Exploratory factor analysis (EFA)
In order to explore the underlying dimensions of consumer perceptions of service quality vis-à-vis life insurance sector (as expressed by performance scores on 26 statements), EFA was performed. The factor analysis results are shown in Tables 4 and 5. First, appropriateness of factor analysis for the data has been checked by Bartlett's test of sphericity and Kaiser-Meyer-Olkin (KMO) measure of sample adequacy. Bartlett's test is used to examine the hypothesis that the population correlation matrix is an identity matrix, that is the variables are uncorrelated in the population. KMO measure indicates the proportion of variance in the variables, which is common variance, that is, which might be caused by underlying factors. The results from Table 4 shows that value of KMO statistic is very high (0.955) and Bartlett's test of sphericity is significant (sig=0.000), which reveals that data are appropriate for factor analysis. The total variance accounted for by all of the six components explains nearly 78.3 per cent of the variability in the original 26 variables (Table 5). So, we can reduce the original dataset by using these six components (Eigen values greater than 1 as shown in Table 5) with only 21.7 per cent loss of information.
The Rotated Component Matrix reveals six factors (which represent the six broad perceptual dimensions of service quality termed as assurance, competence, personalized financial planning, corporate image, tangibles and technology) derived from 26 variables (which represent the consumer perceptions of life insurance policyholders vis-à-vis service quality). To confirm the six-dimensional structure of customer perceived service quality in life insurance sector (as obtained by EFA); we have used confirmatory factor analysis (CFA).
Confirmatory factor analysis
Factor analysis is primarily an exploratory technique because of researcher's limited control over which variables are indicators of which latent construct. Structural Equation Modeling (SEM), however, can play a confirmatory role because the researcher has complete control over the specification of indicators for each construct. So, CFA provides enhanced control for assessing unidimensionality and has more construct validity than EFA.45 CFA is used here, particularly for validation of scales for the measurement of constructs derived from EFA. ‘LISREL 8.8 for Windows’ software was used for this purpose.
The estimates of measurement model and the construct loadings, as provided by the statistical software package LISREL 8.8, are presented in Table 6. As there is no offending estimate in the construct loadings, various goodness-of-fit criteria have been assessed.
Overall model fit
The first assessment of goodness of fit for the model is done for the overall model (Table 7). It provides the degree to which the specified indicators (variables) represent the hypothesized constructs (dimensions). The three useful types overall model fit measures are: absolute, incremental and parsimonious fit measures.
Absolute fit measures
Value of Goodness-of-Fit Index (GFI) is 0.937, which is higher than the recommended value of 0.90 and Adjusted Goodness-of-Fit Index (AGFI) is 0.879, which is though lower than the recommended level (0.90) but close to it. Value of Root Mean Square Residual (RMSR=0.082) is quite low. Therefore, all the measures provide acceptable support for acceptability of overall model.
Incremental fit measures
These measures assess the incremental fit of the model compared to a null model. Null model is hypothesized as a single factor model with no measurement error. Here, both Tucker-Lewis Index (TLI) and Normed Fit Index (NFI) are higher than the recommended level of 0.90.
Parsimonious fit measures
This is the final measure, which assess the parsimony of the proposed model. It evaluates the fit of the model versus the number of estimated coefficients needed to achieve that level of fit. AGFI (0.879) is close to the recommended level of 0.90 and Normed Chi-square (1.82) is within the recommended range of 1.0 to 2.0. These results provide support to model parsimony.
Therefore, we can say that all the measures of overall model goodness of fit provide the acceptability to the proposed model.
Measurement model fit
After analyzing and accepting the goodness of fit for the overall model, all the six constructs (dimensions) are evaluated at two levels:
examining the variable loadings for statistical significance;
assessing the dimension's reliability, variance extracted, unidimensionality and convergent validity.
Results from Table 6 indicate that all the variables are significant (sig <0.05); as the t-values associated with each of variable loadings exceed the critical value for 5 per cent level of significance. Thus, it can be said that all the variables are significantly related to their specified dimensions. This substantiates the proposed relationship among variables and their dimensions.
In the second step, the estimates of the reliability, the variance-extracted measures and the various fit indices for each dimension (to assess the representativeness of each dimension) have been analyzed (Table 8). Results of Construct Reliability show that reliability coefficients of all dimensions exceed the recommended level of 0.700. Further, it has been found out that there is substantial extraction of variance (more than 50 per cent) for all the dimensions. Therefore, major portion of variances of variables have been accounted for by these considered dimensions, as obtained by CFA.
The Comparative Fit Index (CFI) values obtained for all the six dimensions of service quality are more than 0.90 (recommended level) as shown in Table 8. This indicates a strong evidence of unidimensionality, showing the strong representativeness of constructs. This also establishes the Construct Validity of the model. Construct Validity is the assessment of the degree to which an operationalization correctly measures its targeted variables.46
Bentler-Bonett coefficient47 is used to measure Convergent Validity of the model, which is the measure of the degree to which multiple methods of measuring a variable provide the same results.46 Results from the Table 8 show that all six dimensions have Bentler-Bonett Goodness of Coefficient more than 0.90, which is above the required level. This indicates substantial Convergent Validity. Finally, GFI for all of the six dimensions having values more than 0.9 (recommended level) shows best fit of the considered model.48
The overall model goodness-of-fit results and the measurement model fit results provide substantial support for confirmation to the proposed six-dimension model of consumer perceptions of service quality vis-à-vis life insurance sector.
