Relative Leistungsbewertung im dynamischen Kontext — Eine Analyse der Kollusionsproblematik bei wiederholter Delegation

Summary The suitability of relative performance evaluation to achieve better solutions to multiple agent problems is widely known in economic literature. But, unfortunately, relative performance evaluation is susceptible to collusion. Colluding agents can cause considerable damage to the principal. Most authors hence conclude the inappropriateness of relative performance evaluation when collusion is possible. Such an argumentation implicitly assumes that if collusion can occur, collusion will occur. But, as Budde/Göx/Luhmer (1998) have demonstrated using Groves mechanisms, this assertion is not true in general. One goal of this paper is to extend these results to relative performance evaluation. It turns out that in one period models collusion is in fact an unlikely phenomenon. On the other hand, Kunz/Pfeiffer (1999) recently pointed out that the results of Budde/Göx/Luhmer (1998) are no longer valid if a dynamic context is under consideration. In order to get some insight into what conditions then favour collusion, a simulation study was conducted. The results of this study suggest collusion to be in many cases a rational behaviour of the agents. Especially, if the agents manage to have ‘good’ (i.e. collusive) reputation and if their planning horizons are sufficiently large, collusion is indeed a winner strategy.