THE MALAWI CONTEXT

In Malawi, the growth of corruption is a relatively recent phenomenon that has been linked to the introduction of multiparty representative democracy and the liberalisation of the economy during the mid-1990s (Anders Reference Anders2010). The need for campaign funds and the weakening discipline of government employees who exploited the newly available economic opportunities have fuelled the increase in corrupt practices (Anders Reference Anders2010: 124–5). An important influence on the spread of political corruption in particular has been the weak regulatory framework for political party financing, as Dulani (Reference Dulani and Amundsen2019) shows. Since the 1990s, there have been a number of high-level corruption scandals involving ministers and directors of parastatal companies (Anders Reference Anders2010: 122–5) but in contrast to Cashgate, these scandals remained limited to specific ministries and did not trigger sustained law enforcement efforts.

The Cashgate scandal is remarkable for two reasons. On the one hand, it was the biggest corruption scandal ever uncovered in MalawiFootnote ¹ and, on the other hand, the scale of the law enforcement response was unprecedented. The latter was mainly due to the political landscape at the end of 2013 that created a favourable climate for sustained law enforcement efforts. It is estimated that within a six-month period between April and September 2013 over US$32 million were stolen and it is possible that more than US$280 million have been stolen since 2009 (Baker Tilly Reference Tilly2014; RSM 2016). The large-scale theft of public funds was discovered in September 2013 when the country was in the middle of the run-up to the May 2014 elections. As the extent of the corruption racket was revealed, President Joyce Banda ordered a forensic audit, which was funded by the British government (Baker Tilly Reference Tilly2014). By 2014, one hundred suspects had been arrested (the overwhelming majority were released on bail).

The forensic report on Cashgate by the auditing firm Baker Tilly (Reference Tilly2014) covers ‘the extraction of ‘cash’ (with the main currency being that of Malawi Kwacha) using systematic money laundering activities through commercial organizations’ (Baker Tilly Reference Tilly2014: 4). The report confirms that more than 13 billion Malawi Kwacha (~US$32 million) were stolen over a six-month period between April 2013 and end of September 2013. Drawing on the forensic audit (Baker Tilly Reference Tilly2014) and the interviews we conducted in the qualitative study, our study differentiates three types of corruption. The first type is centrally organised theft of government funds, largely facilitated via the Integrated Financial Management Information System (IFMIS) and organised by a relatively coherent group in key positions. The second type is high-level procurement fraud involving ‘overvalued and high value transactions, payment for goods not received and funding transferred internationally’ (Baker Tilly Reference Tilly2014: 5), which appears to be less centralised, but tends to include a number of networks of senior managers in strategic relationships with junior staff – especially accountants and IT-staff. These two types of high-level corruption are differentiated from smaller, localised corruption rackets mainly involving senior and mid-level managers (and sometimes elected councillors) at the district level focusing on bribery, theft of government property as well as allowance and procurement fraud.

At the end of May 2014, Joyce Banda was succeeded in office by Arthur Peter Mutharika and the law enforcement effort gathered steam, resulting in the arrest of dozens of suspects, far-reaching criminal investigations and a series of high-profile criminal trials. Thirteen cases had been concluded at the time of writing in April 2020. The first judgment was handed down against the Principal Secretary in the Ministry of Tourism, who was sentenced in October 2014 to three years in prison for money laundering. Only a month later, an accounts assistant who had been found with US$66,000 in cash was sentenced to nine years in prison. Several other government officials were also found guilty of money-laundering and received prison sentences. One of them, an Assistant Director in the Ministry of Tourism, was sentenced to imprisonment for seven years. In September 2015, a wealthy businessman and functionary in Joyce Banda's People's Party was sentenced to 11 years in prison. In a related case, the Minister of Justice was found guilty of plotting the murder of the Budget Director in the Ministry of Finance and was sentenced to 13 years in prison. At the time of writing, the Budget Director is facing trial for money laundering and theft with 17 co-accused including the former Accountant General and other government officials.

