The working definition in common use is "the abuse of entrusted power for private gain." The weight sits on entrusted: corruption only arises inside a principal–agent structure — someone decides on another's behalf, and that other party cannot see how the decision was made.
Klitgaard's Controlling Corruption (1988) offers a diagnostic formula: corruption = monopoly + discretion − accountability. In engineering terms: a single node provides the service, its handling of requests has large areas of undefined behaviour, and there is no logging or audit — abuse is the system's steady-state output, not an intermittent fault. The cost is also more than money changing pockets; the dearer part is distorted choice, favouring new construction that can be skimmed over maintenance that cannot.
| Form | Typical setting | Main cost |
|---|---|---|
| Administrative corruption small sums, service counter | Permits, customs, fines — paying to speed up existing rules | Turns processing speed into a commodity, hitting hardest those most dependent on public services |
| Political corruption large sums, decision layer | Kickbacks and cartel carve-ups in major procurement and concessions | Project selection is distorted; sums are concentrated while harm is diffuse, so no one has an incentive to pursue it |
The "grease" case at its strongest: Leff (1964) and Huntington (1968) argued that where regulation is rigid, a bribe effectively puts a price on a scarce permit, so the party willing to pay most — often the most productive — gets it first. Not optimal, but better than everyone jamming.
The counter-case at its strongest: Kaufmann and Wei (1999) identified the flaw — red tape is endogenous. If delay can be sold, whoever controls approvals has an incentive to manufacture delay. Their firm surveys found that firms paying more in bribes spent more management time dealing with officials and faced higher capital costs.
Trade-off: the grease case buys throughput on individual cases and sells the stability of the rules themselves — once a price mechanism exists, it reshapes what rules get supplied.
Equating corruption with envelopes of cash. Another form is not even illegal: campaign donations, lobbying and the revolving door are documented and disclosable, yet can produce the same tilt — the rule is drafted to favour one side, after which no amount of strict enforcement helps. Enforcement catches those who break rules; it cannot catch the step where the rule was bent.
A rent is a return above opportunity cost, usually arising from artificial scarcity: licences, quotas, tariff protection. Tullock's 1967 paper on the welfare costs of tariffs, monopolies and theft made the counter-intuitive point that the real waste is not the transfer itself but the resources burned competing for it — every process fighting for the same lock, with the CPU spent entirely on spinning. Krueger coined the term "rent-seeking" in 1974.
Cronyism ties the allocation of rents to personal relationships: acquaintance genuinely reduces default risk, at the cost of making entry non-competitive. State capture is the change in kind — Hellman, Jones and Kaufmann's 2000 World Bank working paper (no. 2444), drawing on the BEEPS firm survey, split corruption into paying to get around the rules and paying to shape the rules themselves. The latter is an investment: one successful piece of legislation pays out for years.
The case for strict limits at its strongest: if lawmaking is persistently occupied by the few organisations able to invest continuously, rules will drift systematically away from majority preferences, and elections cannot correct it — voters have no way to judge technical clauses.
The pluralist case at its strongest: legislators cannot know every industry, expert information has to come from those affected, and association and petition are basic rights. Prohibition does not remove influence; it pushes it into less visible forms. Disclosure beats prohibition precisely because it preserves observability.
Trade-off: strict limits buy neutral rules and sell the informational quality of legislation; permissive-plus-disclosure buys observability and sells the substantive inequality that unequal capacity to participate creates.
Treating rent-seeking as embezzlement. It can be entirely lawful: the research and legal spending poured into securing a favourable technical standard puts not one cent in a private pocket, and the social cost occurs anyway. The mirror-image error is to reason from "more regulation, more rents" to "just deregulate" — deregulating where capture has already happened simply transfers the rent to whoever claims the position first.
Following the diagnostic formula, interventions take three routes: raise the cost of being caught (dedicated agencies, whistleblower protection), cut the opportunity (digitised approvals, narrowed discretion, rotation), and change equilibrium expectations (replacing large numbers of officials at once).
