Raul Sanchez
VC Cinéma le Miroir
Centre de la Vieille Charité
2 rue de la Charité
13002 Marseille
Timothée Demont: timothee.demont[at]univ-amu.fr
Roberta Ziparo: rziparo[at]gmail.com
It is well-known that many state officials sell state assets for private gain. However, their criminal activities are often organized within the state hierarchy. We develop a system to gain access inside the traffic police agency in the DRC in order to uncover the organization of criminal activities. We find that criminal behavior by the police is institutionalized: drivers, police officers, and their commanders developed well-defined, stable, implicit contracts to sustain illicit transfers, enforced through repeated interaction. Such agreements are linked to weaker law enforcement, but better traffic management (through tips for police officers). We also observe the flow of illicit revenue. Police officers and their commanders derive respectively from these arrangements 92% and 99% of their total income. Station commanders keep 75% of the total illicit income generated. We use qualitative data, and simple theory of organizations to explain why such implicit contracts prevail. This framework implies that a commonly advocated policy to combat corruption, higher wages for police officers, could have subtler effects than expected: commanders would likely absorb a large share of the wage, and the residual wage of police officers may reduce traffic management. We double the wage of randomly selected police officers. Their police commanders extract a large share of the wage increase, with a positive marginal "tax" rate. Consistent with a simple income effect, police officers increase consumption and decrease effort, increasing the propensity of traffic jams. We also randomize the terms of the implicit contracts between commanders and police officers, linked to the police officers' efforts to arrest cars. Steeper incentives increase the rate of arrests at the intersections, but lead mostly to higher illicit revenues at the police station. The paper illustrates the value of combining deductive methodology with inductive approaches from other social sciences to learn about phenomena that are usually organized in secrecy.