Information Technology Reference
In-Depth Information
Table 2.2
Factor analysis of all indicators
Variable
Factor 1
Factor 2
Uniqueness
Trust-related scales
Store trustworthiness
0.86
0.02
0.24
Benevolence
0.89
0.01
0.22
Integrity
0.90
0.02
0.17
Competence
0.82
0.05
0.28
Transaction intentions
Item 1
0.18
0.75
0.26
Item 2
0.05
0.92
0.20
Item 3
0.14
0.81
0.21
Item 4
0.10
0.86
0.34
Variance
3.99
3.78
N
¼
751
Proportion
0.50
0.47
Method: principal component factor analysis
Rotation: oblique promax
We used slightly reworded versions of three subscales from the trusting belief scale
(McKnight et al.
2002
), benevolence scale (three items,
a
¼
0.72), integrity scale
(four items,
a
¼
0.89), and competence scale (four items,
a
¼
0.85).
3. Transaction intentions
We used a total of four items: one of three items on the transaction intentions scale
(Item 1 in the
Appendix
) (Pavlou and Gefen
2004
) and three of four items on the
intention to buy scale (Items 2 to 4 in the
Appendix
) (Stewart
2003
)(
a
¼
0.86).
A factor analysis (principal component method) of all of the indicators above
clearly indicated a two-factor structure (Table
2.2
). This result validates our models
with two latent variables; i.e. trust in an e-commerce store and intention to buy.
There were 941 responses to the 1,620 requests to participate in this study. We
excluded data with low trustworthiness by eliminating those from the fastest and
slowest 5 % of respondents in terms of total response time as well as those from
respondents who spent less than 10 s or more than 10 min on steps 2 and 3. In
addition, we eliminated cases that indicated incomplete manipulation.
4
This left
751 subjects for analysis. An ex posteriori sample size calculation (Westland
2010
)
indicated that the sample size of 751 subjects is adequate for hypothesis testing
using structural equation models.
4
A manipulation check was conducted after the dependent variables were measured. Respondents
were asked, “Of the previous buyers who previously bought crab from this store, do you believe
that there are many people who are similar to you, or do you believe that there are more people
who are dissimilar to you, based on the criteria below?” We eliminated the subjects who responded
“dissimilar” to at least one of the two salient values although they were assigned to the similar
salient consumer value condition, as well as subjects who responded “similar” to at least one of the
two salient values although they were assigned to the dissimilar salient consumer value condition.
Sixty-six subjects were eliminated by this manipulation check. This may be because of an interval
of up to 2 weeks between the presurvey and experimental survey and possible differences in
evaluation criteria used for food products in general in the presurvey and for crab only in the
experimental survey.
