Why National Ministries Consider the Policy Advice of International Bureaucracies: Survey Evidence from 106 Countries

Scholars of international relations and public administration widely assume that international bureaucracies, in their role as policy advisors, directly inﬂuence countries’ domestic policies. Yet, this is not true across the board. Why do some countries closely consider the advice of international bureaucracies while others do not? This article argues that international bureau- cracies’ standing as sources of expertise is crucial. We tested this argument using data from a unique survey that measured prevalent practices of advice utilization in thematically specialized policy units of national ministries in a representative sample of more than a hundred countries. Our ﬁndings show that ministries’ perceptions of international bureaucracies’ expertise, that is, specialized and reliable knowledge, are the key factor. International bureaucracies inﬂuence national ministries di- rectly and without the support of other actors that may also have an interest in the international bureaucracies’ policy advice. Our analysis also demonstrates that the effects of alternative means of inﬂuence, such as third-party pressure and coercion, are themselves partly dependent on international bureaucracies’ reputation as experts. The ﬁndings presented in this article reinforce the emphasis on expertise as a source of international bureaucracies’ inﬂuence, and provide a crucial test of its importance.


Introduction
Do international bureaucracies influence national policymakers? As administrative bodies of international organizations (IOs), international bureaucracies 1 advise countries on diverse issues, ranging from designing tax systems to increasing agricultural productivity and adapting to climate change. Is their advice considered by national ministries, and if so, why? We explore the question of whether international bureaucracies' expertise is a direct and important source of their influence, or whether their policy advice is considered for other reasons.
Although there is evidence that not all international bureaucracies are equally influential (see Eckhard and Ege 2016;Margulis 2017), scholars have seldom compared the influence of different international bureaucracies (one exception is . This deficit in the research has been attributed to the lack of a "concept of analysis that would allow for a proper comparison" of influence across different international bureaucracies (Knill et al. 2019, 84). Most of the existing empirical studies are based on case studies. We know far more about a few "influencers" like the International Monetary Fund (IMF) in financial policy (Woods 2006;Chwieroth 2013;Edwards and Senger 2015) or the Organisation for Economic Cooperation and Development (OECD) in education and welfare policy (Armingeon and Beyeler 2004;Dostal 2004;Bieber and Martens 2011) than we do about other international bureaucracies. This is unfortunate, because it is unlikely that all international bureaucracies within a given policy field will be able to influence international and national policy-making in equal measure (Margulis 2017, 517). In addition, we lack systematic knowledge on variations in the influence of international bureaucracies' advice across countries (see also Vabulas 2019) and on their relationships with national counterparts (Fleischer and Reiners 2021, 13).
While previous research has often focused on assessing and explaining international bureaucracies' (potential) influence by way of their formal organizational features, we follow a recent call to "trace the scientific reputation" (Christensen and Yesilkagit 2019, 10) of international bureaucracies based on information from stakeholders. We advance a relational perspective on expertise, which focuses on expertise as a status and not as a capacity, and argue that international bureaucracies' standing as sources of expertise is crucial for their influence. According to the relational perspective, ministries must have the perception that international bureaucracies possess expertise, that is, specialized and reliable knowledge, to consider the advice provided.
Investigating the influence of expertise requires a valid measure. Many previous studies focus on expertise as capacity and draw on cognitive resources (Knill et al. 2019) or simply predefine international bureaucracies as expert bodies. Our large-N empirical assessment of international bureaucracies as policy advisors stems from an elite survey of heads of thematically specialized policy units in ministries across the globe. Not only are these officials the typical, if not primary, addressees of international bureaucracies' advice (Benz, Corcaci, and Doser 2016;Heinzel et al. 2020, 4), but they are also important "gate-keepers of policy research and analysis" at the national level (Doberstein 2017, 398), who decide which knowledge-based outputs and spe-cific policy recommendations should guide policy-making in their countries (Page 1992). Furthermore, for our purpose of assessing officials' self-reports of international bureaucracies' influence on their policy-making, as well as their attributions of expertise to international bureaucracies, we deemed it indispensable to focus on the "small group of individuals" with "first-hand knowledge about the influence of expert actors on a specific decision" (Christensen 2020, 10). These individuals know best whether their policy units have considered the advice provided by various international bureaucracies in a given policy field. We also regard these survey respondents as particularly well positioned to comparatively rate international bureaucracies' expertise in the policy fields in which they are themselves specialized.
