Technical Appendix to Air Demand in a Dynamic Competitive Context with the Automobile (2019)

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.

89 CHAPTER 9. AN ATTTITUDE-BASED MODEL OF AIR VS. AUTO 9(A) INTRODUCTION Our work plan called for the creation of three separate mathematical models to aid in the understanding of the choice of mode between the auto and the airplane in the long-distance trip. The first model is designed to emphasize the importance of “soft” variables, including values, preferences and attitudes in the selection of modes for the long-distance trip; by design, this model does not emphasize the trip-based times and costs. The second model is designed to emphasize the immediately relevant factors of travel time and costs, in addition to other traditional variables such as travel party size: by design, this model does not emphasize the “softer” factors including values, preferences and attitudes. The third model is designed to integrate all relevant factors into the prediction of long-distance travel mode. 9(B) STRUCTURE Technical Appendix Chapter 9 presents the new Structural Equations Model created in this project. Appendix Chapter 10 presents the new Multinomial Logit Model. Appendix Chapter 11 then presents the Hybrid Choice model prepared to be integrated in the larger nationwide model of long-distance trip-making in the United States, and to support the scenario testing process mandated in the Amplified Work Program. 9(C) THE STRUCTURAL EQUATIONS MODEL INCORPORATING VALUES, PREFERENCES, AND ATTITUDES DEFINING THE MODEL The Structural Equations Model of Long-Distance Mode Choice has four basic elements. It is hypothesized that some values which impact transportation behavior are long-term in nature and may influence the location of the traveler; they may influence the shorter-term attitudes and may directly influence the choice of long-distance mode. For example, some basic attitudes toward the ownership of a car might influence one’s location decision, might influence short-term attitudes toward the specific trip, and influence the choice or rejection of the auto for the trip. Location is the second element of the SEM model. Higher levels of ‘urbanity’ of a location often imply that a major airport is close by; the ‘ruralness’ of a location is correlated with longer driving distance to a sizeable airport. In the third element of the model, shorter-term attitudes are formed at a time closer to making the modal decision. For a given trip, the idea of finding satisfactory en-route lodging for the multiday auto trip might be unpleasant for some, while others might not be bothered by it. Some people might feel that ‘people-like-them’ would always choose the air, while others feel the opposite. The fourth element of the model is the modal choice itself. Figure 9-1 presents a basic diagram of how the three explanatory elements are hypothesized to relate to the outcome factor, “Propensity to Choose Car Trip.”

90 FIGURE 9-1: CONCEPTUAL DIAGRAM OF STRUCTURAL EQUATIONS MODEL THE LATENT FACTORS DEVELOPED IN THE MODEL Using the application of factor analysis and the published literature in this area, four latent factors were developed to represent preferences on issues with a longer-term time frame than those concerning the evaluation of a transit trip and its attributes, shown in Figure 9-2. The three categories here have been developed and applied in other CRP studies including NCRRP Report 4: Intercity Passenger Rail in the Context of Dynamic Travel Markets.11 11 RSG, M. Coogan, AECOM, I. Ajzen, C. Bhat, B. Lee, M. Ryerson, and J. Schwieterman. 2016. NCRRP Report 4: Intercity Passenger Rail in the Context of Dynamic Travel Markets. Transportation Research Board, Washington, DC

91 FIGURE 9-2: RELATIONSHIP OF OBSERVED VARIABLES (RECTANGLES) TO LATENT FACTORS (OVALS) Longer-Term Values Values Urbanism. A latent factor representing preferences for the attributes of ‘valuing urbanism’ in neighborhoods was created with the use of three questions about the importance of walking to a commercial district, being outside with people, and having a mix of people from different backgrounds. Auto Orientation. A latent auto factor was created representing hedonic considerations (e.g., love for the auto); an observed variable representing the desire to control one’s own space in the car; and the statement that he/she needs a car to get where he/she needs to go. Values Information Technology. Finally, the importance of being productive, and staying connected all day was explored in one latent factor based on the two related attitudes, and the level of ownership of devices. Location A latent factor was created with the use of two observed variables, one reflecting the density of intersections (“design”), one about the distance to the preferred airport.

