Sem Results Of Theoretical Model After Calibration (Standardization)

The research results show that there is no impact of the variables tangible means (PTHH) and safety and efficiency (ATHQ) because they are not statistically significant, meaning they do not have much impact on the satisfaction of customers using Vietcombank ATM cards (or the explanatory ability of these two variables is the weakest), so hypotheses H3 and H4a are not accepted. Meanwhile, the factors price (GC), network (ML), reliability (DTC) and empathy (SDC) have an impact on customer satisfaction, in which the variable GC has the strongest impact, so hypotheses H1, H2, H6, H7 are all accepted.

Validation of theoretical model after calibration

The research model after adjustment is shown in Figure 4.4. At that time, customer satisfaction using ATM card services only includes 4 components: (1) Price (GC);

(2) Network (ML); (3) Reliability (DTC) and (4) Empathy (SDC). Customer satisfaction (SHL) is assessed through 4 service quality components.

Figure 4.4: SEM results of the theoretical model after calibration (standardization)

The factors price, network, reliability, and empathy explain nearly 46.5% of customer satisfaction using ATM card services of Vietcombank, Vinh Long branch. Of the four factors above, price has the strongest impact on SHL (β = .317), followed by network (β = .235), empathy (β = .139), and reliability.

reliability (β =.135). The model fit indices all meet the requirements, so the model fits the market data shown in the table below:

Table 4.12 Results of some indicators

Index

Results from the model	Good model reference value
Chi-square/df	2,505	<=3
TLI	0.965	>0.9
CFI	0.972	>0.9
RMSEA	0.044	<0.08

Maybe you are interested!

Source: Author's synthesis based on analysis results

The estimated coefficients between the components in the model are all greater than 0.9 and the p-values of these estimates are less than 0.05. Thus, all four hypotheses are accepted. (Table 4.13). And the unstandardized weights are all positive, which also shows that the variables price, network, reliability and empathy have a proportional effect on customer satisfaction.

Table 4.13: Results of hypothesis testing (adjusted model)

The

hypothesis

Relationship	Weight Standardized Beta	SE	CR	P	Result Hypothesis testing
H7	suhailong<---price	0.317	0.070	9,821	0.000	Accept
H6	suhailong<---mangluoi	0.235	0.071	10,734	0.000	Accept
H2	suhailong<---dongcam	0.139	0.073	11,858	0.000	Accept
H4	suhailong<---dotincay	0.135	0.073	11,906	0.000	Accept

Hypothesis H7: Price - Satisfaction

Hypothesis H7 states that : “Price has a significant direct impact on customer satisfaction”. From the standardized Beta weight, we see that the impact of price on customer satisfaction is the largest with a statistical significance level of over 95%. (β = 0.317; sig = 0.000 < 0.05). From that, we conclude that perceived price is the factor that has the strongest impact on customer satisfaction. Hypothesis H7 is accepted.

Hypothesis H6: Network - Satisfaction

Hypothesis H6 states that : “Network has a significant direct impact on customer satisfaction”. From the standardized Beta weight, we see that the impact level of network

network to customer satisfaction with statistical significance level above 95%. (β =0.235; sig = 0.000 < 0.000). From that, we conclude that network is the second strongest factor affecting customer satisfaction. Hypothesis H7 is accepted.

Hypothesis H2: Empathy - Satisfaction

Hypothesis H2 states that : “Empathy has a significant direct impact on customer satisfaction” . The results in Table 4.13 show that empathy also has a certain impact on customer satisfaction (β = 0.139; sig = 0.000 <0.05). In other words, empathy of the bank and bank employees has a positive impact on customer satisfaction with a statistical significance level of over 95%. Hypothesis H2 is supported.

Hypothesis H4: Reliability - Satisfaction

Hypothesis H4 suggests that: “ Reliability has a significant direct effect on customer satisfaction ”. Reliability is a good predictor of customer satisfaction with a statistical significance level above 95% (β =0.135; sig = 0.000 <0.05). Hypothesis H4 is accepted.

4.4.3 Re-testing the estimates using bootstrap

The results of bootstrap estimation with the number of replicate samples N=1,500 in the linear structural model analysis show that bias appears but is not significant.

