximity to destinations among their top three criteria, but only 22% selected transit proximity among their top three choices.

4.2. Comparisons between experimental and control groups Table 4 shows mean values of the four accessibility measures for the properties selected by members in the experimental group and the control group. The results suggest no difference between groups in the distance to bus stops and in the number of transit routes serving the preferred housing units. However, the average network distance to high-quality transit differed significantly between the control and experimental groups. The experimental group located nearly one-quarter of a mile closer on average to transit lines offering a no-transfer trip to their main campus destination at a service frequency of 15 min or better. It is not surprising that the two groups differed in their preference for locating near high-quality transit but not for transit lines in general. First, the overall average distance to a transit stop (0.4 of a mile) was quite small to begin with. For this reason, there was little room for the experimental group to reduce that distance noticeably, even with transit information. Equally important, the information system itself did not seek to guide respondents to locations served by transit in general, but to locations served by transit offering frequent, no-transfer trips to their campus destinations. Thus a property along a direct and frequent bus line could receive an ‘‘excellent” accessibility score even if it were several miles from campus. The difference between the control and experimen-

tal groups in this regard indicate a significant response to this information. In addition to the transit–distance effects, we also find evidence that participants in the experimental group were more likely to choose units closer to one of the four major destinations popular among graduate students. More specifically, participants in the experimental group preferred properties located 0.3 miles (about 30%, measured as the average network distance) closer to either one of the three campuses or downtown Ann Arbor than did the control group. Similarly, participants in the experimental group selected locations that were on average 16% closer to all the four destinations of interest to graduate students (measured as the average sum of the network distance to the four main destinations). This result helps explain why participants in the experimental group also appeared to locate in areas with higher transit service frequency than participants in the control group. Areas close to major destinations in Ann Arbor tend to have higher transit service frequency than areas farther away from those destinations. The fact that the intervention included the accessibility information in a map format may help explain these differences in proximity to major destinations. Participants’ choices may be influenced by both the transit information and the visual cues provided by a map. What are the effects of a 0.3-mile change (from 1.3 miles to 1 mile) in the distance to a destination for walking? Three recent studies examining travel mode choice and walking trips shed light on the question: Cervero and Duncan’s analysis of San Francisco Bay Area commuters (2003), Rodríguez and Joo’s study of commuters to the University of North Carolina in Chapel Hill (2004), and Shay et al.’s (2006) analysis of walking trip frequency by distance to a commercial center in a new urbanist neighborhood. These three studies are useful because they study non-motorized travel separate from all travel, and they use distance-related independent variables to simulate changes in mode shares given changes in commuting distance while holding all other variables constant.5 5 One limitation of using the result of these studies to make inferences on a different population is that these studies are cross-sectional, were conducted in different geographic areas and under different market constraints. As a result, associations may be reflecting partly the importance of self-selection.

310

D.A. Rodríguez et al. / Journal of Transport Geography 19 (2011) 304–312

Table 4 Mean value of outcome measures for properties selected by control and experiment groups. Outcome

** *** a

Mean value

p-Value

Control group

Experiment group

Access to transit Average network distance to transit Average crow-fly distance to transit Average network distance to high-quality transit to home campus Average network distance to high-quality transit to central campus % of preferred properties within 1/4 mile of transit route Total buses per hour within 1/4 mile of properties selected

0.426 0.328 0.837 0.653 86.93 3.304

0.390 0.311 0.596 0.510 91.13 3.655

0.251 0.464 0.000*** 0.001*** 0.302 0.037**

Transit destination diversity Sum of # of routes within 1/4 mile of properties selected

6.401

7.053

0.211

Non-auto accessibility % of properties in locations with: Low accessibility Medium accessibility High accessibility Excellent accessibility

10.2 10.2 19.4 54.8

5.8 13.5 15.9 64.8

0.000***

Pedestrian accessibility to destinations Average network distance to closest destinationa Average sum of network distance to four main destinations

1.325 8.081

1.020 6.894

0.008*** 0.013**

Significance at a 95% level of confidence. Significance at a 99% level of confidence. Closest destination refers to the closest of the following four destinations: Central campus, North campus, Medical campus and downtown Ann Arbor.

