Understanding How Individuals Make Travel and Location Decisions: Implications for Public Transportation (2008)

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.

67 Introduction and Structure of the Chapter This chapter continues the exploration of market sectors that are more likely to be favorable to an urban residential environment, particularly a CN. It also explores the propen- sity for transit use and walking to increase with a change in neighborhood type. This project has created a new source of data that integrates information about personal attitudes and values with more traditional information about travel behavior and neighbor- hood form. The new data set makes possible the examination of the interrelationship between values held by the traveler and the characteristics of the built environment in the for- mation of travel behavior and modal choice. In the first part of this chapter, the relationship between travel behavior and two separate independent variables is examined. First, the relationships between personal values and travel behavior for walking and transit are explored. Then new data on the relationship between the built envi- ronment (in this case, neighborhood type) and travel behav- ior for walking and transit are presented. The second part of the chapter examines the interaction of two of the independent variables, noting their combined effect on the dependent variable of travel behavior. The combined effect of personal values and urban form is examined in terms of a variety of measures of transit and walking patterns. The third part of the chapter examines the revealed rela- tionship between auto availability and travel behavior for walking and transit. The chapter explores the interaction of the three variables on the propensity to walk or take transit. The document reviews what can and cannot be observed from examination of cross tabulations, which reveal the com- bined role of personal values, urban form, and auto availabil- ity on the propensity to walk and take transit. The fourth part of the chapter uses structural equation modeling (SEM) to investigate the relative importance of per- sonal values, urban design, and auto ownership. The fifth part summarizes observations about the role of each of the three variables and the need for further research. Personal Values and Travel Behavior; Urban Form, and Travel Behavior Overview of an Approach to Creating a Personal Values Factor This section examines the relationship between travel behavior concerning walking and transit and the independ- ent variable representing personal values. In a later section of this chapter, travel behavior will be examined in relationship to the interaction of personal values and urban form. Over the past decade, a substantial contribution has been made to the professional literature of travel behavior by those who have argued that travel times and travel costs must be examined in the context of the values, perceptions, and attitudes held by the traveler in making modal decisions (17, 18). This project’s Phase 1 survey was designed to contribute to this liter- ature in several ways. The new database is unique in its basis in a nationwide sample of transit-oriented metropolitan areas and on its use of the TPB in the survey design. In this chapter, the concept of “personal values” is reflected in the use of two groups within the total sample. In a process described below, two groups were defined in terms of their attitudes toward basic conditions of an urban, pedestrian- friendly, and environmentally friendly lifestyle. In this method, a combined factor was created from similarity of re- sponses to 15 key rating statements, as shown in Table 8-1. A combined factor of “urban and environmental values” was created by summing scores on the rating statements shown in Table 8-1. The group whose score was higher than the aver- age (mean) on this combined factor was labeled the high urban/environmental values group; the group with lower than average scores on the combined factor was labeled the low urban/environmental values group. C H A P T E R 8 Travel Behavior by Values, Urban Form, and Auto Ownership

Technical Explanation of the Approach A set of rating statements created for application of the TPB were examined for their role in the creation of a “personal values” factor. Two statements were examined first: “For me to live within walking distance to stores, restaurants, a public library, and a school would be (desirable/undesirable), and “Having a commercial district (with things like a coffee shop, retail stores, and restaurants) within walking distance of my home would be (not important at all/extremely important).” These statements represent the essence of a CN, where walk- ing is a reasonable option. The next step was to add additional statements to the set so that the set would more fully describe values associated with an urban, pedestrian-friendly, and environmentally friendly lifestyle. To create this set of statements, a statistical test known as Cronbach’s alpha was used, which is a way to de- termine if a set of variables is successfully measuring a single construct, albeit one containing different substantive con- cepts. Starting with the original two statements, additional statements that related to urban or environmental values were tested one at a time. When the addition of the candidate statement raised the level of the alpha of the set, that state- ment was accepted for inclusion in the set. With the final list of candidates, each statement was then manually deleted to see if its absence raised the level of the alpha; if so, it was deleted. This process resulted in a final list of 15 statements, reproduced here as Table 8-1. The combined factor resulted in a Cronbach’s alpha of 0.85, which is considerably above the level generally accepted as stable. The 15 statements selected by this process include four that reflect the SN. In short, this set of values represents an integration of personal and inter- personal attitudes. In the language of the TPB, it represents a combination of measures of attitude toward the behavior and SN. The 15 variables were integrated by summing the re- sponses to the 15 rating statements. The sample of respon- dents was divided into two groups, one with higher than average (mean) scoring on the combined factor, labeled as the high urban/environmental values group, and the second with scorings lower than the sample mean, labeled as the low urban/environmental values group. Of the responding sample (865), 467 respondents are categorized as being in the high urban/environmental values group, and 398 are cate- gorized has being in the low urban/environmental values group. Personal Values and Travel Behavior One’s personal values toward urbanity and the environ- ment seem to be strongly associated with the propensity to walk and take transit, as shown in Table 8-2. In this table, green mode is the sum of the transit and walk modes. 68 Rating Statements Having an adequate number of sidewalks in good condition. Having frequent bus or other transit (train or trolley) services. Having buses or other transit services serve areas in which I frequently needed to travel. Having a commercial district (with things like a coffee shop, retail stores, and restaurants) within walking distance of my home. Having access to reliable taxi service whenever I need it. For me, to live within walking distance to stores, restaurants, a public library and a school would be desirable. For me, to be able to take public transportation to work or for other trips would be desirable. For my household to need to own fewer cars would be desirable. I am concerned about global warming and/or climate change. Protecting the environment should be given top priority, even if it means an increase in taxes. I’d be willing to drive less to reduce my use of foreign oil. Friends and family think they should be more active in doing their part to protect the environment. Friends and family are concerned about global warming and/or climate change. Friends and family think that protecting the environment should be given top priority, even if it means an increase in taxes. Friends and family would be willing to drive less to reduce their use of foreign oil. Number of cases: 865, alpha = 0.85. Table 8-1. Fifteen rating statements for urban/environmental values.

