Illuminating the unseen in transit use: a framework for examining the effect of attitudes and perceptions on travel behavior

Transcription

1 Illuminating the unseen in transit use: a framework for examining the effect of attitudes and perceptions on travel behavior (Accepted for publication in Transportation Research Part A) Steven Spears, 1 Douglas Houston, 1 and Marlon Boarnet 2 1 Department of Planning, Policy & Design School of Social Ecology University of California, Irvine Irvine, California USA 2 Sol Price School of Public Policy University of Southern California Los Angeles, California USA Corresponding author: Steven Spears steven.spears@uci.edu +1(804)

2 Abstract This study develops the Perception-Intention-Adaptation (PIA) framework to examine the role of attitudes, perceptions, and norms in public transportation ridership. The PIA framework is then applied to understand the relative importance of socio-demographic, built environment, transit service, and socio-psychological factors on public transit use for 279 residents of south Los Angeles, California, a predominately low-income, non-white neighborhood. Confirmatory factor analysis based on 21 survey items resulted in six transit-relevant socio-psychological factors which were used in regression models of two measures of transit use: the probability of using transit at least once in the 7-day observation period, and the mean number of daily transit trips. Our analysis indicates that two PIA constructs, attitudes toward public transportation and concerns about personal safety, significantly improved the model fit and were robust predictors of transit use, independent of built environment factors such as near-residence street network connectivity and transit service level. Results indicate the need for combined policy approaches to increasing transit use that not only enhance transit access, but also target attitudes about transit service and perceptions of crime on transit. Keywords Public transportation, travel behavior, attitudes, built environment, Los Angeles Research Highlights Examined transit use for 279 residents of south Los Angeles, California Analyzed socio-demographic, built environment, and socio-psychological factors on transit use Conducted factor analysis to develop transit-relevant socio-psychological factors Two factors, transit attitudes and safety concerns, were related to transit use Suggests policies should both enhance transit access and target attitudes and perceptions 1

3 1. Introduction Understanding the factors that facilitate the use of sustainable travel modes like public transportation has become a major focus of transportation planning over the past few decades. Empirical research in this area has resulted in a large number of studies designed to identify characteristics of the built environment, such as residential density, street design, transit service level, and proximity of employment and shopping, that are associated with transit use. Recent meta-analytical reviews of these studies have concluded that there is convincing evidence of a link between the use of public transportation and characteristics of the built environment, including density, diversity, design, destination accessibility, and distance to transit (Cervero and Kockelman 1997; Ewing and Cervero 2001, 2010; van Wee, 2002). These built environmental features are the foundation of a number of planning reform movements, including Smart Growth, New Urbanism, Transit Oriented Development, and Active Living by Design (Ewing et al. 2011), whose principles have been adopted by cities around the world. Though strong evidence exists for a link between the physical environment and travel behavior, a smaller body of research indicates that mode choice and distance traveled have a significant psychological dimension. Studies have shown that individuals travel and adapt their travel behavior depending on their perceptions, attitudes, and preferences (Gehlert et al. 2013; Bohte et al., 2009; Fujii and Gärling, 2003). Many of the studies in this area have utilized well proven attitude-behavior theories from social and environmental psychology, such as the theory of reasoned action (TRA, Fishbein and Ajzen, 1975), theory of planned behavior (TPB, Ajzen, 1991), the norm activation model (NAM, Schwartz, 1977; Schwartz and Howard, 1981) and the transtheoretical model (TTM, Prochaska and Velicer, 1997). Research has shown that attitudes toward transit (i.e. Bamberg and Schmidt, 2001; 2

4 Anable 2005), norms related to car use and the environment (i.e. Nordlund and Westin, 2013; Haustein et al. 2009; Klöckner and Matthies, 2004), perceived control over travel behavior (i.e. Eriksson and Forward, 2011; Gardner and Abraham, 2007; Haustein and Hunecke, 2007), and concerns about traffic and personal safety (i.e. Elias and Shiftan, 2012; Blainey et al. 2012), all play a role in the decision to use public transportation. In addition, studies of travel behavior change have found that transit use can be facilitated by soft measures that are designed to increase knowledge about and positive attitudes toward non-car travel modes (Möser and Bamberg, 2008; Cairns et al., 2008; Fujii et al., 2009). These factors are rarely accounted for in studies of land use and transit ridership patterns. Though attitude-based studies of transit use provide additional insight into the processes that affect transit use, most have a common shortcoming. Just as studies that focus on the built environment do not typically account for psychological factors, most attitudinal studies do not control for the effect of the built environment. Several researchers have identified the need for a more robust approach to travel behavior research that integrates both approaches and bridges this knowledge gap (Van Acker et al. 2010; Handy, 2005). The few studies that have begun to bridge this gap provide evidence that sociopsychological factors are significant predictors of mode choice and/or distance traveled, even after controlling for neighborhood characteristics (Wang and Chen, 2012; Kitamura et al. 1997; Næss, 2005; Cao, Mokhtarian and Handy 2009a; Bhat and Guo, 2007; Hunecke, 2008). However, almost none of the empirical land use - travel behavior studies that attempt to account for spatial, socio-demographic, and socio-psychological aspects of travel have developed a theoretical framework that is tailored to directly measure and account for perceptions and attitudes associated specifically with travel behavior (Van Acker et al. 2010). Exceptions include Van Acker et al. (2011), who examined leisure trip mode choice using a framework that accounted for objective characteristics and individual subjective 3

