Intercity Passenger Rail in the Context of Dynamic Travel Markets (2016)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

46 In Section 4.1, the development of the NCRRP Attitudinal Model for Rail (Figure 34) is docu- mented, as it was developed to help understand the inter-relationship of factors influencing the propensity to take rail in the project sample. In Chapter 7, the Attitudinal Model will be applied to rural areas and individuals’ propensity to take either rail or bus to a major destination city. In Chapter 8, the Attitudinal Model will be applied to the propensity to take the intercity bus. This model has been created by the research team to focus primarily on key explanatory factors which could (in theory) be influenced by the policies and interventions of rail managers and those involved in forming public policy toward investing in rail. Section 4.2 introduces a new model developed by the research team based on the TPB. The new model was developed under the direction and advice of Dr. Icek Ajzen, professor at the University of Massachusetts, and the author of the TPB. Ajzen also had a major role in the creation of the C H A P T E R 4 Understanding Values, Preferences, and Attitudes in the Choice of Rail Figure 34. NCRRP Attitudinal Model for Rail was created using structural equation modeling.

Understanding Values, Preferences, and Attitudes in the Choice of Rail 47 survey instrument itself, and survey data used throughout this project reflect the survey questions highly influenced by the structure of the TPB. Use of the theory helps introduce the concept of normative pressures revealed in the decision to choose rail, as well that of hedonic influences, which are reflections of perceived pleasure concerning the trip. Finally, the reader of Chapter 4 is strongly encouraged to refer to Technical Appendix: Docu- mentation for the Structural Equation Models in NCRRP Web-Only Document 2. In that appendix, the full documentation of the three attitudinal models (for rail, for intercity bus, and for rural use of rail and bus) is presented. In addition, documentation for the TPB model is presented. Models are first presented in diagram form, with presentations of both unstandardized and standardized coefficients for all parameters in each model, with levels of statistical significance for each. Several measures of goodness of fit are presented for each model. The TPB model is documented for the full sample, millennials only, and older groups only. 4.1 The NCRRP Attitudinal Model for Rail 4.1.1 Development of the Attitudinal Model In addition to the demographic variables, the Attitudinal Model can be seen as having four basic elements. Consistent with the analytical framework established earlier in the project, Figure 35 shows longer-term values as having an effect on all three later elements. Most directly, perhaps, these underlying values might influence one’s choice of a residential location. The characteristics of location set the stage for experiences one has in transportation (e.g., a traveler living in a rural area will experience rail in a different way than one living in Manhattan). Those experiences will lead to shorter-term attitudes (e.g., “the bus from my town is full of people I do not like”). These attitudes will almost certainly influence the choice of long-distance mode, in conjunction with both direct and indirect influences from the other factors in the model. Each of the factors in the Attitudinal Model was created from a process called “factor analysis.” The research team used a two-phase process, where all of the relevant variables were first reviewed through exploratory factor analysis, largely based on principal component analysis, as provided in the SPSS software package. Exploratory factor analysis provided a quick suggestion of which variables should be clustered with similar variables in the formation of a list of candidates for inclusion in the creation of the final factors. Figure 35. Diagram of major elements of the analytical framework for attitudes.

48 Intercity Passenger Rail in the Context of Dynamic Travel Markets Four Factors Representing Long-Term Values Observed variables that were identified through exploratory factor analysis were then con- sidered for final acceptance by a process known as confirmatory factor analysis. For example, Figure 36 shows four latent factors (in ovals) linked to 11 observed variables (in rectangles). The strength of the bond between the latent factor and its component observed variables is reflected in its “factor loading” coefficient. In a mathematical equation describing the extent to which the variance in the observed variable I love the freedom and independence I get from owning one or more cars is explained by the latent factor, a standardized coefficient of 0.58 is calculated (located above the arrow). This factor loading is often called the “R” value. The R2 value is shown to the left of the rectangle, in this case 0.34. From this it is deduced that 34% of the variance in the observed variable I love the freedom. . . . is explained by the values auto orientation latent factor. An iterative process was undertaken to ensure that each observed variable was correctly assigned to its latent factor, by confirming that no higher loading factor would be created by connection with a different latent factor. Alternative candidate observed variables were tested to see which combination resulted in the highest quality level of fit in the overall model. The final set of observed variables were chosen based on their overall model fit, provided that the variables reflect the overall theories being tested (e.g., an observed variable representing the price of fish in China would not be accepted, even with a good model fit, if that variable did not reflect the concepts being examined). Values and Location The NCRRP Attitudinal Model for Rail has four latent factors representing four longer- term values held by the respondent. Similarly, a density factor was developed for the residential Figure 36. Four basic latent factors (ovals) and 11 observed variables (rectangles) in the NCRRP Attitudinal Model for Rail.

