SAMPLINGDemographyMeasurement modelBefore assessing structural models, it is vital to ensure that there are no full collinearity issues in the model. According to Kock and Lynn (2012), even if the criteria of discriminant validity (vertical collinearity) is met, lateral collinearity issues (e.g. predictor-criterion collinearity) may sometimes subtly mislead the findings as it is capable of weakening the strong causal effects demonstrated in the model. Table 2 presents the outcome of the full collinearity test. The VIF scores for each individual construct were below the threshold value of 5 (Hair, Ringle, & Sarstedt, 2011), thus implying that there are no issues on collinearity (Kock & Lynn, 2012) and common method bias in this study (Kock, 2015). Table 3: VIF scoresCOLLINEARITY STATS (VIF) INNER MODELFINDINGSfinalize the usability of the questionnaire (Hunt et al., 1982). Purposive sampling technique was adopted to distribute questionnaire to civil servants of Malaysia with grades 41 and above (management and professionals). In all, 1,000 copies were distributed, and 561 completed and usable copies were collected in two weeks in June 2023, indicating that non-response bias was not a major issue. The data were then keyed in into SPSS and imported to Smart PLS to perform latent variable analysis (Ringle et al., 2015). The latter software utilizes structural equation modelling of partial least squares (PLS-SEM) approach to enhance predictive relevance by maximizing the variance of key target variables by different explanatory variables (Hair et al., 2014; Henseler et al., 2009). Most of the respondents were aged between 40 to 54 (56.4%, n = 316), which is expected for a work and study sample. At an age between 40-54, respondents were said to be more enthusiastic about meaningful careers. There are slightly more males (50.4%, n = 282) than females in the sample. As for the race distribution, most of the respondents were Malays (81.4%, n = 456), followed by Others (4.1%, n = 41) and Malaysian-Chinese (6.1%, n = 34). In view that the sample were skewed towards matured workers (40-54 years of age), it was not surprising that most of them were matured in service (11 – 25 years (75.1%, n = 415), all demographics variables were included as control variables in the analyses (Table 4).CG -> OPEB -> OPTD -> OPTE -> OPEB x TE -> OPEB x CG -> OPEB x TD -> OPIn order to assess the moderating effect of a construct in PLS-SEM, interaction terms between the moderator and the predicting variables were created using the product indicator approach suggested in Chin et al. (2003), to examine its effect on the endogenous variable. Table 5 and Figure 2 reports the moderating effect of employer branding on the relationship of talent development, career development and talent engagement. The results indicate that three hypothesised moderation effects of employer branding were supported (H4 to H6). Specifically, the results indicated six significant interaction terms – – Talent 1133.483.0592.6092.92.8555.5354.338
元のページ ../index.html#113