5 evaluations of the built environment, lifestyle, and travel. However, no study we are aware of has employed a similar combined framework to empirically examine transit use in daily travel. This lack of a theoretical framework makes systematic evaluation of transit use more difficult and policy directions less clear (Richter, et al., 2011). The current study contributes to the literature on transit use by bringing together spatial, built environmental and cognitive aspects of travel behavior. Specifically, it expands previous research by providing a theoretical framework for understanding how socio-psychological factors interact with built environment characteristics to affect transit use. Building on work on the determinants of transit use (Cervero and Kockelman, 1997), psychological aspects of mode choice (Gehlert et al. 2013), and the role of personal safety concerns on transit (Elias and Shiftan, 2012), the results of this study contribute to the understanding of how complementary hard (i.e., infrastructure investments and land use changes) and soft (i.e., educational campaigns to encourage behavior change) policy measures can promote the use of public transportation and improve urban sustainability. The remainder of this article proceeds as follows. First, we present a theoretical framework, which we call the Perception-Intention-Adaptation (PIA) model of travel behavior. The PIA model accounts for the contribution of both built environment and socio-psychological factors, and incorporates an expanded version of the theory of planned behavior (TPB, Ajzen, 1991) to model the attitude-behavior relationship. Second, we test the validity of the socio-psychological constructs in the PIA framework using confirmatory factor analysis (CFA). Third, we specify regression models of transit use that include factors derived from the CFA and variables designed to measure the characteristics of the built environment (Ewing et al. 2011) that have been shown to influence travel behavior. To our knowledge, this is the first study to assess the relative importance of built 4

6 environmental and transit-relevant socio-psychological factors on transit use specifically. 2. Theoretical Framework The current study employs a theoretical framework designed to examine how attitudes, norms, and perceptions affect transit use. This framework, the Perception-Intention-Adaptation (PIA) model, builds on earlier models of travel behavior adaptation, such as the model conceived by Fried, Havens, and Thall (FHT) (Fried et al. 1977; Havens 1981), an open system model of travel behavior based on individual adaptive response to imbalances in person-environment fit. In the FHT model, sociodemographic variables such as residential location, lifestyle, and household role patterns are determinants of travel-activity patterns, attitudes, and perceived travel needs. To meet these needs, individuals adapt their behavior based on their perceptions of the urban opportunity structure, financial and time constraints, and household role requirements. Like the FHT model, our PIA model focuses on the role of social and psychological factors in the relationship between the built environment and travel behavior. However, the PIA model moves beyond earlier models by incorporating an expanded version of the Theory of Planned Behavior (TPB, Ajzen, 1991), an empirically tested framework used to understand the relationship between attitudes, behavioral intention, and behavior. The TPB is an expectancy-value theory that focuses on behavioral intention strength as the prime determinant of behavior (Fishbein and Ajzen, 1975; Ajzen, 1991). It posits that individuals form intentions based not only on their attitude toward the behavior and its outcomes, but also on their perceived ease of performing the behavior and the social pressures they feel from others who are significant to them. The combined strength of these three factors (attitude, perceived behavioral control, and social norms) determines the strength of the intention to perform the target behavior. When 5

7 confronted with multiple competing alternatives (for example, the choice of travel mode) the TPB posits that individuals choose the behavior for which they have the highest intention strength (Bamberg et al., 2011). There is strong empirical support for the TPB's ability to predict a wide variety of behaviors, from eating disorders to voting choice to smoking cessation (Armitage and Connor, 2001). It has also been used in transportation studies, including analysis of public transportation use. For example, Thøgersen (2006) found that attitude toward transit, perceptions about whether transit could meet travel needs, and car ownership all predicted transit use among Danish residents. Hunecke et al. (2008) investigated mode choice in three large German cities and found that ecological norms, control beliefs, and attitudes toward different modes were all significant predictors of public transportation use. In a longitudinal study of public transportation use, Bamberg et al. (2003) determined that attitudes, social norms, and perceived behavioral control all predicted transit use among a group of university students. In addition, an intervention in the form of a free bus pass increased transit ridership by affecting the determinants of behavioral intention in a positive way. Expanded versions of the TPB have been used to capture other factors relevant to travel mode choice, such as environmental norms and habit, resulting in enhanced predictive capability (Thøgersen, 2006; Verplanken et al., 1994; Bamberg, 2000; Bamberg et al., 2003; Anable, 2005; Hunecke et al., 2008). Integration of the TPB into the PIA model helps to clarify many of the attitude-behavior relationships identified in the previous theoretical conceptions of travel behavior and allows key constructs to be more easily operationalized. Figure 1 shows our Perception-Intention-Adaptation (PIA) model. The main hypothesis of the PIA model is that both the physical environment and cognitive processes have a direct effect on travel behavior. The physical environment affects behavior through the spatial and temporal distribution of 6

8 activity opportunities and transportation infrastructure. Examples include land use mix, access to employment and services, traffic congestion levels, and public transportation quality. This is the area of focus for typical land use travel behavior studies. For a detailed review of this subject, see Ewing and Cervero, (2010). The second main hypothesis of the PIA is that cognitive processes also have a direct influence on travel behavior. Cognitive processes affect travel behavior through both conscious planning (behavioral intention) and automatic processes (habits formed by repeated past behavior). Cognitive factors that influence travel mode, frequency, and distance include attitudes toward various modes, perceptions about the built environment, social and personal norms, lifestyle desires, and perceived travel necessities. Underlying these attitudinal characteristics are a set of longer-term sociodemographic, social, and lifestyle factors. For example, Salomon and Ben-Akiva (1983) theorize that life cycle and lifestyle comprise the highest level of a hierarchical structure in which daily travel behavior reflects short-term decisions made to satisfy lifestyle desires and demographics that evolve slowly over a long period of time. In addition, demographics and household composition act as constraints that influence the expression of lifestyle desires (Van Acker et al. 2010). Though the PIA model posits that travel behavior is determined by factors related to the physical environment and cognitive processes, individuals also evaluate their desired and actual activity and travel patterns and adapt their behavior as necessary. Person-environment fit theory (French, Rogers and Cobb, 1974; Caplan and Harrison, 1993) states that stress arises when the environment does not meet a person s needs. Coping, which is a response to lack of personenvironment fit, can occur through adaptation to the environment or attempts to alter the environment (French, et al. 1974). In the context of the PIA framework, individuals cope with discrepancies in their 7

