Monitoring corruption: Evidence from a field Experiment in Indonesia
In this article Ben Olken compares the efficiency of two much proposed methods to reduce corruption: increasing audits by government officials and increasing grassroots participation. The results strongly advocate for the first but find no effect of the latter. This article is one of the first carefully designed randomised experiments on the economics of corruption. It constructs a well defined measure of corruption and has a clear policy implication: increasing audits of local public work projects is a very cost effective way of reducing corruption.
Olken used a dataset from 608 rural Indonesian villages benefiting from a World Bank funded development project. Villages received large grants intended for road building, which on average doubled villages’ budgets. To ensure a proper use of the funds, the project included a series of village level accountability meeting, open to all, but usually attended by less than 2% of villagers. Corruption mainly occurred in two ways: inflating the price of materials needed for construction in collusion with suppliers, and manipulating wage payments, which is easy to do since villagers contribute unpaid labour to the projects. The key measure is the difference between expenditure claimed by the village and an independent study conducted by construction engineers estimating what villages actually spent on the project, after observing the constructed road.
The author designed two main interventions, and randomly assigned villages to treatment and control groups. The first intervention was to increase the probability of audits from 4% to 100%. The second was to increase participation in accountability meetings by sending invitations to villagers and asking them to submit an anonymous comment form on how satisfied they are with the project. Villages were told of their treatment group after the budget allocation but before construction began. Increasing the level of auditing from 4% to 100% diminished missing expenditures by about 30%. Whereas increasing grassroots participation had no effect on missing expenditures.
The cost benefit analysis of government monitoring shows huge benefits of an increase in auditing in such one shot interventions. Additional research would be interesting in order to determine the optimal level of auditing and punishment, and how to avoid collusion when auditors and audited parties interact repeatedly.
Citation: B. Olken, "Monitoring Corruption: Evidence from a Field Experiment in Indonesia", Journal of Political Economy 115 (2): 200-249, April 2007
The article can be found on the author’s webpage:
Del.icio.us
Facebook
Twitter
Digg
