Understanding Social Performance: A ‘Practice Drift’ at the Frontline of Microﬁnance Institutions in Bangladesh

This article examines the role of microﬁnance staff and procedures in en- abling microﬁnance’s social mission. It does so primarily through studying institutional ruling relations and practices in rural Bangladesh. Attempting to move away from the linear and deterministic approaches of impact studies, it ethnographically scrutinizes the everyday practices of implementers. Findings point to the emergence of systemic practices that jeopardize microﬁnance institutions’ potential to perform their social mission. These include The results from the survey conducted with ASA’s credit ofﬁcers show that although ofﬁcial ofﬁce time ends between 5 pm and 6 pm, 76 per cent of them ‘regularly’ return to the ofﬁce after 8 pm and 51 per cent ‘regularly’ return after 10 pm. Written statements from loan ofﬁcers indicate that this practice often involves collective action from different members of staff:


INTRODUCTION
Determining the performance of development policy requires vast amounts of energy and resources from planners and researchers (Easterly, 2006). Policy makers often relegate implementers' roles to enacting and applying This article is based on research conducted by the author in rural Bangladesh. The author would like to thank the Brooks World Poverty Institute (now Global Development Institute) for supporting this research and Professor David Hulme and Professor James Copestake for their guidance and support throughout. She is particularly grateful to the respondents who agreed to engage in valuable discussions. Finally, she would like to thank the members of Development and Change's editorial board and the three anonymous referees whose critical comments and feedback helped to substantially strengthen the article. For any further details about, or to discuss, the argument presented in this article, please contact the author. prescribed administrative tasks (Biggs and Smith, 2003). This discursive determinism can exaggerate the role of structures in enacting change (Long and Long, 1992) in that it neglects the crucial part played by human actions and agency in determining the success or failure of policy initiatives (Juma and Clarke, 1995: 126). Similar assumptions about the linear and predictable relationship between policy and practice inform most existing analyses of microfinance interventions. This article explores this phenomenon in relation to social performance, and proposes alternative ways of conceptualizing the relationship between microfinance's mission and its impacts, which help account for the mixed results and diverging experiences associated with microfinance.
The microfinance sector emerged in the 1990s as a revolutionary tool for poverty alleviation (Morduch, 1998;Robinson, 2002). It is implemented through structures called microfinance institutions (MFIs). MFIs are considered institutional 'hybrids' (Labie, 2001: 297), mandated to achieve a double bottom line of financial self-sufficiency and poverty reduction through reconciling market forces with development objectives (Armendáriz and Morduch, 2005). They deliver financial products and services to the bottom of the pyramid, those labelled by mainstream banks as unreliable borrowers incapable of saving (Cull et al., 2009). A tenet of microfinance is that the provision of financial services and products can enable poor households to invest productively in activities that generate sufficient financial returns to improve their condition and free themselves from poverty (Yunus and Jolis, 1999).
Since the early 1990s, the sector has undergone significant structural transformations characterized by increasing commercialization (Otero and Rhyne, 1994;Robinson, 2002). In practice, many MFIs have implemented new, 'more comprehensive management procedures' (Beisland et al., 2014: 277) to bring efficiency and sustainability to their activities through relentless cost-cutting strategies. First and foremost these procedures aimed to protect MFIs' financial performance (Christen and Drake, 2002: Ch. 1;Woller, 2002b). This was seen as all the more crucial in this case because the poor clients were considered to be the principal source of risk. Major stakeholders, academics and development agencies often encouraged the development of frameworks and tools to improve MFIs' ability to cover their costs (Godquin, 2004;Littlefield and Rosenberg, 2004).
A primary explanation for the limited effectiveness of MFIs in reducing poverty has been the commercial and profit-oriented nature of many of them (Ghosh, 2013). Studies by Arunachalam (2011), Taylor (2012) and Guérin et al. (2013) closely examine the gradual commercialization of institutions in Andhra Pradesh, South India, Kenya and Mexico, providing rich countryspecific analyses of the emerging internal tensions and paradoxes within MFIs between their financial and social missions. These in some ways mirror those seen in the Western financial sector's 'subprime crisis' (Mader, 2015), questioning banking institutions' capacity to uphold ethical lending practices (Hulme and Maîtrot, 2014).
Many scholars have explained these dynamics in terms of a 'mission drift' (Copestake, 2007(Copestake, : 1722Woller, 2002b: 14), whereby senior managers, faced with the challenge of microfinance's double bottom line, are said to take deliberate measures to resolve the perceived trade-off between the financial and social missions, at the expense of the latter. According to this argument, the increasing commercialization of the sector incentivizes managers to alter their institutional strategy to better align it with commercially-minded stakeholders' interests. MFIs therefore purposefully target wealthier households, not to cross-subsidize for poorer ones but to secure high financial returns (Armendáriz and Szafarz, 2011). The mission drift therefore denotes an intentional and explicit top-down shift in the institutional practices of MFIs towards securing and protecting financial performance or profits (Fouillet and Augsburg, 2010).
This article contributes to understanding social performance in microfinance. It will argue that existing studies of social performance have reinforced a deterministic understanding of microfinance, which assumes that policy translates linearly into practice. The role of implementers, and implementation processes and mechanisms, remain largely under-studied and under-conceptualized in attempts to understand performance and impact. This article deepens our understanding of how microfinance works, as opposed to whether it works, by examining the role of implementers and implementation. It builds on a central insight of the mission-drift concept, which is that commercialization has changed the way microfinance is practised. The argument developed here, however, is that in the MFIs studied, senior managers have not explicitly shifted away from their stated social mission, nor have they explicitly targeted better-off households. Malpractices observed stem not from a shift in mission translating linearly down into the field, but from a tacit displacement of the decision-making process about social-financial trade-offs to the branch level, fuelled by the need to achieve everyday targets. These targets are often assumed to be an effective way to institutionalize and administer the pursuit of the social mission (increasing the number of clients and monitoring outstanding loan amounts). In fact, as we shall see, they serve the financial performance of branches while warping management practice and the interests and behaviours of staff. The predatory and fraudulent strategies and tactics developed by field-level staff to achieve the targets affect the social performance of the MFI in ways that contradict its stated social mission. This is conceptualized here as a 'practice drift'. These informal yet institutionalized practices shape client recruitment and follow-up, loan renewals, top-up loans and repayment collection procedures. In recognizing the diverse meanings of microfinance for different actors and the variety of practices microfinance embodies, this study nuances the literature which presents microfinance as a uniform and homogeneous development intervention (Armendáriz and Szafarz, 2011;Labie et al., 2009;Rhyne and Otero, 2007). It also suggests an alternative route for explaining its unforeseen and unintended practices and impact on clients (Brigg, 2006;Campbell, 2010;Cons and Paprocki, 2010;Guérin et al., 2013;Karim, 2008).