Understanding Values, Preferences, and Attitudes in the Choice of Rail 49 location of each respondent, taking the form of the logarithmic value of the actual density in persons per square mile. Short-Term Attitudes and the Outcome Factor Four short-term attitudes toward rail were created using the factor analysis process, as shown on the left side of Figure 37. Although it is preferred to have each latent factor associated with several observed variables, both the factors Rail Perceived Less Stressful and Rail Perceived As More Expensive were found to fit into the model best with only one observed factor. Concerning the factor Rail Perceived Unsafe, it is clear from testing alternative observed variables that this factor is about personal safety and disturbing behavior, not safety as related to accidents, etc. The factor Rail Perceived Inconvenient represents both the concern about schedules and the concern about access to and from the train. The latent factor Propensity to Take Intercity Rail is associated with four observed variables, three of which reflect future intent and one on past behavior (i.e., having taken rail on the last reported trip). The factor loadings for the outcome factor are all statistically satisfactory. 4.1.2 Running the Attitudinal Model for Rail The model was run using AMOS structural equation modeling (SEM) software (Version 22), which is part of the SPSS set of modeling software packages. The sample included 5,625 respon- dents, including those from both the NEC and the Cascade Corridor. Several calculations of the Figure 37. Four latent factors (ovals) for near-term attitudes, and outcome latent factor for propensity to take rail, with factor loadings (standardized coefficients).

50 Intercity Passenger Rail in the Context of Dynamic Travel Markets quality of model fit are presented; its root mean standard error of approximation (RMSEA) was 0.046, where values under 0.05 are desired; its comparative fit index was 0.912, where values of over 0.90 are desired; and its Tucker-Lewis index was 0.891, where values of over 0.90 are desired. All coefficients are statistically significant, at p < 0.05. In addition, a “measurement model” was created in which the covariances among all nine latent factors were calculated, with- out the structural regression component of the model, which resulted in similar measures of model fit. 4.1.3 Interpreting the Results of the Attitudinal Model for Rail Some of the highlights of the Attitudinal Model for Rail are presented first. As previewed in Chapter 1, Table 9 shows the relative importance of the nine factors and two demographic variables that were used in the creation of the full model (Figure 34). Table 9 is based on the application of the standardized total effect (STE), which is used through- out each chapter covering the three attitudinal models. The STE represents the sum of both the direct and indirect standardized effects of the independent factor upon the outcome factor. While the STE is usually expressed in the scale of a 100% increase in the independent factor, a more realistic interpretation can be stated in terms of a 10% increase in the independent factor. By way of example, • A 10% increase in the value of the factor Train Trip Inconvenient would be associated with a 7.3% decrease in the outcome factor, Propensity to Take Rail. • A 10% increase in the value of the factor Values Auto Orientation would be associated with a 3.4% decrease in the outcome factor, Propensity to Take Rail. • A 10% increase in the value of the factor Train Trip Less Stressful than Car would be associated with 3.3% increase in the outcome factor Propensity to Take Rail. The consistent use of the STE as an indicator of the importance of candidate explanatory factors in the explanation of rail allows Table 9 to show the rank order of the 11 variables used in the model (rank of “importance” is based on the absolute value of the STE). The first observation from this is the lack of statistical significance of Residential Density in the explanation of rail, and the parallel lack of meaningful importance of Values Urbanism Sociability as a factor. Rank Order (by Absolute Value of STE) Standardized Total Effect (STE) Factor* 1 0.73 Train Trip Inconvenient 2 0.34 Values Auto Orientaon 3 0.33 Train Trip Less Stressful than Car 4 0.29 Values Privacy in Travel 5 0.22 Train Trip Unsafe 6 0.19 Values ICT 7 0.15 Train Trip More Expensive than Car 8 0.09 Educaon 9 0.03 Employed 10 0.03 Values Urbanism/Sociability 11 Not significant Residenal Density * The four basic values are shown in italic bold; the four short term atudes are shown in roman; and demographics are shown in italic. Table 9. Rank order of importance of explanatory factors.

