Page 106 - Contributed Paper Session (CPS) - Volume 8

CPS2205 Abdul Aziz A. Rahaman et al.
From the observed significance levels (p<0.05) in Table 2 below, it can be
seen that four factors out of the six service quality dimensions were statistically
significant in influencing a customer retention. These dimensions include;
tangibility, responsiveness, empathy and trust. Meanwhile, customers who
agree to tangibility are more likely to assign higher ratings on loyalty than
their counterparts who do not disagree. Also, universal banks customers who
agree on the bank’s responsiveness are more likely to assign higher ratings for
loyalty than customers who think otherwise. Interestingly, customers who
disagree on the dimension of empathy are more likely to assign high ratings
for loyalty than those who agree. Moreover, customers who agree on the
dimension of trust in the bank are more likely to assign higher ratings for
loyalty than their counterparts who just disagree.
However, service quality dimensions including reliability and assurance
were each not statistically significant. This means that each of these service
quality dimension marginally influence customer retention. To have a more
rigorous interpretation for the customer retention with the mediation of
customer Satisfaction, the Goodness of fit indices need to be assessed. Also
the GFI = 0.963, NFI = 0.934, CFI = 0.941, and IFI = 0.941. All the incremental
fit measures fulfil the cut-off values (suggested values). Therefore, the model
2
can be said to be a good fit model. However, the statistic of 788.084 (df=39)
is large. The statistic for model fit is still significant, meaning that the null
2
hypothesis of a good fit to the data can be rejected. This could be due the
2
large sample size used here since the test is widely recognized to be
problematic. It is sensitive to sample size, and it becomes more and more
difficult to retain the null as the number of cases increases, which may lead to
the rejection of a good model or the retention of bad ones. The RMSEA
likewise suggests that the fit of the model is just about tolerable. The value of
0.083 exceeds the 0.05 cut-off value for accepting the model fit.

Table 3. Model Fit
Model -2LogLikelihood Chi-Square df P-value

Intercept Only 243.057
Final 125.468 117.589 24 0.000

From Table 3 above, it can be noted that the difference between the two
log-likelihoods with Chi-square distribution has a p-value less than the
significance level 0.05 (i.e. p<0.05). This indicates that there is sufficient basis
to reject the null hypothesis and therefore conclude that the final model gives
a significant improvement over the baseline intercept-only model. Hence the

95 | I S I W S C 2 0 1 9