Table 8-2 shows that while 15% of the trips by all purposes for the low urban/environmental values group were taken by green modes, some 33% of trips of those in the high urban/ environmental values group were taken by green modes. Thus, green mode trip making was over twice as prevalent for the high urban/environmental values group as for those in the low urban/environmental values group. The influence of per- sonal values on the work trip was much less pronounced than for the nonwork trip. However, all of the differences between the high urban/environmental values group and the low urban/environmental values group are significant at p < .05. Table 8-3 shows the relationship between personally held values and travel behavior relative to transit and walking, by trip purpose. Clearly, the effect of the urban/environmental values factor is visible for all trip purpose categories shown in Table 8-3. Again, as expected, the nonwork travelers with low urban/environmental values show the least propensity to use transit of any group with a 4% mode share. All of the differ- ences between groups in this table are significant at p < .05. Looking at walking, the variation between the holders of high urban/environmental values and low urban/environ- mental values is largely consistent with the pattern for green modes as a whole. In general, those who place a positive value on the urban conditions supportive of walking tend to actually walk at a rate roughly 2.5 to 3 times as large as those who place a negative value on those attributes, as shown in Table 8-4. All of the differences between groups are significant at p < .05. Urban Form and Travel Behavior To undertake the analyses in this chapter, it was necessary to create two groups that represent two levels of supportiveness 69 Values Group (15 Variables) Green Mode Share, All Trips (%) Green Mode Share, Work Trips* (%) Green Mode Share, Nonwork Trips (%) High Urban/ Environmental Values 33 50 29 Low Urban/ Environmental Values 15 30 11 Full Sample 24 41 21 High/low pair values significantly different at p <. 05; n ranges from 467 to 341. * Work trips mode share in this chapter is computed for workers only. It differs from the values in Chapter 6, which provide work trip mode share for the entire group of respondents Urban Values Group (15 Variables) Transit Share, All Trips (%) Transit Share, Work Trips (%) Transit Share, Nonwork Trips (%) High Urban/Environmental Values 17 41 12 Low Urban/Environmental Values 8 26 4 Full Sample 13 34 8 High/low pair values significantly different at p < .05; n ranges from 467 to 341. Table 8-2. Personal values and green mode shares. Table 8-3. Personal values and transit mode shares. Values Group (15 Variables) Walk Share, All Trips (%) Walk Share, Nonwork Trips (%) Monthly Utilitarian Walk Trips (No.) Monthly Nonwork Utilitarian Walk Trips (No.) High Urban/ Environmental Values 16 18 17 16 Low Urban/ Environmental Values 6 7 6 5 Full Sample 11 13 12 11 High/low pair values significantly different at p < .05; n ranges from 467 to 398. Table 8-4. Personal values and walking.

