The surprising case of police bribery reduction in South Africa

The paper examines why there was a reduction of almost 15% in police bribery in Limpopo province, South Africa between 2011 and 2015, compared to only a 4% reduction the country overall. Drawing on statistical analysis and in-depth qualitative fieldwork, the research shows that the reduction occurred during an unprecedented anticorruption intervention in the province that did not directly tackle police bribery. Despite this, the intervention ’ s high visibility, along with uncertainty among the police of its mandate, was likely to have made police less willing to engage in bribery during this period. While police sector-specific characteristics (high degree of discretion, peer solidarity and contact with criminals) make fighting entrenched corruption particularly difficult, the research shows how a disruptive event can counteract these factors and how this can happen more quickly than previously anticipated. For long-term impact, however, disruption strategies likely need to be driven by strong leadership and structural changes that will continually disrupt corruption patterns.


Introduction
When South African President Cyril Ramaphosa took office following the resignation of Jacob Zuma in early 2018, he quickly identified fighting corruption and restoring the integrity of public institutions as one of his top priorities. In this maiden speech to parliament, he said, 'A new dawn is upon us and a wonderful dawn has arrived … This is the year in which we will turn the tide of corruption in our public institutions' [1]. This was celebrated in the world's media, even as the scale of the challenges facing him were set out in detail (see, for example, [2][3][4][5]). Perceptions of how South Africa's corruption problem compares to other nations have worsened in the last decade: the country was ranked 43rd out of 179 countries on Transparency International's Corruption Perceptions Index in 2007, in which a lower ranking indicates a control of corruption, but had risen to 71st out of 180 countries a decade later, in 2017 [6,7]. Indeed, South Africa has recently been described as possessing a Bunique politicalcriminal climate of corruption^([8]: 3; [9]).
Fifteen years ago, Transparency International's Global Corruption Barometer (GCB), a household level survey, posed the question to its South African respondents: if given a Bmagic wand^that could eliminate corruption from one institution in the country, what institutions would you wave it at? The police garnered the most votes: almost a quarter of South Africans volunteered the police as their first choice, while only 4 % named the courts, by comparison [10]. Faull [8] writes that the reason why police corruption is such a salient issue is because 'the police represent the only visible point of support and hope for a safer future'.
The South Africa Police Services (SAPS) is mandated by Section 205 of the Constitution [11] to prevent, combat and investigate crimes, including corruption. However, among South Africa's public, there is little faith in the SAPS's ability or willingness to do so. According to the South African Social Attitudes Survey-a nationally representative household level survey of citizens' attitudes-two-thirds of South Africans believe that the most corrupt government officials in South Africa are located in the national police service (as reported by [12]).
Perceptions of corruption in the police have likely been shaped by the many highprofile cases of corruption right at the top of SAPS. For example, in June 2001, the National Head of Organised Crime, Assistant National Commissioner Albert Eksteen, was arrested and faced more than 100 charges related to fraud [13]. A year earlier, KwaZulu-Natal Provincial Head of Organised Crime, Piet Meyer, was charged with bribes and protecting criminal syndicates ( [14]: 5). Former Head of Crime Intelligence Richard Mdluli was also arrested on corruption and murder charges; though he was not eventually convicted, he was relieved of his duties after six years on suspension, throughout which he received his salary [15].
In terms of lower ranking officers, SAPS annual reports show that relatively few police officers are ever arrested for corruption. From 2013/14 to 2015/16, SAPS annual reports show only two dozen police officials arrested for engaging in corruption and registered only eight corruption related criminal convictions (IPID [21] 16). Corruption Watch, a South African anticorruption advocacy organisation, receives reports of corruption from citizens across the country, which it collects in person, or more frequently through telephone, e-mail, regular mail and social media platforms. According to its 2017 annual report, Corruption Watch received 2700 reports of police corruption between January 1st and June 30th 2017. Reports of police corruption accounted for the second highest number of corruption reports received, while bribery was the leading type of corruption reported among the police (see Corruption Watch Annual Report [24]). If police officials can engage in corrupt activities with impunity, low arrest rates will inevitably coexist with high levels of corruption. According to Corruption Watch's report, this seems to be exactly the case.
All of this makes the South Africa police an unlikely sector to emerge as a potential 'positive outlier' of bribery reduction; yet, as the paper shows, this is the case. The term 'positive outlier' is used here to describe an unexpected reduction in bribery in a specific sector within a country relative to all other sectors. The paper looks to contribute to a growing literature that examines how different types of sub-national developmental progress can occur in challenging national governance environments [25][26][27][28][29][30][31][32]. It was through the use of a novel case identification methodology that the SAPS bribery reduction case was unearthed, despite being previously unrecognised in the literature. The methodology employed is unique to the field in its ability to identify cases of 'hidden' developmental progress that have not been documented before.
The police sector, at a national level, was initially identified as a potential positive outlier because, according to data from the GCB, the bribery rate to the police had reduced dramatically from 2011 to 2015, far more than in other sectors the survey asked about. However, it was survey data from Afrobarometer, disaggregated by province, that illuminated the curious case of police related bribery reduction in Limpopo province. The provincial data showed that national figures of police related bribery were likely disguising an especially impressive provincial story. The police related bribery rate had not fallen in all provinces of the country, but it had most dramatically declined in Limpopo. Specifically, Limpopo's provincial police related bribery rate reduced by almost 15% from when the nation was surveyed by Afrobarometer in 2011 to 2015, while on average, the police related bribery rate for the other eight provinces reduced by less than 4%. The paper therefore focuses on explaining why it is that police related bribery likely reduced in Limpopo from 2011 to 2015.