The deeper disagreement is about what kind of problem this is. Mainstream design assumes a principal–agent model: that an uncorrupt principal exists. Persson, Rothstein and Teorell, writing in Governance in 2013 on the basis of interviews in Kenya and Uganda, argued that under systemic corruption this premise fails — the situation resembles a collective action problem: everyone knows everyone else will carry on as before, so unilateral compliance is not virtue but pure loss. That explains why adding one more oversight body so often changes nothing: the new body is drawn from the same pool and faces the same expectations.
| Approach | Mechanism | Cost / limit |
|---|---|---|
| Dedicated agency Singapore's CPIB (1952), Hong Kong's ICAC (1974) | Independent of the existing enforcement apparatus and reporting to the top; the ICAC added prevention and public education alongside investigation | Widely copied with wildly varying results — the structure travels, the independence does not |
| Wholesale replacement Georgia, 2004 | The traffic police (some 16,000 officers) were dissolved and replaced with a far smaller, far better paid patrol force | Those dismissed were not individually found at fault; depends on a post-revolution political window that ordinary politics does not offer |
| Prosecution-led Brazil's "Operation Car Wash," from 2014 | Driven by prosecutors and courts, using plea bargaining (delação premiada) to open up procurement networks | In 2021 the Supreme Federal Court annulled convictions on the ground that the original court lacked jurisdiction, and subsequently found the presiding judge not impartial — procedural dispute can reset results to zero |
The case for strong agencies at its strongest: where corruption is already systemic, ordinary procedure has itself been captured from within — evidence disappears, cases stall, witnesses withdraw. Demanding perfect procedure there amounts to never starting.
The case for procedure-first at its strongest: the power to pursue corruption is inherently usable for eliminating rivals, and the persuasive force of a conviction rests on public trust in the investigators' motives — the scarcest resource in a corrupt society. If a verdict is later overturned on procedural grounds, what the public learns is not "corruption gets punished" but "it was all a fight."
Trade-off: force buys enforcement volume and sells the long-run legitimacy of verdicts; procedure-first buys legitimacy and sells the ability to start at all under deep corruption. A common compromise splits by tier: process redesign and digitisation at the lower levels, stronger judicial review at the top.
Reading "more cases" as "worse corruption." Prosecutions and exposure almost necessarily rise early in any anti-corruption push — the detector has just been switched on, the disease has not worsened. The reverse holds too: a falling perception score may simply mean the press gained more freedom to report. Changes in observation intensity contaminate what is observed.
Transparency is often treated as a synonym for fighting corruption; it is in fact only the first link in a chain: disclosure → someone can read it → someone is motivated to care → a usable accountability channel exists → behaviour changes. Break any link and disclosure is just noise — which explains both why some places publish vast quantities of data to no effect, and why disclosure running ahead of accountability can entrench a sense of futility.
The tools fall into three groups: rights of access (Sweden's 1766 Freedom of the Press Act was the first to establish access to official documents in law; India's 2005 Right to Information Act is among the most heavily used modern equivalents), mandatory declaration (assets and conflicts of interest), and sector-specific disclosure (such as the Extractive Industries Transparency Initiative, formed in 2003). The cross-border dimension is often missed: corruption is frequently taken in one country and hidden in another — the US Foreign Corrupt Practices Act of 1977 moved jurisdiction onto the paying side, and the UN Convention against Corruption, adopted in 2003 and in force from 2005, is the first global instrument.
The maximal-disclosure case at its strongest: informational asymmetry is what allows corruption to persist. Disclosure is cheap and does not require a credible enforcement body to exist first — it distributes monitoring across journalists, researchers, competitors and citizens.
The cautious case at its strongest: the side effects are real and not all in the expected direction. Opening negotiations completely pushes the bargaining into informal channels beforehand, leaving the formal record as theatre; mandatory asset declaration without privacy safeguards transfers risk onto family members.
Trade-off: full disclosure buys distributed monitoring and sells the usability of formal process; selective disclosure buys negotiating efficiency and sells external verifiability. The common compromise splits by stage: the process may be confidential, but outcomes and reasons must be public and challengeable.
First, treating transparency as the whole of anti-corruption: it supplies only the input side, and without courts or elections to convert information into consequences, more disclosure need not mean less corruption. Second, reading a perception index as a description of experience — Transparency International's Corruption Perceptions Index aggregates 13 sources on a 0–100 scale and measures how experts and businesspeople perceive public-sector corruption. It says nothing about any individual's situation.