1. Mounting efforts, limited gains

The cost of corruption greatly exceeds the sum of bribes paid, funds misappropriated, and taxes avoided. Corruption hampers development. It increases inequality, impedes growth, undermines the legitimacy of governments, and weakens the public’s trust in democracy. It impacts everything from learning outcomes in schools to climate change mitigation efforts and is a major impediment to achieving the Sutainable Development Goals. Yet billions of people live in highly corrupt societies, with more than two-thirds of countries scoring below 50 points on Transparency International’s 2020 Corruption Perceptions Index (CPI).

Over the past several decades, the focus on corruption has steadily increased. International conventions and standards have been put in place; laws, institutions, and reforms have been implemented. Academic research on corruption, meanwhile, has taken off as development actors have made tackling corruption an explicit policy objective, backed by funding and capacity building. Protests and polls show that corruption is also a major concern for citizens throughout the world. Yet significantly and sustainably reducing corruption has proved extraordinarily difficult. 

This U4 Issue brings together some of the latest thinking on corruption and how best to curb it. It draws upon research and evaluations that analyse past anti-corruption reforms as well as advances in the theoretical understanding of corruption. It focuses exclusively on broader anti-corruption efforts, not on corruption risk management or on specific processes vulnerable to corruption, such as procurement. The primary intended audience is policymakers and practitioners in the development community. 

The Issue begins by exploring how corruption is defined and perceived, the many forms of corruption, and the factors driving and sustaining corruption. Next, it provides an overview of common anti-corruption interventions and an analysis of the impact of anti-corruption efforts to date. The Issue concludes by providing an overview of potential success factors for macro-level changes as well as concrete guidelines for how to identify, plan, and implement successful anti-corruption interventions.

2. Understanding corruption

2.1. There is no universal definition

One of the most common definitions of corruption is ‘the misuse of entrusted power for private gain’. However, as noted by the Council of Europe, ‘no precise definition can be found which applies to all forms, types and degrees of corruption, or which would be acceptable universally’.774809070ae1 The lack of conceptual precision makes it difficult to measure and compare corruption levels over time and across societies, sectors, and institutions.68ed9f322b32 For programming purposes, the solution has been to complement these broad definitions with a detailed description of the type of corruption in question; the levels, sectors, and institutions where it takes place; the actors involved; and the causal factors. Some scholars have also shifted to defining corruption in terms of its opposites, using terms such as ‘good governance’, ‘quality of government’, ‘ethical universalism’, and ‘state capacity’.

Several studies have found that perceptions of corruption vary from country to country and situation to situation. Smith,8525712f1276 for example, finds that ‘Nigerians’ understandings of what counts as corruption are sometimes quite different than the common Western views’ and that ‘Nigerians themselves often embrace contradictory views’. Small informal payments to service providers are, for example, generally not thought of as bribes in Nigeria. Similarly, while corruption in the abstract is broadly condemned, concrete acts of corruption that benefit a specific social network are often seen by that network as morally legitimate.129fec18d194 However, other studies have concluded that people’s opinions as to what constitutes corruption are relatively similar.a359f453981a Kurer,fd54aff2c020 for example, finds that although there are areas of disagreement, ‘the empirical evidence, patchy as it is, strongly suggests a common understanding of corruption: actions or practices are identified as corrupt even in environments where cultural relativity theory predicts them to be morally acceptable’.

2.2. Corruption takes many forms

‘Corruption is not an aggregate national attribute like GDP per capita.’ – Johnstonc81b507fc0e2

Which country is the most corrupt? This seemingly simple question has no straightforward answer. Corruption can take many forms, occur at all levels of power, and be conducted by all types of agents, including individuals, businesses, public officials, politicians, state agents, and non-state actors.1f60309d680f The scale, focus, and impact of corruption vary by sector and region within a country, and can be influenced by factors beyond the country’s borders.

Common forms of corruption include bribery, facilitation payments, gift giving, embezzlement, favouritism, fraud, extortion, collusion, sextortion, and impunity (see Box 1). Distinctions are often made between political and bureaucratic corruption and between large-scale or ‘grand’ corruption and petty, small-scale corruption. Corruption can also be classified as controlled or uncontrolled. Controlled corruption is hierarchic, disciplined, and coordinated, while uncontrolled corruption is competitive, unpredictable, and chaotic. The causes and manifestations of corruption, as well as the best ways of addressing it, also vary depending on whether the country is stable or fragile, developed or developing, democratic or autocratic.

As discussed in section 2.5 in some countries corruption is also systemic. It is an integral and integrated part of the economic, social, and political system. Most people have no alternative but to engage.

Corruption includes both legal and illegal acts. Scottae546aca9a61 identifies three standards by which to determine whether an action constitutes corruption: public interest, public opinion, and the law. In most cases, the three criteria coincide. In fact, activities that society deems to be corrupt are increasingly illegal.5dc2fa2483d1 However, there are grey areas, as well as differences in law between countries.093d04c7fa8e Corrupt actors are adept at finding new ways of skirting the law. Corrupt acts can also be institutionalised in laws and rules through what is known as state capture.24b0a73f5a3e

Corruption takes place in both the public and private spheres, and in areas where the two intersect. However, most of the corruption literature focuses on the public sector.63efcbd28254 Many definitions of corruption refer exclusively to public officials. Fisman and Golden,83369f4fcec3 for example, argue that ‘corruption always involves a public official’ and that corporate corruption that does not involve public officials is better labelled as corporate malfeasance. This view is in stark contrast to the way in which corruption has previously been defined. In the 1970s, the focus was primarily on corporate abuse of powerb10185cf043d93553d2a8b22 Today there is increasing awareness that while the distinction between public and private is a theoretical ideal in a Weberian state, the two spheres are blurred and overlapping in reality. Some authors, such as Heywood,d6af9ecf60a1 believe that the overlap and interconnections between public and private can lead to new opportunities and modalities of corruption.

2.3. Zero corruption is not a realistic goal

‘Even in the world’s more developed countries ethical universalism remains an ideal, while personal connections and government favoritism still loom larger than anyone officially admits.’ – Mungiu-Pippidic49505bc9495

Anti-corruption efforts are often described as a struggle to fight, root out, and even eradicate corruption. While such language may be useful for drawing attention to the issue, it is important to recognise the historical persistence of corruption. Corruption is not a newly emerged ‘disease’ that can be cured with the right combination of treatments. It is not a deviation, and it cannot be explained away by blaming national culture. Corruption is widespread and has existed for thousands of years. Mungiu-Pippidi2b4b64d3c956 points out that ‘all states have started from being “owned” by a few individuals who control all resources’, and consequently, ‘particularism is a natural inclination – people tend to favor their own’. She adds, ‘The public-private separation in public affairs and the complete autonomy of state from private interest are exceptions in the present world, difficult to reach and difficult to sustain’. In other words, corruption is age-old, entrenched, complex, and resilient. Therefore, ‘unless humans turn into angels, corruption won’t be eradicated any time soon, and even incremental improvements are hard to come by’.979d6fedb522

2.4. Corruption is caused and sustained by multiple reinforcing drivers

‘Corruption is a complex, dynamic, and often contradictory phenomenon. It is hidden yet widely acknowledged, harmful yet at times beneficial, where a victim today can be a perpetrator or beneficiary tomorrow. It is this complexity that enables corruption to systematically undermine development assistance outcomes.’ – Scharbatke-Church and Chigas9f4db6d06236

A number of theories, formulas, and frameworks have been developed to understand corruption. Which of these, or which combination of these, best explains the phenomenon depends on the context. Even within a country, how, why, and to what degree corruption takes place will vary by geographic area, community, sector, and institution, as well as over time.38c7024841e9

Over the past three decades, corruption has mainly been framed in terms of deviation, focusing on bad actors and the incentives that tempt people to act badly. It has been explained by rational choice theory and by the principal-agent problem (described below). Among the equations applied are Klitgaard’s 1988 formula:

Corruption = monopoly + discretion − accountability

The United Nations Development Programme8133efab0216 adds more detail:

Corruption = (monopoly + discretion) − (accountability + integrity + transparency)

There are also many variants of the fraud triangle, often conceived as:

Opportunity + motivation/pressure + rationalisation

However, it is increasingly recognised that our understanding of causation must go beyond individual transactions and incentives. Instead, corruption must be understood in terms of networks, actors, and systems, and as part of the broader socioeconomic context.

One way to visualise this complexity is to distinguish between four categories of drivers that generate and sustain a pattern of corruption: (a) principal-agent/institutional problems; (b) collective action problems; (c) justifying norms, values, and pressures; and (d) the short-term functionality of corruption. These are summarised in Table 1 and detailed at more length below. In any given context, the strength and combination of drivers, as well as the actors involved, will vary. Moreover, the drivers are not always easy to separate from each other and in some case they may overlap. In contexts of systemic corruption, a combination of multiple drivers is likely.

The drivers in turn are influenced by the broader context, such as colonial and Cold War legacies, level of development, and type of economic system. The relative importance of the sector- or region-specific context versus the national and international context will vary depending on the corruption problem in question. International financial flows and tax havens, for example, are more likely to play a role in facilitating political corruption as compared to other types of corruption. Cooley, Heathershaw, and Sharman52f83846029e demonstrate how globalisation enables grand corruption with the aid of ‘professional intermediaries’, shell companies, and new types of financial investments.