The research into why police bribery reduced in Limpopo highlights the role of an unprecedentedif unrelated -national government anticorruption intervention into the province. From 2011 to 2015, the national government intervened in the province by effectively taking control of five influential provincial departments, none of which were directly related to the police sector. As part of this intervention, many high-profile corruption investigations were conducted, though not directly targeted at the police. This points to the effect on police bribery as a potential 'benign side effect' rather that a direct intervention effect. As Perri 6 ([33]: 53), describes: 'Benign side effects are relatively common. Not infrequently, governmentsor at least, lucky onesundertake a policy for one reason and find that its implementation turns out to be useful for other goals'.
The research applies a 'sectored' lens to understand how it is that an ostensibly unrelated anticorruption intervention into the province influenced the behaviour of the police in Limpopo and yet did not lead to similar improvements in the sectors that were directly targeted. The lens illuminates the various police sector-specific characteristics-like a high degree of discretion in police work, police solidarity and the regular engagement police have with criminals-that can make fighting entrenched corruption within the sector so difficult, and yetin this caseworked to contribute to the bribery reduction. The research finds that in light of the intervention, the police in Limpopo likely perceived that the national government had taken effective 'command and control' over anticorruption, and a new fear of being held accountable disrupted bribery patterns that were previously relatively commonplace and enabled. The research therefore suggests that bribery reduction in this case was an unanticipated consequence of this unprecedented high-level intervention. While the long-term sustainability of the impact of this unusual intervention is likely questionable, the case teaches a wider lesson that a disruptive event can counteract sector-specific factors that enable entrenched patterns of corruption, in ways in which may or may not always be anticipated, and that this can happen more quickly than the field generally anticipates.
Methodology used to identify bribery reduction in Limpopo as a potential 'positive outlier' The research involved executing a novel three-stage methodology that aims to identify sectors within a country where bribery has reduced far more than what could be expected, given the bribery patterns observed in other sectors within the same country. Most of the studies that aim to examine exceptional developmental progress identify cases based on their reputations for being success stories [25, 26, 28-31, 34, 35]. Instead, the methodology used promises to identify impressive bribery reduction in sectors that may or may not have been previously identified before [36]. If the sampling strategy followed the reputational sampling path as well, a reduction in police related bribery in South Africa would likely never have emerged, given the sector's abymsal reputation.
The three stages involve: 1) statistical analyses of sectoral bribery rates and statistical identification to identify potential positive outliers; 2) desk research used to vet statistically identified cases; and 3) qualitative fieldwork that aims to uncover and examine how it is that bribery reduced in the cases scrutinised. As discussed below, the first stage of the research highlighted the fact that bribery had likely significantly reduced in the police sector across South Africa, and the second stage flagged the possible reduction of police related bribery in Limpopo, specifically.

Statistical analyses
The statistical analyses that led to the identification of South Africa's police sector as a potential positive outlier used simple regression analyses of sector-specific bribery rates in over 100 countries, constructed from Transparency International's GCB [37]. Comparatively, the GCB dataset contains the largest geographic and temporal reach of individuals' responses to questions probing people's experiences with bribery, across multiple sectors [38]. In its latest wave (2015-2017), 162,136 adults were surveyed in 119 countries/territories [7]. The results of the analyses identified several statistically significant outlying cases, country-sectors that experienced a statistically unexpected change in bribery given bribery patterns associated with other sectors in the same country over the same period of time (see [36] for the full list).
South Africa's police sector was one of 18 potential positive outliers that were identified from this stage of the methodology. According to the GCB, over a quarter of all South Africans had paid a bribe to the police in 2013, but by 2015 the police bribery rate was less than 2%. While the GCB records that the bribery rates for all other sectors in the country over the same time also reduced, other sectors' bribery rates reduced on average only 8 percentage points. Given the other sectors' more modest reduction in bribery, the 25.5% reduction in the bribery rate to the police was flagged by the analyses as being extraordinarily exceptional; analysis estimated that there was less than 5% chance of bribery to the police in South Africa reducing to the extent that the GCB recorded.

Vetting the case
The second stage of the identification methodology involved vetting the case using desk research, and it was in this stage that the dramatic reduction in police related bribery in Limpopo, specifically, was uncovered. The aim of the second stage in the methodology is to assess whether the bribery reduction recorded statistically could be proven to be inaccurate with additional information. This step is important, because statistical outliers are sometimes a product of errors in the underlying quantitative data. Three activities were carried out to vet the case.
First, GCB data on bribery in South Africa was compared with data from Afrobarometer. Like the GCB, Afrobarometer routinely asks a nationally representative sample of South Africans whether they have paid a bribe to the police. Afrobarometer data -which was conducted independently from Afrobarometer in 2013 and then in partnership with Afrobarometer for its latest wave (2015-2017) -also showed an impressive reduction in bribery for the sector [39]. A comparison between Afrobarometer's 2011 and 2015 survey waves suggests that bribery to the police had fallen 6.1% in South Africa (from 7.3% down to 1.2%). Though the reduction Afrobarometer highlighted is not nearly as dramatic as what the GCB documented, the trend found in Afrobarometer data supports the idea that bribery had reduced significantly in the sector over a similar time frame.