9 activity desires by altering their travel patterns, their lifestyle desires, or their environment. The degree of success in coping with the daily desires for activity participation is closely related to an individual s subjective well-being, which has recently been proposed as a complement to utility maximization in travel behavior (Ettema, Gärling Olsson & Friman, 2010). For example, Ettema et al. (2010) propose that travel affects subjective well-being through satisfaction derived from activity participation, affective aspects of travel itself, and the stress relieved through ease of mobility and organization of travel. The adaptive behavior that individuals undertake to cope with this stress and maximize their subjective well-being is represented in the model by a feedback loop between current travel behavior and the cognitive processes that underlie travel decision making. These adaptive processes drive behavior changes in response to modifications in the built environment, and allow people to resolve dissonance between their desired and actual activity participation patterns. Examples of adaptation include short term actions such as changing the timing of activities, or buying or selling a vehicle. In the longer term, individuals may attempt to alter their living environment by, for example, changing residential location to be closer to a workplace or other desirable activities or engaging in neighborhood improvement efforts. 3. Study Design In order to test the empirical validity of the PIA model, we apply the framework to a study of households in several neighborhoods of south Los Angeles, California, as part of the Expo Line light rail study. The study area (Figure 2) was comprised of about 43% African-American, 41% Hispanic, and 9% Non-Hispanic White residents in 2010; approximately 20% were poor, about 30% of residents 8

10 were foreign-born, and about 24% had an educational attainment of a Bachelor s degree or higher (U.S. Census Bureau, Decennial Census, 2010; U.S. Census Bureau, American Community Survey, ). The objective of the Expo Line Study is to conduct the first ever longitudinal, experimentalcontrol group, before-after study of the impact of a major transportation investment in California. Results will illuminate both how the Expo light rail transit line changes driving, walking, and bicycling in ways that can enhance the quality of life in the south Los Angeles neighborhoods traversed by the new rail transit, and to understand how the beneficial impacts of the Expo Line can be enhanced. The current study uses cross-sectional data obtained between September 2011 and February 2012, prior to the opening of the Exposition Line, to test the relative contribution of the PIA model s sociopsychological constructs to the prediction of transit use. The first objective of this study is to examine the extent to which the responses to the attitudinal questions in our survey instrument correspond with the hypothesized transit-relevant constructs in the PIA model. These constructs include core elements of the TPB: attitude toward transit, transit-related social norms, and perceived control over travel behavior. In addition, we include personal environmental norms, perceptions of nearby neighborhood amenities, and concern about personal safety on transit, which have been found to be relevant in previous research (i.e. Anable, 2005; Hunecke et al., 2008). Multiple overlapping questions relating to these constructs were developed through a review of the literature. In order to verify that the questions we chose to operationalize the theoretical constructs of our model are reasonable and reliable, we develop a measurement model using a confirmatory factor analysis (CFA) framework. Factor scores derived from the CFA are then used in the next step of the analysis, which uses regression models to examine the effects of transit-relevant 9

11 socio-psychological constructs on transit use. Because questionnaires that include large numbers of attitudinal variables are not often used in travel behavior research, few examples of factor analysis exist in this literature. However, factor analysis is frequently used in studies of attitude-behavior relationships, such as consumer response to new market innovations (i.e. Hauser and Urban, 1977) and studies of how attitudes toward the environment affect behavior (i.e. Poortinga et al., 2004). Examples in the travel behavior literature include Heath and Gifford (2002), who used factor analysis with a set of normative questions relating to transit use, Anable (2005), who used it as an intermediate step in segmentation study of modeswitching potential, and Hunecke et al. (2008), who used factor analysis to examine the role of attitudes, norms, and beliefs on mode choice. The second objective of the current study is to assess the influence of these transit-relevant attitudinal factors on the probability that a respondent used public transportation, after accounting for socio-demographic and built environment variables commonly used in travel behavior research. These models will help identify psychological factors that are significant to transit use and quantify the predictive power of these factors both alone and in combination with built environment variables. The method extends work by Hunecke et al. (2008), who used hierarchical regression models to identify the significant determinants of travel behavior for a variety of modes among residents in Germany, by using more detailed built-environment and transportation system data. Whereas Hunecke et al. (2008) categorized households only by district type (inner-city, city-border, suburban), our data includes nearresidence measures of walkability, proximity to major roads, employment access, and transit level of service. In addition, our study uses observations from a 7-day travel and mileage dairy rather than selfreports of trip making and mode choice. 10

12 4. Data and Methods Data for this study were obtained through a 7-day travel survey of residents in south Los Angeles, conducted between September 2011 and February In order to identify potential participant households, we purchased a mailing list of all addresses within the study area from InfoUSA, a commercial provider of marketing information. We obtained addresses for a total of 27,275 households. Each household was mailed a postal letter inviting them to take part in the study. Letters were mailed in waves between September and November A total of 651 households (2.4%) indicated they were interested in the study by completing the introductory questionnaire. These households were mailed a complete set of study materials. Of those that received the materials, 279 (42.9%) submitted a complete set of responses and these households comprise the sample for the current study. 4.1 Study Groups and Data Collection Method Based on responses to the introductory questionnaire, the 279 participating households were separated into three groups: web-based (82 participants who completed survey components online), paper-based (58 participants who completed survey components using hard copy materials), and mobile tracking (139 participants who completed survey components using hard copy materials and also participated in supplemental location and activity tracking by carrying portable monitoring devices). All participants were provided with instructions and a 7-day travel log, and they recorded their daily count trips by travel mode (passenger vehicle, public transit, walking, and cycling) during their assigned survey travel week. 11