This article examines the social performance of microfinance in the context of Bangladesh. Bangladesh is often seen as the birthplace of microfinance, which makes it a particularly pertinent context to study the performance and implementation strategies of MFIs. The spread of microfinance in Bangladesh from the 1980s until 2005 was unprecedented and far greater than in any other country. Bangladesh has since become the second largest microfinance market in the world (after India) with 22 million active borrowers in 2016. By 2013, 60 per cent of households in rural Bangladesh reported having taken microcredit at some stage, with credit-based group loans repaid weekly dominating the market (Osmani, 2016). Bangladeshi MFIs including ASA, BRAC, BURO Bangladesh 1 and Grameen Bank 2 account for 46 per cent of the total number of credit officers and 47 per cent of the total number of MFI offices in the whole of South Asia (Khamar, 2016). With a large body of scholarship focusing on measuring outreach and impact in Bangladesh, fundamental questions about the institutional performance and implementation mechanisms of these 'institutional hybrids' remain largely overlooked.
Following this introduction, the first section of the article critically examines the conceptualization of social performance within microfinance, arguing that insufficient attention is given to the implementers and processes of implementation. The next section demonstrates the value of using existing scholarship on organizational theory, institutions and implementation to better understand everyday practices in microfinance (de Certeau et al., 1980;Lipsky, 1980;Mosse, 2005;Smith, 1987). The third section presents the methodology, fieldwork site and original data used and introduces the case study of ASA, renowned for being one of the most cost-efficient MFIs in the world. The subsequent sections constitute the empirical body of the article. They introduce the concept of practice drift; analyse the context within which ASA built its financial discipline; explore the ruling relations that organize everyday relationships; and show the significance for performance of the discretionary power of credit officers and branch managers. The conclusion summarizes the article's main contributions and reflects on their wider significance for the microfinance industry.

SOCIAL PERFORMANCE IN MICROFINANCE PRACTICE
Policies for international development at the end of the 20 th century were characterized by a free-market ideology. The Washington Consensus, in particular, played a central role in embedding international development policies in neoliberal frameworks of privatization, liberalization and deregulation alongside a rolling back of the state (Kamat, 2004). Designing propoor market-based innovations became a rallying call for the private sector to sell products and services to the 'bottom of the pyramid' (Karnani, 2007: 94;Prahalad, 2005). In this context, microfinance, or more accurately microcredit, was considered a revolutionary tool (Morduch, 1998(Morduch, , 1999Otero and Rhyne, 1994). As a result, until the mid-1990s, to meet donors' and stakeholders' interests, much of the literature and research on microfinance focused on supporting MFIs to become financially viable (Hulme and Moore, 2007;Woller and Woodworth, 2001 By the late 1990s and early 2000s, however, multiple studies were warning of the risk that MFIs' mission could shift towards prioritizing profits over poverty-reduction objectives (Woller, 2002a(Woller, , 2002b. Although financial performance is generally well embedded within MFIs, as it enables them to demonstrate their ability to mitigate the financial risks associated with serving the poor, the same is not always true for social performance (Copestake, 2004). The latter is more ambiguous and dependent on the good intentions of senior management; in other words, it is self-regulated (Copestake, 2004(Copestake, , 2007Hashemi, 2007). Because it was often assumed that providing loans to the poor would in itself systematically generate positive outcomes, the perceived need for having tighter social-performance management and monitoring within MFIs was relatively weak.
A dominant explanation for low social-performance achievements of MFIs, as mentioned in the introduction, was that faced with trade-offs between the financial and social mission, MFIs favoured financial performance -the so-called 'mission drift'. Some scholars feared that this drift would weaken the rationale for MFIs to serve the poor (Cull et al., 2007). The opportunity cost of providing products and delivering services to poorer customers would be outweighed by the potential benefits associated with serving better-off households, thereby affecting MFIs' depth of outreach (Schreiner, 2002). Researchers, practitioners, investors and donors use large average loan size as a proxy for mission drift (Armendáriz and Szafarz, 2011;Aubert et al., 2009;Cull et al., 2007;Fouillet and Augsburg, 2010;Frank et al., 2008). A large average loan size, in theory, indicates that an MFI's social mission has drifted because it means the institution has moved away from serving the vulnerable poor in order to mitigate the risks and costs associated with serving this particular segment (Mersland and Strøm, 2010: 29). Outreach to the poor is therefore central to the mission-drift definition and argument.
The proliferation of social-performance initiatives in the early 2000s reflected both rising doubts regarding MFIs' potential to reduce poverty (Malhotra et al., 2002;Zeller et al., 2003) and increasing pressures from investors and funding agencies to establish sound evidence of impact (Cochran, 2007: 451). Under the umbrella of the Social Performance Task Force (SPTF), 3 a number of organizations have, to a substantial degree, reached a consensus on their approach to social performance (Sinha, 2006), defining it as 'the effective translation of an institution's social goals into practice in line with 3. SPTF is a membership-based body that aims to advance understandings and practices related to social performance within MFIs.  accepted social values' (Hashemi, 2007: 3;IFAD, 2006;Woller and Brau, 2004). The aim of the SPTF and other such initiatives is to improve MFIs' responsiveness to clients' needs and increase their accountability to multiple stakeholders in the sector (Doligez and Lapenu, 2006). To date, approaches to social performance have, however, been conceptually and practically limited for three main reasons. First, they offer a narrow and deterministic framework for understanding the diverse means through which MFIs interact with clients. Their cause-effect approach conceptualizes social performance as a linear process that originates in MFIs' mission and is directly translated through internal systems and activities into outputs, outcomes and impact (Jacquand, 2005). This linear structure-conductperformance paradigm (Figure 1) was adapted from industrial organizations and applied to MFIs (Zeller et al., 2003). This perspective potentially ignores the significance of processes of implementation and roles of implementers in determining outcomes and impact. Second, a major pitfall for social performance is that the standards set by such initiatives are very often not institutionalized within MFIs. Furthermore, when they are, they often rely on MFIs' self-assessment processes whereby managers self-administer questionnaires evaluating their own social performance at their own discretion. This assumes that they have access to reliable information and, importantly, that sufficient incentives are in place for staff members to report accurately on their performance. Third, developing universal monitoring tools and targets to prove and quantify social performance (number of poor clients, savings and outstanding loan amounts) can distort the way microfinance is practised and obscure understandings of context-specific processes of impact (Zeller et al., 2003).
By their nature, monitoring data provide little analytical insight into how microfinance's social mission is implemented or if it is indeed implementable. Social-performance monitoring approaches and tools generally are not designed to understand processes of impact but to measure it. Their purpose was to hold institutions accountable to donors and the wider public with regard to their financial and social mission. So far the body of evidence on microfinance's impact is mixed, with outcomes varying from one context to another, one institution to another and one study to another. This begs the question: what determines the institutional performance of MFIs?

