RESULTS

Liking Manipulation Check

As part of the debriefing questionnaire, participants rated the likeability of each division manager. Each statement read, in part, "Based on the information presented, please indicate your impression of how likeable is." The first statement referred to the manager of division A and the second statement referred to the manager of division B. The end points on a seven-point scale for each statement were "Completely Unlikeable" (1) and "Completely Likeable" (7).

A total of 69 participants were assigned to the "Like Manager A” condition. Under the "Like Manager A” condition the mean likeability ratings (standard deviations) of manager A and manager B were 6.0 (0.9) and 3.8 (1.5), respectively. Among the 67 participants assigned to the "Like Manager B" condition, the mean likeability ratings (standard deviations) of manager A and manager B were 3.2 (1.5) and 5.7 (1.0), respectively. A one-way ANOVA was performed using manager likeability, performance measure organization (PMO), and the interaction as independent variables and likeability ratings of manager A as the dependent measure. Manager likeability condition was significant (F=176.37, p<0.0001) and the means were in the expected direction. Neither PMO nor the interaction was significant. A second one-way analysis of variance (ANOVA) was performed using manager likeability, PMO, and the interaction as independent variables and likeability ratings of manager B as the dependent measure. Manager likeability condition was significant (F=71.34, p<0.000 1) and the means were in the expected direction. Neither PMO nor the interaction was significant. Overall, these results indicate that the manipulation of manager likeability was successful.

Hypothesis Testing

Hypothesis l(a) predicts that the BSC instrument will lessen the influence of division manager likeability on performance ratings compared to the NOFORM instrument. A 2 ~2 analysis of variance (ANOVA) was conducted using the rating difference score as the dependent measure. The independent variables were manager likeability at two levels and performance measure organization at two levels. This hypothesis would be supported by a significant interaction between manager likeability and performance measure organization and the pattern of results were consistent with the BSC format acting as a moderator variable. Descriptive statistics for the rating difference scores are presented in Table 3, Panel A. The ANOVA results (see Table 3, Panel B) indicate that manager likeability (F=8.78, p<0.01) is significant. As shown in Table 3, Panel A, the rating difference score was higher in the "Like Manager K' condition (Mean=4.01) than under the "Like Manager B" condition (Mean=-0.58). However, neither the performance measure organization main effect nor the interaction between manager likeability and performance measure organization is significant. Thus, Hypothesis l(a) is not supported.

Hypothesis l(b) predicts that the BSC instrument will lessen the influence of division manager likeability on conviction ratings compared to the NOFORM instrument. A 2×2 analysis of variance (ANOVA) was conducted using the conviction difference score as the dependent measure. The independent variables were manager likeability at two levels and performance measure organization at two levels. Descriptive statistics for the conviction difference scores are presented in Table 4, Panel A. The ANOVA results (see Table 4, Panel B) indicate that manager likeability (F= 11.93, p<0.001) is significant. As shown in Table 4, Panel A, the conviction difference score was higher in the "Like Manager A” condition (Mean=4.49): than under the "Like Manager B" condition (Mean=-l.43). This indicates that strength of conviction is positively associated with manager likeability. However, neither the performance measure organization main effect nor the interaction between manager likeability and performance measure organization is significant. Thus, Hypothesis l (b) is not supported.

Hypothesis l(c) predicts that the BSC instrument will lessen the influence of division manager likeability on bonus allocations compared to the NOFORM instrument. A 2 ~ 2 analysis of variance (ANOVA) was conducted using the bonus allocated to manager A as the dependent measure. The independent variables were manager likeability at two levels and performance measure organization at two levels. Descriptive statistics for the bonus assigned to manager A are presented in Table 5, Panel A. The ANOVA results (see Table 5, Panel B) indicate that manager likeability is significant (F=12.26, p<0.001). As shown in Table 5, Panel A, the bonus assigned to manager A is higher in the "Like Manager A” condition (Mean=$10,684) than in the "Like Manager B' condition (Mean=$9,716). However, neither the main effect for performance measure organization nor the interaction between performance measure organization and manager likeability is significant. Thus, Hypothesis l(c) is not supported.

Supplemental Analysis

In this supplemental analysis, we seek to determine whether rating difference scores mediate the effect of manager likeability on bonus allocations. Mediating variables do not change the relationship between other variables, but increase one's understanding of the process by which evaluators "transform the predictor or input variables" (Baron & Kenny, 1986, p. 1178). Baron and Kenny (1986) assert that several relationships must hold to demonstrate mediation. In the context of our study, an independent variable (subordinate manager likeability) must be found to have a direct effect on the mediator (rating difference scores) as well as the dependent variable (manager A bonus). Both of these effects have been found. In addition, there must be a direct relationship between the mediator (rating difference score) and the dependent variable (manager A bonus). The correlation between these two variables is positive and highly significant (r=0.77, p<0.0001). Finally, when the mediator (rating difference score) is controlled for, the effect of the independent variable (manager likeability) on the dependent variable (manager A’s bonus) should be eliminated or substantially lowered. To test this final aspect of mediation, analysis-of-covariance was conducted with manager A’s bonus as the dependent variable and the rating difference score as a covariate in addition to the two independent variables. The results show that the rating difference score is highly significant (F= 166.4, p<0.001) and that manager likeability remains marginally significant (F=2.95, p<0.09). The overall pattern of results indicates that rating difference scores partially mediate the relationship between manager likeability and compensation. This suggests that manger likeability has a direct effect on compensation over and above the indirect effect resulting from elevated performance evaluations.

结果

喜好处理检验

简要介绍本项数据的内容及取得的方法(问卷的具体问题设计)

给出描述统计数据(样本量、均值、标准差)及方差分析的方法及结果。

统计结果确认变量喜好程度有不同水平，即评价者对A、B的喜好有明显差别

假设检验

重述假设1(a)的内

容，介绍检验方法和检验结果，表3提供方差分析统计量F值，得出结论：假设1(a)不成立。

重述假设1(b)的内容，介绍检验方法和检验结果，表4提供方差分析统计量F值，得出结论：假设1(b)不成立。

重述假设1(c)的内

容，介绍检验方法

[和检验结果，表5提；供方差分析统计量f值，得出结论：假设1(c)不成立。

补充分析

协方差分析变量的相关性，检验变量的中介效应。得出三个变量(rating difference score，subordinate manager likeability, bonus)之间的关

系。