Afrobarometer's dataset also proved to be useful for another reason. Unlike the GCB, Afrobarometer records the province that each South African respondent resides in, and this information allowed the research to scrutinise to what extent the reduction in police related bribery, documented by Afrobarometer, was concentrated provincially. Comparisons across provinces showed that the Limpopo province experienced the most dramatic reduction in police bribery, by far. As noted earlier, according to Afrobarometer, Limpopo's provincial police related bribery rate reduced by almost 15% from when the nation was surveyed in 2011 to 2015, while on average, the police related bribery rate for the other eight provinces reduced by only 3.9%. Notably, in North West province, according to Afrobarometer, police related bribery rate increased by 2% over the same period. The scrutiny of the provincial level data, therefore, highlighted the curious case of police related bribery reduction in Limpopo specifically.
In addition to vetting the case by examining Afrobarometer data, newspaper articles, journal articles and grey literature were also reviewed, with no conclusive evidence found to suggest that bribery had not reduced in the sector, nationally, or in Limpopo, specifically. However, there was also little evidence in these sources to support the notion that bribery had reduced in the sector. No media coverage of the bribery reduction as documented in the GCB and Afrobarometer was uncovered, and the South African chapter of Transparency International did not highlight the reduction. No existing research on the topic was found at any of South Africa's excellent centres for crime and policing research.
Finally, experts familiar with South Africa's police sector were consulted. Experts were identified using a snowball sampling technique that started with contacts drawn from the research teams' personal networks of colleagues and academic contacts, as well as emails to scholars that have published academic research on corruption in South Africa's police. The experts were academic criminologists, anticorruption NGO staff and practitioners in the South African Police Service (SAPS). A few of the experts offered tentative hypotheses for why bribery may have reduced across the nation, but none were fully convinced that the data reflected reality. Instead, a majority of those consulted expressed a deep scepticism that the bribery rate had reduced. Moreover, none of the experts initially consulted, at this stage, offered a hypothesis specific to Limpopo that could help explain the acute reduction in police bribery there. It is important to note that, conversely, none of these experts were able to present compelling evidence that suggested that the reduction recorded in either the GCB or Afrobarometer was inaccurate.
Despite the reservations expressed by experts and NGOs, and the lack of coverage of the bribery reduction trend in academic and news sources, investigation of the case continued. This is because the identification methodology promises to identify previously unrecognised cases of surprising developmental progress. As such, the research was interested in testing the identification methodology to see if the statistical data had indeed pointed to a 'hidden' reform story that could withstand more fine-grained qualitative analysis.

Field work
Six weeks of in-country fieldwork in South Africa was conducted. The identification of respondents again partially relied on the contacts of the extended research team and through snowball sampling initiated in stage two. The process uncovered a large pool of expert respondents willing to share their expertise and personal experiences. Once in-country, participants provided suggestions and introductions to sector practitioners. The research also engaged academics and researchers, relevant government representatives and journalists, as well as members of the South African Police Service (SAPS) itself.
The fieldwork was conducted in Johannesburg, Pretoria, Cape Town, Polokwane (the capital of Limpopo province), and in several small towns (which will not be named to preserve the anonymity of respondents) in the province of Limpopo. A total of 35 key interviewees including senior police leadership, station commanders, mid-ranking police, constables and members of the specialised policing unit known as the Hawks were interviewed through in-depth, semi-structured interviews, with some participants being interviewed several times. Interviews were also conducted with journalists reporting policing and crime, policing researchers, retired police, and key government officials. One member of the research team (Trevor Budhram) is also a police trainer and former police officer.

The 'Limpopo intervention'
Most recent figures suggest that Limpopo is the poorest province in South Africa; it has the lowest average household income, for example, and the highest percentage of citizens living below the poverty line [40]. The province has the smallest percent of white South Africans in the country and the largest percent of black South Africans. About 10% of South Africa's population live in Limpopo [41]. It is the northernmost province in the country, bordering Botswana, Mozambique and Zimbabwe. These bordersand the Limpopo River that links the countries togetherhas been compared to the Mediterranean in terms of refugees, asylum seekers and illegal migration [42,43]. Though only the fifth highest province in the country with regard to overall crime rates [44], rates of crime per capita in Limpopo's largest city, Polokwane, are high: 13,794 crimes recorded in 2017 with an estimated population of 797,127, compared to Johannesburg, with 28,308 crimes recorded for its much larger population of 4,949,347 [45,46]. On the face of it, it is not exactly where one would expect to find anticorruption success.
Afrobarometer's results from their 5th and 6th waves highlighted the fact that police related bribery had reduced most dramatically in Limpopo. Afrobarometer's data also showed that perceptions of corruption in the police had fallen in Limpopo from 2011 to 2015, suggesting that a wider shift had taken place in the police-community relationship in the province. In 2011, according to Afrobarometer, more than half of those surveyed in Limpopo said that 'most or all police are corrupt', and in 2015, that figure had nearly halved (28%). In fact, by 2015, people in Limpopo were the least likely in the country to judge their police as being mostly or all corrupt. In 2015, 43% of South Africans outside of Limpopo judged their police as being mostly or all corrupt, with people from Gauteng province being the least trusting; nearly 70% of respondents from Gauteng thought of their police as mostly or all corrupt. What may have impacted police bribery and community-police dynamics in Limpopo during this time frame?
