Corruption is a significant problem in much of the developing world. Corruption acts like a tax, adding to the cost of providing public services and conducting business. Often, though, the efficiency costs of corruption can be far worse. Indeed, it has been suggested that corruption may be a major contributor to the low growth rates of many developing countries.

Despite the importance of the problem, there is relatively little evidence, and therefore relatively little consensus, on how to best fight corruption. One view is that the best approach is a straightforward combination of greater auditing of public officials and greater punishments for officials who are caught. In practice, however, if the auditors themselves are corrupt, such a policy might just encourage officials to bribe the auditors, and thus perpetuate the corruption.

Recently, the development community has been taking a different tack: grass-roots monitoring of officials by local community members. This approach is now regarded as critical not only to reducing corruption, but to improving public-service delivery generally. For example, the World Bank’s 2004 World Development Report was devoted entirely to the idea of “putting poor people at the center of service provision.” The idea is that if community members are the people who benefit from a successful program, they may have better incentives to carefully monitor them than government bureaucrats. Of course, this approach has potential drawbacks as well—for example, monitoring public projects is a public good, so there may be a serious free-rider problem.

How do these different approaches fare in practice? To test this, starting in 2003 I designed and conducted a randomized, controlled field experiment in 608 Indonesian villages. At the time the study started, each village was about to start building a road as part of a nationwide village-level infrastructure project funded by a World Bank loan.

Villages were randomly assigned to receive different anti-corruption approaches. To test top-down auditing, some villages were told before construction began that their finished project would be audited by a central government agency. The audits were subsequently conducted as promised.

To test the impact of community participation in monitoring corruption, I developed two different experiments around “accountability meetings,” the village-level meetings where project officials account for project funds. In one experiment, hundreds of invitations to the meetings were distributed throughout the village. In the second experiment, an anonymous comment form was distributed along with the invitations, allowing villagers to relay information themselves anonymously. The forms were collected before the meetings in sealed drop boxes, and the results summarized at the meetings. Both interventions successfully raised grass-roots participation levels—and the forms generated hundreds of comments about the project, both positive and negative, in each village. But were they more effective than the audits in slowing corruption?

To determine this, I needed a measure of corruption. I assembled a team of engineers and surveyors who, after the roads were completed, dug core samples in each road to estimate the quantity of materials used, surveyed local suppliers to estimate prices, and interviewed villagers to determine the wages paid on the project. I estimated what each project actually cost to build, and then compared this estimate with what the village reported it spent. Missing expenditures averaged about 24 percent across all the villages in the study.

But in villages associated with the audit experiment, missing expenditures were substantially lower—by about 8 percentage points. This reflected reductions in both unaccounted-for materials and in unaccounted-for labor expenditures. (Interestingly, the number of jobs given to family members of project officials increased in response to the audits, which suggests that alternative forms of corruption may be substituted.) While the auditors’ reports were positively correlated with my independent engineering survey, in the vast majority of cases the auditors did not find the “caught-red-handed” evidence that could be used to prove criminal malfeasance. This may help explain why almost 20 percent of expenditures were still unaccounted for even in villages knowing they would face an external government audit.

By contrast, the participation experiments were associated with statistically insignificant reductions in missing expenditures, even though they raised community participation in the monitoring process. Villages in the participation treatments were more likely to openly discuss corruption problems and to take serious action to resolve them. However, the magnitude of these changes in behavior at the meetings was small. Nonetheless, the small overall effects on missing expenditures in response to increased participation mask several interesting results. First, the invitations experiment substantially reduced missing labor expenditures (community members may have had a strong incentive to monitor wage payments), even if it had no effect on the larger problem of missing materials. Second, the anonymous comment form treatment did reduce missing expenditures, but only when they were distributed entirely through village schools, bypassing the village government and preventing village elites from channeling the forms to their supporters.

In the end, the benefits of the audit approach were most encouraging (although to avoid a new pattern of corruption over time, auditors would have to rotate frequently). The results showed that for the participatory approach to succeed, care must be taken to give villagers a strong incentive to monitor and to prevent village elites from capturing the process.