THE STUDY OF IMPLEMENTATION AND IMPLEMENTERS
This section looks at the relevance of important strands of literature on institutional and organizational theory to advancing the analysis of institutional performance within microfinance. The work of Lipsky (1980) andde Certeau et al. (1980) recognized the importance of studying processes of implementation and problematizing the roles of implementers within institutions. When we apply this to the study of microfinance and its institutions, we do so through a lens that considers the relational positioning and political economy of implementers and their influence on the implementation process and, therefore, on outcomes. As argued in the previous section, this understanding is lacking in existing frameworks for understanding microfinance's social performance and impact.
Policies mobilize institutions that organize and govern practices. At the same time, policies are implemented through multiple channels, which fragment the policies into activities across different institutional actors, thereby potentially distorting the purpose of these policies (Shore and Wright, 1997). Smith (1987: 161-65) points to the difficulties of studying institutions as unitary organizations, arguing that it is people's everyday actions and patterns of 'ruling relations' stemming from structures that determine the outcomes of policies. Ruling relations -defined as an 'internally coordinated complex of administrative, managerial, professional and discursive organization that regulates, organizes, governs and otherwise controls our societies' (Smith, 1999: 49-50) -shape institutional practice in the field. These forms of power relations are often diffuse, pervasive and discursive. They are also mediated, through institutional texts, administrative procedures and policies and, crucially, through social relations. Smith's approach to institutional ethnography identifies institutional actors according to the meaning they give to policies and everyday happenings within an organizational structure. Ruling relations, when mapped locally, provide a framework for understanding wider translocal and systemic power relations within institutional bureaucracies (Smith, 1997: 38-39). The study of ruling relations and interactions can help unearth the systemic dimension of diffuse, personal, implicit or informal institutional practices (Dépelteau, 2008;Smith, 1987). Work on organizational theory by Jepperson (1991)

offers analytical insights into what activities and relations form institutions and what institutions reproduce patterns of relations:
An institution represents a social order or pattern that has attained a certain state or property; institutionalization denotes the process of such attainment. By order or pattern, I refer, as is conventional, to standardized interaction sequences. An institution is then a social pattern that reveals a particular reproduction process. . . . [I]nstitutions are not reproduced by 'action' in this strict sense of collective intervention in a social convention. Rather, routine reproductive procedures support and sustain the pattern, furthering its reproduction. (Jepperson, 1991: 145) As this quote indicates, the outcomes of policies must be understood as also determined by the routine reproduction of implementers' patterns of procedures and relations. The work of Lipsky (1980) demonstrates the significance of these dynamics through an examination of the role that street-level bureaucrats play in shaping institutional practices and policy outcomes. Lipsky argues that the power of discretion of street-level bureaucrats allows staff to absorb the misalignment between standardized public policies and local reality by remaking policy at street level. His argument, like that of Smith (1987), challenges the conceptualization of policy as a predictable, apolitical and linear framework implemented by institutional blueprints rather than humans subjects. Anonymous human subjects, 'common heroes', employ everyday tactical practices and ordinary relations to subvert and 'fool' the dominating power and order (de Certeau et al., 1980: 3-4). This framework emphasizes the need to better scrutinize the significance of the patterns of practices, behaviours and 'ruling relations' to understand institutions, implementers and processes of implementation that can be applied to microfinance.
The argument made here to scrutinize more closely the role of implementers resonates with previous studies within international development. Implementation paradigms have been examined in detail by Mosse (2005). Development policies, he argues, are shaped by the exigencies of organizations as they shape the system of everyday rules and codes, goals and interests that organize implementation. Studying the 'everyday' points to pressures and incentives that condition and motivate staff members. Similarly to scholarship on street-level bureaucracies, the study of the everyday enables problematizing the position of implementers in linear and top-down implementation blueprints. In terms of understanding performance, it opens up the possibility for a bottom-up exploration of delivery mechanisms through studying key human actors.
In the field of microfinance, studies conducted in Asia and Africa have shed an empirical light on MFI loan officers, recognizing the multi-dimensional nature of fieldworkers' functions (Ahmad, 2002a(Ahmad, , 2002bGoetz, 2001). Such work draws attention to studying the 'faces' of the MFI officers who are responsible for delivering and managing 'the social good' (Siwale, 2013: 2), and who come lowest in the MFI organizational hierarchy. Aligned with these arguments, Dixon et al. (2007) have demonstrated how loan officers' behaviour substantially and sometimes unexpectedly influences the outcome and impact of many credit programmes. The above analysis, I believe, constitutes a strong rationale for the focus and methodological approach outlined in the following section.

THE FIELDWORK
The analytical approach adopted in this article focuses on the study of institutions through the lens of implementers, in line with the studies by Mosse (2005), Lipsky (1980) and Smith (1987), who explored the complexity of examining everyday institutional practices and relations between policy, institutions and employees. The research on which this article is based used an abductive research strategy combining a deductive and an inductive approach (Dubois and Gadde, 2002: 559). The logic of this approach is to use an exploratory research methodology from which theorization and further exploration and focused observations are used iteratively. Using mixed methods, I studied the perceptions and experiences of microfinance of diverse groups of people in rural villages and at an institutional level over a period of 12 months between 2010 and 2011. Given the sensitivity of the information collected, data triangulation through different sources of information and means of data collection was essential.
The research was conducted in Tangail district, the same district in Bangladesh in which, in 1979, Mohammed Yunus first piloted what four years later became the Grameen Model. It is generally recognized that Tangail is one of the districts that is most densely served by MFIs (Armendáriz and Morduch, 2005: 128). All the major indigenous Bangladeshi microfinance organizations have branches in Tangail and in the Upazila 4 where this research was conducted. This Upazila does not flood frequently; it has extensive irrigation facilities and extensive cultivation of high-yielding varieties of rice and other crops. Seasonality, however, affects rural livelihoods in Tangail, as in most districts in Bangladesh. Thus, the research site selected captures the impacts on livelihoods of seasonal flooding, which affects most of the country, while avoiding over-representing livelihoods that are extremely vulnerable to environmental hazards (flash floods and river erosion for example). This enables investigation and comparison of ways in which MFIs perform in an area that is reasonably typical of rural settings in much of Bangladesh. The data presented must therefore be interpreted and understood in the context of competitive rural markets in which MFIs including ASA, BRAC, Buro, Grameen Bank, and many smaller local NGO-MFIs operate.
The research was divided into two stages. In the first stage, a survey was administered to 490 rural households, covering four villages in a district where the density of MFIs is particularly high, to determine levels of poverty, life trajectories and microfinance membership. Through analysing these data, three categories of household were delineated based on their livelihood status: improving, stable and declining. Six focus group discussions were then conducted with each category of household, aimed at identifying whether the institutional practices of MFIs could partially explain clients' livelihood status. The purpose was to identify common patterns of experiences. Furthermore, in-depth interviews with nine former clients of MFIs aimed at exploring whether quitting microfinance could be associated with an indicator for improvement in clients' livelihood. During this period, extensive participant observation was conducted in each of the research sites, within communities and institutions alike. Moreover, non-clients of microfinance and authoritative figures within the local community (police officers, village leaders, imams and union chairmen) were also interviewed about microfinance and MFIs' practices. On the basis of this, two MFIs were identified as the best-and worst-performing, as judged by the clients themselves.
