Consider the following hypothetical situation. A new programme providing cash assistance to the poorest households in a society, predominantly in rural areas, is rolled out. A signature initiative of the country’s president, this welfare programme is administered by officials at municipal offices, where eligible citizens line up each month to collect the cash assistance. Local media report recipients queuing with bags of wheat, vegetables, chickens, and other staple foods from the recent harvest. Further investigation reveals that beneficiaries have been told that to receive the assistance to which they are entitled, they should “show some gratitude” in return. Based on reports indicating that public officials indeed demand such “gifts” as a condition of providing the cash assistance, journalists have started to label the administration of the system as “extortive.” The practice contradicts the programme’s intended purpose of providing a basic safety net – indeed, extracting precious foodstuffs would seem to exacerbate the poverty of the beneficiaries. The practice has become so widespread that curbing the abuse is no longer simply a matter of disciplining a few deviant officials. To make matters worse, an external audit reveals the exploitation of financial transfer processes within the programme, with municipal officials skimming cash intended for the beneficiaries.

Shocked by these reports of extortion and embezzlement, the government is pressed into action. Deeming the corrupt behaviour a result of deficiencies in the “integrity framework,” the government focuses on strengthening accountability procedures and monitoring processes. As a response to the extortion, the government decides to issue clear guidelines clarifying that the law prohibits these kinds of conditional exchanges. A code of conduct and integrity training are introduced to ensure that officials abide by basic principles of transparency, accountability, and integrity. Meanwhile, to stamp out embezzlement, the government calls on senior administrators to clean up the programme; simultaneously, it increases wages for all municipal administrators to “change the calculation” against corruption. Despite these measures, one year later the beneficiaries are still turning up with basic foodstuffs, and an external audit finds the same amount of leakage in the programme’s finances.

Although hypothetical, the situation resonates with many real-world cases where corrupt practices, occurring either between a citizen and an official or between officials, afflict administrative processes. How to explain the persistence of such corrupt behaviour and the limited ability of standard anti-corruption measures, such as a code of conduct, internal discipline, and higher wages, to constrain it?

We suggest that focusing on official procedures and sanctions does not capture everything that happens in such a scenario. Beyond the formal rules, other strong reference points can help explain why people might agree to exchange their produce for cash assistance that they are entitled to receive for free, and why public officials might engage in embezzlement schemes. These reference points can be broadly thought of as social norms, “shared understandings about actions that are obligatory, permitted, or forbidden within a society” (Ostrom 2000, pp. 143–44). Such norms provide the unwritten rules of behaviour. Especially when formal rules such as laws fail to regulate conduct, as is often the case in countries riddled with corruption, social norms structure many social interactions by dictating the rules of the game. And there may be social sanctions for violating these norms.

The importance of social norms in sustaining corrupt practices is increasingly recognised in the literature.5d5ee3a3fa74 Many of these helpful contributions are based on empirical cases that illustrate how social norms may intersect with corruption. To add to that body of knowledge, this U4 Issue takes a step forward to think about how the design of anti-corruption and integrity-building policies and interventions can incorporate and benefit from a social norms perspective. While most policy thinking on social norms and corruption draws from a sociological tradition, we make use of a social-psychological perspective to help explain where norms come from and how they might change.

We argue that the threat of social sanctions for social norm violations creates multiple pressures that reinforce and lock in certain behaviours that sustain corrupt practices. The purpose of a social norms approach to policy design, then, is to relieve these social normative pressures so that behaviour can change. And while much of the current discourse speaks of the “norm of corruption,” we recognise that there is not just one corruption norm but rather various social norms that exert diverse influences on corruption. One reason is that people simultaneously belong to multiple social networks in which different, and at times contradictory, norms prevail. We introduce a framework that traces the four most relevant sources of social normative pressures that sustain corruption: sociability, kinship, horizontal, and vertical pressures.

This paper addresses three key questions:

• What are social norms, and why do social normative pressures matter in the context of corruption? (Part 1)
• What methods can practitioners use to diagnose the social normative pressures in a given context? (Part 2)
• What strategies can be deployed to relieve these pressures so that behaviour in regard to corruption can change? (Part 3)

Part 1: Understanding social norms and pressures that play a role in corrupt practices

The first step in using a social norms perspective to understand collective behaviour is to differentiate between individual attitudes and social norms. Attitudes describe personal evaluations of a given behaviour (Fishbein 1967). In the example above, a particular citizen who queues to receive the cash assistance might like or dislike the fact that one has to bring foodstuffs to receive the cash. However, these individual opinions do not fully explain the emergence and persistence of socially embedded, collective behaviour patterns. To understand why many corrupt practices persist we must look beyond individual attitudes and examine shared perceptions, attitudes, and expectations—in other words, social norms.

Across academic disciplines, studies have identified two main types of social norms. First, there are norms based on the perceived frequency of a given behaviour. People engage in a certain practice because they believe (correctly or incorrectly) that it is common: that it is what other people in their community, organisation, or network do. This is called a descriptive norm (Bicchieri 2016). When it comes to bribery, descriptive norms are captured in the explanation “I pay bribes because everybody does” (Köbis et al. 2015). The second aspect of social norms refers to the perceived acceptability of a given behaviour: whether it is considered right or wrong, a socially appropriate course of action or not (Bicchieri and Mercier 2014). This might be captured in a statement like “Giving gifts to officials in exchange for services isn’t wrong because you are showing your gratitude for their help.” This is called an injunctive norm.

Perceived social norms are sometimes aligned with personal attitudes, as when an individual’s opinion (“I disapprove of bribing a teacher”) is in tune with the broader community norm (“Bribery in the education system is socially frowned upon”). But there can also be misalignment. Just because a behaviour occurs frequently does not mean that the majority of people approve of it (Cislaghi and Heise 2018). Many corrupt practices are considered objectionable by large parts of the society. More specifically, two types of discordance between personal attitudes and collective perceptions exist: false consensus and pluralistic ignorance.

False consensus refers to an overestimation of one’s own opinions, beliefs, preferences, values, and habits as normal and representative of what others think and do. For example: “I personally believe that giving gifts to public officials is acceptable, and I think that everybody else does too.” Conversely, pluralistic ignorance describes a situation in which a person privately rejects a norm but incorrectly assumes that most others accept it. The person may therefore go along with the practice, complying with a norm even though she privately disagrees with it. For example: “I bring gifts to public officials even though I dislike it. Because everyone else is doing it, I think they probably consider it to be acceptable.”

Sources of normative pressures that can influence corrupt practices

To better analyse the link between social norms and corruption, it is important to understand that there is not just one source of normative pressure, but several. Social norms theory highlights that a person typically belongs to multiple social networks, some considered essential to a person’s identity, others more peripheral. These different networks follow different, and at times even opposing, norms. People particularly compare themselves to members of their reference group. They want to understand and follow the norms of the group(s) that they closely identify with and care about. A person’s reference group might be people in his immediate physical proximity, but it can also be a geographically scattered group, e.g., diaspora members, who may adhere to social norms of their country of origin. It is the norms of their reference group that people are especially motivated to understand and follow. Hence, not all groups are equally influential in shaping social pressures.

A person therefore does not face a single source of normative pressure but a multiplicity of normative forces. Moral psychology reveals that people often adhere to different (moral) standards depending on the salient network they are operating in (Köbis et al. 2016). Through so-called role distance (Goffman 1959), people can take on different roles and engage in the respective behaviours. In extreme cases this can lead to apparent contradictions: a person can belong to an organised crime network and be a devoted churchgoer at the same time, engaging in behaviour in one realm that he condemns in the other. It helps to bear in mind this ‘compartmentalizing’ ability when trying to understand the multiple accountabilities to which public officials respond (De Herdt and Olivier de Sardan 2015).

We can already see that speaking of a single corruption norm that needs to be tackled is insufficient. People may be subject to multiple normative pressures depending on whom they meet, whom they compare themselves to, and to whom they are accountable. To provide an analytical lens for the complex social normative forces of corruption, we propose a framework that looks at the main sources or types of social normative pressures. It is based on an interdisciplinary literature review with an emphasis on sociological and social psychological research on social norms, corruption, and development challenges.

Figure 1. Four sources of normative pressures

We propose a framework that distinguishes between sources of pressures that stem from society, namely sociability and kinship pressures, and sources of pressures that emerge vertically or horizontally within organisations. Though in the real world some overlap between these sources certainly exists, distinguishing between these pressures in the abstract helps to create a heuristic framework for thinking through the varied pressures that exist.

To outline these different sources of social normative pressures, let us revisit the opening example, which we use to illustrate normative pressures originating both outside and inside the public administration. Why do citizens entitled to cash assistance with no quid pro quo keep turning up with gifts for officials, despite laws expressly prohibiting it? Why have they not demanded an end to these conditional exchanges? To understand such behaviour, we have to look first at social pressures rooted outside of the municipal administration.