Our analysis builds on original survey data from national ministries and central banks (henceforth "ministries"). We used a behavioral frequency question to directly ask how much of international bureaucracies' advice was considered, that is, taken into account in ministerial work over the two years prior to the survey. We received answers from the heads of 354 policy units in 106 countries, which corresponds to a response rate of 38 percent. The survey covered nine global and nine regional international bureaucracies in eight thematic areas in two broad policy fields (agriculture and finance). These policy fields were chosen because a variety of international bureaucracies are currently advising national actors on the diverse challenges that countries worldwide are facing in these areas. Therefore, our sample is composed of many more international bureaucracies than the "usual suspects" of IO research.
To test our argument on the importance of expertise against several alternative explanations, we controlled for three alternative means by which international bureaucracies' policy advice could influence domestic audiences: (1) international bureaucracies provide policy advice that fits addressees' preferences, (2) international bureaucracies rely on intermediaries or "third parties" that can put pressure on the addressees of international bureaucracies' advice, and (3) international bureaucracies rely on the wider means available to IOs to coerce ministries to make specific policy decisions. In addition, we examine the interaction between expertise, the explanation we favor, and each of these three alternative means of influence. We regard this analysis as another critical check for the centrality of expertise: It enables us to examine not only the question of whether expertise is associated with the consideration of policy advice at all levels of the other variables, but also whether it reinforces or undermines these alternative means of influence.
Our results show that those policy units that attributed reliable and specialized knowledge to international bureaucracies were also more likely to consider those bureaucracies' policy advice. Furthermore, perceived expertise can be regarded as the most important explanation overall. The only exception was one subsample, the policy field of finance, in which we found a slightly higher association for the fit of advice with addressees' preferences than for perceived expertise. Taken together, our evidence shows that international bureaucracies influence national ministries directly and without additional input or influence from other actors that may have an interest in seeing ministries follow the advice of international bureaucracies. We even find that the effects of alternative means of influence themselves partly depend on a certain level of expertise. Here, the reputation for expertise functions as a catalyst. Finally, our analysis offers no support for the assumption that expertise only matters when the bureaucracy providing advice is regarded as objective. Our findings are robust to various alternative specifications, including variation in exposure to international bureaucracies' advice and differences at the country level-for instance, in gross domestic product, which can be seen as a proxy for the human resources within a ministry.
This article is organized as follows. First, we discuss the role of international bureaucracies as policy advisors. Second, we derive our central hypothesis and briefly discuss alternative explanations. Third, we describe our research design, the survey, and the statistical models. Fourth, we examine the findings and evaluate the hypotheses. Finally, we summarize our findings and discuss the implications of our argument and findings for research on international bureaucracies as governance actors and the role of expertise in global governance.

International Bureaucracies as Policy Advisors
Historically, international bureaucracies were created to carry out the technical and administrative tasks of IOs that are necessary for the preparation of meetings and conferences. Today, their tasks include generating knowledge and providing policy advice, among other activities, with which they "invisibly" contribute to governance (Mathiason 2007) or exercise "'hidden' influence" (Knill and Bauer 2016, 953). In this article, we focus on between-country variations in the efficacy of their policy advice.
Policy advice refers to international bureaucracies' outputs and policy recommendations, which aim at guiding national policy-making on specific issues. We follow the definition by Halligan, who defined policy advice as an activity that aims to support policy-makers' decision-making by analyzing policy problems and proposing solutions (Halligan 1995). Policy advice is provided through publications, technical cooperation, and learning programs targeted at specific countries or policy-makers (Broome and Seabrooke 2015;Littoz-Monnet 2017b;Stone 2020, 52-53). International bureaucracies also produce annual flagship publications, in which they present their research and analyses on current policy issues and develop recommendations for policymakers. Examples include the Food and Agriculture Organization's (FAO) annual report on "The State of Food and Agriculture" (FAO 2019) and the World Bank's World Development Reports (World Bank 2019). Through these means, international bureaucracies advise national addressees on policies and seek to exercise influence on domestic policymaking.
We study the consideration of this policy advice by national ministries as an instance of international bureaucracies' influence, which is defined as "the sum of all effects observable for, and attributable to, an international bureaucracy" (Biermann et al. 2009, 41).