The province was pushed into the national spotlight in December 2011, when South Africa's National Cabinet surprised many by announcing that they would invoke Section 100 (1) (b) of the Constitution of the Republic of South Africa of 1996 to put five key Limpopo provincial departments under national government administration [47]. The move effectively meant that the national Executive Branch would take over the executive functions of the province that are normally performed by members of the provincial Executive Branch ( [11]: 57). Section 100 (1) (b) mandates that the national government intervene to fulfil a provincial government's constitutional or legislative obligations when it is clear that the provincial government will not be able or is not willing to do so [48]. The national government argued that it had no choice but to intervene because the province was technically bankrupt; it claimed that it was almost R2.7 billion in the red when Section 100 (1) (b) was invoked [47].
The departments put under national administration were the provincial Treasury, Education, Transport and Roads, Health and Public Works, which are the biggest drivers of service delivery in the province, together consuming the largest percentage of the provincial budget ( [49]:8). While putting spheres of government under administration is not a new phenomenon in South Africa, invoking Section 100 (1) (b) in Limpopo attracted significant media attention and some public outcry; it was the first time that a province had had as many as five provincial departments put under administration at the same time [47].
The national government cited the following issues found in the province as reasons for the intervention: supply chain management violations (tenders awarded without bidding processes being followed), poor budget practices, poor asset management, unauthorised expenditures (including over-expenditure, or spending far more than what was budgeted or legislated for) and irregular expenditures (expenditures not at all related to any applicable legislation) [48]. In the 2011/12 financial year, the National Treasury reported that the provincial government had accumulated unauthorised expenditures of over R2.5 billion, and by October of 2011 it had exhausted its overdrafts with both the commercial and reserve banks (PMG, [50]).
The intervention proved to be incredibly disruptive for the province. Not only had the national government descended upon the province to straighten up provincial offices, but the intervention also required South Africa's Directorate of Priority Crime Investigation (DPCI, also known as the 'Hawks')-whose mandate it is to target organised crime, economic crime, corruption and other serious crimes-and the Anticorruption Task Team (ACTT)-an executive controlled institution, tasked with dealing with serious cases of corruption-to probe allegations of corruption that were uncovered during the investigations that led to the intervention [51]. Forensic accounting and other tools were used to demonstrate the poor and sometimes corrupt financial practices rife in the institutions [51].
By the end of the intervention, the ACTT alone filed 43 corruption cases that were directly related to the Limpopo Intervention [52]. However, the total shake up was likely wider. In his 2014 State of the Province Address, the Premier of the province, Chupu Mathabatha, claimed that nearly 300 people had been either charged or faced corruption-related charges since start of the intervention ([53]:22). Official figures may also underestimate the extent of the overhaul as many officials resigned before the conclusion of their respective disciplinary processes [51].
Aside from the anticorruption drive, the intervention also saw the establishment of an electronic accounting system, a review and update of the provincial supply chain management policy and the stabilisation of the budget [54]. By many measures, the intervention appeared to be successful. The Premier of the province, Matabatha, thanked President Zuma for the intervention for Bsalvaging Limpopo from a next to collapse of corporate governance.^([53]:24). By the end of January 2014, the province was in a positive cash position of R4.4 billion [47]. Out of the red, pressure was put on the national government to exit the province. The national intervention into Limpopo ended in early 2015, as full executive powers were returned to the Premier of the province and to departmental heads [55].
Some suspicions have been voiced about the Government's motivations. The Government claimed that Section 100 (1) (b) was invoked because of the province's poor financial position and poor management, with significant evidence to support this. Some, however, claimed that the move was politically motivated rather than a genuine response to the province's financial crisis, including a prominent politician, Cassel Mathale, the province's Premier and the ANC Limpopo Chairman. Mathale and Julius Malema, the former ANC Youth League President who was from Limpopo, fell out of favour with President Zuma following their public criticism of him and their public show of support for Kgalema Motlanthe, then Deputy President and a challenger to the President's rule over the ANC [49,56]. They argued that the move to intervene in the province was retaliatory and was designed to politically discredit their position in the party. Of course, the two motivations are not necessarily mutually exclusive: it is entirely possible that the Government was both concerned about the weak financial position of the province and also sought to use the intervention to politically harm its detractors.

How the Limpopo intervention influenced police bribery
The size of the national government's intervention into Limpopo was unprecedented [47]. It resulted in many corruption investigations, led by national anticorruption bodies, into relatively high-level corruption. However, the intervention itself did not focus on the South African Police Service (SAPS) in the province. Even if there were to have been interest in doing so, Section 100 (1) (b) would unlikely have been appropriate to use for such a matter. This is because Section 100 (1) (b) provides for an intervention specific to a province and not into a national agency, and SAPS is a national agency. How could this intervention-focused on relatively high-level corruption and general financial mismanagement-have influenced the behaviour of the ordinary police officers who interact with the public in the province?
