Three decades after the establishment of the contemporary international anti-corruption architecture, a fundamental question remains: to what extent have its rules, institutions and activities translated into an effective reduction in corruption?

This article examines the gap between institutional activity and anti-corruption outcomes using documentation and data from the European Commission, the OECD, the United Nations, the OAS, the IMF, OLAF and the European Public Prosecutor’s Office. It proposes evaluating outcomes through an analytical chain encompassing the observability of corruption, the receipt and verification of information, access to investigation, prosecution, sanctions, their enforcement and asset recovery, identifying potential losses between the different stages.

The analysis connects international estimates of the macroeconomic dimension of corruption with the limitations of available indicators for determining what proportion of the phenomenon is detected, investigated, sanctioned and reduced. It also examines a specific observability and measurement gap concerning corruption that may involve public officials and officeholders, as well as the importance of effective protection for whistleblowers, journalists and lawyers in the functioning of detection mechanisms.

Finally, it advances the hypothesis that certain anti-corruption systems, when their controls are known but their enforcement remains insufficient, may generate counterproductive effects and indirectly facilitate the adaptation and concealment of some corrupt practices.

The article concludes that the expansion of the anti-corruption architecture does not, in itself, constitute evidence of an equivalent reduction in corruption. It proposes extending evaluation to effective outcomes, measurement gaps and the possible unintended effects of systems established to combat corruption.

Keywords: macroeconomic corruption; systemic corruption; anti-corruption outcomes; investigation gap; public officials; asset recovery; whistleblower protection; corruption measurement; adaptation of corruption.

1. From Institution-Building to the Evaluation of Outcomes

In 2014, the European Commission presented a particularly significant assessment. Member States already possessed most of the legal instruments and institutions necessary to prevent and combat corruption. Nevertheless, the results were unsatisfactory. Rules were not always vigorously enforced, systemic problems were not addressed with sufficient effectiveness, institutions did not always have adequate capacity, and stated intentions remained too far removed from concrete results. The Commission added that genuine political will to eradicate corruption often appeared to be lacking.

This observation was particularly significant because it implicitly distinguished two realities that are frequently confused: the construction of an anti-corruption architecture and its outcomes.

Approximately three decades have passed since the contemporary international anti-corruption architecture began to take shape, and twelve years since that assessment by the European Commission.

What have been the results?

The question requires a change in the unit of measurement.

The adoption of conventions, directives, laws and strategies; the creation of agencies; evaluation visits; reports; recommendations; investigations initiated; and proceedings conducted make it possible to measure institutional activity. They do not, in themselves, establish how much corruption has been reduced.

Measuring outcomes requires reconstructing a different chain:

Information or report → receipt and assessment → verification → access to investigation → detection → investigation → prosecution → trial → conviction or sanction → enforcement → asset recovery → reduction of corruption.

Each transition represents a potential point of loss.

This chain constitutes an analytical model of the possible transitions between the existence of corruption and the outcomes of anti-corruption systems, rather than a single procedural sequence. Some stages may occur in parallel, recur or take place in a different order depending on the procedure and jurisdiction. In particular, detection may precede reporting, and certain measures to secure or recover assets may be adopted before the final resolution of proceedings.

Evaluation also requires determining whether these losses are distributed similarly among different categories of actors or whether the institutional position of potentially implicated individuals is associated with differences in access to successive stages.

There is also a cross-cutting condition: those who provide information, investigate it, publish it or pursue it through legal channels must be able to do so without suffering reprisals capable of interrupting the chain.

2. The First Gap: Rules and Institutions Do Not Equate to Outcomes

The problem identified by the European Commission in 2014 was not primarily the absence of rules. It was the distance between rules, institutions and outcomes.

The Commission also warned that anti-corruption strategies could produce very different results. In some cases, they had brought about genuine progress; in others, apparently ambitious strategies had had little or no effect on existing conditions. Even the creation of specialised agencies could generate negative effects if it created the impression that other institutions were relieved of their own responsibilities.

Twelve years later, the OECD again identifies a similar gap. Its Anti-Corruption and Integrity Outlook 2026 places the average strength of integrity regulations among member countries at 63%, compared with 44% for implementation: a difference of 19 percentage points. Only approximately one in four member countries monitors the implementation of its anti-corruption strategies.

The problem, therefore, does not appear to lie exclusively in the production of legislation.

There is a gap between regulation, implementation and outcomes.

Yet even accurately measuring implementation does not answer the fundamental question. A rule may be formally implemented and an institution may undertake considerable activity without this demonstrating an equivalent reduction in corruption.

3. Existing Institutions, Insufficient Capacity and Independence

The United Nations Convention against Corruption does not merely require institutions to exist. Articles 6 and 36 require preventive and specialised bodies to have the independence necessary to perform their functions effectively and act without undue influence.

The distinction is fundamental.

Institutional existence ≠ institutional capacity ≠ institutional independence ≠ effectiveness.

Institutional independence also requires distinguishing between formally established safeguards and actual operating conditions. The existence of legal autonomy does not, in itself, demonstrate that decisions concerning verification, investigation or prosecution are free from undue influence.

