Monitoring Corruption: Evidence from a Field Experiment in Indonesia

Publication Details

Journal of Political Economy, April 2007, v. 115, iss. 2, pp. 200-249. Available From:

Link to Source
Benjamin A. Olken
Institutional affiliations
None specified
Grant-holding institution
None specified
East Asia and Pacific (includes South East Asia)
Agriculture and Rural Development, Public Sector Management
Rural Roads, Anti-corruption/ Governance
Rural Roads, Anti-corruption/ Governance
Equity Focus
None specified
Evaluation design
Randomised Control Trials (RCT)
Journal Article


This study compares the impact of top-down and grassroots monitoring on corruption in the context of village-level road construction projects funded by the Indonesian government’s Kecamatan Development Project in East and Central Java provinces. Top-down monitoring is carried out by higher-ranking public officials, whereas grassroots monitoring involves direct beneficiary participation. Although both strategies aim to deter corruption, top-down monitoring may be ineffective if high-ranking monitors are themselves corruptible, whereas grassroots monitoring is subject to possible free-rider problems.
The basic methodology is a randomised controlled trial (RCT). To evaluate top-down monitoring, project officials in 283 villages were informed that their projects would be audited by the government audit agency. Audits were carried out 4 months after road construction began for a subsample of one treatment village in each sub-district and 7 months subsequently for all audit treatment villages. The second treatment, designed to test grassroots monitoring, involved distributing invitations for community members in 199 villages to participate in meetings where project officials account for expenditures. The third treatment, also designed to test grassroots monitoring, involved distributing meeting invitations with anonymous comment forms for opinions about the project. This treatment was administered in 202 villages. Invitations were distributed by community officials or through schools, with distribution method randomly assigned by village.
Corruption is measured as the difference between reported project expenditures and an independent estimate of actual spending, on the assumption that “missing” funds have been diverted for corrupt purposes. The actual spending estimate is obtained through an engineering survey of project infrastructure, worker surveys to measure wages and hours worked and price surveys to measure the cost of materials. Reported expenditures are measured using project reports filed with the government. Using these data, the authors identify impact by estimating ordinary least squares (OLS) regressions of the percentage of missing funds on treatment dummies.

Main findings

This study’s main finding is that top-down monitoring through government audits significantly reduced overall missing expenditures, whereas grassroots monitoring had weak and insignificant effects on overall missing expenditures. The government audit treatment, which effectively increased the probability of auditing from 4 percent at baseline to 100 percent, was associated with a statistically significant reduction in missing expenditures of 8.8 percentage points. Nonetheless, even with the audit treatment, about 20 percent of funds remained unaccounted for. Although auditors were able to detect evidence of corruption, in many cases these findings related to procedural issues that did not provide sufficient grounds for prosecution.
The grassroots monitoring treatments were associated with decreases in total missing funds of between 1.5 and 3 percentage points, depending on specification. These findings were not statistically significant. However, the invitations treatment was associated with a statistically significant reduction of missing expenditures for labour of between 14 and 22 percentage points, depending on specification. This reduction was offset by no change or, in some specifications, moderate increases in missing funds allocated for materials. The authors attribute this finding to improved monitoring of expenditures through increased participation in community meetings by labourers with strong private incentives to reduce missing labour spending, but weak private incentives to monitor expenditures on materials.
This study also finds that distribution of meeting invitations with comment forms through schools generated significant reductions in missing expenditures, whereas distribution by community officials had no significant effects. This finding suggests that community officials who benefited from corruption selectively distributed invitations with comment cards to their supporters or other beneficiaries of corruption, thereby reducing the impact of grassroots monitoring. Conversely, distribution through schools was effectively random, ensuring that a wider range of community members were invited to meetings and submitted comments.

Scroll to Top