13 Participants who indicated they preferred to complete the study using the website (the webbased group), were provided with a password and username for the project website. Participants were instructed to log in and complete the baseline survey and to enter the information they recorded on their paper 7-day travel logs. Those who either did not have access to the internet, or preferred to mail their materials to us, received a paper version of the survey materials (the paper-based group). A selfaddressed postage-paid envelope was provided to facilitate return of the completed materials. The survey instruments included in the paper group packet were identical in content to those available on the web-based survey. Participants in the mobile tracking group were provided survey materials during a meeting with our research staff, and their responses were collected and checked for completeness during a second meeting approximately one week later. Mobile tracking data were not analyzed for the current study, but will be used in future research to examine intensity and duration of physical activity, route choice, and activity locations. 4.2 Survey Instrument Primary respondents in each household were asked to complete a questionnaire that included demographic, attitudinal, and personal safety related questions. The attitudinal statements in the survey were chosen to operationalize the constructs of the PIA model. Survey items were adapted from questions shown to affect travel behavior and transit use in previous similar studies (e.g. Hunecke et al. 2008, Anable, 2005; Heath and Gifford, 2002). Specifically, participants were asked questions related to the following: Privacy and Crowding Concerns Attitudes toward privacy and crowding on public transportation versus private vehicles. 12

14 Transit Attitudes Attitudes toward attributes and outcomes related to the public transportation system, such as convenience, travel time, cost, and ease of use. Transit Social Norms Social norms relating to expectations/support for transit use from friends and family. Perceived Travel Behavior Control Perceived control over travel behavior and perceived travel necessities i.e. the need for a car, activity scheduling demands, and the need to travel for others. Personal Environmental Norms Personal norms about conservation and protecting the environment. Transit Personal Safety Concerns Fear of crime in the neighborhood, at transit stops, and on transit vehicles during the day and night. Perceived Neighborhood Amenities Perceptions about the availability of shopping, services, restaurants, and recreation within walking distance of home. In all, participants were asked to rank 34 overlapping socio-psychological statements on a 7- point Likert scale (1=strongly disagree, 2=moderately disagree, 3=slightly disagree, 4=neither agree nor disagree, 5=slightly agree, 6=moderately agree, 7=strongly agree). Participants were also asked to rate their level of fear when walking in their neighborhood, waiting at a transit stop, and riding on transit vehicles during both the day and night. A total of 21 response items specific to transit use were retained for the current study. Socio-demographic variables used in the model include factors shown in previous research to most strongly affect the choice of public transportation as a travel mode (i.e. Hunecke et al. 2008; 13

15 Heath and Gifford, 2002). Sample characteristics reflect the lower-income and non-white character of the study area. Nearly 40 percent of respondents reported not being employed during the study period, and 40 percent had a household annual income of less than $35,000 (Table 1). This compares to a median household income of $56,266 for Los Angeles County and $61,632 for the State of California (U.S. Census Bureau, 2013). Approximately 70 percent of respondents were female, and 50 percent were African-American. We conducted additional advertising in neighborhoods in predominately Hispanic areas in both Spanish and English to enhance participation by Hispanics, but efforts were largely unsuccessful. The age of survey respondents ranged from 17 to 86, with a mean of 50 years old. About one quarter of respondents used transit at least once during their assigned 7-day reporting period. 4.3 Built Environment and Transit Measures In addition to data collected from respondents, we constructed a number of near-residence measures designed to capture built environment characteristics which may be associated with transit use, including measures of local employment, shopping, walkability, and transit service (Table 2). We estimated street connectivity based on the number of street intersections within 0.5 miles of a participant s residence using 2010 Topologically Integrated Geographic Encoding and Referencing (TIGER) roadway data from the U.S. Census Bureau. We estimated nearby traffic volumes within 0.5 miles based on 2005 Highway Performance and Monitoring System data maintained by the California Department of Transportation, which includes estimates of annual average daily traffic on highways and major arterial roads. We estimated near-residence land-use composition with 2011 land-use data from the Southern California Association of Governments (SCAG). We developed and tested several 14

16 other near-residence built environmental factors that did not have an influence on transit use and were excluded in the final analysis: distance to nearest freeway ramp, nearby businesses, and land use composition. We tested these proximity-based built environment measures at two different geographic scales (¼ and ½ mile from the respondent s home). We estimated the distance to the nearest transit stop and the total number of nearby rapid or express bus route stops within 0.25 miles based on the data from SCAG, which included the geographic coordinates of transit stop locations for all public transportation providers in the study area. We also estimated transit service level based on the number of transit vehicles per mile per direction during the morning peak travel time based on data obtained from the Los Angeles Metropolitan Transit Authority (LAMTA), the largest public transportation provider in the study area. Several other nearresidence transit accessibility factors tested at the ¼ and ½ distance thresholds did not have an influence on transit use and were excluded in the final analysis. These included: number of transit stops by service type, total length of nearby transit lines, and potential employment accessible on nearby transit lines. 4.4 Analytical Approach Analysis of the data proceeded in two steps. First, a measurement model was estimated for the factors that comprised the theoretical constructs in the PIA model. In order to verify the convergent and discriminant validity of our measurement model, we conducted a confirmatory factor analysis (CFA). Factor scores obtained from the CFA were then used as independent variables in regression analyses on transit use. In order to examine the effect of attitudinal, socio-demographic, and built environment variables on public transportation use, we developed two regression models, each with different 15

