We construct a one-shot corruption game with three players, a briber who can decide to bribe or not, an official who can reciprocate or not and an inspector who can decide to inspect or not. We employ four penalties that can be distributed asymmetrically, making it possible to punish bribing and bribe-taking as well as reciprocating and receiving reciprocation to different degrees. Penalties apply if corruption is detected. The probability of detection is endogenised, as it depends on inspection. The model differs from other inspection games in that the offence (corruption) can only be completed through a joint effort of the two offending players. This leads to surprising results, especially in conjunction with asymmetric penalties. First, in contrast to Tsebelis' results, we find that, with endogenous detection, higher penalties do reduce the overall rate of offence. Second, this result holds only if the penalty for reciprocating on the official is raised. Surprisingly, and unlike other asymmetric penalty prescriptions in the corruption literature, higher penalties on the briber have the opposite effect. They may reduce the probability of bribery, but they also increase the probability of reciprocation to the extent that the overall probability of reciprocated bribery is increased.

We construct two variants of a three-player one-shot corruption game, one in which reporting on bribers is cumbersome and one in which it is rewarded (profitable). Both variants feature a briber who can bribe or not, an official who can reciprocate or not and an inspector who can inspect or not. In the first variant, the official accepts the bribe by reciprocating or simply rejects the bribe by choosing not to reciprocate. In the second variant, the official either accepts and reciprocates or rejects and reports the bribe. Under successful inspection, offending players receive separate penalties, which can be varied asymmetrically. Under plausible assumptions about the values of payoff parameters, we obtain a mixed-strategy Nash equilibrium in both variants, akin to Tsebelis’ inspection game. We obtain two interesting results. First, marginally changing the penalties moves the equilibrium probabilities in both games in the same directions, suggesting robustness of the model. We find that larger penalties on the briber increase the overall probability of reciprocated bribery, that is, corruption, while larger penalties on the official decrease corruption. Second, when comparing the two models, we obtain the surprising result that the probability of reciprocated bribery (corruption) is higher in the variant where the official is rewarded for reporting on the briber. JEL: K42, H00, C72, O17

While folk theorems for dynamic renewable common pool resource games sustain cooperation as an equilibrium, the possibility of reverting to violence to appropriate the resource destroys the incentives to cooperate, because of the expectation of conflict when resources are sufficiently depleted. In this paper, we provide experimental evidence that individuals behave according to the theoretical predictions. For high stocks of resources, when conflict is a highly costly activity, participants cooperate less than in the control group, and they play the non-cooperative action with higher frequency. This comes as a consequence of the (correct) anticipation that, when resources run low, the conflict option is used by a large share of participants.