An institution may exist, receive reports, compile statistics and close cases while nevertheless failing to detect or prosecute a significant proportion of existing corruption.

Consequently, the number of agencies, officials, cases or actions does not necessarily constitute an indicator of anti-corruption outcomes.

4. Asset Recovery: A Measurable Economic Outcome

The recovery of illicit proceeds is one of the few areas in which the distance between institutional activity and economic outcomes can be observed with relative clarity.

In its 2022 impact assessment on asset recovery and confiscation, the European Commission used an estimate of €139 billion in annual revenues generated by nine major serious and organised crime markets in the European Union, corresponding to 2019. The same assessment estimated that approximately €2.4 billion was frozen annually and around €1.2 billion ultimately confiscated.

In relative terms, the Commission indicated that approximately 2% of estimated criminal proceeds was frozen and 1% confiscated, leaving around 99% in the hands of criminal organisations. These proportions are approximate and derive from estimates covering different reference periods, rather than a homogeneous measurement for a single year.

These figures do not relate exclusively to corruption, but to the broader category of proceeds from serious and organised crime. They therefore cannot be used as an estimate of the recovery of assets derived from corruption.

Their significance for this analysis is different: they show that a developed legal and operational architecture can coexist with an extraordinarily small proportion of criminal proceeds being frozen and confiscated.

The Commission assessed the Confiscation Directive as moderately effective. Although it identified an increase in freezing and confiscation orders, it considered that the results remained insufficient to have a significant impact on the proceeds of organised crime.

Here, once again, the difference between activity and outcome becomes apparent:

More freezing orders ≠ a greater proportion of illicit proceeds recovered.

5. Investigating the Money Before It Disappears

Recovery also depends on time.

Assets may be transferred, transformed, concealed or moved between jurisdictions.

A delayed verification may establish what exists at the time of the inquiry, but not necessarily what existed when the reported transaction occurred.

This introduces a variable that traditional statistics capture poorly: Time to verification.

When a report provides specific financial information — institution, account, amount, date, transfer or beneficiary — the relevant outcome is not simply recording that the report was received.

It is necessary to establish:

What information was verified;
How long the verification took;
Whether the situation existing on the reported date was verified;
What outcome the verification produced;
Whether that outcome enabled an investigation to be opened, assets to be located or the information to be dismissed on substantiated grounds.

A financial investigation initiated after the assets have already disappeared may count statistically as an investigation while representing a substantive failure.

6. Before Investigation, There Is Another Gap: Access to Investigation

Traditional measurement often begins too late.

It counts investigations, charges, proceedings and convictions. But before an investigation exists, decisive institutional decisions have already been made.

Information may be received and dismissed without substantive verification. A report may be considered insufficient because it does not yet contain the evidence that an investigation would be needed to obtain.

It is therefore necessary to distinguish between:

Reports received ≠ reports verified ≠ investigations opened ≠ substantive investigations ≠ prosecutions ≠ convictions.

Two related but distinct variables emerge.

Rate of substantive verification

What proportion of communications containing specific and verifiable information about possible acts of corruption receives documented substantive verification before being dismissed?

Rate of access to investigation

What proportion of communications received leads to a substantive investigation, distinguishing those that proceed directly to investigation from those that do so following preliminary verification?

Evaluating both rates requires distinguishing between communications received, those containing information capable of verification and those meeting the legal requirements for initiating an investigation. The reasons for dismissal and the time elapsed before the decision must also be identified.

This question is particularly important when certain information can be verified only by authorities with access to records, accounts, tax information, communications or other sources inaccessible to the person reporting the matter.

A circularity may then arise:

No investigation because evidence is lacking → the evidence can only be obtained through institutional verification → without verification, the evidence can never emerge.

A statistic recording such a case as “resolved” measures administrative or judicial activity. It does not measure the effective detection of corruption.

7. A Specific Gap: Investigating Corruption Involving Public Officials

Evaluating anti-corruption outcomes presents an additional difficulty when the conduct under investigation involves public officials or members of the very institutions responsible for preventing, detecting or prosecuting corruption.

In such cases, knowing the total number of investigations conducted is insufficient. It is also necessary to determine who is investigated and who remains outside the scope of investigation, and whether the institutional position of potential perpetrators is associated with differences in detection, verification, investigation, prosecution and sanctions.

7.1. The European Union: OLAF and the European Public Prosecutor’s Office

The European Union provides a particularly clear illustration of this problem because it has two specialised bodies with different functions: the European Anti-Fraud Office (OLAF) and the European Public Prosecutor’s Office (EPPO).

Their statistics reveal substantial investigative activity but provide a much less complete picture when attempting to reconstruct specifically how corruption involving officials and members of institutions is handled.

Within the European Public Prosecutor’s Office, active and passive corruption involving public officials — whether national or EU officials — represents a relatively small share of investigative activity. In 2024, corruption accounted for approximately 3% of the offences investigated, while in 2025 it appeared among the offence categories present in 4.69% of active investigations.

