Understanding consumer online shopping behaviour from the perspective of transaction costs

 

---

254
Table 5.19 Discriminant Validity (N=962) 
Note: IAA=Internet Access Availability; EXP= Perceived Internet Expertise; PQC=Product Quality Concern; ESQ=E-Service Quality; REP= Reputation of
Online Store; CONV=Perceived Convenience; PSC= Privacy and Security Concerns; TCS=Perceived Consumer Transaction Costs; CSAT=Customer 
Satisfaction; LOY= Customer Loyalty; RISK= Risk-bearing Propensity; ENJ=Perceived Enjoyment. 
The bold numbers in the diagonal row are square roots of the AVE.
 
 
 
Internal Consistency
IAA 
EXP 
PQC 
ESQ REP CONV PSC TCS CSAT LOY RISK ENJ 
IAA .848 
.807 
EXP .906 
.623 
.841 
PQC .938 
-.353 
-.313 
.868 
ESQ .839 
.599 
.646 
-.753 
.797 
REP 
.886 
.585 .614 -.604 .591 
.812 
CONV 
.876 
.622 .682 -.443 .674 .684 
.765 
PSC .930 
-.496 
-.505 
.766 
-.542 
-.723 
-.623 
.853 
TCS .856 
-.642 
-.645 
.748 
.639 
-.783 
-.559 
.682 
.816 
CSAT .916 .582 
.651 
-.669 
.637 
.730 
.539 
-.668 
-.566 
.885 
LOY 
.923 
.595 .679 -.668 .639 .643 .552 -.699 -.570 .558 
.840 
RISK 
.839 
.479 .485 -.649 .692 .662 .539 -.601 -.639 .514 .605 
.715 
ENJ 
.906 
.552 .589 -.536 .647 .666 .553 -.656 -.641 .716 .736 .658 
.812 

---

255
5.6.6 Common Method Variance 
As single sources of information can introduce spurious relationships among the variables, 
and as this study collected data via the same method (self-report scales), the need to test for 
common method variance was warranted. The Harman
’
s one-factor test through EFA and 
CFA was performed to assess the severity of common method variance.
Firstly, through EFA, all items, presumably measuring a variety of different constructs, were 
subjected to a single factor analysis. Using this approach, sixteen factors were extracted with 
eigenvalues greater than 1 and the variance explained was 78.17 per cent. The first factor 
accounted for 35.53 per cent of the variance with the second factor accounting for 8.79 per 
cent and the remaining fourteen factors sharing 33.85 per cent of the variance. As there was 
not one factor (or a common factor underlying the data) and as the majority of the variance 
was not accounted for by one general factor, a substantial amount of common method 
variance was not evident.
Secondly, the study conducted a CFA to assess the fit of a single factor model (all items 
loading on one factor). The single-factor model showed very poor fit (
χ
2
/df =16.38; GFI= 
0.67; CFI= 0.51; CFI = 0.44; TLI= 0.57; RMSEA=0.18; RMR=0.36). The original twelve 
factor model provided a significantly better fit to the data than a single-factor model (see 
table 5.20). With both tests, it can be concluded that the problem of common method bias 
was not evident in this research. 
The re-verification of convergent and discriminant validity and a test for common method 
variance completed the final phase of the overall measurement model analysis. The next 

---

256
section is dedicated to describing the basic structural model that was established upon the 
new overall measurement model.