The events described above highlight several factors that created favourable conditions for the sustained law enforcement effort. The Cashgate scandal hit the headlines during the election campaign for the May 2014 elections and President Banda was compelled to take action to minimise the fallout for her bid for the presidency. By commissioning the forensic audit and calling for prosecutions she clearly sought to distance herself from the scandal. The international development partners put Banda under considerable pressure by suspending budget support, which had severe consequences on the government budget. Faced with a massive cut in available funding in the midst of the election campaign (Malawi's government budget relies to a large extent on foreign development assistance) she had no other option than to take action. After Banda lost the elections, her successor exploited the opportunity to support the law enforcement efforts against individuals that were mainly associated with his predecessor. Malawi's development partners led by the UK and the EU continued to demand decisive action against corruption. They provided financial and logistical support to the law enforcement agencies that launched a large number of investigations and prosecutions.

Despite these enabling factors in favour of stronger law enforcement there have also been inhibiting factors at work. Malawi's legal framework is characterised by the coexistence of statutes dating back to the colonial period and more recent anti-corruption legislation. The layering of new laws such as the 1995 Corrupt Practices Act and the 2006 Money Laundering Act on top of colonial statutes mainly reproducing English law of the 1950s has not been matched by efforts to address inconsistencies, gaps and ambiguous provisions. This had direct effects on the prosecutions and trials targeting the perpetrators of the Cashgate corruption racket. Since the offences mainly concerned the unauthorised drawing of government cheques and money transfers to private banks, the prosecutors struggled to charge the accused with theft, which requires the perpetrator to take possession of property. Consequently, they mainly relied on money laundering charges, drawing on the untested 2006 Money Laundering Act. However, according to senior law enforcement officials we interviewed, this law ‘was not up to the task’ and it was indeed replaced in 2018 by the Financial Crimes Act. The inadequacies in the law became apparent when one of the main suspects of the Cashgate scandal, a government accountant, was acquitted on appeal. In 2015, the High Court in Lilongwe had sentenced the accountant to eight years in prison for theft and money laundering. In 2018, the Supreme Court overturned the conviction for theft. The conviction for money laundering, in turn, was also overturned because the Supreme Court judges considered theft as predicate offence for money laundering. Thus, without a conviction for theft the conviction for money laundering was void.

Further, law enforcement efforts in Malawi have been significantly affected by a lack of resources. A recent study (Chingaipe Reference Chingaipe2017) found that the Anti-Corruption Bureau (ACB) is understaffed, especially in the prosecution and investigation departments which have significant staff shortages (Chingaipe Reference Chingaipe2017: 162–3). Another major factor affecting the agency's effectiveness is the lack of financial resources. According to the report, ‘the ACB faces acute resource constraints which [are] compounded by erratic funding’ (Chingaipe Reference Chingaipe2017: 154). The situation is similar at the other law enforcement agencies: The Fiscal and Fraud Unit, the Directorate of Public Prosecutions and the Financial Intelligence Authority are all understaffed and under-resourced.

Finally, prosecutions and trials have been hampered by frequent and sometimes lengthy delays. For example, the trial at the High Court in Lilongwe against 18 of the main suspects including the former Budget Director at the Ministry of Finance and the former Accountant General started in December 2015 and had not been concluded at the time of writing in April 2020. One of the main reasons for the slow progress were the frequent adjournments and the small number of sitting days per month. For example, in 2019, the court sat on only three days per month with months of adjournments in between. Similarly, of the initial 60 Cashgate-related cases in 2014, only 13 had been concluded by 2019. About 40 cases never made it to the trial stage. For instance, in August 2019 the Magistrate's Court in Lilongwe ordered the Anti-Corruption Bureau to appear in court to explain why 14 accused charged with Cashgate-related offenses should not be discharged. These accused including high-ranking officials and the former Inspector General of the police were heard in court for a plea in October 2014 but after almost five years the trials had not even started. The backlog of cases and frequent delays are by no means unique to the Cashgate prosecutions. They affect all aspects of the justice system in Malawi (Kanyongolo Reference Kanyongolo2006: 101, 140–1).

