The Biometric Imaginary: Bureaucratic Technopolitics in Post-Apartheid Welfare

Starting in March 2012, the South African government engaged in a massive effort of citizen registration that continued for more than a year. Nearly 19 million social welfare beneficiaries enrolled in a novel biometric identification scheme that uses fingerprints and voice recognition to authenticate social grant recipients. This article seeks to understand the meaning of biometric technology in post-apartheid South African welfare through a study of the bureaucratic and policy elite's motivation. It argues that biometric technology was conceived of and implemented as the most recent in a series of institutional, infrastructural and policy reforms that seek to deliver welfare in a standardised and objective manner. This has, at times, been driven by a false faith in technical efficacy and has involved a playing down of the differential political implications of biometric welfare identification.

protections of the Constitution, another strand of support promotes grants as a means to escape a poverty trap, allowing the poor to enter the market as entrepreneurial risk-takers. 5 In contrast, opponents tend to worry about the cost of the grants, the risk of welfare dependency and perverse incentives, such as the neglect of work. 6 Those 19 million residents -nearly 40 per cent of the country's entire populationprovided their personal details in order to receive their grants through a new payment system. For the government, the system promised a reliable means of delivering millions of rand per month. As the Minister of Social Development, Bathabile Dlamini, would explain, the goal was 'to improve the integrity of our social security system, and to eliminate all forms of fraud and corruption'. 7 Prominent civil society organisations like the Black Sash, too, 'welcomed [the] move to a biometric system'. 8 The new system aimed to offer increased convenience through the provision of a payment card accepted at a large cash distribution network established by the government's contractor, Net1 CPS. 9 The cornerstone of this new system is biometric identification, the use of technologies to recognise specific bodily features. A fingerprint would be the primary means of authenticating recipients, but in case a fingerprint scanner was unavailable, the grant recipients also provided a voice recording to serve as a back-up method of verifying identity at the time of payment. This is not the first time biometric identification has been used widely in South Africa. Both before and during apartheid, government officials were enamoured of the potential for using analogue fingerprinting to identify individuals, most commonly in schemes of racial labour control, and it was used extensively in previous welfare systems. 10 In the post-apartheid era, digital biometric technology has been central to the identification programmes of the Department of Home Affairs.
Yet rarely have these systems functioned as promised, whether for good or ill. 11 The continuing salience of biometric identification among the bureaucratic and policy elite therefore seems curious. Given the contentious politics of grants and the apartheid history of biometrics, what is striking about the case under consideration is the lack of conflict over the massive registration drive. Biometrics have not led to outrage; instead they are considered a commonsensical response to scandals within the grant programme. To explain this technological 'non-scandal', this article narrates the post-apartheid history of welfare state building that culminated in the 2012 -13 biometric registration drive. To do so, I rely primarily on an archive of government documents, civil society reports, court filings, parliamentary minutes, and nearly a dozen interviews with stakeholders. I pay particular attention to the ways in which elites -especially those within welfare bureaucracies and their interlocutors on supervisory panels and commissions -conceptualise and attach meaning to biometric identification. I explicate the reasons it is adopted by policymakers and members of government, especially those within the welfare administration, who face the twin tasks of streamlining delivery and tightening access.
I aim to explicate some of the dynamics behind a post-apartheid culture of registration and identification. 12 I argue that a relatively stable and collective understanding exists among the bureaucratic and policy elite about how to constitute the relationship between welfare state and beneficiary. In brief, this collective understanding -which I propose to call the 'biometric imaginary' -posits biometric technology as a necessary, suitable and effective means of constructing a standardised and objective welfare state.
The will to standardise has arisen as a means of uniting and improving the fragmented and weak bureaucracy inherited from the apartheid regime. South Africa has endeavoured to create equality of citizenship through imposing uniformity. This has included bureaucratic centralisation, first through the establishment of a unitary, national Department of Social Development (DSD) in 1994, and later through the creation of the South African Social Security Agency (SASSA) in 2005, as well as through technological reform, such as the linking of various databases and implementation of new payment infrastructures.
In parallel, an egalitarian ethos has militated against subjectivity in favour of rule-bound practices. In democratic South Africa, subjective discretion has been viewed with suspicion as the state has sought to create equality of citizenship. In the realm of welfare, two forms of discretion have proven particularly troubling: illicit access to grants, and bureaucratic error. A preference for objective practice has deeply influenced the social grants, with biometric identification offering an impersonal and presumably neutral means of grant administration, aligning with a parallel discourse of widespread fraudulent access to the grants.
Both standardisation and objectivity are widespread and productive -though not hegemonic -commitments that shape the understanding of biometric registration and identification. This particular confluence of the will to standardise, the pursuit of objectivity and the use of social grants has given rise to the belief that biometric identification is necessary, suitable and effective. As in Appadurai's discussions of social imaginaries, the biometric imaginary is 'a constructed landscape of collective aspirations' that serves as a 'staging ground for action'. 13 These imaginaries are productive social forces with real-world implications, but they do not always accurately correspond to reality, exhibiting varying degrees of interpretive flexibility. As I will contend, the dominant understanding of biometrics as necessary, suitable and effective is, in part, fantasy and reality.
