Making Gender Known: Assembling Gender Expertise in International Organizations

Abstract

In recent decades, gender equality goals have been adopted widely in global policymaking, creating a demand for specialized knowledge and evidence to support the design and implementation of gender equality policies. Bridging feminist scholarship on gender expertise and practice–theoretical literature on knowledge production, this article examines a knowledge production initiative of the World Bank, the Gender Innovation Laboratories (GILs). While research has examined the position of gender experts and the content of gender expertise in global governance, it has overlooked how knowledge about gender is produced. In this paper, we use a practice–theoretical approach—assemblage thinking—to study the practical work mobilized in the GILs to produce, maintain, and disseminate knowledge about gender inequality. Drawing on interviews with lab researchers, documents, and online materials, and focusing on the epistemic practice of impact evaluations, our analysis demonstrates the work invested in assembling them, such as forging alignments with and securing support among stakeholders, activating repertoires of expertise, and translating results into material objects. These practices produce gender inequality as a governance object, which is amenable to technical policy interventions, which facilitates certain forms of action to address it. Yet, they simultaneously silence more political solutions to gender inequalities.

Durante las últimas décadas, se han ido adoptado, en gran medida, objetivos para la igualdad de género en la formulación de políticas mundiales, lo cual ha creado una demanda de conocimientos especializados, así como de pruebas que apoyen el diseño y la implementación de políticas de igualdad de género. Este artículo estudia, mediante la unión de los estudios feministas sobre el conocimiento en materia de género y de la literatura teórico-práctica de la creación de conocimiento, una iniciativa de creación de conocimiento del Banco Mundial: los Laboratorios de Innovación de Género (GIL, por sus siglas en inglés). Si bien la investigación ha estudiado el posicionamiento de los expertos en materia de género y el contenido de la experiencia a nivel de género en la gobernanza global, la investigación ha pasado por alto cómo se produce el conocimiento sobre el género. En este artículo, utilizamos un enfoque teórico-práctico (pensamiento de agenciamiento, assemblage thinking) con el fin de estudiar el trabajo práctico realizado en los GIL para producir, mantener y difundir conocimiento sobre la desigualdad de género. Nuestro análisis demuestra, mediante entrevistas con investigadores de laboratorio, documentos y materiales en línea, y centrándonos en la práctica epistémica de las evaluaciones de impacto, el trabajo invertido en unirlas que incluye forjar alineamientos y asegurar el apoyo entre las partes interesadas, activar repertorios de experiencia y convertir los resultados en objetos materiales. Estas prácticas dan lugar a la desigualdad de género como si fuera un objeto de gobernanza susceptible de intervenciones de política técnica, lo cual puede facilitar algunos métodos de acción para abordarla. Sin embargo, al mismo tiempo, estas prácticas silencian más soluciones políticas a las desigualdades de género.

Ces dernières décennies, les objectifs d’égalité femmes-hommes ont été largement intégrés dans les politiques du monde entier. Cette tendance requiert des connaissances spécialisées, mais aussi des preuves pour appuyer la conception et la mise en œuvre de politiques d’égalité femmes-hommes. En mettant en relation les travaux de recherche féministes sur l'expertise en genre et la littérature théorique et pratique sur la production de connaissances, cet article s'intéresse à cette dernière par le biais d'une initiative de la Banque mondiale, les laboratoires d'innovation sur le genre (GIL). Bien que la recherche ait analysé la position des experts du genre et le contenu de cette expertise dans la gouvernance mondiale, elle a négligé la production des connaissances sur le genre. Dans cet article, nous utilisons une approche théorique et pratique, le concept d'agencement, pour étudier les travaux pratiques mobilisés dans les GIL afin de produire, conserver et transmettre le savoir sur les inégalités femmes-hommes. En nous fondant sur des entretiens avec des chercheurs en laboratoire, des documents et des contenus en ligne, et en nous concentrant sur la pratique épistémique des évaluations des conséquences, notre analyse met en valeur le travail induit par leur agencement, comme former des alliances avec d'autres parties prenantes et obtenir leur soutien, activer des répertoires d'expertise ou encore traduire des résultats en objets physiques. Ces pratiques transforment les inégalités entre femmes et hommes en objet de gouvernance qui se prête aux interventions politiques techniques et facilite la mise en place de certaines mesures pour y répondre. Toutefois, elles passent aussi sous silence des solutions plus politiques à ces inégalités.

Introduction

Over the past decades, gender equality goals and policies such as gender mainstreaming have been increasingly adopted and integrated in global policymaking and governance (Halley et al. 2018). This has created a demand for specialized knowledge about gender, and about the design and implementation of interventions that can effectively promote gender equality. As a result, gender expertise has emerged as a field of authoritative knowledge in global governance, and gender experts have come to constitute a distinct category of professionals in international organizations (Kunz and Prügl 2019; Kunz, Prügl, and Thompson 2019). A burgeoning literature on the politics of gender expertise has provided nuanced accounts of how gender experts navigate institutional and normative obstacles within international organizations (Ferguson 2015; Miller 2019; Altan-Olcay 2020) and discussed how integration into mainstream governance institutions shapes the content of gender expertise as it tends to depoliticize and instrumentalize gender equality (Chant and Sweetman 2012; Olivius 2016; Olivius and Rönnblom 2019; Prügl and Joshi 2021). However, less attention has been paid to the practices of producing knowledge and expertise about gender in international organizations. That is, how does the work of generating, maintaining, and disseminating knowledge about gender unfold in practice, and what are the effects on how gender inequality is understood and addressed? Further exploring this is important, given the long recognized authority of international organizations to make meaning and define the world through knowledge production (Barnett and Finnemore 2004).

Moreover, exploring the production of gender expertise in global governance is timely and pertinent, as recent years have seen an increase in dedicated initiatives specifically focusing on the generation of knowledge and evidence that can inform gender policies and make gender governable. In this article, we examine a paradigmatic expression of this development—the World Bank Gender Innovation Laboratories (GILs). The World Bank has historically taken a leading role in acting as a global node for knowledge production. Thus, besides being an influential donor, the bank is also a major producer of knowledge about global development problems (St. Clair 2006). Since the turn of the millennium, the World Bank has gradually integrated a focus on gender in its policymaking and knowledge production activities (Prügl 2017; Jones 2018), often justifying this with arguments presenting gender equality as “smart economics” (Chant and Sweetman 2012; Calkin 2015; Berik 2017; Esquivel 2017). Here, the GILs represent a more recent, systematic focus on knowledge production about gender. Over the past decade, five GILs have been established, which all focus on producing rigorous evidence about successful gender equality interventions, that can be used in policymaking by governments, development organizations, and the private sector (World Bank 2016).

In this analysis, we draw on a practice–theoretical approach—assemblage thinking—to study the practical work mobilized in the GILs to produce, maintain, and disseminate knowledge about gender inequality. Drawing on interviews with lab researchers and on documents and online material, we show how the production of gender expertise is dependent on the creation of suitable conditions for knowledge production, the translation of knowledge production into appropriate methodologies, and the packaging of knowledge in fixed dissemination formats. Our analysis thereby adds to existing literature on gender expertise in global governance through a practice-oriented analysis of how knowledge about gender inequality is produced, and how this contributes to rendering gender known as a so-called governance object, meaning a distinct object that is represented, problematized, and rendered governable through knowledge production (Corry 2013). This analysis adds new facets to ongoing debates about the politics and tensions around gender expertise as a potential means of feminist politics (Ferguson 2015; Bastick and Duncanson 2018; Kunz, Prügl, and Thompson 2019; Altan-Olcay 2022).

