پشتیبانی نظارت، روابط مبادله اجتماعی، و عواقب آزار و اذیت جنسی: یک تست از مدل های رقیب

Abstract This article examines the role that the immediate supervisor has in mitigating the negative consequences of sexual harassment experiences when he or she is not the perpetrator of the harassment. We examined a competing mediating/moderating effects model of perceived supervisory support and social exchange relationships on the consequences of perceived sexual harassment experiences. Using survey data gathered from military personnel, we found support for direct effects of both perceived sexual harassment and leadership on individual outcomes but failed to confirm our initial hypothesis of perceived leadership as a moderator. However, we found significant support for a moderating effect when the sample was subgrouped by gender of the leader. We also found partial support for leadership as a mediator of the relationships between sexual harassment and individual outcomes.

Introduction Sexual harassment is an issue of major concern for leaders in organizations. A great deal of attention has been given to leadership in the “diversity literature” where individual differences based on demographics have been shown to influence subordinate perceptions of leader behavior, particularly with regard to gender (Van Velsor & DiTomaso, 1996). Many other studies have noted that perceptions of sexual harassment behavior emanate from the power a leader has over the subordinate (e.g., Knapp et al., 1997 and Lengnick-Hall, 1995). Research in leadership has demonstrated that the leader, and the role that leader's play in the dyadic interaction with her/his subordinates, influences subordinate outcomes both positively and negatively (e.g., Bass, 1990). However, there is a striking paucity of literature that deals with the consequences of the perceived leader's actions relevant to sexual harassment — when the leader is not the perpetrator. What is needed are more studies that examine whether positive perceptions about the leader's role will buffer the negative effects of sexually harassing organizational environments when the supervisor is not the harasser. Therefore, the purpose of this article is to examine the role the immediate supervisor has in mitigating the negative consequences of sexual harassment experiences. Does the subordinate reduce his or her negative outcomes as a result of the social exchange relationship with a nonharassing supervisor? In the following sections, we give a brief explanation as to the nature of sexual harassment, discuss the relationship between gender and leadership, and then argue that effective leadership will reduce the impact of perceived sexual harassment experiences.

5. Results Table 1 provides a summary of the descriptive statistics for the entire sample. Table 2 provides summary descriptive statistics split by gender subgroups. In general, respondents in this sample reported experiencing sexually offensive behaviors rarely over the past 12 months, as indicated by low means for the sexual harassment variables. Nearly 81% of the respondents reported experiencing some form of gender harassment at least once over the past 12 months and about 84% reported experiencing some form of crude behavior over the same period. Sixty-three percent indicated that they were victims of unwanted advances while only 10.3% reported that they were targets of quid pro quo behaviors in their work place. Such sexually offensive behaviors do not seem to occur frequently as indicated by the low means reported for all four types of sexual harassment. Table 2. Descriptive statistics and intercorrelations by gender subgroup Variable Malea Female Intercorrelationb Mean S.D. Mean S.D. 1 2 3 4 5 6 7 8 9 10 11 (1) Gender harassment 1.10 0.97 0.92 0.89 – .53 .36 .24 −.02 −.16 −.30 −.31 −.21 −.29 .11 (2) Crude behavior 0.71 0.74 0.67 0.69 .45 – .64 .38 −.01 −.09 −.28 −.32 −.24 −.26 .12 (3) Unwanted advances 0.47 0.64 0.48 0.64 .32 .64 – .50 −.05 −.03 −.19 −.20 −.18 −.17 .10 (4) Quid pro quo 0.06 0.33 0.08 0.39 .32 .44 .54 – −.01 .01 −.14 −.14 −.10 −.12 .03 (5) Tenure 23.06 20.22 22.64 18.48 .03 −.01 −.02 .02 – .06 −.05 −.04 .02 −.03 −.02 (6) Gender mix 2.66 1.32 4.05 1.49 −.08 −.01 −.01 −.02 .02 – .01 .03 .01 .03 −.03 (7) Social exchange leadership 2.96 0.98 2.89 1.00 −.24 −.22 −.19 −.18 −.04 −.01 – .51 .51 .55 −.16 (8) Supervisory support 0.82 0.44 0.85 0.35 −.23 −.23 −.18 −.11 −.06 −.06 .49 – .30 .33 −.09 (9) Commitment 3.60 0.96 3.52 0.94 −.22 −.18 −.15 −.11 −.02 −.03 .48 .29 – .61 −.28 (10) Satisfaction 3.52 0.67 3.50 0.68 −.28 −.22 −.19 −.18 −.06 .01 .52 .32 .59 – −.26 (11) Turnover intentions 2.33 1.48 2.37 1.50 .09 .05 −.01 .01 .02 .01 −.15 −.10 −.27 −.23 – a Response group identified by whether respondents had a male or female supervisor; significant (P<.001) mean differences between the two groups are in italic (for the group with a greater mean). b Above-diagonal entries for respondents