2.4.1. Principal-agent/institutional problems (‘it is possible’)

When individuals with decision-making power (‘agents’) have access to asymmetric information and the opportunity to use it, they are tempted to pursue private interests at the expense of the ‘principal’ that delegated them this responsibility – and, by extension, at the expense of the public good. This is more likely to occur when the agent has a high degree of monopoly and discretion and when there is insufficient transparency and accountability. Corruption thus flourishes in settings characterised by non-existent, weak, or poorly implemented legal and regulatory frameworks, as well as insufficient oversight by managers, peers, watchdog agencies, and the public.

An alternative way of framing corruption driven by a dearth of formal rules and incentives is in terms of the risk-versus-reward ratio, also known as the cost-benefit ratio. Individuals are more likely to engage in corruption when the gains are significant, the act is relatively low-cost and easy to undertake, and the likelihood of detection and penalty is small. The low likelihood of penalty can stem from a range of factors, such as insufficient capacity to enforce the law or a tacit agreement that certain actors are above the law.

When a principal-agent problem is present, a range of technical solutions can be put in place to strengthen the formal ‘rules of the game’, such as establishing an independent anti-corruption agency, imposing asset and income declaration requirements, and training investigative journalists. The exact details of each tool will be essential. For example, will the asset and income declaration be mandatory or voluntary? If mandatory, will it be enforced? Will the body tasked with reviewing submitted forms have sufficient capacity to verify and analyse the content? And will the public be granted access to submitted forms?

2.4.2. Collective action problems (‘it is inevitable’)

According to collective action theory, when corruption is widespread and accepted under the informal rules of the game, the practice will persist due to lack of trust, lack of information, misperceptions, and free riding. Even though non-corrupt behaviour is in the group’s best interest, individuals will continue to participate in corruption if they expect that others will do so as well. In fact, it is in each individual’s best interest to continue to do so. A person who does not participate risks losing out on income, services, or opportunities, but without the ‘reward’ of knowing that their righteous behaviour is helping to bring about a change in collective practice.

Collective action problems can also stop people from demanding change or holding powerful figures to account for acts of corruption. Unless individuals can trust that enough people will take a stand together and that genuine change is possible, they are unlikely to act. At best, acting as a lone anti-corruption warrior may be a waste of time; at worst, it can lead to adverse consequences for the individual and their family, group, company, or (in the case of a donor agency) country.

Elites, civil servants, civil society, citizens, the private sector, and donor agencies are all potentially subject to collective action problems. Most civil servants, for example, probably would prefer to work in a corruption-free institution if recruitment and promotion were merit-based and salaries were sufficient and paid on time. Similarly, most citizens would likely vote a corrupt government out of office if they believed that the new government would be less corrupt and that any benefits that they currently receive through patronage systems would be exceeded by better public services.

When a collective action problem is present, technical anti-corruption solutions will not be enough. People’s expectations and levels of trust need to be altered. Coordination and cooperation will also be key.

2.4.3. Justifying norms, values, and pressures (‘it is legitimate and expected’)

In some cases, a corrupt practice is considered acceptable or even expected. Such a situation can arise when there is a social norm mandating a corrupt practice that is in conflict with formal rules or laws. ‘A social norm is an unwritten rule, derived through social interaction, that guides behavior within a group’.00e8c2d14a60 Adherence to the norm is maintained through sanctions, such as social shunning or threats, and through rewards, such as a promotion. Jackson and Köbis7b9cc41eff9f identify four social normative pressures: sociability (‘I have to return the favour’), kinship (‘family first’), horizontal (‘my colleagues expect me to’),6257da0f9b51 and vertical (‘I am forced from above’).

The behaviour of individuals is also affected by their morals, ethics, values, and ideologies, as well as their personal circumstances. Individuals and groups may hold mental models that justify or normalise corruption, such as a belief that corruption is normal and expected.

Curbing corruption that is driven by norms and values is not easy. The pressures driving corrupt behaviour need to be relieved, shifted, or countered. For more information on norms, behaviours, and the cognitive psychology of corruption, see Dupuy and Neset,e857f4ad95fb Fisman and Miguel,d8488ab8c1bf Jackson and Köbis,0330aecae2f2 Hoffmann and Patel,7f58721d00ff and Stahl, Kassa, and Baez-Camargo.bf6fb829c550

2.4.4. Short-term functionality (‘it solves a problem’)

Functionality is a driver of corruption when the corrupt act temporarily solves a problem for an individual or group. If an individual makes an informal payment to a health care worker to secure needed medical treatment, the bribe has solved an important problem for that person. If informal payments to health care workers help keep them in a remote, rural community, then such payments arguably serve a social purpose for the community. Similarly, while high levels of corruption can lead to conflict, corruption in the form of economic rents can also reduce the risk of conflict – at least in the short term.4c3f24df6e00

Functionality drivers are more likely in contexts with weak institutions and processes. In some cases, corruption can increase efficiency. However, in the long run, society is collectively better off if the ‘need’ for corruption is eradicated. Controlling corruption will also lead to more institutions becoming functional.

Some corruption experts define functionality quite broadly. Indeed, most corruption is functional in the sense that it serves a purpose. Politicians use corruption to win elections and control opponents; businesses use it to secure contracts and bypass regulations; and students use it to pass exams and get a job. However, many of these ‘solutions’ are in fact symptoms, rather than drivers, of corruption.

There should be a high threshold for defining corruption as functional. Consider the example of a patient who makes an informal payment to a health worker. The informal payment serves a purpose: the patient accesses health care in a reasonable time frame. But why did the need for such a payment arise in the first place? Her action was likely influenced by one of the three drivers noted above: for example, lack of staff oversight and accountability could have created an environment in which health care workers have an opportunity to demand additional payments. Her action may also be traced to corruption elsewhere: for example, grand corruption within the ministry of health may have resulted in a scarcity of health care services. In addition, because the informal payment helped the patient jump the queue and access treatment, it may have resulted in another patient’s treatment being delayed or denied. In short, her gain came at the expense of another’s loss.

Most functionality drivers require long-term solutions such as economic growth, institution building, and peacebuilding. For further details on the functionality of corruption, see Dupuy,049fbf18b89a Marquette and Peiffer,e105f7e44238 and Baez Camargo and Passas.b60d38aa0993

2.4.5. Unpacking, linking, and contextualising the drivers

‘Corruption is not just the behavior of some venal officials in a particular agency; it often represents the operating system of sophisticated –and successful – networks.’ – Chayes82de91371a2c

The four drivers are broad categories, each of which requires further unpacking. Lack of accountability may, for example, stem from lack of a free press, vibrant civil society, strong codes of conduct, or merit-based recruitment and promotion guidelines – to name a few. This diversity is illustrated by Table 2, which provides a sample of the range of potential drivers of widespread bribery, gratuity payments, and off-the-books fees in the health sector. The table makes clear that a given corrupt act can be driven by a multitude of reinforcing factors – some more easily solved than others.

Each driver is also influenced by the political, economic, social, historical, cultural, international, and conflict-related context. These factors not only contextualise corruption, they can help sustain and drive it. Chayes’s7a8c2cea942d case study of corruption in Moldova shows how international factors can contribute to corruption and sustain corrupt regimes. The paper identifies a range of international factors, such as money laundering, sex trafficking, smuggling, and foreign assistance.

For an anti-corruption intervention to have a sustained impact, it needs to either be feasible within the current system or sufficiently alter the system.f0efbbefa58a If not, gains will likely be short-lived and the probability of negative unintended consequences will be high. In Ghana, for example, efforts by the government to double police officer salaries without increasing the punishment for collecting bribes or the salary of other public officials led to an increase – rather than the anticipated decrease – in petty corruption by police.c15af4ba1656

Even when an anti-corruption intervention significantly alters the system, a sustained reduction in corruption is not guaranteed. This is because corruption is not a static process. Corruption is robust and adaptive. Its causes and effects are interconnected and non-linear, and the relationship between cause and effect is not one-way, nor is it always immediate, direct, or proportional.e2ff15fa49a0 Anti-corruption efforts may therefore lead to the emergence of new forms of corruption or to the emergence of corruption in new places, as corruption shifts to or is scaled up elsewhere.

2.5. Corruption can be systemic

In some cases, corruption is not only widespread, but systemic. Systemic corruption is generally characterised by three dimensions:9571bcbbdb5a

• Multi-actor organisation. Systems require coordination among multiple actors. In the case of corrupt systems, each corrupt act is perpetuated not by two individuals but by a set of actors often connected in a network or ordered according to implicit rules. The level of organisation can be loose or tight, and the scale of coordination will also vary. The connections can extend beyond government to penetrate the private sector, criminal elements, and even the institutions that are responsible for tackling corruption. There is always a ‘coordinated coalition of the corrupt’. Because it is organised, systemic corruption is also adaptive; it will react to changes in the context.
• (Partial) institutionalisation. The functioning of a system is not reinvented for each action but is underpinned by rules that can be explicit or implicit. Systemic corruption means that each act is not a one-off transaction but reflects informal rules of the game that define how corrupt actors interact. The level of institutionalisation will vary: the corrupt rules may exist within or alongside the formal rules and conventions, or crowd them out. For example, bribe taking in procurement may be built into the process by which ministries administer tenders. If so, the formal rules will be undermined and enforcement organisations can have little bite. Even if individual politicians or bureaucrats change, or new rules are introduced, the institutionalised rules of corruption persist. This can explain why corruption trends often seem to be stable.
• A broader rationale. A key quality of systems is that they tend to have some internal reinforcement mechanism that sustains them – or at least prevents quick disintegration. Often this means a collective goal or purpose beyond individual self-interest. In corrupt systems the purpose of corruption may be, in part, to serve broader social and political functions rather than only private gain. These will vary across contexts. Individuals engaged in corruption may be driven to fulfil a social norm (e.g., an expectation to help one’s kin and network), to accomplish political goals, or to attain some other end (e.g., a desire to access scarce public services or maintain political stability and peace). If so, corrupt actions will be integrated to varying extents in political/power and social normative logics. Corruption becomes deeply rooted and self-reinforcing. In contexts of systemic corruption, all four drivers of corruption will, to varying degrees, likely be present. The drivers, along with the wider context, mutually reinforce one another, creating a corruption equilibrium that is difficult to pierce.