The biometric imaginary reflects the culture of registration and identification insofar as it is a shared set of understandings with attendant practices and politics, yet it is relevant not merely at the level of ideas; it has material enablers and effects. Biometric welfare is a form of 'technopolitics', Hecht's term for the 'strategic practice of designing or using technology to constitute, embody, or enact political goals'. 14 The will to standardise and the pursuit of objectivity have driven a series of technopolitical programmes within the post-apartheid welfare sector, including institutional centralisation, technological standardisation and bureaucratic automation.
Here, 'post-apartheid' does not mark a total rupture with apartheid's bureaucratic logics, but rather signifies the redirected emphasis. For at least two decades, biometric technology has been harnessed for redistributive welfare rather than apartheid-era controls. In this, the 2012 SASSA contract with Net1 CPS stands as an unprecedented moment: a nationally centralised database of 19 million grant beneficiaries who are biometrically identified for each payment. The emulation of South Africa's system of social grants 15 and biometric methodologies 16 suggests the need for an understanding of the conceptualisation, meaning and drivers of biometric welfare in South Africa. I next turn to the drivers of standardisation and objectivity, before showing how, and with what consequences, biometrics have been used to reach those goals. Section two analyses the institutional and political history of the post-apartheid welfare state; in particular, it traces the emergence of a desire for uniform policy that led to the centralisation of welfare bureaucracies. Section three turns more directly to practice, examining how discourses of 'corrupt' bureaucrats and 'fraudulent' recipients have fuelled the adoption of biometric identification. Together, these two sections excavate the commitments underlying the biometric imaginary, while the closing sections adopt a more critical posture.

The Will to Standardise
As apartheid ended in the early 1990s, an insidious danger lurked in the governing structure of post-apartheid South Africa. Provinces displayed a high degree of variability in their processes, technologies and abilities. To a significant degree, the quasi-independent homelands lacked competent bureaucracies and were riddled with corruption. In 1984, welfare had been designated an 'own affair' within the tricameral parliament, resulting in 'the creation of costly and duplicated administrative structures, with 13 "national" and 4 provincial head offices, plus another 3 coordinating departments'. 17 Not only did the quality of service differ wildly: fundamental definitions of, say, what counted as disability meant that applicants may have qualified for a grant in one province but not in another. 18 Population registration reveals a similar story. Despite the panoptic ambitions of the apartheid government, the basic task of population registration was split between more than 'a dozen discrete yet overlapping and duplicated population registers'. 19 The system's weakness began with a lack of commitment to the registration of births: the Medical Research Council estimated that in 1994, less than one-fifth of births were registered within their first year. 20 Africans found registration and identification particularly onerous, not least because of the unclear and competing structure of 'traditional', 'homeland' and national political authorities. Access to social grants before 1995 required proof of identity, citizenship (or permanent residence) and age. A range of documents -including national ID books, reference books, passports, homeland travel permits or 'farm labourer' stamps -were variously used, but so were less formal systems, such as memories of 'the Great War' or 'the year of apartheid' to establish age. 21 Put otherwise, methods of registration were implemented in uneven and uncoordinated ways.
During the transition, though, the new government recognised the system as woefully insufficient to meet the dual goals of rectifying racial discrimination and the accompanying social and economic inequality. As the government began in earnest to address the situation, it convened a series of national commissions of inquiry. Beginning in 1992, no less than five more major national commissions were convened. These commissions brought together an energetic coalition of academics, activists and state officials who saw a reformed and expanded welfare programme as a crucial project.
Their work reflected the solidifying politics of post-apartheid welfare, concerned primarily with improving the delivery of social protection while reducing fraud and corruption in the system. Although alternatives (such as significant delegation to the provinces) were at least considered, these initiatives rather quickly coalesced around an understanding of the problem as fragmentation that impeded service delivery and created the opportunity for mischief. Increased uniformity required a process of standardisation that, with time, was enacted through the administrative centralisation of social protection. The issue of identifying the recipients occupied the forefront of these processes.
An early and influential source of this will to standardise was the 1996 Committee for the Restructuring of Social Security (CRSS), led by Thabo Mbeki's adviser Frank Chikane. The belief that 'the delivery of social security is in crisis' motivated the Chikane committee. 22 Fragmentation into 14 separate systems, each with particular management, rules and procedures, created the opportunity for loopholes 'which could easily be exploited by unscrupulous officials and members of the public'. Of the R11.5 billion paid out to 2.8 million beneficiaries in 1995 -96, the committee estimated that about 10 per cent was lost to fraud and corruption. As it reports, 'fraud and corruption are rampant as a consequence of a lack of systems, proper internal controls, unduly complex legislation and department rules governing internal disciplinary proceedings'. The committee said that this malfeasance 'represent [ed] the greatest threat to the programme' of welfare, requiring a 'complete re-engineering'. 23 While it mentioned the possibility that some tasks could be decentralised, this potential was not described in further detail; instead, the Committee focused on unifying and standardising a national system which it believed would make the detection of fraud easier. The report also emphasised the need to establish 'linkages with other systems . . . such as the Home Affairs, Population Registration System, other pension insurance funds, South African Police Services (SAPS), Deeds Registry, the provincial financial control systems, and post offices and banks'. 24 Specific recommendations for the social security system were offered, including a nationally organised system, a national human resources strategy, a standardised and integrated management system, a 'national transverse information system' and regulatory simplifications. Thus the committee saw standardisation as a task both within social security and between other government programmes.