Understanding Values, Preferences, and Attitudes in the Choice of Rail 51 As shown in Table 9, the strongest explanatory factor in the explanation of rail choice is the idea that the rail service is inconvenient, which is influenced by finding the schedule frequency unacceptable. The reader will note that this factor is similar to the role of times and costs in more traditional demand modeling, and thus its high ranking is not unexpected. Placing a strong value on privacy in travel and having a pro-auto orientation also negatively affect the propensity to consider rail. The concept that the car trip is more stressful than the rail trip is also a key explanatory factor. 4.1.4 Exploring the Interactions Between/Among Factors Table 10 allows for the exploration of the STE of one factor on an outcome factor. For exam- ple, the bottom row “Rail Propensity” shows the same information as presented in Table 9. Using the same data as in the previous section • The column Less Stressful has an STE on the row Rail Propensity of 0.33. • The column Auto has an STE on the row Rail Propensity of -0.34. Effect of Demographics The higher one’s level of education, the less likely one is to feel unsafe in the rail trip (-0.23) and to value privacy in travel in general, but the more likely one is to take rail (0.09). A person with a higher level of employment is more likely to value staying connected with ICT than a person with a lesser level of employment. Effect of Longer-Term Values • The more one values privacy in travel, the more likely one feels unsafe in the rail trip under consideration (0.64) and the more likely one feels that the rail option is inconvenient (0.45); higher value of privacy is associated with less propensity to feel that the rail option is less stressful than the auto option (-0.26). Higher value for privacy in travel is negatively1 associated with propensity to take the train (-0.29). Impacted Factors Demographics Basic Longer Term Values Locaon Rail Trip Perceived as… Employed Educaon Privacy Auto ICT Urbanism Density Unsafe Inconvenient Expensive Less Stressful Privacy 0.00 0.16 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Auto 0.00 0.04 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 ICT 0.15 0.06 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Urbanism 0.00 0.09 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Density 0.00 0.03 0.00 0.39 0.00 0.16 0.00 0.00 0.00 0.00 0.00 Expensive 0.02 0.08 0.00 0.13 0.10 0.00 0.00 0.00 0.00 0.00 0.00 Unsafe 0.00 0.23 0.64 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Inconvenient 0.01 0.13 0.45 0.30 0.04 0.01 0.08 0.47 0.00 0.16 0.00 Less Stressful 0.02 0.09 0.26 0.16 0.16 0.09 0.03 0.14 0.30 0.03 0.00 Rail Propensity 0.03 0.09 0.29 0.34 0.19 0.03 N.S. 0.22 0.73 0.15 0.33 N.S. = not significant Table 10. Standardized total effect from explanatory factors (columns) on impacted factors (rows). 1 In this chapter, every attempt has been made to make the explanatory text follow the direction of the relationships reported in the tables. Unfortunately, the result is, very often, the use of double negatives in the text.

