بررسی پیش بینی نظریه رهبری تحولی و مسیر-هدف در سطح فردی تجزیه و تحلیل

This study tested the recent path-goal leadership theory prediction [House, R.J., 1996. Path-goal theory of leadership: Lessons, legacy, and a reformulated theory. The Leadership Quarterly, 7, 323–352] that leader contingent reward behavior negatively moderates relationships between transformational leadership and subordinate performance and job satisfaction at the individual level of analysis. Also tested was the prediction that transformational leadership would positively augment the effects of leader contingent reward behavior [Bass, B.M., 1985. Leadership and performance beyond expectations. New York: Free Press]. Confirmatory factor analyses, hierarchical linear multiple regression, and within- and between-entity analyses were employed, along with a sample of 169 social services workers in 40 groups. No evidence was found supporting either the path-goal or additive augmentation hypotheses. However, a positive moderator effect was found for some transformational leader behaviors and the leader contingent reward behavior variable, supporting a form of “augmentation effect” that is not incongruent with Bass's [Bass, B.M., 1985. Leadership and performance beyond expectations. New York: Free Press] approach to transformational leadership. Additionally, strong support was obtained for the level of analysis prediction. Future research directions are briefly considered.

The results of this study provide strong support for Hypothesis 3 and for interpreting the relationships among the leadership variables, employee performance, and satisfaction as generally occurring at the individual level of analysis (i.e., both within- and between-units). With few exceptions, both the variance and covariance analyses (WABA I and WABA II, respectively) supported this conclusion. This, of course, is in agreement with the specification of House (1996) that the path-goal theory applies at the individual level of analysis. This finding also agrees with other theoretical positions, such as many of those dealing with charismatic or transformational leadership (e.g., Bass, 1985 and Conger and Kanungo, 1987). Practically speaking, an individual level effect means that relationships between supervisors and subordinates are defined according to each subordinate's perception of his/her leader's behaviors, and that the work unit or group has no empirical effect. Finding effects to occur almost entirely at the individual level of analysis allows the raw-score regression results to be interpreted without concern for possible misrepresentation of the nature of the obtained effects. The specific moderator results with respect to the Proposition 24 of House (1996) (Hypothesis 1) provide no support for the most recent revision of path-goal theory. For leader contingent reward behavior, although one of the four performance and three of the four satisfaction interactions were statistically significant, positive moderator effects were obtained. In fact, examining the moderator regression equations presented earlier shows that under conditions of low contingent reward behavior, the slopes of the lines were negative or, at best, positive but trivial (the most positive slope, for goals and satisfaction, was only + .03X). In contrast, under high contingent reward behavior, all of the regression slopes were positive and substantially larger-ranging from a low of .16X (vision and performance) to a high of .58X (goals and satisfaction). Furthermore, statistically significant positive moderator effects were obtained for vision with both performance and satisfaction. However, the revised path-goal theory and its underlying logic would appear to predict that the vision dimension should yield the strongest support for the theory (since it is ideologically based)—not the strongest support against it. The positive moderator results obtained for contingent reward can possibly be explained using distributive justice theory. Distributive justice is defined as the fairness of the ends or rewards received (Greenberg, 1987). Previous research has found a positive relationship between distributive justice and job performance (see the review by Greenberg, 1990) and between distributive justice and job satisfaction (e.g., Dailey and Kirk, 1992 and McFarlin and Sweeney, 1992). Possibly, those individuals who report low levels of contingent reward may have felt that they were not being fairly rewarded (i.e., they perceived distributive justice to be low). If followers do not feel fairly rewarded for effort they are currently exhibiting, it seems reasonable to expect that transformational leader behaviors (asking for additional effort and performance) would not be very successful. Indeed, if low levels of contingent reward reduce or decrease perceptions of distributive justice, it is perhaps not surprising that some of the regression slopes were negative under the low contingent reward conditions. Additionally, perceptions of contingent reward may have impacted the subordinates' trust of their supervisors, so that transformational leader behavior was most credible (and impactful) when accompanied by leader contingent reward behavior (cf. Pillai et al., 1999 and Podsakoff et al., 1990). On the other hand, the positive moderator effects which were obtained for contingent reward may be viewed as offering support for a different augmentation effect than has traditionally been hypothesized and tested by Bass and his associates (e.g., Bass, 1985, Bass, 1990 and Bass and Avolio, 1993). Traditionally, augmentation is tested via linear multiple regression, entering transactional leadership at the first hierarchical step. Then, in the next step, transformational leadership is added and the change in explained variance of the dependent variable is tested for statistical significance. Augmentation thus predicts a significant increase due to the main effect of transformational leadership (and no interaction is required). In the current study, this type of traditional or “additive augmentation” was not supported, contrary to Hypothesis 2. Re-running the regressions to enter contingent reward behavior first and each TLI transformational leadership dimension second did not yield any significant increases in explained variance at the second steps in the regressions. This is in direct contrast with previous studies which have found support for the augmentation effect using the MLQ of Bass and associates (Bass, 1985, Bass, 1988, Bass and Avolio, 1988 and Bass and Avolio, 1989; see Bass, 1985, Bycio et al., 1995, Dubinsky et al., 1995, Koh et al., 1995, Seltzer and Bass, 1990 and Waldman et al., 1990, for studies of augmentation). Our findings, when coupled with the significant positive moderator effects noted above, appear to suggest that effective transformational leadership was dependent upon, or at least enhanced by, the supervisor's display of contingent extrinsic reward behavior. This might reflect the effect of perceived distributive justice or trust in the supervisor, particularly