from the built environment, in relation to the propensity to walk and take transit. The Phase I survey offers a wide variety of questions in this area, ranging from neighborhood type to size of the town or city; it also offers the chance to examine sep- arately the question of whether your neighborhood has multi- unit housing, which is different from whether your personal residence is in multiunit housing. Several of these categoriza- tion methods to create the two groups were examined, and one of these (mix of housing) is presented later in this section. However, it became clear that the most comprehensive defini- tion to use was the project’s definition of a CN. As noted in the definitions presented in Chapter 1, a location is referred to as a “CN” when (a) there is some form of housing other than a single-family home within one-third mile, (b) there is a com- mercial district within one-third mile, and (c) the neighbor- hood has transit service. If any of the preconditions are lacking, the location is categorized as not in CN. The data set divides the 865 respondents into two groups: living in CN and not in CN. Of the total sample, some 222 reside in a CN and 643 do not. Table 8-5 shows the strong relationship between the type of neighborhood inhabited and the use of transit and walking, Logically, the use of walking and transit would be higher in a CN than in other locations. Table 8-5 shows the relationship between the built environment and the propensity to take trips by walking or transit, summarized in the green mode share. Table 8-5 also shows that while 18% of the trips by all purposes in areas other than CNs were taken by green modes, some 44% of trips in CNs were taken by green modes. Thus, green mode trip making was almost 2.5 times as prevalent in the CNs as in the non-CNs. The differentiation attributable to the conditions of the built environment is far more prevalent in the analysis of the nonwork trip than for the work trip. The differences between respondents in CNs and outside CNs are significant at p < .05. Transit can be examined separately from walking. From Table 8-6, it can be observed that total public transportation trip making was more than 2.5 times as prevalent in CNs as in non-CNs. Furthermore, it can be observed that the effect of neighborhood type on the propensity to take transit is far more pronounced for the nonwork trip than for the work trip. As ex- pected, this reflects the wider distribution of transit over trip purposes in the CN neighborhoods than in the rest of the re- gion: outside of the transit-rich areas, nonpeak hour, nonwork travel is far less important than in the CN. Turning our attention to the relationship between neigh- borhood form and the propensity to walk, Table 8-7 shows that one-fifth of all utilitarian trips in the CN are taken by walking, which is about two times the rate of areas outside of the CN. Interestingly, the walk mode share for the nonwork trip is only slightly higher than the walk mode share for all trip purposes. Looking at the metric of walk trips in terms of absolute numbers, rather than shares, respondents from the CNs reported taking 22 utilitarian walk trips per month, while those outside of the CNs reported taking nine utilitar- ian walk trips per month. As noted, the CN has three supportive components that come together to facilitate the choice of transit and walking. The CN has supportive attributes in terms of housing mix, walking destinations, and available transit. To understand the relative importance of various elements of the CN, travel behavior has been calculated as a function of variation in 70 Location Green Mode Share, All Trips (%) Green Mode Share, Work Trips (%) Green Mode Share, Nonwork Trips (%) Living in CN 44 56 41 Not in CN 18 35 14 Full Sample 24 41 21 CN/Not CN pair values significantly different at p < .05; n ranges from 643 to 197. Location Transit Share, All Trips (%) Transit Share, Work Trips (%) Transit Share Nonwork Trips (%) Living in CN 24 44 20 Not in CN 9 31 4 Full Sample 13 34 8 CN/not CN pair values significantly different at p < .05; n ranges from 643 to 222. Location Walk Share, All Trips (%) Walk Share, Nonwork Trips (%) Monthly Utilitarian Walk Trips (No.) Nonwork Utilitarian Walk Trips (No.) Living in CN 20 22 22 20 Not in CN 9 10 9 8 Full Sample 11 13 12 11 CN/not CN pair values significantly different at p < .05; n ranges from 643 to 197. Table 8-5. Neighborhood and green mode shares. Table 8-6. Neighborhood and transit mode shares. Table 8-7. Neighborhood and walking.

housing mix only. The relationship between housing mix and green mode use can be seen in Table 8-8. Table 8-8 shows the importance of mixed housing charac- teristics, separated out from the importance of transit service and commercial destinations within walking distance. A total green mode share of 30% is reported for the mixed neigh- borhoods, compared with 41% in the full CNs (as shown in Table 8-8. In terms of reported utilitarian walking trips, those from neighborhoods with a variety of housing types report making 15 walking trips a month, versus seven trips from those in neighborhoods with only single-family housing. Clearly, the existence of a mix of housing types is a major (though partial) component of the total effect of CNs in terms of travel behavior. The Combination of Personal Values, Urban Form, and Travel Behavior This section of the chapter will examine the relationship between travel behavior concerning walking/transit and the combined forces of personal values and the nature of the built environment. To visualize the combined impact of both pos- itive and negative influences on the use of walking and transit, a simple four-cell matrix will be used throughout this section. This format allows the user to examine variation separately (looking along either the rows or the columns) or together (looking at the relationship of the four cells to each other). Creating the Four-Cell Matrix The four-cell matrix shows the modal share associated with a combination of personal values (represented by the two columns) and supportive neighborhood conditions (repre- sented by the rows). Table 8-9 shows the basic table, as well as the percentage of respondents that fell into each cell. As Table 8-9 shows, when both the values and the envi- ronment are positive, the highest green mode share results (51%). When both the attitudes and the environment are negative, the lowest green mode share results (12%). For each “conflicted” group, when one factor is positive and the second factor is negative, the value looks something like the average value for the sample (24% green mode share). The four-cell matrix shows how the two independent vari- ables interact. For example, if someone who holds a set of high urban/environmental values were to move into a CN, and if she were to act like others in the CN with similar val- ues, the result would be a high mode share for transit and walking. On the other hand, if someone with low levels of urban/environmental values were to move to a CN, and if she were to act like others in the CN with similar values, the re- sultant mode share would be similar to the average for the entire sample. Relationship to the Theory of Planned Behavior Seen in terms of the TPB, the urban/environmental values can be viewed as a combination of the attitude and the subjec- tive norm. Similarly, the extent to which the built environment either facilitates or impedes the adoption of the behavior can be viewed as relating to self-confidence. In our admitted sim- plification of the factors included in the TPB, the subjects with positive attitude/SN, whose decision is facilitated by the built environment, end up with 51% adoption of the behavior. The conflicted subjects with positive attitude/SN, but whose envi- ronment impedes the adoption of the behavior, end up with a near-average 24% adoption of the behavior. On the other hand, the group whose attitude/SN is nega- tive, and whose environment tends to impede adoption of the behavior, winds up with only 12% adoption; the conflicted 71 Neighborhood Mix Green Mode Share, All Trips (%) Transit Share, All Trips (%) Walk Share, All Trips (%) Monthly Utilitarian Walk Trips (No.) Mixed Housing Types 30 13 16 15 Single-Family Houses Only 15 8 8 7 Total Sample 24 13 11 12 Mixed housing/single housing paired values significantly different at p < .05; n ranges from 450 to 415. Table 8-8. Neighborhood housing mix and mode shares. Low Urban/ Environmental Values High Urban/ Environmental Values Living in CN 26% 51% Not in CN 12% 24% Differences between row pairs and column pairs significant at p < .05; n= 333 to 65. Table 8-9. Location and values together—green mode share, all trips.

