Knowing corruption is an equilibrium sustained by weak enforcement (module 3), and knowing how to measure it (module 4), we can finally ask the practical question: what actually reduces it? The honest answer is that most anti-corruption fails, but a growing body of rigorous evidence identifies what works, for what kind of corruption, and under what conditions. This module is that evidence.
Top-down audits
The most direct approach is external audit with consequences. Olken's Indonesia experiment (module 4) showed that a credible audit threat cut missing expenditure in road projects — top-down monitoring works on theft that an outsider can detect. But audits only bite if findings lead to sanctions; an audit that reports irregularities to no consequence (the pattern in many countries' annual audit reports) changes nothing. The power of audit is realised only when it is coupled to enforcement or to information that mobilises a sanction.
Audits plus information plus elections (Ferraz-Finan, 2008)
Brazil randomly audited municipal governments and released the results publicly. Ferraz and Finan found that where audits revealed corruption, mayors were significantly more likely to be voted out — and the effect was much larger where local radio could broadcast the findings. The mechanism is the accountability chain: an audit detects the corruption, information transmits it to voters, and the election sanctions it. Remove any link — no audit, no broadcast, or no electoral accountability — and the corruption is unpunished. Detection alone is not enough; it must reach someone who can and will act.
Bottom-up community monitoring
The opposite approach asks citizens to monitor their own services. The evidence is mixed and conditional. Olken found community monitoring weak against theft of materials (citizens can't easily verify what's buried in a road, and face a collective-action problem). But Björkman and Svensson (2009) found that community monitoring of health clinics in Uganda — giving citizens scorecards and facilitating engagement with providers — improved attendance, service quality, and even child health. The difference: bottom-up monitoring works when citizens can observe the thing being monitored (whether the nurse is present), when the collective-action problem is solved by facilitation, and when they have a channel to act. It fails when the corruption is hidden or citizens are powerless.
Technology and the removal of discretion
Take the official out of the loop (Muralidharan-Niehaus-Sukhtankar, 2016)
The most promising recent lever is technology that removes the discretionary human gatekeeper. In a large randomised rollout in India, paying welfare and workfare benefits through biometric smartcards — which authenticate the recipient directly and pay them, bypassing the officials who previously skimmed — substantially reduced leakage and delays, with recipients getting more of what they were owed. The principle generalises: e-procurement, electronic tax filing (eTIMS/iTax), direct digital transfers, and automated systems reduce corruption by eliminating the discretion and face-to-face contact where bribes and skimming happen. You cannot bribe an official who is no longer in the transaction.
But technology is not magic. Systems can be captured (the biometric database manipulated), exclude legitimate beneficiaries (failed fingerprints, no connectivity), and simply relocate corruption to whoever controls the system. The evidence supports technology that genuinely removes discretion over a transaction, not technology layered on top of an unchanged process — the isomorphic-mimicry warning from module 2 applies to digital reforms too.
Wages, deterrence, and the big push
Two more levers, both with caveats. Higher wages (the efficiency-wage idea — pay officials enough that the bribe isn't worth the risk of dismissal) have mixed evidence: necessary perhaps, but not sufficient, and they work only alongside real monitoring and a credible threat of losing the job. And because corruption is an equilibrium (module 3), incremental deterrence often fails to move it — which is the argument for occasional coordinated 'big-push' campaigns that try to flip expectations all at once. The synthesis (Olken-Pande): there is no single cure; the effective strategy combines detection (audits, technology), transmission (information, free media), and sanction (enforcement, elections), matched to the specific corruption and its measurability.
Exercise
A national programme pays a monthly stipend to elderly citizens through local officials, and large sums are being skimmed — recipients report receiving less than they are owed, and some 'ghost' beneficiaries exist. (1) Diagnose which anti-corruption tools fit this specific problem and which don't. (2) Explain why simply auditing and publishing findings might not reduce the skimming on its own here. (3) Make the case for a biometric direct-payment reform, citing the evidence, and state two risks it carries. (4) Design a combined strategy across detection, transmission, and sanction, and explain the sequencing.