These figures relate to different statistical units and should not be interpreted as a directly comparable time series. Nor do they indicate what percentage of investigations involves public officials, since a single investigation may encompass several offences and the involvement of officials may arise in cases classified under other categories.

In 2025, the EPPO had 3,602 active investigations. The Office itself also notes that corruption may be associated with other offences within its jurisdiction, meaning that these percentages do not represent all investigations in which a public official may be involved.

Another relevant indicator appears at an earlier stage of the chain. In 2024, only 1.7% of reports of possible criminal offences received by the European Public Prosecutor’s Office originated from EU institutions, bodies, offices and agencies themselves.

The EPPO explicitly noted that this low percentage demonstrated the need to improve the detection and reporting of suspected fraud by EU institutions.

The problem may therefore arise even before investigation: if certain facts are neither detected nor reported, they cannot subsequently appear among investigations, charges or convictions.

OLAF presents a complementary problem.

Its mandate includes internal investigations into serious misconduct by officials and members of EU institutions. Between 2021 and 2025, it issued 107 disciplinary recommendations, on which the competent authorities had taken 65 decisions.

During the same period, it issued 145 judicial recommendations. National judicial authorities had taken decisions on 73 of these, and 27 had resulted in indictments.

OLAF may issue both disciplinary and judicial recommendations following its internal investigations.

However, although OLAF provides separate information on its disciplinary and judicial recommendations, its aggregate judicial statistics do not allow a complete and comparable reconstruction of what proportion of investigations, recommendations and criminal justice outcomes concerns officials or members of EU institutions, and what proportion concerns economic operators or other private individuals.

Nor do they systematically identify the losses occurring between the receipt of information, investigation, recommendations and their final consequences for each category of actors.

Consequently, they do not allow a complete and comparable chain to be reconstructed for public officials:

Information received → assessment → verification → investigation → disciplinary or judicial recommendation → indictment → sanction or conviction → enforcement.

This statistical limitation is significant in itself.

An anti-corruption architecture cannot be evaluated solely on the aggregate number of investigations or recommendations it produces. It is also necessary to know which categories of actors pass through each stage of the system and with what outcomes.

The available data do not, in themselves, establish that public officials receive favourable treatment.

The low statistical share of corruption involving officials may reflect several factors: the limited jurisdiction of the bodies concerned, the classification of offences under other categories, lower detection rates, lower reporting rates, evidentiary difficulties, referral of cases to national authorities or differences in effective access to investigation.

The inability to distinguish adequately between these explanations is itself part of the evaluation problem.

A specific gap in the observability, investigation and measurement of corruption involving public officials can therefore be identified: the distance between corruption that may involve, require or benefit from actions performed from public positions and the portion of that phenomenon that is detected, verified, investigated, prosecuted and ultimately sanctioned.

7.2. From the European Union to the Member States

This gap should not be analysed exclusively as a characteristic of EU institutions.

The European architecture for combating fraud and corruption is closely connected to national systems.

OLAF refers judicial recommendations to the competent national authorities and subsequently depends on them for a substantial part of the criminal justice consequences.

The European Public Prosecutor’s Office has broader investigative and prosecutorial powers, but operates through European Delegated Prosecutors and before the national courts of participating Member States.

An institutional chain thus operates in both directions:

Member States → detection and reporting → OLAF/EPPO → investigation and referral or prosecution → national authorities and courts → sanctions, convictions and enforcement.

A national gap in the detection, investigation or prosecution of corruption involving public officials may therefore be transmitted to the European level and reappear subsequently when the outcomes of European investigations depend on national action.

The outcomes observed in OLAF and the European Public Prosecutor’s Office therefore raise a more general question: to what extent does the limited observable presence of corruption involving public officials in certain European statistics also reflect gaps within Member States’ anti-corruption systems?

Answering this question would require comparable national statistics capable of distinguishing, at a minimum, between public officials, political officeholders, economic operators and private individuals across successive stages of reporting, verification, investigation, prosecution, conviction and enforcement.

In the absence of such information, aggregate numbers of proceedings may conceal substantial differences depending on the position of the individuals concerned.

7.3. From Member States to the Global Dimension of Corruption

The issue acquires an even greater dimension when connected to international estimates of the scale of corruption examined in the first article of this series.

As shown there, estimates disseminated within the United Nations system have placed the cost of corruption at approximately 5% of global GDP, while the United Nations has subsequently indicated that up to 25% of global public expenditure may be lost to corruption.

These magnitudes make it possible to regard corruption not merely as an accumulation of individual acts, but as a phenomenon of macroeconomic dimensions. [Marzal, 2026, Article 1].

Comparing this scale with anti-corruption outcomes raises a fundamental question.

If corruption has a macroeconomic dimension, the effectiveness of the systems established to combat it should not be evaluated exclusively through the absolute number of rules adopted, agencies created, investigations opened, charges brought or convictions obtained.

This relationship presents a methodological difficulty: macroeconomic information on corruption and administrative or judicial statistics do not necessarily use the same units of measurement, reference periods or populations.