While the GILs constitute an illustrative example of the integration and production of gender expertise in global governance, they also exemplify a turn toward evidence-based governance in international development aid more generally. Over the past decade, international development aid has experienced a surge of practices that serve to make development work increasingly professionalized and evidence-based. Through practices and artifacts such as performance measurement and log frames, the development industry seeks to improve the possibilities for more rigorous monitoring. At the same time, practices that aim to generate evidence on results and impacts of development interventions have found their way into development work (Eyben et al. 2015). A central part of this turn toward evidence-based agendas is an increase in knowledge production on “what works” in development aid. Here, experimental evaluation methods from the medical sciences have become more common. Donovan (2018) charts this development to the rise of the “randomistas,” a collective of economists who contributed to making randomized controlled trials (RCTs) widely adopted in development aid. The appeal of this method is its purported ability to determine the causal impact of development interventions. The turn toward evidence-based governance and experimental methods thus promotes a belief in objective knowledge, which tends to obscure the performative and political character of knowledge production: how it does not merely respond to, but also constructs the meaning of, development problems.

This problem has been examined in a burgeoning practice-oriented literature in international relations (IR). The field of IR is not alien to the study of knowledge production, as primarily constructivist IR has contributed to theorizing the role of international organizations in producing knowledge. However, this research has a history of being actor-based, focusing on how particular actors exert their power on the production of knowledge through “agenda-setting, policy-making, implementation, enforcement, evaluation, and monitoring” (Robinson 2018, 424). As a response to this focus, research has begun to direct attention to the practical work of generating, maintaining, representing, and disseminating knowledge, as well as how these practices create epistemic objects and make them governable (Bueger 2015; Robinson 2018). This shift in focus is the result of the increasing impact of “practice theory” in IR. In contemporary IR research, practice theory is now widely used to analyze practices of ordering and stabilization, thereby constituting an alternative to the dominant focus on actors as the primary enforcers of order in the international realm (Bueger 2015; Bueger and Gadinger 2018).

This article contributes to an emerging research agenda that centers practices of knowledge production and explores how these make international phenomena knowable as objects of governing. Our mobilization of assemblage thinking introduces a new approach to analyzing knowledge production, which directs attention to the laborious efforts that go into constructing, maintaining, and legitimizing knowledge. Adding to recent analyses of phenomena such as migration (Robinson 2018), climate change (Allan 2017), and piracy (Bueger 2015), our analysis demonstrates how an increasing integration of gender concerns into global governance is bound up with practices of knowledge production that make gender and gender inequalities known and governable in particular ways. This analysis also makes possible gaining a deeper understanding of the silences and exclusions inherent in knowledge production. Our analysis, we argue, contribute to bridging a broader literature in IR deploying practice-based approaches to knowledge production and a more recent, growing feminist literature on gender expertise, which has largely developed in isolation.

The article is structured as follows. Next, we contextualize the study by reviewing previous research on knowledge production in IR, particularly emphasizing the so-called practice turn in research on knowledge production. Here, we also situate our study in feminist research on knowledge production and gender expertise. We thereafter introduce the case of the GILs and situate them in the context of policymaking and knowledge production on gender in the World Bank. We then construct an analytical framework relying on the mobilization of assemblage thinking, which gives us the theoretical tools to study the practice of knowledge production, and subsequently present the research methodology and material of the study. In the analysis, we then present our findings on how knowledge is produced in the GILs. In conclusion, we reflect on the contribution of our analysis to practice-oriented research on knowledge production generally, and to literature on gender expertise specifically.

Practices of Knowledge Production: Making the World Governable

The recent interest in IR in how objects and phenomena are enacted through practices of knowledge production has not emerged in a vacuum but can be traced to Michel Foucault's critical investigations of the performativity of knowledge. In his work on the history of sexuality, Foucault (1990) shows how sexuality became an object of thought, resulting in the production of discourses and truth claims that enabled an increasing management and optimization of populations. Drawing on Foucault's insights, a range of seminal contributions on the performativity of knowledge production in IR have been made, for example, Mitchell's (2002) work on how the capitalist economy in Egypt emerged as an object of thought through the production of maps and statistics and Escobar's (2012) work on how underdevelopment was made into an object of intervention through scholarly knowledge production in development economics.

As a range of initiatives destined to produce and disseminate knowledge on a myriad of global problems have emerged, Foucault's discussion on the performativity of knowledge production seems even more pertinent. In contemporary global governance, global problems such as state failure, poverty, and violation of human rights are continuously subject to various forms of knowledge production, for example, through indicators and reporting mechanisms arranged by various expert bodies, which transform them into so-called governance objects. IR scholars argue that practices of knowledge production are always bound up with a particular “object,” which is to be represented, designated, and problematized (Corry 2013; Allan 2018). When the object goes through the process of representation, designation, and problematization, it also becomes governable and a range of actors can orient themselves toward the object and make it a target of steering efforts (Corry 2013).

During recent years, research on how practices of knowledge production contribute to the emergence and problematization of particular governance objects has increased. Consequently, research has examined how objects ranging from migration (Robinson 2018) to education (Papanastasiou 2020), piracy (Bueger 2015), climate change (Allan 2017), poverty (Freistein 2016), and human rights (Broome and Quirk 2015; Merry 2016) are rendered known and governable through practices of knowledge production. In this research, it is shown how knowledge production always entails processes of simplification, abstraction, and translation, which, in turn, create the objects that they purport to represent. The role that indicators and benchmarks play in the production of knowledge about areas such as states’ adherence to human rights obligations, level of economic competitiveness, and the extent of state fragility has given rise to its own research strand. This research illustrates how international organizations translate and simplify complex phenomena into quantified and comparable categories. The opacity of indicators and benchmarks is often emphasized, as they tend to conceal the normative and political assumptions and agendas informing their construction (Broome and Quirk 2015; Freistein 2016; Merry 2016). In addition, indicators are performative in the sense that they do not describe the world but perform and enact it. As Merry (2016, 21) puts it: “Those who create indicators aspire to measure the world but, in practice, create the world they are measuring.” For example, the work of Farias (2019) shows that indicators shape how the world is understood, entrenching particular categorizations and binaries and shaping subjectivities and behaviors. When the World Bank decided to phase out the distinction between “developed” and “developing” countries in its annual World Development Report (WDR), Farias (2019) argues that even when the Bank no longer used them, these categorizations had been entrenched in such a way that particular countries had embraced them and made them part of their foreign policy identity.

