Does widespread collusion undermine the case for using peer-assessment schemes with assessed group work?

Peer assessment is one way of adjusting assessed group work grades into individual marks that may better reflect the different contributions of each student. However, people question the extent to which the peer-assessment scores reflect the actual contribution of each team member. One factor that might cause them to diverge is the presence of collusive behaviour. This paper develops a method to formally quantify the extent of this problem: Firstly, we identify student pairs in groups that award each other high peer-assessment scores (mutual high scoring - MHS). Secondly, we exclude cases where MHS appear to reflect higher actual contributions by these two students. Applying this method to peer-assessment data from an assessed group work assignment of an 1st year applied economics module, we find evidence of potential collusion in only 4-7 percent of all student pairs. However, the share of groups that contain at least one pair of students that were potentially colluding was quite high at 35-55 percent. This suggests that although collusion was not widespread across the cohort, it did occur between two team members in a relatively large number of the groups. Therefore, when implementing peer-assessment schemes in assessed group work, the potential issue of collusion needs to be considered by tutors.