تشریح ساختار رهبری تحول گرا: نقش تاثیر

Despite being the most widely used and popular measure of transformational leadership, the MLQ has not received much attention in terms of possible method biases, either at the item or structural level. Based on a priori theory and empirical work, we examined the influence of affect on the measurement of transformational leadership, as assessed by the MLQ, and its structural relationships with important organizational outcomes in two separate studies. Results of the first study (N=307) indicated that while temporary mood states have little impact either on measurement or structural relationships, target-specific affect (i.e., liking) influenced both. The effect of target-specific affect was fully replicated in a second study (N=120) using an independent sample and a more conservative estimate of the liking effect. Overall, the results of this research suggest that transformational leadership, at least when assessed with the MLQ, is highly influenced by the interpersonal affect raters feel towards the target being rated (i.e., liking). These results are discussed in terms of their implications for future work on transformational leadership and affect.

In the current paper, we examined the extent to which liking and mood impacted both the measurement of TL and its structural relationships with important organizational outcomes. Overall, our results provided little support for the role of mood, suggesting that raters' mood states do not seem to be a source of method bias in survey research assessing TL with the MLQ. These findings coincide with previous examinations of TL and mood (Bass & Avolio, 1989 and Lewter & Lord, 1992) as well as a growing body of work in other domains (e.g., Keeping & Levy, 2000 and Williams & Anderson, 1994). For example, Keeping and Levy found that, although positive and negative affect produced minor method effects, these effects did not bias the meaning or measurement of performance appraisal reactions. In contrast to mood, our results demonstrated that liking not only substantially influenced raters' item level responses and the interrelationships between the TL subscales, but also influenced the substantive relationships that TL had with outcomes (i.e., job satisfaction and AOC). A possible explanation for why liking induced a bias in the TL indicators and relationships, while positive and negative affect did not, may lie in the specificity of liking. We offer this possibility based on the work of Ajzen and Fishbein (1977), who, within the context of attitude–behavior consistency, demonstrated that the lack of strong relationships between attitudes and behaviors is often due to a low correspondence between target and action. More specifically, they suggested that researchers have measured general attitudes as predictors of specific behaviors, thus producing a low correspondence between target and action. Applying this to the current context, it makes sense that target-specific affect should exhibit a higher relationship with TL than general affect, and thus present a greater bias in TL than general affect. In terms of the implications associated with the findings regarding liking, one conclusion that can be drawn is that liking is a key component of TL and that its exclusion is potentially problematic insofar as it results in the misspecification of the underlying model (James, Mulaik, & Brett, 1982). This is potentially problematic because misspecified models can substantially bias parameter estimates (Hayduk, 1987). In terms of the TL literature, the relationships reported between TL and organizational outcomes may include covariance due to liking, presenting a misrepresentation of previously reported relationships. Coinciding with our findings, a growing number of scholars have called for the investigation of affective processes in TL research (e.g., Ashkanasay & Tse, 2000 and George, 2000). As George notes, TL is fundamentally an emotional process, one that is possibly mediated by subordinates' affective responses to their leaders. To assess George's (2000) suggestion, we ran a model in which liking fully mediated the relationship between TL and the substantive outcomes. In both samples, the liking mediation model provided a good fit for the data (Sample 1: χ2=346.76, p<0.01; χ2/df=2.39; TLI=0.943; CFI=0.952; RMSEA=0.067; Sample 2: χ2=270.00, p<0.01; χ2/df=1.56; TLI=0.911; CFI=0.924; RMSEA=0.085). Furthermore, in both samples, TL was significantly related to liking, liking was significantly related to each of the outcomes, and a significant indirect effect emerged between TL and the outcomes. To further assess whether the relationship between TL and the outcomes was fully mediated, we ran a series of partial mediation models. For each of the partial mediation models, we included a distinct direct path (e.g., TL→OCB). To assess whether each of the partial mediation models was a significant improvement in terms of fit over the fully mediated model, we used a chi-square difference test. For only a single variable, AOC, in one sample (Study 1), did a significant improvement result by including a direct path between TL and the outcome. Overall, these supplementary analyses were consistent with George's suggestion that affective reactions to a supervisor mediate TL and relevant organizational outcomes. Future research needs to replicate and extend these findings. If, in fact, liking plays a substantive role in TL, models of TL will need to take this into account. Beyond the mediational explanation, it is important to note an alternative interpretation for our results. In this regard, our findings may indicate that liking is a source of method bias, one that substantially alters important relationships (Doty & Glick, 1998). That is, liking may be little more than a nuisance factor that contaminates subordinate-generated responses to TL instruments. If this is the case, it may be quite problematic, given that we found the liking effect to be pervasive. Viewing our findings through this lens, it is important to understand the precise process by which liking works. That is, liking may simply reflect a summary judgment, one that is based on the behaviors in which a subordinate observes a leader engaging (Srull & Wyer, 1989). Alternatively, subordinate liking may precede a leader's actual behavior and color a subordinate's encoding, memory, and judgement processes (Dasborough & Ashkanasy, 2002). The distinction between these two possibilities has important implications for the basic assumptions of the TL model. In this regard, TL work has been largely premised upon a leader-centric perspective (Brown & Lord, 2001 and Meindl, 1995), one that assumes that TL behavior causes significant outcomes. Thus, whether liking represents a summary judgement or precedes and constrains subordinate information processing is an issue that requires future consideration. 