subject whose attitude/SN is negative, but whose environ- ment is encouraging of the behavior, ends up with a near- average adoption of the behavior (26%). Note that the ap- proach taken in this chapter is highly influenced by the logic put forward in the paper, Changing Individual Travel Be- haviour: From Policy to Perceived Behavioral Control, by S. G. Stradling (50). Applying the Four-Cell Matrix to Walking and Transit Table 8-10 shows the four-cell matrix applied to the green mode share for work and nonwork trips. In general, each of the cells of the matrix appears as expected, with both the con- flicted cells reflecting a value near the average work trip green mode share of 41%. As noted before, the work trip shows less variance associated with the two independent variables than does the nonwork trip. The matrix for green mode share for the nonwork trip is also shown in Table 8-10. The lack of use of green modes for the nonwork trip for those outside the CNs and with low urban/environmental values is contrasted with the robust share for the opposite group. The values for the two con- flicted groups mirror the full sample nonwork mode share of 21%. When the four-cell matrix is limited to transit mode share (Table 8-11), there are similar patterns as with the green mode share, except that many of the differences are no longer significant. In this table the differences that are not significant are indicated by pairs of percentages with the same superscripts. The CN residents take transit at a rate sev- eral times those in non-CNs. For the work trip, transit attracts a significant share from all four market segments, while for the nonwork trip transit mode split declines more sharply either for those in non-CNs or with low urban/ environmental values. Most work trip transit share differ- 72 Green Mode Share, Work Trips Green Mode Share, Nonwork Trips Low Urban/ Environmental Values (%) High Urban/ Environmental Values (%) Low Urban/ Environmental Values (%) High Urban/ Environmental Values (%) Living in CN 41 63 22 49 Not in CN 27 43 9 19 Difference between row and column pairs significant at p < .05; n= 333 to 58. Table 8-10. Location and values together, by trip purpose. Transit Share, All Trips Transit Share, Work Trips Transit Share, Nonwork Trips Low Urban/ Environ- mental Values (%) High Urban/ Environ- mental Values (%) Low Urban/ Environ- mental Values (%) High Urban/ Environ- mental Values (%) Low Urban/ Environ- mental Values (%) High Urban/ Environ- mental Values (%) Living in CN 16x,y 27y 34a,c 47b,c 12 23 Not in CN 7x 12 24a 38b 2 6 Differences between row and column pairs (except subscripted pairs) significant at p < .05 n = 333 to 58. Differences that are not significant are indicated by pairs of percentages with the same superscripts. Table 8-11. Location and values together—transit share by trip purpose.