In research on indicators, there is often a focus on how they contribute to simplifying reality and reducing complexity through quantification and numbers. However, the type of simplification that is at play when constructing indicators and benchmarks does not only entail quantification. Here, Papanastasiou's (2020) work on how best practices are generated in the European Commission's work on education policy illustrates how this form of knowledge production relies on the same form of simplification but does not mobilize numbers as a primary translation technology. Drawing on how best practices are constructed from a range of country-specific policies on education, Papanastasiou (2020) shows that universal principles in the form of “commonalities,” “key elements,” and “mechanisms” are produced by international organizations.

The specific technologies and epistemic infrastructures needed to produce and stabilize knowledge have also been studied. Bueger (2015), drawing on the example of how piracy is rendered known, shows how knowledge is produced in multiple “laboratories,” such as in monitoring groups and by special advisors. Here, “laboratories” as spaces for the production, packaging, and transportation of knowledge are important entry points for the study of knowledge production. Knowledge is also dependent on continuous translation and stabilization in order to appear coherent. Thus, research has delved deeper into the practical technologies of “objectifying” knowledge. As climate increasingly has been transformed into an object of knowledge production, technologies such as tables, charts, and models are used as a way of rendering climate visible and knowable (Allan 2017). Similarly, in the field of migration, Robinson (2018) notes how the capacity of states to manage migration is made visible in material frameworks and benchmarks.

In parallel, an emerging literature on gender expertise has drawn attention to how feminist knowledge is incorporated in processes and institutions of national and global governance (Bustelo, Forest, and Ferguson 2016; Kunz and Prügl 2019). This literature has often focused on gender experts as a category of actors, and examined how they navigate mainstream institutions seeking to produce feminist change from within (Ferguson 2015; Altan-Olcay 2020). Highlighting the tensions and dilemmas that this entails, this research echoes debates about the co-optation of feminist ideas (Eschle and Maiguashca 2018). While acknowledging the dangers of feminism in “governance mode” (Halley et al. 2018) and the potential of such feminism “doing violence itself” (Zalewski and Runyan 2013, 299), research positioning gender experts as feminist insiders, like similar work on feminist bureaucrats and on gender mainstreaming (True 2003; Eyben 2010; Chappell and Mackay 2021), also emphasize the subversive potential and small victories in the pursuit of incremental change from within. In this context, the adoption of instrumental arguments for gender equality, or simplified indicators as a means of documentation, can be strategic choices in order to legitimize gender expertise (Ferguson 2015; Bastick and Duncanson 2018; Altan-Olcay 2022).

A second strand of research on gender expertise has analyzed discourses of gender expertise and problematized how gender expertise produced by international organizations makes possible particular problem definitions and solutions (Prügl 2017; Prügl and Joshi 2021). Studies in this vein have demonstrated how ideas and goals originating from feminist theorizing and movements merge with mainstream practices and rationalities of governing in fields such as peacekeeping (Reeves 2012), development (Calkin 2015), and humanitarian aid (Olivius 2016), thus resulting in new forms of gender knowledge. These studies highlight how the ways in which gender equality goals are interpreted and applied in global governance are shaped by dominant governing rationalities, and therefore limit the potential for transformative change. For example, representations of gender inequality as an expression of the “backwardness” of beneficiary communities draw from and reify global inequalities and colonial imageries and establish a clear hierarchy between the knowledge of external gender experts and the knowledge of local women (Reeves 2012; Olivius 2016). Neoliberal economic logics frequently shape representations of women's empowerment centered on fostering entrepreneurial, productive subjects while obscuring structural constraints and power dynamics (Calkin 2015; Prügl and Joshi 2021), and the reduction of gender equality work to the application of measurable indicators and procedural checklist exercises has clear depoliticizing effects (Olivius and Rönnblom 2019). Yet, as many of these studies emphasize, multiple governing rationalities frequently coexist, creating contradictions, ambivalences, and tensions that open up space for contestation.

However, in contrast to the practice-oriented IR literature, the tangible practices through which knowledge about gender is produced have not been the focus of analysis in literature on gender expertise. Rather, research on gender expertise has been confined to the first two “generations” of research on expertise in international politics. According to Bueger (2014a), these first two generations either focus on the role of experts as actors having a causal influence on international politics or on expertise as a form of discourse. The third generation, however, focuses on the practices through which expertise is produced. Research on gender expertise could more clearly, we argue, draw on the advances made in this generation of research, and further explore the practical work needed to produce and maintain gender expertise as a body of knowledge. In this article, we contribute to addressing this gap through our analysis of the production of gender expertise in the World Bank GILs.

Gender and Knowledge Production in the World Bank

The World Bank is not only a major global development donor, but also a major producer of knowledge. The bank has evolved into a hub with high capacity to generate and disseminate knowledge about development problems and about what works to solve them. This has led St. Clair (2006, 77) to argue that the World Bank “is a transnational expertised state-like institution that sets the scene for both global politics and global knowledge.” For years, the bank has, for example, developed different instruments, such as concepts, indices, world maps, and graphs with the aim of measuring and producing knowledge about phenomena such as poverty and state capacity (Nay 2014; Freistein 2016).

As part of a broader shift toward integrating social dimensions of development, the World Bank has increasingly adopted the goal of gender equality and engaged with concerns around gender inequality in its policies from the mid-1990s and onward (Bergeron 2003). As a result, gender has become subject to various forms of knowledge production. The first gender strategy was adopted in 2001 and to step up its implementation, a gender action plan (GAP) was launched in 2007 with the aim of strengthening women's economic empowerment, gradually moving gender issues into the mainstream of World Bank affairs (World Bank 2016; Jones 2018). In 2007, a World Bank report coined the much-cited terminology of a “business case” for gender equality as “smart economics” (World Bank 2007, 145). This instrumental representation of gender equality as a means to economic progress not only helped legitimize the adoption of gender equality goals within the bank, but also gave rise to a still ongoing debate about the co-optation of feminist ideas, and the dilemmas of working as a gender expert within international organizations (Chant and Sweetman 2012; Ferguson 2015; Altan-Olcay 2020).

In 2012, the World Bank devoted its main flagship report, the WDR, to the topic of Gender and Equality. Building on extensive background research and being widely disseminated, the WDR 2012 constituted a key milestone in the integration of gender as a core focus of bank policy and knowledge production. Its publication generated a flurry of feminist debate and critique (Elson 2012; Harcourt 2012; Razavi 2012; Ferguson and Harman 2015). Overall, feminist scholars welcomed the fact that the report presented a more nuanced conception of gender equality as not only instrumentally good for growth but also a development goal in its own right. It also extended the earlier focus on women's unequal participation in wage labor to also discuss issues such as reproductive labor, violence against women, and political participation. In this way, the WDR 2012 widened the space for feminist engagement with the World Bank and pushed the boundaries for what could be said about gender and development (Razavi 2012).