17 dependent variables derived from reported transit use by respondents during the study period. The dependent variables in each model are as follows: Model 1 Transit use during the 7-day diary period (1/0) Model 2 Mean daily transit trips during the study period For each dependent variable, three regression model iterations were tested. The first iteration included only socio-demographic variables as predictors. The second iteration added land use and transit system attributes to the base model, as is typical in travel behavior studies (Handy, 2005). The final iteration added socio-psychological factors derived from the CFA. The relative magnitude and significance of variables were generally consistent across iterations. Therefore, for each dependent variable we report only the final model iteration which includes socio-demographic, land use and transit system, and socio-psychological factors. 5. Results 5.1 Measurement Model Confirmatory Factor Analysis First, a measurement model was estimated for the factors that comprised the theoretical constructs in the PIA model using confirmatory factor analysis (CFA). The 21 socio-psychological items pertaining to the constructs in the PIA theoretical framework were subjected to confirmatory factor analysis in order to evaluate the fit of our measurement model. Table 3 shows the number of items included for each of the seven transit-relevant socio-psychological measures along with descriptive statistics for the response items that make up each measure. To evaluate the internal consistency of the factors, Cronbach's alpha was calculated for each. All factors had a Cronbach s alpha 16

18 of 0.61 or higher, above the recommended minimum of 0.60 for exploratory research (Hair et al., 2006). This agrees with similar work done by Anable (2005) and Hunecke et al. (2008) which retained factors with α > 0.65 and α > 0.58 respectively. Coefficient estimates and fit statistics for the CFA model were calculated using MPLUS Version 5 (Muthen and Muthen, 2012). The fit statistics for the model (Χ 2 = df = 50 (p < ), Χ 2 /df = 2.40, CFI = 0.94, RMSEA = 0.068) indicate that the measurement model has acceptable fit. Factor scores derived from the CFA were used as predictors in regression models of transit use. Table 4 shows the items included in each measure and standardized factor loadings for each item. All items in the CFA had standardized factor loadings of 0.35 or greater, and all were significant at p< Regression Models of Public Transportation Use We developed two models to examine the influence and significance of socio-demographic, built environment, and socio-psychological factors on transit use and frequency of use (Table 5). The first is a binary logistic regression which examines factors associated with the probability that a respondent used transit for at least one trip during the 7-day study period. The second model is a tobit model which examines factors associated with the mean number of daily transit trips taken by the main respondent. Both models include the same set of socio-demographic, built environment, and sociopsychological independent variables. One socio-psychological factor, privacy and crowding concerns, was dropped from the analyses due to multicollinearity problems, particularly with transit attitudes and personal safety concerns. Diagnostic testing of models including this factor showed variance inflation factors (VIF) in excess of the cutoff of 5 to 10 recommended by Stine (1995). This was not completely unexpected since the 17

19 factors in a CFA are allowed to co-vary, and the constructs relating to transit attitudes, personal safety, and privacy concerns are distinct but also interrelated. Removal of the privacy and crowding factor resulted in VIF for all variables of less than 3. Results indicate that a fairly consistent set of socio-demographic, built environment, and sociopsychological independent variables were associated with the two different specifications of transit use (Table 5). Among socio-demographic predictors, respondents who were older, had more household vehicles, and had children under 16 years of age had lower levels of transit use. Households in our middle income range ($35,000 to $75,000 per year) were less likely to use transit, and African- Americans were more likely to use transit. Models examined the influence of built environment and transit measures that were designed to capture street design, traffic levels, commercial activity, and transit service level at the neighborhood level. The number of intersections within ½ mile of the respondent s home, which captures nearby street connectivity, was significant in all models and suggests that more walkable neighborhoods with more pedestrian street crossings and smaller block sizes may be associated with more transit ridership. Shorter block lengths have been found in previous studies to be associated with more walking trips (Lee and Moudon, 2006; Ewing and Cervero 2001, 2010), as they allow more direct access to destinations. Street layouts with short block lengths may also promote public transportation use by reducing walking distance to transit stops. Interestingly, transit service level was negatively associated with transit use across all three models. This factor represents the number of transit vehicles per mile per direction which pass within ¼ mile of a resident s home during the morning peak travel time. Its negative influence on transit use is counterintuitive, but this result may be driven by the fact that stops with higher service level tend to 18

20 be located in areas with a higher number of travel destinations, such as shopping and employment centers. Households living close to these locations may therefore be able to complete some of their activities in these areas, either through trip chaining on journeys to and from home, or by making nonvehicle trips. Further study is needed to understand how respondents balanced trips by all travel modes in the areas with high transit service, and whether respondents living nearby took fewer transit trips because they have local access to services and employment. Two socio-psychological constructs derived from the CFA were consistently significant in both models after controlling for socio-demographic and built environmental variables. As expected, individuals with a more positive evaluation of transit were more likely to use transit, and those who expressed higher levels of concern about their personal safety on transit were less likely to use public transportation. Interestingly, transit social norms, transit perceived behavioral control, personal environmental norms, and perceived neighborhood amenities were not significant factors in either model of transit use. Likelihood ratio tests of the full models against the models with only sociodemographic and built environmental measures indicate model fit was significantly improved by the inclusion of transit-relevant socio-psychological constructs (p < for both models). In order to further examine the relationship between demographic characteristics and the sociopsychological factors, we regressed each of the six attitudinal factors on the demographic variables used in the full model (Table 6). Although previous research has proposed that attitudes (expressed through lifestyle desires) are determined by socio-demographic characteristics (Havens, 1981; Salomon and Ben-Akiva, 1983), the six models explain relatively little of the variation in attitudes, perceptions and norms in our sample. The R-squared value for transit attitudes is highest at 0.28, and sociodemographics explain 17 percent of the variation in transit perceived behavioral control. For each of 19