52 Intercity Passenger Rail in the Context of Dynamic Travel Markets • Higher levels of auto orientation are associated with lower levels of density at the location of residence (-0.39); a higher propensity to feel that the rail service is inconvenient (0.30); and lower belief that rail is less stressful than auto for the trip (-0.16). Higher levels of auto orientation are negatively associated with propensity to take the train (-0.34). • Higher levels of preference for urbanism are associated with higher residential density (0.16) and higher belief that the rail trip is less stressful than the auto trip (0.09). As noted previously, the level of belief in urbanism is only weakly associated with propensity to take the train (0.03). • The need for ICT is positively associated with belief that the train is less stressful than the car (0.16) and with the perception that the train is more expensive than the car (0.10). This desire for productive connection is associated positively with the propensity to take rail (0.19). Effect of Shorter-Term Attitudes • The perception that the train trip is unsafe is strongly associated with stating the train trip is inconvenient (0.47), and negatively associated with reporting that the train trip is less stressful than the car (-0.14). The belief that the trip is unsafe is negatively associated with propensity to take the train (-0.22). • The belief that the train trip is inconvenient in terms of schedules and access problems is the strongest factor in the explanation of propensity to take rail (–0.73). It is also negatively associ- ated with perception that a rail trip is less stressful than car trip (–0.30). • The more one perceives the train to be expensive, the more likely one is to report that the train is inconvenient (0.16), and the less likely one is to take the train (-0.15). • The more one concludes that the rail trip is less stressful, the higher one’s propensity to take rail (0.33). 4.1.5 Conclusion: Results of the Attitudinal Model for Rail As shown in Table 9, and in more detail in Table 10, the structural equation modeling showed the importance of factors in addition to convenience, which reflects the relative quality of the rail service compared with its competitors. The next two highest ranking factors in the explanation of rail choice both concern attitudes toward the automobile, reflecting the nature of the compe- tition in a society where automobile trips dominate for most travel purposes. Even in the project sample, which was specifically and purposefully designed to find travelers between well-served large metropolitan areas, about six times as many trips are reported for auto as are for rail. Thus, looking beyond the convenience issue, the next largest factors are about attitudes toward own- ing an automobile and attitudes about the level of stress associated with using an automobile. Potentially more troubling is that the next two explanatory factors both concern feelings of discomfort or fear regarding train travel; discomfort or fear not only about crime per se, but also about disruptive behavior. All of these factors need to receive policy attention. These concepts will be explored in Section 4.2, with a new TPB model, and later in the discussion of intercity bus in Chapter 8. 4.2 The TPB Model for Rail Building upon the results of the Attitudinal Model for Rail described in the first part of this chap- ter, the research team created a working application of the TPB for the intent to take intercity rail. 4.2.1 What is the Theory of Planned Behavior? The TPB is the most applied social psychology model in the field of transportation behavior. The theory posits that change in behavior occurs after the formulation of intent, which is formed

Understanding Values, Preferences, and Attitudes in the Choice of Rail 53 after the individual quickly reviews three categories of salient information. These three catego- ries are considered direct predictors of intent and are as follows: 1. Attitude toward the behavior (ATB) concerns the individual’s conclusions about the good- ness of the proposed behavior for his/her own self, often determined on both rational and emotional bases. 2. Perceived norm (Norm) concerns the conclusion drawn by the individual that (a) the proposed behavior will be approved by persons important in his/her own social network (injunctive norm) and (b) such behavior is undertaken by those important persons them- selves (descriptive norm). 3. Perceived behavioral control (Control) reflects the potential difficulty associated with car- rying out the behavior, including the conclusion that she/he has the power to undertake the behavior. The theory suggests that these three factors have to be taken into consideration to understand a person’s intent to engage in a behavior. The theory posits that, after intent is established, the subject will revisit perceived behavioral control to reconsider the extent to which the behavior can actually be accomplished, given possibly increased understanding of the difficulty in doing so, as shown in the TPB diagram presented as Figure 38. In many ways, the TPB diagram is consistent with the logic of the analytical framework developed for this NCRRP project. In this diagram, the three kinds of near-term beliefs parallel the use of near-term attitudes and beliefs in the analytical framework. The list of background factors on the left side of Figure 38 contain both demographics and longer-term values, personality, and more general attitudes referenced in the value–attitude–behavior hierarchy discussed earlier in Chapter 1. 4.2.2 Factors in the TPB Model for Rail Figure 39 presents a simplified conceptual diagram of the TPB Model for Rail in which longer- term background factors have an influence on near-term beliefs, which have an influence on the three direct predictors in the TPB. There are 11 latent factors, including an outcome factor, included in the TPB Model for Rail. The logic presented in the simplified diagram in Figure 39 is developed in further detail in Figure 40, which shows the interaction among factors as one- directional arrows. Source: Fishbein and Ajzen (2010, Figure 1.1: Schematic presentation of the reasoned action model). Figure 38. The Theory of Planned Behavior as presented by Fishbein and Ajzen in 2010.