92 FIGURE 9-3: SEVEN SHORTER-TERM ATTITUDES ABOUT THE TRIP AND THE MODES Shorter-Term Attitudes about the Trip Seven latent factors were created in the model to reflect shorter-term attitudes and trip conditions. A total of 19 variables were used in development of the seven latent factors. Cost. A latent factor dealing with the perceived additional cost of the air trip was created with three observed variables. The first sought agreement with the estimate that this trip would be more expensive by plane. The second two were based on a) the actual airfare per mile, and 2) the airfare per hour of driving. Stress due to air travel. Reported attitudes were used for seats on planes being too close, dealing with crowds at the airport, and stress from the airport security procedures, to create the latent factor. Stress due to driving. A latent factor was created from two concerns about roadway congestion. Safety and disturbing behavior was reflected in a latent factor, based on three survey questions, each of which was scaled so that higher values represent higher levels of concern with air travel. In addition, three more factors were created, which were influenced by (but do not exactly operationalize) the Theory of Planned Behavior, a theory commonly used in the adaptation of principles of social psychology to the subject area of transportation. Affective attitudes toward trip. The first represents the concept of whether this behavior (e.g. choosing the car) is in the interest of the participant and is seen as a pleasant experience. These

93 attitudinal questions are similar to how “Attitude Toward the Behavior” in the Theory of Planned Behavior is usually operationalized. This Latent Factor was created from three observed variables, two of which concerned the unpleasantness of the drive, and one about the level of uncertainly associated with it. Social support. The next latent factor was created to help explore the influence of normative factors, that is, to what extent important others would support the participant’s modal decision, and whether they travel by that mode. In the Theory of Planned Behavior this factor is called Subjective Norm (and sometimes, Social Norm). Control – perceived inconvenience of the trip. The final latent factor was created to reflect the concept that choosing to travel by air might be difficult or even impossible, that is, something that one would not have the control to do. For this latent factor, three observed variables were used which reflect the perceived difficulty or impossibility of going by air, including “the airlines just do not go where I need to go.” Needing a car at the destination, and difficulty of getting to the airport are also included in the creation of the latent factor. In the Theory of Planned Behavior, the concept that one may not be able to carry out the behavior is called “Perceived Behavioral Control.” The Outcome Factor: Choice of Mode A latent factor was developed to reflect “The Propensity to Choose Car Trip” as shown in Figure 9-1, which was based on a combination of mode selection for the reference trip, for the short- distance trip, and for a summary of all trips mentioned by the survey participant. RUNNING THE MODEL The model was run as a SEM, using AMOS Version 22 software, which is part of the SPSS set of modeling software packages. The sample included 4,232 respondents. The final model shows the relationships between and among the 12 latent factors of the model, based on the application of the 33x observed variables portrayed as rectangles in the three figures above. The Long- Distance Mode Choice SEM which results has positive overall evaluative characteristics. The model has a Root Mean Square Error of Approximation of 0.04, where any value below 0.05 is considered a good fit. It has a comparative fit index and a Tucker Lewis Index of 0.95 and 0.93, where values above 0.90 are considered to reflect a good model fit. The model has 33 coefficients, whose unstandardized and standardized coefficients and p values are presented in Table 9-2. All the coefficients in the model are significant with p values of less than or equal to 0.01, except one, which is under 0.02. INTERPRETING THE RESULTS OF THE LONG-DISTANCE MODE CHOICE SEM The model is not primarily designed to predict behavior, but to contribute to understanding of the relationship between and among factors, given the relationships hypothesized in Figure 9-1. SEM models allow the ability to look at the combination of direct and indirect effects of one factor upon another, called the ‘Standardized Total Effect’ (STE). For example, to explain the meaning of the bottom left cell in Table 9-1, the AMOS software program states:

