Factor analysis was conducted on the items comprising the service quality construct. A total of 22 service quality items were subject to the analysis. Table 5.21 reveals that the KMO measure of sampling adequacy is 0.925, with the Bartlett’s test rendering a significant result (p=0.000), which statistics indicate that a factor analysis will, therefore, be appropriate.
Table 5.21: KMO and Bartlett’s Test for Service Quality Factors
Kaiser-Meyer-Olkin Measure of Sampling Adequacy. .925 Bartlett's Test of Sphericity Approx. Chi-Square 3153.652
Df 231
Sig. .000
Table 5.22 reveals the outcome of Principal Axis Factoring using Varimax rotation, which procedure was used to extract and rotate the factors resulting in five factors being extracted.
Cumulatively, these factors contributed 50.198% to the total variance. It is also evident from Table 5.22 that the first factor contributed 13.852%, the second factor contributed 10.372%, the third
164 factor contributed 10.251%, the fourth factor 8.360% and the fifth factor 7.363%, to the total variance.
Based on the rotated factor matrix depicted in Table 5.23, Factor 1 loaded strongly on a combination of two service quality dimensions, Assurance and Empathy with eight items or variables. However, variable loadings pertaining to Empathy were higher. Factor 1 can therefore be called to ‘Empathize and Assure’.
Factor 2 had three items/variables, which loaded strongly on issues pertaining to Tangibles, and is called ‘Tangibles’. Factor 3 loaded strongly on four Reliability items and is called ‘Reliability’.
Factors 4 and 5 loaded heavily on the “Responsive-related” dimension, with only one Reliability- related item/variable included. Therefore these factors combined are called ‘Promptness and Accuracy’ (Factor 4) and ‘Helpfulness’ (Factor 5).
Question A4 loaded very weakly and was therefore not included into a particular factor. Question A14 loaded onto two factors, indicating a lack of convergent and discriminant validity and was not included in any factor.
165 Table 5.22: Total Variance Explained for Service Quality
Factor
Initial Eigenvalues
Extraction Sums of Squared Loadings
Rotation Sums of Squared Loadings
Total
% of Variance
Cumulative
% Total
% of Variance
Cumulative
% Total
% of Variance
Cumulative
%
1 8.895 40.430 40.430 8.431 38.323 38.323 3.048 13.852 13.852
2 1.455 6.614 47.044 .931 4.230 42.554 2.282 10.372 24.224
3 1.181 5.370 52.414 .689 3.133 45.686 2.255 10.251 34.475
4 .992 4.509 56.923 .532 2.416 48.103 1.839 8.360 42.835
5 .956 4.346 61.269 .461 2.095 50.198 1.620 7.363 50.198
Extraction Method: Principal Axis Factoring.
Table 5.23: Rotated Factor Matrix Service Quality
Service Quality Variables Factor
1 2 3 4 5
A1 State of Equipment .727
A2 Visual Appeal - Physical Facilities .657
A3 Appearance of Employees .427
A4 Visual Appeal of Materials
166
A5 Keeping Promises .644
A6 Sympathetic to Solving Student Problems .451
A7 Providing Service Right First Time .521
A8 Providing Service at Promised Time .722
A9 Keeping Accurate Records .469
A10 Informing Students of when Service will be Performed .648
A11 Promptness of Service .473
A12 Willingness to Help .552
A13 Employees never too Busy to Help .700
A14 Confidence instilled by Employees .430 .442
A15 Feeling Safe in Transacting with Institution .489
A16 Courteous Employees .518
A17 Employee Knowledge in Answering Questions .507
Q18 Providing Individual Attention .400
Q19 Convenience of Operating Hours .536
Q20 Personal Attention Provided by Employees .688
Q21 Institution Having My Best Interests .535
Q22 Employees Understanding My Specific Needs .555
Extraction Method: Principal Axis Factoring.
Rotation Method: Varimax with Kaiser Normalization.
167
Rotation converged in 6 iterations.
Based on the factor analysis, a Confirmatory Factor Analysis (CFA) using AMOS version 23 was conducted for the service quality construct, which is based on the SERVPERF model with 22 items/variables. The CFA revealed a five-factor service quality model as depicted in Figure 5.7 hereunder.
Figure 5.7: Measurement Model for Service Quality
168 Figure 5.7 represents the service quality model based on SERVPERF dimensions through conducting a CFA where Empathy=Empathy, TANG=Tangibles, RELIA=Reliability, RESP=Responsiveness, and HELP=HELPFULNESS. Reliability analysis (Table 5.24) revealed the Cronbach Alpha scores for each dimension (above 0.7) that were deemed reliable (Andrew et al., 2011:202).
Table 5.24: Reliability Scores for SERVPERF Construct Dimensions Confirmed by CFA
Variable Cronbach’s Alpha
Empathy 0.836
Tangibles (TANG) 0.737
Reliability (RELIA) 0.805
Responsiveness (RESP) 0.715
HELPFULNESS (HELP) 0.734
The model fit indices for the Service Quality model appear in Table 5.25.