Sociability pressures:“I have to return the favour”

One possible reason that this transaction might have emerged is that a general unwritten rule may specify that when one receives benefits from a person in authority, it is “sociable” to offer something in return. This social norm taps into broader notions of reciprocity within one’s immediate society, that is, within the community of people around the individual within which interactions occur (Torsello and Venard 2016). Such reciprocal gift-giving rules and norms exist in societies and cultures around the world (Fiske 1992). Practices emanating from reciprocal norms exist not just to serve a function or a purpose, but also represent something substantially “social” beyond their instrumental use (Graycar and Jancsics 2017).

Such a sociability norm may exert a persistent pressure by becoming an injunctive norm, specifying “the right thing to do.” One reason why reciprocity norms are ubiquitous and have an “ought to” aspect lies in the strong negative responses to reciprocity violations. Extensive behavioural research shows that it is widely considered unacceptable not to reciprocate favours (Fehr and Fischbacher 2004), resulting in emotional responses such as anger and resentment. In his seminal book Bribes, John Noonan engages in a historical analysis of the origins of bribery and traces related practices back to ancient times. Based on early records dating back as far as 3000 BC, he specifies different forms of punishment for non-reciprocation of favours (Noonan 1987).

Given its deep roots in human psychology, the concept of reciprocity plays a key role in many corrupt practices. The social significance of the norm creates pressures to conform, and violating this norm may result in a guilty conscience, an undermining of self-identity, a loss of face, harsh criticism, even censure or punishment. In our example, sociability pressures do not lead to the corrupt practice directly, but they make the extortion possible. With exchanges generally conditioned by the reciprocity norm, ambiguities arise over whether the cash assistance is a right or a favour, and public officials can exploit these uncertainties to extract “gifts.” Without the reciprocity norm in place, there would be much less social pressure to conform to the extortion demanded by public officials: in short, it would be much easier for people to say no.

Kinship pressures: “Family first”

In addition, both the embezzlement and the extortive practice in our hypothetical example might respond to pressure stemming from a sense of obligation to kin. The social norm of providing for your kin (from immediate family to extended family to clan or tribe) can be strong, trumping norms of integrity, especially when public salaries are so low that officials are considered to need corruption in order to support family members who depend on them (Bauhr and Nasiritousi 2011). In our example, public officials who engage in embezzlement may do so in part because of family pressures they are under. Interestingly, an analysis of the World Values Survey by Alesina and Giuliano (2010) found that, in general, the greater the importance of family ties in a society, the higher the level of corruption. Providing for one’s family through illicit means can earn a public official status and respect, a positive reward for adhering to kinship pressures (Baez-Camargo et al. 2017a).

This kinship norm (also called in-group favouritism) stems from the fact that humans as a species have spent most of the ancestral past in small groups and tribes. Evolution has favoured norms that ensure cooperation within these groups (Trivers 1971). Seminal experiments in social psychology show that people today still readily make distinctions between in-group and out-group members (Diehl 1990; Sherif 1936), and cooperate—some say instinctively—with members of their group (Greene 2014; Rand 2016).

This norm of kinship and in-group favouritism frequently has an injunctive element attached to it. That is, around the world it is considered morally right to be loyal towards one’s in-group and kin (Dungan, Waytz, and Young 2014). For many officials this norm of kinship favouritism and loyalty might trump any norms of impartiality and integrity (see also Dungan, Waytz, and Young 2014; Köbis et al. 2016). In fact, not providing for kin might be seen as a loyalty violation. These kinship pressures are particularly strong when the distinction between the public and private is only loosely configured. Summed up in a common aphorism used in some pre-1989 Soviet countries: “Those who do not steal from the state steal from their families” (Misangyi, Weaver, and Elms 2008).

In the same extortion scenario, why doesn’t the public official just say no to the “gift” that the beneficiaries are providing? To understand the additional social norms motivating bribe takers, we have to look at pressures within the public administration.

Horizontal pressures: “My colleagues are doing it too”

One important influence could come from within the office, where colleagues may deem it normal to demand “gifts” from beneficiaries, creating within-institution peer pressures. The emergence of social norms within public organisations is well documented. Anders, for example, describes how beneath the layer of official rules in the Malawi civil service lies a complex web of interpersonal relationships amounting to a “parallel structure…with its own rules in regard to corruption” (2008, p. 15). It is not easy to escape the strength of this “unofficial code of conduct,” as resistance or rebellion can lead to social isolation, diminished career opportunities, and restricted access to attractive posts and workshops (Anders 2008).

Extensive research in organisational psychology shows that within organisations, horizontal pressures, especially peer pressures, can lead to the normalisation of corruption (Ashforth and Anand 2003). Seminal conformity experiments in social psychology by Solomon Asch (1951) have demonstrated that people follow social cues, even if they are clearly misleading. Since then, ample studies have provided additional evidence that the (observable) behaviour of others can sway people who consider themselves as moral and ethical to commit atrocities that they never thought possible, such as mass killings (Welzer and Christ 2005).

Such peer pressures can have a particularly strong influence within well-established peer groups. Over time, local social norms emerge, and these local norms can dictate corrupt practices even though many individuals within the group may personally perceive such practices to be unethical and wrong. These group members will often adhere to these local social norms rather than risk peer punishment for nonconformity. Taken together, horizontal pressures – based on injunctive and/or descriptive norms – can exert a strong toll on public officials to engage in corrupt practices.

Vertical pressures: “I am forced from above”

Whereas horizontal pressures come from one’s peers, vertical pressures emanate from people at higher levels of an organisational hierarchy. To illustrate, let’s turn to the second form of corruption found in the opening example: embezzlement. Why do public officials get involved in skimming schemes? One reason – though certainly not the only reason – may relate to pressures from higher-ups. Subordinates have certain formal obligations to follow orders from their superiors. When a superior requires conduct that is illegal or unethical, the subordinate’s obligation to comply may enter a grey area legally, but pressures to comply will still be strong. Saying no may well result in loss of a job or promotion. This is true even when – as is likely in the case of corruption – superior pressure takes the form of an implicit expectation or demand rather than a formal order.

Extensive social psychological research corroborates the notion that lower-ranking individuals typically comply with orders or pressure from above. Stanley Milgram’s experiments in the 1960s are considered a cornerstone of the rich literature on obedience to authority. In one of the most famous studies in psychology, participants were instructed to administer electric shocks to another person, the “learner,” whenever the learner failed to fulfil a task. The voltage of the shocks gradually increased and the learner became increasingly agitated, leading to loud screams of pain. However, in the initial study 65 percent of the participants carried on after being instructed to do so by the experimenter, administering deadly shocks of up to 450 volts. No one died in the experiment because, unbeknownst to the participant, the learner was an actor and the shocks were fake (Milgram 1963). The study has recently received criticism (Griggs and Whitehead 2015), with new attempts to replicate its findings calling into question the blind obedience that the original results suggested. However, the main finding – that people frequently obey orders from authority figures even when it means engaging in behaviour that they deem unacceptable – remains broadly accepted.

Why would superiors require (explicitly or implicitly) lower-ranking officials to engage in corruption? One reason may be that they are getting a cut of the proceeds. Wade (1982) describes how canal irrigation engineers in South India raised vast amounts of illicit revenue from the distribution of water and contracts. The engineers faced extensive pressures to redistribute the revenues “up the chain” to superior officers and politicians. In another example, Smith’s (2003) study of a donor-led family planning programme in southeastern Nigeria demonstrates how local staff appointed to the project were expected to appropriate and channel some resources upwards to their “patrons” in the ministry. Many public officials find themselves enmeshed in vertical social networks that function to ensure the upward movement of public goods (Scott 1972).

The varied influence of social pressures

These normative pressures differ in strength. As illustrated by Cislaghi and Heise (2018), the spectrum starts with the weakest normative pressures, which merely create the impression that a practice is possible. For example, paying a bribe may appear as one possible way to obtain a driver’s license. Slightly more binding are normative forces that indicate that a given practice is tolerated. If others look the other way or even actively sustain the practice by condoning it, weak social norms in favour of the act exist. One step further along the spectrum are acts that are considered expected. Once something has become “the right thing to do,” and a substantial proportion of a group appears to regard the conduct as appropriate and expected, normative pressures to conform exist. It is important to note that a person’s perception of the group’s approving attitude does not have to be accurate in order for a social norm to produce normative pressures on that person. Finally, we get to the strongest end of the spectrum: obligatory behaviour. These are acts that a person is required to do, or believes that he/she is required to do.

Social norms do not have to directly dictate the corrupt practice to play an important role in encouraging or condoning it. Broader norms may serve as a background factor that influences the calculation as to whether to engage in corruption or not. Consider a social norm that can be described as “support the family above everyone else.” Imagine if you believed everyone in your reference group (e.g., your family network) deemed this norm both typical and desirable. In and of itself, this perception would not need to lead to corruption. But if you are a public official, it could lead to your family network exerting pressure on you to skim off money or provide jobs to relatives for the benefit of the extended family. Resisting this pressure may result in a social sanction – being shunned at family gatherings, for example. This social norm is not the only reference point; it may be balanced out by norms of public service or concerns around being caught. But it can nonetheless create pressures to engage in fraud, embezzlement, nepotism, extortion, and so on, even though the norm – in this case family loyalty and obligation – is far from being a corruption norm per se.

