3.2. Intervention Effects on PsyCap
To test the first hypothesis, which proposed that the intervention would effectively improve the PsyCap level, we performed a mixed between-within ANOVA. The results revealed a significant interaction effect for groups by time for PsyCap,
F = 3.77,
df = 3.439,
p = 0.009, η
p2 = 0.069. As a post hoc analysis, we conducted MANOVA to test whether the three groups differed significantly regarding their means at T1, T2, and T3 for PsyCap. The results indicated no significant differences between the groups (
Table 2).
In addition, we conducted ANCOVA for a more rigorous test for mean differences. Specifically, PsyCap data at T2 and T3 were compared among the experimental, placebo, and control groups, controlling for PsyCap at T1. In addition to controlling for the effect of PsyCap at T1, we also included the covariates of age, gender, job tenure, education, and marriage. The results (
Table 3) suggested that the group variable was a significant predictor of PsyCap at T2 and T3 (
p < 0.05), whereas age, gender, job tenure, education, and marriage were not (
p > 0.05).
We also analyzed whether the means in the experimental group were significantly distinct from T1 to T2 and from T1 to T3, expecting an increase across this time frame. Pairwise comparison showed a significant promotion in PsyCap over time in the experimental group (
Figure 3): the means changed significantly from T1 to T2 (Δ(T2–T1) = 0.198,
p = 0.001,
d = 0.501, 95% CI [0.044, 0.958]) and from T1 to T3 (Δ(T3–T1) = 0.224,
p = 0.000,
d = 0.557, 95% CI [0.098, 1.015]). For the placebo group, pairwise comparison showed no significant distinctions from T1 to T2 (Δ(T2–T1) = 0.043,
p = 1.000,
d = 0.136, 95% CI [−0.363, 0.634]) or from T1 to T3 (Δ(T3–T1) = 0.03,
p = 1.000,
d = 0.098, 95% CI [−0.4, 0.596]). For the control group, pairwise comparison also showed no significant differences from T1 to T2 (Δ(T2–T1) = −0.043,
p = 1.000,
d = −0.143, 95% CI [−0.613, 0.326]) or from T1 to T3 (Δ(T3–T1) = 0,
p = 1.000,
d = 0, 95% CI [−0.469, 0.469]). Thus, only the experimental group saw a significant promotion in PsyCap level, which supported Hypothesis 1.
3.3. Intervention Effects on Work-Related Attitudes
For job satisfaction, the mixed between-within ANOVA showed no significant interaction effect of group by time, F = 2.015, df = 3.643, p = 0.101, ηp2 = 0.038. However, the main effect of time was significant, F = 4.571, df = 1.821, p = 0.014, ηp2 = 0.043, whereas the main effect of group was not significant, F = 0.242, df = 2, p = 0.786, ηp2 = 0.005.
Furthermore, we conducted ANCOVA for a more rigorous test for mean differences of job satisfaction. The results (
Table 3) suggested that the group variable was not a significant predictor of job satisfaction at T2 (
p > 0.05) but was a marginally significant one at T3 (
p = 0.050), whereas age, gender, job tenure, education, and marriage were not significant at either interval (
p > 0.05).
In addition, pairwise comparison showed a significant promotion in job satisfaction over time in the experimental group (
Figure 3): the means changed significantly from T1 to T2 (Δ(T2–T1) = 0.228,
p = 0.002,
d = 0.559, 95% CI (0.1, 1.017)) and from T1 to T3 (Δ(T3–T1) = 0.272,
p = 0.001,
d = 0.634, 95% CI (0.173, 1.095)). For the placebo group, pairwise comparison showed no significant differences from T1 to T2 (Δ(T2–T1) = 0.038,
p = 0.642,
d = 0.092, 95% CI (−0.406, 0.59)) or between T1 and T3 (Δ(T3–T1) = 0.070,
p = 0.424,
d = 0.15, 95% CI (−0.349, 0.648)). For the control group, pairwise comparison showed no significant differences from T1 to T2 (Δ(T2–T1) = 0.048,
p = 0.531,
d = 0.099, 95% CI (−0.369, 0.568)) or from T1 to T3 (Δ(T3–T1) = 0.014,
p = 0.082,
d = 0.037, 95% CI (−0.431, 0.506)). Thus, the experimental group saw a significant promotion of job satisfaction level, which supported Hypothesis 2.
