The influence of leader humor on relationships between leader behavior and follower outcomes

Journal of Managerial Issues, Summer, 2009 by Robert P. Vecchio, Joseph E. Justin, Craig L. Pearce

In order to test the proposed moderator hypotheses, hierarchical regression analysis was conducted for the dependent variable of teacher performance. As stated earlier, a multiplicative interaction term was created from the mean-centered predictor variables for use in the final step of a set of comparative equations. The first step included the two control variables of job satisfaction and years of teaching (as these variables have been identified as correlates of the likelihood of being rated as a superior performer due to differential attrition (Schneider et al., 1995)). Following the initial control variable step, the main effect predictors were added in a second equation, followed by the interaction term in the third equation. The results of these analyses, as displayed in Table 4 (top portion), indicate that contingent personal reward interacted with the use of humor in accounting for significant increments in variance on the outcome measure of teacher performance (change in [R.sup.2] of .042, beta of -.222).

In order to interpret the form of the significant interaction, procedures outlined by Aiken and West (1991) were followed, wherein cuts were made at plus and minus one standard deviation from the mean on the leader behavior predictor variable. The form of the graphed interaction reveals that the present finding for leader contingent personal reward only partially replicates the findings of Avolio et al. (1999) for the outcome variable of teacher performance. Concerning the predictor variable of contingent personal reward, performance ratings were lower with the joint combination of low use of humor by the leader and low use of contingent personal reward (see Figure I), thereby supporting Hypothesis 2 more clearly at one extreme combination of humor and contingent personal reward.

To test Hypothesis 3 (wherein leader use of humor was predicted to moderate the effect of leader integrity on follower job performance), hierarchical regression analysis was again employed. Following the same procedures used to test Hypothesis 2, mean-centered predictor variables were used along with a multiplicative interaction term in a series of three equations that initially included the control and main effect variables. The key issue is whether the introduction of the interaction term in the final equation significantly adds to accounting for an increment in criterion variance. As indicated in Table 4 (bottom portion), integrity significantly interacted with the use of humor (change in [R.sup.2] = .049, beta of -.253).

A graph of this interaction (employing cuts at plus and minus one standard deviation from the mean on the leader behavior predictor variable of integrity) reveals that leaders who were judged to be less honest and made less use of humor did, in fact, provide lower ratings of follower performance (see Figure II), thereby supporting Hypothesis 3.

While follower gender and leader gender were not associated with ratings of the use of humor, the gender-interaction hypothesis (i.e., Hypothesis 4, which predicted that same-gender leader-follower pairs would have reports of higher levels of humor) was tested by contrasting the mean levels of humor for the four possible combinations of leader-follower gender (i.e., male-male, N = 58; female-female, N = 29; male-female, N = 29; and female-male, N = 63). Because of the inherently dichotomous nature of the predictors, analysis of variance (ANOVA) was conducted for the dependent measure of leader's use of humor. The results of this analysis (which were the same as the results yielded by a moderated regression analysis coding gender as 0, 1) indicated a partial eta-squared (increment in [R.sup.2]) value of .017, with a significance level (p < .10) that did not attain a conventional level of statistical significance. Hence the observed pattern of means (see Table 5) should be cautiously interpreted as being suggestive of the proposal that same-gender leader-follower pairs report higher levels of humor relative to different-gender pairs. Therefore, Hypothesis 4 received only directional support.