For an example of systemic corruption, see the Corruption, Justice and Legitimacy Program’s systems map of the dynamics driving police and judicial corruption in northern Uganda.96a154c0b31e

3. Anti-corruption efforts, past and present

3.1. Types of anti-corruption interventions

Good governance and anti-corruption rose to prominence on the development agenda in the late 1980s and early 1990s, resulting in massive global efforts to curb corruption starting in the mid-1990s. While this intense focus continues, the thinking about anti-corruption has evolved, along with types of anti-corruption interventions.

Safeguarding aid funding from being stolen or misappropriated is still the first priority for donor agencies and their implementing partners. However, donor agencies, governments, the private sector, civil society, and citizens around the world are also working to combat corruption more broadly. Anti-corruption interventions can be global, regional, national, subnational, sectoral, or institution-specific. They can be explicit (e.g., an asset declaration law) or implicit (e.g., civil service reform). They can be stand-alone (e.g., development of a national anti-corruption strategy) or integrated (e.g., training for parent-teacher associations as part of a larger education programme). They can be top-down or bottom-up, technical or political, internal or external.6c9c44a6da65

A distinction can also be made between direct and indirect approaches. Indirect approaches generally are not designed with anti-corruption as a goal, but research suggests that they may contribute to a reduction in corruption. Such approaches include decentralisation, increasing access to education, promoting gross domestic product (GDP) growth, and expanding internet access.

The vast majority of anti-corruption interventions are designed to address principal-agent drivers of corruption. However, a few interventions aim to address social norms, collective action problems, and the short-term functionality of corruption. There is no toolbox for these three drivers. However, examples of potential approaches include supporting trend setters and reformist politicians; values, ethics, and anti-corruption education; integrity pacts; rotating staff; and formalising the economy.

Considerable resources have been applied to develop and implement a range of international anti-corruption conventions, norms, and standards, including the United Nations Convention against Corruption (UNCAC). There are also a number of multi-country interventions, such as joint stolen asset recovery efforts. In addition, developed countries are increasingly recognising the role that their own governments and companies play in enabling and driving corruption in developing countries, as well as the need for a whole-of-government approach to anti-corruption. However, ‘although global attention on combating [illicit financial flows] has increased, the scale of donor support is relatively modest’.35baaa6c0950

Box 2 provides a sample of the range of anti-corruption tools that have been employed at national and subnational levels. As noted, these tools have primarily been developed to solve principal-agent problems. For alternative ways of categorising these tools, as well as slight variations in the tools themselves, see Kirya;3571fb2da9c8 Jackson and Köbis;fb9b41808be6 the United Nations anti-corruption toolkit;c5dbeb327f9b Curbing Corruption’s reform approaches and experiences; and the Index of Public Integrity website (especially the section headed ‘What can the IPI be used for?’).

An overview of potential approaches to social norms–related drivers of corruption can be found in Jackson and Köbis.14266c60988c To aid in the development of anti-corruption strategies that go beyond traditional approaches, Khan, Andreoni, and Roy2a4d46720f26 have identified four clusters of strategies: (a) aligning incentives, (b) designing for differences, (c) building coalitions, and (d) resolving rights.

As discussed in the remaining sections of this paper, anti-corruption interventions should employ a problem-centric approach whereby tools are selected and tailored based on the specific drivers and actors involved, as well as the overall context.

3.2. The impact of decades of anti-corruption efforts

A commonly held assessment of anti-corruption is that despite decades of effort, only a handful of countries have achieved significant and sustained reductions in corruption.1d3ca625d658 Few, if any, of these national ‘successes’ are attributable to development assistance.f215bf9653f4 Encouragingly, there are examples of anti-corruption interventions that have had more targeted impact. Still, the overall conclusion is that anti-corruption efforts have not delivered sufficient results.8fc84dd1102b

Many observers attribute the disappointing results to weaknesses in conventional anti-corruption approaches. Indeed, the frequency and scale of success is likely less than what could have been achieved had our current understanding of corruption and anti-corruption been applied. However, whether a specific anti-corruption tool or intervention is effective is not a simple yes or no question. A number of key factors need to be taken into account. An appraisal of the anti-corruption landscape must also consider time horizons and assess the extent to which our understanding of corruption and our approaches to anti-corruption have evolved.

3.2.1. A complex, never-ending process

Curbing corruption is immensely difficult. It also takes time. The institutions, laws, practices, and norms of the world’s least corrupt countries evolved over decades, if not centuries. The process itself is not linear, nor is it ever complete.4b17ca6123aa Long periods of no progress can precede a ‘big bang’,daed4ed3dac8 and steps in the right direction can be followed by setbacks. Compromise and opposition are inevitable. The complexity of this journey is vividly captured by Grindle,f83fc5273dea who documents the difficult transition to merit-based systems across a number of countries.

An assessment of impact also needs to consider the context in which anti-corruption efforts are being implemented. It is far more difficult to reduce corruption in fragile states and in countries where corruption is not only high but systemic.46dcba12380d As Philpa1f9455a763c points out, ‘The more embedded corrupt activity is within a society’s broader social and cultural mores the more difficult it is to deal with, because there will be more things to change, more reflexive and organised resistance to such change, less legitimacy attaching to those who promote the changes and considerable collective action problems (such as who is to change first).’

In some cases, given trends towards rising corruption, maintaining current levels of corruption may be considered tantamount to success. Those who benefit from corruption have an incentive to both enhance their gains and undermine anti-corruption efforts. Shocks such as civil wars, financial crises, the toppling of dictatorships, and natural disasters present opportunities for reform. However, they also offer opportunities to increase corrupt activities, as checks and balances are often lower in times of crisis.

Anti-corruption efforts are also vulnerable in times of stability. Reformers can be removed from their posts, arrested, smeared in the papers, threatened, co-opted, or even killed. Similarly, reform-intent institutions can have their power curbed, especially if they start to become too great a threat. Corrupt actors and networks can also respond to anti-corruption interventions by finding ways to either directly skirt or undermine them or by shifting their corrupt practices to elsewhere. This phenomenon is often referred to as ‘backlash’, ‘squeezing the balloon’, or the ‘boomerang effect’, and efforts to repress corruption where it pops up has been compared to a game of whack-a-mole (Box 3).

It is therefore not surprising that only 16 countries experienced statistically significant improvements in their CPI score between 2012 and 2018.36c7874fe97b The resilience, adaptability, and ingenuity of corrupt actors and networks is further explored in Box 4 and in the U4 article ‘Managing a Hostile Court Environment: Common Challenges and Recommendations’.1111c48715a4

3.2.2. Lack of impact versus lack of documented impact

It is important to distinguish between lack of impact and lack of documented impact. Several studies have synthesised existing knowledge on what works in anti-corruption, including Rocha Menocal et al.,3c0be2a587ab Johnsøn, Taxell, and Zaum,51e731b29712 Hanna et al.,2cf168a8fbc6 and McGee and Gaventa.042666c6a7d2 While these studies find that many conventional anti-corruption tools appear to be ineffective, they also conclude that there is insufficient credible evidence on what works and why. The studies also find that some tools show promise, as detailed in Box 5. Encouragingly, researchers are increasingly shifting their focus from studying corruption itself to studying how best to combat it. The Anti-Corruption Evidence (ACE) programme, funded by the UK Department for International Development (DfID), and the Targeting Natural Resource Corruption programme, funded by the United States Agency for International Development (USAID), are two such examples. There are also a number of donor-funded evaluations that, if made public, could likely provide insights into the extent to which anti-corruption initiatives have had an impact, especially at the sector, institution, and programme levels.

Evaluations and research depend, however, on quality evidence. Although the situation is improving, there is still a need for more and better baseline and monitoring data for anti-corruption interventions.31fd36f5d21f This is especially the case for interventions that are implicit or mainstreamed. Not all anti-corruption programmes are labelled or advertised as such, as in some circumstances an approach viewed as innocuous – perhaps because its stated goals do not mention corruption – can be more effective. As Robillard and Robillarda67235a6cc66 point out, while there are advantages to making anti-corruption interventions explicit, being visible also ‘makes it easy for people in power to applaud these initiatives in public – and to avoid them, or even undermine them, in private. By the time the project reports are written, the systems that facilitate corruption will have shifted, adapted, and survived.’ The challenge with the implicit approach, on the other hand, is that in order to keep such efforts ‘under the radar’ and avoid pushback, the results may be less rigorously documented; this in turn makes the impact harder to determine. The lack of sufficient evidence on mainstreamed anti-corruption interventions is discussed in the subsection below.