In line with the varying strength of social normative pressures, different sanctions follow norm disobedience. External sanctions are imposed by others to punish deviance from the norm. They can take different forms such as shunning, shaming, or social exclusion – or even direct violence. A crude example of external enforcement of social norms in the corruption context is the case of Frank Serpico, a police officer who entered the New York Police Department with idealistic views. It did not take long for him to be confronted with a corrupt reality, as bribe taking by police officers seemed to be the norm. His attempt to resist this norm led to harsh social sanctions and eventually to severe bodily harm. In addition to such visible sanctions, social sanctions can also happen out of sight of the target, like gossiping about a norm violator.

These social pressures certainly are not the only explanation for how citizens and public officials act and interact, but they belong to the constellation of factors that may determine how people behave. For example, people might adhere to a code of conduct outlining certain duties and responsibilities, which serves as an alternative reference point. In her research in East Africa, Baez-Camargo (2017, 19) found many individuals who felt burdened by the overlapping and often conflicting expectations and liabilities stemming from social norms as well as from their legal duties and responsibilities. Therefore, we emphasise the relative influence of social pressures over other drivers of behaviour.

How normative pressures can frustrate anti-corruption reforms

Just as these pressures explain why corruption persists, they also help explain why standard anti-corruption reforms may fail. In the introductory hypothetical example, we saw that the interventions did little to stop the corrupt behaviours from recurring. We would argue this is due to a shortcoming in the standard repertoire of anti-corruption actions: namely that they generally ignore the influence of social norms and pressures, an oversight that may explain why seemingly sensible solutions may often have a limited effect. Let us examine two popular interventions that illustrate this point.

Salary increases

One of the most commonly proposed policies to reduce corruption among public officials is salary increases (Fisman and Golden 2017). According to the fair-salary hypothesis, if public officials could earn enough to make ends meet and support their families, then they would have fewer incentives to ask for bribes (Becker and Stigler 1974; Van Rijckeghem and Weder 2001). This policy has had mixed results, in part due to the persistence of social norms. Higher wages have been tried in the tax administration in Uganda, for example, where norms of providing for the family are very strong (Baez-Camargo 2017a). The introduction of higher wages actually increased expectations about the extent of family support that public officials could provide, leading to a net loss for some officials once they had met social demands for redistribution. To make up for such a loss, public officials who have received salary raises may seek to extract even more bribes (Fjeldstad 2005).

Another reason for the failure of salary increases might lie in the persistence of horizontal pressures. If within an organisation, such as the police force, accepting bribes has become a widespread and maybe even acceptable practice, higher salaries do little to change these pressures. Indirect evidence for this perspective stems from a study conducted by Foltz and Opoku-Agyemang (2015), using direct observable data on bribes. The United States Agency for International Development (USAID) had been recording bribes paid by truck drivers in Ghana and Burkina Faso between 2006 and 2012. During that period, the salary of the Ghanaian road police force was doubled, which allowed the authors to test whether a salary increase indeed reduced the elicitation of bribes. The results are disappointing. If anything, increasing the salary led to an increased effort by police officers to collect bribes, and the size of bribes increased as well. One reason appears to lie in the lack of punishment for bribe taking, which was unchanged by the reform. Another, possibly more important, reason is that the social norms within the police force remained the same. Asking for bribes remained a common – and acceptable – practice. Although police salaries increased, the social pressures remained in place, and bribe taking continued.

Code of conduct

Introduction of a code of conduct is a typical reform within public administrations. Intended to provide clarity about expected behaviour, duties, and responsibilities, such codes seek to provide a normative reference point to which employees should adhere. But as we have seen, there may be horizontal pressures on employees not to abide by the official code. If the parallel “social code” on the office floor tolerates corrupt actions and is tacitly upheld by a majority of the group, then a formal pledge to comply with the official code will be unlikely to change behaviour. The introduction of the official code merely stipulates required behaviour without relieving the normative pressures embodied in alternative, unwritten, unofficial social codes.

Official codes of conduct also need enforcement from above, that is, from figures at higher levels of an organisation. But as described earlier, these higher-ups may be receiving a share of the proceeds of corruption schemes and therefore have little interest in enforcing rules about integrity. Indeed, the bosses may be the ones issuing directives to subordinates to “skim” or otherwise engage in corruption, in which case the attempt to discipline from above will have limited success. Vertical relations within an organisation are often based on feelings of dependency and indebtedness. For instance, in Malawi, junior civil servants draw a sharp line between themselves and “the bosses,” believing that as lower-ranking workers they depend on protection by their superior officers (Anders 2008). Going against vertical pressures may result in sanctions – loss of job, position, or earnings – and thus may constrain a change to more honest behaviour.

The point is not that these two policies – salary increases and codes of conduct – cannot work or are misguided per se, but rather that they are unlikely to be sufficient in and of themselves, given the persistence of social norms. Understanding these norms and then devising complementary strategies to address them could create space for movement towards new, more honest behaviours. In any given context, understanding the type and strength of social pressures is essential to figuring out which interventions can help change, manage, or circumvent these pressures. Part 2 will help practitioners employ this approach.

Part 2: How to map and diagnose normative pressures

Imagine you are asked to advise on what can be done in the situation of the corrupt municipality described at the beginning. Designing successful interventions requires a thorough understanding of the social forces that perpetuate the corrupt practices. In this section we combine the insights from various academic disciplines with the current trends in social norms programming in other domains to provide an overview of the available methodologies (see for example, Stefanik and Hwang 2017). First, we outline a schematic stepwise process of designing an anti-corruption policy. While we are aware that policy implementation often does not neatly follow these steps (see, for example, Mosse 2011), we find that this schematic framework helps illustrate the key components of policy design. Next, we provide an overview of the methods available to execute each step in the process. Hence this section provides guidance on how to map and diagnose social pressures, as well as methods to generate the necessary data.

Deploying these tools may at first glance seem daunting, especially at the planning stage of a project. However, do not despair: it may helpful to consider what information might already be available in your networks or among staff, and to consider whether there are local universities, institutes, or consultants that can be commissioned to generate information using the methods suggested below.

Stepwise process of policy design for an anti-corruption intervention

Step 1: Assessment of corruption scheme(s)

Before one can understand and mitigate the social pressures driving corruption, a first step is to specify the corrupt behaviour in question (see, for example, Heywood 2017). In other words, we need to ask, “What is going on?” As described in our opening example, it may well be that multiple schemes, at times interlinked with each other, are at play. Grasping the complete picture requires extensive research. A mapping of the corruption scheme(s) can help to visualise the interlinkages; Woodrow provides a useful overview on how to use such systems maps.

The following questions can be useful in the initial assessment of corruption schemes:

• What kind of corrupt action is occurring?
• Who is involved?
• Who instigates it?
• Who benefits from it?
• Who loses out?
• What transactional forces are at play, that is, who gets what, from whom, for what?
• How do actors communicate with one another?
• How is information distributed?
• How are the proceeds of the scheme(s) distributed?

Step 2: Formative research

A second step is to identify which (if any) social norms are contributing to the persistence of the corrupt scheme(s) identified (Stefanik and Hwang 2017). This requires research to identify important reference groups for key actors, the norms that prevail within these groups, the channels for social pressures to enforce these norms, and the anticipated sanctions for deviance. Such research can detect potential leads for the baseline assessment in step 3. To achieve this goal, the formative research phase mostly draws on three information sources: a literature review, (informal) interviews, and insights from existing surveys (see more details below). The following questions can serve as a guideline for data collection.

Leading questions for data collection:

1. Is there a social norm pertaining to the corruption scheme? If so, which practices does it support?
   1. Is a given corrupt practice perceived as common?
   2. Do people think that others engage in the practice?
   3. Do people approve of the practice?
   4. Do people think that others approve of the practice?
2. Which are the influential reference groups for each norm? What are the main sources of normative pressures?
   1. Do the reference groups include society, peers at work, superiors, kin?
   2. What practice is considered common in the reference group?
   3. What practice is considered acceptable in the reference group?
3. What kinds of social sanctions are anticipated in response to deviation from the norm?
   1. Who enforces norm compliance?
   2. How severe are the punishments?
4. Does norm deviation occur?
   1. Who are the people who deviate from the norm?
   2. Are there people who are exempt from the norm?
   3. Are there circumstances when it is more acceptable to deviate from the norm?

Step 3: Baseline

In the third step, practitioners seek to obtain more generalizable and robust insight into the corrupt practices and to identify entry points for interventions. Hence, this baseline assessment phase employs more structured methods. By following up on the leads identified in the formative research stage, it seeks to more closely examine the existence of the outlined corruption schemes, as well as the sources and strength of social norms pertaining to it. The goal is to provide a baseline measure of both the existing social norms as well as the respective behaviour (i.e., corruption). The choice of methods for the baseline measures should be in accordance with the choice of the endline measures. Besides the tools mentioned in the formative research stage, the most useful methods in this step are more formal interview techniques, focus groups, vignettes, social network analysis, and bespoke surveys.