For turnover intention, the mixed between-within ANOVA showed a marginally significant interaction effect of group by time, F = 2.476, df = 3.607, p = 0.052, ηp2 = 0.047. However, neither the main effect of time, F = 0.842, df = 1.804, p = 0.422, ηp2 = 0.008, nor the main effect of group, F = 0.318, df = 2, p = 0.728, ηp2 = 0.006, was significant.
Nevertheless, the mixed between-within ANOVA revealed a significant interaction effect of group at two time points (T1 and T2),
F = 4.022,
df = 2,
p = 0.021, η
p2 = 0.074, which indicated that the effect of the PsyCap intervention on turnover was significant at T2, but the significant effect did not persist to T3. We conducted ANCOVA for a more rigorous test for mean differences of turnover intention; the results shown in
Table 3 suggested that the group variable was a significant indicator of turnover intention at T2 (
p < 0.05) but not at T3 (
p > 0.05), whereas age, gender, job tenure, education levels, and marriage status were not significant at either interval (
p > 0.05).
We also analyzed whether the means of turnover intention in the experimental group were significantly changed from T1 to T2, indicating a decrease across this time frame. However, pairwise comparison showed no significant reduction in turnover intention in any of the groups between T1 and T2: the experimental group, Δ(T2–T1) = 0.200,
p = 0.094,
d = −0.273, 95% CI (−0.724, 0.179) (
Figure 3); the placebo group, Δ(T2–T1) = −0.213,
p = 0.107,
d = 0.261, 95% CI (−0.239, 0.761); or control group, Δ(T2–T1) = −0.229,
p = 0.067,
d = 0.304, 95% CI (−0.167, 0.776). Although the experimental group did not exhibit a significant reduction in turnover intention level, it appeared that the intervention did prevent participants from increasing turnover intention, which was observed in both the placebo and control groups. Thus, Hypothesis 3 was partially supported.
For job embeddedness, the mixed between-within ANOVA showed a marginally significant interaction effect of group by time,
F = 2.422,
df = 4,
p = 0.050, η
p2 = 0.046. However, neither the main effect of time,
F = 1.794,
df = 2,
p = 0.169, η
p2 = 0.017, nor main effect of group,
F = 0.067,
df = 2,
p = 0.936, η
p2 = 0.001, was significant. Additionally, the ANCOVA results (
Table 3) suggested that the group variable was not a significant predictor of job embeddedness at T2 (
p > 0.05) but was a marginally significant one at T3 (
p = 0.053), whereas age, gender, job tenure, education, and marriage were not significant at either interval (
p > 0.05).
In addition, despite expecting an increase over time, pairwise comparison of the means of job embeddedness from T1 to T2 and from T1 to T3 showed no significant promotion in job embeddedness over time in the experimental group (
Figure 3): the means did not differ significantly from T1 to T2 (Δ(T2–T1) = 00.034,
p = 0.572,
d = 0.167, 95% CI (−0.284, 0.617)) or from T1 to T3 (Δ(T3–T1) = 00.023,
p = 0.712,
d = 0.109, 95% CI (−0.341, 0.559)). However, for the placebo group, the pairwise comparison showed significant changes from T1 to T2 (Δ(T2–T1) = 0.152,
p = 0.023,
d = 0.327, 95% CI (−0.174, 0.828)) and from T1 to T3 (Δ(T3–T1) = 00.157,
p = 0.027,
d = 0.305, 95% CI (−0.196, 0.806)). For the control group, the pairwise comparison showed no significant differences from T1 to T2 (Δ(T2–T1) = 0.016,
p = 0.794,
d = 0.058, 95% CI (−0.41, 0.527)) or from T1 to T3 (Δ(T3–T1) = −0.118,
p = 0.075,
d = −0.281, 95% CI (−0.752, 0.19)). Thus, the experimental group did not see a significant improvement in job embeddedness. Thus, Hypothesis 4 was not supported.