The challenge is not just the lack of evidence. It is also the difficulty of accurately measuring corruption.e654adb1165b Corruption is a complex phenomenon, whose drivers and manifestations vary both across and within countries, as well as over time. It therefore cannot be fully captured in a global index. Nor is there any agreed universal definition of corruption. Without conceptual precision and strong empirical measures, the level and degree of corruption cannot accurately be compared over time or across societies.079938cbd890 An agreement to jointly develop a corruption measurement framework was taken at the 2019 Conference of the States Parties to the United Nations Convention against Corruption (COSP) through Resolution 8/10.

However, even with continued advances, the extent to which global indexes can be used to gauge incremental progress towards curbing corruption will remain limited. They are simply too aggregated and imprecise. For an exploration of how global corruption indexes may fail to capture significant anti-corruption progress at the country level, see Göbel.657c839ac221

3.2.3. Determining what impact entails

As discussed above, accurately measuring corruption is difficult. In addition, reductions in the magnitude or frequency of corruption may not be the best measure of impact. Under certain circumstances, relatively small bribes can do more damage than larger bribes elsewhere. In other cases, one large act of corruption can do more damage than many small ones. What ultimately matters is reducing the harmful impact of corruption on development, and an impact assessment therefore needs to estimate the effect of a reduction in corruption on specific development outcomes. Depending on the end goal, such outcomes could be anything from economic growth to improved learning outcomes in schools.

The impact of a particular reform may also be modest, both in scale and in recognition. A 2020 World Bank report, ‘Enhancing government effectiveness and transparency: The fight against corruption’, highlights 21 case studies from around the world. The authors conclude that many of these reforms have not led to quantifiable savings, jumps in global survey rankings, extensive press coverage, or high-profile convictions. Yet in their own way, each reform has made a positive contribution to the fight against corruption.

More guidance on how to measure anti-corruption successes, as well as progress towards curbing corruption, would be useful.

3.2.4. Managing expectations

The impact debate needs to consider the difficulty of determining attribution, as well as what can be considered within the sphere of influence of development actors, agencies, and interventions. As already discussed, it is an immensely difficult and complex undertaking to reduce corruption in a way that is significant and sustainable. The overall level of corruption is also an inappropriate measure of success.3ba3190b311f Programme-specific indicators carefully selected to gauge the intended impact of a particular intervention are far more useful.

Expectations also need to be aligned with an intervention’s budget, time frame, activities, and degree of political support. Anti-corruption agencies (ACAs) are a case in point. As Schütte40773266d139 points out, ‘ACAs are often seen as a last resort to reduce corruption, but these unrealistic expectations have quickly given way to perceptions of failure when the institutions have not delivered the hoped-for results.’ Yet the mandate, manpower, degree of independence, and context in which these agencies operate all vary significantly. An assessment of ACAs’ impact should not, therefore, ask whether they are effective, but whether, why, to what extent, and in which contexts they are effective.15268c485cc1

The same argument can be extended to all interventions and tools. Rather than a binary analysis of their impact – as effective or not – we need to understand whether, why, to what extent, under what circumstances, in which contexts, in which combinations, and for whom anti-corruption efforts have a direct or indirect impact on corruption levels and, ultimately, on development outcomes. Some analysis is already available. For example, Mungiu-Pippidi and Dadašov6102fd498dd7 find that freedom of the press and independence of the media can affect the effectiveness of anti-corruption legislation. Similarly, Adam and Fazekas599ee04cbc87 find that information and communication technology can be used to both enhance and undermine anti-corruption reforms. Additional examples of such research are provided in Box 5 and Box 6. However, more work is needed to shift the discourse away from ‘What works?’ and then provide a sufficient evidence base for a more nuanced discourse.

While anti-corruption agencies vary in terms of their mandate, staffing, and support, they are all explicit, stand-alone anti-corruption interventions. However, as illustrated in Box 2, many anti-corruption interventions are implicit, indirect, or mainstreamed. Their primary goal is not necessarily to reduce corruption. An education programme, for example, may include anti-corruption activities such as establishing community monitoring mechanisms, strengthening parent-teacher associations, raising awareness about students’ rights to free education, building teacher housing, and making budgets more transparent. However, if the goal of the programme is to increase school enrolment, completion, and learning rates, its results framework will likely have few if any outcome or impact indicators specific to corruption. Even if the framework does include corruption-related indicators such as leakage of funds and teacher attendance rates, the scale of the programme’s impact will likely not be sufficient to bring about major changes in overall sectoral or national corruption levels. Indeed, most anti-corruption interventions can only be expected to contribute to change at the project level.

An assessment of the effectiveness of anti-corruption efforts also needs to take into account stakeholders’ motivation and commitment levels. In some cases, anti-corruption strategies and institutions are created in response to internal and external pressures and do not reflect a genuine commitment to reform. A country may, for example, need to appear to be making progress on good governance in order to receive international aid. Anti-corruption can also be used as a smokescreen for actions to neutralise political opponents. Donor governments and aid agencies also have conflicting motivations and incentives. Foreign aid is both a tool for development and a tool to strengthen military alliances, commercial access, and cultural influence. In addition, donor governments and aid agencies may prioritise anti-corruption initiatives designed to ensure that loans are repaid and that funds are not misappropriated. More broadly, anti-corruption interventions are subject to the general challenges and constraints of development assistance, including short-term funding cycles, pressure to show immediate results, and preference for quantifiable outcomes.

The impact of a given anti-corruption intervention depends upon a number of factors beyond the approach itself. Significantly and sustainably reducing corruption is particularly difficult in fragile states and in settings where corruption is systemic. In contrast, it is more feasible when enabling conditions are in place, such as having an educated and autonomous professional group that demands good governance.abf55f8aac45

3.2.5. From whether to when a tool is effective

A red thread throughout this paper is that anti-corruption efforts are more likely to be effective when they are selected and designed to fit the specific context; when there is an enabling environment for reform; and when they are part of a broader, tailored package of interventions. This has a number of implications for gauging impact.

Determining whether a specific anti-corruption tool works can be difficult given that the same ‘symptom’ in two different contexts cannot be addressed with the same combination of tools. Lack of results in an anti-corruption project could indicate either that the tool itself is ineffective or that the tool was misapplied. Some tools are only effective in certain contexts.65d95840ee1a As Khan, Andreoni, and Royfccbbc8226bc put it, ‘Policy effectiveness depends on the presence or absence of supporting policies and governance capabilities but […] just saying that ‘context matters’ does not help. Societies are differently constituted, politically, institutionally and culturally. […] [We need to use] frameworks like political settlements to generate evidence on combinations of policies that are effective in countries with identifiable types of political settlements.’

Isolating the impact of a particular intervention from the effects of wider reform efforts and determining the direction of causality is also difficult.da9b3cb9bb86 Given interdependencies, it is not surprising that many of the econometric studies of anti-corruption have found no or mixed effectscfaa6964164724c366a8503b Salary reforms, for example, are associated with both a reduction and an increase in corruption (see Box 6).

A more nuanced approach to understanding ‘what works’ requires particular types of methodologies at both the design and evaluation stages. ‘Generating useful evidence depends on effective strategies for making the problem of interdependence tractable for analysis’.998d7cf84044 Promising approaches during the evaluation stage include process tracing and qualitative comparative analysis. Guidance on how best to evaluate and determine the impact of anti-corruption interventions has been produced by Johnsøn and Søreide.2c623d281344

3.2.6. Gaps between knowledge and practice

‘Despite the difficulties that have surfaced, the one conclusion we should not draw is that we would be better off simply giving up.’ – Mason4bfbc7489f37

An assessment of the field of anti-corruption needs to go beyond an analysis of the extent to which corruption levels, and the concomitant impact of corruption on development, have been reduced. We must also look at how our understanding of corruption and our approaches to anti-corruption have changed.be148e4d2b0e

Today, corruption is widely recognised as an impediment to sustainable development, as evidenced by the inclusion of corruption in Sustainable Development Goal 16. While some countries with high levels of corruption have achieved economic and social development (e.g., the ‘Bangladesh paradox’), and while corruption can serve a short-term function, a wealth of evidence clearly documents that corruption increases inequality, impedes growth, undermines the legitimacy of governments, and weakens the public’s trust in democracy.

Corruption is also widely recognised as a global problem, fuelled by cross-border linkages. The perspective that corruption is primarily a problem of and in the global South still exists, but it is increasingly acknowledged as outdated. The focus on anti-corruption within development assistance has also expanded from protecting one’s own programmes to additionally working to address corruption in aid-recipient countries more broadly – though the primary focus remains on corruption risk management.

A number of important global standards and norms have been developed. UNCAC has been ratified by 187 states parties, including 181 United Nations member states. Most countries have also put a range of formal anti-corruption institutions and laws in place. However, the extent to which these are functional and operationalised varies.

Researchers and development actors are increasingly recognising the limitations of donor agencies’ traditional methods of work, as well as the limits of viewing corruption solely through a principal-agent framework and ‘state modernisation’ lens. Research on social norms, collective action, and systemic corruption has increased. The need for tailored, multi-pronged approaches is widely acknowledged. The World Bank’s 2017 World Development Report, for example, admits the shortcomings of anti-corruption efforts that attempt to replicate the reforms undertaken by low-corruption countries. It also highlights the importance of addressing underlying causes, entrenched power structures, and social norms.40af86893f79

The next step is to take this emergent understanding from paper to practice, as applying a fragmented ‘toolkit’ approach based on the principal-agent theory and traditional approaches to programming is still a common anti-corruption tactic.7a7dc29c24c2 There is also a need to keep anti-corruption on the international and national agenda, as the overall momentum appears to have waned in recent years.