ences in Table 8-11 are not significant. The exception is that, for those not living in a CN, there is a significant difference in transit mode share to work between the high values group and the low values group. Turning to the walking patterns (Table 8-12), it becomes clear that, for each cell of our matrix, walk share does not vary much by trip purpose. Looking at the data in terms of absolute number of walk trips per month, the most positive group takes about 26 walk trips per month, while the least positive group takes about five trips per month. The two conflicted groups each report about 12 trips per month, as shown in Table 8-13. Auto Availability and Travel Behavior In the previous section, modal behavior describing transit and walking was summarized in terms of its relationship to the combination of two independent variables. This section examines the impact on modal behavior of a third factor— auto availability—and its interaction with the original two variables. Lower-than-average levels of auto availability is a key characteristic of life in a CN, and it is associated with higher use of walking and transit. Several variables were examined in the project for use as a proxy for the latent condition of automobile orientation. The calculations in this section of the chapter were based on the cre- ation of two groups—one group with fewer cars in their house- hold than adults, and a second group with an equal or greater number of cars than the number of adults. The first group is referred to as having a low auto availability level, while the sec- ond group is labeled the high auto availability level. Other concepts could be used in this analysis. Based on responses to the statement “I need to drive my car to go where I need to go,” two groups were formed, one with below- average values and one with above-average values on the statement. When it was applied to the analysis below, the im- pact of this variable on travel behavior was extremely similar to the results derived from the use of the two auto availabil- ity level groups chosen for this analysis. In the end, the variable of auto availability level was chosen because of the simplicity of its logic and because this variable can be created from readily availably sources, such as the National House- hold Travel Survey and (in aggregate) the U.S. Census. 73 Walk Share, All Trips Walk Share, Nonwork Trips Low Urban/Environmental Values (%) High Urban/Environmental Values (%) Low Urban/Environmental Values (%) High Urban/Environmental Values (%) Living in CN 10a 24 10b 26 Not in CN 6a 12 6b 13 Differences between row and column pairs significant at p <.05 except for pairs a,b; n=333 to 65. Table 8-12. Location and values together—walking, by trip purpose. Number of Monthly Utilitarian Walk Trips Number of Nonwork Utilitarian Walk Trips Low Urban/Environmental Values High Urban/Environmental Values Low Urban/Environmental Values High Urban/Environmental Values Living in CN 12 26 10 24 Not in CN 5 12 5 12 Differences between row and column pairs significant at p < .05; n = 333 to 65. Table 8-13. Location and values together—number of walk trips.

Auto Availability as a Characteristic of Neighborhood Type A lower level of auto availability is a key characteristic of life in a CN. In our sample, the majority of persons in CNs (51%) come from a household having less than one car per adult (Table 8-14); for those outside of the CNs, only 25% have less than one car per adult. The average auto ownership rate is 1.2 cars per household within CNs and 1.9 outside CNs. Clearly, low levels of auto availability are associated with location in a CN. Auto Availability and Travel Behavior Table 8-15 shows the relationship between the two levels of auto availability and the propensity to take a green mode for dif- ferent trip purposes. Those from households with less than one car per adult have nearly two to three times the propensity to take a green mode than those who have at least one car per adult. From Table 8-15, the relationship between auto availability level and travel behavior for green modes (transit and walk) can be observed. Auto Availability, Personal Values, and Travel Behavior The interaction between auto availability and personal val- ues in the formation of travel behavior is shown in Table 8-16. Those with two conditions supportive of green mode use have a 49% share, while those with two conditions unsupportive of green mode use have a 12% share. Those in the two conflicted cells act as expected. Auto Availability, Urban Form, and Travel Behavior The relationship between travel behavior and the combi- nation of auto availability and neighborhood type is shown in Table 8-17. When the two conditions supportive of walking and transit are present, the green mode share is 62%; when the two conditions unsupportive of walking and transit are present, only a 14% share is reported. 74 Location Low Auto Availability (%) In CN 51 Not In CN 25 Table 8-14. Percentage of group with low auto availability, by neighborhood type. Green Mode Trips (%) Transit Share,Share, All All Trips (%) Transit Share, Work Trips (%) Walk Share, All Trips (%) Monthly Utilitarian Walk Trips (No.) Low Auto Availability 43 24 49 19 21 High Auto Availability 16 8 27 8 7 Differences between column pairs are significant at p < .05; n = 591 to 249. Table 8-15. Auto availability and modal behavior. Green Mode Share, All Trips Low Urban/Environmental Values (%) High Urban/Environmental Values (%) Low Auto Availability 27 49 High Auto Availability 12 21 Differences between row and column pairs significant at p < .05; n = 324 to 74. Table 8-16. Green mode share, based on auto availability and personal values.

