The measures to assess the construct’s reliability are shown in Table 32. All loadings of the items turned out to be high (.73 to .88) but did not show extreme values over .90. All t-values associated with each of the loadings exceeded the critical values for .01 significance, implying that the included items are reliable indicators of their hypothesized constructs. This is also supported by data from the validation sample, as all items also achieved similar loadings (ranging from .66 to .90) and significance levels. The items also demonstrated notable corrected item-to-total correlations, varying from .62 (.56) to .84 (.85). Further evidence for the reliability on the item level was provided by the items’ squared multiple correlation (R^{2}), which were all equal to or greater than the recommended cut-off value of .50 (Fornell & Larcker, 1981).

To assess reliability of the constructs on the first-order level Cronbach’s alpha, construct reliability, the average inter-item-correlation and the average variance extracted were evaluated. All first-order constructs evidenced remarkable alpha levels: ability: .92 (.91), integrity/honesty: .94 (.91), benevolence: .81 (.79), willingness to rely: .92 (.93), and willingness to depend: .83 (.82), meaning that the internal consistency of almost all constructs (except benevolence) was above the more restrictive alpha level of .80 proposed for newly developed constructs (Netemeyer et al., 2003). Construct reliabilities ranging from .81 (.81) to .93 (.94), as well as average item-to-total correlations varying from .59 (.55) to .74 (.77), provided additional evidence for the reliability of the five constructs. The AVE were all equal to or greater than .59 (.56), hence the majority of the sub-dimensions variance was explained by the higher-order construct. Every first-order construct was reasonably and significantly related with the second-order construct. Here, factor loadings fluctuated between .70 (.67) and .93 (.87). The squared multiple correlations were also mainly the limit of .50. Only the benevolence dimension showed a lower level in the development sample (.34), but only a slightly lower level in the validation sample (.45). On the second-order level, the reliability of the first-order indicators was also assessed by the index of construct reliability and the AVE. The internal consistency measures exceeded the .80 threshold (Netemeyer et al., 2003) with values of .90 (.88) and also the .50 cut-off value (Fornell & Larcker, 1981) (.65/.61). This implied that the first-order constructs are commonly reliable measures of the higher-order construct.

As presented in Table 32, factor loadings of all items turned out be greater than .71 (.66), and every one was statistically significant at the .01 significance level. A similar pattern was observable for the relationship between the first-order and the second-order constructs, supporting on both levels evidence for convergent validity (Bagozzi & Yi, 1991). More stringent evidence for convergent validity is provided where the squared factor loading between the item and its hypothesized sub-construct and/or the loadings between the sub-construct and the higher-order construct is .5 or better. On the item level, the lowest squared loading (.50) was observed for Wi2. The remaining items, however, showed loadings typically in the .70 range. In the validation sample, the same item (besides In3) also achieved the lowest loading but which was slightly above the one in the development sample (.53) encouraging the item’s validity. On the first-order level, almost all constructs turned out to have a squared factor loading equal to or greater than .50. However, benevolence showed a slightly lower loading (.45) only in the validation sample. Second, validity on the first-factor level was supported by significant correlations among the five constructs (see Table 32), implying that they all measure the same construct.

Table 32: Psychometric Properties of the eWT-S (Modified Scale, Samples 4 and 5)

Notes: Results for cross-validation sample 5 in parentheses; у = Completely standardized second-order loading; X = Completely standardized first-order loading; AYE = Average variance extracted.