METHODS EMPLOYED IN THE STUDY

The research builds on the in-depth anthropological study on the Malawi civil service by Anders (Reference Anders, Blundo and Lemeur2009, Reference Anders2010) and the mixed-methods research by Seim (Reference Seim2014) and Robinson & Seim (Reference Robinson and Seim2018) focusing on motivations, rationalisations and practices of Malawian public officials who perpetrate corruption. The secretive and illegal nature of corruption is a considerable barrier to rigorous empirical research on this topic (Blundo Reference Blundo, Nuijten and Anders2007; Anders Reference Anders, Blundo and Lemeur2009). The researchers addressed these challenges by designing a mixed-methods research approach that factors in the sensitive nature of the information sought and triangulates the different data sources to understand the potential for corruption deterrence in Malawi comprehensively.

The sociocultural context of Malawi provides a compelling environment to examine the effect of law enforcement on government officials’ perception of deterrence. As in the rest of Africa, the modern state in Malawi is a product of the colonial encounter resulting in a disconnect between the moral values promoted by the state apparatus and prevalent moral values in society (Ekeh Reference Ekeh1975). In Malawi, as in other African societies (Olivier de Sardan Reference Olivier de Sardan1999; Blundo & Olivier de Sardan Reference Blundo and Olivier de Sardan2006; Smith Reference Smith2008), there is a lack of deeply entrenched moral values and social sanctions that compel compliance with the law of the modern state (Anders Reference Anders, Blundo and Lemeur2009). In contrast, there is a heavy emphasis on the norm of sharing with others, a social obligation felt in particular by the more affluent members of society who are expected to share their wealth. Specifically with regard to Cashgate, it seems those who were involved in the theft of government funds often made a point of being generous to improve their social standing in the community. These values and the lack of social sanctions lend support to the assumption that any perceived subjective deterrent effect can be mainly attributed to the impact of law enforcement.

The qualitative research consists of 51 interviews with government officials on Capital Hill, the area in Lilongwe where most of the government ministries are located, and key informants in the law enforcement agencies and the judiciary, conducted between 25 November 2016 and 12 January 2017. In addition, the qualitative research included observations and informal conversations at key sites relevant to Cashgate and two follow-up visits to Malawi in May and August of 2019. All interviews were semi-structured, allowing the interviewees to expand on issues or introduce themes they deemed important and enabling the researchers to adapt to specific topics that came up in conversation.Footnote ²

To determine the perception of government officials working with IFMIS, sampling specifically focused on controlling officers in senior management roles, the Common Accounting Service and junior accountants. Sampling covered most government ministries including the Office of the President and the Cabinet, the Ministry of Finance, the Ministry of Education, the Ministry of Health and the Ministry of Agriculture.Footnote ³ Thanks to the personal introductions, assurances of anonymity and the team members’ familiarity with the Malawi context, the team was able to build rapport with the interviewees who were all willing to openly discuss the Cashgate scandal, corruption and the impact of the law enforcement response.Footnote ⁴

During each interview, the interviewer discussed the interviewee's understanding of the two key concepts employed by the study: corruption and deterrence. Across Africa, there are rich vernaculars of corruption and widely shared indigenous moral norms that justify corrupt practices by invoking social and moral obligations to kith and kin (Mbembe Reference Mbembe2001; Smith Reference Smith2001; Hasty Reference Hasty2005; Blundo & Olivier de Sardan Reference Blundo and Olivier de Sardan2006; Anders Reference Anders, Blundo and Lemeur2009). However, the existence of multiple moral discourses of corruption does not imply that government officials’ definition of corruption is at odds with the legal definition of corruption and corruption-related offences. Even though officials might justify some corrupt practices by invoking social and moral obligations, especially with regard to petty corruption (Anders Reference Anders, Blundo and Lemeur2009; Blundo & Olivier de Sardan Reference Blundo and Olivier de Sardan2006), they usually do not condone high-level corruption on moral grounds. This was confirmed by our study. No government official we interviewed offered a cultural interpretation of the high-level and centralised theft of government funds. In each interview the interviewer confirmed that deterrence referred to the omission of corrupt practices due to the fear of punishment. The interviews unpacked the concept by differentiating certainty, swiftness and severity of punishment.