It directed this logic particularly at the means of identifying grant recipients. The committee argued that the 'benefits of a biometric system can only be reaped if there is a uniform system for the country as a whole'. 25 They were deeply concerned that the Eastern Cape, Gauteng, Northern province, KwaZulu Natal and Mpumalanga were in various stages of issuing biometric contracts, enrolling beneficiaries and using different proprietary systems for delivery, perhaps permitting fraud: '[w]ithout a national fingerprint database this gives no assurance that the person is who they claim to be'. Thus their opposition to biometrics was pragmatic, concerned with the difficulty of creating a functioning system, especially as the amalgamation of different systems might compound data errors.
They also found the costs deeply worrying, especially in light of early indicators of ineffectiveness and corruption. As early as 1993, biometrics were tested in the Cape province, but the results included considerable false positives. Despite this, biometrics reportedly cost more than double the alternative methods. In the Western Cape, the biometric system was said to cost R1.9 million per month, compared with the savings of R16,500 per month from identifying duplicate recipients. 26 Throughout, concerns emerged about corruption in provincial tendering for biometric welfare systems; ultimately these concerns would result in a full investigation by the Office for Serious Economic Offences (OSEO) and ensnare the National Party's senior welfare official, Abe Williams. 27 In the years following the Chikane committee's deliberations, the depiction of the problem and solution solidified into a relatively consensual view that (a) a crisis existed, and (b) standardisation was the necessary response. The archive yields few dissenting voices. The 1997 White Paper on Social Welfare declared that fragmentation led to 'gross inefficiencies' and 'loopholes . . . which could be exploited by officials and the public'. 28 The White Paper specifically cited as problems the 'manual system' for accounting and 'the lack of an integrated national ID system'. 'A uniform social grants system', it reasoned, required 'the rationalisation of computer systems and the development of a National Social Grants Register and automated fingerprint technology'. 29 It noted that a national re-registration effort might be required to produce an accurate database.
The sense of calamity grew when the Mail and Guardian newspaper revealed in 1997 the loss of an estimated R1 billion to pension fraud, from an annual budget of R14.3 billion. 30 A major report in the following year from a government watchdog, the Public Service Commission (PSC), detailed the state of service delivery and proposed a variety of institutional and technical reforms. The PSC's report argued, like its predecessors, that the troubles of the social security programme resulted from a lack of standards in the data captured from applicants, the complexity of regulations, the 'varying interpretations of eligibility' and the methods of payment. It recommended the 'development of a national policy to standardise' paperwork, bureaucratic procedures and identity documents. 31 Further, it promoted increased centralisation of the welfare bureaucracy, rejecting a decentralised model where 'each province develops its own social security service', due to limited accountability and the recognition that 'all citizens of South Africa should be treated equally'. 32 When the report received additional attention at a series of parliamentary hearings in February and March 1998, many of the findings and arguments were affirmed. For example, members of the Executive Councils from Free State and the Eastern Cape complained of the identification troubles facing their provinces. In the Free State, it was alleged that citizens from Lesotho claimed South African pensions; in the Eastern Cape, former Transkei and Ciskei citizens lacked the new ID books, while others had both the new ID books and older identity documents that were still accepted, and thus benefited twice. 33 The 2002 Report of the Committee of Inquiry into a Comprehensive System of Social Security for South Africa (led by Vivienne Taylor) provided an important boost to these recommendations by advocating the establishment of the South African Social Security Agency -an entity introduced in 2004. The goal of SASSA is to serve as 'the sole agency that will ensure the efficient and effective management, administration and payment of social assistance'. 34 As Selwyn Jehoma, the head of grants for the Department of Social Development, told parliament in February 2003, integrating social security within a national agency would, in addition to providing clear accountability, permit 'standardisation and uniform business processes', which would 'reduce costs of service delivery'. 35 Further, '[p]ractices of double dipping into funds would also not be possible with an Agency as the institution would have a better grip on social services'. Social Development Minister Skweyiya reiterated this, saying that 'financial leaks had necessitated the centralisation of control and payment of grants, leading to the formation of a social security agency that would do the job'. 36 As Mr Jehoma told parliament a few months later, this was necessary because 'it was not possible to give all nine provinces a set of guidelines and expect them to interpret and implement them in the same way [so] a nationally guided process was needed'. 37 While few in parliament advocated otherwise, at least some civil society organisations raised doubts about centralisation: the Black Sash worried that SASSA would merely 'replicate the existing dysfunctions of the system' and that 'the new system overlooks the political, social and technological nuances in each province'. 38 Their protest, however, was ineffectual.
SASSA, then, has been tasked with homogenising the policy and implementation of grants. While the fragmentation was particularly troubling for the manner in which it facilitated illicit behaviour, it was also at odds with the egalitarian ethos of the newly democratic nation. In the context of hard-won equality, the grants, then, became a means of aspiring to and establishing uniformity of citizenship.

The Pursuit of Objectivity
In addition to bureaucratic structures, fragmented and discretionary practice has come under scrutiny. This section examines the manner in which everyday practices were problematised as a 'crisis' that justified biometric identification. As the second underlying force of the biometric imaginary, the pursuit of objectivity has sought to implement procedures and techniques that minimise the opportunity for subjective practice. In particular, techniques of objectivity have been directed at two forms of behaviour that have continued to trouble the bureaucratic and policy elite: illicit access to grants and inappropriate or inept bureaucratic action.