with male supervisors (N=6332). Below-diagonal entries for respondents with female supervisors (N=1761). For male supervisor-group, absolute value of correlation above .040 are significant at P<.001; For female supervisor-group, absolute value of correlation above .076 are significant at P<.001. Table options We compared the two subgroups defined by supervisor gender. Respondents reporting to a male supervisor did not differ from those reporting to a female supervisor in all measures excepting gender harassment, social exchange leadership and commitment. As expected, work groups led by female supervisors had a higher proportion of women in that work unit (mean 4.05 vs. 2.66, P<.001). Respondents led by a male supervisor reported experiencing significantly higher levels of gender harassment (mean 1.10 vs. 0.92, P<.001), as well as organizational commitment (mean 3.60 vs. 3.52, P<.001). The pattern of intercorrelation for male and female supervisor subgroups was quite similar. The four sexual harassment factors were moderately to highly intercorrelated, and the two leadership variables were significantly correlated with sexual harassment factors and outcome measures in the anticipated direction. The four harassment factors were also significantly correlated with the outcome measures. Results also indicate that perceptions of leadership were not influenced by either tenure or gender composition of the work group for each subgroup. 5.1. Effects of sexual harassment on outcomes Our first set of hypotheses stated that sexual harassment behaviors will be negatively related to individual outcomes of job satisfaction and organizational commitment, and positively related to a respondent's intentions to leave the organization. Results of the intercorrelations in Table 1 and Table 2, and the LISREL analysis reported in Table 3, provide support for our hypotheses across the entire sample and the subgroup sample. Table 3. Results of the path analysis — moderated effects of leadership on sexual harassment outcomes N for males=6332; N for females=1761. Parameter estimates are standardized to a common metric for comparison purposes. Significant differences between the regression estimates (P<.01) for the two subgroups are in italic. The degrees of freedom for the F test were 15, 8063. Incremental R2 for the interaction terms are in bold. Variables Commitment Job satisfaction Intentions to leave Male Female Male Female Male Female Harassment Gender harassment (GH) −0.018 −0.099 and −0.116 and −0.152 and 0.029* 0.057* Crude behavior (CB) −0.071 and −0.030 −0.073 and −0.030 0.049*** 0.022 Unwanted advances (UA) −0.046 and −0.040 0.010 −0.021 0.057 and −0.059** Quid pro quo (QP) −0.011 0.018 0.003 −0.030 −0.046 *** 0.003 Social exchange leader behavior (SEL) 0.477 and 0.431 and 0.506 and 0.458 and −0.137 and −0.132 and Supervisory support (SS) 0.044 and 0.058** 0.043 and 0.065** 0.006 0.002 Interactions SEL×GH 0.016 0.031 −0.025 0.025 −0.023 −0.058 SEL×CB 0.010 −0.002 0.021 0.013 −0.015 0.013 SEL×UA −0.016 −0.044 −0.029 0.001 0.025 −0.062 SEL×QP −0.022 0.002 0.021 −0.025 0.005 0.067* SS×GH −0.003 0.019 −0.008 −0.004 −0.010 −0.002 SS×CB −0.054* −0.007 −0.041* −0.021 0.018 −0.093* SS×UA 0.031 −0.013 0.037* −0.022 0.027 0.055 SS×QP −0.006 0.007 −0.014 0.014 −0.028 0.012 0.002* 0.002 0.005*** 0.001 0.003* 0.005 Total R2 0.239*** 0.203*** 0.277*** 0.247*** 0.029*** 0.033*** Chow test, F 1.80* 0.97 1.89* *** P<.001. a Power of the test at α=.01: Power=0.95. * P<.05. b Power of the test at α=.01: Power=0.90. ** P<.01. Table options In general, perceptions of sexual harassment experiences were significantly associated with the three outcomes. Of the four types of sexual harassment behaviors, the less severe forms of harassment such as gender harassment or crude behavior were negatively related to commitment and satisfaction, and positively related to intentions to leave the organization. Unwanted advances were negatively related to organizational commitment and positively related to intentions to leave the organization only for the male supervisor subgroup. The most severe form of sexual harassment, quid pro quo, was not related to commitment or satisfaction, and was related to turnover intentions in a direction opposite to that predicted for the respondents led by a male supervisor. To ensure that our findings of significance were not influenced by the large sample size, we also performed a power analysis and have indicated in Table 3 when the power of the test was at least 90%. While all significant parameter coefficients for the first two models predicting commitment and satisfaction exhibited adequate power, only the male supervisor subgroup met our statistical power criterion for unwanted advances predicting intentions to leave. There were significant differences in the effects of perceived sexual harassment experiences on outcomes for those who worked for a male supervisor vs. those who worked for a female supervisor. In general, sexual harassment variables explained a greater proportion of the variance in the outcome variables for male supervisors than for female supervisors. Gender harassment had a greater impact on organizational commitment and satisfaction for respondents who worked for a female supervisor, while crude behavior had a greater negative effect on commitment and satisfaction, and unwanted advances had a greater positive impact on intentions to leave for those who worked for a male supervisor. 