Consequently, the identified gap cannot be quantified directly by comparing their aggregate figures. It is, first and foremost, a knowledge and measurement gap requiring the development of indicators and methods capable of approximating the different stages of the phenomenon.

The discrepancy is particularly significant in relation to public officials and officeholders.

Corruption capable of affecting such a high proportion of public expenditure can hardly be analysed exclusively through the actions of companies and private individuals.

Public procurement, subsidies, budget management, authorisations, regulation and numerous administrative decisions necessarily involve public institutions and officials.

The United Nations identifies public procurement as an area particularly exposed to corruption because of the interaction between public officials and private operators.

Nevertheless, there is no equivalent global measurement allowing us to reconstruct what proportion of this corruption is detected, what proportion is substantively verified, which public officials or officeholders are investigated, how many are prosecuted and sanctioned, and what proportion of the resources affected is ultimately recovered.

A much broader gap therefore emerges:

Macroeconomic dimension of corruption → observable corruption → detected corruption → verified information → investigated corruption → prosecuted corruption → sanctioned corruption → enforced sanctions and recovered assets → effective reduction of corruption.

The observability and measurement gap identified in OLAF and European Public Prosecutor’s Office statistics may therefore constitute a particularly visible European manifestation of a broader problem: the distance between the estimated macroeconomic dimension of corruption and the portion of the phenomenon that anti-corruption systems manage to convert into effective outcomes, particularly when actors within public institutions are involved.

European data do not allow us to infer that the same gap has an identical magnitude across all Member States, much less globally.

They do, however, demonstrate the need to incorporate a variable that has so far been insufficiently visible in the evaluation of outcomes: the institutional position of the individuals investigated.

The relevant question is therefore no longer simply how much an anti-corruption system investigates. It must also ask:

What proportion of a phenomenon of macroeconomic dimensions is detected, verified, investigated, prosecuted, sanctioned and ultimately reduced, and how do these outcomes vary when possible corruption involves public officials, officeholders or other actors situated within institutional structures themselves?

This question directly connects the macroeconomic dimension of corruption examined in the first article of the series with the outcomes studied here.

It also identifies a different causal question that lies beyond the scope of the present paper: why certain reports, actors or dimensions of corruption gain access to effective investigation while others do not.

Analysing the institutional mechanisms that may produce these differences requires a separate study.

8. Actual, Observable, Detected and Sanctioned Corruption

This distinction leads to a deeper methodological problem.

Four magnitudes must not be confused:

Actual corruption ≠ observable corruption ≠ detected corruption ≠ sanctioned corruption.

These differences are not necessarily distributed uniformly among different actors.

The institutional position of potentially implicated individuals must be incorporated as an analytical variable: aggregate outcomes may conceal differences between public officials and officeholders, economic operators and private individuals in the transition from information to investigation, prosecution and sanctions.

Detected corruption depends on the capacity to observe it.

And the capacity to observe it depends, among other factors, on reports received, the willingness of authorities to verify them, access to financial information, institutional independence, whistleblower protection, freedom of journalists, the work of lawyers and the quality of statistical systems.

A decline in detected cases may therefore indicate less corruption.

But it may also indicate less detection.

Without additional information, one situation cannot automatically be equated with the other.

Estimating unobserved corruption requires methods that complement administrative and judicial records.

Consequently, the proportions of corruption detected, investigated or sanctioned relative to corruption actually occurring cannot be calculated directly from the number of cases or convictions.

Their estimation requires additional sources, explicit assumptions and an assessment of uncertainty.

These distinctions make it possible to represent the problem as a chain of transformation and loss.

Between actually existing systemic corruption and the final outcomes of anti-corruption efforts lie successive gaps in observability, verification, access to investigation, investigation, prosecution, conviction and enforcement.

The system may reduce corruption, prove ineffective or, under certain conditions, indirectly facilitate its adaptation, persistence or expansion.

Figure: From the Macroeconomic Dimension of Corruption to Anti-Corruption Outcomes

General gap and specific gap in the investigation of public officials and officeholders.

The figure in the original manuscript represents three interconnected levels -global, European Union and Member States- and the successive stages between macroeconomic and systemic corruption and effective anti-corruption outcomes.

Analytical diagram of transitions and potential losses. Stages may overlap and do not represent a single procedural sequence. Macroeconomic magnitudes and institutional outcomes do not constitute directly comparable statistical series.

9. The Information Deficit Concerning Outcomes

International organisations depend to a considerable extent on information produced by States themselves.

This creates two different dependencies:

Production dependency: the State must generate, preserve and structure the information.
Cooperation dependency: it must also provide that information to the organisations evaluating its outcomes.

If information is not produced, is produced in a fragmented manner, ceases to be published or is not provided, the international organisation cannot adequately reconstruct the evolution of corruption either.

Three deficits can therefore be distinguished:

Non-existent information;
Fragmented, incomplete or non-comparable information;
Informational regression, when previously available variables disappear or progressively lose detail.

A further distinction must be made between information that is not produced, information that exists within institutions but is not published, and data that are published without the level of disaggregation necessary to evaluate outcomes.