The second phase of the research consisted of institutional ethnographies of the two selected MFIs. The purpose of this was to understand whether the perceptions and experiences of clients and former clients related to the institutional structures of these MFIs. During this period, extensive participant observation and informal interviews were conducted with field-level staff. More formally, access to field-staff members who interact directly with clients was informally negotiated with four branch managers who were themselves subsequently interviewed. A self-administered questionnaire, which included closed and open-ended questions on their experiences of delivering microfinance to clients and their personal relationships with their employing institutions, was completed by 36 credit officers. Thereafter, 12 semi-structured interviews were conducted with individuals higher up in the institutional hierarchy, including regional and district managers based in rural and semi-rural settings and senior managers located at the institutions' headquarters. This article uses the data from ASA, one of the two institutional case studies, to explore the implications of institutional practices for the relationship between credit officers and clients. ASA was selected as a case study based on consistent negative accounts from the community regarding ASA's institutional practices. The impact of the identified practices on clients will be the focus of a future article.
These interviews and surveys sought to understand the social positioning, power relations and inner politics within the institution. To do this, data from non-clients were collected before data on clients, data on clients before loan officers, and data on loan officers before managers and senior staff members. This was important to maintain independence from the perceived hierarchical relations and to establish a rapport of trust with participants. By shifting from a traditional top-down process of data collection, this reorganization was particularly valuable to investigating how power relations were constructed and reproduced by actors and subjects of microfinance. Further methodological details can be found in Maîtrot (2014). All interviews were audio-recorded and transcribed from Bengali to English. Names of individuals and the exact location of Upazilas and villages have been anonymized to protect respondents' identities.

UNDERSTANDING PRACTICE DRIFT
The case study of ASA, one of the most cost-efficient MFIs in the world, focuses on social performance by analysing the processes of implementation and the everyday roles and relations of implementers from a perspective that sees policy as translating into sets of practices in a non-linear fashion. The empirical findings presented below demonstrate that a set of informal practices has been developed and reproduced by field-level staff and has become institutionalized within ASA. These practices determine ruling relations at the field level by enabling staff practices to drift in a way that undermines the institution's social performance. This is enabled by a heavily decentralized structure of governance and management, serving a low-cost expansion strategy.
The dynamics described in the following sections are conceptualized here as a 'practice drift'. This both builds on the concept of 'mission drift' and points to its limitations, by problematizing how microfinance is practised at the field level and by better explaining its mixed results. The concept of a practice drift shares a central claim of the mission-drift argument: that there is a relationship between the commercialization of microfinance, the way in which microfinance is practised and outcomes for clients. It differs, however, in four important respects. First, mission drift refers to an intentional, top-down change in MFIs' strategy (to respond to commercial pressures), whereas practice drift argues that without any alteration in claims about the mission of microfinance, practices in the field can shift in a way that contradicts its social mission. The dynamics observed at the field level appear to be disjointed from the social mission of microfinance, which has remained unchanged, despite commercialization.
Second, a key characteristic of mission drift is the argument that MFIs move away from poorer clients in favour of less risky clients and larger loans to better-off households (Copestake, 2007;Mersland and Strøm, 2010). Practice drift places emphasis on a set of informal practices through which MFI field staff achieve financial targets. This includes the opportunistic targeting and retaining of poor and vulnerable households. MFIs' depth of outreach therefore seemingly remains aligned with the social mission of microfinance: serving the poor.
Third, the notion of practice drift points to loan officers and managers using their discretionary power to achieve targets. This often entails a drift in the behaviour and attitudes of field-level staff subjected to everyday targets and the need to maintain good relationships with colleagues and superiors. Loan officers' practical and immediate interests lie in limiting time-consuming procedures and in disciplining clients. Spending time with clients (assessing their creditworthiness, repayment capacity and investment purposes) is therefore devalued and discouraged as it represents a high opportunity cost for managers and credit officers who are not rewarded for it. They are encouraged to be inflexible about repayments and to put financial targets before social achievement, which they do through sometimes violent and abusive means.
Fourth, the notion of mission drift does not capture the opportunities for implementers to move away from formal practices regulated by MFIs and underestimates the significant role played by implementers in outcomes. The practice-drift argument highlights that it is at the margins that the tradeoffs between the social and financial missions of microfinance are being brokered by fieldworkers (Siwale, 2013;van den Berg et al., 2015). In highly decentralized structures this can remain unnoticed or ignored by senior managers.
In summary, then, while the practice-drift and mission-drift arguments both recognize the impacts of commercialization practices on microfinance's outcomes and impacts, practice drift demonstrates how field-level structures and management systems create the conditions for practices to drift, and social performance to suffer, without there having been a deliberate, topdown shift in the mission of the MFI. The mission-drift argument describes a top-down decision-making process, in which changes in mission lead to a change in field practices. The practice-drift argument points to a potential displacement of that decision making down to the branch level. The case study of ASA will demonstrate that the pursuit of low-cost, streamlined delivery creates conditions within which the pursuit of social performance is actively discouraged or even punished at branch level. Under such conditions, to 'perform well' within ASA means to meet specific targets that are assumed by senior managers to reflect social performance but which in fact serve the financial performance of the branch. Evidence from the case of ASA highlights this contradiction between these targets' intended purpose and what they achieve in practice.

ASA'S FINANCIAL PERFORMANCE: A SKILFUL BUREAUCRACY
This section analyses the means through which ASA built its good financial performance. By doing so it demonstrates how rigid financial targets and incentive structures coupled with continuous and pervasive low-cost monitoring systems can effectively serve financial performance. This section contextualizes the evolution of ASA within Bangladesh's political and development history.
In the late 1980s, international aid donor agencies and foreign governments became increasingly pro-market, opposing the Bangladeshi government's socialist political-economic thinking (Rutherford, 2009: 37-52). During this period, Bangladeshi NGOs emerged as a solution to the perceived decline in the efficacy of the government in reducing poverty (Lewis, 2004;Shamsuddoha, 2003;White, 1999). With direct support to NGOs rising from 6 per cent of total aid disbursed to Bangladesh to 18 per cent between 1990 and 1995 (Devine, 2003: 229), NGOs became crucial implementers of state services to citizens through dense rural and urban networks (Zaman, 2004). Wood (1996) characterized this phenomenon as the 'franchise of the state'. As a consequence of the growing degree of competition among domestic non-state actors (Ghosh and Van Tassel, 2011), many NGOs sought to build financial self-sufficiency (Fernando, 2006) in an attempt to maintain a degree of autonomy from the political patronage of donor agencies and governments (Lewis, 2017). This domestic political economy provided an auspicious terrain for NGOs to use microfinance as a strategy to expand their outreach (Wood, 1996;Wood and Sharif, 1997). By targeting poor women, microfinance was also able to appropriate and capitalize on the women-indevelopment paradigm advocated and supported by multilateral donors and Western development agencies (Karim, 2008(Karim, , 2011. ASA emerged in 1978 pursuing a radical social-mobilization agenda, but this collective-action agenda provided limited results. In line with the wider dynamics of 'development as delivery' (Rutherford, 1995: 70) in the development sector in Bangladesh, ASA shifted to microcredit and became one of the country's most skilful administrators of financial services and products to the poor (Zaman, 2004). Its rapid and cost-effective scaling-up capacity, and its administration of simple and standardized operational procedures on a massive scale with a 'vision unmatched in its clarity and relentlessness' (Morduch in Rutherford, 2009: ix), gained international recognition.
