Descriptive statistics
Respondents’ demographic information was analyzed in terms of 4 variables: gender, age, education and years of service (work experience). According to the descriptive analysis, 68.1 percent of respondents were women and 31.9 were men; 32.5 percent aged between 20 and 30, 38.6 percent between 31 and 40, 22.9 percent between 41 and 50, 4.8 percent between 51 and 60, and 1.2 percent above 61; 15.7 percent were high school graduates, 18.7 junior college graduates, 42.8 percent university undergraduates, 12.7 percent university graduates, and 10.2 percent Ph.D. graduates; 28.9 percent had less than 5 years of working experience, 21.7 percent between 5 and 10 years, 21.1 percent between 10 and 15 years, 12 percent between 15 and 20 years, 12.7 percent between 20 and 25 years, and 3.6 percent worked more than 25 years.
Table 2 presents Cronbach’s alpha, mean response and the respective standard deviation of each variable. Note that the Cronbach’s alpha for job satisfaction was initially .42, but after the exclusion of one item it rose to .82.
Insert Table 2
As it can be observed in the above table, the mean responses for all variables are in an appropriate mean, among which the highest amount belongs to ethical leadership.
As the presence of a pairwise linear correlation between variables is a necessary assumption in applying the latent variables method in structural equation modeling (SEM), first, for each pair of variables, Pearson correlation test was run and the result is presented in Table 3.
Insert Table 3
Note: ** [one-tailed] correlation at significance level of p < 0.01; * correlation at significance level of p < 0.05
As Table 2 shows, the strongest correlation is that of job satisfaction and SWB (r = 0.560), and the smallest correlation exists between SWB and ethical leadership (r = 0.103). In addition, all the estimated paired correlation coefficients are positive and significant.
Validity and reliability of measurement and structural model
The research model was analyzed by Smart PLS 3 employing structural equation modeling (SEM). The validity and reliability of the constructs was estimated using factor loadings, Cronbach's Alpha, composite Reliability average variance extracted (AVE) shown in Table 4.
Insert Table 4
As shown in Table 4, all factor loadings were more than 0.5, shown appropriate reliability. Cronbach’s α incidents were above 0.7 value showing satisfactory reliability. Moreover, the value of composite reliability and AVE were more than 0.7 and 0.5 respectively, showing satisfactory reliability [34].
To test the hypotheses, the partial least squares structural equation modeling (PLS-SEM) by Smart PLS 3 was employed. To test the fitness of structural model, R2 and Q2 measures were shown in Table 5.
Insert Table 5
As shown in Table 5, The first criterion for examining the structural model is the coefficient of determination R2 related to the endogenous (dependent) latent variables in the model and shows the effect of an exogenous variable on an endogenous variable. The strength of this effect interpreted with three values of 0.19, 0.33 and 0.67 as weak, medium and strong values [34]. Accordingly, the result shows that the model can predict 0.213 percent of job satisfaction changes, measured as a mediate effect. Moreover, 49 percent of subjective wellbeing changes predicted by the model, showed strong effects of exogenous variables of the model on subjective well-being. Q2 value determines the strength of the model in predicting dependent variables. Hair et al., (2014) considered three values of 0.02, 0.15 and 0.35 as low, medium and strong predictive strength [34]. As shown in Table 4, the value of Q2 for all dependent variables were moderate.
Test of hypotheses
To test the hypotheses, PLS-SEM by Smart PLS 3 was employed. Figure 2 shows the SME model in T-value mood:
Insert Figure 2
According to Figure 2, the relationship between variables are significant when the T-value was more than 1.96. As shown in Figure 2, all hypotheses of the research were accepted. The result of hypothesis tests is shown in Table 6.
Insert Table 6
According to Table 4, all hypothesis was supported. Thus, the impact of ethical leadership (β= 0.155, T-value= 2.420) and job satisfaction (β= 0.619 T-value= 11.338) on subjective wellbeing are significant. Moreover, the effect of ethical leadership on job satisfaction is supported (β= 0.462 T-value= 7.445).
Testing mediation effects
To measure mediation effect of job satisfaction in the relationship between ethical leadership and subjective wellbeing, the indirect effect shown in Table 7.
Insert Table 7
As shown in Table 4, the indirect effect of ethical leadership on subjective wellbeing through job satisfaction is confirmed (β= 0.286, T-value= 7.160), shown the mediation effect of job satisfaction. The result also shows that the indirect effect of ethical leadership on subjective wellbeing through job satisfaction is more than the direct effect of ethical leadership on subjective wellbeing (β= 0.155, T-value= 2.420). Accordingly, the total effect (direct effect* indirect effect) of ethical leadership on subjective wellbeing is 0.443.