Table 4.14: Bootstrap estimation results with N =1500

Relationship

SE	SE-SE	Mean	Bias	SE-Bias
suhailong<---dotincay	0.046	0.001	0.137	0.003	0.001
suhailong<---dongcam	0.057	0.001	0.138	-0.002	0.001
suhailong<---mangluoi	0.064	0.001	0.24	0.005	0.002
suhailong<---price	0.066	0.001	0.312	-0.005	0.002

Note: se-se: standard deviation of standard deviation; bias: bias; se(bias): standard deviation of bias

4.3.4 Testing the alternative hypothesis of difference

This section presents a multigroup analysis method to test the secondary hypotheses about the differences between the relationships in the theoretical model between customer groups.

male and female; young customer group and middle-aged customer group; high-income customer group and low-income customer group.

4.3.4.1 Multi-group model testing method

The multigroup structural analysis method is used to compare theoretical models according to a certain group of a qualitative group: gender, age, income. The multigroup analysis method used in this study includes 2 models: variable model and invariant model. In the variable model, the estimated parameters in each group model are not constrained. In the invariant model, the estimated parameters in each group model are constrained to have the same value.

The ML (Maximum Likelihood) estimation method is used in multi-group analysis. The Chi-square difference test is used to compare 2 models. If the Chi-square difference test shows that there is no difference between the 2 constant models and the variable models (p-value>0.05), the constant model will be selected (because it has higher degrees of freedom). Conversely, if the Chi-square difference is statistically significant (p-value<0.05), the variable model will be selected (has higher compatibility).

Choose the constant or variable model. We test the following hypothesis: Ho: Chi-square of the variable model is equal to Chi-square of the constant model. H1: Chi-square of the variable model is different from Chi-square of the constant model.

4.3.4.2 Testing the Sub-Hypothesis of Gender Differences

As introduced, the research sample is divided into 2 groups: 369 male customers and 410 female customers. There are 4 sub-hypotheses established that the relationship between factors to female customer satisfaction will be stronger than male (as shown in Figure 4.5).

The results of the multi-group test of the variable model show that the theoretical model has 382 degrees of freedom. The SEM results show that the model achieves compatibility with market data: Chi-square = 894.589; P-value = 0.000; Chi-square/df = 2.342(<2.5); TLI = 0.939 (>=0.9); CFI = 0.949 (>=0.9); RMSEA = 0.042 (<0.08).

Network

Reliability

P1'

P6'

Sex

P7'

P2'

ATM card service customer satisfaction

Empathy

Price of perceived service

Figure 4.5 Multigroup research model with gender control variable

The results of the multi-group invariance model test show that the theoretical model has 386 degrees of freedom. The SEM results show that the model achieves compatibility with market data: Chi-square = 898.482; P-value = 0.000; Chi-square/df = 2.328; TLI = 0.939 (>=0.9); CFI = 0.949 (>=0.9); RMSEA = 0.041 (<0.08).

Table 4.15 Chi-square test between gender-group invariant and heteroscedastic models

Model

CMIN	df	P	NFI	RFI	IFI	TLI
Variable	894,589	382	0.000	0.916	0.898	0.950	0.939
Invariant (partial)	898,482	386	0.000	0.915	0.898	0.950	0.939
Differentiation value	3,893	4	0.421	-0.001	0.000	0.000	0.000

P-value= Chidist(3,893;4) = 0.421

We have P-value =0.421>0.05 accepting the Ho hypothesis. In other words, there is no difference in Chi-square between the constant model and the variable model. When choosing the constant model, we can conclude that there is no difference in the influence of reliability, empathy, network and price on customer satisfaction between male and female customers. The model does not change according to gender. So the hypotheses P1', P2', P6', P7' are all rejected.

4.3.4.3 Testing the Sub-Hypothesis of Age Difference

Similarly, the age research sample was divided into 2 groups: young customers (from 18 to under 35 years old) with 383 customers and middle-aged customers (from 35 to over 55 years old) with 386 customers. There are 4 hypotheses established that the relationship between factors to the satisfaction of young customers will be stronger than that of middle-aged customers.