ASA's institutional model and strategy of expansion evolved over the years to adjust to the levels of borrowers' delinquency, especially from 2008. Structurally, ASA is decentralized and has adopted standardized and low-cost human resources policies to maintain high financial performance (MacDonald, 2012). It was this model that saw ASA named as the world's leading MFI by the Microfinance Information Exchange's MIX Market report in 2005, 5 and the world's best MFI by Forbes in 2007 (out of 641 microfinance service providers worldwide) and led to its winning The Financial Times and International Finance Corporation's 'Banking at the Bottom of the Pyramid 2008' (out of 129 institutions across 54 countries) (ASA, 2012). Not only has ASA grown in Bangladesh, its model, seen as an exemplar of efficiency, has been replicated by other MFIs in India, Nigeria, Pakistan, Sri Lanka, Philippines, Uganda, Tanzania, Myanmar, Kenya and Ghana.
This NGO achieves impressive financial efficiency. In Bangladesh it multiplied the number of active borrowers more than 3.5 times within less than a decade (from 2001 and 2008) and the size of its portfolio more than 200 times in 20 years -from US$ 2.97 million in 1992 to more than US$ 608.08 million in 2012 (MIX Market, 2013). In 2011, ASA's financial self-sufficiency reached 118.32 per cent and operational self-sufficiency 182.48 per cent while its operating cost ratio has been halved (from 21.89 per cent in 1992 to 9.82 per cent in 2012) (ibid.). The NGO declared itself donor-free in 2011. It envisaged disbursing US$ 2.5 billion in loans among 6.6 million clients over 2015-16 in Bangladesh only , and overachieved this target 5. A non-profit based in Washington, DC, the Microfinance Information Exchange (MIX) provides market data and intelligence on financial service providers catering to the poor. Their MIX Market platform consolidates organizational data on financial and social performance in the microfinance industry.
by actually disbursing US$ 2.68 billion to 7.4 million clients (ASA, 2016). Although ASA has developed a large set of products and services (savings, credits and life insurance among others), 96 per cent of its product portfolio depends on one credit-based product -the 'primary loan' at 15 per cent flat interest rate (29 per cent annual percentage rate), aligned with the ceiling set on MFIs in 2009 by the Microfinance Regulation Authorities (MRA) 6 (ASA, 2010; MIX Market, 2013). The rates used by MFIs are expected to be higher than those of commercial banks due to the size of loans and the high transaction costs. ASA follows a growth-focused strategy and uses target-based financial monitoring systems spread across all levels, from branches to regional and district offices. According to interviews with regional and district managers, it is regional managers who balance surpluses and deficits across the region. They do so by moving financial resources from one branch to another in a way that ensures sufficient liquidity for branches to function properly and respond to demand. Minimal costs are borne by the headquarters of ASA as the branches assume the human resource and administrative costs of their staff members. Branches function as financially self-sufficient profit centres that follow rules standardized across the MFI. Regional managers allocate funds across branches (between four and six) by assessing the reports on projected loans and savings and anticipating planning for the liquidity needs in each branch of their region. District managers regularly meet regional managers to discuss previous and current lenders' figures, outstanding amounts, recovery rates, number of clients, problems faced and measures taken after submitting the monthly report. Social performance in terms of poverty reduction, economic empowerment or well-being of clients is rarely, if ever, discussed. There are no incentives or mechanisms within ASA for field officers to measure, report or represent the interest of clients. The issue that comes closest to social performance that is sometimes discussed is 'conflict' between clients and branch level staff, and these conflicts are only considered significant when they threaten the financial performance of branches.
Regional managers reported experiencing pressures from district managers to improve the performance figures of the branches under their supervision. One regional manager explained that district managers (respectfully referred to as 'the sirs') visit branches to 'motivate' the staff and write a review which 'depends on the profit. In 2008, the sir gave a good review but in 2009 and 2010 the profit was less and sir sent a circular to motivate us to improve ourselves, but this year it is better'. 7 Time use and profits are central to branch performance. One branch manager explained to me 8 that when 23 clients out of 100 cannot repay, they manage to keep their financial 6. The MRA was established by the government of Bangladesh in 2006 to monitor and supervise microfinance operations of NGO-MFIs. 7. Interview, ASA regional manager, Tangail District, April 2011. 8. Interview, ASA branch manager, Tangail District, April 2011. performance indicators high and generate profits if they disburse more and bigger loans rapidly: 'If the speed at which we can disburse loans is fast then we can make profits, but if it's slow then losses are faced. Initially our total loan amount outstanding was low but now it's about BDT 1.5 crore '. 9 Branch managers explained how they keep human resource management costs to a minimum within their branches. It is the branch staff members themselves who handle recruitment and the training of loan officers. Training at ASA is mainly informal and occurs on the job through a process called 'one-teaches-one-learning'. There is no formal training course or centralized training centre. District and regional managers provide information to new recruits for one or two days before they are sent to the field for a week to observe how senior colleagues at branch level interact with clients, report in passbooks and carry out other financial management procedures. The following week, new recruits apply what they have learnt under the supervision of that same colleague. This decentralization achieves several purposes. It enables the MFIs to avoid the costs and time associated with formal staff training and ensures a continuum in institutional practices. It also means branches are directly responsible for their new recruits' performance.
Branches' organization is standardized through a book developed by ASA, known simply as the Manual. The Manual explains institutional policy, the rules and protocols regarding staff promotion, recruitment, transfer, branch default, staff misconduct and the remuneration scale. To ensure and safeguard financial integrity, ASA applies a policy of regular staff transfer. Loan officers who work for many years in the same villages become familiar with clients and local elites and could establish informal relationships and 'deals' with them. Top managers interviewed reported that there was a risk of loan officers' financial performance declining as a result of informal and personal relations limiting their capacity to enforce repayment. By rotating loan officers between branches every three years, ASA aims to reduce fraud and maintain financial discipline.
One has to recognize ASA's commendable capacity for financial management and monitoring. With approximately 14,000 loan officers handling cash daily, the scope for misappropriating money is considerable. The Manual facilitates managers' decision making, reduces their discretionary power on human resource issues and financial management and decreases opportunities for fraud, money misappropriation and mismanagement. Branch managers verify daily transactions that loan officers record by entering data into ASA's computerized system, which also enables higher-level managers to monitor liquidity. ASA's monitoring system is so efficient that it reportedly detects fraud and default within 1 to 15 days. Beyond tight monitoring, administrative sanctions and penalty systems safeguard ASA's financial performance. Mistakes and faults have negative repercussions for multiple 9. That is BDT 15 million or approximately US$ 210,000 at 2011 exchange rate. employees throughout the hierarchy. Branch managers, regional managers and district managers oversee the loan register and are personally fined when mistakes are found. Financial sanctions vary according to their position and proportionally according to their salary and size of the error but can represent up to 10 per cent of monthly salary. Moreover, bad financial performance, mistakes and transgressions are noted in the staff members' personal files and are likely to have negative effects on promotion prospects.