Thus, since the turn of the millennium, the World Bank has stepped up its research capacity related to gender issues and expanded its conception of gender equality beyond a narrow, instrumental “good for growth” model. Over time, bank reports focusing on gender demonstrate both increasing attention to equality as an intrinsic goal and increased diversity in terms of the issues discussed (Prügl 2017). While this shift has created openings for feminist engagement (Razavi 2012; Prügl 2017), feminist critics have continued to point out the persistent limitations of World Bank approaches to gender. Despite the bank's “social turn,” economistic neoliberal discourses continued to shape understandings of gender, privileging instrumental legitimations for gender equality as “good for growth” (Bergeron 2003; Esquivel 2017) as well as efficiency arguments premised on essentialist ideas of heterosexual gender complementarity (Bedford 2007, 2009). In addition, feminists have been critical of the lack of recognition of the relationship between macroeconomic policy and gender inequality—for example, the gendered effects of austerity policies—and thereby the role of the World Bank itself as a key player in the production of structural conditions leading to gender inequality (Elson 2012; Razavi 2012). As Griffin (2009, 2007) has argued, the World Bank has continuously failed to acknowledge the gendered underpinnings of its dominant economic logics and theories, such as the masculinized homo economicus of neoclassical macroeconomic theory. In effect, “the incorporation of feminist ideas in the World Bank's research has taken place without a fundamental rethinking of the neoliberal economic agenda of commodification, privatization and [ … ] austerity” (Berik 2017, 565; see also Prügl 2017). As a result, Esquivel argues, gender equality interventions continue to “rely on micro-level fixes to improve gender-equal access without challenging the neoliberal macroeconomic framework” (Esquivel 2017, 547).

Relatedly, despite the inclusion of a broader range of economic, social, and political issues, World Bank policy and strategies are argued to be characterized by the reduction of complex issues, such as reproductive health and rights, into simplified and measurable indicators, such as maternal mortality, and as a result it fails to account for the complex linkages between issues such as trade, economic policy, reproductive rights, and gendered violence (Harcourt 2012). Thus, rather than fundamentally challenging gendered power relations, the World Bank's embrace of feminist ideas is argued to illustrate a broader contemporary “neoliberalisation of feminism” (Prügl 2015), where some feminist ideas have become key building blocks in the crafting of new forms of neoliberalism in the wake of the erosion of the Washington Consensus (Bergeron 2003; Calkin 2015; Berik 2017; Prügl 2017).

Thus, feminist research on the World Bank has generated rich and nuanced accounts of the discourses and assumptions that shape its gender policies (e.g., Griffin 2007; Berik 2017; Prügl 2017), and their effects when development programs are implemented in various contexts (e.g., Bedford 2007, 2009; Ferguson and Harman 2015). In this article, we build on this body of research, but shift the research focus away from the content of gender policies and their implementation to the processes and practices of knowledge production about gender within the World Bank. We thereby add new insights into the labor that goes into producing, sustaining, and disseminating particular ideas about gender inequality, which then come to shape development strategies as well as lived experiences.

To do so, we focus on the World Bank GILs, as these are relatively new institutions with a mandate to produce knowledge about gender. To produce evidence-based knowledge that can make gender equality work as effective as possible, the Umbrella Facility for Gender Equality (UFGE) was created in 2012 as a node for efforts to produce and disseminate knowledge on how to close gender gaps in four thematic areas: human endowments, employment and entrepreneurship, asset ownership and control, and voice and agency (Umbrella Facility for Gender Equality 2020). The UFGE supports five so-called GILs: the Africa Gender Innovation Lab, the East Asia and Pacific Gender Innovation Lab, the Middle East and North Africa Gender Innovation Lab, the Latin America and the Caribbean Gender Innovation Lab, and the South Asia Gender Innovation Lab. All GILs focus on producing rigorous evidence on “what works” to close gender gaps. To this end, the GILs rely on the production of systematic reviews that synthesize existing knowledge and experimental impact evaluations that produce knowledge on the effectiveness of particular development interventions.

The GILs are examples of “analytic institutions” in international organizations: “specialist units, departments, committees, adjudicatory bodies and others housed by or linked to IOs that develop the cognitive framework for understanding and solving policy problems” (Broome and Seabrooke 2012, 5). Analytic institutions are important for how international organizations “see” and perceive the world upon which they act, as they accumulate “actionable knowledge” through practices of statistics and documentation. In this way, the GILs make it possible for the World Bank and its various partners to diagnose problems and prescribe solutions regarding gender inequality. In other words, their work makes it possible to “see” the world and make it legible as an object of governing. In this article, we focus on how the GILs function as spaces for generating and disseminating policy-relevant knowledge on gender inequality, and discuss the political effects of their knowledge production in terms of how they render gender knowable and governable. This is important in order to critically examine the power effects of knowledge production in international organizations as these organizations “orient action and create social reality” (Barnett and Finnemore 2004, 6).

Analyzing Knowledge Production through Assemblage Thinking

Exploiting the recent turn to practice theorizing both in IR generally and in studies on knowledge production specifically, we mobilize assemblage thinking as our analytical framework for studying knowledge production. With its focus on the practical aspects of assembling different objects and phenomena, it is particularly suitable to study the work of producing knowledge.

During the recent decade, assemblage thinking has exploded and found its way into multiple disciplines in the social sciences, such as anthropology (Li 2005, 2007; Ong and Collier 2005), human geography (Anderson and McFarlane 2011; McCann 2011), sociology (Sassen and Ong 2014; Clarke et al. 2015), and political science (Acuto and Curtis 2014; Bueger 2018). A common denominator for this research is that it engages with the philosophical work of Gilles Deleuze and Félix Guattari. In A Thousand Plateaus (Deleuze and Guattari 2013), they introduce a conceptual apparatus that rejects taken-for-granted axioms of Western philosophy, such as logic, universalism, and linearity. Part of this conceptual apparatus is the concept of assemblage, which can be used as a tool for thinking about the world as in a state of permanent becoming—the world before us is a product of complex “process[es] of arranging, organizing [and] fitting together” (Wise 2005, 91, emphasis in original).

Thus, the world can be thought of as being continuously assembled as heterogeneous components are drawn together to form temporary alliances. Assemblages are characterized by the striving for consistency and homogeneity, while at the same time being prone to dissolve and fall apart. (Deleuze and Guattari 2013). Moreover, assemblages should not be understood as products of predetermined parts being put together, but rather as “ad hoc groupings” of various elements (Bennett 2010, 23) that come from both discursive and material registers.

As assemblage thinking has gained traction in the social sciences, it has been considered an invitation to conduct empirical research with the aim of deconstructing phenomena that appear unified, homogeneous, and ordered. Thus, rather than accepting the ostensible unity of social life, assemblage thinking provides the conceptual vocabulary to direct the focus onto how unity and order is produced through processes of assembling. In this last phrase, assemblage has been transformed into a verb that has been an analytical strategy for different forms of research during the last decade. This research refrains from treating assemblage as a noun, which would make it a “preexisting thing in the world with pre-given properties” (Rabinow 2014, 206), in favor of conceptualizing the term as a verb that captures the laborious work of drawing together heterogeneous elements into a coherent, yet provisional, whole. This focus directs the research gaze onto the labor being invested in drawing together heterogeneous elements into ostensibly coherent assemblages (Newman and Clarke 2009; Anderson and McFarlane 2011). As Li (2007, 264) notes: “Assemblage flags agency, the hard work required to draw heterogeneous elements together [and] forge connections between them.” At the same time, this process of drawing together elements is fragile as “that which has been assembled can more or less easily come apart, or be dismantled” (Newman and Clarke 2009, 9).