since the organization in which our study was conducted is a large state bureaucracy in which cynicism abounds and in which complaints about employee compensation are widespread. Quite possibly, how employees interpret the receipt or non-receipt of contingent extrinsic rewards may be affected by the organizational context, in addition to individual difference factors such as cognitive dissonance (Festinger, 1980). Thus, perhaps a boundary condition needs to be added to the new path-goal theory of House (1996) to reflect justice, trust, and/or contextual organizational factors. In particular, and with respect to the Proposition 24 of House (1996), it may be that only extrinsic financial rewards produce the hypothesized effect. Extrinsic nonfinancial rewards (praise, recognition, etc.) may enhance the credibility of transformational leader behaviors such as vision and not cause cognitive dissonance processes to be evoked. Finding significant positive interactions (instead of only positive main effects) may not indicate lack of support for Bass's (1985) framework. Main effects should not be interpreted in the presence of significant interactions (Cohen & Cohen, 1983). While the simple additive effect proposed by Bass, 1985 and Bass, 1990 was not found, the results clearly support augmentation through positive moderation. Our findings support the notion that transformational leadership enhances the relationship between transactional (contingent reward) leadership and subordinate performance and satisfaction. Thus, managers engaging in transformational behaviors (e.g., articulating a vision, role modeling, fostering the acceptance of group goals, and communicating high performance expectations) concurrently with strong contingent reward behaviors will have significantly better results than those engaging in contingent reward behaviors alone. 4.1. Limitations One issue regarding this study is the extent to which common method variance should be viewed as problematic (cf. Podsakoff et al., 2003 and Richardson et al., 2005). We did not include a theoretically appropriate marker variable in our survey that would have allowed us to make post hoc corrections for common method variance (Richardson et al., 2005). However, we should mention four things that should at least somewhat reduce concerns about our findings being only artifacts and due to common method bias. First, we used a number of procedural study design remedies that should have at least somewhat reduced the susceptibility of our data to common source bias (for example, very strong assurances of respondent confidentiality and use of different questionnaire sections, instructions, and response scales for different measures; Podsakoff et al., 2003). Second, our CFA results (comparing multiple measurement models) clearly supported the five-factor model. Substantial common source variance might very well have resulted in supporting four or fewer factors (Podsakoff et al., 2003 and Podsakoff and Organ, 1986). (Parenthetically, we did not use the method factor CFA approach recommended by Podsakoff et al. [2003, Fig. 1, Situation 7] because Richardson et al. [2005, p. 39] specifically recommended “…that the unmeasured latent factor method not be used to correct for CMV [common method variance] in leadership data.”) Third, one of our dependent variables was supervisor-rated performance, a variable not affected by common source variance (Podsakoff et al., 2003). The pattern of results for this variable was no different than that obtained for the satisfaction dependent variable. Finally, it has been suggested by Hunt, Osborn, & Larson (1973), Kerr & Schriesheim (1974), Schriesheim & DeNisi (1981), and others that the testing of moderator predictions is, in itself, a partial control for this bias. As noted by Cummins (1972), “There is no a priori reason to suggest that the …difference between correlations of two variables measured at…levels of a third [moderator] variable…is influenced by method specificity [common method or source variance]” (p. 657). Some of the idiosyncrasies of our particular sample may have influenced the results we obtained. The sample was comprised of state social service agency employees, and, given the low pay and bureaucracy that the employees had to endure, we should probably assume that the majority of these employees were at least somewhat altruistically motivated. This ties in to the self-interest boundary condition of House (1996, p. 327). House proposes that employees will exert effort based on the expectation that they will receive valuable outcomes. Typically, we would think of “self-interest” as related to things such as pay raises, training opportunities, promotions, working conditions, etc. However, in this particular sample, most of these things were not generally available or alterable. Thus, the actual act of providing services to people in great need and the resulting satisfaction may have been the “valuable outcome” these employees sought. Given these circumstances, we must caution that our results may not generalize to other samples, and therefore call for similar studies to be conducted in other organizations. 4.2. Future research There are several directions for future research suggested by the current study. First, as there appears to be an inconsistency between our findings and those predicted by the Proposition 24 of House (1996) with respect to the contingent reward moderator, replication of this investigation is highly desirable. We should note that any such replication(s) should be attentive to both measurement and level of analysis issues, as we have tried to be in this investigation. On a related note, there is a need for further theoretical development to address the positive moderator effect of contingent reward on satisfaction and performance. The literature on transformational leadership and the augmentation effect should be considered and possibly integrated with the path-goal theory. At the very least, future research needs to explore the conflicting predictions of these two theories, perhaps by adding additional variables to further reflect both individual differences and organizational context effects. 4.3. Summary and conclusion In conclusion, while the current research provided support for the level of analysis prediction that transformational leadership–subordinate performance and satisfaction relationships occur at the individual level, there was no support for Proposition 24 of the revised path-goal theory of leadership (House, 1996). In particular, for leader contingent reward behavior, a number of statistically significant positive moderator effects were obtained, conflicting with the revised path-goal theory ( House, 1996). Additionally, our analyses indicated that the traditional additive augmentation model of Bass (1985) and his associates was also not supported, although positive interactive effects were obtained. Clearly, since these findings are potentially important, future research that replicates and further extends the current investigation appears quite desirable.