Historically, grant fraud has occurred in various ways, from a 2008 case where just three people absconded with R22 million, 39 to more everyday acts of dissimulation, such as 'looking poor' to pass the means test. 40 In conversations with and public statements by SASSA officials, I found that two practices are considered particularly widespread. First, officials worried about 'phantom twins', the practice of registering fictional children in order to receive extra grants (one town reportedly had more than 100 twins in 2010). 41 Secondly, officials felt that deceased pensioners are not reported as dead, and relatives continue to collect money on their behalf. This form of fraud is often linked to the use of bank accounts, where money is traditionally deposited without verifying that the recipient is alive. 42 The wayward behaviour of middle-and low-level bureaucrats has also been particularly troubling, with the widespread belief that error and incompetence undermine the effectiveness and equality of service. For pro-poor civil society organisations, bureaucrats have historically been found to be an impediment to improving legitimate access to the grants. This was most evident during the early years of the Child Support Grant, when organisations such as the Children's Institute and Black Sash highlighted 'worrying . . . reports that applicants are often dissuaded from persisting with applications because of the attitude of welfare officials'. 43 They called for welfare officials to be educated to improve their delivery behaviour. As Simon Kimani of the National Association of Democratic Lawyers told parliament, 'some welfare officials are arrogant, rude and abusive'. 44 At the time, in 2002, the process of applying for grants was described as 'torturous', because there were 'no uniform standards, assessment guidelines and procedures, and some officials themselves [did] not know of or understand the procedures'. 45 For some, outsourcing to private payment firms was particularly problematic owing to the inability or unwillingness of private firms to provide quality services. 46 As Francie Lund and colleagues noted, 'administrative discretion appears to be subverting the aim of the broader social policy'. 47 The work environment has exacerbated these problems. Offices were poorly maintained, without electricity and other infrastructure. Paperwork and documents have been particularly troubling, with necessary forms variously lost, in the wrong language and open to forgery. 48 For post-apartheid welfare, the documentary mediation of state and citizen has often been a source of discretion and error, undermining the aspirations to fixed objectivity pinned upon them. 49 During the 1990s, there was 'no way of verifying the authenticity' of applicants' documents, especially if they were issued by entities with which the welfare administration had little interaction (most importantly Home Affairs). 50 In 2001, a report from the Auditor-General found 225,471 computer-generated ID numbers. These were formally supposed to be used for recipients who had only apartheid-era identity documents, but in practice unscrupulous bureaucrats used them to create false recipients. Nearly a decade later, an audit of SASSA highlighted similar documentary problems, noting deficiencies 'including information technology controls' and 'poor-filing management'. 51 For beneficiaries, a lack of identity documents has been one of the primary barriers to accessing the grants. Budlender et al. detail how 'officials were requiring documents and other evidence far beyond what the law dictated'. 52 Early in the use of the Child Support Grant this was recognised as a barrier, and lobbying removed some of the requirements for children, but it still took a lawsuit -decided only in 2008 -to permit a broader array of documents, including sworn affidavits. 53 Yet a 2010 survey found that problems with documentation remained the most significant reason for not receiving legitimate grants. 54 As Breckenridge noted, biometrics remove bureaucracy 'from the world of paper-based documents and -more importantly -from the domain of human agency'. 55 In Daston and Galison's words, '[i]nstead of freedom of will, machines offer . . . freedom from will'. 56 In the case of post-apartheid welfare, the goal continues: in a recent outline of work for 2012 -2015, SASSA emphasises that The automation of systems for improved service delivery is non-negotiable . . . The constant use of manual systems not only limits the number of applications that can be processed in a day, but also contributes significantly to fraud and corruption in the grants administration system. 57 South African Biometric Welfare under SASSA South Africa has a certain consistency in welfare policy discourse. Since the 1990s, social grants have been depicted as 'in crisis', owing to the spectre of these forms of bedevilling 48 Kimani, 'Social Security Delivery'. 49 Although it is not traditional to consider inanimate paperwork an influential mediator, a growing amount of scholarship has emphasised the vitality of mundane artefacts. 'Just as discourse has long been recognised as a dense mediator between subjects and the world, we need to see graphic artifacts not as neutral purveyors of discourse, but as mediators that shape the significance discretion. The illegal and diffuse nature of this behaviour makes estimating the extent of the problem quite difficult, but numerous entities have tried. 58 These figures filter through parliamentary hearings and budget speeches, news reports and opinion pieces, and daily discussions in South African society. When the Democratic Alliance said in parliament that the DSD was 'crippled by managerial dysfunction' and facing 'a management crisis', 59 or when the Inkatha Freedom Party labelled 'all SASSA offices across the country as . . . breeding grounds for corruption', 60 their rhetoric was not particularly uncommon, nor was it merely opposition politicking. The leaders of DSD and SASSA are also quick to admit widespread fraud and corruption, with minister Dlamini calling it 'endemic'. 61 Despite the difficulties of accurately assessing the amount of fraud and corruption, the ambiguity is rarely noted. It is the perception of crisis that is productive -based on, but not congruent with, facts. Janet Roitman has called for an understanding of 'the kinds of work the term "crisis" is or is not doing'. 62 In this case, at least one result is the adoption of biometric identification, conceived as a way to deliver grants impartially and uniformly, lowering costs and boosting efficiency. As a standard and objective technology, biometrics are understood as the solution to crisis. The 2012 contract and re-registration into a centralised biometric database is the highwater mark of the welfare administration's adoption of biometrics, propelled in part by the weakness in national identification infrastructure. In the 1990s, welfare policy-makers began promoting the Home Affairs National Identification System (HANIS) as a key enabler of welfare expansion. 63 Its proponents meant HANIS to be a multipurpose national identity card; however, it became a long-running fiasco that only began to issue identity cards in mid 2013. 