A large-scale survey conducted in January and February of 2017 provided a test of causal relationships between dimensions of law enforcement and perceived corruption deterrence.Footnote ⁵ The survey was executed among a random sample of 524 government officials across Malawi. The survey sampled 444 government officials posted in the headquarters of government agencies in Lilongwe, Blantyre and Zomba, and 80 government officials posted in decentralised agencies in eight randomly selected districts: two districts in the Northern Region (Mzimba and Nkhata Bay), three districts in the Central Region (Mchinji, Salima and Ntcheu), and three districts in the Southern Region (Blantyre, Phalombe and Zomba).Footnote ⁶ This multi-level sampling strategy facilitates the comparison of corruption patterns and perceptions across centralised and decentralised government offices.Footnote ⁷

Embedded in the survey was a conjoint experiment to test how aspects of the law enforcement response affect officials’ perceptions of corruption deterrence. We followed the procedure laid out in Hainmueller et al. (Reference Hainmueller, Hopkins and Yamamoto2013) in designing and analysing the conjoint experiment. In such an experiment, the participant is asked to examine two hypothetical law enforcement environments (aka profiles) that differ along the law enforcement dimensions identified above as possible drivers of deterrence: the certainty and swiftness of detection and punishment, and punishment severity.Footnote ⁸ In this article, we focus on a subset of the dimensions varied in the conjoint experiment, presented in Table I, along with the text for the profile descriptions. In each profile, the description of each dimension was randomly selected from the options displayed in Table I.

Table I

Conjoint experiment dimensions and values.

After examining the two environments, the respondent was asked to evaluate them on a set of outcomes: which one of the two environments would result in higher corruption, or harder working government officials, or a more respected judiciary, for example. Because each environment's descriptions of the dimensions are randomly selected from a set of options (see Table I), the researcher was able to empirically identify the causal effect (specifically, the average marginal component effect, or AMCE) of each dimension on the respondent's choice of one environment over the other.Footnote ⁹ In total, the conjoint experiment documented 17 outcomes that group into three areas: bureaucratic performance, accountability and the effectiveness of anti-corruption systems, though we focus only on a subset of the outcomes within bureaucratic performance in this article.Footnote ¹⁰

The conjoint experiment captured the respondents’ perceptions of what would happen in the hypothetical law enforcement environments presented to them, rather than their perceptions of what has actually happened in a real-world scenario (e.g. post-Cashgate). By adapting the features of the hypothetical law enforcement environments, the conjoint experiment is a more malleable method but is also more abstract, thereby providing a complement to the more concrete but also more context-specific interviews. Further, by disaggregating the analysis of the conjoint experiment according to salient characteristics of the government officials (e.g. level of seniority or type of role), we can assess patterns of corruption deterrence perceptions within different sub-groups. We include such sub-group analysis in the following two sections.

CORRUPTION IN GOVERNMENT AGENCIES: SURVEY RESULTS

While the primary goal of the survey was to probe perceptions of corruption deterrence, it also provided an opportunity to document patterns and trends in corruption pre- and post-Cashgate from the perspective of government officials in Malawi. This in-depth focus on the perceptions of government officials supplements previous surveys on corruption in Malawi by Chinsinga et al. (Reference Chinsinga, Dulani, Mvula and Chunga2014) and Chunga & Mazalale (Reference Chunga and Mazalale2017).