5.2. Moderating effects of leader behavior We hypothesized that perceptions of social exchange leadership behavior and supervisory support for preventing sexual harassment will moderate the effects of sexual harassment on outcomes. Results of the moderated regression analysis are also reported in Table 3. We first tested the LISREL model with all main effects estimated. We then included the paths from the interaction terms to the three outcome measures, and examined the significance of the incremental R2, as well as the change in χ2. As previously indicated, the variables were mean centered before computing the interaction terms to minimize multicollinearity problems. Even though the χ2 change was significant for all six models, incremental R2 was significant only for the male supervisor subgroup for all three outcomes. Moreover, the practical significance was questionable as the additional variance explained by the interaction terms was less than 0.5%. Hence, we decided not to offer any interpretations of the few significant interaction terms. Our results provide support for conceptualizing the two leadership measures as suppressors rather than as moderators (Sharma, Durand, & Gur-Arie, 1981). Leadership variables explained more than 15% of the variance in commitment and satisfaction, and nearly 2% of the variance in turnover intentions for both subgroups. Both supervisory support and social exchange leadership variables exhibited levels of power of at least 0.99 for the male supervisor subgroup while only perceptions of social exchange leadership met our power criterion for the female supervisor subgroup. We found significant differences between male supervisor and female supervisor subgroups. Social exchange leadership had a significantly greater impact on commitment (0.486 vs. 0.444, P<.001) and satisfaction (0.506 vs. 0.458, P<.001) for the male supervisor subgroup. Supervisory support had no impact on turnover intentions, while social exchange leadership had an equal impact on intentions to leave for both subgroups. The total R2 explained by the moderator effects model was significant for all three outcomes and represented the contributions of the two leadership measures. Our second hypothesis was that leadership would moderate the effects of sexual harassment and was not supported. 5.3. Post hoc moderator analysis Our test of moderator effects was under the assumption that leadership variables will influence the form of relationship between sexual harassment behaviors and the outcomes. The suppressor effects we reported above suggest that leadership variables may moderate the strength of the relationships between sexual harassment and outcomes rather than the form of the relationship (Arnold, 1982). To test whether leadership variables moderate the strength of the relationship, we performed a subgroup analysis. We first split the sample into three roughly equal groups based on the social exchange leadership score. We excluded the middle-third from our analysis after ensuring that all variables were linearly distributed across the range of social exchange leadership scores. We regressed the three outcome measures on the four sexual harassment variables separately for the high and low social exchange leadership subgroup. We tested for model equivalence using the Chow test. Significant F statistics provide support for the moderator hypothesis. We performed a similar subgroup analysis for the dichotomous supervisory support variable. Since over 80% of our sample members indicated that their supervisor provided such support, we had an unequal distribution in the two subgroups. We found a similar pattern of results and chose not to report those post hoc results in the interest of space. Results are summarized in Table 4. Table 4. Moderating effect of high vs. low social exchange leadership on the relationship between sexual harassment and outcomes The degrees of freedom for the F test were 5, 4955 for the male supervisor subgroup and 5, 1271 for the female supervisor subgroup. Commitment Satisfaction Turnover intentions Male supervisor Female supervisor Male supervisor Female supervisor Male supervisor Female supervisor High SELa Low SELa High SEL Low SEL High SEL Low SEL High SEL Low SEL High SEL Low SEL High SEL Low SEL Gender harassment −0.029 −0.098*** −0.072 −0.213*** −0.149*** −0.173*** −0.156*** −0.247*** 0.005 0.072*** 0.027 0.140*** Crude behavior −0.068** −0.081*** −0.070 −0.016 −0.054 −0.126*** −0.015 −0.055 0.029 0.066*** 0.025 0.056 Unwanted advances −0.050 −0.105*** −0.067 −0.052 −0.017 −0.016 −0.041 −0.065 0.053* 0.041 −0.088 −0.019 Quid pro quo −0.004 0.060*** −0.019 0.003 0.028 −0.012 −0.053 −0.048 −0.012 −0.046 0.088*** −0.091 R2 0.015 0.045 0.027 0.060 0.034 0.077 0.038 0.116 0.005 0.018 0.010 0.023 Chow test, F 212.89*** 58.32*** 270.18*** 72.95*** 12.81*** 5.48*** a High vs. low social exchange leadership (SEL) groups formed by dividing the sample into roughly three equal groups on the basis of their SEL score and considering the top and bottom groups for this analysis. *** P<.001. ** P<.01. * P<.05. Table options The subgroup analysis provides significant support for the moderator hypothesis. The four sexual harassment variables collectively explained a significantly larger proportion of variance under the low social exchange leadership condition than under the high leadership condition for all three outcomes. Our results also indicate that sexual harassment variables had a far greater impact on outcomes when the respondents did not trust their supervisor than when they trusted her/him. The Chow test results also support our conclusion that social exchange leadership moderates the strength of the relationship between sexual harassment and outcomes. 5.4. Mediating effects of leader behavior We hypothesized the competing explanation to the moderating effects model as a mediated effects model. Leadership variables were hypothesized to mediate the relationship between perceived sexual harassment experiences and individual outcomes. We estimated the direct and indirect effects of sexual harassment on outcomes using LISREL. As in the previous analyses, we used tenure and gender mix as antecedents to both sexual harassment experiences and leadership perceptions so as to control for possible extraneous influences on leadership perceptions. Standard errors of indirect effects were estimated using the procedure recommended by Sobel (1982). Estimated effects of the relationships between sexual harassment, control variables, and leadership perceptions are summarized in Table 5. All four types of sexual harassment experiences were significant predictors of social exchange leadership perceptions for those supervised by a female while only the less severe forms of sexual harassment predicted the two leadership variables for the male supervisor subgroup. In general, sexual harassment experiences had a greater negative impact on leadership perceptions for the male supervisor subgroup than for the female supervisor subgroup. The less-severe forms of sexual harassment behaviors also met our criterion of exhibiting at least 90% power of the test of significance. Crude behavior had a significantly greater negative impact on leadership perceptions for the male supervisor subgroup as compared to the female supervisor subgroup. Table 5. Effects of sexual harassment experiences and control variables on leadership perceptions: results of LISREL analysis Parameter estimates are standardized to a common metric for comparison purposes. Significant differences between the regression estimates (P<.01) for the two subgroups are in italic. The degrees of freedom for the F test were 7, 8079. Variable Social exchange leader behavior Supervisory support Male supervisor Female supervisor Male supervisor Female supervisor (1) Gender harassment −0.213 and −0.178 and −0.199 and −0.166 and (2) Crude behavior −0.167 and −0.101 and −0.219 and −0.144 and (3) Unwanted advances 0.001 −0.049 * 0.011 −0.055 ** (4) Quid pro quo −0.025* −0.055** −0.014 0.031 (5) Tenure −0.053 and −0.040 −0.041*** −0.061** (6) Gender mix −0.031* −0.023 −0.014 −0.065 *** R2 0.078*** 0.043* 0.089*** 0.060*** Chow test, F 2.55** 3.22** *** P<.001. a Power of the test at α=.01: Power=0.95. b Power of the test at α=.01: Power=0.90. * P<.05. ** P<.01. Table options Tenure was negatively associated with leadership perceptions indicating that respondents with shorter tenure had a more positive perception of their supervisors. We expected gender mix in the work unit to be positively associated with leadership perceptions for both subgroups under the premise that female respondents might develop trust in their supervisor when their work unit had a greater proportion of females. However, gender mix was negatively associated with leadership perceptions indicating that respondents rated their supervisors higher when their work unit had a higher proportion of men. R2 for all four models was significant. The effects of sexual harassment on the three outcome measures were highly significant and in the expected direction when leadership measures were not entered in the model. When leadership variables were included in the model, the effects of sexual harassment on outcomes either became insignificant or considerably diminished. Table 6 summarizes the results of the test for mediated