Multi-group analysis is a type of analysis that can employ standard errors for path coefficients is what is often referred to as a multi-group analysis (Kock 2013). This section deals with the results obtained from the country comparison, gender comparison, and online experience of using internet for online luxury shoppers’ samples. Kock (2014) argues that differences in the path coefficients between the compared models could be caused artificially by measurement differences. In fact, even though common method bias has already been assessed in this study, it was only checked individually for each group and hence could go unnoticed, and bias the comparison when multi-groups are involved. To avoid such a scenario, equivalence of measurement models needs to be checked and established before
comparing the structural models. In this case, p-values should be greater than the significance threshold.
Comparing two groups in two different countries, comparing gender, and looking at online experience of using the internet, is conducted in a similar way at both the measurement and structural models. First, a pooled standard error is calculated for each path coefficient pair (at the structural model) and weight pair (at the measurement model) using the following equations. If the standard errors are similar in both compared models (pooled method):
Where:
𝑁1 is the sample size for the UK, 𝑁2 is the sample size for the USA, 𝑆1 is the standard error for the path coefficient in the UK, and 𝑆2 is the standard error for the path coefficient in the USA.
If the standard errors are different in both compared models (Satterthwaite method):
Second, the critical ratio T s calculated using the following formula:
𝑇12= (𝛽1−𝛽2)/𝑆12
The obtained T ratio is then used to identify the p-value associated with it. This p-value reveals whether there is any difference between the path coefficients (Keil et al., 2000, Kock, 2014).
1) Country-based MGA
The comparison of the results obtained from UK and USA online luxury shopping samples is conducted at both measurement and structural models. In the present study, the Satterthwaite method (Kock 2014) is used to calculate the pooled standard errors. This is owing to the fact that the standard errors in the UK and the USA samples were found to be different (0.063, 0.066 respectively). However, Kock (2014) recognises that although such a method is not
widely used as it yields slightly higher values for the pooled standard errors, the differences are generally minor. Table 5.1 shows the weight comparison of the constructs included in the final model. In addition, Appendix G shows the weight comparison of the constructs included in the final model.
As it could be seen from Table 5.1, the paths recording statically similarities between the two investigated countries were the relationship between the external factors, attitude with purchase life cycle and purchase life cycle on the COLS model. It can therefore be argued that both countries have similar results. This leads to the conclusion that online luxury consumers in the UK had similar impact on the buying online luxury shopping as in the USA.
Concerning the differences, only the AP to eS and the eS to eL are significantly different between the two countries. In the USA the effect is significantly higher than in the UK.
As can be seen from Table 5.1, all the p-values were statistically non-significant meaning that there was invariance between the measurement models applied in the two countries. This confirms that the measures used in the survey were equal in both countries. Hence, the comparison of the path coefficients can be conducted. Table 5.1 illustrates the path comparison and their p-values.
Table 5.1 Country Comparison
Relationships Path Coefficients UK
UK SE Path Coefficients US
USA SE P-Value External Factors
Electronic Service Quality (eSQ) Attitude (Att
0.48 0.06
0.47 0.06 0.45 NS
Electronic Word of Mouth (eWOM) Attitude (Att
0.06 0.06 0.04 0.06 0.40 NS
Perceived Usefulness
(PU) Attitude (Att 0.11 0.06 0.18 0.066 0.20 NS
Perceived Ease of Use (PEU) Attitude (Att)
0.00 0.063 -0.02 0.066 0.40 NS
Social Network Site Usage (SNNAU) Attitude (Att)
0.10 0.063 0.15 0.066 0.27 NS
Social Media Marketing activitites (SMMA) Attitude (Att)
0.06 0.06 0.13 0.06 0.20 NS
Perceived Brand Value (PBV) Attitude (Att)
0.07 0.06 0.06 0.06 0.45 NS
Attitude with Purchase Life Cycle Attitude (Att)
Intention (Int)
0.68 0.06 0.75 0.06 0.20 NS
Attitude (Att) Actual Purchase
0.36 0.06 0.42 0.06 0.24 NS
(AP)
Attitude (Att) Electronic Satisfaction (eS)
0.52 0.06 0.45 0.06 0.20 NS
Attitude (Att) Electronic Loyalty (eL)
0.47 0.06 0.45 0.06 0.40 NS
Purchase Life cycle Intention (Int)
(AP) Actual Purchase (AP)
0.18 0.06 0.27 0.06 0.14 NS
Actual Purchase (AP) Electronic Satisfaction (eS)
0.18 0.06 0.39 0.06 0.00***
Electronic
Satisfaction (eS) Electronic Loyalty (eL)
0.27 0.06 0.40 0.06 0.00***
***Significant at 1%; **Significant at 5%; *Significant at 10%; NS Non-significant 2) Gender-based MGA
The gender comparison of the results obtained from online luxury shopping samples is conducted at both measurement and structural models. This study used the Satterthwaite method (Kock 2014) to calculate the pooled standard errors. This is owing to the fact that the standard errors in the male and the female samples were found to be different. As Table 5.2 shows, the path results record a statically significant difference between the two investigated gender samples. However, some of coefficients are statically different than zero, for example perceived brand values with attitude, attitude with actual purchase and intention with actual purchase.