For walking, Cervero and Duncan’s results suggest that a decrease of 0.3 miles from the mean commuting distance is associated with an increase of 5.6% in the share of walking trips, from 12.5% to 18.1% of all trips.6 Rodríguez and Joo’s results suggest that a similar decrease in distance, from the mean distance in their sample of 2.5 miles, is related to an increase of 8.3% in the share of walking trips. Average distances in this study are significantly shorter than what Cervero and Duncan (2003) and Rodriguez and Joo (2004) observed; for this reason expected changes based on their results are likely to underestimate the effect for our sample. Indeed, Shay et al.’s (2006) results suggest that a decrease of 0.3 miles (from 1.3 to 1.0 miles) in the distance to commercial areas increased walking trip frequency by 57%. Taken together, the results suggest that participants in the experimental group chose locations that are better served by non-automobile modes, as compared with participants in the control group. These locations are characterized by high-quality transit and increased proximity to desirable destinations. 4.3. Comparisons between subgroups of the experimental group and other participants The various tests revealed three significant findings, two of which related to age. First, even though the mean distance between the selected housing units and a bus stop did not differ between the full experimental and control groups, we found that within the experimental group, older participants located closer to transit routes. Looking at the mean network distance between their preferred housing units and a transit stop, for every additional year of age, participants in the experimental group located between 0.02 and 0.03 miles closer to transit (8.2% closer) than younger individuals in the experimental group and individuals of any age in the control group. Second, for bus service frequency we also found that older participants in the experimental group located in areas with higher transit service frequency than did younger members of the experimental group and individuals of any age in the control group. For every additional year of age, participants in the experimental 6 Cervero and Duncan (2003) do not report the mean walking or cycling distance. In their study, they considered only trips of less than 5 miles.

group chose properties that were 4% more likely on average to be within one-quarter mile of a transit stop. Thus, for a 10-year difference in age of participants, the older participants would be 40% more likely to select properties within one-quarter mile of a transit stop than younger participants, holding everything else equal. Third, frequent university shuttle riders in the experimental group made use of the accessibility information offered to them by locating closer to major destinations and in areas that are well-served by transit relative to non-shuttle users in the experimental group and to all individuals in the control group. Participants who self-reported as shuttle users chose housing units that were closer to one of the four major destinations in the area, to transit stops, and to AATA bus routes that offered higher service frequency. Specifically, for an increase in one percentage point in the percentage of UM trips taken by bus, participants in the experimental group chose properties that were:  located near a bus route with 0.4 more buses per hour (an 11% increase in frequency, evaluated at the mean frequency),  9% more likely, on average, to be within one-quarter mile of a transit stop,  0.37 miles closer to one of the four major destinations (or 30% closer, evaluated at the mean distance to the closest major destination), and  1.6 miles closer to all major destinations considered simultaneously (20%, evaluated at the mean distance to all destinations). By contrast, we found no differences in location decisions when we compared participant groups segmented by prior exposure to transit, perceived transit inconvenience, reported ease of using the search tool, being a regular auto driver, or being an AATA transit user. As a result, we conclude that habitual UM transit riders are more open to using information to choose locations that support non-auto travel modes. It may be that UM shuttle riders represent a middle ground between regular AATA bus riders (who will seek out transit-friendly units, even without help) and car drivers (who would be unlikely to use a bus under any circumstances). In other words, UM bus riders may be the ‘‘fence-sitters” who will choose transit-accessible units if they are told which ones those are.

D.A. Rodríguez et al. / Journal of Transport Geography 19 (2011) 304–312

311

4.4. Effect on rents of properties selected

5. Conclusions

A comparison of the average rent of properties selected by participants in both groups shows that the rent for properties selected by participants in the experiment group was $37 per month lower (4.7% lower, p = 0.89) than properties selected by the control group. This rent difference is confirmed by the fact that properties selected by each group (not shown) were similar to each other in their characteristics (other than accessibility). Therefore, experimental group members selected properties that were no better or worse than the control group members, except for having better accessibility.

