Conventional wisdom tells us corruption is bad (except for a few people). The evidence is surprisingly mixed (see old post), but that has not deterred every government or international agency to at least pretend to do something about it. Eradicating or attenuating corruption rarely works, maybe it is because the approach is wrong.
Martin Dufwenberg and Giancarlo Spagnolo verify an innovative angle to attack bribing, as proposed by Kaushik Basu and widely discussed in the Indian press: make bribing legal, but levy heavy fines on accepting bribes. That is somehow similar to the situation in some countries where consuming some drugs is legal, but their commerce is not. The conundrum the proposal tries to resolve is that briber and bribe-taker are usually partners in crime and unlikely to report the activity. By guaranteeing immunity to one party with return of the bribe, it is more likely that the corruptive transaction will be reported. But it turns out to optimize it, immunity should only be granted to those who report a bribe. Indeed, this gives stronger incentives to report. Also, granting immunity to any briber would have made it morally acceptable to bribe. To this I would have added that one also offer the repented briber a commission on top of the returned bribe. Then it would become profitable to report bribes, even when the requested service is ultimately not provided.