64 Instead, welfare officials began to take steps to develop a uniform identification method for grant recipients in particular. As early as 2003, a senior DSD official told parliament that the proposed SASSA 'would not have nine different contractual arrangements across all the provinces'. 65 In 2007, shortly after its formation, SASSA tried to standardise payment and identification infrastructure through a tender whose intent was 'to ensure that service providers appointed in the nine provinces provided a standardised payment service in line with the norms of service delivery approved by government'. 66 It was ultimately (and grudgingly) cancelled, because the bids received offered no such regularity. As the adjudication committee wrote, the bids did not offer 'standardised payment services' or appropriate norms and standards of security and integrity, nor were they cost-effective. 67 In the interim, SASSA used the inherited provincial systems, many of which were technically incompatible and ineffective. Furthermore, for technical and contractual reasons, SASSA has not had ownership of a unified biometric database for recipients. 68 The system that resulted from the amalgamation of provincial databases has been routinely criticised for being out of date, filled with inaccuracies and open to fraud. Where fingerprint verification did occur, it was not always reliable, such as the case in 1999 when one town was found to be home to 1,650 identical fingerprints. 69 More recently, though, biometrics have been turned inward in an effort to remove the opportunity for bureaucratic misbehaviour. As SASSA reported to parliament in March 2010 when justifying their budget, '[s]taff were more strictly controlled through biometric access systems and clearer controls' that recorded employee activity, to search for fraud and 'ensure that these staff could not easily access sensitive programs where they could manipulate information or create "ghost beneficiaries" in the system'. 70 This is part of a broader effort to automate grant delivery, removing unskilled or corrupt bureaucrats. For example, the acting chief executive officer (CEO) of SASSA illustrated the biometric imaginary's conception of objectivity in an early 2011 complaint that '[t]he lack of automated business processes make [sic ] activities extremely labour-intensive and error-prone'. 71 Later that year, the head of SASSA's internal audit and fraud management unit highlighted automation as a means of overcoming 'poor employee work ethic'. 72 Moving to electronic systems also saves money, and the acting CEO of SASSA noted during a time of financial limitations that automation also helped to reduce personnel, their 'second largest portion of the budget'. 73 The continuing salience of these twin goals is clearly evident in the issuing of the new grant payment contract in early 2012 to Net1 CPS, a South African technology firm. The government sought to address the same issues as in the failed tender mentioned above. Its request for bids from contractors called for 'significantly improved services' with 'sameness of Beneficiary experience'. 74 'The minimum acceptable requirement during bulk and ongoing enrolment', they said, 'is that all ten finger prints of Beneficiaries must be captured. The Biometric Data capturing during enrolment will be used for matching and authenticating during payment process'. In addition to requiring this data from grant recipients, beneficiaries -such as children -would also be incorporated. The goal, it explained, was to ensure 'that a Beneficiary is not enrolled more than once'. 75 The award in early 2012 of the R10-billion contract for the entire country to Net1 CPS prompted an acrimonious court challenge by a losing bidder, AllPay. 76 As a subsidiary of Barclays' ABSA bank, AllPay was the incumbent payment provider in the Western Cape, Free State, Gauteng and parts of the Eastern Cape, while Net1 CPS operated in the other provinces (except Mpumalanga, where a third firm -Epilweni -was used). AllPay alleged improprieties in the tendering, including a potentially last-minute change in requirements. These allegations were quickly supported by subsequent media reports, which suggested the influence of bribery. 77 These assertions led to a brief investigation in South Africa, one ongoing (as of late 2014) in America (where Net1 is listed on the stock exchange) and a series of lawsuits, which came to the fore with a November 2013 decision by the Constitutional Court, requiring that a new tender be offered without procedural irregularities.
SASSA's court filings in response to AllPay's initial February 2012 complaint make clear the importance of standardisation and biometric authentication to their new contract. Their filings reiterate the established understanding of delivery failure and abuse discussed above. For example, the SASSA representative explained, 'that the six different [payment] methodologies employed in the Eastern Cape Province has directly led to substantial fraud and other abuses'. 78 In contrast to SASSA's response, AllPay's complaint depicted a different vision for the contract. 'AllPay's focus', it asserted, 'is to facilitate access to financial services and products to beneficiaries of social grants, with particular focus on the rural and semi-urban communities'. 79 Given the extensive network of ABSA bank branches and automated teller machines (ATMs), AllPay maintained that they could provide formal financial services where the non-bank Net1 could not. An insistence on biometric identification for each payment (instead of just registration) would render ATMs useless, because they do not have fingerprint scanners. SASSA, however, rejected the pre-eminence of 'banking the unbanked', contending that AllPay 'did not make provision for adequate biometric verification and standardisation of services', and thus 'fell short of requirements imposed by SASSA'. 80 Net1 CPS, in contrast, put biometric identification at the core of its offerings, and, as it gleefully revealed in its submissions to the court, its new offering has been able to combine its proprietary biometric payment technology with the standard national payment system (such as that used by bank cards and ATMs). Because the standard system would still not allow fingerprint verification, Net1 CPS would 'conduct proof of life verification telephonically, as an alternative to fingerprint technology'. 81 This new form of voice biometric verification would be used in cases where recipients received their grants into a bank account, requiring a brief monthly call to certify that they had not died. 82 This seems to have caught AllPay by surprise. Not completely without merit, they had believed that fingerprinting was the biometric method de rigueur. Indeed, in declaring the contract invalid in late 2013, the Constitutional Court affirmed that a last-minute shift to requiring monthly biometric verification (rather than only at registration) rendered 'the process entirely uncompetitive'. 83 It was not the case that AllPay was opposed to biometric identification -indeed, they spent considerable time detailing their technical acumen -but only after seeing their opening salvo in favour of 'banking the unbanked' stall. It was a difference in emphasis arising from technical and institutional legacies. Net1 CPS stakes its work on universal biometrics; for AllPay, it has been secondary to -even sometimes at odds with -its banking and ATM system.
