Learning and Corruption on Monitoring Chains

“But who will guard the guardians?” I revisit this age-old question under the following assumptions: (i) guardians are devoid of ethical motives and have quasilinear preferences, (ii) guardians monitor one another through a monitoring chain, and (iii) any two consecutive guardians in the chain can bargain away “inefficient” punishments through corruptive arrangements. Under these assumptions, monitoring is impossible unless rewards or punishments are unbounded. When material incentives are bounded and local corruption is feasible, the answer to the initial question is “no one.”