Over the past twenty years, researchers in PA (Trondal et al. 2010;Eckhard and Ege 2016;Knill and Bauer 2016) and IR (Barnett and Finnemore 2004; have argued that international bureaucracies play an important role in world politics. Their work has established the view that international bureaucracies are actors in their own right that enjoy autonomy as well as authority and shape global and domestic policies. Given the substantial evidence-mainly from case-study research and small-N studies-that international bureaucracies influence governance at multiple levels, academic attention has now shifted to observing and explaining this influence in a more systematic and generalizable manner (Ege, Bauer, and Wagner 2020). A few studies have suggested that the influence of the same international bureaucracy could differ de-pending on the client or addressee. Kramarz and Momani (2013), for instance, report that stakeholders hold "contradictory assessments" of the World Bank's knowledge role. As described below, data from the survey that we conducted with national ministry officials confirm this variation in influence. We sought to better understand this variation and add to the literature by providing a large-N assessment of international bureaucracies' influence by focusing on a critical group of their addressees.
What makes international bureaucracies influential? Many regard expertise as a primary source of international bureaucracies' power (Barnett and Finnemore 2004, 29;Dostal 2004;Fang and Stone 2012, 5;Chou and Riddervold 2015;Boswell 2017;Littoz-Monnet 2017b). This research has also highlighted that international bureaucracies are composed of professionals with specific skills and experience that allow them to claim expertise (Barnett and Finnemore 2004;Trondal et al. 2010;Chwieroth 2013;Henriksen and Seabrooke 2016;Knill et al. 2019). International bureaucracies advise actors at the international and national level and seek to inform the decisions of intergovernmental bodies, such as the legislative bodies of IOs, and the foreign and domestic policies of member states. It is widely acknowledged that publishing and distributing policy recommendations is an important means of entering into domestic policy-making (Armingeon and Beyeler 2004;Stone 2004;Béland and Orenstein 2013). Case-study research indicates a link between expertise, on the one hand, and influence, on the other hand (e.g., . This link has not, however, been systematically assessed in quantitative studies. Hence, we sought to complement existing research by carrying out a systematic, theory-guided test of whether international bureaucracies' standing as sources of expertise is associated with national ministries' consideration of their policy advice.

Explaining Why National Ministries Consider International Bureaucracies' Policy Advice
Why do national ministries consider international bureaucracies' policy advice? We first introduce our preferred explanation-expertise-and then outline alternative explanations. Our explanation attributes a direct and relatively high influence to international bureaucracies, whereas the alternative explanations refer to more indirect channels of their influence.

The Argument and Our Hypotheses: Expertise Is the Most Important Source of International Bureaucracies' Influence
We argue that international bureaucracies' policy advice is mainly considered when the addressees of this advice perceive the international bureaucracy to possess expertise. We follow the widespread definition of expertise as specialized and reliable knowledge in a given domain that is derived from training and experience (Barnett and Finnemore 2004, 46;Gornitzka and Holst 2015, 2;Eckhard and Ege 2016, 971;Littoz-Monnet 2017a, 50). Since scholars first began studying the influence of international bureaucracies, expertise has been considered as an important factor ). International bureaucracies often cite expertise in their descriptions, where they frequently describe knowledge transfer as crucial for their work. The FAO, for example, describes itself as a "knowledge network" that uses the "expertise of our staff […] to collect, analyse and disseminate data that aid development" and aims at "turning knowledge into action," "sharing policy expertise," and "bringing knowledge to the field" (FAO 2020). And the OECD claims to "provide knowledge that Member countries and a range of partners can use as they make policy decisions" (OECD 2020). In line with this understanding of their role, many international bureaucracies have set up their own research departments. Staff members have specialized academic qualifications, many hold doctorates, and some contribute to academic debates. As members of scientific networks, they also mobilize the knowledge of scholars and other experts, for example, as consultants or speakers.