Step 4: Monitoring

Upon implementation of an intervention, the next step is to monitor whether the intervention goes as planned. Are there signs of norm change? Do people respond to the newly introduced intervention in the way that was anticipated? Are there any signs of backlash or boomerang effects? These findings can be used to potentially make adjustments to the policy design, helping to increase the chances of a successful anti-corruption intervention. Methodologically, this step can draw on any or all of the outlined techniques, depending on the specific context.

Step 5: Evaluation

The final step is to evaluate the intervention. The assessment as to whether social norms around a certain corrupt practice and the practice itself have changed should draw on similar if not the same methodology as the baseline step. Such methodological consistency allows the most valid pre-treatment versus post-treatment comparisons. Below we provide more detail on the different methods available for each of these steps.

Methods

To investigate social norms of corruption, a practitioner can choose between qualitative and quantitative tools. While qualitative methods seek to gain deeper insights into particular cases, quantitative methods seek to draw comparative conclusions across multiple cases. As we will outline below, often the best choice is a combination of both types of tools. So-called mixed-methods approaches help reduce the risk of biases inherent in a single-method assessment. Moreover, certain tools can be used either quantitatively or qualitatively.

Literature review

Reviewing the relevant academic literature as well as local news sources on the subject marks a good starting point for any project whose purpose is to change collective behavioural patterns. The academic literature may at times include extensive and high-quality reports on corruption schemes, such as published papers and books on the local political and social context. Non-academic resources can include reports and commentary on the scheme in the local news media as well as in social media and other online sources. As an example, the website Ipaidabribe.com has compiled a large collection of bribery reports that might help to explain common features of corruption schemes that are occurring in various countries. In our hypothetical example, local media reports might already hint at the sociability norm that seems to spur the bringing of foodstuffs. A careful review of what has already been done ideally provides first answers to some of the most crucial questions about the dynamics of social norms pertaining to corruption.

Interviews

Four main types of interview techniques can be used, differing in their degree of formality. Informal interviews are casual conversations that can occur without being explicitly labelled as interviews, e.g. by chatting in a bar. Unstructured interviews resemble informal interviews in that they do not draw on a fixed set of questions, but they differ in being unambiguously labelled as interviews. Semi-structured interviews make use of predefined topics or questions while leaving room to explore leads that may emerge during the interaction. Finally, structured interviews closely follow a set of pre-specified questions and hence represent the most controlled and closed-ended type of interview.

The choice of interview technique depends both on the particular corruption scheme and on the amount of information already available. Both the content of the questions (open and explorative vs. closed and confirmative) and the interview technique determine how much additional information one can obtain. Conducting interviews on such delicate topics as the social norms of corruption requires confidentiality, so that trust between interviewer and interviewee can emerge. Ensuring privacy additionally helps to reduce social desirability concerns on the part of the interviewee (see Box 1). Interviewees should receive information on how and which data from the encounter will be stored, shared, and/or published, before giving informed consent.

Interviews have immense potential to provide insights into a particular corrupt practice, including insights that may be unobtainable through quantitative methods. In the example outlined above, asking citizens who have received the cash assistance whether and how they were urged to “show some gratitude” can help reveal who enforces norms, and how (see leading questions for data collection, questions 3a–3b). Interviews with public officials, especially in an informal setting, can help identify whether any horizontal or vertical pressures exist. For instance, officials can be asked how their colleagues would react if, all of a sudden, the corrupt salary “top-up” ceased to be distributed among them. Besides the actual answer (most likely heavily influenced by social desirability concerns), other cues might prove insightful: a surprised facial expression or laughter might suggest that this option was never considered. A long pause before giving an answer might suggest that the respondent has to carefully choose her words or has to think about what actually would happen. Body language can provide clues as well: a tensing of the posture could suggest that such peer norm violations would entail severe repercussions. Such qualitative cues often entail some ambiguity, so corroborating these indications by means of additional information sources is advisable. For more details on how to capture (social norms of) corruption using interview techniques, see Varraich (2017).

Focus groups

Focus groups are another qualitative research method that can expand the number of people who serve as information sources. This method involves organised discussion with a selected group of individuals to gain information about their views and experiences around a topic. It is particularly suited for obtaining several perspectives on the same topic. Focus groups draw out respondents’ attitudes, feelings, beliefs, experiences, and reactions in a way that may be less feasible using other methods such as observation, one-to-one interviewing, or questionnaire surveys (Gibbs 1997).

The literature provides extensive guidance on how to conduct effective focus groups. One could, for example, conduct a focus group on sociability pressures, with guided discussion of when giving a gift to a public official is considered right and acceptable. Such a discussion might reveal that most people don’t really want to give gifts to public officials but do so because of the perception that everyone else does it. This would provide evidence that the norm is more descriptive than injunctive, a finding that could provide scope for interventions around more far-reaching reforms.

Focus group research can generate a lot of data in a short amount of time. However, it faces the challenges of group dynamics, such as peer pressure, confrontations, dominant personalities, and social desirability bias. As with all methods, triangulation of the data generated by focus groups with other sources such as survey data or other interviews is advisable. Research in East Africa, for example, has effectively used two different focus groups, one for citizens and another for public officials, to understand the relationship between social norms and corruption as part of a research strategy that also included vignette-based surveys and participant observation. Like one-to-one interviews, focus groups provide qualitative insights that should be followed up by additional (quantitative) research to confirm the leads obtained.

Vignettes

Vignettes represent another complementary method that can be used to assess social norms of corruption. Typically vignettes consist of “short stories about hypothetical characters in specified circumstances, to whose situation the interviewee is invited to respond” (Finch 1987, p. 105). They provide an engaging and subtle instrument with which to tap into highly complex behavioural frameworks and address the situational elements of behavioural choices. As vignettes allow for different situations to be built into the research design, they can reveal people’s likely choices in the kinds of situations they actually face (Finch 1987). Vignettes encourage respondents to engage with sensitive topics. In contrast to talking about one’s own experience, vignettes allow for a more detached assessment of the social norms of corruption: respondents feel less threatened when commenting on a vignette story that involves hypothetical characters outside their own lives. Such story-based methods avoid the possible intrusiveness of face-to-face interviews, creating a comfortable distance between the researcher and respondent (Renold 2002).

Vignette studies offer a stepping-stone to a deeper inquiry into the logic and strength of social normative pressures (see Box 2). With respect to the opening example, vignette research could assess the potential enforcement of sociability norms pertaining to the extortion scheme. Asking citizens which group most strongly enforces the norm of bringing foodstuffs (see leading questions for data collection, questions 2a–2c),or what happens when someone violates the norm by not bringing it (questions 3a–3b),can uncover the mechanisms of social norm enforcement. This way, vignettes enable a subsequent assessment of the key components of social norms outlined above (see Jackson 2018). They can be used qualitatively, with open-ended questions about the scenarios that let respondents answer as they wish, or quantitatively, by standardizing the answer options on fixed numeric scales. Moreover, using incentives for accuracy can help researchers obtain more valid responses to these vignette scenarios (see Box 3).

Vignettes also allow an experimental investigation: manipulating parts of the vignette and testing differences in responses provides a look at causal links. If change in the manipulated variable leads to changes in our dependent variable of interest, then some evidence of causality exists. As an example, a vignette experiment could change the family member who enforces a kinship norm, and then assess whether the responses differ, in order to check whether it makes a difference who enforces the norm (questions 3a and 3b). Such information can help identify the most influential individuals in reference networks, setting the stage for potential interventions using trendsetters (see part 3).

Corruption games

Recently, corruption games, i.e., behavioural measures of corruption, have gained popularity as a means to measure (social norms of) corruption quantitatively both within and outside the lab (Abbink, Irlenbusch, and Renner 2002; Köbis et al. 2017; Serra and Wantchekon 2012). When carefully crafted to model realistic corruption schemes, such games can offer a controlled measure of people’s inclinations to engage in corruption. Like vignette experiments, corruption games allow researchers to manipulate individual variables in order to isolate causal variables. With this methodological control, corruption games have been used to test the workings of social norms on corrupt behaviour (see leading questions for data collection, questions 1a–1d) (Abbink and Serra 2012; Barr and Serra 2010; Köbis et al. 2015).

Corruption games can also serve as a tool to compare the effectiveness of interventions. Recent research has tested the effectiveness of social norms interventions in reducing corruption by combining corruption games with vignettes (Köbis, Soraperra, and Troost 2018). The team of researchers distributed posters containing descriptive norms messages (“less and less people bribe”) widely throughout a medium-sized town in South Africa. They also set up a mobile lab in the town’s centre. In this lab, they measured perceived social norms with incentivised vignettes (see Box 3) and used a behavioural measure of corruption played for real money, during baseline, monitoring, and endline periods. The combination of methods provided empirical evidence that the messages decreased the influence of perceived descriptive norms during the period in which the poster was up. Also, people were found to be less willing to engage in bribery (in the context of a bribery game) during that period.

