Two decades of donor effort in anti-corruption have produced only very limited success. We are still far from confident about ‘what works’. Most analysis has been on WHAT donors have been trying to do. But could part of the answer be in HOW they work?