Factors extraction results
As obtained in the Table 6, the CFA provided the construct loading for various factors (dimensions).
Factor 1 incorporates the variables – ‘Trained and well-informed agents’, ‘Approaching from customer's point of view’, ‘Trusting agents when explaining policies’, ‘Clarity in explaining policy's terms and conditions’ and ‘Understanding intimately specific needs’. These variables reflect that policyholders are assured that they are being dealt with representatives who are well versed with the nitty-gritty's of the service and intend to act as someone who would help them to identify their specific needs, and also provide solution accordingly. As all these variables assure the policyholder of knowledge of agents and their ability to inspire trust and confidence, this factor has been labeled as assurance.
Factor 2 consists of variables – ‘Provision of flexible payment schedule’, ‘Availability of flexible product solution’, ‘Provisions for convertibility of products’, ‘Supplementary services’. Life insurance involves long-term association, hence policyholder moves through different life cycle stages in this long period and his needs and preferences change accordingly. Here, all these variables are depicting handling of these changing preferences by providing flexible solutions and convertibility options and giving personalized services. Therefore, this factor can be labeled as personalized financial planning.
Factor 3 has variables – ‘Staff dependable in handling customer's problems’, ‘Efficient staff’, ‘Easy access to information’, ‘Prompt and efficient grievance handling mechanism’ and ‘Prompt and hassle free claims settlement’. These components talk about the ability of the service provider to perform service dependably and efficiently and also about their willingness to provide hassle-free and prompt services. Therefore, this factor can be labeled as competence.
Factor 4 incorporates variables – ‘Adequate no. of branches’, ‘Accessible location of the branch’, ‘Good ambience of the branch’, and ‘Possessing good certification and credentials’. As all these components are related to providing physical facilities and communication materials, therefore this factor has been labeled as tangibles.
Factor 5 has variables – ‘Innovativeness in introducing new products’, ‘Courteous agents’, ‘Value for money’, ‘Simple and less time consuming procedure for purchasing a policy’, and ‘Financially stable company’. All these components are related to creating an overall image of the organization in the eyes of the customers. Therefore, this factor can be labeled as corporate image.
Factor 6 incorporates variables – ‘Easy online transaction’, ‘Prompt complaint handling, online’ and ‘Proactive information through e-mail or SMS’. As all these components are related to use of modern aids in providing service, therefore this factor has been labeled as technology.
Structural equation model for customer satisfaction with service quality
SEM is a multivariate technique combining aspects of multiple regression and factor analysis to estimate a series of interrelated dependence relationships simultaneously. Here, SEM has been used to represent the multiple and interrelated dependence relationships among six dimensions of consumer perceived service quality (exogenous variables); three satisfaction components: ‘satisfaction with the life insurance agents’, ‘satisfaction with functional services provided by the life insurance sector’ and ‘satisfaction with the insurance company’ (endogenous variables); and finally with the ‘overall satisfaction with the life insurance services’ (endogenous variable). In the model, ‘overall satisfaction’ has been considered as a function of ‘satisfaction with contact person’ (agent), ‘core service’ (functional services) and ‘institution’ (company); as proposed by Gronroos49 and Lehtinen.50 The services literature49, 50 distinguishes between the quality produced as the customer interacts with the contact resources of the organization, what the customer actually gets as a result of interaction (core elements) and the image of the company.
SEM is useful for estimating the interdependent multiple relationships and representing the latent concepts (satisfaction) in these relationships. The prime objective is to understand and measure this complex relationship between service-quality dimensions, satisfaction dimensions and overall satisfaction with life insurance services.
Path diagram has been constructed by LISREL 8.8 to depict the series of causal relationship (Figure 1). This diagram represents predictive relationships between constructs and associative relationships among constructs. The path diagram forms the basis of path analysis for the SEM. It is the procedure for empirical estimation of the strength of each relationship (path) depicted in the path diagram. After developing the model and portraying it in path diagram, the model has been specified in a more formal terms (in the structural equations, and in the measurement model).
The desired structural equation model was estimated with LISREL 8.8 and expressed in Table 9. The corresponding measurement model specifies the correspondence of indicators to constructs at two stages. At the first stage of structural equation model, associative relationship of the six exogenous constructs (quality dimensions) with the three endogenous constructs (satisfaction dimensions) has been considered. At the second stage, it has been attempted to define endogenous construct (overall satisfaction) as the function of the three satisfaction dimensions.
Evaluating goodness-of-fit criteria
First, the inconsistency in the estimated path coefficients has been checked, in terms of offending estimates (coefficients exceeding the acceptable limits, that is standardized coefficients exceeding 1). Results from Table 9 clearly indicate the absence of any offending estimate.