Personal Values, Urban Form, Auto Availability, and Travel Behavior Table 8-18 shows the derivation of the matrix showing green mode share for all trip purposes, based on the interac- tion of all three major variables. At the extremes, the best case scenario (i.e., positive attitudes, supportive built environ- ment, and low levels of auto availability) produces a 64% green mode share; the worst case scenario produces an 11% green mode share. Figure 8-1 presents those data in a graphic form. Most of the pairs in the rightmost column are signifi- cantly different. Table 8-19 presents a summary of the all of the categories with transit share, walk share, and the absolute number of walking trips per month in addition to the green mode total (walking plus transit). Examination of Relationships Using Structural Equation Modeling A final analysis examines the relative importance of each of the three factors. It is clear that there is a certain amount of correlation between the urban/environmental values and the choice to reside or not to reside in the CN; there is correlation between the choice of the neighborhood and the number of autos owned therein; and there is correlation between num- ber of autos owned and the urban/environmental values. To understand the separate roles of each of the three variables, a statistical process known as SEM was used. This helps to iso- late the degree of explanatory power for each of the variables. A Structural Equation Model with the Three Variables A path diagram from a structural model using the three variables of personal values, neighborhood type, and auto availability is shown in Figure 8-2. In this model, the binary categories of high urban/environmental values and low urban/environmental values were replaced by the continu- ous variable representing all reported values on the com- bined scale. The binary categories of high auto availability and low auto availability were replaced by the continuous variable autos per person, representing all reported levels of autos per adult. For the question about living in a CN, the binary values were retained. Several demographic variables were tested, and income per person was retained. Figure 8-2 reveals how the three observed exogenous vari- ables relate to each other and to the endogenous variable of green mode share. In relation to the propensity to walk or take public transportation (green modes), the standardized coefficient for personal values is 0.20, while the coefficient for neighborhood is 0.23. The coefficient for autos per person (−0.43) is the largest in absolute value, reflecting a negative relationship with the number of cars owned per adult in the household and the propensity to take green modes. The signs for each of the correlations between the three exogenous variables (as shown in the double-ended arrows on the left side of the diagram) are logical. Having positive views towards urban attributes and environmental concerns is positively correlated with the decision to live in a CN. Com- pact neighborhood location is negatively correlated with the number of cars owned, and the pro-urban, pro-environmental personal values are negatively associated with the number of cars owned. All coefficients from the exogenous variables (including income per person) are significant at p < .001. The r2 equiva- lent is 0.44. There is no one measure to determine the “good- ness of fit” for a structural equation model. The following set of measures indicates that this model performs only moder- ately well: the normed fit index is 0.898, which is above the minimum level of 0.8; the comparative fit index is 0.9, with 1 being the most desirable value. However, the Tucker-Lewis index is 0.5, when it should be at least 0.9. A root mean squared error of approximation (RMSEA) of 0.16 is above the desired maximum of 0.05. All of these indices combine to suggest that there are more explanatory factors to be identi- fied in future research. A Better Model Figure 8-3 shows a path diagram from a structural equa- tion model derived by adding additional exogenous variables. Figure 8-3 expands upon the exogenous variables of Figure 8-2 to look at factors associated with autos per person, then at the choice of green modes (the share of total trips by tran- sit or walking). 75 Green Mode Share, All trips Not in CN In CN Low Auto Availability 30% 62% High Auto Availability 14% 25% Differences between row and column pairs significant at p < .05; n = 482 to 109. Table 8-17. Green mode share, based on auto availability and location.

Urban/ Environmental Values Group Current CN Status Auto Availability Index Share (% ) Low Auto Availability 64 High Auto Availability 31 In CN Total 51 Low Auto Availability 35 High Auto Availability 18 Not in CN Total 24 Low Auto Availability 49 High Auto Availability 21 High Urban/Environm ental Values Total Total 33 Low Auto Availability 50 High Auto Availability 17 In CN Total 26 Low Auto Availability 20 High Auto Availability 11 Not in CN Total 12 Low Auto Availability 27 High Auto Availability 12 Low Urban/Environm ental Values Total Total 15 Total In CN Low Auto Availability 62 High Auto Availability 25 Total 44 Low Auto Availability 30 High Auto Availability 14 Not in CN Total 18 Low Auto Availability 42 High Auto Availability 16 Total Total 24 n = 276 to 17 Table 8-18. Green mode share, all trip purposes.

77 The Full Sample Green Mode Share = 24% N = 865 High Urban/ Environmental Values Group Green Mode Share = 33% Low Urban/ Environmental Values Group Green Mode Share = 15% N = 398 N = 467 Those with High Urban/Environmental Values Who Live in a Compact Neighborhood Green Share = 51% N = 157 Those with High Urban/Environmental Values Who Do Not Live in a CN Green Share =24% N = 310 Those with Low Urban/Environmental Values Who Live in a CN Green Share = 26% N = 65 Those with Low Urban/Environmental Values Who Do Not Live in a CN Green Share = 12% N = 333 High Urban/Env. Values, CN Location, Low Auto Availability Green Share = 64% n = 96 High Urban/Env Values, non-CN, Low Auto Availability Green Share = 35% n = 104 High Urban/Env Values, non-CN Location, High auto availability Green Share = 18% n = 206 Low Urban/Env.Values, CN Loc. Low Auto Availability Green Share = 50% n = 17 Low Urban/Env. Values, CN Location, High Auto Availability Green Share = 17% n = 48 Low Urban/Env. Values, non-CN Location, Low Auto Availability Green Share = 20% n = 57 Low Urban/Env Values, non-CN Location, High Auto Availability Green Share = 11% n = 276 High Urban/Env. Values, CN Location, High Auto Availability Green Share = 31% n = 61 Figure 8-1. Green mode share with all three independent variables combined.