94 “The standardized total (direct and indirect) effect of Auto Orientation on Car Chosen is - .449. That is, due to both direct (unmediated) and indirect (mediated) effects of Auto Orientation on Car Chosen, when Auto Orientation goes up by 1 standard deviation, Car Chosen goes down by 0.449 standard deviations.” Table 9-1 shows the total effect of each column factor in each row factor, expressed as the STE. Focusing on the effects of various factors on the model’s key outcome, choice of mode, it becomes clear that the factor concerning the issue of the comparative “expense” of the two modes has the highest STE (at .72). The results of the bottom line in Table 9-1 are expressed graphically in Figure 9-4 in vertical bar chart format. TABLE 9-1: STANDARDIZED TOTAL EFFECT, IMPACT OF COLUMNS ON ROWS Auto  Orientation  Urbanism ICT Location Expensive Air Trip  Stressful  Car Trip  Stressful  Crime Attitude‐  Unpleasant  Social  Support  Inconven ient Location 0.37 ‐0.31 ‐0.09 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Air Trip Stressful 0.39 ‐0.02 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Trip Expensive 0.60 0.14 ‐0.21 0.10 0.00 0.21 0.00 0.00 0.00 0.00 0.00 Crime/Disturbing Behavior  0.42 0.03 ‐0.07 0.03 0.34 0.62 0.00 0.00 0.00 0.00 0.00 Car Trip Stressful  ‐0.06 0.06 0.40 ‐0.01 ‐0.09 ‐0.02 0.00 0.00 0.00 0.00 0.00 Social Support ‐0.14 0.18 0.26 ‐0.07 ‐0.70 ‐0.18 0.30 0.15 0.00 0.00 0.00 Attitude‐ Unpleasant ‐0.21 0.09 0.31 ‐0.02 ‐0.15 ‐0.10 0.73 0.00 0.00 0.00 0.00 Control ‐ Inconvenient 0.60 0.06 ‐0.20 0.14 0.66 0.43 0.00 0.30 0.00 0.00 0.00 Outocme ‐ Car Chosen  0.45 0.02 ‐0.26 0.08 0.72 0.22 ‐0.24 0.02 ‐0.26 ‐0.17 0.15

95 FIGURE 9-4: TOTAL EFFECT ON MODE CHOICE, RANK ORDERED BY THEIR ABSOLUTE VALUES Interpreting the SEM format The data in Table 9-1 can be interpreted in several ways. 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. For example, the table shows that:  A 10% increase in the value of the factor “Auto Orientation” would be associated with a 4.5% increase in the Outcome Factor, “Car Chosen.”  A 10% increase in the value of the factor “(Car) Attitude Unpleasant” would be associated with a 2.6% decrease in the Outcome Factor, “Car Chosen.”  A 10% increase in the value of the factor “Air Trip Stressful” would be associated with 2.2% increase in the Outcome Factor, “Car Chosen.” Comparing the role of the factors While the dominance of expense of the modes (expressed in the question as the expense of air) is totally expected, the role of other factors is noteworthy. The long-term values held by the traveler concerning his/her love of and need for cars emerges as the second strongest predictor of mode choice. Beyond the concern for costs, attitudes about the car seem to be more powerful explainers of the outcome than are attitudes about the air trip—with measurement of trips being unpleasant and stressful given higher ranking than equivalent questions about the air experience. In our interpretation, the SEM model shows that the dominant characteristic of air travel is its cost – attributes about stress at the airport, or poor coverage of destinations (“inconvenient”) simply did not emerge as explanatory factors the way that price does; an increase in air price is strongly associated with an increase in auto mode share. Figure 9-5’s horizontal bar chart format (same data as above) helps to reveal the positive and negative characteristics of the STE data. It 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 St an da rd ize d  To ta l E ffe ct  

