Abstract: We employ an at-scale randomized experiment to evaluate how one of the world's largest citizen-led monitoring efforts, India's ``social audits'' initiative, impacted corruption and access to social protection in one of India's poorest states. A unique feature is the program's reliance on female auditors drawn from village self-help groups. Combining administrative data on program expenditures with survey data, we document how the nature of corruption varies across the two distinct schemes linked to India's workfare program. Submitting names of villagers who did not work was the main form of leakage in the public works scheme, whereas kickbacks dominated in the national housing scheme, which provides a conditional cash transfer for home construction. Audits reduced corruption in both programs, with differing implications for participants by program. Reduced corruption in public works was accompanied by lower levels of implementation and therefore less access -- survey data show the share of citizens participating reduced by more than 50% in audited communities. Conditional on participating, amounts received by citizens did not change. In contrast, audits reduced kickbacks paid to local leaders under the national housing program by 28% and as a result income retained by participants increased by 7%.

Sponsorship of an event does not constitute institutional endorsement of external speakers or views presented.