This version of assemblage thinking¹ has informed research that uses the concept to analyze how different phenomena are assembled into coherent formations that, at the same time, can be disassembled. As an example, assemblage thinking has been mobilized in order to reveal the laborious work of assembling that goes into the construction of unitary and “black-boxed” policies and reform projects. In an attempt to understand the complex character of policymaking, scholars have found in assemblage thinking a possibility to analyze the work required to make policies hold together, such as the enrollment of actors, the enlistment of expertise from national and international sources, the invention of technologies, and the mobilization of institutional capacities (Prince 2010; McCann 2011). Clarke (2015, 98) argues for the use of the concept of “repertoires” to denote the “raw material” in the form of “discourses and devices, texts and techniques, roles and resources” that can be appropriated and translated into policy work. In Clarke's (2015) analysis of the managerialization of British universities, this concept makes it possible to capture how repertoires of management vocabularies, techniques of governing, and different sets of identities provide the raw material from which a managerial organization can be assembled. Ong and Collier (2005) have added to this discussion, as they argue that these repertoires are composed of “global forms,” meaning globally circulating scripts, models, and technologies that can be translated into the work of making policies and reform projects come true.

Assemblage thinking has also been taken up widely in IR research. Most recently, there has been a turn toward using assemblage thinking as a tool for studying practices of assembling. Through studies of, for example, maritime piracy (Bueger 2018), palm oil production (Köhne 2014), and development aid projects and interventions (Li 2005, 2007), research has shown how efforts to govern these phenomena are dependent on the work of assembling, including the forging of alliances between stakeholders, the establishment of objects and fields for technical intervention, the mobilization of expertise, and the depoliticization of potentially contentious issues.

Assemblage thinking as it has been described above is mobilized as a theoretical resource to go deeper into the work of assembling that is invested in the production of knowledge in the GILs. As such, assemblage thinking is, for us, an “empiricist project” (Bueger 2014b, 58) that values empirical proximity to the mundane practices of assembling knowledge. In our analysis of knowledge production in the GILs, we, therefore, aim to get closer to the practical work of producing, maintaining, and disseminating knowledge. This means that we direct our research gaze toward the work of assembling that has been described above, such as the work of aligning interests of stakeholders and the use of repertoires that provide the necessary resources that can be translated into knowledge production.

In the following analysis, it will become clear that we have opted for analyzing a particular practice of knowledge production, that of impact evaluations. We approach impact evaluations as epistemic practices, which, according to Bueger (2015, 6), “are concerned about knowing a distinct (epistemic) object and aim at building universals out of particulars.” Thus, we approach impact evaluations as epistemic practices that are fundamentally assembled, meaning that they build on the drawing together of particular elements in order to create a universal story about the epistemic object of gender inequality.

In order to examine how impact evaluations are assembled as epistemic practices, this study relies on an empirical investigation of the GILs conducted during the autumn of 2021. Approaching the GILs as “analytic institutions” that make the world legible and governable through knowledge production, we generated data through eight semi-structured zoom interviews with researchers in the GILs. We also conducted a focus group with five selected GIL researchers. In parallel to these interviews, we collected a range of documentation and online material (such as webinars, white papers, PowerPoint presentations, policy briefs, and evaluation reports). In the analysis of the material, we have been steered by a desire to zoom in on the practical aspects of knowledge production such as the design of research projects, collection and analysis of data, and the presentation of research findings.

Making Gender Known in the Gender Innovation Laboratories—A Process of Assembling Knowledge

In this section, we focus on how gender inequality is made knowable in the GILs through assembly work. We do so by approaching the impact evaluation as a central epistemic practice.² Being a resource-intensive form of research, impact evaluations account for the largest share of the budget of the GILs and are described as their main “comparative advantage.”³ An impact evaluation is a form of research aiming to assess the effectiveness of a particular development intervention through the use of experimental methodologies. These methodologies aim to assess if particular changes can be attributed to a specific intervention. In the GILs, impact evaluations are used to assess the effectiveness of development interventions aiming to increase women's empowerment in various respects, such as education, earnings, and labor market participation. Impact evaluations can be understood as epistemic practices in the sense that they draw together data and claims about particular interventions in order to make general statements on “what works.”

We structure the analysis according to a linear understanding of how impact evaluations are produced, meaning that we see impact evaluations as occurring in three phases: the planning and preparation phase, the implementation phase, and the dissemination phase. We recognize that impact evaluations as forms of knowledge production do not fully conform to these linear phases in practice, but we use it as a narrative device to reconstruct the assembly work needed to produce impact evaluations.

Planning and Preparing for Impact Evaluation—Identifying Projects and Recruiting Support

Before the work of executing an impact evaluation can start, the GILs work with identifying interventions that are deemed particularly suitable for an evaluation. This is a matter of identifying “projects.” As a way of organizing this process, the GILs have set up a project application system with which potential stakeholders (nongovernmental organizations, ministries, donors, or private sector firms) can engage in order to present projects containing interventions that are suitable for an evaluation. Applicants should describe the project and the intervention to be evaluated in a so-called expression of interest (EOI). As a help for the applicants in formulating a coherent project, the GILs formulate “calls” for EOIs in which detailed information on what the EOI should contain is provided. Here, application forms with particular questions to be answered are common.

For example, in a call for EOIs that focuses on attracting proposals to evaluate interventions destined to strengthen women's economic empowerment in rural Nigeria, applicants are asked to provide information on the project team submitting the application, a project description detailing the interventions to be evaluated, the geographical location of the project, the time period of the project, and to which extent men and women are targeted by the project. In addition, the EOI should detail the impact objectives of the interventions, propose an appropriate methodology that takes into account the specific considerations characterizing an impact evaluation, list the possible risk factors with the project, and state the willingness to form functioning partnerships with the GILs (Africa GIL 2020). The document containing the call thus provides detailed instructions on how to formulate the EOI and can be seen as an effective “technique for interesting” (Freeman and Maybin 2011, 159). Understanding documents as techniques for interesting implies that a document “connects actors and coordinates their actions” (Freeman and Maybin 2011, 159). In this case, call documents align the applicants to the project application system and bring order to their work in formulating the EOI in a coherent manner.

Once projects have been assessed as suitable for an evaluation, the GILs must form a functioning cooperation with the project stakeholders in order to construct appropriate research questions. Here, the GILs rely on specific activities that work as techniques for coproducing research questions. Workshops, including both GIL researchers and project stakeholders, work as spaces for constructing questions that can be widely accepted by everyone:

And then this two-day-workshop, we do a few hours on “What's impact evaluation? What's the methodology? What are the different methods? Why would you do one? What can you learn with it? How does it work? What are the different steps?” And then we start taking apart the project, figure out what are all the components of the project, what's the project's theory of change, and what are the questions? What are the questions that the team has? Trying to get them to the point where the question mark is. So, you know: “You're doing these activities and you think they are having this impact, but do you really know that for sure?” And we … sort of introduce them a little bit to the literature and the evidence gaps. And what we know and what we don't know; and what would be interesting learnings that could come from the project. And then together, we sort of come up with the research questions: “What will we evaluate and how?”⁴

The work done during the workshops can be understood as a process of forging alignments and recruiting support. Forging alignments (Li 2007) entails the work of linking together the objectives of the various parties involved. Second, this very act also entails the recruitment of support. Mosse (2005) argues that the level of success of a particular activity is dependent on the recruitment of support from possible opponents. Thus, the workshop becomes a technique for exposing possible divergences in understandings of the project to be evaluated (e.g., its logical construction) in order to impose a workable evaluation design. The workshops work as a way for GIL researchers to act as “interpretive communities” (Mosse 2005) that impose their interpretations and representations of the project to be evaluated in order to secure participation from the stakeholders. This does not only work as a way of enlisting stakeholders into the practice of impact evaluations but also functions as a way of aligning them to a specific neoliberal discourse of evaluation and measurement that now pervades feminist gender equality work.