21 the other socio-psychological factors, the total variance explained is only about 10 percent. Looking more closely at the two factors that are significant in the models of transit use, being female is significantly associated with perceived personal safety threat on public transportation. For transit attitude, the highest education levels and greater car access are associated with negative attitudes toward transit. African-Americans in our sample hold significantly more positive attitudes toward transit compared to White respondents, though no other racial or ethnic group differed significantly. 6. Discussion and conclusion Our results show the value of including socio-psychological factors in studies of transit use. We found that attitudes toward transit and concerns about personal safety have a significant and consistent effect on the decision to use public transportation in our south Los Angeles study neighborhoods. These cognitive factors were significant even after controlling for demographic and built environmental variables included in typical land use travel behavior studies and their inclusion improved the overall performance of the models. Results generally agree with those of Hunecke et al. (2008), who found that age, number of household vehicles, employment status, perceptions of public transport scheduling control, attitudes toward various modes of transportation, and privacy concerns were all significant predictors of transit use among residents of three metropolitan areas in Germany. In addition, our findings on the insignificance of personal environmental norms in transit use add to evidence that the impact of environmental concern may depend on context. For instance, Elias and Shiftan (2012) found that environmental concern and problem awareness had no impact on transit use among a sample of predominantly low socio-economic status individuals, whereas other studies (i.e. Klöckner and Matthies, 2004; Hunecke et al. 2008) with a more diverse sample found 20

22 environmental norms to be significant. In our study personal desire to protect the environment was almost universally rated as important (mean = 6.22, SD = 1.04 on a scale of 1-7). Again, this was similar to data obtained by Elias and Shiftan (2012) (mean = 3.9 on a scale of 1-5). In examining the relationship between environmental concern and the use of transit, the specificity of our questions to the behavior under consideration (transit use) may have a considerable impact on explanatory power (Bohte et al., 2009). Our questions about environmental norms were phrased in a general way and were not directly linked to transit, but we may have found that environmental norms played an important role if we had asked respondents specifically about their willingness to endure inconveniences of transit use for the sake of environmental benefit. The fact that only one built environment factor (total near-residence intersections) and only one transit factor (total near-residence transit service level) were significant in any of the models was somewhat unexpected. Findings of previous research, particularly the meta-analysis of Ewing and Cervero (2010), indicate that distance to transit, street network characteristics, and land use diversity significantly impact transit use. Also, Hunecke et al. 2008, using a similar model to the one used in our analysis found that city center and suburban residents used transit significantly less than those who lived in districts bordering the city center. Unlike most land-use travel behavior studies, which typically focus on metropolitan areas or regions, our study area was relatively small (about 12 square miles) compared to the Los Angeles metropolitan area. For this reason, the built environment and transit factors in our study area tended to be more homogeneous in terms of street layout and neighborhood design. The area largely has a typical urban grid street pattern with similar block sizes throughout (Figure 2). The same is generally true of traffic levels, as the pattern of collector and arterial streets is relatively uniform throughout the study area. In addition, the study area is transit-rich. More 21

23 than 95 percent of study households lived within ¼ mile of at least one transit stop, with a mean distance to the nearest stop of 604 feet (184 meters). This relative lack of variation (as compared to a metropolitan region) may help to explain why built environmental factors appear to play a smaller role in our models of transit use. The current study has several additional limitations. First, although our study adds important insights to the literature regarding travel patterns and preferences for residents of a largely low-income and non-white community, the study s small sample size and geographic and socio-demographic scope limit the generalizability of findings. Though gender was not a significant predictor of travel behavior in our models, nearly three-quarters of respondents in our study were female which could also limit the generalizability to other populations. Second, although we provide a rigorous assessment of the role of several socio-psychological and neighborhood perception factors which likely influence residential location choices, it was beyond the scope of the analysis to fully investigate the role that residential self-selection may play in the observed patterns - particularly whether the study households moved to the study area to suit their travel and activity preferences (Cao, Mokhtarian and Handy, 2009b). Third, our study design allowed us to evaluate the influence of the near-residence built environment on travel, but we did not collect information about the built environment near other locations that might influence transit use, such as the workplace. The travel data are from trip logs with counts by mode only. We did not query respondents about destinations and hence could not assess the role of land use characteristics at workplaces or other destinations. Our future research will examine how the PIA model can be applied to help understand the role of cognitive factors in travel behavior change. Data for the current study were collected before the Expo Line opened, and travel outcomes, socio-demographics, and attitudes have been collected for the 22

24 same households after the opening of the line. These data include not only transit use, but also trips taken by all modes, time spent walking and bicycling, and vehicle miles traveled. We will compare changes of households in the experimental neighborhoods adjacent to the new line with those of comparable control neighborhoods located at varying distances of up to two miles away (Figure 2). This quasi-experimental study design will allow us to evaluate whether the new line had significant impacts on travel mode choice and travel-related physical activity. In addition, the panel data obtained in the before- and after-opening time periods will be used to evaluate aspects of the PIA model that are not easily assessed in a cross-sectional setting, and are therefore not covered in the current study. For example, the results the models presented here indicate correlations between attitudes and travel behavior, but the causal direction of this relationship is unclear. Panel data obtained before and after the opening of the Expo Line will allow the use structural equation modeling or other techniques to determine if attitudes do indeed have a causal relationship with transit use or if, for example, transit use causes more positive attitudes toward transit. Similarly, we will use longitudinal data to provide insights into the role of habit in transit use through analysis of the effect of past behavior on current mode choice. Overall, our results indicate that attitudes and perceptions of the built environment and transit system attributes appear to play a significant role in transit use that is independent of objectively measured attributes such as level of service, employment accessibility, or security. This finding is consistent with previous research that indicates that attitudes tend to matter in travel behavior regardless of whether built environment measures are significant. For example, Kitamura et al. (1997) found that the addition of attitudinal variables decreased the explanatory power of built environmental variables in explaining mode choice. Similarly, Handy, et al. (2005) determined that once attitudes and 23

