سابقه و پیامدهای فراوانی مقایسه اجتماعی رو به بالا و رو به پایین در محل کار

Abstract The current paper examines the dispositional and situational antecedents, as well as the attitudinal and behavioral consequences, of the frequency of upward and downward social comparisons. We predicted social comparison frequency would be influenced by uncertainty-related antecedents, and that social comparisons in organizations would be characterized by contrast, not assimilation, effects. A large and occupationally diverse sample of 991 employed adults was surveyed at three separate points in time over a 12–16 week period. Our results, based on structural equation modeling, indicated that (a) role ambiguity, task autonomy, and core self-evaluations were significant predictors of upward social comparison, (b) upward social comparison was significantly negatively related to job satisfaction and affective commitment, (c) downward social comparison was significantly positively related to job satisfaction and affective commitment, and (d) upward and downward social comparisons had significant positive and negative indirect effects on the frequency of job search behaviors, respectively. The findings are discussed in terms of their general implications for understanding the importance of directional social comparison processes in organizational settings.

Results Table 1 presents means, standard deviations, alphas, and intercorrelations among the study variables. Notably, upward social comparison was significantly related to each of the hypothesized antecedents (i.e., CSE, job ambiguity, task autonomy) and each of the hypothesized outcomes (i.e., job satisfaction, affective commitment, and job search behavior). Similarly, albeit more weakly, downward social comparison was also significantly related to many of the antecedent and outcome variables. It is noteworthy that upward and downward social comparison were positively correlated (r = .63), a result that largely replicates prior social comparison research (e.g., Buunk et al., 2003). The positive correlation shows that individuals who make upward social comparisons also tend to make downward social comparisons. Overall, the pattern of bivariate correlations provides encouraging evidence that upward and downward social comparisons at work are related to a number of well established organizational variables. Table 1. Descriptive statistics, zero order correlations, and alphas Mean SD 1 2 3 4 5 6 7 8 9 10 11 12 1. Age 41.39 8.88 — 2. Gender .46 .49 −.14⁎⁎ — 3. CSE 3.67 .59 .05 −.03 .85 4. Job ambiguity 2.97 1.19 −.07⁎ −.05 −.47⁎⁎ .92 5. Task autonomy 5.82 1.13 .07⁎ −.01 .28⁎⁎ −.39⁎⁎ .71 6. POS 4.54 1.41 −.02 .03 .34⁎⁎ −.52⁎⁎ .42⁎⁎ .94 7. LMX 2.92 .68 −.03 .01 .27⁎⁎ −.49⁎⁎ .38⁎⁎ .64⁎⁎ .92 8. Upward social comparison 2.50 .72 −.15⁎⁎ −.05 −.27⁎⁎ .21⁎⁎ −.18⁎⁎ −.17⁎⁎ −.19⁎⁎ .80 9. Downward social comparison 2.26 .78 −.14⁎⁎ −.02 −.12⁎⁎ .12⁎⁎ −.09⁎⁎ −.06 −.07⁎ .63⁎⁎ .85 10. Job satisfaction 3.91 .74 .12⁎⁎ .02 .44⁎⁎ −.51⁎⁎ .44⁎⁎ .50⁎⁎ .43⁎⁎ −.23⁎⁎ −.08⁎ .94 11. Affective commitment 4.34 1.41 .14⁎⁎ .03 .27⁎⁎ −.40⁎⁎ .34⁎⁎ .58⁎⁎ .47⁎⁎ −.20⁎⁎ −.05 .61⁎⁎ .88 12. Job search behavior 1.66 .74 −.08⁎ −.02 −.17⁎⁎ .28⁎⁎ −.22⁎⁎ −.28⁎⁎ −.26⁎⁎ .25⁎⁎ .10⁎⁎ −.31⁎⁎ −.36⁎⁎ .88 Note. N ranges between 972 and 991, alphas are on the diagonal in bold. Gender: 0, female and 1, male. ⁎ p < .05 level. ⁎⁎ p < .01 level. Table options Next, we assessed our 10-factor measurement model. The results of this analysis indicated that all of the indicators loaded significantly on their respective latent factors (0.65–0.95). Table 2 displays the indicator loadings and Table 3 reports the fit statistics for the measurement model. When evaluated in terms of the recommended cutoffs or the combination cutoff approach (Hu & Bentler, 1999) the fit indices indicate that measurement model fit the data adequately For purposes of comparison we contrasted the hypothesized measurement model to a 1-Factor measurement model in which all of the hypothesized factors were set to load on a single underlying factor. Additionally, as indicated in Table 3, we also compared the fit of the hypothesized measurement