Table 5.25: Model Fit Indices for SERVPERF Dimensions (Service Quality Construct)
Measure Threshold Indices for Model Comment
Chi-square/df (cmin/df) <3 good, <5 sometimes allowed (Hu & Bentler, 1999).
2.250 ACCEPTABLE
p- value >0.05 (Hu & Bentler). 0.00 NOT ACCEPTABLE
CFI >0.9 (Hu & Bentler, 1999).
0.965 ACCEPTABLE
169
Measure Threshold Indices for Model Comment
GFI 0.9 minimum (Hu &
Bentler, 1999).
0.951 ACCEPTABLE
AGFI Equal to or >0.9 (Hooper et al., 2008 cited in Kats, 2013:103).
.923 ACCEPTABLE
NFI >0.9 (Bentler, 1995). 0.939 ACCEPTABLE
RMSEA <0.06 (Hu & Bentler, 1999).
0.056 ACCEPTABLE
PCLOSE >0.05 (Hu & Bentler, 1999.
0.196 ACCEPTABLE
Although the p-value of the Service Quality (SERVPERF-based) model is not in accordance with the recommended threshold, however based on the values of the other fit indices, which are in accordance with the recommended thresholds, it is inferred that the model is a good fit.
To determine whether the factors identified in the model display convergent and discriminant validity, appropriate analyses were conducted. According to Esposito (2010:696) convergent validity exists when AVE is greater than 0.5. In addition, when MSV is less than AVE and ASV is less than AVE, discriminant validity can be claimed (Hair et al., 2009 cited in Ernst, 2015:38).
The Table 5.26 was generated by a template put together by Professor Gaskin to test for convergent and discriminant validity in confirmatory factor analysis (Statwiki, n.d.). From the table, AVE values for each factor in the model is greater than 0.5 and hence convergent validity can be claimed for each factor (Esposito, 2010:696). Furthermore, for each factor, the MSV and ASV values are less than AVE and hence discriminant validity can be claimed (Hair et al., 2009 cited in Ernst, 2015:38).
170 Table 5.26: Convergent and Discriminant Validity Indices
CR AVE MSV ASV RESP EMP TANG RELIA HELP
RESP 0.725 0.572 0.546 0.461 0.756
EMP 0.841 0.516 0.450 0.374 0.671 0.718
TANG 0.737 0.583 0.365 0.288 0.586 0.497 0.764 RELIA 0.814 0.596 0.503 0.409 0.709 0.606 0.604 0.772 HELP 0.746 0.599 0.546 0.394 0.739 0.658 0.442 0.634 0.774
Please note, for clarification pertaining to the table, RESP=Responsiveness, EMP=Empathy, TANG=Tangibles, RELIA=Reliability and HELP=HELPFULNESS.
The data in the model was tested for normality. According to Kline (2005), cited in Harrington (2009), it is suggested that variables having absolute values of greater than 3 for the skew index and absolute values of greater than 10 for the kurtosis index indicate normality problems. The skewness and kurtosis values in the dataset for each variable is well within the specified range and hence the assumption of normality is met.
Table 5.27: Assessment of Normality
Variable min max skew kurtosis A12_1 1.000 7.000 -.669 -.271 A13_1 1.000 7.000 -.415 -.722 A10_1 1.000 7.000 -.993 .210 A11_1 1.000 7.000 -.514 -.222
171 Variable min max skew kurtosis
A5_1 1.000 7.000 -.516 -.643 A7_1 1.000 7.000 -.401 -.561 A8_1 1.000 7.000 -.373 -.644 A1_1 1.000 7.000 -.293 -.159 A2_1 1.000 7.000 -.275 -.216 Q18_1 1.000 7.000 -.404 -.742 Q19_1 1.000 7.000 -.758 -.232 Q20_1 1.000 7.000 -.330 -.598 Q21_1 1.000 7.000 -.436 -.503 Q22_1 1.000 7.000 -.054 -.699
Multivariate 95.494
Source: AMOS 23 output generated by Researcher.
As can be seen therefore, the service quality model, based on the analysis conducted comprises of five dimensions – Empathy, Tangibles, Reliability, Responsiveness and Helpfulness, which is applicable to the higher education institutions in the sample studied. The Assurance dimension is not applicable to the sample in this study and is not an underlying dimension of service quality because it did not load into the model. In addition, of the 22 SERVPERF items or variables used in this study, only 14 were relevant. This means that the other eight variables are not relevant to Service Quality (SERVPERF-Dimensions) for the sample studied. However, the Empathy dimension of service quality was the only dimension for which all five SERVPERF variables
172 used to measure Empathy, were relevant. Therefore, Empathy is a key underlying factor or dimension of Service Quality in this study.
The Factor analysis and Confirmatory Factor analysis for the Brand Equity construct is discussed in the next section.