78 Current Compact Neighborhood Status Auto Availability Index Green Mode Share, All Trip Purposes (%) Transit Share, All Trip Purposes (%) Walk Share, All Trip Purposes (%) Monthly Utilitarian Walk Trips (No.) Low Auto Availability, n = 96 64 36 2 31 High Urban/ Environmental Values, Currently in CN High Auto Availability, n = 61 31 13 18 19 Low Auto Availability, n = 104 35 20 15 14a High Urban/ Environmental Values, Not in CN High Auto Availability, n = 206 17 8 10 11a Low Auto Availability, n = 17 50* 30* 20* 27* Low Urban/ Environmental Values, Currently in CN High Auto Availability, n = 48 17* 11* 6* 6* Low Auto Availability, n = 57 20 11 9 9 Low Urban/ Environmental Values, Not in CN High Auto Availability, n = 276 11 6 5 5 High/low auto availability pairs significantly different at p < .05, except a. * = not computed due to small segment size. Table 8-19. Mode share by urban form, values, and auto availability. Urban/ Environmental Values Compact Neighborhood Autos per Person .44 Green Mode Share Err1 .20 .23 -.43 Income .21 -.29 -.25 .17 Figure 8-2. Structural equation model with values, neighborhood type, and auto ownership.

The Variables This structural equation model uses the three variables used previously: (a) a composite variable based on 15 statements reflecting attitudes toward features of an urban neighborhood and toward the environment (urban/environmental values); (b) a CN variable, which is one if the respondent lives in a CN and zero otherwise; and (c) autos per person, which is the num- ber of autos per household divided by the number of adults in that household. In addition, the model uses new variables, in- cluding affective love for several cars, which is based on agree- ment with the statement “I love the freedom and independ- ence that comes from owning several cars.” Another variable is auto dependence, which is based on agreement with the statement “I need to drive my car to get where I need to go.” The survey text for these new auto-related variables is as follows: How strongly do you agree or disagree with the following? When I think of things that are important to me . . . 1 2 3 4 5 6 7 STRONGLY STRONGLY DISAGREE AGREE I love the freedom and independence that owning several cars provides for my household. [affective love for several cars] I need to drive my car to get where I need to go. [auto dependence] The model shown in Figure 8-3 also provides a way to interpret much of what was learned in the first phase of the study. At the top of the diagram lies a surrogate for our val- ues and beliefs, the factor called urban/environmental values. It is a composite of concerns about one’s personal attitudes about neighborhood characteristics (e.g., “I would like to live where I can walk to the coffee shop”), one’s personal attitudes towards environmental issues, and one’s belief about the environmental attitudes of one’s family and friends. This set of values is applied in the choice of mode (as in Figure 8-2) and also to the question of the number of autos per person in the household. In the Figure 8-3 path diagram, personal val- ues are used to directly explain the green mode share and the dependent variable autos per person. Having these personal values is positively associated with a high green mode share and negatively associated with having more cars. The same path diagram reveals that having these pro-urban and pro-environmental values is positively corre- lated with living in a CN and negatively correlated with a feel- ing of auto dependence and a feeling of freedom and inde- pendence from having several cars. Components of the Model The structural equation model shown in Figure 8-3 can be seen as the simultaneous application of two component sub- models. 79 Gender (f) Auto Dependence Compact Neighborhood .35 Autos per person .54 Green Mode Share err1 Income per person Urban/Environmental Values err2 Affective Love for Several Cars .46 -.24 -.07 -.07 -.31 .16 .23 .12 -.15 -.07 .27 -.29 -.34 .21 -.10 -.14 .14 -.40 .27 -.24 Figure 8-3. Structural equation model with the three variables, plus other values and constraints.