Beyond the court filings, SASSA officials have validated the new system to the public. For example, in response to an inquiry about grant fraud from parliament, the CEO of SASSA called the new biometric scheme 'the greatest risk reduction tool that SASSA had engaged in its history'. Emphasising the appeal of removing bureaucratic weakness, she has specifically highlighted 'automation of business processes'. 84 Following the Constitutional Court's unanimous declaration that the contract between SASSA and Net1 CPS was 'constitutionally invalid', where this proclaimed success stands, however (as of late 2014), is less clear. 85 The Court found that SASSA had contravened its constitutional duties in two ways during the tender process. First, the last-minute shift to seemingly requiring biometric verification at each payment created vagueness and uncertainty that undermined procedural fairness. Secondly, SASSA did not ensure that Net1 CPS was, as constitutionally required, empowering previously disadvantaged people. And while it thus found the contract 'fatally defective', it recognised the importance of continued grant delivery. 86 Therefore, it required a new tender process be conducted with regular oversight by the judiciary. For SASSA, the task of distributing millions of rand per month has required a sociotechnical network that minimises the capacity for intermediaries to serve as anything but passive infrastructure. In the words of the PSC, the goal is that 'policy and execution are not divorced'. 87 Biometric identification has been understood as a crucial means of achieving this goal, serving as what Rose calls a 'key fidelity technique'. 88 In many ways, the new grant payment and identification infrastructure instituted in 2012 is the result of nearly two decades of reform efforts. The Chikane committee's call for biometrics to be addressed in a unified, national manner has been accomplished. In minister Dlamini's words, '[t]he improved biometric-based payment solution was long overdue and will go a long way towards minimising fraud and corruption so prevalent in the previous systems'. 89 But if the literature on audit 90 is an indicator, there will need to be a social solution to the lack of trust, for solutions based on technique alone 'fail to immunise the assemblages they govern from doubt'. 91 Indeed, already the biometric technology has been questioned and subject to fraud, such as a case where fraudsters were found to be in possession of three biometric registration machines. 92 The rest of this article considers these shortcomings more directly.

The Myth of Mimesis
Policy-makers and their peers in civil society, the media and elsewhere have presented biometrics as a material means of improving service delivery, uniformly identifying recipients and removing undesirable activity. Speaking of the new biometric grants programme, Social Development Minister Dlamini unequivocally stated that it 'will eliminate incidents of fraud and corruption in the social grants system', which, she said, resulted from the 'manual system that SASSA has been using'. 93 SASSA's chief, Virginia Peterson, claimed 'the new SASSA card . . . will make it virtually impossible for fraudsters to defraud the Agency'. 94 This appeal of biometric identification rests on its presumed uniqueness and universality, and its ability to labour reliably free of error. 95 This faith builds on the presumption of universality and algorithmic objectivity, especially when compared to alternative means of authentication, such as a personal identification number (PIN), which is considered vulnerable to fraud because the number can be shared. In contrast to this view of imperviousness, Magnet argues that biometric failure is not an aberration or exception, but rather an endemic feature. 96 The experience in South Africa adds to her thesis, suggesting that the faith in biometrics is at least partly misplaced, and that they are unable to remove the unevenness for which they are adopted. The mistake arises from a simplistic, overly technocratic understanding of real-world implementation. Too close a focus on the technical means through which the state identifies individuals often underlies the biometric imaginary. Even the otherwise thoughtful Chikane committee fell prey to defining problems and solutions in terms of technological systems, arguing that the 'principal cause . . . [of] serious and widespread' fraud and corruption was 'the lack of proper systems of internal control'. 97 Such a view ignores the role that poverty, inequality and limited opportunity play in generating fraud. Instead, it focuses particularly on reforming systems of surveillance and audit. Biometrics here become another technical solution to social problems. But as the repeated difficulties encountered when instituting such a system demonstrate, one cannot isolate technical efficacy from the bureaucratic technopolitics of South Africa.