Common to both international bureaucracies' selfdescriptions and the labels assigned to them by scholars is the idea that expertise and policy-making are distinct concepts. Given that we are interested in the impact of expertise on policy-making, we have treated them as distinct in our research design. We nevertheless acknowledge the mutual reinforcement or "co-production" of expertise and policy, in particular, over longer periods of time (Jasanoff 2004). The scientification of politics and the politicization of expertise occur in parallel (Christensen 2020, 1). International bureaucracies do not formulate their policy advice independent of their respective environments, but rather within a socially constructed field that privileges or even demands specific epistemic structures (Littoz-Monnet 2017b, 5). By the same token, the production of expertise can be seen as a way of doing policy (Boswell 2009). For our comparative endeavor and time-bound snapshot, these arguments are, however, less relevant. Although international bureaucracies advise member states on many issues, they do not make domestic policies.
The argument that international bureaucracies' influence depends on expertise is informed by Weber's concept of bureaucratic power (Weber 1991). International bureaucracies are said to possess specialized and reliable knowledge that other actors lack but demand (Barnett and Finnemore 2004, 29; Littoz-Monnet 2017b, 2). Societies expect that policy-makers base their decisions on available expertise and thus rationalize them. Research on the role of expertise in domestic policy-making processes substantiates this function of expertise (e.g., Daviter 2015). Confronted with uncertainty over complex policy problems and the need to solve these issues, domestic policy-makers turn to international bureaucracies (Littoz-Monnet 2017b; Christensen 2020, 7). This reasoning implies that the unique status of international bureaucracies as expert bodies prompts other actors to consider what the international bureaucracy says. We argue that this status depends on whether international bureaucrats are actually perceived as having expertise, because the "perception that the bureaucrats possess reliable, relevant and detailed knowledge creates confidence that their proposals are well founded" (Littoz-Monnet 2017b, 9). By focusing on the attribution of expertise by recipients of international bureaucracies' advice, we follow a relational conception of expertise (see Grundmann 2017, 27). This posits that expertise is "a status that is attributed to someone," while a realist conception would argue that expertise is "a capacity a person possesses" (Collins and Evans 2019, 85). We find both concepts in the research on international bureaucracies. Scholars applying the realist concept argue that not only individuals but also the entire staff of an organization may possess expertise. Expertise is thus understood as a property, skill, and resource that international bureaucracies may own (Biermann et al. 2009, 55;Knill et al. 2019) and that they may use to exert influence (Knill and Bauer 2016). However, the fact that one possesses specialized knowledge does not necessarily mean that others know this and that they trust the source. Many scholars, therefore, prefer the relational conception and argue that expertise needs to be socially recognized to matter (Barnett and Finnemore 2004, 24;Littoz-Monnet 2017a, 587).
In our view, the attribution of expertise to international bureaucracies by the addressees of their advice (here, ministry officials) is the focus best suited to the purpose of our analysis. For the argument, it matters less whether international bureaucracies indeed possess and share specialized and reliable knowledge than whether ministry officials believe that the advice communicated to them is based on expertise. This relational perspective on expertise has been widely applied in various academic disciplines. For example, in the philosophy of science, Goldman uses the notion of a "reputational expert" who "is widely believed to be an expert (…), whether or not he really is" (Goldman 2001, 91 our emphasis). Consequently, we focus on the reputation for expertise to explain influence (see also Jinnah 2010, 60; Christensen and Yesilkagit 2019, 955) and follow Grundmann (2017, 27) in arguing that experts "are primarily judged by their clients, not necessarily by peers." We hypothesize that ministry officials consider an international bureaucracy's policy advice when they ascribe expertise to them. To our knowledge, this perspective has seldom been applied empirically in research on IOs or international bureaucracies.
From the perspective of ministry officials, this argument leads to the following hypothesis:

H1:
The more ministry officials attribute expertise, i.e., specialized and reliable knowledge, to an international bureaucracy, the more of the international bureaucracy's policy advice they will consider.
Our hypothesis is open to variations in what counts as expertise. It thus acknowledges that expertise is socially, culturally, and historically contingent. This is important in our research context, given that the scale of our study is global. Jasanoff and Long Martello (2004), for example, argue that "claims of expertise in global forums are themselves valid only within particular 'situated' frameworks of presumptions and practices" (345), thereby referring to local contexts across the globe (see also Kramarz and Momani 2013). Scholars point to "embedded expertise" and "knowledge orders" that vary across states, including in the ways that expertise is attributed to actors and in how actors arrive at the judgment that someone is an expert (Jung, Korinek, and Straßheim 2014, 411-13;Daviter 2015, 497-98). Consequently, we expect that it is relatively unlikely that all ministry officials agree on how much expertise an international bureaucracy has.