Interviews with several officials involved with and deeply familiar with the intervention suggested that the intervention likely reduced the willingness of the police in Limpopo to ask for bribes. This is because the highly visible and disruptive intervention-albeit focused on other forms of largely high-level corruption-likely raised awareness of a general push to fight corruption in the province. The intervention also received a high level of coverage in the media, which could have also reduced the willingness of the police to engage in bribery. In the same way that crime has become a 'spectacle' in South Africa, where politics and perception often collide, as argued by Super [57], the Limpopo Intervention's prominence brought politics, corruption and crime together in a highly visible way. In the words of one of the police officials we interviewed, who worked on the intervention, there was …a lot of hype around it, there was a lot of initiatives from the Special Investigations Unit, from the Police Joint Chiefs on that, so obviously [the intervention] created some awareness, so people became a bit more intent that they wouldn't accept or solicit bribes…there was a fear factor from Section 100. There were a lot of activities and clampdowns that were localised. They weren't necessarily orchestrated, but the anticorruption vibe was there. It's a perception that I might be caught (interview, August, 2017).
The disruption of the intervention was likely felt most in Polokwane, the province's capital, where the province's departmental offices are based. About 12% of the people in Limpopo reside in Polokwane's municipal district. Despite being the capital, Polokwane was described by one interviewee as having an almost 'small town' feel, where most people knew each other or each other's families. In the words of one police official involved in the intervention, Polokwane was a Bplace where you still greet the neighbours…whereas here [Pretoria] you don't even know the neighbours^(interview, August, 2017).
During the intervention, an influx of administrators and investigators from Pretoria descended upon the provincial capital. According to a Senior Police Official, who travelled in and out of Limpopo during the intervention: So there was a lot of senior people from government traveling to Polokwane on a daily basis because they were mostly situated [in Pretoria]… So in town, or in the city, there was a lot of strange faces moving around, in government places. And then besides that, the Hawks also started investigation there. And even though we don't wear uniforms or drive marked cars, in a small town like that, you notice strange faces (interview, August, 2017).
Moreover, police based in Polokwane would have had many opportunities to interact with many of the ACTT and DPCI investigators and prosecutors. The city boasts only one police station. A joint operation centre was set up from designated space provided at the police station in SAPS. According to one Senior Police Officer involved in the intervention, B[Polakwane is] the main name or station under which we register cases...So there was a lot of us going in and out of there [the police station] as well( interview, August, 2017).
Indeed, a 2015 Parliament Committee Report shows that all of the ACTT's corruption cases associated with the intervention were processed in Polokwane [52].
In addition to an anticorruption drive being highly visible, there may have been a generally low level of knowledge about the mandate of the intervention itself. Some of our interviewees observed that police officers in Limpopo may not have understood the scope of Section 100. One senior public servant who had been involved in the intervention noted, You must understand the people in South Africa, they don't understand how the government is arranged, including those who are working in government. The police will think they are next on the line and only much later find that they are not as they're not part of the provincial administration. They wouldn't realize they aren't part of the provincial government because they are stationed in a province. They think that they are part of the province. They don't see a difference between a national government, the provincial government and a local government, or a municipality -they don't see the difference. It's just one government to them (interview, August, 2017). This view was echoed by a number of interviewees. According to a senior police officer engaged in the intervention, local police were largely unaware of what the DPCI and ACTT were investigating; this lack of understanding is said to have meant that they were fearful that they too might be affected (Senior Police Official, August, 2017). This was partly because the ACTT and the DPCI based themselves in offices that were closed off from regular SAPS in the province. The same interviewee suggested that because of the uncertainty, local police may have resorted to unusual tactics to understand the nature of the investigations; they reported that the local police installed a monitoring device in the house they were renting in Limpopo during the intervention. They suspected that this was done to gain insights into the investigations.
Finally, it was also suggested that traffic police in the province may have been more reluctant to pull over and request bribes from drivers during the intervention. In the words of one of the officials involved in the intervention, Bbecause we don't drive marked cars and wear uniform, there's now the risk that [they] might pull over a Hawks officer or SIU member of a senior government official^(Senior Police Official, August, 2017). Fearful of unknowingly pulling over a high-ranking officer and/or someone investigating corruption, it was suggested that the traffic police in the province may have pulled fewer cars over during the intervention, and of the cars that were pulled over, the police would have been less likely to request a bribe because they would have been fearful that they would be doing so to an officer working on the intervention.
This particular dynamic would have likely influenced officers patrolling the N1, the province's main highway; it directly links Polokwane to Pretoria and is the highway national agents would have frequented most in their commute to Polokwane. Arguably, traffic police in South Africa have the greatest opportunities, of all police, to request bribes. They can find cause to charge drivers for infractions, can use pulling over cars in low traffic areas to gain privacy as they request a bribe, and can use legitimate official fines as leverage to ask for a bribe or as an incentive for drivers to offer bribes. If the intervention indirectly deterred traffic police from asking for bribes, the total provincial police related bribery rate may have been significantly affected. This, in turn, one expects, would also affect public perceptions of police bribery.
It is also worth mentioning, however, that rumours were circulating across Limpopo's mainly rural police stations that traps had been set to catch police officials involved in bribery [58]. These were fuelled by evidence that the national office of the SAPS had conducted its own intelligence driven operations in the province. During the 2011 to 2015 period nineteen police officers were arrested for corruption in the Limpopo district, all arrests that coincided with but were unconnected to the unrelated national intervention [58].