These situations raise different problems of statistical capacity, transparency and comparability.

This problem introduces a fundamental limitation into any international evaluation.

Knowledge of anti-corruption does not necessarily equate to knowledge of corruption.

Disaggregation by institutional position is part of this problem.

An aggregate statistic may report the total number of investigations, proceedings or sanctions while simultaneously making it impossible to determine whether public officials, officeholders, economic operators and private individuals pass through the different stages of the system with different outcomes.

The absence of such information limits not only statistical transparency, but also the possibility of evaluating whether the anti-corruption system itself operates equally across different categories of actors.

We may acquire increasingly detailed knowledge of the laws, strategies, institutions and procedures established to combat corruption while remaining unaware of a substantial proportion of the corruption actually occurring.

10. The Americas: The OAS and the Belated Need to Measure Impunity

The evolution of MESICIC is particularly illustrative.

The Inter-American Convention against Corruption was adopted in 1996, and its follow-up mechanism began operating subsequently.

For decades, it produced reports, recommendations, visits and cooperation tools.

However, the development of indicators specifically designed to measure outcomes relating to the prevention, detection and reduction of impunity is much more recent.

The OAS approved a methodology for developing these indicators in 2021 and adopted the first-phase indicators in 2022, which were subsequently published in 2025. That same year, it published a second phase devoted to the protection of whistleblowers and witnesses.

The chronology is significant: approximately a quarter of a century after the adoption of the Inter-American Convention against Corruption, its follow-up mechanism moved towards specific instruments for measuring outcomes.

The methodological shift is significant.

After decades of anti-corruption activity, it became necessary to construct indicators capable of answering questions much closer to substantive outcomes:

Are cases investigated?
Are they prosecuted?
Are they brought to trial?
Are sanctions imposed?
Are sanctions enforced?
Are those who report corruption protected?

The evolution of MESICIC confirms the need to distinguish between measuring the anti-corruption architecture and measuring its outcomes.

11. The IMF: Greater Institutional Knowledge and Persistent Obstacles

In 2018, the International Monetary Fund strengthened its framework for addressing governance and corruption, incorporating a systematic assessment of vulnerabilities relating to core State functions.

Its 2023 review considered that the Fund’s engagement had become broadly more systematic, explicit and effective.

However, the same review identified significant limitations.

Compliance with governance-related benchmarks was considered somewhat disappointing.

Implementation of certain commitments undertaken during the pandemic was uneven.

The Fund itself acknowledged limited capacity, lack of ownership and vested interests as persistent obstacles.

The IMF’s 2023 review also identified a particularly revealing limitation: during the period examined, only 13 countries had volunteered for the Fund to assess their frameworks concerning the transnational aspects of corruption, a level of participation below expectations.

This again demonstrates the difference between the capacity to generate knowledge and effective access to the information required to exercise that capacity.

An international organisation may develop increasingly sophisticated methodological instruments while simultaneously depending on the willingness of States to access part of the reality it seeks to evaluate.

12. The Most Vulnerable Link: Those Who Make Corruption Detectable

Corruption does not detect itself.

This dependency is particularly relevant to the portion of corruption that may occur within institutional structures themselves.

The less externally observable such corruption is, the more important the individuals capable of providing, investigating, publishing or legally presenting information that would otherwise never reach the anti-corruption system may become.

Before an investigation begins, there are often people capable of observing, communicating, investigating, publishing or legally presenting information:

Sources → whistleblowers → journalists → lawyers → investigators.

The effectiveness of anti-corruption efforts depends on this chain being able to function.

Whistleblower protection is therefore part of the infrastructure of detection, not merely a matter of individual rights.

The same applies to investigative journalism and, in a different capacity, to lawyers who make it possible to turn information or reports into legal action capable of activating administrative and judicial controls.

When these actors suffer reprisals, the effect is not necessarily limited to the individual concerned.

It may change the behaviour of other potential whistleblowers, journalists or professionals.

13. From Formal Protection to Effective Protection

The existence of protective legislation does not, in itself, demonstrate that protection works.

A distinction must be made:

Legal protection of whistleblowers ≠ effective protection of whistleblowers.

Its anti-corruption significance is particularly high when protected individuals constitute essential channels for making information about possible misconduct by public officials or officeholders observable, where such information would be difficult to obtain from outside institutions.

Similarly:

Legal protection of lawyers ≠ effective protection of lawyers.

The case law of the European Court of Human Rights provides a particularly relevant example.

In Rodríguez Ravelo vs Spain, the Court examined the criminal conviction of a lawyer for statements made in the course of his professional activities and considered the chilling effect that a criminal sanction could have on the practice of law.

The systemic significance of this reasoning extends beyond the individual lawyer.

Protection that becomes effective only after years of proceedings, appeals and costs may ultimately overturn a conviction while having already produced a substantial part of its chilling effect.

The evaluation of outcomes should therefore incorporate an additional variable: Rate of effective protection against reprisals

What proportion of reported reprisals against individuals involved in detecting and disclosing possible acts of corruption receives effective protective measures, what proportion of substantiated reprisals is stopped, annulled or remedied, and how much time elapses before those measures are adopted?