Although expertise is considered an essential means of influence for international bureaucracies, it is often not regarded as sufficient. Many authors instead follow Max Weber's notion of bureaucratic expertise (Weber 1991) and argue that experts have to appear as neutral or "objective" [for individual experts and scientific organizations, see Goldman (2001) and Rietig (2014); for international bureaucracies, see Weiss (1982), Barnett and Finnemore (2004), and Littoz-Monnet (2017b)].
International bureaucracies strive to maintain the appearance of objectivity by generating and disseminating information and policy recommendations that appear to be efficient, impersonal, and embodied in abstract rules and, hence, not as serving specific ideologies (Barnett and Finnemore 2004, 68-69). They spend considerable time presenting their policy advice in objective and evidencebased ways, e.g., by quantifying information. Many scholars hold the view that this pays off for them: "Indeed, it is the objective and depoliticized nature of these policy recommendations that allows them to garner political support […]" (Barnett and Finnemore 2004, 69).
However, objectivity is "more myth than reality" (Weiss 1982; see also Barnett and Finnemore 2004, 21). It is widely known that international bureaucracies' outputs are informed by specific ideologies and economic paradigms, normative orientations, and organizational cultures (Fang and Stone 2012, 550-51;Chwieroth 2013;Kramarz and Momani 2013;Heinzel et al. 2020). One may even argue that international bureaucrats belong to a group of "issue professionals," who derive their expertise "from professional experience linked to an extended commitment to the issue that can be traced from their careers," and not from "independent objectivity" (Henriksen and Seabrooke 2016, 723). What should ultimately count, then, is how objectivity is judged by the addressees of international bureaucracies' policy advice. Can the perception of ideological bias indeed undermine an international bureaucracy's credibility, while being seen as objective can increase it, as scholars have claimed (Barnett and Finnemore 2004, 24, 68-71)? Following the prominent argument in the literature, we expect that the appearance of international bureaucracies' expertise as objective in the eyes of their addressees increases the degree to which addressees consider their policy advice.
This leads to our second hypothesis:

H2:
The more ministry officials attribute expertise, i.e., specialized and reliable knowledge, to an international bureaucracy and the more they regard the international bureaucracy as objective, the more of the international bureaucracy's policy advice they will consider.

Alternative Means of Influence: Congruence, Third-Party Pressure, and Coercion
Although we argue that expertise is the primary source of influence for international bureaucracies, we also consider the fact that they "have various tools of influence at their disposal" (Reinalda and Verbeek 1998, 6). We therefore additionally controlled for three alternative explanations or means of influence that have been widely discussed in the literature, and that may vary between the selected policy fields of agriculture and finance. First, the addressees of international bureaucracies' policy advice could consider this advice because it is congruent with national preferences. Scholars have explained the relevance of congruence or "fit" in various ways. Some have argued that the consideration of policy advice depends on how closely the advice fits with dominant discourses (Kramarz and Momani 2013) or knowledge orders (Jung, Korinek, and Straßheim 2014). From this point of view, discourse-related cognitive frames already filter what advice gets through to policy-makers and "sympathetic elites" (Broome and Seabrooke 2012, 3;Sending 2015). Others contend that international bureaucracies' recommendations are more likely to be considered if they are highly compatible with national preferences or entail only moderate changes (Joachim, Reinalda, and Verbeek 2008;Fang and Stone 2012). Finally, some scholars argue that international bureaucracies' advice is considered when it helps to justify what policy-makers had already planned to do in any case (Boswell 2009;Edwards and Senger 2015, 320). In this view, national bureaucrats instrumentalize the international bureaucracy's advice.
Second, other actors may exert pressure on national policymakers to consider international bureaucracies' policy advice. These third parties may be other member states, lobby groups, nongovernmental organizations, or the public at large (Checkel 1999;Carraro 2019). Such third-party pressure might emerge because an international bureaucracy uses "orchestration" (Abbott et al. 2015) to actively engage with intermediaries to ensure that its policy advice is considered. Such pressure is also conceivable without any active involvement of the international bureaucracy. In the latter case, third parties push for the consideration of the international bureaucracy's policy advice because it supports their own preferences.