As researchers and policy makers search for ways to broaden alternatives to automobile use, one avenue they have begun to explore is the use of new information technologies. The hope is that these might enable or even encourage travelers to more often bike, walk, and take transit. We investigated one innovative approach to information provision, that of providing people searching for new housing with detailed map-based information about the transit and walking options to popular destinations from each available unit. We found that providing housing seekers with map-based information about the ease of using transit or walking raised the probability that some of them would choose a housing unit where those modes would be relatively convenient. Respondents in the experimental group selected locations with excellent accessibility about 10% more often than those in the control group. Compared to the control group, experimental group participants also chose units that were nearly one-quarter of a mile closer to high-quality (that is, direct and frequent) transit service and that were closer to major destinations by 27–30%. We also found that experimental group members selected properties with slightly lower rents than control group members. Although we are not sure whether people choosing more accessible locations would travel differently, there is a strong possibility that increased walking and transit use would result. Empirical evidence to date suggests that travelers find out-of-vehicle time (e.g., walking, waiting or transferring) several times more onerous than in-vehicle travel time and that transfers between vehicles carry a perceived ‘‘penalty” beyond the sheer time involved in switching between vehicles (Goodwin, 1993; Koppelman and Wilmot, 1982), so it is reasonable to assume that people who choose to live close to high-quality transit service will indeed use transit more than people living farther from such service. Other research suggests that people are more likely to walk when distances are shorter (Cervero and Duncan, 2003; Rodriguez and Joo, 2004; Shay et al., 2006), which suggests that people will walk more if they end up living closer to their major destinations. Overall, the positive findings of this experimental research underscore the importance of identifying planning opportunities that help those who value accessibility to easily choose such locations. Many localities are currently developing highly accessible city spaces, but little change in travel behavior may result if high-accessibility areas are not identified as such by those members of the public who prefer to reduce their driving. The study’s findings also contribute a new angle to the ongoing debates within the transportation research community about the strength and nature of the transportation-land use relationship. Rather than asserting the existence of a strong or weak relationship, our findings suggest a third option: the relationship between transportation and land use can be either nurtured or thwarted by policy. Appropriate interventions can increase the capacity of transportation accessibility to guide locational decisions, thus strengthening the transportation-land use relationship. And as this research suggests, integrated transportation and housing information, offered at the time of a residential locational decision, may constitute one of these interventions.

4.5. Limitations Two chief benefits of our experimental research design are that it overcomes sample selection bias, since we randomly assigned participants to the two groups, and that it provides a clear indication of causality. For these reasons, we believe the results provide support for policies that provide home-seekers with multi-modal transportation information in a map format as a strategy to have them choose locations with higher accessibility. By giving more ‘‘fence-sitters” the opportunity to find homes that will allow them convenient transit, walk, and bike options, prior research suggests that use of non-auto modes might increase accordingly. At the same time, four features of the research design limit our ability to conclusively generalize the finding that providing housing seekers with information about multi-modal accessibility led participants to choose more accessible units. More research is needed to confirm whether the study findings do indeed apply to the wider population. The first limitation is our reliance on stated preferences in a laboratory simulation, rather than using revealed preferences. This is a common limitation of experimental research (Kitamura et al., 1995; Mahmassani and Liu, 1999). However, stated preference surveys about housing choice can be fairly accurate, as suggested by Earnhardt’s (2002) findings that actual and stated preferences for housing choices follow similar decision processes with respect to key attributes of the property. A second limitation of the research design is that the laboratory simulation forced individuals to make decisions within a shortterm horizon and with limited information. When considering housing location decisions, and particularly when making housing purchasing decisions, individuals may not only ponder their choices over a longer time, but also are likely to visit each location. Such visits would provide decision-makers with detailed information about accessibility to destinations over and beyond what was provided to the control group in our simulation. Nevertheless, providing accessibility information could determine people’s final choice of housing by influencing the set of units they choose to visit in the first place. A third limitation is that our intervention bundled the housing and accessibility information together by displaying it in a map format, while the control group had neither a map nor the accessibility information. Rather than attempt to isolate the effect of the information from the effect of the map, we view the intervention as the bundling of both in a single, map-based instrument. A fourth and final limitation of the research design is that it relied on participants, graduate students, who do not accurately reflect the larger population. Although this factor limits our ability to predict whether the effects found in the study can be translated to the broader population or to other sub-populations (Zellner and Rossi, 1986), similar studies of transportation behaviors in student populations have revealed insightful behavioral patterns (Bamberg and Schmidt, 2003; Collins and Chambers, 2005).