Time once again becomes an outcome variable.

Effective protection becomes particularly important when the information reported concerns public officials or officeholders, especially where the alleged conduct is difficult to observe from outside institutions.

Whistleblowers, journalists and lawyers may make it possible for such information to become known, verified and investigated.

Reprisals and disproportionate professional costs may reduce not only their individual capacity to act, but also the likelihood that other people will report similar information.

Effective protection is therefore a relevant variable in anti-corruption outcomes, rather than merely an individual safeguard.

14. The Paradox of Anti-Corruption Knowledge

There is one final possibility that has received little consideration in evaluations of thirty years of institutional development.

The anti-corruption architecture generates knowledge.

International organisations, governments, prosecutors, police officers, judges, companies, universities and civil society organisations study how corruption occurs, which indicators enable its detection, which financial transactions raise suspicion, which control systems work, what reporting mechanisms exist and how investigations are conducted.

This knowledge is necessary to combat corruption.

But a substantial proportion of that knowledge is necessarily public.

Rules, protocols, reports, evaluations, recommendations, case law, methodologies, indicators, identified vulnerabilities and institutional reforms progressively explain how the anti-corruption system works.

The information is not available only to those combating corruption.

It may also be available to those seeking to evade its controls.

A possible paradox of anti-corruption knowledge thus emerges:

Greater institutional knowledge about how to combat corruption does not necessarily imply greater public knowledge of the corruption actually occurring and may, under certain conditions, facilitate learning by those engaging in corrupt practices.

The possible mechanism would be dynamic:

Corruption → detection → investigation → institutional learning → new controls → public knowledge of controls → possible adaptation of corrupt practices → reduced detectability → new investigative techniques.

The argument is not that the development of anti-corruption systems necessarily produces more corruption.

The hypothesis is different.

Certain configurations of anti-corruption systems may indirectly facilitate the adaptation, sophistication, concealment or reproduction of corrupt practices when the knowledge generated by control systems increases more rapidly than their effective capacity to detect, investigate and sanction adapted forms of corruption.

Under these conditions, it is possible to hypothesise that certain anti-corruption systems may generate processes of adversarial learning, through which those engaging in corrupt practices use information about existing controls and their limitations to modify their behaviour.

The existence, intensity and consequences of these processes require empirical verification.

15. Can Anti-Corruption Efforts Favour Certain Corrupt Practices?

The preceding possibility requires broadening the initial question.

Until now, the analysis has generally been framed in binary terms: anti-corruption policy either works or does not work.

But there is a third possibility: it may produce counterproductive effects.

This already appears partially in the European Commission’s 2014 assessment.

The Commission observed that certain anti-corruption agencies could even play a negative role when their existence created the impression that other authorities no longer needed to assume their share of responsibility.

The principle can be extended.

Under certain conditions, an anti-corruption architecture could:

Create an appearance of control greater than the control actually exercised;
Concentrate attention on easily measurable conduct while other conduct remains outside the observable field, or certain actors remain insufficiently distinguished within aggregate statistics;
Indirectly reveal which behaviours are detected and which remain beyond control mechanisms;
Displace corrupt practices towards more sophisticated legal, financial or organisational mechanisms;
Discourage reporting when those who communicate information bear greater costs than those reported;
Facilitate the persistence or reproduction of certain practices when the perceived probability of investigation, sanction and loss of illicit gains remains low.

Under these conditions, insufficiently effective anti-corruption policy may indirectly favour certain corrupt practices, not because that is its objective, but because it changes incentives and provides information about the effective limits of the control system.

These potential effects become particularly relevant when there is a substantial distance between the formal severity of rules and their effective enforcement.

In such circumstances, legal prohibition may signal a high level of risk while the perceived probability of detection, sanction and loss of illicit gains remains low.

The evaluation of outcomes must also examine this difference between formally declared legal risk and effective risk.

16. The Outcome That Remains Unmeasured

Thirty years later, the anti-corruption architecture makes it possible to measure in considerable detail what institutions do.

We know how many conventions exist.

How many recommendations are issued.

How many laws are adopted.

How many agencies are created.

How many reports are published.

How many complaints and reports are received.

In certain areas, we also know how many investigations are opened, how many convictions are obtained and how much money is frozen.

But we still have far more limited information with which to answer different questions:

What proportion of existing corruption is detected?
What proportion of sufficiently specific information is verified?
How much reported corruption actually gains access to a substantive investigation?
How does that access vary according to the institutional position of the individuals potentially implicated?
What proportion of investigations, prosecutions, convictions and sanctions concerns public officials and officeholders, economic operators and private individuals?
What proportion ultimately results in sanctions that are effectively enforced?
How much illicit wealth is actually recovered?
How many whistleblowers suffer reprisals?
How many of those reprisals are stopped and remedied?
How many lawyers or journalists refrain from acting because of the anticipated cost?
How much corruption changes form to evade existing controls?
And how much has corruption ultimately been reduced?

These are questions about outcomes.

