Measurement model and validation

To assess the probability of common method bias, we applied Harman’s single-factor test. It showed that the maximal variance of the model explained by one variable was 38.6%. Hence, the explained variance fell below the upper limit of 50% and indicated no major threat to common method bias (Podsakoff et al. 2003). Additionally, the applied marker variable showed no to minor correlations with the other variables, strengthening the case against common method bias (Semin et al., 2005; Simmering et al. 2014).

We then evaluated the convergent validity of the measurement model using the factor loadings (λ), Cronbach’s α, composite reliability (CR), and average variance extracted (AVE). The loadings reflect how far the manifest variables are deemed a part of the latent variable. Loadings for each item should be at least .5 to signify indicator reliability (Chin 1998; Hair et al. 2010). Table ​Table11 shows that the loadings were above the recommended threshold. The Cronbach’s α and CR are indicators for the construct reliability. The lowest Cronbach’s α was .76, and the lowest CR was .77. Thus, both exceeded the recommended threshold of .7 (Fornell and Larcker 1981; Kline 2015). The AVE indicates convergent validity and reveals how far one latent construct explains the associated indicators. The AVE did not fall below the recommended threshold of .5 for any of the constructs (Fornell and Larcker 1981).

Validity and reliability for constructs

¹Response format: Items were rated on a 7-point Likert scale (strongly disagree – strongly agree)

^RInversely worded items were recoded prior to the analyses

Subsequently, we assessed the discriminant validity. Discriminant validity is given when one variable that is theoretically unrelated to another is also statistically unrelated. In this case, discriminant validity can be measured using the Fornell-Larcker criterion. Discriminant validity exists when the square root of the AVE of a latent construct (bold diagonal in Table ​Table2)2) exceeds its correlation with the other latent constructs within the model, as this indicates that the modeled constructs can be reliably separated (Fornell and Larcker 1981).

Discriminant validity

*p < .05 level (2-tailed). **p < .01 level (2-tailed). Values on the diagonal (bold) are the square root of the average variance extracted (AVE), while the off diagonals show Pearson’s product-moment correlation coefficients

Our data revealed discriminant validity as the square root of each AVE was always higher than the correlation between the respective construct with the other ones. To strengthen the robustness of the discriminant validity, we further conducted the heterotrait-monotrait (HTMT) method. The results are summarized in Table ​Table3.3. The values of the HTMT do not exceed the threshold of .85 and thus meet the recommended criteria (Henseler et al. 2015).

Heterotrait-monotrait

Shaded boxes are the standard reporting format for the HTMT procedure

Overall, the measurement model showed reliable and valid results. Hence, we continued with the second stage of the structural equation model analyses. We computed fit indices, analyzed regression and determination coefficients, effect sizes, and presented indirect and total effects.