Acknowledgments We gratefully acknowledge the support of the Mineta Transportation Institute for this study. We also especially thank Katja Irvin for her contributions throughout the research process, as well as Melissa Goldstein, Jessica Good, Alan J. Levy, Richard Murphy, Craig Snyder, Julie L. Stetten, Chris Weisen, and Jessica Zgobis for

312

D.A. Rodríguez et al. / Journal of Transport Geography 19 (2011) 304–312

their help with various parts of the study. Finally, we thank the two anonymous reviewers for their thoughtful and detailed suggestions. References Aarts, H., Verplanken, B., Van Knippenberg, A., 1997. Habit and information use in travel mode choices. Acta Psychologica 96, 1–14. Abdel-Aty, M.A., 2001. Using ordered probit modeling to study the effect of ATIS on transit ridership. Transportation Research Part C 9, 165–277. Abraham, J.E., Hunt, J.D., 1997. Specification and estimation of a nested logit model of home workplace and commuter mode choice by multiple worker households. Transportation Research Record 1606, 17–24. Al-Deek, H.M., Khattak, A.J., Thananjeyan, P., 1998. A combined traveler behavior and system performance model with advanced traveler information systems. Transportation Research Part A – Policy and Practice 32, 479–493. Balcombe, R.J., Vance, C.E., 1996. Information for Bus Passengers: A Study of Needs and Priorities. Transportation Research Laboratory, Crowthorne, UK. Bamberg, S., Rolle, D., Weber, C., 2003. Does habitual car use not lead to more resistance to change of travel mode? Transportation 30, 97–108. Bamberg, S., Schmidt, P., 2003. Incentives, morality, or habit? Predicting students’ car use for university routes with the models of Ajzen, Schwartz, and Triandis. Environment and Behavior 35, 264–285. Bhat, C.R., Guo, J., 2004. A mixed spatially correlated logit model: formulation and application to residential choice modeling. Transportation Research Part B – Methodological 38, 147–168. Cao, X.Y., Handy, S.L., Mokhtarian, P.L., 2006. The influences of the built environment and residential self-selection on pedestrian behavior: evidence from Austin, TX. Transportation 33, 1–20. Cervero, R., Duncan, M., 2003. Walking, bicycling, and urban landscapes: evidence from the San Francisco Bay Area. American Journal of Public Health 93, 1478– 1483. Chorus, C.G., Arentze, T.A., Molin, E.J.E., Timmermans, H.J.P., Van Wee, B., 2006a. The value of travel information: decision strategy-specific conceptualizations and numerical examples. Transportation Research Part B – Methodological 40, 504– 519. Chorus, C.G., Molin, E.J.E., Van Wee, B., 2006b. Travel information as an instrument to change car-drivers’ travel choices – a literature review. European Journal of Transport and Infrastructure Research 6, 335–364. Chorus, C.G., Molin, E.J.E., Van Wee, B., Arentze, T.A., Timmermans, H.J.P., 2006c. Responses to transit information among car-drivers: regret-based models and simulations. Transportation Planning and Technology 29, 249–271. Collins, C.M., Chambers, S.M., 2005. Psychological and situational influences on commuter-transport-mode choice. Environment and Behavior 37, 640–661. Dimitriou, H.T., 1992. Urban Transport Planning: A Developmental Approach. Routledge, London, New York. Earnhart, D., 2002. Combining revealed and stated data to examine housing decisions using discrete choice analysis. Journal of Urban Economics 51, 143– 169. Ewing, R., Cervero, R., 2001. Travel and the built environment. Transportation Research Record 1780, 87–114. Giuliano, G., Small, K.A., 1993. Is the journey to work explained by urban structure? Urban Studies 30, 1485–1500. Goodwin, P.B., 1993. A review of new demand elasticities with special reference to short and long run effects of price changes. Journal of Transport Economics and Policy 26, 155–163. Handy, S., Cao, X.Y., Mokhtarian, P.L., 2006. Self-selection in the relationship between the built environment and walking – empirical evidence from northern California. Journal of the American Planning Association 72, 55–74. Heatherton, T., Nichols, P., 1994. Personal accounts of successful versus failed attempts at life change. Personality and Social Psychology Bulletin 20, 664–675. Hickman, M.D., Wilson, N.H.M., 1995. Passenger travel time and path choice implications for real-time transit information. Transportation Research Part C 3, 211–226.