Taken together, the spectacle and visibility of a national level intervention, the influx of national agents investigating and prosecuting corruption in the province, a lack of awareness about the mandate of the national intervention, as well as the SAPS' own separate investigations, are likely to have caused many police officers in the province to be extraordinarily cautious. Perceptions of concentrated anticorruption activity and fear of being caught out by the unprecedented level of the intervention may have led many police officers to act with caution and to avoid engaging in bribery. Clearly, the change in police behaviour was noticeable enough to shift public perceptions in an almost unparalleled way.

Sector characteristics that helped make bribery reduction in the Limpopo SAPS a positive unanticipated consequence
Our research suggests that in certain cases where corruption is difficult to control, a sufficiently disruptive event can push against such corruption and its enabling factors. One of the most interesting things about the Limpopo intervention, though, is that the 'disruptive event' was focused on five sectors where the data shows no significant improvements in bribery reduction despite being the direct targets of a large-scale anticorruption intervention. So why did bribery reduce so dramatically in the police, despite not being the subject of the intervention?
In his recent assessment of the state of research on corruption, Heywood [59,60] reports that much of corruption research has been focused on the nation state as a unit of analysis and, in doing so, has largely failed to differentiate between different types of corruption, different localities of corruption and the differences in how corruption behaves across sectors. With respect to the later of these observations, Heywood [59] calls for a more Bmeso-level approach^which can draw attention to corruption's sector specific characteristics, so as to provide better understanding about the Bmodalities of corruption and corruption related risks in key areas.^Heywood [59] argues that although there have been some attempts to explore the link between particular sectors and corruption (notably, for example, public administration, the energy sector, the judiciary, defence and security and so forth), these have often taken the form of seeking to account for the extent or overall level of corruption within a given polity. What is needed now is a more detailed understanding of how and why corruption takes place within these sectors: what it looks like in practice, what particular characteristics it has, and how we can better identify risks.
To this end, Mcloughlin and Batley's [61,62] 'sector characteristics' approach argues that the specific characteristics of a service shape a range of issues surrounding the provision of the service, accountability with respect to service quality, the power dynamics between policy makers and other actors in relationship to the service, and the extent to which citizens can make demands on how services are delivered. They argue that the Bmeasurability, transaction-intensity and level of discretion involved in performing different functions influences the degree to which policymakers and bureaucrats can control the behaviour of delivery organisations^( [62]: v).
In addition to the unique role they play in terms of anticorruption, the police are unique in a number of ways as a sector. For example, they operate with a clear command and control hierarchy, engage regularly with criminals as part of their dayto-day work, and police work, by its nature, can be extremely dangerous. A sectored lens helps to shed light on why-unlike in other sectors not included in the Intervention-police bribery reduced as a positive unanticipated consequence of the Limpopo intervention.
Like the military, the police operate with a clear command and control hierarchy thatin theorystarts at the top, with a National Commissioner, and finishes up with junior desk officers. As a General in the SAPS explained, regarding his experience in exerting command and control over several stations, BAt the end of the day, it boils down to command and control, whatever you're trying to do…You take stock of people, supplies, and close the loopholes. Corruption has many forms…So the commander has to be vigilant. They must be hands on and make sure that you as an officer they control everybody( interview, August, 2017).
An academic and expert on South Africa's police also noted the critical nature of strong leadership, BIt's all about command and control. If you have a good commander who is able to pull the command structure together at the police stationand that includes regular inspections, simple stuff…this creates conditions where it is more difficult to be corrupt^(interview, August, 2017).
For most police work, station commanders, with the level of control and authority they can exercise within their stations, are incredibly important. Specific environments for station commanders shape their interests and incentives to try to control corruption. If corruption is widespread at lower levels, it is unlikely (if not impossible) that it is occurring without senior officers knowing about it. Such widespread corruption would not, however, necessarily imply that station commanders and other senior police approve of such corruption. Gareth Newham, a leading expert on South Africa's police, explained, There are many station commanders who wouldn't take bribes and are deeply concerned with the fact there is bribery going on at the station, but they don't have the support from their juniors to deal with it. If they start taking action against these guys, they are threatened. Cops have been killed. While I was in the [redacted] police station, a detective investigating a colleague involved in organised crime was shot dead. And according to the research I did in the early 2000s, every internal investigation officer I worked with had been intimidated, threatened, their cars vandalised, their offices vandalised, getting threats on their cell phones. And that was investigating the relatively petty stuff sometimes. So, if you're a station commander who wants to fix corruption, you kind of have a death wish (interview, August, 2017).
The responsibility that senior officers hold for the activities of their juniors can create a disincentive for even well-meaning senior officers to report or prosecute corruption among their ranks. In exposing the corruption that happens 'under their noses' they also admit their own inability to maintain effective command and control.