8.1. Limitations and implications In the current article, we chose to model affect as a method factor as a means of delineating its effect on TL and the relations between TL and several outcome variables. We recognize that liking may not be considered a traditional method effect by some researchers. However, it is important to note that the use of affective variables in this capacity is not unusual in the organizational behavior literature (e.g., Munz et al., 1996, Schmitt et al., 1996, Williams & Anderson, 1994 and Williams et al., 1996). For example, both Munz et al. and Williams and Anderson examined the effects of positive and negative affectivity on organizational variables (e.g., organizational commitment and job satisfaction) and modeled affectivity as a method effect. Moreover, conceptualizing liking in this manner does not detract from the basic conclusions that can be drawn from our results: that liking exerts a strong influence on TL, at least when TL is assessed with the MLQ. As such, liking should not be ignored and should be the subject of future research by scholars (George, 2000). Given the cross-sectional nature of our studies, it was not possible to ascertain the precise causal relationship that liking has with TL. Thus, future experimental and longitudinal work is needed to delineate the precise nature of the TL-liking relationship. A second limitation of the current investigations lies in the nature of the samples that we employed. In both studies, we recruited employed undergraduate and college students. An advantage of this sampling was that it allowed us to obtain individuals who worked in a number of different occupations and across a wide variety of organizations, thereby decreasing potential contextual constraints (Johns, 2001). However, this approach did result in a fairly young sample, comprised of individuals in mostly entry-level positions, and who evaluated organizational leaders at the lower echelons of an organization. It is important to note, however, that TL scholars view TL to be robust and generalizable across all organizational levels (Avolio, 1999, Bass, 1997 and Bass et al., 1987). Thus, although the findings and conclusions drawn from our paper should be considered in light of the sample characteristics, it seems unlikely that sample characteristics would have compromised our findings. However, future work should examine the role of liking for other sample populations. Third, it should be noted that the SEM techniques we employed are not without limitations. Podsakoff, MacKenzie, Lee, and Podsakoff (2003) have noted that these techniques only allow for a single source of method bias and assume that the method factor does not interact with the substantive constructs. However, it is noteworthy that the methodology employed in this study does possess clear advantages over the partial correlation techniques previously employed (Bass & Avolio, 1989). The SEM techniques we used allowed us to take into account measurement error in the method factor, examine the biasing impact of the method factor at both the measurement and construct level, and allowed the method factor to differentially influence the substantive indicators. Despite these benefits, however, conclusions drawn on the basis of any single technique must be tentative until they can be corroborated. Future research should use alternative statistical techniques, such as those highlighted by Podsakoff et al., as well as disentangle liking and TL in the laboratory. Although this paper is primarily methodological in nature, we suggest that it also has managerial implications. Having employees complete the MLQ is analogous to a performance appraisal, and, therefore, many of the same challenges involved in appraisal ratings apply to the assessment of TL with the MLQ. Similarly, many of the suggestions for improvement to performance ratings can be applied to the use of the MLQ. For example, because liking can play a role in MLQ ratings, managers should ensure that they have a number of raters complete the questionnaire. In addition, it is important that respondents be trained to fill out the MLQ properly, paying attention to the fact that the questions inquire about behaviors, not general impressions or emotions. We believe the current investigation contributes to the TL literature as well as the more general organizational behavior literature for both substantive and methodological reasons. Substantively, our results broaden contemporary understanding of TL, thus responding to calls from leadership researchers to focus on the construct of TL. For example, Hunt (1999), utilizing a framework provided by Reichers and Schneider (1990), suggested that the investigation of TL has entered the second stage of construct development, concept evaluation/augmentation. We believe the results of our studies contribute to this second stage of development. Second, our results support contemporary calls to investigate affective processes in leadership (e.g., Dasborough & Ashkanasy, 2002, George, 2000 and Hall & Lord, 1995). Although the precise role of liking requires further explication, our studies advance the understanding of how affect is related to TL by illustrating the significant role that liking plays in TL, at both measurement and structural levels. This is important because although liking has been explored in other leadership domains (e.g., leader–member exchange), ours is one of the few studies to do so in the TL literature. Methodologically, the current investigation illustrates the utility of applying the techniques introduced by Williams and his colleagues to the leadership domain to obtain a more comprehensive understanding of the influence of method effects on substantive constructs. In addition, it represents an important addition to the growing body of research utilizing these procedures (e.g., Munz et al., 1996, Schmitt et al., 1996, Williams & Anderson, 1994 and Williams et al., 1996). Similar to other studies using these procedures, we modeled diffuse affect as a method effect. Consistent with these studies, we did not find that diffuse affect induced a method bias in any of the relationships. However, unlike previous research, we also modeled target-specific affect in our study and found that it induced method effects on many of the relationships. Future research should investigate the implications of using target-specific method effects in other content domains to determine if this can uncover important relationships. TL is currently one of the most popular leadership theories and provides the foundation for many leadership training interventions (Avolio, 1999). As a consequence, further elaboration and elucidation of the TL construct is imperative, not only for research purposes, but in order to better serve organizations and employees. After all, a better understanding of what TL is can only improve our ability to provide effective leadership training programs and interventions. To this end, we hope that our studies represent one step toward a more comprehensive understanding of the construct of TL, particularly as it is assessed with the MLQ, its most popular measure.