• No results found

SEM can be used to test the full TPB model, as shown in Figure 9-1. The components of this model include the following:

Dependent Variable: Attitude Toward the Behavior Independent Variables:

Behavioral Beliefs (If I moved to a compact neighborhood)

Coefficient t-statistic Probability

Constant -2.30* -8.63 .0001

I would exercise by

walking or bicycling. 0.10* 2.93 .0035

I would make friends with

more of my neighbors. 0.10* 2.57 .0103

It would be easy for me to get to stores, restaurants, a

library and other activities. 0.27* 5.76 .0001 I would take public

transportation to work or

for other trips. 0.13* 4.18 .0001

My household could own

fewer cars. 0.10* 3.63 .0009

The streets would be noisier than where I live

now. -0.19* -5.73 .0001

I would have less living

space in my home and lot. -0.12* -3.45 .0006 *significant at probability level indicated

n = 822, R2= 0.32.

Table 9-4. Regression for attitude toward the behavior.

Normative Beliefs (1 = extremely unlikely to 7 =

extremely likely) Mean (SD)

Motivation to Comply (1 = not at all to 7 = very much)

Mean (SD)

My family thinks that I should move to a compact

neighborhood.

2.4 (1.6) Generally speaking, how much do you care what your family thinks you should do?

4.3*(1.8)

Other people who are important to me think that I should move to a compact neighborhood.

2.4 (1.6) Generally speaking, how much do you care what other people who are important to you think you should do?

3.9*(1.6)

*significantly different at p < .05 n = 822

Table 9-5. Mean and standard deviation for normative beliefs and motivation to comply.

• Seven factors hypothesized to influence attitude towards the behavior that are the products of

– behavioral beliefs and – outcome evaluation.

• Two factors hypothesized to influence SN that are the products of

– normative beliefs and – motivation to comply.

• Four factors hypothesized to influence self-confidence that are products of

– control beliefs and – power of control.

Within each intention and direct measures box shown in Figure 9-1, the ratings on each statement are averaged to cre- ate a single score. With respect to the indirect measures, each element in the “belief” box is multiplied by its corresponding element in the “relevance” box (outcome evaluation, moti- vation to comply, and power of control).

The structural equation model attempts to predict intent from the direct measure scores while also attempting to predict each direct measure score from its corresponding set of indi- rect measure products. The key results that are produced are the coefficients (and significance levels) for (a) direct measures predicting intent and (b) indirect measures predicting the cor- responding direct measures (and, consequently, intent). Results

In Table 9-9, the columns represent the following (left to right):

• Endogenous variables • Direction of association • Exogenous variables

Dependent Variable: Subjective Norm

Independent Variables:

Normative Beliefs Coefficient t-statistic Probability

Constant 1.67* 23.8 0.0001 My family thinks that I

should move to a compact

neighborhood. 0.28* 5.70 0.0001 Other people who are

important to me think that I should move to a compact

neighborhood. 0.35* 7.01 0.0001 *significant at probability level indicated

n = 822, R2= 0.45.

Table 9-6. Regression for subjective norm.

Control Beliefs

(1 = very unlikely to 7 = very

likely) Mean (SD)

Power of Control (1 = strongly

disagree to 7 = strongly agree) Mean (SD) How likely is it that you could

find an affordable home in a compact neighborhood?

3.8 (1.9) It would be easier for me to move to a compact

neighborhood if I could find an affordable home there.

4.6 (2.0)

How likely is it that you could get by with less living space in the coming year?

3.0 (2.1) It would be easier for me to move to a compact

neighborhood if I required less living space.

4.1 (2.0)

How likely is it that you would lose touch with current friends if you moved to a compact neighborhood?

3.2 (2.0) It would be easier for me to move to a compact neighborhood if I was sure I would not lose touch with my current friends.

3.5 (2.0)

How likely is it that you could get by with fewer household cars in the coming year?

2.9 (2.2) It would be easier for me to move to a compact

neighborhood if I didn’t need so many household cars.

3.5 (2.1)

n = 822

Table 9-7. Mean and standard deviation for control beliefs and power of control.

• Operation between exogenous variables (for indirect measures)

• Second exogenous variable (for indirect measures) • Regression weights (estimate)

• Standard error (SE) of the estimate

• critical ratio (CR), analogous to a t-statistic—higher is better • Probability from significance test (P)—lower is better. This

is the chance that the coefficient could have been zero. The values in the estimate and probability columns are the key results to focus on; the bolded text represents relation- ships that test as significant at the 5% level, which means there is a 5% or less chance that the coefficient is zero.

Based on this model, ATT and SN are more important than SCF as influences on intent to move to a compact neigh- borhood (although SCF is still a significant factor). Note that the relationship shown between intent and ATT, SN, and SCF is very similar to what was shown in Table 9-2.

With regard to the indirect measure of ATT, it was found that several of the features thought to be important were in- deed significant. These included the ability to walk to stores, the ability to take public transportation, and the ability to do

with one less car. The ability to make friends with neighbors was also significant. This was a surprise to the researchers, but had come up in the focus groups as a possible advantage of a CN. Surprisingly, the ability to exercise by walking or bicy- cling did not turn out to be significant. Neither did the two negative factors (noise on the street and less living space).

Based on the output in Table 9-9, the most important relationships are between

• SN and others’ opinions,

• ATT and access to commercial districts, • SCF and affordability, and

• SN and family’s opinion.

Relationships that were not significant are between • ATT and exercise,

• ATT and noise,

• ATT and living space, and

• SCF and requiring less living space.

Thus the SEM result shown in Table 9-9 finds many, but not all, of the same factors to be significant as the regression analyses shown in Tables 9-4, 9-6, and 9-8.

Overall, the fit of the structural equation model is poor. The Tucker-Lewis index is 0.22, when it should be at least 0.9. The comparative fit index is 0.32, when it should be at least 0.9. The RMSEA is 0.21, when it should be less than 0.06. Thus while many of the hypothesized factors do affect the ATT, SN, and SCF, clearly more research is needed to more fully describe the factors affecting choice of a CN.

Outline

Related documents