Police work also uniquely involves, by necessity, close engagement with criminals and criminal networks, through investigations, informants, arrests and so on. This can both provide the police with unique opportunities to engage in corrupt activities and makes police much more vulnerable to claims of abuse of power. Vigneswaran and Hornberger [63] write about the Binformality and ethical ambiguity of real police work^; in order to do effective police work, officers arguably need Bimmense discretionary power^( [8]: 3) relative to other sectors, as they need to make judgment calls on a regular basis about whether or not following the rules to the letter is the best decision to make. According to Newham, BPolicing is a very particular kind of profession, and it comes with a lot more risk than other kinds of professionsteachers, doctors etc. Because of the powers they have, they also have great discretion. When they're out on the streets they are not being directly monitored or supervised, and so [they] face very low levels of accountability^(interview, August, 2017).
Neither discretion nor informality are, in and of themselves, a bad thing, but as police work also involves close interactions with criminals, the combination of these factors creates a challenging environment for ethical behaviour. Indeed, Klockers et al. ( [64]: 1) describe police integrity from a negative perspective: Bthe normative inclination among police to resist the temptations to abuse the rights and privileges of their office^.
Another unique characteristic of police work is that it is inherently dangerous and high risk. This comes through the interviews as well: police respondents expressed frustration that the public perceive them to all be corrupt but do not adequately appreciate that every day they go to work they literally risk life and limb. They flag that this is the case for the police everywhere, but that police work is especially dangerous in South Africa. One senior police officer explained, We read about the corrupt policeman or the policemen who are committing robbery. That sells newspapers. But on a daily basis there's good work being done by policemen in South Africa. Really, to put on a uniform and go rescue the life of people you don't know, in a situation where you don't know what's coming… You're called to a domestic violence incident, and when you get there, it's a house robbery, and they start shooting at you. These guys are putting their lives on the line (interview, August, 2017).
This particularly violent and dangerous working environment helps to create another fairly unique sector characteristic: group solidarity among the police can be especially strong, with suspicion of outsiders and codes of silence being prevalent ([8]: 3; [65]). Crank [65] summarizes the 'dangerous-solidarity hypothesis', which suggests that Bthe sheer danger of police work, like combat, encourages strong loyalties, an 'all for one and one for all' sense of camaraderie, and a military sense of combat-readiness and general spiritedness.Ŝ olidarity can also translate into a perceived sense of insularity for the sector. This is because solidarity among police is also commonly sourced from a shared sense that police authority is wrongfully challenged by several societal and governmental groups, including citizens, the courts and the press ([65]: 198). A professor of policing described the seemingly universal sense that the police are an insular sector, BWithin all police services in the world you get this culture of a closed unit, a sense that we are separate from the public. It becomes a culture, and if they think they can get away with it, if they can protect each other, they will get along irrespective of what is being said in the media. They will take the opportunity. Police officers, by virtue of the fact that they are enforcement officers, do have a lot of power^(interview, August, 2017).
Solidarity can be expressed in positive events, from protecting other officers from harm to attending funerals of colleagues ( [65]: 197). However, it can also make anticorruption work particularly difficult. Lester and Brink [66], for example, found that in certain contexts, police solidarity may lead to greater tolerance for misbehaviour by fellow police officers. When police are primarily loyal to their colleagues, and their colleagues engage in corruption, the duty the police have to enforce anticorruption laws can become compromised. Moreover, given how tight-knit social relationships between police officers can become, there can be real social consequences for exposing corruption committed by 'one of their own.' Newham [67], for example, talks about the importance of Ba fear of both personal and professional retaliation^for any anticorruption activity in the SAPS. Retaliation can be threatened against one's family, with spouses also isolated from social relationships. The cost of stepping outside the social networks and violating the codes around solidarity and silence can be fierce.
When viewed through a sectored lens, the challenges to fighting corruption and in reducing bribery, specifically among the SAPS, can seem almost insurmountable. The sector has suffered from several high-profile corruption scandals, and ordinary South Africans, who most likely interact with the sector's rank and file, rate the sector to be one of South Africa's most corrupt. In combination, the seemingly systemic corruption within the sector and the more generic sector specific factors discussed here-that police enjoy a high degree of discretion, engage regularly with criminals, and that there is often a sense of insularity and solidarity among the police, which can express itself in the police 'protecting their own'-highlight why fighting police corruption -in the South African setting and elsewhere -is so incredibly difficult. The sector specific factors discussed imply that the SAPS may be especially tempted to engage in bribery. Their work is difficult to monitor, and there may be a lack of will to enforce anti-bribery mandates within the sector because of the strong loyalties actors are prone to having to each other in the police. As each of these factors represent challenges to accountability within the sector, they also make clear why it is that an effective 'command and control' can be difficult to achieve within a station, and beyond the station environment, further up the sector's hierarchy.
The Limpopo intervention's likely unanticipated impact on police behaviour suggests, however, that seemingly entrenched bribery patterns can be disrupted, and the sectored lens helps explain why. While the Limpopo intervention did not alter the official SAPS command and control structures in the province, the intervention seems to have likely introduced a perception among the local police that national level actors had effectively taken command over corruption concerns in the province and were making visible steps to effectively control corruption in the province. Solidarity may have meant that officers were restrained in their own behaviour so as not to negatively impact upon their fellow officers. In such a fearful environment, which officer would want to endanger the others in their team by getting caught asking for a bribe? The visibility of the Hawks and other anticorruption actors, and the multiple corruption cases being processed in the Polokwane police station, as well as the uncertainty among the police about the intervention's mandate, likely all worked to produce this perception of effective 'command and control'. Any perceived impunity to engage in bribery may have dissipated for a period of time because of the disruption that the intervention caused among the police. Indeed, the case suggests thatdespite the sector's unique challenges with regard to bribery and corruptionit was surprisingly easy to disrupt these practices. Whether the impacts on corruption caused by such an unusual disruption are sustained over a longer period of time, however, is questionable and is worthy of future research.