Overall model fit
The overall fit of the structural equation model has been examined, at the first stage, to ensure that model is an adequate representation of the entire causal relationship (from six service quality dimensions to three satisfaction dimensions to the final overall satisfaction).
Absolute fit measures
The value of the likelihood Chi-square is 119.422 and is statistically significant (Table 10). Values of GFI and AGFI are both more than the required level of 90 per cent, both RMSR (RMSR=0.082) and Root Mean Square Error of Approximation (RMESA=0.071) are at moderate acceptable level of 80 per cent. These results clearly indicate that the considered model is substantially acceptable.
Incremental fit measures
At this stage, the considered model is evaluated in comparison to a null model. The value of Chi-square is 427.54, therefore there is substantial improvement in terms of reduction in the Chi-square value, because of the estimated coefficients in the proposed mode (Table 10). Results (Table 10) show that both TLI and NFI have values more than the required level of 0.90.
Parsimonious fit measures
Lastly parsimonious measures are considered. Results (Table 10) show that CFI (0.914) is above the threshold limit but both Parsimonious Normed Fit Index (PNFI=0.872) and Parsimonious Goodness-of-Fit Index (PGFI=0.822) are near to the desired threshold limit. So, the considered model can be considered as marginally acceptable.
Measurement and structural model fit
After ensuring the goodness of fit for overall model, measurement and structural model fit are examined. It has been found that all the construct loadings for the model are significant at 5 per cent level of significance (Table 9). The value of R-square for the entire model is obtained as 0.726, which is substantially high and significant (Table 10).
The overall model goodness-of-fit results and the measurement model fit results provide substantial support for the acceptance of the considered structural equation model for service quality and customer satisfaction in the life insurance industry.
Structural equation model for customer satisfaction
After examining the goodness of fit, reliability and validity measures, path coefficients of different causal relationships of the proposed customer satisfaction model have been considered. The results from the structural equation model (Table 9 and Figure 1) show that ‘satisfaction with life insurance agents’ can be significantly explained by ‘assurance’. ‘Personalized financial planning’ has significant impact upon ‘satisfaction with life insurance agents’ and ‘satisfaction with functional services provided’, but there is substantial difference in the strength of the relationship. ‘Personalized financial planning’ has a major impact upon ‘satisfaction with functional services’. ‘Competence’ significantly affects both ‘satisfaction with functional services’ and ‘satisfaction with the life insurance company’ with almost equal magnitude. ‘Corporate image’ has significant impact upon both ‘satisfaction with the company’ and ‘satisfaction with functional services’. While comparing the strength of relationships, we can say that ‘corporate image’ has higher impact upon ‘satisfaction with the company’ as compared to that on ‘satisfaction with functional services’. Further, as compared to ‘corporate image’, ‘competence’ has significantly higher impact upon ‘satisfaction with functional services’. ‘Tangibles’ significantly affects ‘satisfaction with the company’. Similarly, ‘technology’ has influence upon ‘satisfaction with functional services’. But both of these relationships are moderate in nature.
Therefore, results from the path analysis suggest that ‘assurance’ and ‘personalized financial planning’ are significant predictors of ‘satisfaction with the insurance agent’, who is the main channel of communication and business between customer and the insurance company. Here, ‘assurance’ is the major predictor of this satisfaction. Further, ‘personalized financial planning’, ‘technology’, ‘competence’ and ‘corporate image’ are significant predictors of ‘satisfaction with the functional services provided by the insurance company', which is the main driver of relationship between the customer and the company. Here, ‘competence’ and ‘personalized financial planning’ are the major predictors. Lastly, ‘competence’, ‘corporate image’ and ‘tangibles’ are the significant predictors of ‘satisfaction with the insurance company’, with ‘competence’ being the major predictor, followed by ‘corporate image’.
At the second stage, customers’ ‘overall satisfaction’ with the service quality in life insurance industry can be explained in terms of three satisfaction components (endogenous variables): ‘satisfaction with the life insurance agents’, ‘satisfaction with functional services provided by the life insurance sector’ and ‘satisfaction with the insurance company’. Here, the all three relationships are significant. ‘Overall satisfaction’ is best explained in terms of ‘satisfaction with functional services’ (0.883), closely followed by ‘satisfaction with agents’ (0.728). On the other hand, ‘satisfaction with the insurance company’ has much lower influence upon ‘overall satisfaction’ (0.402). So, it can be said that these three satisfaction components are significant determinants of overall customer satisfaction.
The relationship between the endogenous variables representing the three satisfaction components has also been examined. There is a strong and significant relationship between ‘satisfaction with agents’ and ‘satisfaction with functional services’ (0.808). Similarly, ‘satisfaction with financial services’ and ‘satisfaction with the insurance company’ are correlated with marginally lower value (0.711). Relationship between satisfaction with life insurance agents and satisfaction with life insurance companies is significant but has substantially lower magnitude (0.534).