model to the independence model. Finally, we also compared the fit of the hypothesized measurement model against a 9-factor model in which the two social comparison factors were combined. In all three cases, the hypothesized 10-factor measurement model fit the data best, both in terms of the fit statistics and when directly contrasted using a Δχ2 test. Table 2. Measurement model indicator loadings for the hypothesized model Indicator Loading Ambiguity 1 .906 Ambiguity 2 .840 Ambiguity 3 .887 CSE 1 .843 CSE 2 .798 CSE 3 .778 Autonomy 1 .660 Autonomy 2 .836 POS 1 .910 POS 2 .909 POS 3 .949 LMX 1 .924 LMX 2 .881 LMX 3 .879 Upward 1 .872 Upward 2 .778 Upward 3 .655 Downward 1 .864 Downward 2 .867 Downward 3 .718 Affective commitment 1 .837 Affective commitment 2 .865 Affective commitment 3 .829 Job satisfaction 1 .892 Job satisfaction 2 .942 Job satisfaction 3 .869 Job search 1 .746 Job search 2 .956 Note. All loadings were significant at the p < .01 level. CSE, Core self-evaluations; POS, Perceived organizational support; LMX, Leader-member exchange. Table options Table 3. Model fit statistics χ2 df Δχ2 χ2/df TLI RMSEA CFI SRMR Model Hypothesized 10-Factor Measurement Model 1214.99⁎⁎ 305 — 3.98 .94 .055 .96 .038 Independence Model 20596.09⁎⁎ 378 19381.10⁎⁎ 54.49 — .232 — 1-Factor Measurement Model 11504.07⁎⁎ 350 10289.08⁎⁎ 32.87 .40 .179 .45 .137 9-Factor Measurement Model 1454.63⁎⁎ 306 239.64⁎⁎ 4.75 .93 .062 .94 .137 Hypothesized Structural Model 1446.06⁎⁎ 322 — 4.49 .94 .059 .94 .055 Alternative Structural Model 1280.11⁎⁎ 319 165.95⁎⁎ 4.01 .94 .055 .95 .046 Note. TLI, Tucker–Lewis Index ( Tucker & Lewis, 1973); RMSEA, root-mean-square error of approximation ( Steiger, 1990); SRMR, standardized root-mean-square residual; CFI, comparative fit index. In the 10-Factor model the relationships between the latent constructs were freely estimated. For the measurement models, the Δχ2 was calculated by independently contrasting the Independence model, the 9-Factor model, and the 1-Factor Measurement Model against the hypothesized 10-Factor Measurement Model. For the alternative structural model, the Δχ2 was calculated by contrasting it with the hypothesized structural model. ⁎⁎ p < .01. Table options Given the acceptable fit of our measurement model we proceeded to test our structural model (Fig. 1). The fit indices suggested that the structural model provided a good fit, both when evaluated against the recommended cutoffs and when evaluated in terms of the combination cutoff approach (Hu & Bentler, 1999). Table 4 displays the direct and indirect relationships (controlling for all other variables in the model) as well as the amount of variance in the endogenous variables accounted for by the exogenous constructs (i.e., R2). Table 4. Standardized direct effects, indirect effects, and R2 for the Hypothesized Model Outcome Upward social comparison Downward social comparison Job satisfaction Affective commitment Job search behavior R2 .151 .026 .376 .451 .182 Predictor Direct Indirect Direct Indirect Direct Indirect Direct Indirect Direct Indirect Core self-evaluations −.25⁎⁎ −.10⁎ .07⁎⁎ .05⁎⁎ −.02⁎⁎ Job-ambiguity .09⁎ .05 −.02 −.01 .01 Task autonomy −.15⁎⁎ −.05 .05⁎⁎ .03⁎⁎ −.02⁎⁎ POS .38⁎⁎ .50⁎⁎ −.21⁎⁎ LMX .15⁎⁎ .15⁎⁎ −.07⁎⁎ Upward social comparison −.17⁎⁎ −.24⁎⁎ .13⁎⁎ Downward social comparison .12⁎ .15⁎⁎ −.08⁎⁎ Job satisfaction −.13⁎⁎ Affective commitment −.33⁎⁎ Note. The parameters and significance of the indirect effects were determined using a bootstrapping analysis and the bias corrected percentile method. ⁎ p < .05. ⁎⁎ p < .01. Table options Turning to the hypothesized direct relationships, job ambiguity and task autonomy were, respectively, significantly positively and negatively related to upward social comparisons, but neither variable was significantly related to downward social comparisons. These findings provide support for Hypotheses 1a and 2a, but not for Hypotheses 1b and 2b. Hypotheses 3a and 3b, which predicted that CSE would be negatively related to upward and downward social comparisons, were supported. Consistent with the predicted contrast effects (Hypotheses 4a, 