Conclusion
Police related bribery reduction in Limpopo occurred at the same time that the national government led an unprecedented high level anticorruption intervention into several provincial departments. The research suggests that the police in Limpopo may have been especially reluctant to engage in bribery during the intervention because of an uncertainty about whether the police were also under investigation for corruption and because of the heightened salience of anticorruption action that took place.
The drivers of bribery reduction in this case were likely indirect and unanticipated, a 'benign side effect' of a separate anticorruption intervention. The intervention in Limpopo targeted high-level corruption problems in the province and was not targeted at local police; yet police likely feared that they were under greater scrutiny from the perceived efficient degree of 'command and control' national level agents were exerting during the intervention over corruption issues. The sustainability of the impact on bribery reduction that the intervention in Limpopo had is questionable as a result. The intervention was unprecedented and has since resolved. Only by looking again at the GCB/Afrobarometer data in the future would one be able to say for sure whether or not the reduction in bribery was sustained; however, because the reduction was the result of a temporary, unrelated intervention that did nothing to alter the underlying structural problems that work as enabling factors, a long-lasting impact on policerelated bribery patterns in the province is unlikely. This case thus suggests that certain types of disruption may work to reduce bribery patterns but may do so only across a relatively short time frame. It also suggests that for longer impact, disruption strategies likely need to be targeted to their sector, to be continuously inventive and reinventive and/or driven by strong leadership able and committed to continually disrupting corruption patterns.
The research also highlights the ways in which new insights can be garnered from an approach to anticorruption research that focuses on specific forms of corruption in a specific place, sector and time. As Heywood [59] notes, corruption has far too often been researched at the nation-state level; however Bif we are to address corruption for the purposes of effective policy, we need to disaggregate into different types, as well as between different levels and locations in which it occurs.^This call to researchers builds on what is common knowledge in the field, which is that different types of corruption likely have different drivers and consequences, anticipated or otherwise. Corruption occurring in an urban area, for example, can look different to corruption occurring in a largely rural province, much less the differences in corruption patterns that seemingly occur across national borders.
To this end, this paper has focused on a specific type of corruption: bribery, in a specific location: Limpopo, and has applied a sectored lens to help understand why the Limpopo intervention likely had a welcome, if unanticipated, impact on the behaviour of local police. A sectored lens proves to be important as it helps to highlight Bthe pathologies of corruption risks in specific fields^([60]: 11). Bribery in the police is understandably difficult to control, and especially so in a country where bribery, more generally, is relatively commonplace. This is because there is a high degree of discretion in police work; the citizens they engage with who have committed infractions are more prone to want to pay a bribe to avoid a harsher punishment; and police solidarity and insularity may make officers reluctant to investigate or turn in 'one of their own' for engaging in bribery. The research suggests, however, that a sufficiently significant disruptive event-in this case, in the form of the Limpopo interventioncan push back against these specific enabling factors that are likely ever-present in police sectors plagued by seemingly entrenched patterns of corruption, especially if able to use the sector's own unique characteristics to support this. In terms of police corruption, while there are other examples in the literature where corruption has been reduced in the sector, these tend to be standalone studies (e.g., [68][69][70], among others). Indeed, there is a paucity of comparative research on the sector, which makes it difficult to assess whether the disruptive effect seen in Limpopo is unusual and how it might compare to cases elsewhere.
The research has not uncovered other cases where bribery has reduced as an unintended consequence of another governance intervention. In writing about antipoverty programmes, Klitgaard tells us, BIn policy circles, we have become perhaps too accustomed to think that all unintended results must be perverse; in science we are much more tolerant, and indeed we are fond of thinking about serendipity, good things that happen as the unanticipated outcomes of trying our hardest^( [71]: 1963). This is important because it reminds us that unintended consequences can be welcome as well as unwelcome. However, the anticorruption field gives little-if any-attention to studying these welcome unintended consequences. Without knowing much about when and why such benign side effects may have occurred in the past, researchers and practitioner are unable to begin to predict when they may occur again or what factors could be put in place to facilitate them. This is an important potential area for future research, if such cases can first be identified.
It was only through using a novel three stage identification methodology that the research was able to locate the unexpected and previously unrecognised impressive case of police related bribery reduction in Limpopo. By explicitly scrutinising sectorspecific bribery patterns the methodology presents one way forward for researchers to explicitly adopt the Bmeso-level approach^to researching corruption that Heywood describes the field as needing ( [59]:42). However, the methodology contributes more than just this. By not relying on reputational assessments to identify progress, the methodology was able to identify a positive trend that was, for all intents and purposes, 'hidden' or unrecognised. The case, therefore, also demonstrates that the methodology can be useful for identifying progress where it may be least expected to occur. While the field has learned a lot from well-recognised and researched cases of corruption reduction, it is perhaps from those unexpected places-perceived to be fraught with intractable problems in the fight against corruption-that lessons about how corruption can be effectively disrupted are arguably most valuable.