In some evaluations, GIL researchers come in earlier in the implementation process in order to influence the design of the intervention so as to make it possible to evaluate at a later stage. Although this is not the formal role of the GIL researchers, they sometimes provide suggestions on how to translate the abstract goals of the project into activities that can be evaluated:

So, these details so far are not yet very well defined which makes it difficult for us to design the evaluations. Because you cannot really design an evaluation if you do not know […] the details of what will happen. You need to know more about the intervention. And then that sort of carries through for almost all the interventions. […] I mean, it's not really our role to design the intervention. But we were sort of telling them like: “Look, I mean, it's important for you to, you know, to maybe think about these things.”⁵

Thus, this process can be said to be characterized by the work of “maximizing evaluability,” meaning that GIL researchers “nudge” the project team to adopt an intervention design that can be subject to evaluation. This means that these processes include the work of defining and refining the meaning of gender inequality so it can be made amenable to an impact evaluation, and this work shapes how gender is produced as a governance object that can be known and acted upon. First of all, statistical comparison requires that the categories of men and women are treated as binary and fixed. As Griffin (2007, 228) notes, although many bank publications describe gender as socially constructed and variable, “men” and “women” at the same time remain as biological and constant categories of measurement. Gender inequality is thereby treated as measurable “gender gaps”—a term frequently used by GIL researchers⁶—between women and men. Gender gaps, and the potential success of gender inequality interventions in narrowing them, can be captured through data on women's income, productivity, education, or health relative to men's. This stabilizes and reproduces the idea that gender inequality is mainly due to a lack within women, who need to be “empowered” to “catch up” with men (Prügl 2015). This echoes the critique of Esquivel (2017, 547), that World Bank gender equality interventions tend to rely on technical “micro-level fixes” within an unchallenged economic and political order, and also brings to mind the long-standing critique of Women in Development approaches that have been argued to simply “add women and stir” (Chant and Sweetman 2012; Calkin 2015).

Thus, already in the process of identifying projects to evaluate, the logic of impact evaluations as a specific epistemic practice has effects for how gender, and gender inequality, is rendered knowable and governable. As the GILs “nudge” stakeholders into designing projects that would fit an impact evaluation, gender inequality is given a sufficiently static meaning to be measurable and evaluable. This means that gender inequality must be simplified and complexities erased to fit the experimental logic of impact evaluations. This shapes not only the knowledge about gender inequality that is produced, but also the development interventions to which beneficiaries in the Global South are exposed.

Carrying out Impact Evaluations in Practice—Activating Repertoires of Expertise

If the preceding activities are about securing the basic conditions for an impact evaluation to take place, the next step is to create a methodology that allows for its implementation. The methodological part of the impact evaluation entails a construction of a solid research framework and data collection strategies. Here, it becomes clear that although carrying out an impact evaluation is messy, there are measures that can be taken to bring order to this research practice. In order to fulfill the basic aim of an impact evaluation—which is to determine if particular interventions destined to improve women's empowerment lead to the intended effects—GIL researchers draw on multiple resources. The way GIL researchers construct the methodological part of an evaluation does not occur in a vacuum but entails the activation of a set of repertoires that provide the necessary “raw material” for the evaluation methodology, such as specific vocabularies, methodological techniques, and strategies of legitimization (Clarke 2015).

In particular, the GILs take advantage of a methodology that provides the necessary repertoires to construct a rigorous impact evaluation: the “randomized controlled trial.” Although not the only method available, it is, as one informant argues, “the golden standard of impact evaluations.”⁷ The RCT can be understood as a “global form,” having a “distinct capacity for decontextualization and recontextualization, abstractability and movement” (Ong and Collier 2005, 11), meaning that it is generic enough to be broadly applicable but at the same time adaptable to specific contexts. Thus, the RCT has developed into a generic methodology that can be used for impact evaluations in a variety of contexts. The RCT is made up of repertoires that can be activated for the sake of bringing order to the design and execution of the impact evaluation. Here, the RCT offers both specific resources in the form of methodological techniques and standardized concepts and general strategies of legitimization and justification.

RCTs build on the idea that the effects of interventions can be measured by subjecting a group to a “treatment” (e.g., letting the group benefit from particular activities), while another group does not receive the treatment. Often, the participants in an intervention are randomly assigned to either a “beneficiary group” receiving a particular treatment or a “comparison group” not receiving treatment. To trace change over time, “baseline data” are collected in order to establish the status of the beneficiary group prior to the intervention. The so-called end-line data are collected toward the end of the intervention. This serves as a basis for using various statistical calculation methods for determining the effects of the intervention on the target groups. The GILs use RCT methodology to evaluate a plethora of interventions that aim to strengthen women's empowerment in various ways, such as how road maintenance programs can lead to the strengthening of women's economic empowerment (East Asia and Pacific GIL n.d.). Through this methodology, GIL researchers argue that

[W]e can distinguish causation from causality. So, basically probably we can say with greater evidence that problem X affected the outcome B, rather than there is a correlation between the two […] for example, we have a project on cookstoves which exposed the impacts of cookstoves on health and women's time use.⁸

Thus, in this impact evaluation project, women's health outcomes and time use are compared between a group of women who were exposed to the “treatment” of receiving cookstoves and a group of women who did not receive this treatment. Exemplifying the logic of RCTs, if the group of women who used the cookstove is found to have better health outcomes and use less time for tasks related to cooking, the cookstove project will be considered a gender equality intervention that works. Here, attention to context and diversity among women will necessarily be limited, as the treatment group and control group are considered comparable with the exception of being exposed to the intervention or not. While the value of these types of “micro-level fixes” (Esquivel 2017) for women's everyday lives should not be discounted, this example also illustrates how the knowledge that is produced is blind to structural economic and political conditions that cause women to be overburdened by unpaid reproductive and subsistence work (Razavi 2012; Berik 2017).