This false and narrow faith in biometrics has a lengthy pedigree in South Africa. 98 It continued as problems emerged in the 2012 -13 registration initiative. For example, the children to be enrolled often reacted negatively to the experience, protesting and crying. 99 Elderly and disabled recipients found it difficult to travel to the registration sites. Early results from the use of the voice recognition system, too, suggest technical difficulties, including poor recording environments during re-registration, and delayed implementation. 100 The result is a chaotic and confused scene at payment locations, not a standardised process. 101 A more significant failure of biometric technology is the portion of the population unable to enrol. In September 2013, thousands of elderly and ill pensioners did not receive their grant because they had been unable to re-register, and the home visits SASSA promised had not occurred. 102 Biometrics have proven especially problematic for domestic labourers whose fingerprints are eroded from years of washing dishes. Additional worries exist for farm and mine labourers, many of whom have lost fingerprints or even whole fingers and hands. Because biometric identification is compulsory in the new system, these populations will require a procurator, an individual who will enrol and withdraw money on their behalf. 103 In October 2012, with only 15 per cent of the recipients re-registered, 13,000 procurators were already required. 104 This unsettles the stated goal of 'sameness of Beneficiary experience', and creates a form of dependency that, at the very least, will introduce complicated negotiations and conflicts for the individuals. 105 Importantly, this disjuncture between presumed (and marketed) universality and the diverse reality of bodily features occurs along class, age and gender lines: fingers typical of white-collar labour are rarely problematic, compared to low-income populations engaged in manual labour.
For all their mimetic aspirations, neither biometric technologies nor the dominant ways in which they are conceived accurately reflect reality. Thus the term 'biometric imaginary' is particularly apt: an imaginary is not a fantasy -it is grounded in reality and is actually productive -but it may be decoupled from certain on-the-ground realities. Put another way, despite biometric universality and objectivity being mythical, the biometric imaginary is productive. The technology does, viewed en masse, by and large enable an enormous and impressive welfare programme. But like other technopolitical efforts, biometric identification can also be a productive failure. 106 The promises embedded in the technology -such as ending corruption -need not be fully realised in order to justify the expense and effort of biometrics. This interpretive flexibility enables a wider range of support for biometric identification, fuelling the biometric imaginary. Because biometric technology is presumed to be effective it can serve certain needs.
This was clear when minister Dlamini was 'pleased to report' in March 2013 that the biometric re-registration process had resulted in 44,000 grants being cancelled and 66,000 lapsing due to non-collection. 107 In both her speech and the subsequent media reports, this represented a successful example of using biometric identification to, as she put it, 'root out fraud and corruption whenever and wherever it appears [,] to ensure that social grants only go to eligible beneficiaries'. She does not note that, as a portion of beneficiaries (less than half a per cent), this is quite small.
Thus welfare surveillance and auditing did not need to prove technically efficacious in order to produce its effects here. These individuals were not caught; rather, they preemptively cancelled their grants or allowed them to lapse. Even if all the lapsed grants were fraudulent (an heroic assumption given widespread beneficiary confusion and frequent turnover normally), the promised de-duplication or identity verification was not responsible. The mere perception that the technology works produces this result. The authoritative presentation of biometric technology as effective is performative; rupture-talk helps to create the desired disjuncture. 108 A bit more speculatively, one can imagine that the presentation of such anti-corruption efforts -especially those utilising innovative and not particularly wellunderstood technology -performs a different reality to another audience: it demonstrates a particular seriousness about the 'crisis' in grant fraud and corruption on the part of DSD and SASSA to a population and political elite deeply concerned with cost overruns and illicit grant access.
Finally, biometrics also 'work' in another way, even if they do not meet their promise of ending fraudulent access. In receiving the SASSA contract, Net1 CPS secured a large and lucrative market of around R2bn annually. Had they lost the contract, 50 per cent of their business reportedly would have followed. 109 Instead, they have revealed that they are currently in talks with 11 other countries, no doubt benefiting from the prestige and experience of this contract. Serving as payment provider for so many has also opened up ancillary markets, such as (controversially) offering financial services. 110 For those who place credence in the corruption allegations (which they deny), the importance of this contract to the firm is taken as motive, but a more general point is also true: the fervent commitment to combating the 'crisis' with biometric identification -and not, say, identity documents -has boosted a nascent technology with few other markets.

The Political Ambiguities of Delegation
In addition to the means, the ends of this process also raise questions. The goal to end improper use of discretionary powers and uneven service delivery has delegated responsibility increasingly away from humans and toward machines. This has happened largely in two ways: (a) a move from poorly scrutinised or verified grant applications towards interoperable databases, in order to remove those deemed undeserving, and (b) a move from street-level bureaucrats towards electronic cash dispensers. Most South Africans consider this transformation legitimate, but in neither case is it an unalloyed good. Instead, as von Schnitzler has discussed with regard to prepaid metering in South Africa, these infrastructures implement particular ethical and political regimes. 111 In the case of biometric welfare in South Africa, they may even expand the distance between the state and its most vulnerable citizens.
The emphasis on technologies of surveillance and audit serves to constrict what Scott calls 'infrapolitics', the everyday weapons of the weak. 112 Such infrapolitics occurs in at least two ways. First, many poor South Africans share knowledge about how to pass the means test and qualify for a grant. 113 Condemning this as fraud ignores the dire situation of those who live above the qualifying income but still have considerable needs due to ongoing rampant underemployment. Second, many who at one point qualified for the grant continue receiving it after their financial status improves; the Special Investigation Unit considers such behaviour one of the primary forms of grant 'fraud'. 114 Yet the simplistic, binary means test on which such reasoning relies ignores the fluid, fluctuating incomes of the poor. The dominant bureaucratic approach to fraud, then, ignores what are actually infrapolitical tactics and necessary livelihood strategies. 115 Instead of recognising these as legitimate, state officials may actually increase the surveillance of grant usage, perhaps through digital transaction monitoring or payment systems that block purchases such as alcohol. 116 As detailed above, the new grant system has sought to replace street-level bureaucrats with machines, and to increase recipient convenience by allowing payments at ATMs and third-party merchants (equipped with biometric readers). While 60 per cent of beneficiaries previously received their grant at SASSA pay points, by April 2013 only 22 per cent did, with a significant shift toward ATMs and retailers. 117 Subsequently, ATM providers have reported 'a huge injection' of new users in the form of SASSA beneficiaries. 118 In contrast to the traditional SASSA pay points, managed by government representatives, the new system requires beneficiaries to interact with a third-party merchant or faceless machine.