25 demographics were accounted for, the role of the built environment mostly disappeared. Næss (2005) found that attitudes, demographics, and neighborhood characteristics all had an impact on travel behavior. Bagley and Mokhtarian (2002) demonstrated that among urban residents, attitude was a significant predictor of the decision to commute by car. Adequately accounting for the role of psychological factors in travel behavior could have significant implications for planners and policy makers. Clearly identifying specific attitudes, norms, and feelings of control that facilitate or hinder transit use allows the use of targeted information campaigns to address barriers to change. For example, changing perceptions about the speed, convenience, and incidence of crime on transit may be at least as important in increasing ridership as changes in transit service level. Further research is needed to assess whether tailored, individualized information on transit versus car travel times and destination accessibility helps to change perceptions about the convenience of car use. Reporting of transit-related crime information in conjunction with efforts to improve security on transit may help to align personal safety perception and reality for potential new users. Such targeted efforts could potentially help increase ridership in a relatively costeffective manner and help maximize the impact of built environment and transit service enhancements. Acknowledgements We thank the following organizations for their support of this research: Haynes Foundation, Lincoln Institute of Land Policy, San Jose State Mineta Transportation Institute, Southern California Association of Governments, University of California Transportation Center, University of California Multi-Campus Research Program on Sustainable Transportation, and the University of Southern California Lusk Center for Real Estate. We also thank all of the research assistants who were integral to 24

26 data collection and processing, particularly Dongwoo Yang, Wei Li, Gavin Ferguson, Hsin-Ping Hsu, Sandip Chakrabarti, and Carolina Sarmiento. 25

27 References Ajzen, I The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50, Anable, J `Complacent Car Addicts or `Aspiring Environmentalists? Identifying travel behaviour segments using attitude theory, Transport Policy 12, Armitage, C. J., and M. Conner Efficacy of the theory of planned behavior: A meta-analytic review. British Journal of Social Psychology, 40, Bagley, M.N. and P.L. Mokhtarian The impact of residential neighborhood type on travel behavior: a structural equations modeling approach. The Annals of Regional Science 36, Bamberg, S The Promotion of New Behavior by Forming an Implementation Intention: Results of a Field Experiment in the Domain of Travel Mode Choice. Journal of Applied Social Psychology, 30, Bamberg, S. and P. Schmidt Theory-driven evaluation of an intervention to reduce the private car-use. Journal of Applied Social Psychology, 31 (6), Bamberg, S., I. Ajzen, and P. Schmidt Choice of travel mode in the theory of planned behavior: The roles of past behavior, habit, and reasoned action. Basic and Applied Social Psychology, 25, Bamberg, S., S. Fujii, M. Friman and T. Gärling Behaviour theory and soft transport policy measures. Transport Policy, 18(1), Blainey, S., A. Hickford, and J. Preston Barriers to Passenger Rail Use: A Review of the Evidence. Transport Reviews 32 (6) (November): doi: / Bhat, C.R., and J.Y. Guo A Comprehensive Analysis of Built Environment Characteristics on Household Residential Choice and Auto Ownership Levels. Transportation Research Part B, 41(5), Bohte, W., K. Maat and B. van Wee Measuring Attitudes in Research on Residential Self Selection and Travel Behaviour: A Review of Theories and Empirical Research, Transport Reviews: A Transnational Transdisciplinary Journal, 29:3, Cairns, S., L. Sloman, C. Newson, J. Anable, A. Kirkbride, and P. Goodwin Smarter Choices: Assessing the Potential to Achieve Traffic Reduction Using Soft Measures. Transport Reviews 28 (5) (September): doi: / Cao, X.Y., P.L. Mokhtarian, and S.L. Handy. 2009a. The relationship between the built environment and nonwork travel: A case study of Northern California. Transportation Research Part A, 43 (5) (June): Cao, Xinyu, Patricia Mokhtarian, and Susan L. Handy. 2009b. Examining the impacts of residential self-selection on travel behavior: A focus on empirical findings. Transport Reviews 29 (3), Caplan, R. D., and R. Harrison Person-Environment Fit Theory: Some History, Recent Developments, and Future Directions. Journal of Social Issues 49 (4) (January): doi: /j tb01192.x. Cervero, R., and K. Kockelman Travel Demand and the 3Ds: Density, Diversity, and Design. Transportation Research D 2: Elias, W., and Y. Shiftan The Influence of Individual s Risk Perception and Attitudes on Travel 26

28 Behavior. Transportation Research Part A: Policy and Practice 46 (8) (October): doi: /j.tra Eriksson, L., and S. E. Forward Is the Intention to Travel in a Pro-environmental Manner and the Intention to Use the Car Determined by Different Factors? Transportation Research Part D: Transport and Environment 16 (5) (July): doi: /j.trd Ettema, D., T. Gärling, L. E. Olsson, and M. Friman Out-of-Home Activities, Daily Travel, and Subjective Well-Being. Transportation Research Part A: Policy and Practice 44 (9) (November): doi: /j.tra Ewing, R., and R. Cervero Travel and the Built Environment. Transportation Research Record, no. 1780: Ewing, R., and R. Cervero Travel and the Built Environment: A Meta-Analysis. Journal of the American Planning Association 76 (3): Ewing, R., Meakins, G., Bjarnson, G., and Hilton, H Transportation and Land Use. In Making Healthy Places (pp ). Island Press/Center for Resource Economics. Fishbein, M., and I. Ajzen Belief, Attitude, Intention and Behaviour: An Introduction to Theory and Research. Ed. Denise C Park and Linda L Liu. Reading MA AddisonWesley. Addison-Wesley. Fujii, S., S. Bamberg, M. Friman, and T. Gärling Are Effects of Travel Feedback Programs Correctly Assessed? Transportmetrica 5 (1) (January): doi: / Fujii, S. and T. Gärling, T Application of attitude theory for improved predictive accuracy of stated preference methods in travel demand analysis, Transportation Research Part A, 37, French, J. R. P., Jr., Rodgers, W. L., & Cobb, S Adjustment as person-environment fit. In G. Coelho, D. Hamburg, & J. Adams (eds.), Coping and adaptation. pp New York: Basic Books. Fried, M., J. Havens and M. Thall Travel Behavior A Synthesized Theory, Final Report, NCHRP, Transportation Research Board, Washington D.C. Gehlert, T., K. Dziekan, and T. Gärling Psychology of Sustainable Travel Behavior. Transportation Research Part A: Policy and Practice 48 (February): doi: /j.tra Gardner, B., and C. Abraham What Drives Car Use? A Grounded Theory Analysis of Commuters Reasons for Driving. Transportation Research Part F: Traffic Psychology and Behaviour 10 (3) (May): doi: /j.trf Hair, J. F., B. Black, B. Babin, R.E. Anderson and R.L. Tatham Multivariate Data Analysis (6th ed.). NJ: Pearson Education Inc. Handy, S. L Critical Assessment of the Literature on the Relationships among Transportation, Land Use and Physical Activity (Washington, DC: Transportation Research Board). Handy, S., X. Cao, and P. Mokhtarian Correlation or causality between the built environment and travel behavior? Evidence from northern California. Transportation Research Part D 10, Haustein, S., C. Klöckner, and A. Blöbaum Car use of young adults: the role of travel socialization. Transportation Research Part F: Traffic Psychology and Behaviour 12 (2), Haustein, S., and M. Hunecke Reduced use of environmentally friendly modes of transportation caused by perceived mobility necessities: an extension of the theory of planned behavior. Journal of Applied Social Psychology, 37,