The officials in the survey estimated that the average corrupt official in their institution obtains 24% of his or her income from corruption.Footnote ¹¹ Answers to this question are highly correlated with answers to a question asking respondents to estimate the percentage of officials in their institution who have engaged in corruption in the last year (correlation coefficient: 0.80, P = 0.0000). This suggests that highly corrupt institutions in terms of the proportion of officials involved are also highly corrupt in terms of the monetary impact. To understand the prevalence of various corrupt practices, officials were also asked whether they had committed various forms of corrupt acts in the six months preceding the interview. Three forms of corruption were examined: the acceptance of bribes, the issuance of false receipts, and the favouritism of friends and family in awarding them government contracts or jobs. These three questions represented the low-level, routinised corruption, the mid-level, more sophisticated corruption and the most organised and high-level corruption, respectively. Table II compiles the information provided by officials in response to direct questions about their own behaviour.

Table II

Per cent of officials who acknowledged engaging in corruption.

On average, 28% of all officials accepted a bribe in the past six months (Table II, first column, first row). On average, 25% of officials stated they have falsified a document in order to pocket some funds (Table II, second column, first row). Finally, on average, 27% of the sampled officials have selected a friend or family member for a government job or contract (Table II, third column, first row).

These results were less pronounced among senior officials across all three types of corruption (Table II, second row). Non-HQ officials report higher rates of corruption than those in headquarters in terms of accepting bribes (representative of low-level corruption), falsifying documents (representative of mid-level corruption), and favouritism for family or friends in government jobs and contracts (representative of the high-level, organised corruption endemic in Cashgate) (Table II, fourth and fifth rows).Footnote ¹² In other words, low-level government officials and those working outside of headquarters report higher rates of corruption. This should not come as a surprise as staff outside of the headquarters tend to have more interaction with members of the public who might be tempted to pay bribes. By contrast, bribery is less likely to occur in the ministerial headquarters, where officials have only limited interaction with the public. At the ministerial headquarters, it is more likely to find corruption requiring a high level of organisation such as Cashgate.

Another set of questions in the survey asked specifically about trends in corruption rates pre- and post-Cashgate. Eighty-five per cent of officials state that corruption in their institution had remained the same or increased since 2012 (immediately before Cashgate). This perception is consistent regardless of post, employer location or level in the public-sector hierarchy. While one interpretation of this finding is that the law enforcement efforts responding to Cashgate were unsuccessful, it is equally possible that these efforts simply ‘shone a light’ on corruption in Malawi such that the survey responses merely reflect the intensity of the public debate about corruption when actually it remained the same or even decreased. Unfortunately, the survey data do not allow us to adjudicate between these two plausible explanations, so this will remain an avenue for future research. In sum, the survey data indicate that corruption in Malawi is still prevalent after Cashgate. Probing the deterrent effect of the law enforcement response, we examine perceptions of deterrence among government officials.

PERCEPTIONS OF DETERRENCE

One important aspect of deterrence is publicity and communication (Kennedy Reference Kennedy1983: 6). Without knowledge of the sanctions and information about punishment meted out, no deterrent effect can be expected, as likely perpetrators need to be aware of the threat of punishment. Our research shows that government officials were well-informed about the extent of corruption and the law enforcement efforts. All interviewees stated that the perpetrators of Cashgate and other high-level corruption scandals knew that what they were doing was illegal and against regulations. As one of them put it: ‘They knew. People know the law. They know there are protocols. They know the three Ps: The Public Audit Act, the Public Finance Management Act and the Public Procurement Act.’ Another one pointed out that ‘there is no excuse’ as ‘in terms of government procedures everything is in black and white’.

The findings present a nuanced picture of government officials’ perceptions of the effect of the law enforcement efforts. The analysis of the 51 interviews of the qualitative study paints a detailed picture of the potential for corruption deterrence. According to 32 interviewees, the law enforcement response to Cashgate had a deterrent effect on the civil service but the interviews also highlight the problem of selective justice due to political interference and the ability of corrupt government officials to remain undetected. For example, a mid-level manager in the Ministry of Trade stated:

The remaining 19 interviewees stated that the law enforcement effort had no or only a limited deterrent effect whilst also highlighting that the arrests and punishment of corrupt government officials and businesspeople did have an effect on government officials who became ‘very cautious’. However, six of the 19 government officials who had a more pessimistic outlook denied any deterrent effect and claimed that corrupt officials merely have become more adept at avoiding detection. According to one of them, a mid-level official in the Office of the President and Cabinet, ‘people have become more shrewd’ rather than deterred due to the law enforcement effort.