• The Mode Choice Submodel: The path diagram describes several variables associated with green mode share: urban/environmental values, neighborhood type, auto dependence, autos per person, and the demographic val- ues of income per person and gender. This is similar to the content of the Figure 8-2 path diagram, with the addition of auto dependence as a direct independent variable. • The Auto Ownership Submodel: Autos per person in the household is affected by urban/environmental values, neighborhood type, auto dependence, affective love for several cars, and income per person. Detailed Model Results • The Mode Choice Submodel. The independent variable with the strongest association with green mode share is auto dependence. Green mode share is negatively associ- ated with auto dependence, with a standardized coefficient of -0.40, and negatively associated with autos per person, with a standardized coefficient of -0.29. Green mode share is positively associated with urban/environmental values, with a standardized coefficient of 0.14, and with living in a CN, with a standardized coefficient of 0.16. Green mode share is positively associated with income per person (0.12) and negatively associated with being female (-0.07). The dependent variable green mode share has an r2 equivalent of 0.54. • The Auto Ownership Submodel: Auto dependence and in- come per person have the strongest association with autos per person. Autos per person has positive associations with standardized coefficients of (0.27) for auto dependence and (0.23) for affective love for several cars. Autos per person has a negative coefficient (-.10) with urban/environmental values; it is also negatively associated with living in a densely settled neighborhood (-0.14). Autos per person is also pos- itively associated with income per person (0.27). The auto ownership model has an r2 equivalent of 0.35. All of the coefficients described above are significant at the p < .001 level, except the coefficient between the independent variable female and the dependent variable green mode share, which is significant at p < .03. As a whole, this model performs well. The Tucker-Lewis index is 0.986, which is close to 1.0, as desired. The compar- ative fit index is 0.997, also close to 1.0. The RMSEA is 0.026, which is considerably better than the desired maximum value for good fit of 0.05. In the structural equation model, one variable can be asso- ciated with another through direct effects, as shown in Figure 8-3 and also indirectly through its association with interven- ing variables. For example, auto dependence is associated with green mode share directly and indirectly through its association with autos per person. The structural equation model thus provides the means to examine the total associa- tion between variables. Table 8-20 shows the standardized total effects for the model shown in Figure 8-3. As can be seen, auto dependence has the greatest total effect on both of the dependent variables. Conclusion: Learning from the Revised Model The revised structural equation model represents the simultaneous examination of several variables that interact in the behaviors that ultimately effect the choice of transit and walking for utilitarian travel. The SEM process allows the cumulative examination of the direct and indirect “standard- ized total effects” on all endogenous variables. Importantly, the strongest total association with the decision to take green modes comes from (in order) auto dependence (-0.483), autos per person (-0.292), CN (0.204), and urban/environmental values (0.166). The dominance of auto dependence suggests that this variable is deserving of further research. Summary Observations A key theme in the development of this research is the understanding of how people’s residential decisions are in- terrelated with their decisions concerning trip making by walking and transit. This chapter has emphasized the impor- 80 Affective Love for Several Cars Urban/ Environ- mental Values Auto Depend- ence Income per Person CN Female Autos per Person Autos/ Person 0.226 -0.104 0.273 0.267 -0.139 0.000 ---- Green Mode Choice -.066 0.166 -0.483 0.046 0.204 -0.073 -0.292 Table 8-20. Standardized total effects for two dependent variables.

tance of the simultaneous interaction of key variables con- cerning neighborhood characteristics and their association with various indices of travel behavior concerning walking and transit. Based on the presentation of the data up to this point, three summary observations about the key variables can be made. Personal Values Consistent with other recent work in this field, it is clear that one’s personal values have an impact on one’s decision to make trips by walking and transit. An urban infrastructure has an independent effect on walking and transit, but the impact of values is of similar magnitude. The creation of a single, combined scale of urban/environ- mental values derived from responses to 15 separate ques- tions allows for a quick and easily replicated categorization of groups. This project has demonstrated that the analysis group with high urban/environmental values has more than two times the propensity to choose green modes as the group with low urban/environmental values. This observation alone is enough to encourage the further study of the effects of personal values on travel behavior concerning walking and transit. Urban Form (Compact Neighborhood) A built environment designed to facilitate and support the use of walking and transit is associated with significantly higher use of these modes than experienced in other areas. The definition of a CN as one with mixed housing, with a com- mercial area within walking distance, and with transit avail- able provides a consistent, easily applied definition for a set of conditions representing the concept of a well-designed built environment. As summarized in this chapter, those in the sample who lived within a CN had walking/transit rates more than twice those reported for those who lived outside of a CN. Noticeably, transit plays a key role in nonwork trips within the CN that do not occur outside of the CNs. For the conflicted user with high urban/environmental values, but not supported by the conditions of the CN, the com- bination of positive factors does not occur, reflected in green mode share that mimics the average of the sample. But when the high urban/environmental values group is supported and fa- cilitated by the physical environment, the majority of trips are undertaken by walking and transit. Auto Orientation A third phenomenon associated with the propensity to walk and take transit is the orientation of the subject toward his/her automobile. Of the three variables reviewed in this chapter, the one most deserving of future research may be the influence of an individual’s orientation toward the automobile. The difference in the use of walking and transit between (for example) a couple with two cars and a couple with one car is marked. For those living in a CN, green mode selection is much associated with car availability: CN residents with less than one car per adult report a 62% green mode share, which is 2.5 times the share for those with higher car availability, at 25%. A similar relationship by auto availability level occurs in all the cells of the eight-cell matrix (refer to Table 8-19). For example, for the population with high urban/environmental values and location within a CN, the subgroup with low auto availability reports a 64% green mode share, compared with a 31% share for the high auto availability subgroup. A key variable defined for this research is auto dependence, which is the respondent’s agreement with the statement “I need to drive my car to get where I need to go.” This variable has the strongest association with a respondent’s mode share for walking and taking transit, as well as with the respondent’s auto ownership level. (It also has a strong negative correlation with a location in a CN.) The dominance of auto dependence suggests that further research should explore the extent to which the perception that “I need to drive my car to get where I need to go” can be affected by public policy interventions that actually do lower the level of auto dependence, and the extent to which the perceptions themselves can be altered. 81