Surveys

Surveys represent another quantitative method to assess social norms of corruption. A rich collection of large and open-access surveys on corruption already exists (for a comprehensive overview see Richards 2017). Although many surveys offer a rough first estimate of existing norms at best, others zoom in more closely on different practices and differentiate between different types of corruption. One example is Transparency International’s Bribe Payers Index, which asks whether the respondent (or anyone in her family) has paid a bribe in the last 12 months. Although mostly aggregated on the national level, these surveys can help answer the question of how frequent certain types of corruption are perceived to be. Related to our introductory example, existing survey results that show a high frequency of reported side payments to obtain public services, such as cash transfers, can point towards the existence of a descriptive norm.

It is much less frequent, however, for existing surveys to take a close look at social norms. Although they might give an indication of the perceived frequency of a corrupt practice (answering question 1a in leading questions for data collection), they rarely assess other crucial elements such as the respective reference group, injunctive norms, and beliefs about the expectations of others (leaving questions 1b–4c unanswered). To get an answer to these questions, one has to design and conduct a survey tailored to this purpose.

Polling a large number of people can happen via personal interviews, telephone interviews, or online surveys (each has pros and cons; see Bowling 2005). The choice of survey method depends on the scope, type of question, and financial means available. For instance, researchers may consider conducting a survey across several communities to compare the extent to which kinship pressures underlie embezzlement schemes like the one in the example. Ensuring that the schemes indeed are similar across different regions is an important precondition for comparable survey results (see Box 4). Another challenge has to do with social desirability bias (see Box 1). For instance, asking respondents directly whether their family members expect to receive a portion of the revenues obtained by embezzlement will likely yield inaccurate results. Creative measures to reduce social desirability response bias exist, such as masking respondents’ individual responses using random devices (see Box 2 for more details). When they manage to meet these challenges, surveys can offer a uniquely wide view, revealing information about social norms of corruption on a large scale.

Social network analysis

Social network analysis focuses on the investigation of social structures by using networks and graph theory. It is a tool that can be used qualitatively or quantitatively. It draws on extensive research in social psychology showing that humans constantly compare themselves to others. Trends currently emerging in social psychology employ new methodologies to assess the extent to which these comparisons shape a given behaviour. Here we introduce this methodology to investigate how social connections may foster or prevent corruption. In order to gain insights into this question, identification of the reference network marks a key step.

A reference network is the group of individuals to which a person feels most closely connected. It is this group that most strongly shapes the social norms influencing a person’s behaviour. Reference networks are commonly classified as tight networks when the contact is frequent and close. This typically increases the strength of social normative pressures. On the other hand, loose networks leave more freedom to the individual, and the normative pressures in such loose networks influence behaviour less strongly. Horizontal and kinship pressures usually stem from tight social networks, as the average individual is in close contact with both colleagues and family members. Indeed, colleagues and other peers at the workplace frequently represent a reference group that shapes the social norm pressures underlying corrupt practices.

Social network analysis then enables a closer look at these networks, e.g., by analysing the communication chains – who talks to whom about what, and how fast information spreads. The analysis provides several types of information. First, it helps estimate the tightness of the social network and thus provides a first clue to the strength of the social norms. The analysis can also reveal who is connected to whom, who spends time with whom, and who influences whom – an indication of crucial sources of pressure. Hence, such assessments help identify the reference groups that crucially shape the social norms perceived by a given individual. The results of social network analysis also help identify people with heightened influence when it comes to changing social norms – so-called trendsetters or norm entrepreneurs. These individuals can serve as entry points for social network–based interventions. Recent approaches that focus on these opinion leaders show success in several domains, such as changing the social norms of conflict (Paluck, Shepherd, and Aronow 2016).

Attempts to combat corruption by using social networks have gained popularity. An example is the mapping of a corruption scheme in the criminal justice system in Lubumbashi, Democratic Republic of Congo. By engaging citizens, the programme was able to plot the network ties that sustain corruption and identify entry points for change. In line with the reasoning put forth in this paper, the citizens pointed towards the alleviation of social normative pressures as a key path towards reducing corruption. This indicated that strengthening the ability of actors in the system to resist corruption and providing information to citizens to enable them to understand the corrupt system was of immense importance. Related to our hypothetical example, the first point ties in with the general attempt to relieve horizontal, vertical, and kinship pressures. The second point corresponds to the approach aimed at reducing information asymmetry between citizens and public officials, informing citizens about their right to obtain cash assistance for which they are eligible without offering gifts in return. Research by Betsy Levy Paluck provides a good example of how social network analysis can be used to assess and change social norms (Paluck 2009; Paluck, Shepherd, and Aronow 2016).

Choosing a method

The noted development economist Esther Duflo (2017) compares the work of economists who apply their theoretical insights to change behaviour in the field to the work of a plumber. The plumber has to decide which tools to employ during the different phases of a project, choosing from among the items in his or her toolbox. The same applies to the anti-corruption practitioner, who must select from among the qualitative and quantitative research methods available. It is at least as important to know which tool to use as to know how to use the tools correctly. If all you have is a hammer, every problem looks like a nail, but having a full toolbox enriched by the methods outlined above requires the practitioner to choose the right one. The corruption scheme at hand determines which diagnostic tool to use. Each method has its strength and weaknesses. Luckily, picking a tool is not necessarily an either-or decision. Mixed-methods approaches are becoming increasingly popular, so one might opt for a combination of tools.

Diagnosing the pressures stemming from social norms requires careful planning, including thoughtful choice of method(s), and has to overcome several methodology challenges (see Boxes 2, 4, 5, and 6). As outlined in detail here, the insights obtained into the corruption scheme and the contributing social norms can offer entry points for policy interventions – the subject of the next section.

Part 3: Anti-corruption strategies for relieving normative pressures

Realising a social norms approach to anti-corruption means developing sets of interventions that are distinct from the standard repertoire of interventions. Nevertheless, the approach is complementary: the purpose of social norms strategies is to relieve and shift social pressures so the other kinds of interventions – such as codes of conduct, salary increases, legal reform, enforcement, and civil society oversight – can be effective. Therefore, social norms strategies should always be part of a policy mix. The following list (adapted from Jackson and Salgado-Moreno 2016) provides an overview of the standard repertoire of strategies. We have appended social norms strategies at the end as a distinct tool that nonetheless forms an integral part of the whole.

Standard repertoire of strategies + social norms strategy:

• Rule-changing strategies focus on changing formal rules and institutions, such as specific anti-corruption laws and regulations, or on broader institutional changes, such as decentralisation or changes to electoral systems.
• Enforcement strategies focus on the specific political practices and organisations that enforce and monitor the anti-corruption elements of the law, such as audit institutions, the judiciary, and anti-corruption watchdogs.
• Transparency strategies focus on generating information that would otherwise be unavailable, using various mechanisms such as rules on public disclosure and right to information.
• Organisational and managerial strategies focus on the day-to-day practices of political actors. For example, they can aim to change individual behaviour through soft laws, such as a code of conduct, or to change economic incentives through a new wage structure.
• Social accountability strategies involve efforts by social actors – e.g., civil society organisations, political pressure groups, social movements, newspapers, or religious leaders – to hold government accountable through monitoring, disclosure, and/or advocacy.
• Social norms strategies focus on relieving and shifting some of the social pressures that sustain corruption. Such strategies make use of methods such as dialogue, signalling, value change, information provision, trendsetters, and other mechanisms discussed below.

Social norm strategies are oriented to how things are rather than how they ought to be. In this sense, we eschew reverse-engineering solutions: that is, we do not simply import generic anti-corruption or integrity institutions from other countries in the hope that they can instil new norms. For example, establishing a meritocratic hiring and promotions system could prevent some of the horizontal and vertical pressures that people are under. But if meritocracy could simply be instituted then there would be no need to relieve the pressures, because implicit in the establishment of meritocracy is the proper functioning of administrative rules, undisturbed by these pressures. So, we need to think about how to get to meritocracy by relieving pressures rather than simply suggesting meritocracy as a solution.

We present here not prescriptions but policy approaches, clarifying why each approach may address a particular pressure. These are intended as general guidelines for practitioners: the particular intervention will have to be developed according to the features of each case and the respective entry points. They are drawn from a review of the literature and from real-world interventions designed to address social norms, where there is some evidence of effectiveness. However, using a social norms approach means recognizing that no “one-size-fits-all” solution exists. Effectiveness in one location does not imply effectiveness elsewhere, and we reiterate the importance of conducting research and pilot-testing interventions separately for each case.Moreover, the existence of social pressures in a community is a sensitive matter, and there may be a number of risks in attempting to change norms. It is therefore important to assess and address these risks so that practitioners first “do no harm” (Alexander-Scott, Bell, and Holden 2016, p. 19).