4. Basic principles of anti-corruption approaches

As we have seen, controlling corruption is immensely difficult, and the process is neither linear nor ever complete. A unique combination of approaches, tools, and actors is needed to collectively address the root causes of corruption in a given context. There is no single blueprint. However, quantitative and qualitative studies of countries, cities, ministries, and sectors that have achieved significant reductions in corruption have identified some potential pathways to success. Studies of what has not worked have also yielded useful insights. This section aims to collate these overarching principles, a number of which are further explored by Jackson.a4a1196d62af

4.1. Let the context determine the approach and scope

Because there is no single pathway to reform,f870d81f369e we need to talk about ‘best fit’ rather than ‘best practice’ approaches. The feasibility of reform is determined in part by whether there is an enabling environment. In many cases, the focus may need to be on laying the groundwork for future anti-corruption efforts. In fragile contexts, the natural starting point may be to build trust through the provision of basic services.2508b584fec6 Similarly, in neopatrimonial regimes and societies where leaders are unwilling to address corruption, interventions that build demand for good governance, empower citizens, strengthen civil society, and work with civil society and the private sector are potential entry pointsd5aaabb67244c7ae884c0901 However, even similar contexts will require different approaches.

4.2. Think in terms of corrupt systems and act accordingly

When corruption is deeply entrenched or systemic, it has to be recognised and dealt with as such. Treating corruption as if it were a series of individual bad acts will not work. Similarly, treating manifestations of corruption rather than the underlying drivers will not work. The system underpinning corruption needs to be understood and either altered, co-opted, or countered.

A recent review of the state of knowledge on anti-corruption policies identified the ‘importance of developing comprehensive, integrated anti-corruption strategies’ as one of the key lessons learned.0f25d9a3d686 Stand-alone interventions, fragmented approaches, and reforms that view corruption as a collection of individual transgressions are likely to fail in the long term. Instead, broad-based, multi-pronged action is needed.

At the same time, implementing across-the-board anti-corruption strategies can be difficult in light of limited resources and vested interests. The difficulty increases when structural factors such as a complex, developed economy are not in place.c6ff34382983 Many therefore argue for sequentially targeting institutions, sectors, cities, or specific situations where reform efforts are both politically feasible and have the potential to have a high impact.e178f8256873 Proponents believe that such an approach will not only directly benefit targeted areas, but ultimately contribute to a more enabling environment for national reform. More evidence on the potential spillover effect is still needed.98c256f9370d However, it is clear that even targeted reform efforts need to employ a holistic approach and be linked to wider plans. As Taylor28506952c284 argues, ‘Bursts of anticorruption policy seldom develop into lasting shifts in the overall corruption equilibrium if these policies are not embedded in a broader accountability effort.’

4.3. Shift the equilibrium

Social norms, entrenched interests, and collective action problems make corruption highly resistant to reform. To achieve and sustain gains in corruption control, researchers suggest, the existing corrupt equilibrium must be disrupted. But scholars disagree on whether this requires a ‘big bang’ or a series of incremental, sequential changes that can bring about ‘virtuous circles’ of reform.

Big-bang proponents argue that gradual reform efforts are hampered by collective action problems and by individuals and interests that benefit from the status quo. They maintain that best way to bring about a new equilibrium is to take advantage of a crisis or other major shock to the system to unleash comprehensive, rapid reform. A crisis can enable reformists to shift norms, change expectations, and implement the necessary changes before vested interests have time to block or undermine them.656a4e55e2da Proponents of sequencing, on the other hand, argue that the big-bang approach is risky, expensive, and difficult to bring about. They also maintain that a gradual approach can be more flexible and sustainable, as systems, practices, and support are built step by step.

Historically, some countries have controlled corruption through a gradual process, while others have made headway through a combination of shocks and sequencing.c677ce616f84 Regardless of whether big-bang or incremental approaches are best, it is clear that shocks to the system – such as corruption scandals, economic crises, regime changes, and elections – offer potential windows of opportunity for reform.29c1fdfe3552

4.4. Collaborate, coordinate, and build trust

Collaboration, coordination, and trust building make it easier to disrupt a corrupt equilibrium.810a9f7719df Collaboration allows for multi-pronged approaches, enhances the bargaining power of reformers, makes more people part of the solution, and increases the likelihood of overcoming collective action problems. According to Johnston,add207f0c2a7 broad-based action, trust, and commitment are particularly important when ‘reform amounts to mobilising the weak and divided against the strong and entrenched’. With regard to coordination, Klitgaard481a3b7326ff identifies three priority approaches: (a) coordinate government institutions, (b) involve public servants, and (c) mobilise citizens and the business community. Coordination among external actors is also important.

4.5. Maintain momentum

Changing from a high- to a low-corruption society takes time. Even in countries such as Georgia and Estonia, where transitions were spurred by a big bang, subsequent reform efforts were sequenced over many years.e6ae9f80726d The process is never complete. No country has ever achieved zero corruption. In addition, even when an acceptable equilibrium is reached, there is always a potential for backsliding and backlash. Maintaining the momentum is therefore critical.

Momentum is essential for both seemingly simple and complex changes. As Taylor5650d15a41f1 points out, ‘The marginal effects of anticorruption policy bursts often diminish over time.’ The passage of a freedom of information law, for example, may lead to an increase in transparency. But over time there is a risk that bureaucrats will find ways to block unwanted inquiries. Long-term investments in building citizen demand for good governance and a well-functioning judiciary, therefore, are a necessary complement.723da7629021 Reform efforts need to be continuously renewed, but sustaining engagement and momentum is not easy. See chapter 8, ‘Staying power: Building and sustaining citizen engagement’, in Johnston0c19f89ef38b for details.

4.6. Build and harness political will and commitment from elite groups

Human agency and political will are essential factors for reform.3ccdf583f90c Many of the contemporary anti-corruption ‘success cases’ were spearheaded by top politicians, including, among others, the current president of Rwanda, Paul Kagame; the former prime minister of Estonia, Mart Laar; the former prime minister of Singapore, Lee Kuan Yew; and the former mayor of La Paz, Ronald MacLean-Abaroa. The agencies considered to be ‘islands of integrity’ within Uganda are similarly headed by leaders intent on reducing corruption within their agencies.0b8131d0c4df

Powerful and committed groups inside and outside the government – including bureaucrats, citizens’ groups, civil society, and the private sector – have an important role in pressing for change. Mungiu-Pippidi3d002a4ccc98 states that ‘without educated and autonomous professional groups fighting for good governance because it is in their best interest, sustainable progress in fighting corruption will not occur’.

They may pay a high price for their commitment, however. As discussed in section 3.2.1, reformists can be fired, transferred, arrested, smeared, threatened, and even – in extreme cases – killed. Support systems and networks for those putting their careers, reputations, and safety on the line can help counter this risk.

Reducing corruption is not always the primary objective of those championing reform. Politicians may pursue an anti-corruption agenda for a range of reasons, including a desire to stimulate economic growth, maintain power, appease voters, or attract foreign aid and investment. The genuineness of their commitment also varies. History suggests that leaders who spearhead anti-corruption reform can succumb to disillusionment, corruption, and autocratic tendencies if they stay in power too long. To determine whether there is adequate leadership for a good-governance coalition, Mungiu-Pippidi54ae9b3234b3 poses the following question: “Who are the plausible agents of change and how long would they remain so if they were to gain power?” To better understand political will and how to assess it, see Brinkerhoff25fe4d44b69d and Kukutschka.1c8b72b5896e

4.7. Combine top-down and bottom-up approaches

A combination of top-down and bottom-up approaches to anti-corruption may be the most sustainable path to reform. Anti-corruption efforts can be driven from the top, as was the case in Botswana and Singapore. However, without citizen demand for good governance, such top-down reforms are particularly vulnerable to reversal. There is also a risk that politicians will prioritise tackling petty corruption over grand corruption. Alternatively, bottom-up action and demand for reform can be the driving force for change. Grassroots and civil society organisations were behind right-to-information legislation in India221ae11b5298 and the ‘clean record’ bill in Brazil,2cc75b25beb0 and civil society also played a major role in anti-corruption efforts in the Republic of Korea.4fa6529a8f3d However, for sustained, macro-level changes to occur, bottom-up efforts appear to be insufficient on their own; a degree of political will at higher levels is essential.883bd2c4863d

The importance of bottom-up approaches is further confirmed by a recent review of security sector reform programmes, which found that initiatives ‘from below’ were the most likely to improve police accountability.1d12252db171 However, while bottom-up approaches can be effective, creating and sustaining this pressure is difficult due to social norms and collective action problems. Emerging research suggests that these impediments can potentially be overcome through a combination of efforts to raise awareness of the importance of good governance, strengthen media and civil society, increase access to credible information, expand social media usage, build trust and coalitions, facilitate coordination, empower citizens, and increase their role in the policy arena.a9fe0d702c5b In short, expectations must be changed.c6a91b2b8ebb More research on how to overcome collective action problems and change social norms is, however, needed.

