Write the summary as if the reader has never read the text

Descriptive Statistics Table 1 reports the means, standard deviations, correlations, and reliability coefficients of all variables used in the study. We found that leaders' moral attentiveness (r = .21, p < .01), and moral identity (r = .23, p < .01) were positively related to ethical leadership. Similarly, ethical leadership was positively related to followers' moral attentiveness (r = .2$,p< .01), and moral identity (r= .34, p < .01). Additionally, the coefficients between leaders' moral attentiveness and followers' moral attentiveness (r = .33, p < .01), between leaders' moral identity and followers' moral identity (r = .10,/? < .05), were statistically significant. We also calculated skewness and kurtosis statistics for all variables in Table 1 ; none indicated violations of normality assumptions. Hypothesized Model The M-PLUS program provides the correct parameter estimates and standard errors, and yields more accurate Type I error rates than non-hierarchical methods. M-PLUS not only provides a general way to deal with non-independence of ratings of ethical leadership among different followers within one unit under the leadership of the same manager, it also permits intercepts and slopes to vary randomly across leader clusters (Preacher, Zyphur, & Zhang, 2010). Therefore, we used multilevel path analysis Table 1: Descriptive Statistics, Reliability Coefficients, and Correlations amongst Study Variables Variable 1. Leader moral attentiveness 4.52 1.18 .84 2. Leader moral identity 4.55 .43 .26** .75 3. Ethical leadership 4.13 .71 .21** .23** .92 4. Follower moral attentiveness 4.47 1.25 .33** .24** .28** .88 5. Follower moral identity 4.41 .63 .07 .10* .34** .22** .87 Note. N = 460. * p < .05; **p < .01 ; *** p < .001 Values shown in the diagonal are Alpha reliability scores. This content was downloaded from 83.143.248.20 on Tue, 11 Jan 2022 16:04:03 UTC All use subject to https://about.jstor.org/terms Ethical Leaders and Their Followers 105 model to accommodate the cross-level nature of this study and to model top-down relationships (Preacher et al., 2010). The multilevel path analysis model decomposes the variance of a variable into the within-group variance and the between-group variance (Muthn & Asparouhov, 2009). The multilevel path analysis model allows modeling the different variance components independently at different levels. At the within-group level, variables can have random intercepts and random slopes can differ between groups. At the between-group level, these random intercepts are latent variables with each team member as the indicators. The multilevel path analysis avoids such problems as conflated within- and between-group relationships and is able to calculate the indirect effects more precisely than the traditional multilevel approach (Nohe, Michaelis, Menges, Zhang, & Sonnta, 2013; Preacher et al., 2010). The two-level model is able to partition the variance of a variable in clustered data into two orthogonal latent components (i.e., the between component and the within component) (Asparouhov & Muthen, 2006). The level 2 variables (leader moral identity and moral attentiveness) have only a between-group variance. Level 1 variables (ethical leadership, follower moral identity, and moral attentiveness) usually have both between and within components (Preacher et al., 2010). In our proposed model, independent variables such as leaders' moral identity, and moral attentiveness are assessed at level 2, the mediator ethical leadership is assessed at level 1 but with both within-group and between-group variances, and the outcomes such as followers' moral identity, and moral attentiveness are assessed at level 1. Therefore, in our model, X (leader moral identity, and moral attentiveness) are assessed at Level 2 and M (i.e., ethical leadership) and Y (i.e., follower moral identity, and moral attentiveness) are assessed at Level 1 (i.e., 2-1-1 design). We also calculated the intra-class correlation coefficients (ICCs) for ethical leadership (ICC(l) = .31 ,p < .01; ICC(2) = .71, p < .01). The results show that 31% of variance resides in the between-group level, which means that between-level (i.e., group level) variance of ethical leadership cannot be ignored. The ICC(2), which represents that the group means has reliability higher than the recommended level (Bliese, 2000), shows that it is appropriate to treat ethical leadership as a group-level construct. Therefore, it is appropriate to categorize the total variance of ethical leadership into the within-group and between-group components. Nevertheless, through the command of "Cluster" M-PLUS, we are able to adjust for clustering or classification in data analysis taking