After recognizing which normative forces exert pressure for a given corrupt practice, practitioners should tailor the intervention to that pressure. Therefore, we arrange interventions around the four principal types of pressure identified above: sociability pressures, kinship pressures, vertical pressures, and horizontal pressures. In practice, however, some of the interventions described below are applicable across types. For example, certain interventions aimed at sociability pressures would also be appropriate for kinship pressures, and vice versa. This should be judged on a case-by-case basis.

The sections below present sample strategies for addressing and reducing each of the four main pressures, with suggestions for one or more methods that can be used to support the strategy. Each section begins with a table that summarises the policy paths.

Sociability pressures

Strategy: Strengthen alternative norms around civic rights

One way to relieve sociability pressures is to strengthen an alternative reference point that challenges these expectations, around which new forms of behaviour can emerge. With regard to the introductory hypothetical example, this could mean establishing the receipt of cash assistance as a “civic right” (of those households that meet eligibility criteria) and not as a gift that needs to be reciprocated. Below we outline several potential strategies to achieve this goal.

Method: Transformative dialogue techniques. One way to socially strengthen a counter-norm is through a pedagogic technique that emphasises participation, dialogue, and problem solving to help participants imagine new social realities; this is sometimes referred to as collective deliberation (Bicchieri and Mercier 2014). Cisaghli (2018) studied this technique as applied to a “transformative human rights education” (THRED) programme led by an NGO called Tostan in a rural community in Senegal. One of the programme’s aims was to challenge existing gender norms in the community that often prevented women from participating in public life. Through sustained weekly interactions, the programme helped participants use a human rights curriculum to analyse their relationships. Not only did existing norms relax, but new norms around women and participation emerged. Whereas a public role for women was previously frowned upon, women involved in the programme began to access public space and demand participation in political decision-making processes (Cislaghi 2018, p. 264).

Though we cannot be sure that this pedagogic model would work well for other kinds of normative change, some of its elements could feed into modules on civic education for both citizens and public officials that aim to socially strengthen the norm that establishes the receipt of assistance as a civic right. Such an approach represents an alternative to standard awareness-raising or messaging campaigns in which information about corruption is “dropped from above.” Applied to the initial example, policy interventions aimed at using such a collective deliberation technique could engage citizens, public officials, and local authorities. Based on social network analysis the programme could identify individuals or groups who might influence the discourse about the sociability norm of gift giving and point towards alternative means of showing gratitude.

Key practical considerations. As illustrated in the example described by Cislaghi (2018), successful programmes typically are:

• Sustained over time. Such a model defies expectations that donors often have about short-term social change. In the Tostan programme, for example, two cohorts of 25 to 50 participants each – one for adults and one for adolescents – met three times a week for 12 months.
• Conciliatory rather than conflictual. The technique did not emphasise oppositional “say no” techniques, but helped participants reimagine existing relationships and power dynamics through participation, encouraged by a facilitator. This ensured that participants could develop mutual understanding, trust, and respect.
• Context-specific rather than based on models imported from outside. Rather than imposing new cultural models, the human rights curriculum offered participants a new critical perspective through which they could reflect on their lives.

Strategy: Create coordination signals

Why do social expectations of reciprocity persist? One reason is that a person who questions this norm has no way to know what others in the society really think about it. Such a person will be understandably reluctant to change her own behaviour if it seems uncertain or unlikely that others will change as well. Social norms change when enough people in a reference group believe that enough people are changing. “Joint attention” – the capacity to witness an event while also witnessing others witnessing that same event – shapes expectations of how others will react when that same event happens again (Cislaghi, Manji, and Heise 2017). Mechanisms that can tease out and reveal collective attitudes can help relieve sociability pressures: people tend to feel safer in rejecting a norm when they are sure others will also do so. The aim of this strategy, then, is to build up mechanisms that help people signal to others that they would like to relieve social pressures around corruption. Two types of mechanisms can help send these signals.

Method: Small-torches approach. When individuals signal their support or disapproval of an existing norm, they can help create a critical mass. A well-reported example from India features worthless zero-rupee notes created by an NGO and engraved with the words, “I promise to neither accept nor give a bribe.” These were essentially a signalling mechanism: in handing over the notes to would-be bribe takers, citizens were signalling to public officials that they no longer tolerated norms of corruption, but instead expected integrity. Using the notes in public settings was also a way of signalling to fellow citizens that the individual refused to accept corruption as the norm (World Bank 2014). There have been other projects using similar “small signalling” approaches. In Serbia, service providers were requested to wear pins stating, “I work for the salary, not for the gift!” (Baez-Camargo 2017, p. 5).

Method: Lighthouse approach. This approach relies on a more publicly visible repudiation of a corruption norm, typically through public acts such as performances or visual demonstrations, that encourage the public to reject corruption. An example of the lighthouse approach comes from Paraguay. Tired of how politicians pocketed public money, the owner of a chain of tailor shops created a suit without pockets, dubbed “the anti-corruption suit.” The “Ibáñez Collection” of men’s pocketless suits, named for José María Ibáñez, a Paraguayan politician known for his abuses of power, was widely showcased in the Paraguayan media and abroad. The stunt stimulated a public reaction that allowed for a visible, widespread rejection of corrupt norms (Zúñiga 2018).

Key practical consideration. In supporting the creation of these signal mechanisms, practitioners can work with civil society. In addition, it is helpful to reach out beyond NGOs to members of the arts community. Effective signals often need inspiration, and they also need to be grounded in specific features of the local context, and so the process of developing signals should be led by people within the community.

Strategy: Invest in a self-image of integrity at the community or national level

In our initial example, gift giving seems to be aligned with the society’s self-image: many people may believe “it is the right thing to do because in this country or community this is how we should behave.” Strategies that build up an alternative self-image in which reciprocity is superseded by norms of integrity can help relieve sociability pressure (Baez-Camargo 2017). There are different mechanisms by which to cultivate pro-integrity self-images. The approach here is based on a long-term perspective, as the creation of a new self-image relies on the inculcation of new values. We outline three methods.

Method: Values training. Reports from Rwanda suggest that the government’s attempt to instil an integrity-based self-image has played an important role in strengthening anti-corruption norms (Heywood et al. 2017). This self-image was cultivated through itoreros, precolonial-style training camps where participants spend several weeks learning Rwandan history, precolonial values, and national policies in order to recover traditional values and a “Rwandan way.” Itoreros cultivate values around patriotism, integrity, heroism, leadership, commitment, dignity, self-esteem, creativity, entrepreneurship, rights, and how to live with others. With such a shift in self-image, corruption becomes associated with a lack of values, lack of dignity, and betrayal of the nation; it becomes an enemy of development and peace (Heywood et al. 2017).

Method: Cultural interventions. Self-images can also be constructed and deconstructed by culture, such as music, theatre, and literature. An example of this is Gbagba, an anti-corruption children’s book steeped in a folkloric tradition, where children learn about virtues and vices. Through Gbagba, author Robtel Neajai Pailey hopes to build a movement of children who question corruption and to embarrass adults into living more authentic, ethical lives. The book has been placed on the list of supplemental readers for 3rd to 5th graders in Liberia and for Primary 3 in Ghana, piloted in 30 schools, and turned into a song, music video, radio programme, and stage play. Such an approach is different from one-off artistic interventions – the point here is to find ways to cultivate norms that become internalised by tapping the transformative potential of creative methods such as forum theatre.

Method: Education in schools. Including values, integrity, and anti-corruption education in school curricula is another long-term approach to self-image construction. In fact, school-based education is an intervention mandated by Article 13 of the United Nations Convention Against Corruption. It may involve offering specific courses, revising curriculum frameworks, and developing specific learning tools and reading material (OECD 2018). Learning from interventions in other areas may be useful too. The Gender Equity Movement in Schools (GEMS) in India is a school-based approach that seeks to foster more gender-equitable norms among female and male students ages 12–14 (Alexander-Scott, Bell, and Holden 2016). Group activities include role-playing games, interactive extracurricular activities, and critical reflection-centred lessons that explore topics like girls attaining higher education, delaying marriage, and more equitably sharing household tasks with men and boys. The Maharashtra state government has integrated key elements of GEMS in gender education in all of its nearly 25,000 public schools. Evaluations of the programme showed that after two years of participation, students were more likely to support higher education for girls, openly express opposition to gender-based violence, and champion a higher marriage age.

Key practical considerations. Research into ethics training suggests that a “rules and principles” approach to school-based interventions may be insufficient if it is not matched with competency-based training that focuses on building up subject-matter knowledge, reasoning, and problem-solving skills, as well as advocacy and consensus-building capabilities. Case-scenario didactic methodologies may be the most effective way of building up these capabilities (Whitton 2009).