Khattak, A., De Palma, A., 1997. The impact of adverse weather conditions on the propensity to change travel decisions: a survey of Brussels commuters. Transportation Research Part A 31, 181–203. Khattak, A., Yim, Y., Stalker, L., 1999. Does travel information equally influence commuter and non-commuter behavior? results from the San-Francisco Bay Area TravInfo project. Transportation Research Record 1694, 48–58. Kitamura, R., Reddy, P., Vaughn, K.M., Jovanis, P.P., 1995. Transit pre-trip information systems: an experimental analysis of information acquisition and its impacts on mode use. In: Proceedings of the Second World Congress on Intelligent Transportation Systems. Koppelman, F.S., Wilmot, C.G., 1982. Transferability analysis of disaggregate choice models. Transportation Research Record 895, 18–24. Levine, J., Inam, A., Torng, G.W., 2005. A choice-based rationale for land use and transportation alternatives – evidence from Boston and Atlanta. Journal of Planning Education and Research 24, 317–330. Levine, J., Rodriguez, D.A., Song, J., Weinstein, A., 2006. Can Consumer Information Tighten the Transportation/Land-use Link? A Simulation Experiment. Mineta Transportation Institute, San Jose, CA. Levine, J.C., 1998. Land use solutions to transportation problems? rethinking accessibility and jobs-housing balancing. Journal of the American Planning Association 64, 133–149. Levinson, D.M., 1998. Accessibility and the journey to work. Journal of Transport Geography 6, 11–21. Mahmassani, H.S., Liu, Y.-H., 1999. Dynamics of commuting decision behaviour under advanced traveller information systems. Transportation Research Part C: Emerging Technologies 7, 91–107. Matthies, E., Kuhn, S., Klockner, C.A., 2002. Travel mode choice of women – the result of limitation, ecological norm, or weak habit? Environment and Behavior 34, 163–177. Molin, E., Oppewal, H., Timmermans, H., 1999. Group-based versus individualbased conjoint preference models of residential preferences: a comparative test. Environment and Planning A 31, 1935–1947. Molin, E., Timmermans, H.J.P., 2003. Accessibility considerations in residential choice decisions: accumulated evidence from the Benelux. In: Presented at the Annual Transportation Research Board Meeting, Washington, DC. Reed, T., Levine, J., 1997. Changes in traveler stated preference for bus and car modes due to real-time schedule information: a conjoint analysis. Journal of Public Transportation 1, 25–47. Rodriguez, D., Joo, J., 2004. The relationship between non-motorized mode choice and the local physical environment. Transportation Research Part D 9, 151–173. Saelens, B., Handy, S., 2008. Built environment correlates of walking: a review. Medicine & Science in Sports & Exercise 40, S550–566. Sermons, M.W., Koppelman, F.S., 1998. Factor analytic approach to incorporating systematic taste variation into models of residential choice. Transportation Research Record 1617, 194–202. Shay, E., Fan, Y., Rodriguez, D., Khattak, A., 2006. Drive or walk? Utilitarian trips within a neo-traditional neighborhood. Transportation Research Record 1985, 154–161. Shen, Q., 1998. Spatial technologies, accessibility, and the social construction of urban space. Computers, Environment and Urban Systems 22, 447–464. Speare, A., 1974. Residential satisfaction as an intervening variable in residential mobility. Demography 11, 173–188. Staats, H., Harland, P., Wilke, H.A.M., 2004. Effecting durable change – a team approach to improve environmental behavior in the household. Environment and Behavior 36, 341–367. Verplanken, B., Aarts, H., Van Knippenberg, A., 1997. Habit, information acquisition, and the process of making travel mode choices. European Journal of Social Psychology 27, 539–560. Wood, W., Witt, M.G., Tam, L., 2005. Changing circumstances, disrupting habits. Journal of Personality and Social Psychology 88, 918–933. Zellner, A., Rossi, P., 1986. Evaluating the methodology of social experiments. In: Munnell, A. (Ed.), Lessons from the Income Maintenance Experiments. Federal Reserve Bank of Boston, Boston. Zondag, B., Pieters, M., 2005. Influence of accessibility on residential location choice. Transportation Research Record 1902, 63–70.