Attributional comparisons across biases and leader-member exchange status.

Leadership is a phenomenon involving a leader, a follower, and the relationship between the two (Graen and Uhl-Bien, 1995). Two theories that address the relational nature of leadership in organizational settings are attribution theory (Green and Mitchell, 1979) and leader-member exchange (LMX) theory (Dansereau et al., 1975). Attribution theory asserts that people continuously attempt to develop causal explanations for events experienced by themselves (Kelley, 1967) and others (Kelley and Michela, 1980), and that the resulting perception of causality influences future behavior (Abramson et al., 1978; Campbell and Martinko, 1998; Sabini et al., 2001). LMX theory suggests that supervisors develop high quality "in-group" exchanges with some subordinates and develop low quality "out-group" exchanges with other subordinates (Dansereau et al., 1975). The quality of the exchange relationship is related to positive outcomes for in-group members (Vecchio and Gobdel, 1984) and negative outcomes for out-group members (Tanner et al. 1993).

With respect to attributions, an important issue is the target of the attribution, or the person about whose behavior an attribution is made, in that there are asymmetries in attributions one makes about oneself and others (Weary et al., 1989). The self-serving bias, for example, is the tendency for individuals to take credit for their successes and to blame situational factors for their failures (Kelley, 1967), while the actor-observer bias is the tendency to attribute one's own performance to situational factors and to attribute others' performance to internal factors (Jones and Nisbett, 1972; Martinko and Douglas, 1999).

Findings regarding these two biases are robust (Wilson and Levine, 1997); however, a comparison of the self-serving and actor-observer biases indicates a contradiction in the attributions expected when performance is positive (Martinko and Gardner, 1987). The self-serving bias predicts that an actor will make an internal attribution for his or her positive performance, while the actor-observer bias predicts that an actor will make an external attribution for his or her positive performance. Although numerous studies provide support for the occurrence of the actor-observer bias in negative performance situations (cf., Bernardin, 1989; Jundt and Hinsz, 2002), there has been little examination of the actor-observer bias when performance is good, providing little evidence to resolve the contradiction in predictions between the self-serving and actor-observer biases.

Leader-Member Exchange (LMX) theory (Dansereau et al., 1975) may help to resolve this contradiction, in that leaders may be more likely to make attributions for in-group members' performance that are similar to attributions that they would make for their own. In other words, the self-serving bias may extend from leaders to their in-group followers (Davis and Gardner, 2004). Simultaneously, leaders' attributions for out-group members may still be subject to the actor-observer bias.

The purpose of this study was to explore the relationship of LMX status with the self-serving and actor-observer biases. Two additional goals of the current study were to obtain both the leader's and the member's point of view about LMX status and to examine both positive and negative performance incidents. The following section summarizes the research related to attributional biases and details how leader-member exchange theory may assist in resolving the contradictory predictions inherent in the self-serving and actor-observer biases. A survey study exploring attributional biases on the basis of LMX status is then described, and the results and their implications are then discussed.

CONCEPTUAL BACKGROUND

The foundational assumption of attribution theory is that people engage in cognitive activity, developing explanations for events (Kelley, 1967). The resulting attributions can be classified by their presumed locus of causality, either internal (ability and effort) or external (task difficulty and luck) to the actor (Weiner et al., 1971). Dyadic relationships are a primary focus in attribution theory (Heider, 1958), and Green and Mitchell (1979) specifically address the leader-member dyad in their suggestion that leaders' perceptions of the cause of followers' behavior influence the leaders' future behaviors. Subsequent research has upheld this prediction (Gibson and Schroeder, 2003; Swift and Campbell, 1998).

The attribution process, however, is not completely objective, as biases can occur. One such bias is the self-serving bias (Kelley, 1967), which occurs when actors take credit for their success and blame situational factors for their failure (McAllister et al., 2002; Miller and Ross, 1975). This bias extends to cross-cultural work relationships (Peterson et al., 2002). In the actor-observer bias, actors make internal attributions for others' performance and external attributions for their own performance (Floyd, 2000; Jones and Nisbett, 1972). The focus of exploration of the actor-observer bias in work settings, though, has primarily been on poor performance (Bernardin, 1989; Martinko and Gardner, 1987; Nelson and Beggan, 2004). In addition, the attributions that have been examined in these studies have not been consistent across studies, making comparisons of results difficult.

The previously-mentioned contradiction in predicted attributions under these biases may be resolved by Leader-Member Exchange (LMX) theory, with its supposition that leaders do not relate to all members in the same manner (Dansereau et al., 1975). Subordinates with whom leaders have high-quality relationships are known as "in-group" members, while subordinates with whom leaders have lower quality relationships are known as "out-group" members (Dansereau et al., 1975). The results of relationship quality are generally positive for in-group members (Erdogan and Liden, 2002) and negative for outgroup members (Duval and Silvia, 2002; Vecchio and Gobdel, 1984).

One positive outcome for in-group members is the attributions that their supervisors make about them. A few prior studies have explored supervisor attributions when subordinate performance is positive, consistently finding that supervisors were more likely to make internal attributions for the effective performance of their in-group subordinates and to make internal attributions for the ineffective performance of their out-group subordinates (Gibson and Schroeder, 2003; Heneman et al., 1989; Swift and Campbell, 1998). However, all of these studies compared supervisor attributions across LMX status (i.e., "in" versus "out") but not within groups (i.e., "in" positive performance compared to "in" negative performance), thus their data do not permit an evaluation of attributional biases.

In-group dyads are expected to have "shared conceptions of their environments" (Scandura et al., 1986: 580). Schneider's Attraction-Selection-Attrition (ASA) theory argues that similarity, or homogeneity, on psychological attributes attracts people to others by reducing the psychological distance between individuals (Schneider, 1987). In fact, people who sense that they do not fit the group will leave the organization (Pendergrass, 2002). LMX status, therefore, as a psychological attribute, may result in reduced psychological distance for in-group dyads and greater psychological distance for out-group dyads. This implies that supervisors may make attributions about in-group members that are more like self-target attributions and may make attributions about out-group members that are more like other-target attributions. If both in-group supervisors and in-group subordinates are making self-target types of attributions, we would expect in these dyads to see both leaders and members displaying the self-serving bias when making attributions about in-group members' work performance. This expectation is reflected in the following hypotheses:

H1. When performance is positive, supervisors will make greater internal than external attributions for in-group subordinate performance.

H2: When performance is negative, supervisors will make greater external than internal attributions for in-group subordinate performance.

H3: In-group subordinates will make greater internal than external attributions for their own positive performance.

H4: In-group subordinates will make greater external than internal attributions for their own negative performance.

These hypothesized relationships are diagrammed in Figure I, illustrating our expectation that supervisors and subordinates would demonstrate the same attributional pattern when explaining the positive and negative performance of in-group subordinates.

A different situation occurs when supervisors and subordinates are not psychologically similar to each other (Schneider, 1987), a phenomenon associated with differing perspectives on workplace occurrences, as in the actor-observer bias. Therefore, from the supervisor's perspective, we expected evidence of the actor-observer bias, such that supervisors, as observers, would make internal attributions for both positive and negative out-group subordinate performance, as stated in the following hypotheses:

H5: When performance is positive, supervisors will make greater internal than external attributions for out-group subordinate performance.

H6: When performance is negative, supervisors will make greater internal than external attributions for out-group subordinate performance.

Because of their psychological distance from their supervisors, we expected evidence of the actor-observer bias such that subordinates, as the actor, would make external attributions for both their positive and negative performance, as stated in the following hypotheses:

H7: Out-group subordinates will make greater external than internal attributions for their own positive performance.