Third and finally, one could argue that national ministries consider international bureaucracies' policy advice when they anticipate that they will be obliged to follow the bureaucratic advice in the future. Intergovernmental bodies could pick up the advice and use coercion or weaker forms of enforcement like arbitration, adjudication, or conditionality to ensure that it is followed (Cao 2009(Cao , 1101. While international bureaucracies typically lack authority to enforce decisions or even coerce states to adapt their behavior, we assume that if intergovernmental bodies possess such competencies, this might indirectly affect the likelihood that international bureaucracies' policy advice is considered (Martin 2006).
Although we treat these means of influence as alternative explanations, we are aware that IOs often employ different means simultaneously (Reinalda and Verbeek 1998;Stone 2004). Consequently, it is conceivable that expertise might further enhance or inhibit the other means of influence, or vice versa. In order to substantiate our argument that expertise is most important, we also assessed interactions between expertise and alternative explanations.
International bureaucracies are likely to differ in their ability to employ these different means of having their voices heard. Cognitive and coercive pressures are typically observed among international bureaucracies active in global financial policy (Larmour 2002;Babb and Carruthers 2008;Fortan 2018). Some of our survey respondents in the thematic area of debt management explicitly stated that the BIS and the ECB have "disproportionate legal powers and therefore influence," as one respondent phrased it. Research has also shown that orchestration is prevalent in global financial regulation (Rixen and Viola 2020). The influence of international bureaucracies has been studied less, however, in other policy fields such as agriculture. We therefore cannot draw policy-specific expectations from the literature: We simply do not know how the extent and form of pressure (or coercion) vary between finance and agriculture and whether the demands for congruence with domestic preferences are higher in one of the two fields. We, therefore, tested our hypotheses for both the full sample and the two policies separately.

Research Design
To examine the consideration of international bureaucracies' policy advice, we surveyed a representative sample of heads of policy units of national ministries and central banks (henceforth "ministry officials"). 2 While survey-based approaches do not allow for examining the influence on specific policy decisions, they are regarded as "appropriate for studying expert influence on policy-making in general" (Christensen 2020, 10) and have recently been used in innovative studies on IO influence (Carraro 2019). Furthermore, for the arguments we make, the crucial factor is the actual attribution of expertise to an international bureaucracy by the ministry officials, since they are the ones who eventually consider the policy advice. Accordingly, we measure our explanations based on the assessments of international bureaucracies' addressees instead of using attributions that we would make as external observers.

Survey Design
For our survey, we approached the heads of ministerial policy units that work in one of the eight selected thematic areas in agricultural or financial policy (see table 1). 3 International bureaucracies' advice is usually targeted at and in the overall sample. A response rate for each UN region and each World Bank income group can be found in the online appendix (table A.1), where we also explain why the number of respondents and policy units differ. 3 Our survey focused mainly on national ministries of agriculture and finance.
In finance, we also targeted a few public agencies and (national) central banks.
Our specific choice depended on the thematic area and organizational structure of the national authorities. In the following, for the sake of simplicity, we use the term "ministry" to refer to all targeted national authorities, and "ministry official" for all survey respondents.
decision-makers in member state bodies, such as ministries, who are responsible for the respective policy areas. Ministry officials are regarded as important actors in domestic policymaking, as they provide policy advice to political executives (Page 1992, 61-62). Consequently, they constitute an important group of domestic "interlocutors" (Woods 2006, 72-73) of international bureaucracies and represent an important channel by which international bureaucracies' policy advice may influence domestic policy-making. We were interested in the behavior of policy units in national ministries rather than in individual behavior. Hence, our questionnaire explicitly asked the respondents to selfreport their units' frequency of considering policy advice. The heads of policy units are best positioned to provide information on the practices of their units, so we addressed the survey to them. 4 We consider our respondents as highly qualified to answer our survey: Almost half of the respondents had more than ten years of work experience in the respective thematic area and had been working in the respective ministry for at least five years. Only 3 percent of the respondents had less than one year of thematic work experience. These data reassure us that we have identified the "small group of individuals [that] has first-hand knowledge about the influence of expert actors" (Christensen 2020, 10) and have thereby reduced the risk of our chosen single-informant approach failing to provide valid assessments. 5 To enable a large-N research design and to increase the external validity of our findings, we selected the bureaucracies of nine global and nine regional IOs active in up to four different thematic areas in the policy fields of agriculture and finance (table 1). Many other studies are limited to one particular policy field, but policy fields are not functionally equivalent, and international bureaucracies might find it easier to be active and influential in some than in others (Broome and Seabrooke 2012, 5). The coverage of our sample is remarkable insofar as it includes different types of IOs, such as funds and programs, and specialized agencies, as well as IOs with global and regional membership. In addition, the selected policy fields are vast and include a variety of different thematic areas. Finally, the survey captures the assessments of ministry officials from a representative sample of countries around the globe.