non-independence of observations into account (Bliese, 2000; MacKinnon, Lockwood, & Williams, 2004; Preacher & Hayes, 2008). Doing so offers more accurate coefficient estimation and significance testing by controlling for the non-independence effect at the group level (Bliese, 2000; Muthn & Muthn, 1998-2010), which also means that the confounding influence is due to the fact that employees are in the same organization (note: we have two firms) is also controlled for. We conducted the path analysis using Mplus 6.0 (Muthn & Muthn, 1998-2010) with robust maximum likelihood (MLR) estimation. The fit indexes of the proposed theoretical model are: f = 4.36, df= 2, CFI = .97, RMSEA = .05, SRMR (within) = .00, SRMR (between) = .07. Although the SRMR (between) is slightly higher than This content downloaded from 83.143.248.20 on Tue, 11 Jan 2022 16:04:03 UTC All use subject to https://about.jstor.org/terms 106 Business Ethics Quarterly the recommended score, it is still within the acceptable range of up to. 10 (Browne, & Cudeck, 1993). Also, the score of /df (2. 18) falls below the adequate fit range of less than 5 (Marsh & Hocevar, 1985). Table 2 shows the regression path coefficients, supporting hypotheses 1 through 4. Finally, to test for mediation (Hypothesis 5), we followed Mackinnon, Lockwood, Hoffman, West, and Sheets (2002) and calculated the product of coefficients with M-PLUS 6.0 (Muthn & Muthn, 1998-2010). The statistically significant indirect path coefficient implies that the relationship between the predictor and the outcome variables occurs through the mediator (Iacobucci, Saldanha, & Deng, 2007; Mackinnon et al., 2002; Preacher & Hayes, 2008). The indirect effect corresponding to the chain relationship (leader moral identity - ethical leadership - follower moral identity) is significant ( = .077, standard error = .043, p < .05), as shown in Table 3, and the direct relationship between leader moral identity and follower moral identity is not ( = .084, standard error = .082, p > .05), which is no longer considered a necessary condition for claiming a mediating effect by some scholars (e.g., Kenny, 2014; MacKinnon, Fairchild, & Fritz, 2007). Therefore, H5a, which posits that ethical leadership mediates the relationship between leaders' moral identity and followers' moral identity, was supported, which means that ethical leadership is a full mediator. The indirect effect corresponding to the chain relationship (leader moral attentiveness - ethical leadership - follower moral attentiveness) is not significant ( = .02, p > .05) as shown in Table 3. Therefore, H5b, which posits that ethical leadership mediates the relationship between leaders' moral attentiveness and followers' moral attentiveness, was not supported. However, the direct relationship between leader moral attentiveness and follower moral attentiveness is significant ( = .29, p < .01). To summarize, we found that ethical leadership mediates only the relationship between leader moral identity and follower moral identity, but not the relationship between leader moral attentiveness and follower moral attentiveness. We conducted further analysis by reversing the relationship paths among these variables. Specifically, we treated follower moral identity and follower moral attentiveness as IVs, and ethical leadership, leader moral identity and leader moral attentiveness as DVs. This model exhibits a poorer model fit (x2 = 29.91, df- 6, CFI = .73, RMSEA = .09, SRMR (within) = .04, SRMR (between) = .21) compared Table 2: Structural Equation Path Coefficients Est. S.E. p-value Significant Leader moral attentiveness - Ethical leadership .08* .04 < .05 YES Leader moral identity- Ethical leadership .21** .10 .10 NO Note. N = 460. The standardized coefficients are reported. * p< .05; ** p < .01 (1- tailed) This content downloaded from 83.143.248.20 on Tue, 11 Jan 2022 16:04:03 UTC All use subject to https://about.jstor.org/terms Ethical Leaders and Their Followers 107 Table 3: Mediation of the Effect of Leader Moral Characteristics on Follower Moral Characteristics through Ethical Leadership Est. S.E. P-Value Significant Leader moral identity- Ethical leadership- Follower moral .08 .03 .01 YES identity Leader moral identity- Ethical leadership-Follower moral .07 .03 .05 YES attentiveness Leader moral attentiveness-Ethical leadership - Follower .02 .01 .05 YES moral identity Leader moral attentiveness-Ethical leadership-Follower .02 .01 > . 05 NO moral attentiveness Note. N = 460. The standardized coefficients are reported. * p < .05; ** p < .01 (1- tailed) to the hypothesized model. This analysis casts doubt on the alternative explanation that followers with higher levels of moral attentiveness and moral identity rate their leader higher on ethical leadership.