With regard to top-down approaches, none of the success stories identified by Mungiu-Pippidi9da1330aa9cd and her colleagues took place in a federal state, suggesting that a strong central government may be an enabling factor for combating corruption at a national scale. Hong Kong’s anti-corruption agency is an example of how an institution’s ability to fulfil its mandate depends in large part on political support from above.59d7e2e9a11c Similarly, the experience of Bogota, Colombia, is an example of how a mayor can both implement formal changes and propel a shift in social norms.0025fc864615

A combined top-down/bottom-up approach may be able to overcome self-reinforcing corruption equilibria by changing incentives, shifting expectations, and overcoming collective action problems. As the World Bank points out, citing the example of Uruguay, ‘Important changes in history have been driven by coalitions between reforming elites and organized citizens that support reform initiatives’.6b001200fded This hybrid approach can be further expanded to include pressure from the side, via horizontal accountability bodies.34e661a563c0

The top-down versus bottom-up debate can also be framed in terms of whether to prioritise conventional national anti-corruption strategies or more targeted approaches. Evidence suggests that attempting to replicate the laws, institutions, and practices of today’s least corrupt countries in not sufficient to significantly reduce corruption levels where they are high. In fact, quantitative studies indicate that anti-corruption tools such as party finance legislation, autonomous anti-corruption agencies, whistle-blower protection, supreme audit institutions, and laws criminalising corruption as a special offence have, on their own, had little impact.1d22c165f2db This can be explained in part by the fact that policies and legislation are not always sufficiently operationalised. As Mason30b1382241d8 puts it, ‘Adequate law is the starting point, not the end. Implementation is everything.’ The effectiveness of an institution also depends upon its mandate, its manpower, and the context in which it operates. Top-down approaches are more likely to be effective when there is a credible commitment from those in power to address corruption and when such measures trigger or are complemented by a shift in social expectations and norms.980472af2927 See the section above, ‘Think in terms of corrupt systems and act accordingly’, for details on sequential and strategic targeting. In addition, see Jackson228160c0ab0a for an overview of the debate around the so-called state modernisation approach and the critique that transitions to integrity are not brought about by ‘reverse-engineering anti-corruption measures from those found in low-corruption countries’.

4.8. Change the institutional culture

As argued by Fisman and Golden,77b1b36e1f0b ‘To reform a culture of corruption, […] we need somehow to change everyone’s beliefs about how to behave all at once.’ Klitgaard06d517b508c5 identifies three complementary ways to change expectations and, in turn, the institutional culture:

• Punish major offenders, including senior officials within the political party in power;
• Build momentum by prioritising visible problems that can be addressed relatively quickly; and
• Bring in ‘new blood’.

Although more research is needed, the potential importance of changes that are comprehensive and shocking enough to shift institutional norms is further confirmed by Rothsteinb96257be51d5 and the World Bank.4c53d53aebc0 Indeed, corruption in the Republic of Korea’s bureaucracy was immensely reduced in the 1960s through such an approach. Within a month of former president Park Chung Hee coming to power, the top 10% of bureaucrats were dismissed while the rest were sent for two weeks of training. In addition, performance monitoring and staff rotations were put in place, and several leading businesspersons were jailed.13b36396db4a

Changing expectations is also vital to overcoming the collective action problems inherent in many bottom-up anti-corruption efforts. See the section above for details.

4.9. Harness external pressure and support

Countries that have successfully moved from a high- to low-corruption equilibrium have primarily done so by means of internal forces. However, external actors can play a role. Based on an analysis of past success cases, Klitgaard8fb0c0960a7b concludes that ‘providers of development assistance can contribute resources, knowledge, convening power and leverage that may help recipient countries reform’.

Mungiu-Pippidi34c4b3159356 and her colleagues find that context shapes the role that donor agencies can play. In neopatrimonial regimes, donors’ best approach may be to build demand for good governance, empower citizens, foster collective action, and strengthen media and civil society. In contrast, in countries characterised by competitive particularism, the best approach may be to support rising demand for good governance, identify plausible agents of change and help them build coalitions, and support the government in efforts to develop and build anti-corruption institutions and practices. These authors also find that ‘borderline countries, where particularism and ethical universalism wrestle for supremacy, present the best opportunity for international anticorruption efforts to make a serious impact.’ In such cases, their research suggests that donors should support the building of a sound good-governance programme that increases constraints and reduces opportunities for corruption. However, the potential role of development assistance depends upon domestic agency.

Fisman and Golden34329a1e5692 find that external forces can act as a trigger for change when a society is stuck in a high-corruption equilibrium. Unfortunately, their main illustration of external forces contributing decisively to broader change is Hong Kong in the 1970s – an example that is highly context-specific. They also highlight the external triggers of Italy’s 1970s Clean Hands investigation and the role that the International Commission against Impunity played in bringing charges against the president, vice president, and 26 other members of a customs corruption ring in 2015.

Social pressure to follow international standards and desire to gain membership in international organisations such as the European Union and World Trade Organisation may also encourage governments to implement anti-corruption reforms.02479ed2a968 However, as discussed above, there can be a mismatch between the priorities given to anti-corruption in speeches, on paper, and in practice. Countries and agencies on both the receiving and giving ends of development assistance can display such inconsistencies. According to Bauhr and Nasiritousi,d2e25f63f554 international organisations are best able to encourage public officials to follow international standards when they themselves prioritise good governance. Donor and inter-agency coordination and coherence is also important.8b2bb94804fd

It is important to remember that corruption is transnational. External actors that want to reduce corruption must also address international drivers of corruption. They need to go beyond technocratic solutions and, importantly, must be willing to ensure that they themselves do not inadvertently increase resources for corruption or contribute to a ‘theater of delusion’.79e892dd3358 See section 5.2 for a discussion of the principle of ‘do no harm’.

4.10. Reshape the policy arena

Elites intent on reform are not always able to initiate anti-corruption initiatives. And even successful initiatives are sometimes defunded or otherwise derailed. Why? Because technical solutions are not enough.42beac38a131 Anti-corruption interventions need to either be feasible in the context of existing power structures and social norms, or seek to alter them. The World Bank’s World Development Report 2017 recommends reshaping the policy arena to expand the set of policies that can be feasibly implemented. To do so, they identify three levers: change who participates in decision-making processes (i.e., enhance contestability); transform elite actors’ incentives; and shift elite preferences and beliefs. This approach is similar to the Problem Driven Iterative Adaptation (PDIA) approach to solving complex problems, which emphasises the need to cultivate authority and acceptance.

The importance of thinking politically is echoed by Mason,49000ec53cb6 who concludes that a ‘combination of technical and political interventions, adopted coherently in each location, could significantly change the extent of donor impact’. This means ‘working to reduce elites’ room for manoeuvre rather than relying on their ‘political will’ to reform themselves’.

4.11. Use a context-specific combination of direct and indirect approaches

As discussed above, some anti-corruption interventions are direct, while others are indirect and implicit. In some cases a direct approach is called for in order to put in place needed institutions, laws, and procedures and/or to demonstrate commitment to reform. However, in other cases, indirect and implicit approaches may be a better way to address the underlying causes of corruption, create a more enabling environment for reform, and/or reduce the likelihood of interventions being blocked or undermined by vested interests.

Research also indicates that anti-corruption interventions are not the only way to control corruption; indeed, reduced levels of corruption can come about as a by-product of other reform efforts. Estonia, for example, did not officially prioritise anti-corruption. Yet today it is considered to be the least corrupt of the post-enlargement European Union states.b4f9c195c2c1

According to Fisman and Golden,ed4fe607f309 approximately 60% of a country’s corruption level – as measured by the CPI – can be explained by its GDP. Mungiu-Pippidi (2015) similarly finds that nearly half the variation in control of corruption can be explained by a country’s life expectancy, education level, and income. The direction of causality is, however, not clear. It may be two-way, with corruption control and demographic factors reinforcing each other.4960387ffb51 Many countries also have a significantly higher or lower level of corruption than governance models would predict. Mungiu-Pippidic7c70a1da871 herself points out that ‘education and economic development have increased over the past twenty years without bringing better governance’ – a phenomenon referred to elsewhere as the ‘Huntington paradox’. Still, this line of research indicates that accelerating structural transformation is a potential avenue for long-term reform.

5. How to identify, plan, and implement successful anti-corruption interventions

‘If corruption was simply an incentive problem, it would have been resolved long ago.’ – Bo Rothsteinb34db60d4f64

‘Expecting those who benefit from the status quo actively to assist in organising its demise looks a decidedly optimistic basis on which to build an external donor’s anti-corruption approach.’ – Phil Mason7b724055488c

‘One should not underestimate the importance of political will in driving reforms [...] Nor should one underestimate the power of vested interests to block them.’ – Jesper Johnsøn41de51ba198d

The literature on corruption and development provides a number of insights on how to best identify, plan, and implement anti-corruption interventions. Many of these insights are theoretical and based on an assessment of ‘what doesn’t work’. They therefore require further testing. However, collectively they still encompass the best available advice for the field.

For suggestions on how to evaluate anti-corruption interventions, see Johnsøn and Søreide,8a33a236bf7f Befani, Wathne, and Devine,83707358f75f and Wathne and Devine.de259c416d46

5.1. Identification

The first insight is obvious yet fundamental: ‘To tackle corruption, we need to understand it’.3d7a20f63ea5 The identification stage needs to be both broad and deep, as well as context-specific and feasibility-sensitive. The scope of inquiry should go beyond the institution or sector being targeted for reform. It should identify the various types of corruption, their effects and level of harm, who is involved, what the drivers are, and how these factors interact. Such an inquiry involves exploring collective determinants of corrupt behaviour; the types and intensity of social pressures; the legal and institutional frameworks and practices (formal and informal) that facilitate or discourage corruption; the short-term functionality of corruption; and how corruption is understood and experienced by citizens, particularly marginalised groups.c0279f29b4fb

The identification stage should also identify possible entry points and map which stakeholders are for or against reform, why they hold those positions, and their capacity and level of power and legitimacy (Box 7). An understanding of political will, power dynamics, and ‘no-go areas’ (untouchable corruption issues) is also essential.8a51a6f0d4fb In short, what are the potential leverage points? And ‘how might the system ‘push back’ against efforts at reform?’.062aa2c6e248 The broader political, social, economic, and cultural context needs to be understood as well.

