Empirical analysis and findings

Validity and reliability

Standardised factor loadings, squared multiple correlations (SMC) and model fit indices are considered to be the key statistical criteria for an acceptable measurement model. Table 3 summarises the statistical criteria of the measurement model. R2 represents the squared multiple correlations and all the values are above the recommended level of 0.50. In addition, all standardised factor loadings are above the recommended level of 0.70 and statistically significant. Furthermore, a good measurement model fit is indicated as the Tucker Lewis Index (TLI) and Comparative Fit Index (CFI) are both well above the recommended level of 0.90.

Table 3 Summary Results of Measurement Model

Latent Construct Observed Indicators Unstandardized factor loadings Standardized factor loadings Standard error Z-value R2 (item reliability)
QPI QPI 0.23 1.00*** 0.01 24.61 1.00
LP LPIAT 0.17 0.93*** 0.01 24.62 0.86
LPICQ 0.18 0.97*** 0.01 29.75 0.94
LPIEA 0.12 0.85*** 0.01 19.40 0.73
LPIEC 0.20 0.95*** 0.01 27.93 0.90
LPIFS 0.13 0.87*** 0.01 22.55 0.76
LPIQT 0.22 0.97*** 0.01 29.42 0.95
ST CT 1.46 0.98*** 0.07 20.28 0.95
LSC 0.78 0.88*** 0.04 19.24 0.77
NE PGDP 0.99 1.00*** 0.04 23.16 1.00

Model-fit: χ2 (31) = 81.93, CFI = 0.98, TLI = 0.97, RMSEA = 0.09, SRMR = 0.02

*p < 0.05, **p < 0.01, ***p < 0.001

In addition to the model fit indices and SMC, we also checked the reliability of the latent constructs. Reliability is referred to by the value of Cronbach's Alpha. The Cronbach's Alpha values of the constructs LP and ST are 0.97 and 0.86, respectively; both values exceed the required level of 0.70 as suggested by Nunnally. This also confirms internal consistency of the latent constructs.

To test for convergent validity, we examined the statistical significance of the factor loadings through their z-values (also stated as t-values). As a rule of thumb, acceptable estimates should have z-values higher than 2 or less than − 2. As depicted in Table 3, all z-values of indicators are higher than 2, which means that all indicators measure their respective latent construct, and confirm the uni-dimensionality and convergent validity of each construct. As all the R2 values are above 0.50, item reliability is also confirmed. To assess discriminant validity, a series of pairwise confirmatory factor analyses (CFA) were conducted. In this process, a non-constrained CFA of one pair of constructs was compared with a constrained CFA at a time to avoid the influence of construct pairs with significant values over non-significant ones. All the chi-square difference test results were statistically significant (p < 0.001), providing evidence of the discriminant validity of the constructs.