Kinship pressures

Strategy: Coordinate collective change

Relieving the pressures from kinship norms entails a basic coordination problem. Why should one kin group refrain from corrupt practices – such as demanding resources, jobs, and favours from their relatives in the public service – when they cannot be sure other kin groups will do the same? Part of the challenge lies in getting all actors to move away from this norm at the same time, a shift that requires some kind of coordination mechanism around which collective change can take place. This policy path, then, is in part is about finding people or mechanisms to coordinate behaviour so that family norms can be collectively reinterpreted. As a result, interventions need to convince each kinship group to refrain from corruption and convince them that others will refrain as well.

Method: Trendsetters. One specific mechanism for coordinating collective behaviour that is becoming increasingly popular consists of targeted interventions aimed at trendsetters (Bicchieri 2016; Paluck, Shepherd, and Aronow 2016). Trendsetters are “first movers,” individuals who break free from established norms in a way that can inspire and mobilise others to follow suit. This approach recognises that change often comes about through imitation of successful role models, and so identifying and supporting these leading individuals, as opposed to engaging indiscriminately with a group, may be a more effective means to ultimately change collective behaviour. Working through trendsetters (also referred to as early adopters, norm entrepreneurs, and influencers) is often suggested as a solution to many challenges, including bullying in schools, reproductive health, and smoking.

Trendsetters may hold political office but are more likely to derive their potential to inspire others outside the lines of formal authority, through a shared identity, trust, and credibility. Religious leaders are one such example. The India Heritage Research Foundation,for example, engages spiritual leaders in an attempt to change behaviour around sanitation. Even fictional characters can shape the audience members’ perceptions about what behaviour is common or appropriate in their society (Singhal et al. 2003). The highly successful Peruvian television soap opera Simplemente María helped change norms around education. A storyline involving a young maid who learns to read and gets ahead through hard work established a norm and aspiration adopted by some viewers, leading to increased enrolment at adult literacy classes and a rise in the social status of maids in Peru and neighbouring countries (Hoffmann and Patel 2017).

Key practical considerations.The first important step in any successful programme around trendsetters is to identify these individualsusing social network analysis. Although this might seem a daunting task, research in psychology suggests that such influential individuals are widely known and recognised within a social network (Paluck, Shepherd, and Aronow 2016). Once identified, trendsetters can then be supported with capacity building, training, or other kinds of resources.

Strategy: Construct social spaces for horizontal negotiations around norms

Reducing the influence that kinship norms exert on corrupt practices requires collective change. This in turn is less about a new contract between state and society and more about a social covenant reached through horizontal negotiation between the different kinship groups in society. It requires a space in which alternative norms can be considered and discussed. This strategy thus aims to facilitate a dialogue and create a space in which to demonstrate alternative norms.

Method: Provide an infrastructure for normative dialogue. One example of a programme designed to construct social spaces for horizontal negotiations is the Voices for Change (V4C) programme in Nigeria, funded by the UK Department for International Development (DFID). It set out to strengthen the environment for gender equality and to empower young women and men (aged 16–25) by changing social norms in three key areas: women in leadership, women’s role in decision making, and violence against women and girls. As part of the approach, V4C aimed to trigger attitudinal change through an engaging radio drama series that portrayed positive attitudes and behaviours on gender equality and empowerment of women. The radio drama combined a mix of real-life practical scenarios in the lives of young people with entertaining storylines that provided a distinct alternative to existing norms around gender. In engaging with the show, viewers could consider alternative forms of behaviour and debate them with friends and family. At the same time, V4C provided virtual safe spaces for young people to discuss the show through an online portal, drawing on the logic that if media is consumed with a discursive element, it can amplify shifts around norms. These online spaces also offered users support and information relating to their physical and emotional well-being.

Key practical considerations.The evaluation report on this intervention offers important practical advice. Chief among them is that it is important to carefully adapt an intervention to its users and their environment. The evaluation found that people in the 14 intervention sites were at different starting points in terms of gender norms, and that this may have affected how V4C interventions were received and ultimately how effective they were. It is therefore important to keep these kinds of interventions at a small scale so that contextual factors can be taken into account.

Strategy: Support the collective establishment of pro-integrity norms

Kinship norms create pressures around social status. The more generously you provide for your kin and community, the higher status you may receive. Shunning corrupt practices can leave a public official as a social pariah (Hoffmann and Patel 2017). A further strategy to relax this norm can focus on changing notions of status from those associated with providing for family to those around public service.

Method: Imihigo public service mechanism. One interesting mechanism to shift notions of status away from fulfilling kinship obligations and towards public service has been used in local government in Rwanda. It is called imihigo, a traditional term that means “vow to deliver.” Villages are asked to identify the activities they consider to be priorities for service provision, after which public agencies each year sign formal public service agreements to deliver key specific outputs (ADB 2012). In one sense, imihigo public service agreements amount to a planning and monitoring system; yet included in the system are norm-building mechanisms around which post-kinship norms of integrity and public service can be nurtured.

First, the very concept of imihigo emphasises a clear normative sense of commitment to action and personal responsibility. The concept is rooted in the traditional Rwandan cultural practice whereby two parties would publicly commit themselves to the achievement of a particularly demanding task. Those who fulfilled their pledges became role models in the community, and their feats were remembered in history, while failure led to dishonour (ADB 2012). Just as in former iterations of imihigo, there is an important ceremonial aspect, as all districts participate in a ceremony during which the previous financial year’s targets and performance are appraised and performance agreements for the upcoming year are signed with the president. The top three mayors are rewarded for their exemplary achievements, and the top ten mayors get a chance to take a group photo with the guest of honour.

Key practical considerations. The social visibility aspect of this model is important, imparting a sense of collective normative standards. Political support is also significant: the conferral of prestige by the president strengthens the sense of strong public backing for the norm of public service. Most importantly, this model is locally led, drawing on locally embedded cultural understandings. This is not just branding, but a way to connect behaviour within public administration to personal and communal identities.

Vertical pressures

Strategy: Change norms from the top down

Vertical pressures are often enabled by a strong hierarchical mode of governance whereby people at the top of an organisation dictate the norms upon which institutional practice is based. Although change from the top is challenging, under certain circumstances vertical networks can be a source of change from within (Roll 2011). Leaders at the top can provide a strong signal that there is a “new order” under which rent seeking is no longer tolerated. Baez-Camargo (2017) finds anecdotal evidence that Tanzanian public officials who refuse to succumb to network pressures often simply state the name of the current president, Magufuli, to communicate the idea that under this leader favouritism or other forms of administrative corruption are no longer tolerated. The question then becomes how to identify agents with power to instigate change and how to induce them to use their power to shift their hierarchies towards integrity norms.

Method: Support reformist politicians. The aim of this method is to support leaders at the top of hierarchies who are willing to initiate a flipping of norms in the network – a removal of corruption-inducing hierarchical norms and a cascade of pro-integrity norms. This support can be effective even in unfavourable contexts. For example, though Nigeria’s political system is heavily influenced by vertical patron-client networks, President Obasanjo shielded the reform of a public agency, the National Agency for Food and Drug Administration and Control (NAFDAC), from vested interests. NAFDAC has the mandate to regulate and control quality standards for imported, as well as locally manufactured, food and drugs. Although the agency had been influenced by criminal elements, Roll’s analysis of the case suggests that by 2009 it had become a pocket of effectiveness, led by a dynamic director and unaffected by corruption pressures. The president’s pro-integrity stance enabled this change in the NAFDAC’s administrative hierarchy, in part because it was consistent with his administration’s political goals – in particular, Obasanjo’s desire to improve Nigeria’s international image, especially as he sought an international debt repayment and relief deal. To ensure new kinds of norms within the organisation, the president enforced new procedures for hiring the director of NAFDAC, elevating integrity and public service mission as key selection criteria (Roll 2011).

Method: Take advantage of windows of opportunity. Some change agents may be genuinely committed to reform but lack the opportunity to initiate reforms. Occasions to shift hierarchical norms may be rare but can happen, generally when specific events open a window in which to take advantage of social discontent with the status quo.

Opportunities can arise from revolutions or other “tectonic” movements in politics. An insightful interview with a key participant in the successful anti-corruption reform efforts in Georgia highlights the importance of the preceding Rose Revolution, which provided a strong popular mandate for the government. Seizing the opportunity, reformers swept away old hierarchies across all sectors of government, creating new hierarchies led by non-corrupt leadership willing to establish pro-integrity norms within the institutions (Kovziridze 2017). The upheaval enabled quick and radical reforms, and because results were swift, the public started to appreciate the tough tactics. Windows of opportunity are not always triggered by revolutions – reformist momentum can be created by a whistle-blower exposing corruption, a major political scandal, or public frustration with ineffective anti-corruption efforts (Panth 2011).

Key practical considerations.Windows of opportunity may be wide or narrow; the point is that when these chances arise, it is important to seize the moment. Incentives and interests matter for potential reform, and leaders may only support reform if it is in their interest to do so (Kelsall 2016). Practitioners need a strong understanding of the underlying political economy of a country and of the interests and incentives that influence key leaders who may be in a position to flip networks to integrity. It’s also important to constantly monitor political developments, as shifting alliances or turnover of individual ministers may bring about opportunities for change.