To concretise corruption in a given context, several lists with guiding questions have been developed, including those by Heywood,de658a4c5c95 Jackson and Köbis,442318820737 Hart,088e75ff504f Scharbatke-Church and Barnard-Webster,bc83fe8aea14 and USAID.f0688719a719 A selection of such questions is provided in Box 7. The list is not exhaustive, nor is the grouping under topic headings meant to be precise. In practice, these questions should be asked, if not simultaneously, then at least in a non-linear way.

A number of corruption measurement and analysis tools are available to aid in answering these questions. Static analysis tools include indices and surveys such as the Global Corruption Barometer, which disaggregates data by institution and service, and the Index of Public Integrity, which disaggregates data by the dimensions likely to reduce opportunities for corruption and increaseconstraints. Dynamic studies include political economy analysis, everyday political economy analysis, political settlement analysis, context analysis, institutional analysis, corruption vulnerability analysis, and corruption risk assessment.

Many donor agencies have developed their own specific tools and resources. These include, among others, Sida’s Power Analysis tool, BMZ’s Anti-Corruption WORKS workshops, GIZ’s Guidelines for Integrating Anti-Corruption into the Planning and Implementation of Technical Cooperation Projects and Programmes, UNDP’s530fa79b39fa Capacity Assessment of Anti-corruption Agencies, and USAID’s Anticorruption assessment handbook.1f548e52d162 In addition, Scharbatke-Church and Chivasb77003c6ff2e have produced a tool to determine whether corrupt patterns of behaviour are influenced by social norms, while Walton and Jacksonedd58cac9857 have developed a three-tiered analysis for understanding how informal systems of reciprocity function. Mungiu-Pippidi and Fazekasf891001ff518 have also released a very promising framework for assessing current measurements and offer a universal method for organising those measurements.

There are also a number of other identification and design tools that are well suited for anti-corruption interventions, such as stakeholder mapping, ethnographic studies, outcome mapping, theory of change, power and influence analysis, and Problem-Driven Iterative Adaptation. A detailed methodology for identifying entry points that are both high-impact and feasible can also be found in Uberti.79b307dc7de2

Practitioners and agencies can also complement intervention-specific context and needs assessments with secondary information. The ACE-SOAS Consortium, for example, has produced a number of country-level political settlement analyses. Other insightful studies include an ethnographic study of illicit and habitual corruption in Mozambique by Tvedten and Picardob255f8b06b11 and the study of social norms in Nigeria by Hoffmann and Patel.cc22a93df639 In addition, the Curbing Corruption website provides a sector-by-sector overview of different types of corruption that can be used to inform a context-specific analysis.

The 2019 U4 Guide to using corruption measurement and analysis tools for development programming08372615a853 and accompanying Measurement and assessment tools table and Reference table on matching measurement and assessment tools to corruption diagnostic questions provide a detailed overview of available tools for both static and dynamic analysis. Key take-aways from this guide are as follows: (a) multiple sources of information are needed; (b) administrative statistics, target surveys, proxy indicators, and other ‘homegrown’ data are usually more informative – albeit more costly – than internationally generated data; (c) data should be disaggregated by sector, institution, and marginalised group to the extent possible; and (d) dynamic analyses of the drivers of corruption should continue throughout the project cycle.fa24647f1da8

Encouragingly, there is a trend towards more robust analysis of corruption.2e8b323737bc However, many studies are too broad and generic to provide adequate guidance for specific anti-corruption intervention planning processes. Corruption risk assessments, for example, may lead to the selection of standardised anti-corruption tools without sufficient consideration of the specific context.89add8e11bf2 Similarly, political economy analyses may identify a ‘lack of political will’ without digging ‘more deeply into the many layers of incentives, rules and constraints – formal and informal – that characterise the host country’s institutional make-up’831d0339167e3dcdb2e11024

The importance of maximising the usefulness of an approach is further highlighted by Booth, Harris, and Wild.c3645503d245 These authors argue that political economy analysis is most useful when (a) it is problem-focused, (b) it takes place early enough and feeds into key design decisions, and (c) is used to inform programmes employing adaptive approaches.

There is also an emerging call to employ systems thinking or a systems-based corruption analysis in contexts of systemic corruption. Indeed, reformers thinking in terms of corrupt systems was a common factor in many of the anti-corruption success cases explored by Klitgaard.98a5f7754f07 According to CDA,36c6faf3f261 ‘Systems thinking is a mental model. It is a way of seeing interconnections among structures, behaviors and relationships that can help us identify the underlying causes and uncover opportunities for creating positive change.’ Scharbatke-Church, Barnard-Webster, and Woodrow67bce93ef381 similarly define systems thinking as ‘a way of understanding the world (or any particular problem) as a series of complex interactions among multiple factors that, together, form a constantly shifting whole’. The process consists of identifying the factors that generate and sustain corruption and organising them into ‘causal loops’ that depict how they interact. The resulting ‘systems map’ can help practitioners identify the parts of the system that are susceptible to reform, visualise how corruption in a given sector or institution is embedded in the wider socio-political economy, test theories of change, and identify likely resistance to reform.6a91a1549971

Regardless of the methods and tools used, it is essential that the identification process be locally owned and led.c5d74fa95c56 It may also be necessary to explore both domestic and international drivers of corruption, as well as the role of external actors. For information on how to identify and analyse international drivers of corruption, see OECD.80dda74a65b6

5.2. Planning and implementation

Researchers’ and practitioners’ recommendations on how to plan and implement anti-corruption interventions are increasingly converging around several key tenets. Many of these coincide with the basic principles of anti-corruption approachessummarised in section 4.

Interventions should be locally defined and anchored. The process of identifying, designing, implementing, and adjusting anti-corruption interventions should be inclusive and locally grounded. To the extent possible, local actors should take the lead, and the backing of powerful stakeholders should be secured.e3152853996b As Johnston0ebeb427da96 puts it, ‘Even the best anti-corruption ideas need strong political and social foundations – the support of people and groups with lasting reasons, and the ability, to defend themselves politically against abuses by others.’ Interventions should also be aligned with existing efforts and strategies.

The importance of local ownership and endogenous demand is confirmed by Borges et al.14e2e6a3ee1a It is also recognised across the development community. Indeed, three of the six principles identified by the Doing Development Differently (DDD) Manifesto emphasise the need for interventions that solve locally defined problems, are locally owned, and work through local conveners (see Box 8).

While local ownership is vital, it is important to recognise that not all stakeholders will be genuinely interested in controlling corruption. Indeed, one of the lessons highlighted by Mason9e96671bb734 is that ‘the governments we deal with are part of the problem, and need to be seen as such. Donors need to stop deluding themselves that their “partners” share an equal ambition to tackle corruption.’ When this is the case, the choice of intervention and stakeholder engagement should reflect this reality.

Interventions should be designed based on the identification stage. Even when a corruption ‘symptom’ is the same across contexts, the factors that cause and sustain corruption, as well as the challenges and opportunities for anti-corruption efforts, will vary across societies and institutions. One-size-fits-all solutions should therefore be avoided.7324f0e05c22 As Johnston68a52ce605c9 notes, ‘What might seem to be a good reform idea in country A may well be impossible in B, irrelevant in C, and downright harmful in D.’

Interventions should draw on the existing knowledge base. The PDIA approach identifies four types of knowledge: existing practice, latent practice, positive deviance, and external best practice. While it is essential that programming be driven by the local context and local actors, there is still value in making use of the expanding literature on how to reduce corruption, as well as the literature on whether, under what circumstances, in which contexts, and in which combinations existing anti-corruption tools can be effective.

Interventions should, in combination with other initiatives, be comprehensive enough to bring about sustainable change. Practitioners need to design a complexity-responsive approach that recognises the entrenched, interconnected, and, in some cases, systemic nature of corruption. This will require a multi-pronged intervention, as no tool on its own is sufficient to significantly shift an equilibrium. It may also require venturing beyond the standard toolkit, as these tools are likely insufficient in cases of systemic corruption.9a530087891d

Sustainable change requires identifying, working with, and supporting agents of change, as well as creating trust and building coalitions.8f9f1856c2e2 Interventions should ideally support broader reform efforts and coordinate with other actors and initiatives – both domestically and, potentially, abroad, including in donor countries.e55aa072d928 Even when targeting a discrete element of the system, it should be clear to those designing and implementing the anti-corruption intervention how the programme contributes to systems change.

The DfID-funded programme Strengthening Action Against Corruption in Ghana is an example of a multi-dimensional, multi-institutional, adaptive anti-corruption initiative. Although the programme focuses on just one driver of corruption, namely principal-agent problems, it targets a range of institutions in order to strengthen the detection function, the investigation function, and the prosecution and adjudication function. The programme also works with both state and non-state actors and is implemented using a problem-driven and adaptive approach. It is complementary to additional anti-corruption programmes in Ghana, such as Strengthening Transparency Accountability and Responsiveness in Ghana, Phase II, funded by DfID, DANIDA, and the European Union, which works to further develop a well-informed and assertive civil society.