As can be seen from Table 5.2, the paths recording a statistically significant difference between gender (male and female) were the relationships between the external factors, attitude with purchase life cycle, and the purchase life cycle on the COLS model. It can therefore be argued that the effect in the male was significantly greater than effect in the female. This leads to the conclusion that luxury male consumers had greater impact in buying online luxury shopping than luxury female consumers.
Online sellers should be aware that the gender gap is narrowing, but more importantly, they should be aware of the factors that might affect male and female shoppers differently. Online sellers must acknowledge that these factors also depend on product type, and should be conscious of these differences when designing marketing plans (San-Martín et al. 2015).
Table 5.2 Gender Comparison
Relationships Male SE Path Female SE Path P-Value
The external Factors Electronic Service
Quality (eSQ) Attitude (Att)
0.07 0.54 0.07 0.44 0.29 NS
Electronic Word of Mouth (eWOM) Attitude (Att)
0.07 0.05 0.08 0.00 0.66 NS
Perceived
Usefulness (PU) Attitude (Att)
0.07 0.098 0.08 0.044 0.62 NS
Perceived Ease of Use (PEU) Attitude (Att)
0.07 0.018 0.07 -0.091 0.33 NS
Social Network Site Usage (SNNAU) Attitude (Att)
0.07 0.147 0.079 0.115 0.77 NS
Social Media Marketing Activities (SMMA) Attitude (Att)
0.076 0.156 0.080 0.044 0.31NS
Perceived Brand Value (PBV) Attitude (Att)
0.077 0.095 0.077 -0.233 0.00***
Attitude with Purchase life Cycle Attitude (Att)
Intention (Int)
0.067 0.738 0.069 0.712 0.78 NS
Attitude (Att) Actual Purchase (AP)
0.075 0.234 0.073 0.502 0.01***
Attitude (Att) Electronic Satisfaction (eS)
0.071 0.499 0.073 0.486 0.89NS
Attitude (Att) Electronic Loyalty (eL)
0.071 0.461 0.075 0.343 0.25 NS
Purchase life Cycle Intention (Int)
(AP) Actual Purchase (AP)
0.074 0.293 0.079 0.106 0.08 *
Actual purchase (AP) Electronic Satisfaction (eS)
0.074 0.320 0.076 0.294 0.80 NS
Electronic
Satisfaction (eS) Electronic Loyalty (eL)
0.073 0.327 0.074 0.409 0.43 NS
***Significant at 1%; **Significant at 5%; *Significant at 10%; NS Non-significant
3) Online Experience level -based MGA
The internet experience level comparison of the results obtained from intermediate level and advanced level from the online luxury shopping samples is conducted at both measurement and structural models. This study used the Satterthwaite method (Kock 2014) is used to calculate the pooled standard errors. This is owing to the fact that the standard errors in the intermediate level and the advanced level samples were found to be different.
As can be seen from Table 5.3, the paths recording a statistically significant difference between the two investigated online experiences of using internet (intermediate and advance) were the relationship between the external factors, attitude with purchase life cycle, and purchase life cycle on the COLS model. It can therefore be argued that the effect in the intermediate level of experience in using internet had greater impact of buying online luxury shopping than advance level.
Table 5.3 Online Experience of using Internet Comparison
Relationships Intermediate SE
Path Advance SE Path P-Value
The External Factors Electronic Service
Quality (eSQ) Attitude (Att)
0.093 0.548 0.074 0.442 0.29NS
Electronic Word of Mouth (eWOM) Attitude (Att)
0.107 0.053 0.081 0.004 0.66 NS
Perceived
Usefulness (PU) Attitude (Att)
0.106 0.098 0.080 0.044 0.62 NS
Perceived Ease of Use (PEU) Attitude (Att)
0.105 0.018 0.079 -0.091 0.33 NS
Social Network Site Usage (SNNAU) Attitude (Att)
0.106 0.147 0.079 0.115 0.77NS
Social Media Marketing Activities (SMMA) Attitude (Att)
0.103 0.156 0.080 0.044 0.31NS
Perceived Brand Value (PBV) Attitude (Att)
0.077 0.095 0.077 -0.233 0.00***
Attitude with purchase life cycle Attitude (Att)
Intention (Int)
0.067
0.738 0.069 0.712 0.78 NS
Attitude (Att) Actual Purchase (AP)
0.075 0.234 0.073 0.502 0.01***
Attitude (Att) Electronic Satisfaction (eS)
0.071 0.499 0.073 0.486 0.89 NS
Attitude (Att) Electronic Loyalty (eL)
0.071 0.461 0.075 0.343 0.25 NS
Purchase life cycle
Intention (Int) (AP) Actual Purchase (AP)
0.074 0.293 0.079 0.106 0.08 *
Actual Purchase (AP) Electronic Satisfaction (eS)
0.074 0.320 0.076 0.294 0.80 NS
Electronic
Satisfaction (eS) Electronic Loyalty (eL)
0.073 0.327 0.074 0.409 0.43NS
***Significant at 1%; **Significant at 5%; *Significant at 10%; NS Non-significant