Not all these questions can be answered using the same instruments.

Some require disaggregated administrative and judicial statistics; others require information on how reports are handled, the duration of proceedings, the enforcement of sanctions and asset recovery.

Estimating unobserved corruption, unreported reprisals and behavioural changes designed to evade controls requires complementary methods.

Evaluation must therefore distinguish between directly observable outcomes, outcomes that can be reconstructed using additional information, and dimensions whose estimation is subject to considerable uncertainty.

17. Thirty Years Later

The contemporary international anti-corruption architecture has brought about a considerable transformation of the legal and institutional framework.

Conventions, monitoring mechanisms, specialised agencies and cooperation instruments have expanded capacities for prevention, detection, investigation, prosecution and asset recovery.

It cannot be concluded that these advances have had no effects.

However, the expansion of the anti-corruption architecture cannot be equated with an equivalent reduction in corruption.

The evidence examined repeatedly reveals the distance between rules and implementation, between institutional activity and effective outcomes, between information received and access to substantive investigation, between investigations and economic recovery, and between legal protection and effective protection.

The European Commission warned as early as 2014 that stated intentions remained too far removed from concrete results.

In 2026, the OECD continues to identify a substantial gap between regulation and implementation.

The IMF acknowledges obstacles relating to institutional capacity, ownership of reforms and vested interests.

The OAS has developed indicators intended to evaluate outcomes relating to prevention, detection, prosecution, sanctions and impunity.

These limitations acquire greater significance when related to the scale of the phenomenon that anti-corruption systems are intended to combat.

The first article in this series examined international estimates that make it possible to regard corruption as a phenomenon of macroeconomic dimensions.

The present article shows that there is no equivalent, sufficiently comprehensive and comparable measurement capable of determining what proportion of that phenomenon becomes observable, detected, verified, investigated, prosecuted, sanctioned and ultimately reduced.

A general knowledge and measurement gap thus emerges between the estimated dimension of corruption and the anti-corruption outcomes that can be identified.

This gap does not constitute a quantified difference between directly comparable magnitudes, but rather a fundamental limitation of the instruments available for evaluating the effectiveness of anti-corruption systems.

Within this general gap, a specific issue arises: what happens when possible corruption involves public officials, officeholders or other actors situated within institutional structures themselves?

The statistics examined, particularly those of OLAF and the European Public Prosecutor’s Office, make it possible to identify relevant activities and outcomes, but not to reconstruct the successive stages of the system in a sufficiently complete and comparable manner according to the institutional position of the individuals potentially implicated.

This limitation does not, in itself, demonstrate that public officials or officeholders receive favourable treatment.

It does, however, prevent an adequate determination of whether detection, verification, access to investigation, prosecution and sanctions exhibit systematic differences between categories of actors.

Institutional position must therefore be incorporated as a specific variable in the evaluation of outcomes.

The analysis also shows that the effective protection of whistleblowers, journalists, lawyers and others who help reveal possible corrupt practices is not merely a matter of individual rights.

Their capacity to act influences which information can become known, verified and investigated.

Reprisals and the costs associated with disclosing information may affect both those directly involved and the behaviour of other potential sources.

These limitations are accompanied by a hypothesis requiring further investigation.

Certain anti-corruption systems may produce counterproductive effects when their control mechanisms are known but their enforcement remains insufficient.

Under such conditions, information generated by the anti-corruption architecture itself, together with differences between formal prohibition and effective risk, may indirectly facilitate the adaptation, sophistication or concealment of certain corrupt practices.

This does not mean that combating corruption necessarily produces more corruption.

It means that the evaluation of anti-corruption policy must consider, in addition to its direct outcomes, its possible unintended effects and the capacity of corrupt practices to adapt to the mechanisms established to combat them.

Thirty years of international efforts against corruption therefore allow us to formulate one conclusion and two questions.

The conclusion is that the expansion of the anti-corruption architecture does not, in itself, constitute evidence of an equivalent reduction in corruption. Although legal instruments, institutions and operational capacities have been developed, a fundamental gap persists between available knowledge of the macroeconomic dimension of the phenomenon and the information required to measure its effective outcomes. Nor do we have sufficiently comprehensive and comparable information to determine how these outcomes vary according to the institutional position of the individuals potentially implicated.

The first question is what proportion of corruption actually occurring becomes observable, detected, verified, investigated, prosecuted, sanctioned and ultimately reduced, and how this progression varies when possible corrupt practices involve public officials, officeholders, economic operators or other private individuals.

The second question is why these gaps arise and persist, which institutional and social mechanisms influence them, and under what conditions anti-corruption systems may indirectly facilitate the adaptation, concealment or reproduction of certain corrupt practices.

Answering the second question requires moving from the analysis of outcomes to the mechanisms that produce them.

And it leads directly to the next question in this research: Who corrupts the corrupt, and what structures allow corruption to reproduce itself?

Author’s Experience and Accumulated Knowledge

This research programme is the result of a cumulative process spanning more than fifteen years of practical experience, observation and research into corruption and the functioning of institutional and anti-corruption mechanisms.

