Does Power Corrupt? Evidence from random recruitment of police in Papua New Guinea

Research Method

Blocked recruitment lottery with 45 police candidates and 1,383 citizen respondents in 39 villages. Behavioral games and surveys.


Bougainville, Papua New Guinea




Bougainville Police Service, New Zealand Police, J-PAL (funders)

Research Question

How does giving people the power to police their community change their behavior and attitudes?


Governments frequently face a tradeoff when delegating powers to subsidiaries. On the one hand, giving distributive and punitive powers to local actors may improve the efficiency of service provision. On the other, empowering local actors might enable them to abuse their position for private gains. While a large literature has studied the effects of different incentive structures on agents already in situations of power, very few can isolate the effect of conferring power on actors per se. I leverage the random recruitment of candidates into the police force in Papua New Guinea in early 2016 to examine whether power itself produces a corrupting or virtuous effect on their behavior. Empirical predicitions arise from a theory of how delegation affects the impunity of government agents tasked with allocating public goods. I measure the behavior through random allocation dice games with candidates that were and were not hired into the police. These games randomly vary whether dishonesty can be observed by a community member of the candidate, providing an estimate of the importance of community monitoring. I also measure the degree to which attitudes toward the state and perceptions of power and police impunity differ between community members and police candidates in the recruitment lottery.