The results of the multi-group variable model test show that the theoretical model has 382 degrees of freedom. The SEM results show that the model achieves compatibility with market data: Chi-square = 888.106; P-value = 0.000; Chi-square/df = 2.325(<2.5); TLI = 0.938 (>=0.9); CFI = 0.949 (>=0.9); RMSEA = 0.041 (<0.08).

The results of the multi-group invariance model test show that the theoretical model has 386 degrees of freedom. The SEM results show that the model achieves compatibility with market data: Chi-square = 893.118; P-value = 0.000; Chi-square/df = 2.314; TLI = 0.939 (>=0.9); CFI = 0.949 (>=0.9); RMSEA = 0.041 (<0.08).

Network

Reliability

P1''

P6''

Year old

P7''

P2''

ATM card service customer satisfaction

Empathy

Price of perceived service

Figure 4.6 Multi-group research model with age control variable

The results of the group variable and partial invariance tests for the two groups of young and middle-aged people are presented in Table 4.16. Because the difference between the two models is not significant (p=0.286>0.05), we choose the partial invariance model. This means that age does not change the relationships between service quality factors and customer satisfaction. Therefore, hypotheses P1'', P2'', P6'', P7'' are all rejected.

Table 4.16 Chi-square test between age-variable and age-invariant models

Model

CMIN	df	P	NFI	RFI	IFI	TLI
Variable	888,106	382	0.000	0.914	0.896	0.949	0.938
Invariant (partial)	893,118	386	0.000	0.914	0.897	0.949	0.939
Differentiation value	5,012	4	0.286	0.000	0.001	0.000	0.001

P-value= Chidist(5,012;4) = 0.286

4.3.4.4 Testing the secondary hypothesis of income differences

Similarly, the income research sample was divided into 2 groups: the group of customers with low average income (under 5 million VND/month) with 458 customers and customers with high income (from 5 million VND/month or more) with 321 customers. There are 4 hypotheses established that the relationship between factors to the satisfaction of customers with low average income will be stronger than that of customers with high income.

Network

Reliability

P1'''

P6'''

Income

P7'''

ATM card service customer satisfaction

P2'''

Empathy

Price of perceived service

Figure 4.7 Multi-group research model with income control variable

The results of the group variable and partial invariance tests for the two low- and high-income groups are presented in Table 4.17. Because the difference between the two models is not significant (p=0.717>0.05), we choose the partial invariance model. This means that income does not change the relationships between the service quality factors and customer satisfaction. Therefore, hypotheses P1'",P2''',P6''',P7''' are all rejected.

Table 4.17 Chi-square test between variable and constant models of income

Model

CMIN	df	P	NFI	RFI	IFI	TLI
Variable	877,146	382	0.000	0.917	0.900	0.952	0.941
Invariant (partial)	879,264	386	0.000	0.917	0.901	0.952	0.942
Differentiation value	2,118	4	0.714	0.000	0.001	0.000	0.001

P-value= Chidist(2,118;4) = 0.714

4.4 DISCUSSION

 Perceived price :

Based on the SEM results (Figure 4.4) of the theoretical model after adjusting (standardizing) considering the standardized Beta weight, we see that this factor has the strongest impact on customer satisfaction because it has the largest Beta coefficient (with β = 0.317 at the significance level Sig = 0.000). This means that under the condition that other factors remain unchanged, the issuance fee, annual fee, cash withdrawal fee, transfer fee, savings interest rate or non-term interest rate for payment accounts, loan interest rate increases (or decreases) by 1 unit, customer satisfaction also increases (or decreases) by 0.317 times. Thus, customers using banking services consider the interest rate factor (perceived price) as the most important factor affecting customer satisfaction.

Network:

Next is the network, this factor has the second influence on customer satisfaction in the research model with a standardized Beta coefficient (β = 0.235) at the significance level Sig = 0.000. Thus, under the condition that other factors remain unchanged, when customers feel that the bank's ATM network increases (or decreases) by 1 unit, customer satisfaction also increases by 0.235 and vice versa.

Empathy:

Next is empathy, this factor has the third influence on customer satisfaction in the research model with a standardized Beta coefficient (β = 0.139) at the significance level Sig = 0.000. Thus, under the condition that other factors remain unchanged, when customers feel that the bank's empathy increases (or decreases) by 1 unit, the customer's SHL also increases (or decreases) by 0.139 times and vice versa.

Comment