29 Havens, J New Approaches to Understanding Travel Behavior: Role, Life-Style, and Adaptation. In P. Stopher, A. Meyburg, and W. Brög (eds.) New Horizons in Travel-Behavior Research. LexingtonBooks, Lexington, Massachusetts, Hauser, J. and G. Urban A Normative Methodology for Modeling Consumer Response to Innovation. Operations Research 25: Heath, Y. and R. Gifford Extending the Theory of Planned Behavior: Predicting the Use of Public Transportation. Journal of Applied Social Psychology 32,10: Hunecke, M., S. Haustein, S. Bohler, and S. Grischkat Attitude-Based Target Groups to Reduce the Ecological Impact of Daily Mobility Behavior. Environment and Behavior 42 (1) (October 10): Klöckner, C.A., and E. Matthies How habits interfere with norm-directed behaviour: a normative decision-making model for travel mode choice. Journal of Environmental Psychology 24 (3), Kitamura, R., P.L. Mokhtarian, and L. Laidet A micro-analysis of land use and travel in five neighborhoods in the San Francisco bay area. Transportation 24, Lee, C., and A.V. Moudon The 3Ds+ R: Quantifying land use and urban form correlates of walking. Transportation Research Part D: Transport and Environment, 11(3), Moser, G., and S. Bamberg The Effectiveness of Soft Transport Policy Measures: A Critical Assessment and Meta-analysis of Empirical Evidence. Journal of Environmental Psychology 28 (1) (March): doi: /j.jenvp Muthén, L. K., and B.O. Muthén Mplus User's Guide. Fifth Edition. Los Angeles, CA: Muthén and Muthén. Næss, P Residential location affects travel behaviour: but how and why? The case of the Copenhagen metropolitan area, Progress in Planning, 63(1), pp Nordlund, A., and K. Westin Influence of Values, Beliefs, and Age on Intention to Travel by a New Railway Line Under Construction in Northern Sweden. Transportation Research Part A: Policy and Practice 48 (February): doi: /j.tra Poortinga, W., L. Steg, and C. Vlek Values, Environmental Concern, and Environmental Behavior: A Study into Household Energy Use. Environment & Behavior. 36 (1): Prochaska, J. O., and W.F. Velicer The transtheoretical model of health behavior change. American Journal of Health Promotion, 12, Richter, J., M. Friman, and T. Gärling Soft Transport Policy Measures: Gaps in Knowledge. International Journal of Sustainable Transportation 5 (4) (March 22): doi: / Salomon, I., & Ben-Akiva, M. E The use of the life-style concept in travel demand models. Environment and Planning A, 15, Schwartz, S. H Normative influences on altruism. In L. Berkowitz (Ed.), Advances in Experimental Social Psychology (Vol. 10, pp ). San Diego: Academic Press. Schwartz, S.H. and J.A. Howard A normative decision-making model of altruism, in: J. P. Rushton and R. M. Sorrentino (Eds.) Altruism and Helping Behaviour: Social, Personality, and Developmental Perspectives, pp (Hillsdale, NJ: Lawrence Erlbaum). Stine, RA Graphical Interpretation of Variance Inflation Factors. The American Statistician 49 (1):

30 Thøgersen, J Understanding Repetitive Travel Mode Choices in a Stable Context: A Panel Study Approach. Transportation Research Part A: Policy and Practice 40 (8) (October): doi: /j.tra United States Census Bureau State and County QuickFacts Los Angeles (city), California. Accessed 10 September 2013 at Van Acker, V., P. Mokhtarian, and F. Witlox Going Soft: on How Subjective Variables Explain Modal Choices for Leisure Travel. European Journal of Transport and Infrastructure Research 11 (2): Van Acker, V, B Van Wee, and F Witlox When Transport Geography Meets Social Psychology: Toward a Conceptual Model of Travel Behaviour. Transport Reviews 30 (2) (March): Van Wee, B., H. Holwerda, and R. Van Baren Preferences for modes, residential location and travel behaviour: The relevance for land-use impacts on mobility. European Journal of Transport and Infrastructure Research, 2, Verplanken, B., Aarts, H., van Knippenberg, A., and van Knippenberg, C Attitude versus general habit: Antecedents of travel mode choice. Journal of Applied Social Psychology, 24, Wang, T., and C. Chen Attitudes, Mode Switching Behavior, and the Built Environment: A Longitudinal Study in the Puget Sound Region. Transportation Research Part A: Policy and Practice 46 (10) (December): doi: /j.tra

31 Fig. 1. The Perception-Intention-Adaptation (PIA) Theoretical Framework 30

32 Fig. 2. Study Area Map, South Los Angeles, California 31