4b, 5a, and 5b), upward social comparisons were negatively related to job satisfaction and affective commitment, while downward social comparisons were positively related to these two attitudinal variables. Importantly, these effects were over and above the significant effects of POS and LMX on affective commitment and job satisfaction, suggesting social comparison processes predict over and above these established antecedents of job attitudes.5 As well, Hypotheses 6a and 6b, which predicted that the impact of upward and downward social comparisons on job search behavior would be mediated through job satisfaction and affective commitment, were supported, as evident in the significant indirect effects in Table 4.6 In addition to our hypothesized direct and indirect relationships, it is also noteworthy that with the exception of job ambiguity, all of the indirect relationships predicting job satisfaction, affective commitment, and job search behavior were significant. These significant indirect effects suggest that the impact of the dispositional and situational factors on relevant job attitudes work in part through the frequency of directional social comparisons. To follow-up on the significant indirect relationships a series of supplementary analyses were conducted. Although our findings suggest that upward and downward social comparisons mediated the relationships between CSE, autonomy, and the job attitude and job search constructs, it is unclear whether these mediation effects were full or partial. For example, self-concordant goals have been shown to mediate the relationship between CSE and job satisfaction (Judge, Bono, Erez, & Locke, 2005). Similarly, other mechanisms, such as met expectations and person-job fit, have been proposed to mediate the relationship between characteristics of the work setting (e.g., autonomy) and commitment (Meyer & Allen, 1997). Given other, unmodeled, mechanisms we assessed the extent to which partial mediation may be present by running several additional models. In each of these models a distinct direct path was freed between CSE/autonomy and the attitudinal variables. Each new model was then contrasted with the fully mediated model (Table 5). Table 5. Partial mediation tests χ2 df Δχ2 χ2/df TLI RMSEA CFI SRMR Hypothesized Structural Model 1446.06⁎⁎ 322 — 4.49 .94 .059 .94 .055 Freed path Autonomy → Affective commitment 1444.69⁎⁎ 321 1.37 4.50 .94 .059 .94 .056 CSE → Affective commitment 1438.82⁎⁎ 321 7.24 4.48 .94 .059 .95 .056 Autonomy → Job satisfaction 1377.57⁎⁎ 321 68.49⁎⁎ 4.29 .94 .058 .95 .050 CSE → Job satisfaction 1354.12⁎⁎ 321 91.94⁎⁎ 4.22 .94 .057 .95 .050 Note. TLI, Tucker–Lewis Index ( Tucker & Lewis, 1973); RMSEA, root-mean-square error of approximation ( Steiger, 1990); SRMR, standardized root-mean-square residual; CFI, comparative fit index. ⁎⁎ p < .01. Table options In the top row of Table 5 the results of the hypothesized structural model test is presented. In each subsequent row the results of a less constrained, partial mediation model is presented.7 As can be seen in Table 5, both CSE and autonomy had a significant direct relationship with job satisfaction, indicating that upward and downward social comparisons only partially mediated the relationship between CSE/autonomy and job satisfaction. In contrast, none of the direct paths between the CSE/autonomy and affective commitment were significant, indicating that upward and downward social comparison processes fully mediate these relationships. As a final step, we computed an alternative structural model, freeing the three paths between the exogenous variables (i.e., CSE, task autonomy, role ambiguity) and job satisfaction.8 As can be seen in Table 3, the alternative structural model fit the data quite well. However, with the exception of the emergence of a significant indirect relationship between job ambiguity and job search behavior, the pattern of significant direct and indirect relationships reported in Table 4 did not change in the alternative structural model. Moreover, this alternative model was very similar to the hypothesized model in terms of its fit indices.