A large strand of the economic literature proposes to investigate how the structure of attorney fees and the allocation of litigation costs affect the efficiency of legal systems. A significant effort has been particularly devoted to explore the performance of two major cost allocation system: the British rule and the American rule. Under the British rule, the party who loses in court pays (part of) the attorney fees of the winning party. In contrast, the American rule provides that each party is responsible to pay its own attorney fees. In continental Europe, attorney compensation that is contingent on success at trial is increasingly being used in combination with the British fee-shifting rule. Although the contingent fee and the fee-shifting mechanism have been widely studied separately, few is known on their performance when they coexist in the same legal system.

In this study, we implement an original experiment to explore how the structure of attorney fees (fixed or contingent on success at trial) and the allocation of litigation cost (with or without fee-shifting) affect the performance of legal systems. More precisely, we focus our attention on three performance criteria: the effort of the attorney in charge of the case, the deterrence of frivolous suits (i.e. suits that should not have been filed based on their merit) and the promotion of meritorious cases. Our experimental setting involves three games.

In our experimental framework, participants can either take the role of plaintiff or the role of attorney. Subjects are randomly assigned to an eight-person experimental legal market comprising four attorneys and four plaintiffs. Participants interact exclusively within this group throughout the entire experiment. The experiment is divided into three games. In the first game, participants in the role of plaintiffs take part in a real-effort task. Their performances in the task determine their own initial endowments. The second game consists in a repeated litigation game. In the litigation game plaintiffs incur an exogenous loss and can decide to hire an attorney to attempt a full recovery of this loss. Each period is a different case, where the amount of the loss and the merit of the case (probability to recover the loss in case of trial) vary. Attorneys who have been hired in a suit decide of a level of effort to provide. This level of effort affects the probability of the suit to be successful. Finally, the third game of our experiment is a standard lottery game that provides us with an individual measurement of risk attitude.

We implement four treatments in a 2x2 factorial design. Between our treatment, two main dimensions vary. First, we vary the nature of attorney's fee that can either be fixed or contingent. In the fixed-fee treatments, fees are expressed in ECU and are not conditional on the outcome of the trial. In the contingent fee treatments, fees are expressed as a share of the recovered loss, and are therefore conditional on the outcome of the trial. Second, we vary the allocation of litigation cost, that can either allow for fee-shifting or not. In the fee-shifting treatments, an additional amount is added to the plaintiff's payoff if the suit succeeds, i.e. the defendant incurs part of the plaintiff's attorney fee. If the suit fails, an extra cost is withdrawn from the plaintiff's payoff, i.e. the plaintiff incurs part of the defendant's attorney fee.

We establish the following findings. (1) Plaintiffs' willingness to enter trial is the highest in the contingent fee-shifting setup, where attorneys have a clear incentive to provide effort. This finding entails two phenomena. While the risk of losing the case prevents plaintiffs to suit meritorious cases in the other treatments, it is less the case in the contingent fee-shifting setting. However, plaintiffs tend also to file a suit when the case is non-meritorious. Limiting the presence of frivolous suits is a key property of an efficient legal system. (2) Surprisingly, we observe that the fee-shifting mechanism, which only concerns plaintiffs' payoff, affects attorney's effort. More precisely, attorneys provide less effort when a fee-shifting mechanism is introduced in a fixed fee setting. In sharp contrast, the effort of attorneys increases when a fee-shifting mechanism is introduced in a contingent fee setting. We discuss these findings and provide a comparison of the efficiency of our four experimental legal markets.