Strategy: Disrupt networks

Vertical pressures are transmitted through hierarchies that are often highly organised. A key reason why these pressures persist is that figures at lower ranks of the hierarchy, once they have been co-opted, may find it difficult to opt out. Indeed, a low-level worker may owe her job to someone at the top and thus remain in a permanent state of indebtedness. A strategy to disrupt patron-client ties between persons at different levels of a hierarchy may relieve some of these pressures.

Method: Rotation system. In certain ministries, senior managers may rotate into different managerial posts. This strategy can potentially sever existing patron-client ties and undermine vertical pressures. This remains a largely unexplored mechanism. There are, of course, possible downsides: it is disruptive and costly, requiring resources for training. However, under certain conditions it may free frontline civil servants from the co-opting pressures of vertical networks, creating an autonomy that, in the end, might outweigh the costs. Even so, the rotation system itself could be co-opted into the vertical network, with senior management exploiting discretion over the system to build up their own patronage network (Fjeldstad 2003).

Method: Expatriate staff. Another mechanism to reduce vertical pressures is to bring outsiders into management roles within an institution. The logic behind this strategy is that strong expatriate leadership, free from patron-client pressures, may be more effective in providing an enabling environment within which systemic changes and new forms of staff behaviour can be implanted (Fjeldstad 2009). This intervention has been tried in revenue authorities in Tanzania and Zambia, for example. The experience of the Zambia Revenue Authority suggests that expatriate senior advisors and top managers who are in place for a limited period can contribute to effective change by building integrity and professionalism in the organisation through systemic changes (de Wulf and Sokol 2005).

Key practical considerations. More research is needed to understand the conditions under which these interventions could work. New initiatives along these lines should also take local circumstances into account.

Horizontal pressures

Strategy: Support collective shifts within organisations

Establishing an ethos of public service can help create a pivot in which everyone on the office floor begins to move away from corruption. Fashioning such a counter-norm within an organisation does not normally require starting from scratch; despite assumptions to the contrary, a public service ethos certainly exists in developing countries. For example, research in Nigeria suggests that public servants are motivated to serve the public good (Roll 2011). How to encourage collective shifts within organisations without resorting to a top-down approach?

Method: Support managers with autonomy and resources. Given the autonomy to do so, pro-integrity leaders and managers within organisations can demonstrate exemplary behaviour, build up organisational values, and create environments where it is safe to challenge norms (Heywood et al. 2017). Even with autonomy, managers cannot induce shifts on their own. Collective behaviour change also does not arise from confrontational exchanges – it is often more effective to bring everyone in an organisation to the same point together. Shared responsibility and a sense of moving together are important in creating a new institutional culture. Getting everyone on board may require some additional resources. For example, support for a new human resources management system may be important in establishing more secure employment arrangements. Civil service jobs can be insecure, either because they are tied to the incumbency of a particular politician or because contracts are short-term. When an individual cannot be certain of future income, this insecurity may heighten the influence of horizontal pressures to engage in self-enriching acts (Fjeldstad 2009) – a situation that is not conducive to normative shifts within organisations. Strengthening employment security may provide a basis for nurturing a culture of integrity over the long term.

Key practical considerations. As a starting point, researchers have found that it is important for organisations to set their own goals and integrity standards rather than importing them from the outside, or having this process micro-managed by central ministries (Grindle 1997). Nonetheless, leaders may need training, resources, and other forms of support towards this end. Furthermore, it is important to support the right managers and to identify whether the managers themselves might resist the changes due to benefits they reap from existing corrupt arrangements (see vertical pressures, above).

Strategy: Overcome pluralistic ignorance

Horizontal pressures may also result from a person’s inflated notion of how many of his co-workers are engaged in corrupt acts. Survey data show that people frequently overestimate the willingness of others to tolerate corrupt practices (Hoffmann and Patel 2017). This is what social psychologists call “pluralistic ignorance,” a form of mistaken cognition that may be a common aspect of corruption perceptions within an organisation. The aim of the intervention is to address this collective ignorance by providing credible information and reshaping perceptions about how much corruption is tolerated within an organisation. Once again, we present two methods.

Method: Provide credible information about the frequency of corruption.Compiling and disseminating information about how much, or how little, corruption actually occurs in peer organisations may be one way of overcoming pluralistic ignorance. Besides encouraging people to refrain from exaggerating the prevalence of corruption, providing reliable information can help change descriptive norms – and subsequent behaviour. This approach has been used successfully in several domains unrelated to corruption: for example, summary information about fellow students’ low endorsement of racial stereotypes has been used to reduce racial bias on campus. However, messages describing disapproval of racial stereotypes at a student’s own campus had a stronger signalling effect than messages outlining disapproval of racial stereotypes at other campuses. In other words, the effect was stronger when the information referred to an in-group than when it referred to an out-group, underlining the importance of thorough research to identify the relevant reference group (Stangor, Sechrist, and Jost 2001).

Method: Highlight integrity acts within organisations. Another way to overcome an “everybody does it” perspective is to organise high-visibility actions within a peer group to challenge perceptions. Accountability Lab’s Integrity Idol is a campaign that rewards honesty among public officials in many different countries, from Nepal to Liberia. Local teams of volunteers travel across their countries seeking “idols” by gathering nominations from citizens, in addition to hosting public forums and generating a national discourse on the need for public officials with integrity. Idols then become part of a competition that is shown on television and played on the radio, with the winners crowned in a national ceremony in the capital. Such a public celebration of integrity sends a message that not everyone is corrupt and that integrity will be rewarded with public acclaim. Analogous interventions that publicly celebrate integrity in a particular government department or district office, with smaller versions of the national event, could further help relieve horizontal pressures.

Key practical consideration. Messaging strategies to overcome pluralistic ignorance should be tailored towards a specific reference group. Such messaging is most effective when the recipient knows that others have received the same information. For example, studies on reconciliation in post-conflict Democratic Republic of Congo show that people who listened to a radio programme in a group setting showed a stronger shift in norms than people who listened to the same programme alone (Tankard and Paluck 2016). This was probably because, first, those who listened in a group knew that others heard the programme too, and second, collective media consumption set the stage for collective deliberation, discussion, and renegotiation of norms. In addition, recent findings suggest that talking about norms using the language of “trends,” instead of a more static description, might be especially promising. That is, instead of saying that “few public officials extort bribes,” one can frame the story as “fewer and fewer public officials extort bribes” (see Köbis, Soraperra, and Troost 2018).

Conclusion: A social norms approach to anti-corruption

A social norms approach to anti-corruption focuses on relieving and shifting some of the social pressures that sustain corruption. The strategies outlined above make use of methods such as dialogue, signalling, value change, information provision, trendsetters, and other mechanisms. Rapid transformational change in norms is unlikely; instead these strategies aim to relieve and shift the pressure exerted by a norm, opening up space for new forms of behaviour and new interventions.

Different types of normative pressures suggest that a mix-and-match approach is necessary: that is, multiple normative pressures require a multipronged response. Sequencing is an important consideration. Returning to the example presented at the beginning, a strategy to relieve and shift social pressure could start off with interventions that try to overcome pluralistic ignorance within the municipality, which could help dislodge some entrenched norms. Building on this, a subsequent intervention that provides autonomy and resources to municipal managers could help instigate a collective change within the municipality towards a stronger public service norm. Because kinship pressures might persist in spite of these efforts, support could be given to trendsetters in the broader community to help shift norms around favouritism to the family. It goes without saying that these strategies should take place in the context of wider efforts to build integrity through better checks and balances, support to civil society, and increased enforcement of anti-corruption laws.

Across these different strategies certain themes recur. The first is that norms can fight norms. Many of the strategies involve constructing or drawing upon alternative reference points. The aforementioned Imihigomodel draws on alternative norms of public service to try to eliminate negative, corruption-inducing norms. The strategy therefore invokes what Cislaghi calls “protective norms” in a society, meaning those values that are generally geared towards positive outcomes (Cislaghi, Manji, and Heise 2017). Second, networks can be used to fight networks. Social networks enforce the sanctions that can lead to the social pressures to engage in corruption; yet networks can also be employed to mitigate the same pressures. Top-down network “flipping” is one strategy. This relies on individual leaders, but many other strategies require collective shifts and negotiation.

In general, these strategies should be deployed within specific contexts where there can be intensive engagement, such as within a community or sector. Such an intensive focus may mean that social norms can shift in a relatively short time frame of two to three years (Cislaghi, Manji, and Heise 2017). All interventions should be accompanied by an analysis of the political economy and should be locally led to avoid oversimplifications such as “rendering societies technical,” meaning that easy fixes are proposed for complex socially embedded phenomena (Li 2007). A locally grounded policy intervention is also less likely to have unintended side effects (Fisman and Golden 2017). When practitioners recognise and avoid these pitfalls, a social norms approach can provide a novel lens to understand, diagnose, and eventually change the social forces that sustain corruption.