In contrast to most organizations in Bangladesh, ASA's human resource management system, especially recruitment procedures, salaries and internal promotion, is perceived as meritocratic and performance-based, rather than nepotistic or clientelistic. This finding is in line with the arguments of Ahmad (2002b) and Uphoff (1996) that the culture and management of large-scale NGOs in Bangladesh has professionalized around strict rules and policies on promotion or transfer. Branch managers and regional managers reported starting their careers within ASA as loan officers and getting steadily promoted. ASA's Executive Vice President explained: 'In ASA we have only one entry position which is loan officer. . . . We don't directly recruit branch manager or upper-level staff. Gradually we promote them to senior positions'. 10 Interviews with staff members suggest that decent wages and transparent promotional paths were a prominent part of their reason for joining ASA. They reported being satisfied with the salary structure in place and motivated by the career prospects within the NGO. 11 As mentioned above, in order to avoid potential conflict of interests at the frontline of microfinance delivery, ASA applies a strict transfer policy according to which credit officers switch branch every three years. This mechanism aims to ensure high repayment rates, reduce fraud and maintain financial performance. In theory, frequent and automatic transfers and punitive transfers in cases of low financial performance provide strong incentives for credit officers to maintain a professional distance from their clients.
In practice, staff performance is assessed through an evaluation of financial performance combined with managers' recommendations and test results. Staff members compete among themselves to be rated highly according to their transactions records and financial achievements. It is not surprising therefore that in the staff survey, 47.6 per cent of credit officers at ASA reported being motivated by bonuses and rewards such as 'thank you' letters. 12 When I asked what data were used to evaluate employees, one branch manager at ASA explained: From the beginning of each month, each employee's number of new loan clients, the loan collection and their outstanding amount is calculated each and every day on the computer. At the end of the month we can get an idea about their performance and take steps accordingly.
You must note that our [the branch's] earnings and success depends on these loan officers' performance levels and evaluation. 13

EVERYDAY RULING RELATIONS: TEXT AND TARGETS
Decentralization, tight financial monitoring and low-cost human resource management are efficient devices that secure financial performance and growth for ASA. Within ASA the configuration of these ruling relations is a great source of stress and strain for the frontline staff involved in implementing microfinance and enforcing financial performance.
Financial performance is a priority of the management and organizational culture at every level of ASA. The head office communicates non-negotiable financial targets to branch managers who are responsible for anticipating the demand for loans (loan uptake). Each loan officer is assigned individual financial targets, with a determined number of clients to recruit and retain and a set loan amount to lend out to clients. Monthly disbursement targets are set twice a year in six-month reports.
It is important to note that the majority of loan officers sleep in ASA branch dormitories. This is particularly common for men, while women generally have the choice to live at the branch or outside with their family. All the five branches studied displayed, in the main common room where loan officers eat and rest, organizational rules and posters exhibiting yearly financial figures and, on blackboards, daily and weekly cash flow targets and achievements. The blackboards highlighted three daily targets, including the number of borrowers visited, the daily disbursement amount and the daily repayment to be collected. The everyday routine of loan officers is to write the amount to be collected that day on the board in front of their colleagues, then depart for the villages by 8 am to collect kisti (repayment) before returning to the branch by 2 pm, filling out daily financial transaction reports together and writing the amount they have collected on the board beside the target amount.
This institutional practice is not temporary or specific to the locality, but constructs translocal institutional practices. As argued by Smith (2001), it is the textuality of ruling relations that fix those ruling relations, regardless of variabilities of place, time and people. In large-scale organizations such as ASA, the 'text' -here the Manual, the posters and the blackboard -organizes the way in which employees are socially connected. By its presence, it transcends the social reality to create formal, standardized and acontextual relations ruling individual performance. Its public display, highlighting the gap between targets and achievement, is a powerful tool to reinforce financial discipline. This is especially effective as individual financial performance determines overall staff performance and promotions. Achieving these set targets relies on the loan officers' ability to recover all their loans.
What if targets are not achieved? Underachievement is punished through financial penalties, pressures and sanctions put on branch managers and loan officers. Regular and abrupt staff transfers and direct financial fines are designed to enforce timely repayment collection. When loan officers are unable to collect instalments, they are warned about the consequences 'bad repayments' have for the institution and strongly rebuked by their managers (sometimes in public). The following written accounts from loan officers illustrate these deeply stressful institutional patterns: The worst part is that in every position the subordinates suffer mental harassment from superiors. 14 The rules in ASA and the mental harassment faced by the employees are the worst part of ASA. 15 If their repayment performance does not improve, employees get a written warning and must pay a fine. Moreover, their yearly salary increment can be cancelled and their holidays go unpaid. Staff survey results and written statements reveal credit officers' stress and anxieties within the institution. Half the credit officers surveyed reported that managers get angry very often in the organization and 76 per cent reported being motivated by the fear of punishment and exclusion. The fear of being socially excluded or in conflict with colleagues is reportedly strong, especially given that most credit officers live together at the branches.

PERFORMING IN THE FIELD
Building on findings presented above, I argue here that ruling relations embedded within the institution shape the subjectivity, agency and everyday practices of frontline staff in a way that achieves financial performance but undermines social performance. Daily client recruitment, top-up loans, follow-up procedures and repayment collection practices tacitly drift at the field level.
In top-down MFIs where management is decentralized, fieldworkers subjected to high financial-performance pressures are often marginalized within the organizational structure of their large-scale institution (Agier and Szafarzy, 2010). In ASA, the structure of the field administration and the high levels of decentralization mean that credit officers are often isolated when they are in the field, without close supervision from branch managers. As such, they have considerable discretionary power that they are incentivized to use to achieve their financial targets. This shapes practices and attitudes 14. Self-completed questionnaire, ASA credit officer, Tangail District Branch, March 2011. 15. Self-completed questionnaire, ASA credit officer, Tangail District Branch, March 2011. towards clients in ways that contradict microfinance's stated social mission, theory of practice, and the formal policies of ASA itself.