96 shows that attitudes toward the auto are more nuanced: survey responses about one’s love for the freedom and independence from auto ownership emerge as strong factors in the explanation of variance in mode share. At the same time, statements that long trips are unpleasant are part of a factor with strong negative explanatory power. Attitudes toward the auto seems therefore to be ambivalent, simultaneously receiving both positive and negative evaluations. FIGURE 9-5: STANDARDIZED TOTAL EFFECT, EXPRESSED AS POSITIVE AND NEGATIVE VALUES ‐0.4 ‐0.2 0 0.2 0.4 0.6 0.8 Air Expensive Auto Orientation Car Unpleasant Likes ICT Car Trip Stressful Air Trip Stressful Social Support for Air Air Inconvenient Location more rural Crime on air trip Pro Urbanism Expressed as Standarized Total Effect

97 TABLE 9-2: UNSTANDARDIZED AND STANDARDIZED COEFFICIENTS IN THE SEM Standardized  Estimate S.E. P Estimate Air Trip Stressful <‐‐‐ Auto Orientation 0.493 0.035 *** 0.389 Location <‐‐‐ Auto Orientation 0.088 0.009 *** 0.364 Location <‐‐‐ Urbanism ‐0.079 0.009 *** ‐0.303 Location <‐‐‐ ICT ‐0.018 0.007 0.008 ‐0.096 Expensive <‐‐‐ Auto Orientation 0.586 0.052 *** 0.495 Expensive <‐‐‐ Air Trip Stressful 0.197 0.025 *** 0.211 Expensive <‐‐‐ Urbanism 0.219 0.038 *** 0.17 Expensive <‐‐‐ ICT ‐0.204 0.029 *** ‐0.219 Expensive <‐‐‐ Location 0.463 0.156 0.003 0.094 Car Stressul <‐‐‐ Expensive ‐0.11 0.028 *** ‐0.094 Car Stressul <‐‐‐ ICT 0.413 0.03 *** 0.378 Crime/Disturbing <‐‐‐ Air Trip Stressful 0.766 0.029 *** 0.549 Crime/Disturbing <‐‐‐ Expensive 0.504 0.034 *** 0.337 Car Stressul <‐‐‐ Urbanism 0.12 0.039 0.002 0.08 Inconvenient <‐‐‐ Expensive 0.686 0.044 *** 0.591 Inconvenient <‐‐‐ Air Trip Stressful 0.15 0.028 *** 0.138 Attitude ‐ Unpleasant <‐‐‐ Auto Orientation ‐0.182 0.052 *** ‐0.094 Attitude ‐ Unpleasant <‐‐‐ Air Trip Stressful ‐0.111 0.035 0.002 ‐0.073 Inconvenient <‐‐‐ Location 0.525 0.125 *** 0.092 Attitude ‐ Unpleasant <‐‐‐ Expensive ‐0.131 0.044 0.003 ‐0.08 Attitude ‐ Unpleasant <‐‐‐ Car Stressul 1.011 0.042 *** 0.725 Social Support <‐‐‐ Crime/Disturbing 0.15 0.033 *** 0.149 Social Support <‐‐‐ Expensive ‐1.103 0.072 *** ‐0.732 Social Support <‐‐‐ Auto Orientation 0.529 0.06 *** 0.297 Attitude ‐ Unpleasant <‐‐‐ Urbanism 0.117 0.042 0.005 0.056 Social Support <‐‐‐ Car Stressul 0.382 0.032 *** 0.296 Social Support <‐‐‐ Urbanism 0.477 0.044 *** 0.246 Social Support <‐‐‐ Air Trip Stressful ‐0.152 0.045 *** ‐0.108 Inconvenient <‐‐‐ Crime/Disturbing 0.224 0.025 *** 0.288 Car Chosen <‐‐‐ Expensive 0.214 0.027 *** 0.494 Car Chosen <‐‐‐ Social Support ‐0.049 0.007 *** ‐0.169 Car Chosen <‐‐‐ Attitude ‐ Unpleasan ‐0.069 0.005 *** ‐0.258 Car Chosen <‐‐‐ Inconvenient 0.047 0.019 0.015 0.125 Unstandardized