The way an RCT is conducted by the GILs reveals how it provides useful repertoires that can be activated for the sake of doing impact evaluations. First, it provides a range of standardized concepts. Concepts such as “baselines,” “beneficiaries,” “comparison groups,” and “end-line data” work as “boundary objects,” “which are solid enough to circulate back and forth between social worlds undamaged as ‘immutable mobiles,’ but also flexible enough to be adaptable to various local contexts” (Rottenburg 2009, 104). Thus, these concepts contribute to making an impact evaluation an ordered process. When the GILs activate these concepts in impact evaluations, a messy reality can be coded according to universal categories that are adaptable to a specific context. Second, RCTs provide a set of methodological techniques. With RCTs comes a certain way of collecting and analyzing data that makes it possible to determine effects of an intervention. The practice of randomization makes it possible to divide people into beneficiary and comparison groups. Data collection techniques are often different forms of surveys, which are administered and facilitated by survey firms. In the analysis, specified statistical analyses are used, such as regression analyses. Third, RCTs provide legitimacy. RCTs have a long history of use in clinical trials and have also spread widely in the social sciences. Since RCTs are based on the ideals of positivist science in which reality can be objectively measured, this way of doing science becomes important to fulfill the aim of impact evaluations, which is to generate evidence on “what works.” As a form of rigorous science, RCTs can lend scientific legitimacy to the conclusions and recommendations emanating from impact evaluations. As Altan-Olcay (2022) notes, the production of quantitative data and the technicalization of women's empowerment in decontextualized indicators and measures are central to the authorization and recognition of gender knowledge in international organizations. While these practices may be employed strategically by gender experts seeking to expand the space for feminist politics, they also generate their own effects, which limit how gender inequality can be known and addressed, and simultaneously de-legitimize and obscure other forms of knowledge. This simultaneous production of knowledge and ignorance (Bakonyi 2018) is pointedly illustrated by a GIL researcher, describing the importance of a quantitative study of perceptions of safety in a number of Global South locations:

You have the evidence now which you did not have before. No one knew how women and men of Dakar city or Bangladesh feel about safety. There was no measurement. Now you have it measured. Now you know how much is the percentage.⁹

This example illustrates that when gender inequality is produced as an objective and quantified object of knowledge, it simultaneously entails that previous knowledge is discounted. Clearly, women and men in the locations mentioned had knowledge about safety and insecurity in their everyday lives prior to the GIL study, but these forms of knowledge are silenced and rendered invisible, as they are not systematic and actionable in the right way.

The central role that RCTs play in the GILs reflects a wider turn toward experimental methods in the production of global knowledge. The proliferation of experimental methods has resulted in the exclusion of various voices and views from knowledge production, most notably those that do not live up to scientific standards (e.g., Littoz-Monnet 2022). We observe similar effects in the way the GILs understand RCTs as embodying a superior form of scientific knowledge production. RCTs and the repertoires of expertise they provide are examples of the hegemony of “the monoculture of knowledge,” which, according to Santos (2003), advances the ideals of modern science at the expense of other knowledge regimes, effectively rendering them nonexistent. Consequently, RCTs and the repertoires of expertise they provide are not neutral methodologies but reproduce and perpetuate a “monoculture of knowledge” that silences forms of knowledge that are rooted in subjective and “nonmeasurable” experiences. However, for gender inequality to be made known and governable, it must be inscribed into a western monoculture of knowledge, thereby making other forms of knowledge about gender inequality that are rooted in subjective experiences invisible and nonexistent.

Disseminating Results—Translating Knowledge from Impact Evaluations to a Broader Audience

A goal for the GILs is for their findings to gain traction in policymaking circles. Consequently, the GILs put significant resources into disseminating results from impact evaluations to a broader audience. Different forms of dissemination require a significant amount of assembly work as the results from complex impact evaluations need to be drawn together, translated, and represented in an accessible way for policymakers. The GILs have multiple dissemination formats at their disposal that help them tailor findings for different audiences, such as policy research working papers, scholarly publications, policy briefs, and PowerPoint presentations. Describing their practices of disseminating results, one GIL researcher explains that:

We produce evidence for policy briefs that are two- or four-pagers, that are very practitioner-friendly with the main results. And we do seminars. But most of the dissemination happens in direct interaction with the practitioners. So, if I have ties with some staff in a ministry, for example. And they are starting a new project, or a new policy and they ask us for inputs. That's when we say… we might present to them as a PowerPoint, or have a discussion with them, or write a customized brief for them with the evidence. And by evidence I mean: “On this topic, in this country, we found – you know, one paragraph – this worked! This particular intervention worked. It increased women's productivity by 20% when we did extension services in this way.”¹⁰

As illustrated here, a key factor for successfully promoting “research uptake,” as GIL researchers term their work with dissemination, is to present concise, clear, and causal narratives of impact: “if you say something, you know: ‘I have scientifically measured this and it increased women's income levels by 30 percent.’ That speaks to people.”¹¹

Although different forms of dissemination vary in length and technical detail, they all entail an “objectification” of knowledge, meaning the translation of knowledge into material form that gives knowledge the appearance of being stable and coherent (Robinson 2018). Objectification requires translation that has the effect of both reducing and amplifying the findings from impact evaluations (Rottenburg 2009). Although different written publication formats differ in length and technical detail, they make use of general techniques for visualization and representation. In order to present findings from methodologically complex evaluations, many of the mentioned dissemination formats make use of “inscriptions,” which are devices that “translate the messiness of the world [ … ] into tangible knowledge that is concrete and visible [ … ]” (Adler-Nissen and Drieschova 2019, 534), such as graphs, diagrams, tables, scientific instruments, surveys, and charts. Through these inscriptions, “reality is made stable, mobile, comparable [and] combinable. It is rendered in a form in which it can be debated and diagnosed” (Miller and Rose 2008, 66–67).

Inscriptions, such as diagrams, are useful for visualizing results from RCTs as they can describe the various effects interventions have on beneficiaries and comparison groups. For a more technically interested audience, tables give information on results from statistical analyses (see Appendix in East Asia and Pacific GIL n.d.). The way these findings are visualized and represented is a process of both reduction and amplification. Through inscriptions, findings from impact evaluations are condensed to fit the format of visual representations. However, while this is certainly a matter of reducing the complexity of findings, this is also an amplification. This means that the representation transcends the inscription device itself and makes possible conveying findings in new ways, thus creating a powerful message. For example, it is possible to visualize “success stories” of particular interventions that can give inspiration on how to design interventions in new contexts. Thus, findings from impact evaluations, represented and visualized in a particular way, can move beyond their zone of communication as decontextualized bodies of knowledge (cf. Martin de Almagro 2021).

A central ambition in the communication efforts of GILs is to provide examples of “what works” when working with preventing gender inequality. By conveying results about successful interventions in brief and “punchy” communication formats, the aim is for these results to be taken up and “scaled up” (Africa GIL 2022) in other contexts. Although well intentioned, this risks having the effect of understanding gender inequality as it can be counteracted and prevented through standardized solutions. Again, we can observe how complexities of the roots and effects of gender inequality are downplayed through these forms of dissemination, as they can only accommodate brief messages about successful “best practices,” stripped off context and complexity. As several GIL researchers explained to us, presenting gender inequality problems as tangible, technical, and amenable to intervention is critical to rendering them actionable in the eyes of policymakers:

So, there may be underlying constraints that are a lot harder to change. Like, you know, norms around what a woman's work is, and whether a woman should work outside the home and that sort of stuff. These are hard to change, and they are sticky, and they take a long time to change. But, if you find that women farmers in Nigeria use a lot less fertilizers than men farmers, then you actually have an entry point for policy. Like: “oh let's work on ways to help women access more fertilizers or use more fertilizers!” That helps policy makers think about what they could do, rather than thinking: “well, we need to change everything and nothing is going to improve if we don't change the culture”, which is something we often hear. We say: “well, you don't need to change everything to make improvements. You can start with this, and this, and this. And this can reduce gaps by this much.”¹²

This example points to the importance of rendering policy problems technical (Li 2007) and measurable, simplifying complex realities in order to amplify key messages to target audiences. In the process, more complex, and political, dynamics of gendered norms and power relations are obscured. As the interview extract shows, GIL researchers are well aware that they are suggesting technical fixes in response to complex problems rooted in unequal relations of power. As they argue, these practices do make some aspects of gender inequality visible and amenable to intervention, thus making some forms of change possible. At the same time, however, the dissemination of technical, simplified narratives about “what works,” which then comes to shape policy, renders the broader context of gendered hierarchies and norms invisible and thereby leaves them unchallenged.