effects of leadership on outcomes. Both direct and indirect effects of sexual harassment on outcomes are reported for comparison purposes. The indirect effects of sexual harassment experiences on outcomes were either similar to or larger than the size of direct effects, indicating that leadership measures do mediate the relationship between sexual harassment and outcomes. However, this is not a complete mediation; not all the direct effects of sexual harassment became insignificant when the leadership measures were included in the model (Baron & Kenny, 1986). Our results, therefore, provide partial support for our third set of hypotheses. Table 6. Results of the LISREL analysis — mediated effects of leadership on outcomes NM=6332; NF=1761. Variable Commitmenta Job satisfaction Intentions to leave Direct Indirect Direct Indirect Direct Indirect M F M F M F M F M F M F (1) Gender harassment −0.020 −0.103*** −0.101*** −0.081*** −0.075*** −0.107*** −0.076*** −0.061*** 0.056** 0.105** 0.042*** 0.042*** (2) Crude behavior −0.075*** −0.037 −0.107*** −0.064*** −0.060*** −0.026 −0.080*** −0.048*** 0.096*** 0.084 0.042*** 0.034*** (3) Unwanted advances −0.073*** −0.039 0.001 −0.034*** 0.007 −0.016 0.001 −0.026*** 0.107*** −0.139** −0.001 0.018*** (4) Quid pro quo 0.074* 0.061 −0.032* −0.056*** 0.005 −0.043 −0.024* −0.044*** −0.171*** −0.111 −0.014* 0.028*** a Unstandardized coefficients of direct and indirect effects of sexual harassment reported; indirect effects are the estimates of the effects of sexual harassment on outcomes mediated by social exchange leadership and leadership support. *** P<.001. ** P<.01. * P<.05. Table options It should be noted that the direct effects of unwanted advances on turnover intentions were not in the predicted direction for the female supervisor subgroup nor were the effects of quid pro quo on turnover intentions for the male supervisor subgroup. In addition, the magnitude of indirect effects of sexual harassment for the male subgroup was usually larger than for the female subgroup. These results are consistent with the results summarized in Table 3 and Table 5. Hypothesis 3a and Hypothesis 3b were partially supported as we observed only partial mediation effects of sexual harassment rather than complete mediation effects. 5.5. Impact of leader gender We had hypothesized that leader gender would have a significant impact on the relationships among sexual harassment experiences, leadership perceptions, and outcomes. Specifically, we expected the effects to be greater for male supervisor subgroup than for female supervisor subgroup. The results of the t tests reported earlier provide support for concluding that the impact of sexual harassment experiences and leadership on outcomes are greater for male supervisor subgroup. We also applied two other tests to seek more conclusive support for the hypothesis. First, we applied the Chow test on the two subgroups to test whether the two sets of regression equations were equivalent. The F test results reported in Table 3 and Table 5 indicate that supervisor gender moderated the strength of the relationships in the case of organizational commitment (F=1.80, P<.05), turnover intentions (F=1.89, P<.05), social exchange leadership (F=2.55, P<.01), and supervisory support (F=3.22, P<.01). In all cases except turnover intentions, the set of predictors explained a significantly greater proportion of variance in the male supervisor subgroup than in the female supervisor subgroup. In addition to the Chow test, we conducted a two-sample LISREL analysis to test whether the two models were equivalent. We reestimate the LISREL models after constraining all parameters for the female subgroup to be equal to the male subgroup. The χ2 statistic increased significantly for this re-estimated model thus indicating that the equality hypothesis was not supported by the data. These tests support our fourth hypothesis that victims reporting to a male supervisor would suffer from greater negative consequences of sexual harassment than those reporting to a female supervisor. Of the sexual harassment behaviors, only gender harassment had a negative impact on all three outcomes for those who worked for a female supervisor. The full range of sexual harassment behaviors, however, affected those who worked for a male supervisor. Social exchange leadership perceptions had a significantly greater impact on outcomes for the male supervisor subgroup than for the female supervisor subgroup. Thus, it appears that social exchange leadership may be more important in male supervisor-led work groups to control the negative impacts of sexual harassment experiences. We also observed a similar pattern in the results predicting leadership perceptions. Leadership perceptions were affected by sexual harassment experiences to a greater extent for male subgroups than for female subgroups. Results thus provide significant support for our fourth hypothesis.