54 Intercity Passenger Rail in the Context of Dynamic Travel Markets Figure 40. The TPB Model for Rail. Figure 39. Conceptual diagram for the TPB Model for Rail. As shown in Figure 41, each latent factor is depicted in an oval, and each is associated with a set of observed variables, of which there are 26, plus two demographic variables. Figure 41 shows that the three latent factors representing the direct predictors reflect seven observed variables (shown in rectangles); the outcome factor, Intention to Take Intercity Rail, is based on two observed variables. The latent factors representing the four basic values are the same as used in the Attitudinal Model for Rail, and were shown in Figure 36, earlier. The model is designed to be run for two age groups (millennials and older respondents) to deal with questions of the impact of age. Income was deleted from earlier models after it was found to not have a statistically significant impact on the intent to choose rail (the lack of this linear relationship tends to reflect the fact that rail use tends to be somewhat higher for individuals under age 35,

Understanding Values, Preferences, and Attitudes in the Choice of Rail 55 then somewhat lower from ages 35 to 45, then somewhat higher for those above age 45). Density was also deleted from the model to preserve its parsimony. Model Fit The model was run on the AMOS SEM program (Version 22), in which model estimation was accomplished with the maximum likelihood method. The model showed reasonable standards of performance. The model has a comparative fit index of 0.92, and a Tucker-Lewis index of 0.91; in both evaluative measures, a value of more than 0.90 is considered essential to be labeled a good fit. The model has a RMSEA of 0.045, where a value of under 0.05 is considered to be a good level of model fit. For the full model shown in Figure 40, all parameters were found to be statistically valid, and a lack of an arrow from any latent factor to any other latent factor signifies that the parameter had been tested, and rejected as not significant. Complete documentation of model performance, including the value of all coefficients for all attitudinal models in this report, is included in NCRRP Web-Only Document 2 as Technical Appendix: Documentation for the Structural Equation Models. 4.2.3 Results of the TPB Model Table 11 shows how the factors affect each other, expressed as the STE on the impacted (row) factor. Consistent with the results of the Attitudinal Model in Table 10, Table 11 shows that the deci- sion to choose rail is influenced by the perception of inconvenience of a rail trip. In addition, the TPB model reveals the following influences: • The perception that a car trip is more stressful and less pleasant is powerful in explaining rail choice (ATB). Figure 41. Relationship between observed variables (rectangles) and latent factors (ovals) in the prediction of Intention to Take Rail.

56 Intercity Passenger Rail in the Context of Dynamic Travel Markets • The belief that friends and peers take rail is a strong predictor of rail choice (Norm). • The value placed on privacy is a strong negative factor (Privacy). – This is associated with concern for personal safety on the rail trip (Unsafe). • The value placed on auto ownership has a strong, negative relationship with rail choice (Car). • The desire to have connected technology is positively associated with rail choice (ICT). • Neither values about urbanism nor demographics have much explanatory power. 4.2.4 Interpretation of Results, Total Sample The application of the TPB to the question of rail choice for intercity trips provides a broad- ened base for understanding how mode choice decisions are made. First, Table 11 shows that all three direct predictors (right three columns) influence intention to take rail. In the model, per- ceived behavioral control primarily reflects the perceived inconvenience of the train trip; stated differently, the column factor Inconvenient has a very high STE (-0.80) on Control. Beyond the well-established importance of convenient service, the TPB model is examined for the potential influence of perceived norm and ATB. Understanding Normative Influences in Mode Choice Among the patterns that were revealed in the development of the model is that the sample as a whole, and the millennial group in particular, seems to be influenced more by descriptive rather than injunctive norms. Phrased more simply, this means that the decision to choose rail is more influenced by the observation that one’s valued peers and equals are also using the train (or are not using the train) than by a belief that those valued individuals are making an evaluative judgment and approve or disapprove of the respondent taking the train. The impact of normative pressure on intercity mode choice was not entirely anticipated at the time of the work plan development. In the literature of the TPB, it is often found that social pressure from peers is more important concerning emotional issues than for utilitarian issues (e.g., “what kind of dress do I wear to the dance” would be more susceptible to peer influence than “should I rotate my tires this week”). The results suggest that the choice of an intercity travel mode is somewhat less “utilitarian” in nature than perhaps expected and more subject to feelings of identification with one’s peer group. Further research could be undertaken, for example, about the role of peer influences in the use of curbside buses, many of which serve college campus areas. Finally, it should be re-iterated that the dimension of peer influence as a factor in the choice of mode applies in both directions. On the one hand, the feeling that “all the other students in my college are choosing the curbside bus” represents the positive context of application. On Table 11. Standardized total effect of each explanatory factor (column) on each impacted factor (row).