In line with MacDonald's (2012: 102) findings, clients and non-clients interviewed described ASA as an MFI that lends money 'easily'. Lowering standards for selection and follow-up of clients allows loan officers to meet their short-term financial targets. Poor client selection refers to the targeting of households that are likely not to have the capacity to invest in income-generating activities and who are likely to use the loan for immediate consumption. The geographic location of such clients allows loan officers to reduce the time dedicated to kisti collection. Spending time with clients to explain the purpose and implications of loans is not incentivized within the institution. Some 66 per cent of loan officers reported that their clients 'do not understand the concept of credit', and often this was identified as a consequence of time constraints faced by credit officers. During five focus group discussions, clients explained that misreporting the intended purpose of the loan on application forms with the full knowledge of credit officers was common. Clients would state that loans were for 'business purposes' (bebsha kora, in Bengali), when in fact they were intended for a range of other uses, thereby enabling clients to access loans, and credit officers to lend money and recruit new borrowers.
One of the oldest tenets of microfinance is the feasibility of collateralfree loans to the poor (Morduch, 1999). The data suggest, however, that outside the conventional joint-liability setting, credit officers, to securitize credit transactions, exploit their discretionary power and make informal judgements of loan applicants' ability to access sources of repayment. Material and immaterial forms of what I call 'micro-collaterals' compensate for thorough, time-consuming screening and follow-up procedures. Assets (pots, pans, chickens, roofing material), MFI clientship and social connections (wealth of family and friends) are three prominent sources of liquidity that influence credit officers' decisions. This practice, also documented by Fernando (2006), Uddin (2013) and White and Alam (2013), enables credit officers to compel clients to maintain timely kisti repayments and mitigate their risk of underperforming.
Households with long-standing relationships with MFIs (former and current clients) described credit officers as commercial agents aiming to persuade any household to borrow increasingly large amounts from their MFI regardless of the intended (or actual) use of the loan or capacity to repay. They repeatedly said 'MFIs give and take money, nothing else' ('taka dai khali, taka nei . . . ar kicchu nai') to explain the interaction between clients and credit officers. According to clients, these sorts of practices are increasing. A former client of ASA and BRAC for 15 years explains why she quit microfinance: Banks [MFIs] give money to everyone, they don't worry about helping anymore; they only care about interest and repayments. . . . People misuse the money now and the officers do not check on them like they used to. The relationship was better then. . . . They only talk about money and instalments; before they were very light hearted. They would advise us about our mistakes but now it's nothing like this. 16 Many poor clients reported that credit officers pressured them to take up loans. This practice relates to both the recruitment of new clients and encouraging existing clients to take new and larger loans. This finding directly challenges the common perception that there is an unquestionable demand and need for formal credit, that top-up loans serve to make financial products more flexible for clients (Laureti and Hamp, 2011) and that large loan size indicates high social-performance achievement. Credit officers are reported to go door to door to 'persuade' households to borrow, and use forceful methods to make household members feel ignorant and imprudent if they do not borrow, or borrow larger amounts. Such a practice constitutes a form of hard selling (tsap dawa).
During an in-depth interview, Zoshim, who was a client of ASA for many years, claimed that once clients manage to repay their loan, credit officers force them to borrow larger amounts 17 regardless of their needs, income or ability to repay: 'Then at times they try to exert force. They knock down the doors and slam doors, such kind of pressure . . . they coerce us into taking loans. They say that if we do not take loans then they shall take inappropriate action and even violence'. 18 Some women clients reported that men were encouraged by loan officers to use their wives to access loans. Eight informal discussions and one focus group discussion with women clients indicated that women were sometimes violently reprimanded or threatened by their husbands when they refused to borrow from MFIs. On several occasions women reported that it was the credit officers themselves who suggested such violence.
There are a number of tactics loan officers use to achieve timely repayment from client households. 19 An analysis of these points to some of the everyday ambiguities and constraints branch-level employees face in attempting to achieve high repayments rates. In their accounts both loan officers and clients consistently reported that MFIs, in general, do not tolerate delays in kisti repayment.
Micro-collaterals, both material and immaterial, are used as leverage by credit officers in case of non-repayment. It is common that a client's husband, brother-in-law, father or friend is asked by credit officers to intervene and lend money to the client so the debt can be paid off. When needed, influential external parties are also often informally pulled in to these financial dealings by the MFIs. In the case study area, the head of the police, a Union 16. Interview, Parveen, former microfinance client, Tangail District, February 2011. 17. Having repaid one loan, clients are usually eligible for larger loan amounts. 18. Interview, Zoshim, former microfinance client, Tangail District, December 2010. 19. While there is no space to go into these tactics in depth here, a detailed analysis can be found in Maîtrot (2014).
Parishad member, 20 the village head and Upazila chairman explained that MFIs commonly use them as a means to pressure defaulting clients: they might visit the MFI clients' houses, issue notices for the MFI, mediate between loan officers and clients, and pressure clients' families and neighbours to repay kisti for them. The practice of 'unauthorized, though tacitly accepted, asset confiscations' identified by Cons and Paprocki (2010: 645) in another district of Bangladesh was reported by two of the six focus groups in the study area. Clients described valuable assets (such as chickens, ducks, chairs, pots and pans or tin roofing sheets) being seized by MFIs operating in the area and sold at the market to get sufficient cash to cover the kisti due on that day. Mobilizing material forms of micro-collateral enables credit officers to meet their daily targets. Clients and credit officers reported across multiple interviews in different study sites that even in the event of the death of a client's close relative (a husband, a son or a daughter), credit officers sat in clients' houses until they repaid their kisti. In some cases, this left the client without enough money to bury the body and pay for a funeral ceremony.
Another means to collect kisti from defaulting clients is to use their MFI savings as loan collateral to conceal clients' default. This practice was reported by the clients and later confirmed by branch managers as a system called 'savings withdrawal'. 21 This common practice often generates conflicts between loan officers and borrowers, who described being reluctant to save if their savings are used as loan collateral. These informal practices, when they become systematic, can often contradict the purpose of voluntary or mandatory savings (providing clients with a security buffer against shocks). To avoid such drastic measures, which would attract the attention of their manager, credit officers often try to mitigate problems of non-repayment amongst themselves. A quarter of ASA loan officers interviewed reported having repaid the money due for repayment themselves, or relying on colleagues who have sufficient liquidity on that day to maintain a steady repayment record. Loan officers would then report to their manager that the kisti had been successfully collected, later recovering the money informally.