Conclusions: Knowledge Production and Depoliticization

In this article, we have shown how the production, maintenance and dissemination of knowledge and expertise in the World Bank GILs is dependent on different forms of assembly work. With a focus on impact evaluations as epistemic practices, our analysis demonstrates the work invested in assembling the different phases of impact evaluations, such as forging alignments with and securing support among stakeholders, activating repertoires of expertise, and translating results into material objects. This process of assembling knowledge also has political effects in the way gender inequality is made known as a governance object. Through the impact evaluations conducted by the GILs, gender inequality is represented and understood as a measurable, objectively existing phenomenon that can be observed as “gender gaps” in issues ranging from income to entrepreneurial success, or health outcomes. In this way, gender inequality is produced as a governance object that is amenable to technical and expert-based policy interventions. The way in which impact evaluations are assembled and create gender inequality as a governance object reveals how other forms of knowledge are silenced and rendered nonexistent. Impact evaluations, with their roots in positivist knowledge production, make knowledge about gender inequality that is rooted in subjective, and thereby not easily measured, experiences invisible. As such, impact evaluations privilege a form of knowledge which is rooted in a dominant, Western “monoculture of knowledge” (Santos 2003).

This study of knowledge production in the World Bank GILs adds new insights of relevance to several research fields. First, this article contributes to a practice-oriented field of research on knowledge production through an exploration of how the practical work of generating, maintaining, and disseminating knowledge about gender and gender inequality unfolds in the World Bank GILs. By focusing on gender, it adds a new empirical focus to previous analyses of how governance objects such as migration (Robinson 2018), climate change (Allan 2017), and piracy (Bueger 2015) are rendered knowable and governable. These findings have been facilitated through a fruitful engagement with assemblage thinking. Although the practice-oriented literature on knowledge production is steadily growing, we note that there has been a more thorough engagement with governmentality (Bakonyi 2018) and approaches related to actor-network theory and material semiotics (Bueger 2015; Allan 2017; Robinson 2018). Thus, we argue that our analytical mobilization of assemblage thinking introduces an approach to analyzing knowledge production, which facilitates attention to the laborious efforts that go into drawing together multiple components to make knowledge production an ordered and stable process.

Second, this article contributes to feminist scholarship on gender expertise in international politics and governance and on gender and the World Bank. Through our focus on impact evaluations as one specific epistemic practice used to produce knowledge about gender inequality and about what works to address it, we highlight how gender expertise is produced, sustained, and legitimized. This deepens our understanding of how depoliticized, instrumental representations of gender equality, frequently criticized by feminist scholarship, come about through specific knowledge production practices. Our analysis highlights how mundane and seemingly neutral practices of knowledge production—planning, data collection and analysis, and dissemination—have profound effects on how gender inequality is understood and addressed. For example, the logic of RCTs produces a binary and static conception of gender, where context, complexity, and intersectional diversity cannot be accommodated either in explanations for or in the design of solutions to gender inequality. These findings underscore the challenges of gender expertise as a means for advancing feminist change from within international organizations. A practice–theoretical perspective, we argue, thus facilitates new analytical entry points in the study of gender expertise, generating new insights into the labor that goes into producing, sustaining, and disseminating it.

As a final concluding remark, we want to reflect on how the kind of knowledge production analyzed in this article reflects a wider rationality that has implications for how political solutions to gender inequality in the Global South can be imagined. Drawing on Wendy Brown's (2015) notion of the “neoliberal rationality,” we argue that this type of knowledge production that seeks to offer technical and standardized solutions to different problems are part of a central depoliticizing trait of neoliberalism: the quest for generic “best practices” that provide scientific and “right” solutions beyond the messy realm of politics. The practices of knowledge production conducted in the GILs—with their focus on producing best practices “that work”—obscure other, possibly more political, solutions to gender inequality that go beyond the application of technical and standardized knowledge. This type of knowledge production, thus, risks closing important political debates on how development in the Global South can be imagined beyond science and managerial solutions. Thus, political solutions that are grounded in the complex lived realities of women in the Global South—and that are not easily incorporated in the technical best practice production of the GILs—are silenced and ignored.

Footnotes

1

Here, we dedicate the space to describe the version of assemblage thinking that is important for our own analysis. This presupposes that there are other versions which we do not cover in depth here, such as the more coherent assemblage theory put forward by DeLanda (2006) and the more “thin” version of assemblage thinking as an “analytical tactic” associated with Sassen (e.g., Sassen and Ong 2014).

2

This focus is justified by the status of impact evaluations in the GILs, although they also engage in a variety of other types of knowledge production, and by our analytical agenda to get close to the practices invested in generating knowledge.

3

Focus group interview with GIL researchers, November 23, 2021.

4

Interview with GIL researcher, November 8, 2021 (our italics).

5

Interview with GIL researcher, December 7, 2021.

6

For example, the Africa GIL (2022) describes its mission as “To generate evidence on how to close gender gaps in earnings, productivity, assets and agency.”

7

Interview with GIL researcher, November 19, 2021.

8

Interview with GIL researcher, November 19, 2021.

9

Interview with GIL researcher, November 3, 2021.

10

Interview with GIL researcher, November 8, 2021.

11

Interview with GIL researcher, November 8, 2021.

12

Interview with GIL researcher, November 8, 2021.

Funder Information

This work was supported by the Swedish Research Council under grant no. 2015-01756.

Conflict of Interest

The authors declare that there is no conflict of interest.

Author Biography

David Scott holds a PhD in Political Science and works as a University Lecturer at Karlstad University, Sweden. His research explores the marketization and managerialization of politics, a topic he has examined in research on the “projectification” of development aid.

Elisabeth Olivius is an Associate Professor in Political Science at Umeå University, Sweden. Her research explores the politics of gender expertise in international governance, and the gender dynamics of postwar peacebuilding and development with a focus on Myanmar.

Notes

Authors’ note: We are grateful to our colleagues at the political science departments at Umeå and Karlstad University and to participants at the 2022 European Conference on Politics and Gender for comments on previous drafts. We would also like to thank the reviewers for encouraging and constructive feedback. We thank the Gender Innovation Lab researchers who participated in the study for sharing their time and their experiences with us.