Group activities have become a common pedagogical technique in accounting classes. Besides contributing effectively to the learning process, group activities are an excellent vehicle to assess and improve a student's interpersonal skills (Accounting Education Change Commission 1990). These interpersonal skills are often key determinants in a person's future professional success (Messmer 1998).

     By their nature, group activities introduce an evaluation problem, because it is hard to measure an individual's contribution to the group's final product. This problem of "social loafing" (also called the "free rider" problem) is commonly combated through the assignment of individual grades for group work (Harkins, Latane, and Williams 1980). The assignment of individual grades for group work can be accomplished in several ways but a common method is the use of peer evaluations where each group member assesses the individual contributions of his or her peers (Rinehart and Berry 1999). Peer evaluations are particularly effective in reducing the inherent limitations of instructor evaluations of individual contributions. Since a class may be comprised of several small groups, the instructor may have a limited time to spend assessing each group. Furthermore, many group activities may be conducted outside of class, unmonitored by the instructor. Because group members should be intimately involved in the activities of the group, each group member should have a better understanding of the group dynamics than the instructor. Peer evaluations attempt to access this rich source of information for evaluation purposes.

     The adoption of peer evaluations requires the instructor to decide whether self-evaluations should be part of the peer evaluation process. Two strong arguments support the use of self-evaluations. First, the use of self-evaluations may provide quantitative and qualitative increases in the information that can be used in the evaluation process. In a four-member group, peer evaluations without a self-evaluation provides the instructor with twelve evaluations, whereas including self-evaluations provides the instructor with sixteen evaluations, a 33% increase in the amount of data available. Potentially, the self-evaluator is the best source of information about his or her contributions and if this information is accurately revealed, the instructor receives high quality information. Secondly, the use of self-evaluations may enhance the perceived fairness of the evaluation process, because the self-evaluator is allowed an opportunity to comment on his or her own contributions (Kilpatrick, Linville, and Stout 2000). Research shows that if a process is perceived as fair, the outcome of that process, whether positive or negative, is more likely to be accepted (Folger and Konovsky 1989). On the other hand, self-evaluations may introduce bias into the evaluation process. Students may deceive themselves about their own contributions or may behave opportunistically in order to increase their personal grades. The purpose of this paper is to determine if bias exists in self-evaluations.

     As reported by Sullivan and Hall (1997), research on this topic has generally found agreement between self-evaluations and the evaluations of others. However, enough studies have found a lack of agreement between self-evaluations and the evaluations of others that a consensus on this issue cannot be formed. Boud and Falchikov (1989), in their review of fifty-seven articles on this topic, found seventeen (29.82%) studies in which self-evaluators overestimate their performance and eleven (19.30%) studies in which performance is underestimated. They identified eight methodological problems (discussed in the next section) that may have led to the differing results. Only one of the fifty-seven studies examined by Boud and Falchikov (1989) involved business students (Burke, 1969) and none involved accounting students.

Methodology

     Data was collected over four semesters from students in a senior-level accounting class at a mid-sized public university. This results in 85 self-evaluations and 260 peer evaluations. Small group activities were an integral part of the class pedagogy. Some class time was allocated to group activities although completion of the group projects required time outside of class. Generally, groups of four were maintained.

     Students knew from the first day of class that peer evaluations would be required, but they did not know the format of the final peer evaluation instrument. All final evaluations were completed at the same time, either on the last day of regular class or at the final exam depending on the semester. Students were not given an opportunity to communicate with one another while completing the peer evaluation instrument so that "game playing" with the process could be kept to a minimum. Opportunities for peer assessment in order to improve individual skills were given during the semester. The final peer evaluation instrument was for the purpose of assigning final grades.

     Students knew that the peer evaluations were confidential so that they could be totally honest about a peer's performance without fear of reprisal. Students evaluated themselves and their group peers on several specific criteria using a five-point Likert scale and were asked to provide an overall evaluation using a scale of 0.0 to 4.0 with 0.1 increments available. Space for additional written comments was also provided. Evaluators were instructed to give each student in the group a unique overall score. The method for breaking ties between two equally performing students, if any, was left to the evaluator's discretion. Whether recognized by the students or not, this method forces the evaluator to rank each student's contributions to the group. Further, it prevents an evaluator from giving each group member the same score and essentially opting out of the evaluation process while in turn providing the instructor with little or no information.

     Boud and Falchikov (1989) identify eight methodological concerns common to earlier studies of self-evaluations. These methodological concerns are summarized below along with parenthetical comments about how the methodology of this study addresses the concern.

1. The self-evaluator and other evaluator use different rating criteria. (In this study, the self-evaluator and the peer evaluator use identical evaluation instruments.)
2. Differences between self-evaluators and other evaluators are measured in terms of magnitude or frequency but seldom both. (In this study, both the magnitude of difference and the frequency of difference are reported.)
3. Explicit criteria are often missing. (In this study, five detailed criteria are provided on each evaluation instrument.)
4. Studies seldom include different groups taking a class over a period of time or following the same group over time. (This study includes subjects from the same class over four different semesters.)
5. Evaluation criteria are sometimes based on effort alone. (In this study, effort is not one of the explicit criteria on the evaluation instrument.)
6. Ill-defined scales are often used. (In this study, each evaluation instrument uses a five-point Likert scale with each point on the scale clearly defined.)
7. Self-evaluations are often based on expectations of future performance before learning has occurred. (In this study, the evaluations are done at the end of the semester.)
8. Studies often fail to provide a grading norm, which leads to subjective grading. (In this study, the specific criteria on the evaluation instruments provide a norm against which students are evaluated.)

Findings

     The 85 self-evaluations collected are compared to the 260 peer evaluations to determine if any bias exists in the self-evaluations. Little evidence of bias is found in the results. The most interesting results are listed below with a discussion following each point.

1. No overall bias in self-evaluations exists.

Each self-evaluation can be compared to at least three peer evaluations to provide a measure of the magnitude of the differences in the evaluations. In the total sample, the difference between the average self-evaluation score and the average scores given to the same person by the group peers differ by approximately 0.5%, an obviously immaterial amount. In the four individual classes, the percentage difference between the average self-evaluation and the average scores given to the same person by the group peers are 1.40% understatement (by the self-evaluator), 1.95% understatement, 1.65% overstatement, and 2.12% overstatement. Taken as a whole, self-evaluations are remarkably consistent with the peer evaluations and suggest that self-evaluators provide a realistic assessment of their contributions.