The irony is that the distance between citizen and state has been extended in an effort to remove discretionary middle-men between the Treasury and the pockets of the poor. In Ferguson's felicitous phrase, the result is perhaps 'asocial assistance' rather than 'socially "thick" recognition . . . between state and citizen'. 119 When something fails -as it did during a technical error in January 2013, when many pensioners in the Western Cape did not receive their full grant -the delegation to machines and third parties means that those on the ground cannot address the errors. 120 The shift to objective technology is not neutral, but biased towards those who control it.
These acts of delegation and redistribution of power raise the stakes elsewhere. The countless street-level negotiations, manoeuvres and deceptions that the biometric imaginary condemns as illicit fraud (rather than ambiguous infrapolitics) decline in importance while the influence of the central policy and standards rise. 121 As discussed earlier, a purposeful centralisation of policy-making has taken place, most notably through the creation of SASSA. This shift away from the autonomy of provinces and low-level bureaucrats has been considered an effort to reduce discretionary powers that would introduce inequalities of service delivery; it reflects SASSA's desire for a singularly defined delivery mechanism. When the rise of biometric welfare administration is viewed as redistribution of subjectivity, however -rather than (as the biometric imaginary has it) an absolute reduction -then it is less surprising why the current biometric contract between SASSA and Net1 CPS has been the subject of intense legal dispute and allegations of corruption. If the allegations are true, the effect of the biometric system may have been to reduce petty corruption but increase grand corruption. 122 Although SASSA and Net1 strenuously deny the allegations (and have gone forward with the contract), the episode suggests that the reduction of subjectivity at street level increases subjectivity at the centre. This centralisation also creates fragility, illustrated during the court battles when the judiciary found the contract invalid but refused to set it aside owing to the disruption it would cause to the millions of grant beneficiaries.
Indeed, the centralisation of welfare decision-making in post-apartheid South Africa has created a situation of concentrated judgement about who qualifies for grants and who does not, both in policy and technical performance. The opposition to street-level subjectivity magnifies the influence of decisions made by SASSA. Given the pathetic state of social assistance delivery at the end of apartheid, some kind of centralised uniformity was clearly necessary, 123 but numerous observers have argued that the South African state is now too centralised and technocratic. 124 The biometric imaginary may be understood similarly, but, as I have argued, it also results from political liberalisation, an opening up that means that poor administrative capacity is no longer free from the agitations and demands of the public. The technopolitical regime that reached its height with SASSA's 2012 biometric payment contract is closely tied to the founding principles upon which most agree: post-apartheid social policy should be equitably and competently administered, in a manner that is impersonal and objective, free from the informal and formal discriminations of the previous era.

Conclusion
The social grants are one of the flagship interventions of the new South Africa, even if, in many ways, they grew out of the old South Africa. Their importance and scale make them the subject of politically divisive debates, on aspects ranging from their affordability to their impact. Throughout these contests, though, runs a commitment to delivering grants reliably and equitably. This realm of implementation is one of civil servants and paperwork, fingerprint ridges and accounting techniques. A complex interchange between state and citizen that occurs millions of times each month, the implementation of social grants relies on a relationship in which biometric identification technology is pivotal. Importantly, it is a deeply political technology that has largely been depoliticised, freed from the scandals that emerge in other contexts. 125 On one level, this article has asked what explains the investment in this identification infrastructure in the post-apartheid bureaucracy. A colloquial answer would point to fraud, corruption, and administrative incompetence. And, indeed, these ongoing grant scandals have been a key motivating factor for the non-scandalous adoption of biometric technology. But this does not exhaust the particular reasons and manners through which this history has unfolded.
Instead, my argument traces a shared commitment -the biometric imaginary -that has emerged and solidified during nearly 20 years of welfare implementation. While not wholly distinct from the apartheid history of biometric registration, the emphasis on welfare provision is an important change that has facilitated the expansion of biometrics since the 1990s. The biometric imaginary positions biometric technology as a necessary, suitable and effective means of achieving standardised and objective welfare administration. Broader than simply the purchase and use of fingerprint scanners, it builds upon programmes of institutional reform, policy-making debates and infrastructural development. Yet, the implementation of a nationally centralised biometric identification scheme is perhaps the apogee of the biometric imaginary, a technopolitical goal that has emerged during the postapartheid era as various forms of delivery weakness and fraud have bedevilled the initiative.
Unification through a process of centralised standards-setting has been the order of the day, necessitating institutional reorganisation, regulatory alignment and infrastructural interoperability. For the government, biometric identification represents an automated -and thus impersonal -means of identification, a way to end illicit behaviour while paying grants uniformly. As I have argued, these commitments to standardisation and objectivity are deeply held, influencing welfare policy and practice well beyond biometric identification. As the new biometric grant system unfolds, they will continue to influence its trajectory, though they are unlikely to do so in uniformly effective or positive ways.