The interviews revealed significant changes in perception since 2012. The interviewees in the qualitative study stated that in 2013 the perpetrators were quite confident that they would not face punishment. They were assured protection by the key conspirators and were tempted by greed to benefit from the illegal transactions. As one interviewee put it succinctly: ‘Maybe they were deceived. They thought their bosses would protect them. It was pure self-interest, greed, to get the posh lifestyle.’ However, the impunity they were hoping for did not materialise due to the efforts pushing for prosecutions under both Joyce Banda and Arthur Peter Mutharika. It seems government officials who became involved in the illegal activities benefitted and acquired large amounts of cash, which they typically spent on vehicles and real estate. According to 23 officials interviewed in the qualitative study, political influence greatly affected the fraudulent transactions until September 2013, with certain officials being fast-tracked and promoted to strategic positions. According to a senior official in the Ministry of Education, ‘it was like a virus which was bound to spread across many Ministries’. All interviewees pointed out that the enrichment was conspicuous and unchecked, resulting in a feverish atmosphere around the government offices on Capital Hill in Lilongwe with lavish consumption, new cars and unexplained wealth openly displayed. According to our interviewees, the perpetrators felt very confident indeed. They were ‘very generous’ and explained their wealth was acquired by doing ‘business’, a common source of income for government officials (Anders Reference Anders2010: 70–98). They expected to enjoy impunity and wealth from what one interviewee dubbed ‘the Fountain’.

One of the interviewees, a senior lawyer in the Ministry of Justice, drew a comparison with the situation in 2010/11 when large-scale procurement fraud estimated at roughly 400 million Kwacha was discovered in several government agencies including the Malawi Police Service and the Accountant General's Office. However, the discovery by the Financial Intelligence Unit did not lead to any significant investigations and prosecutions at the time. He pointed out that participants in this scheme felt encouraged by the lack of sanctions and had a sense of impunity that directly fuelled the large-scale theft using the Integrated Finance Management Information System (IFMIS) in 2013. It is noteworthy that some of the key actors had also been implicated in the procurement fraud discovered in 2011.Footnote ¹³ This is backed up by several interviewees in the qualitative study who highlighted the low certainty of punishment in 2011 as a major contributing factor to the palpable sense of impunity in the government quarter on Capital Hill in 2013.

The mixed-methods evidence surrounding perceptions of deterrence among the government officials enrolled in our study in Malawi complements the findings of the qualitative study. As discussed previously, the conjoint experiment presented participants with hypothetical law enforcement scenarios and asked them to evaluate the potential for corruption deterrence in these scenarios. By randomly assigning the characteristics of the law enforcement scenarios, we were able to identify which characteristics make a perceived deterrent effect more or less likely. However, this approach is also quite abstract. To facilitate a more context-specific analysis of law enforcement and corruption deterrence, we compare the results from the conjoint experiment to the findings from the in-depth interviews.

Before we examine in detail the government officials’ perceptions with regard to certainty, swiftness and severity, we display the results of the conjoint experiment in Figure 1. In this figure, the horizontal lines represent the effect of each dimension on the outcome variable listed above the plot (point estimate with 90% confidence interval).Footnote ¹⁴ Lines below zero indicate lower corruption, and therefore a perceived deterrent effect, whereas lines above zero indicate higher corruption, and therefore the opposite of a perceived deterrent effect. Panel (A) shows how the various dimensions of the law enforcement environment affect perceived deterrence of corruption specifically among government officials, and Panel (B) shows how the various dimensions of the law enforcement environment affect perceived deterrence of corruption more generally (not only among government officials). In the discussion below, we will refer back to Figure 1 as we discuss certainty, swiftness and severity one by one.

Figure 1

Law enforcement causes of corruption deterrence in the conjoint experiment.