Understanding Values, Preferences, and Attitudes in the Choice of Rail 57 the other hand, the feeling that “no one in my country club would be caught dead on the bus” is simply a response at the opposite end of the same spectrum. The TPB model makes known the importance of normative influences in the selection of mode; at this point, the research has not clarified to which end of the spectrum is most powerful in the explanation of the positive or negative influence. Understanding Hedonic Influence in Mode Choice While all three direct predictors in the TPB model were found to be relevant to the choice of mode, ATB was found to have the strongest STE value of the three. As shown in Figure 41, ATB was formed from only two observed variables: the conclusion that a rail trip would be “pleasant” and less stressful than an auto trip. These individual observed variables describe the basic condition of personal pleasure on the part of the survey participant. The idea that basic decision such as choice of mode are influenced by hedonic considerations has been a major theme in the social psychology of transportation; see for example Figures 3 and 8 in Chapters 1 and 2, respectively. The argument that seemingly utilitarian decisions are in fact influenced by a variety of motiva- tions was explored in the landmark article “Car Use: Lust and Must. Instrumental, Symbolic and Affective Motives for Car Use” (Steg 2005). A more recent article by Steg et al. (2014) provides additional exploration of these kinds of motivations. The authors conclude that “interventions aimed to promote pro-environmental actions should consider hedonic consequences of actions, as these may be important barriers for behavior change.” In the TPB model, ATB (which is hedonic in nature) has an STE second only to that for the inconvenience of the rail trip, in the list of 12 explanatory factors in Table 11. Taken together, this supports the argument that factors in addition to times and costs need to be explored to better understand the nature of transportation demand. 4.2.5 Comparing Millennials with Older Respondents Because the TPB model is designed so that it can be run for two age groups in a single com- bined model, the luxury of a high-level comparison between the sub-samples is possible. (Based on this same method, the model was run simultaneously for males vs females; in general, differ- ences were less distinct than those revealed in the age-based comparison.) Table 12 shows how the factors affect each other, expressed as the STE on the impacted (row) factor. Compared to the experience of the older group, the decision to choose rail by millennials may be based somewhat more on • Feeling that rail is less stressful and more pleasant (ATB) and • Feeling that their peers would take the train (descriptive norm), rather than feeling that their peers would approve of their taking the train (injunctive norm). (This observation was not derived directly from Table 12, but from the model testing process.) Table 12. Standardized total effect on intent to choose rail for two age groups.

58 Intercity Passenger Rail in the Context of Dynamic Travel Markets Compared to the experience of the millennials, decisions not to choose rail by the older group may be based somewhat more on • Perception that rail is inconvenient, • Concern for personal safety on the trip (Unsafe), • Perception that rail is more expensive, and • Peer pressure about the mode choice (Norm). 4.2.6 Conclusion: The Theory of Planned Behavior For the purposes of NCRRP 03-02, the TPB should be regarded as background information for further analyses of possible psychological factors relevant to the choice of intercity mode, and to transportation behavior more generally. Using the additional tools provided by the method, it becomes apparent that further research could investigate the idea (1) that individuals choose what they conclude to be, all in all, the more pleasurable option and (2) that the choice of the optimal mode might be very much influenced by what individuals see as a desired behavior by members of their social network. These two concepts merit additional exploration above and beyond the scope of this project. Social psychologists, and other social scientists, are committed to exploring basic concepts such as quality of life and the related concept of “subjective well-being,” which describe what people generally refer to as happiness. Others are exploring the concept that descriptive norms are a force which impacts much of our daily lives. Both are worth exploring because of their potential relationship to travel behavior.