During this period, the author’s direct experience in reporting and communicating information concerning possible cases of corruption to Spanish, European and international institutions was broadened through the establishment and leadership of various civil society initiatives: as president of an association representing people affected by institutional actions, as the sole founder and president of a federation of associations, and as the sole founder and president of an alliance comprising associations and a social movement.

These activities enabled him to become directly involved in situations affecting other individuals and organisations, undertake actions before different institutions, and share projects and experiences with other civil society leaders.

This trajectory produced a progressive and cumulative process of knowledge acquisition.

Practical experience raised new questions that led to further research; this expanded the available knowledge and, in turn, prompted the investigation of other problems, the reconsideration of previously examined issues and the identification of relationships between phenomena that are commonly studied separately.

This process has influenced not only the questions addressed by the research, but also how those questions have been formulated, the sources and data examined, the relationships established between them, and the analytical framework through which the evidence is interpreted.

The resulting research has progressively connected the macroeconomic dimension of corruption with its systemic nature, the gaps between the phenomenon and its criminalisation, the gaps in the functioning of anti-corruption systems, their measurable outcomes, and the mechanisms through which corruption may adapt, persist, reproduce itself and contribute to institutional, economic, political and societal transformations.

The author’s accumulated knowledge has thus helped shape a research programme that goes beyond the study of individual acts of corruption to examine jointly the scale and nature of the phenomenon, the structures within which it operates, and the systems established to combat it.

The specific cases known to the author through this experience do not form part of the empirical evidence used in the seven articles comprising Part I of the series.

Their empirical basis consists of the documentary sources, official data and published evidence cited in each article.

Instead, the author’s experience and accumulated knowledge have contributed to identifying the questions to be asked, selecting the dimensions to be investigated, connecting evidence produced across different institutional and disciplinary contexts, and integrating these elements into the analytical framework developed throughout the series.

Declaration on the Use of Artificial Intelligence

During the preparation of this article, the author used ChatGPT (OpenAI) as a tool to support research and manuscript development.

Its use included locating and examining additional information, analysing and comparing documents and data, assessing relationships between sources, exploring alternative interpretations, organising materials, reviewing the structure and coherence of arguments, assisting with writing, and supporting translation.

The tool was also used as an analytical interlocutor in an iterative process directed by the author.

The research questions, interpretative framework and a substantial initial body of sources and evidence originated from his own research and accumulated knowledge.

The author guided the selection of sources and the incorporation of additional documentation, including reports from OLAF and the European Public Prosecutor’s Office.

AI-assisted responses were subjected to critical evaluation by the author.

The tool does not constitute a source of evidence or an authority replacing the documentary sources cited.

The author has reviewed and approved the final manuscript and assumes full responsibility for its content.

SERIES ON CORRUPTION. Part 1

Without Corruption There Would Be No Public Debt. https://doi.org/ 10.5281/zenodo.21993280
SYSTEMIC CORRUPTION. From Institutional Consequences to Systemic Transformation. https://doi.org/ 10.5281/zenodo.21994617
The Criminalisation Gap in Corruption. Impunity Can Begin Before Prosecution. https://doi.org/ 10.5281/zenodo.22729133
Three Gaps in the International Fight against Corruption. Thirty Years of Rules, Non-Compliance and Insufficient Implementation. https://doi.org/10.5281/zenodo.23019185
Thirty Years of Fighting Corruption: What Have the Results Been? https://doi.org/10.5281/zenodo.23333997
Toward a Structured Definition of Corruption. https://doi.org/10.5281/zenodo.23684696
Who Corrupts the Corrupt? From the Individual Corruptor to the Selection and Reproduction of the Corrupt (Forthcoming)

The first seven articles develop an integrated examination of the scale, systemic dimension, legal framework, institutional response, results and definition of corruption, concluding with an analysis of the actors and structures that corrupt and reproduce corruption.

Previous Research Underlying the Series

Normative Asymmetry in the Inter-American Anti-Corruption Framework: Structural Diagnosis and Conduct-Based Enforcement. https://doi.org/10.5281/zenodo.18851233
Integrating Structural Risk Analysis into GRECO Evaluations: Methodological Opportunities within the Council of Europe Anti-Corruption Framework. https://doi.org/10.5281/zenodo.18871431
Bridging the Structural Gap in the United Nations Convention against Corruption. A Juridical-Economic Analysis of Systemic Corruption in Advanced Economies. https://doi.org/10.5281/zenodo.19021241
Structural Calibration in European Union Anti-Corruption Law: A Contextual Mechanism for Judicial Sentencing Individualisation. https://doi.org/10.5281/zenodo.19021587

Related Historical-Structural Framework

Marzal, J. (2026). The Rise of the Macro-State: A Historical-Structural Analytical Framework for the Convergence of Economy and Power. https://doi.org/10.5281/zenodo.19089595.

Javier Marzal

ORCID: 0009-0009-4373-8068

5.- THIRTY YEARS OF FIGHTING CORRUPTION:

What Have Been the Results?

Table of Contents

Abstract