Understanding consumer online shopping behaviour from the perspective of transaction costs

4.10 CHAPTER SUMMARY 
 
This chapter has described the methodological approach used to conduct the present study. A 
sample of Chinese online shoppers was surveyed. A four-stage process was employed for the 
development of the questionnaire. The data collection was conducted on the streets in 
Shanghai and Nanchong based on a face-to-face survey. The survey instrument was 
developed based on both published survey instruments and the broader extant literature. The 
major issues pertaining to reliability and validity in the design of the study, as well as 
comments on the appropriateness and soundness of the methods utilized were discussed in 
this chapter. A discussion of the analytical technique employed (SEM) highlighted the 
important considerations in the interpretation of results from such an analytical technique. 
Specifically, in this study, the two-stage model-building approach, where the measurement 
model is adjusted and fitted to the data prior to testing the structural model, was employed, 
and six goodness-of-fit measures (
χ
2
, GFI, CFI, TLI, RMSEA and RMR) were used to assist 
in data interpretation. 

---

205
CHAPTER 5: DATA ANALYSIS 
5.1 CHAPTER OVERVIEW 
Chapter five consists of eleven sections and is structured as follows. The chapter begins with 
data cleaning and screening to check for incompleteness and inconsistencies, followed by an 
overview of the respondent profile. Then, preliminary analysis of the data was conducted via 
correlation analysis, exploratory factor analysis (EFA), confirmatory factor analysis (CFA) 
and reliability estimates. Following this, results of the preliminary analysis of the items 
measuring each multi-item construct were presented. The purpose of the preliminary analysis 
is to identify a set of indicators that represents each multi-item constructs in the overall 
measurement model. Next, using the two-stage model-building approach recommended by 
Anderson and Gerbing (1988), the overall measurement model was first confirmed and then 
the structural model (n = 962) for controlling for age, gender, income and education level 
were assessed, respectively. The structural model was to test the causal research hypotheses 
of this study (i.e., H1a, H1b, H1c, H1d, H1e, H2b, H2c, H3a, H4a, H4b, H4c, and H5). The 
mediation testing was conducted in order to address H6 which proposed that customer 
satisfaction partially mediates the effect of consumer TCs on loyalty. Following this, the 
moderating effects of consumer’s risk-bearing propensity and perceived enjoyment of online 
shopping (i.e., H7a, H7b, H8a, and H8b) were tested using multiple group analysis via 
structural equation modelling. Lastly, a comparative analysis was undertaken to determine if 
the proposed model differed across two product categories (i.e., search products vs. 
experience products), which aimed to test the final hypotheses H9a and H9b of this study. 
The chapter concludes with a summary of the findings associated with the tested hypotheses.