Interventions should address the root causes of corruption and employ a long-term perspective. When corruption is embedded or systemic,traditional anti-corruption tools will be insufficient to bring about a sustainable reduction in corruption. As Jackson, Tobin, and Eggertb5bf169bbaac point out, ‘Tackling systemic corruption requires alternative approaches; these need to go beyond the sorts of standard interventions that target more isolated forms of wrongdoing. Systemic corruption can only be curbed effectively by seeking to challenge, counterbalance, or provide alternatives to the underlying system, rather than by trying to ‘catch’ individual acts of corruption.’ This requires interventions that contribute to changing incentives, attitudes, expectations, and social norms.2b90a209899e When corruption serves a function, interventions should also seek to provide alternative solutions – ideally by addressing the underlying cause itself.ddf68e5ff03d

While there is still insufficient evidence on how best to design such programmes, some information is available, including Hoffmann and Patel’s5b870574eee4 paper on collective action and corruption in Nigeria; Jackson, Tobin, and Eggert’s53a8e3863258 paper on countering ‘wasta’ in Jordan; and chapter 9 of Fisman and Golden’s 2017 book Corruption: What everyone needs to know.

While broad-based action and an appreciation of the complexity of corruption are important, interventions also need to be targeted, sequenced, and realistic. The pace, scale, and level of actions should be determined by the opportunities and resources available. At the same time, there is a minimum threshold that should be adhered to. Interventions that are too limited or focused purely on the technical side are unlikely to contribute to a shift in a corruption equilibrium and may do more harm than good – a risk explored below.

With regard to what to prioritise, there are two main considerations: impact and feasibility. With respect to impact, one should ask: Which corrupt practices are causing the most harm? Which corrupt practices are particularly harmful to poor and marginalised households? Which interventions can best complement existing efforts? And what is the most urgent need?

When working to address systemic corruption, there are three levels of potential impact to consider: (a) reducing or eliminating a specific corrupt practice; (b) weakening the larger system of corruption (to avoid an outcome where the system simply adapts, with a new form of corruption emerging to take the place of the old one); and (c) contributing to deeper systems of accountability in a society.27dd32cad069

It is not always possible to implement the most impactful intervention. Sometimes, second-best solutions – less ambitious but more feasible measures – are in fact the best choice.24f05c48ed43 Two aspects determine an intervention’s feasibility. First, what is feasible given the political settlement and the wider context? Potential questions to consider are as follows: Is there an enabling environment and/or window of opportunity for reform? Is there sufficient demand for reform (i.e., pressure from below, above, or the side)? Which actors are for or against reform and why? Do the informal norms and structures provide those who are in favour of, or amenable to, reform with the agency to act? How can we work with the political grain? What high-impact corrupt practices can be addressed without threatening the prevailing political settlement? Under what circumstances have past interventions succeeded? Can incentives be altered, or the policy arena reshaped? And where there is not an enabling environment, what actions can be taken to enhance the feasibility of broader and more impactful interventions in the future?

This final question stems from a theory of change proposed by Khan, Andreoni, and Roy3f64368e9d3e for how to approach anti-corruption when the context does not yet allow for broad-based reform:

• IF anti-corruption strategies can sequentially attack corruption in sectors and activities where anti-corruption is both feasible and has a high impact,
• THEN, corruption levels will decline at an accelerating pace,
• BECAUSE these targeted strategies will enhance developmental outcomes, helping to create a more broad-based economy with a growing number of powerful organisations that will want rule enforcement in their own interest. This will make possible successively more ambitious anti-corruption strategies. Eventually, strategies targeting higher-level institutional characteristics like the enforcement of a rule of law or society-level transparency and accountability become more likely.

The second feasibility dimension has to do with donor and implementing agencies themselves. As discussed in sections 3.2.4 and 4.9, donor governments, aid agencies, and implementing partners have an incentive to maintain good relations with aid-receiving governments. They also face a number of internal constraints, including strategies to prioritise certain sectors and sub-sectors, short funding cycles, and the pressure to spend funds and document short-term results.

It is also important to consider the mandate and expertise of implementing agencies, as well as the resources available. To successfully implement anti-corruption interventions, practitioners need to have both the technical know-how and ‘soft skills’ such as an ability to influence people, change mindsets, and form strong coalitions. The available budget and manpower also need to be considered, along with the commitment level of actors who are expected to champion and defend reform efforts. Continued authority and acceptance cannot be assumed.00eb18d3b3f0

For further insights on politically smart programming and the impact/feasibility trade-off, see Uberti,08403ba043ca Khan, Andreoni, and Roy,e6fc79ba802e Levy,d8e769e9ef15 Kirya,6c2a91e34258 World Bank,97e4c1bc27fd and Step 3 of Curbing Corruption’s guidance for strategy development.

Interventions should have clear objectives and an underlying theory of change that anticipates backlashes and other unintended consequences. A well-defined objective and strong theory of change (ToC) can increase the likelihood of anti-corruption interventions having an impact.f04cf6b671c6 ‘Constructing a ToC enables government and donor staff to identify the logic underpinning their programmes and clarify how interventions are expected to lead to the intended results’.6d8cb50a1ed5 For assistance in developing and clarifying the underlying logic of an intervention, see ‘How change happens in anti-corruption: A map of policy perspectives’cc361cef2a82 and ‘Theories of change in anti-corruption work: A tool for programme design and evaluation’.36e6802c50a9

As implied by the discussion above, theories of change need to consider power dynamics and the specific context in which a measure is implemented. Interventions also need to be designed based on an understanding that corruption is complex, caused by interconnected drivers and enablers, resistant to change, and adaptive. Potential questions to consider include: ‘How might anti-corruption efforts unleash dynamics that could undermine the sustainability of reforms in the long run? What are the broader (unintended) effects anti-corruption efforts might have in other areas – such as security of citizens, recruitment of judges, social cohesion, etc.?’.41a778daeb8a

With a realistic and well-thought-out theory of change, practitioners can put in place measures to minimise the likelihood of resistance, co-option, and negative spillover effects. These measures will be intervention-specific but may include tactics such as securing the backing of powerful stakeholders, putting in place safety nets and support systems for change agents, garnering popular support, forming strong coalitions, adding project components, and addressing the root causes of corruption. A strong theory of change will allow for the early identification of ‘no-go areas’ and a reflection on whether an intervention risks doing more harm than good. It will also support the development of clear and feasible goals by recognising, for example, that corruption eradication is not a realistic objective.

A smartcard programme for liquid propane gas subsidies in India provides a classic example of the resistance that anti-corruption interventions should anticipate. The introduction of smartcards to the scheme led to a significant reduction in the administrative cost of providing subsidies as well as the leakage of funds. Yet the use of smartcards was dropped abruptly in the run-up to elections. This unexpected change is presumed to be the result of lobbying by black marketeers, whose profits were undermined by the intervention’s success.54600d3f4dc0 For details on the likelihood of backlash and unintended consequences, see sections 2.4 and 3.2.1.

Interventions need a learning-focused monitoring and evaluation (M&E) plan that balances the need for clear objectives with the need for flexibility and, at times, discretion. A well-considered plan is essential. It is also important to have a quality monitoring and evaluation system that contributes to the needed evidence base on anti-corruption.223808bea61f However, the M&E plan should also facilitate regular learning and adjustment.b9eb9e413899 Causal assumptions and context assessments need to be reviewed and updated throughout the implementation phase. Because change is not linear, the project design tools need to allow practitioners to experiment and change course.ab13913e5910 The importance of flexibility is increasingly recognised both within and outside the anti-corruption community, with a call for greater use of approaches and tools such as Doing Development Differently, thinking and working politically, adaptive management, and Problem-Driven Iterative Adaptation.

There is a wealth of material on politically smart and adaptive development assistance available. Vähämäki and Verger,357c79dde0f6 for example, provide a useful comparison of classic results-based management with alternative management approaches. Derbyshire and Donovana97f65daa54f highlight the lessons learned from two development projects that used adaptive approaches. In addition, Jackson and Dolve87bbf51e1af2 have produced two blog posts on adaptive approaches to anti-corruption.

Depending on the context, there may be a need to be discrete about an intervention’s objectives and impact. See section 3.2.2 for details.

The intervention should do no harm. In the section on theory of change, we explored the possibility of unintended consequence and backlash within programmes. It is also important to ensure that anti-corruption interventions do not negatively affect the wider economic, political, and social dynamics.e34275fce2aa

‘Do no harm’ can be defined as ‘avoiding premature or poorly-thought-out reforms that can do more harm than good – notably, steps that overwhelm a society’s capacity to absorb aid and put it to effective use, and that risk pushing fragile situations and societies into particular kinds of corruption that are severely disruptive’.b22c2821059e As Johnston and Johnsøna0f1d9fd0826 explain:

‘Attempts at reform may shift a society’s politics, alter relationships between leaders and followers, invite countermoves from those with a stake in the status quo, and introduce new uncertainties. Anticorruption initiatives can be co-opted or captured by venal and repressive regimes to distract the international community from their abuses or to serve as a pretext for locking up critics and leaders of opposition groups. Often, corruption is not reduced, merely displaced. Seeing grand proclamations but few results, citizens may come to distrust the government, reform leaders, and each other. Serious collective action problems may result if disillusionment replaces the initial enthusiasm for reform.’

The ‘do no harm’ principle is particularly important in fragile contexts.ca96e0c8bdac

6. Conclusion

Corruption is complex and resilient. Even incremental improvements are difficult to bring about, much less sustain and build upon. This paper sets out to synthesise the latest thinking, and these lessons and insights can be summarised in the following checklist for anti-corruption interventions (Box 9). Some are relatively straightforward, while others may require a change in development practice.