In a dense and deeply competitive MFI market, collecting kisti can become a source of conflict between credit officers and clients. An ASA credit officer wrote: 'the organization is not ready to accept [a] delay in instalment. So we have to be inhuman and treat the clients in an inhuman way' to convince them to pay. 22 More than 70 per cent of credit officers surveyed reported that collecting repayments from clients was difficult and responses collected from clients, former clients, non-clients and MFI staff members depict relationships as hostile. Clients reported dreading the day when the 20. Each Union Parishad is composed of one elected chairman and nine elected members. 21. Interview, ASA branch manager, Tangail District, April 2011. 22. Self-completed questionnaire, credit officers, ASA Tangail branches, April 2011. kisti is due because if they 'fail' to repay on that day credit officers 'change their colours'. 23 This colloquial Bengali expression has strong negative connotations, referring to MFI employees' rapid mind-set change, to attitudes described as 'abusive', 'threatening' and 'publicly humiliating'. 24 All six focus group discussions conducted with clients reported having experienced or observed such practices. 25 During informal discussions, loan officers reported that managers encouraged and trained them (through 'one-teachesone') to collect kisti punctually. Survey results report that 47 per cent of credit officers admit having threatened clients to force them to repay and 12 out of 21 credit officer reports gave accounts of ASA's hard-line approach to client repayment. 26 Credit officers and managers exploit their discretionary power to enforce financial discipline and timely repayment. Some managers interviewed reported using religion to discipline clients, invoking verses of the Quran. A regional manager described telling a client: 'God will make you pay for this someday. . . . Because of this you will be cursed for life. Maybe I will not come to you again, but another Manager who will be in my position later will come for you once again. This will go on throughout your life. You will pay for this'. 27 As a result of credit officers' attitudes, some women clients said they feared being alone on repayment day. A woman interviewed reported that her credit officer made explicit sexual threats in public such as 'We will stay in the house today, make a bed for me!' aimed at humiliating her. 28 Another woman reported that her credit officer instructed her to go hang herself if she could not repay. 29 It was commonly reported that credit officers would make clients feel guilty about their medical, school and food expenses and advise them to reduce their medication, children's education and diet (Maîtrot, 2014).
An informal branch-level rule bans credit officers from returning to their office -which is also their home -without the expected amount. As a result, all tiers of staff, including branch, district and regional managers, reported returning to clients' homes at night or on Fridays (which is a holy day) when clients failed to repay. Molida, a former client who borrowed from Grameen, BRAC and ASA for three years, five years and one year respectively, reported that she stopping borrowing from all MFIs 10 years ago as she and her husband experienced regular disrespect and day and night harassment from credit officers. 30 Credit officers also report suffering from this rule: When I do not get an instalment then I inform my boss that 'sir, there is a problem in this house and they cannot repay today'. Then my boss orders me to sit in that house until my clients give the money. 'If you have to sit there throughout the night you will but do not come back without the instalment' he says. So if I leave without the money and I face this kind of mental and physical torture I feel like quitting the job. 31 The results from the survey conducted with ASA's credit officers show that although official office time ends between 5 pm and 6 pm, 76 per cent of them 'regularly' return to the office after 8 pm and 51 per cent 'regularly' return after 10 pm. Written statements from loan officers indicate that this practice often involves collective action from different members of staff: If I do not get an instalment I inform the manager. Then he comes with all the staff and we stay in the client's house up to 12 or 1 o'clock at night. And we are not authorized to enter the office without the instalment. Whatever happens I have to collect the instalment and then can go to the office. 32 This practice can also involve regional managers, one of whom reports: 'Yesterday [Friday] I went to seven such people who do not cooperate properly with us. We went in three groups, four people per group and one Team Leader in each group. We had target of going to at least 15 clients. . . . There are about 200 defaulters in total in this branch'. 33 Working during the night and on weekends to collect repayments is compulsory to avoid disciplinary measures such as personal financial sanctions and the loss of promotion prospects. It is also necessary to circumvent the negative collective implications that non-repayment could have for their branch, which would threaten their relationships with their manager and colleagues. Credit officers are strongly incentivized to solve problems by themselves, and do 'what works', since involving members of staff higher up in the hierarchy reflects negatively on their capabilities and often generates resentment among colleagues.
Despite this, clients often expressed empathy toward MFI credit officers. They sometimes justified credit officers' attitudes, claiming that they could not be blamed for these practices and that such outcomes clearly stem from institutional pressures. They stated that officers are 'simply following orders of the top officials', that they are 'scolded at work by their managers' who reportedly say 'terrible words to them in public'. 34 Some clients reported stories of temporary default that led to extreme situations for loan officers. Once a pregnant female credit officer came at night to collect the overdue repayment, begging the client to find a way to repay, saying she dreaded her manager's reaction if she returned to the MFI office empty-handed. The client was incapable of finding the money that evening and the client and officer spent the night under the same roof until the officer's waters broke. 31. Self-completed questionnaire, ASA credit officer, Tangail Branch, March 2011. 32. Self-completed questionnaire, ASA Credit Officer, Tangail Branch, March 2011. 33. Interview, ASA regional manager, Tangail District, June 2011. 34. Focus group discussion, stable and declining clients, Tangail District, February 2011.
The client had no means to take her to the nearest hospital and had to mobilize other ASA clients.

CONCLUSION
This article advances the concept of 'practice drift' to denote the development of everyday practices at the field level that undermine microfinance institutions' social performance. Joining the dots between findings emerging from institution-focused studies (MacDonald, 2012;Shekh, 2006;Siwale and Ritchie, 2011) and impact studies (Aoki and Pradhan, 2013;Attanasio et al., 2015;Dattasharma et al., 2016;Thibbotuwawa et al., 2012;Waelde, 2011) it sheds light on the unfolding of institutional practice in the specific cultural and organizational context of a non-profit NGO in rural Bangladesh called ASA. In so doing, it suggests an alternative way of understanding the diverse outcomes of microfinance and its varying impact on poverty reduction. The approach developed acknowledges and problematizes processes of implementation and the roles of implementers for institutional performance and microfinance's impact.
The analysis of implementation processes and of the power of implementers strongly challenges deterministic policy frameworks and demonstrates how frontline staff engage with everyday brokering activities to enforce financial performance. The study deconstructs common assumptions made about social performance and demonstrates that it is not naturally or systematically achieved through the provision of financial products and services to the poor. This suggests that depth and width of outreach are therefore insufficient and misleading proxies for social performance.
The use of ethnographic data collected at the village and institutional level permitted an in-depth analysis of the multiple everyday roles, ruling relations and experiences of clients, former clients and field-level staff members in microfinance activities. This approach enabled the study of relationships between systemic informal practices and institutional performance. Applying the work of Lipsky (1980), Jepperson (1991) andde Certeau et al. (1980), the article revealed some of the everyday tactics and routines that allow the street-level bureaucrats of microfinance institutions to use their discretionary power to serve their interests by achieving financial targets in a timely manner. The evidence presented in the last two sections demonstrates how informal practices such as forceful recruitment procedures, hard selling of larger loans, abusive and sometimes violent repayment collection practices are developed and replicated by field-level staff. These, I argue, are systemic and constitute a practice drift that contradicts the social mission of microfinance.
There are strong reasons to think that these everyday practices of negligence, violence and abuse through which policies are redefined are not 'micro' and specific to the context in which the study was conducted, but 'macro'. In other words, they are characteristic of the enabling organizational structures and management systems of commercial, standardized and low-cost models of microfinance implementation with insufficient social-performance monitoring and framework. The external validity of the practice-drift phenomenon is therefore likely to go beyond the specific context of the villages and institution examined here, and have bearing on contexts where similar low-cost models are operating and where analogous institutional practices have been reported (van den Berg et al., 2015;Takahashi et al., 2010). If microfinance is to achieve its social mission there is a need, as illustrated in this article, for new approaches and efforts to understand how commercialization shapes the practice of microfinance in the lives of its clients.