We applied a stratified random sampling process based on all United Nations (UN) member states. 6 This led to a sample of 121 countries that mirror the global distribution of countries according to the World Bank income groups and UN world regions.
The decision to examine these thematic areas and bureaucracies was based on several important considerations. First, to ensure reliability and comparability of answers, the survey questions had to refer to specific thematic areas. Second, to include various types of international bureaucracies and thereby to increase generalizability, we chose broader policy fields, to which the thematic areas belong.
Therefore, in the first step, we specified a variety of policy fields, including agriculture, education, finance, and health, each of which includes a range of different thematic areas that may be relevant for ministry officials in countries worldwide. Based on this approach, we analyzed publications, mission statements, and task descriptions to identify those policy fields and thematic areas in which several international bureaucracies were engaged in providing policy advice to domestic policy-makers. We found the largest number of international bureaucracies providing policy advice in the eight selected thematic areas. We also included regional IOs in the policy field of finance given that they published policy advice on the respective thematic areas. In our analysis, we tested our hypotheses for both the full sample and the policy-specific subsamples. This allowed us to be sure that the influence of expertise is not limited to a specific policy field only.
We carried out a self-administered survey to increase reporting accuracy and to decrease social desirability and acquiescence bias. For similar reasons, we granted anonymity to survey respondents, offered answer options such as "prefer not to say," and avoided questions or answers that require self-reporting of abilities or compliance with expectations.
Within this self-administered format, the survey questions were formulated with extreme care and pretested on a sample of respondents from different world regions, notably bureaucrats from ministries other than those included in our survey, along with public officials at the subnational level and scholars in political science and PA. Additionally, re-5 The use of multiple respondents would not only have been costly but would also have been prone to additional risk of bias. Given the lack of knowledge on the internal working procedures of all ministry units worldwide, we would not have been able to select an appropriate aggregation approach (Enticott, Boyne, and Walker 2009). 6 We excluded Nauru due to missing data on gross national income and San Marino because there is no ministry of agriculture or comparable institution. spondents were given various opportunities to resolve comprehension problems and other questions that arose in the process of participation. The invitations to participate in the survey were personally addressed to the heads of the ministry units responsible for the formulation of policies in one of the eight thematic areas covered by our study. Most were sent out by mail, because unsolicited email surveys could be considered as spam and because email addresses of public officials are seldom publicly available (Enticott 2003, 57). 7 To decrease the response burden, several ways to participate in the survey were offered. Invitations included a link to the online questionnaire as well as a paper questionnaire that could be returned by mail, fax, or email. Most respondents used the online questionnaire. Online and paper questionnaires were available in English, French, German, and Spanish. Invitations to participate were in the same four languages. Also, in Arabicspeaking countries, invitations were sent in Arabic. All this helped us to achieve a response rate of 38 percent, which corresponds to completed questionnaires from 354 policy units in 106 countries (online appendix A).

Operationalization and Measurement
We conceptualized the dependent variable consideration of policy advice as all instances in which ministry officials took an international bureaucracy's policy advice into account in their ministry's work. We measured this variable by the following behavioral frequency question on a scale from 1 ("no advice") to 7 ("all advice") (figure 1).
As figure 2 shows, the extent to which ministry officials considered policy advice varies across international bureaucracies, policy fields, and ministry officials from different countries with different thematic areas under their responsibility. While the IMF and two regional banks, the AfDB and the ECB, proved to be relatively successful in making their voices heard, UNCTAD's and UN DESA's advice was seldomly considered.
Data on the independent, alternative, and some control variables were also collected through the survey (all survey questions in online appendix C).
We measured international bureaucracies' expertise by asking respondents to what extent they regard the international bureaucracies to be experts whose policy advice is based on reliable and specialized knowledge that can be considered without further examination. This operationalization of expertise follows the relational conception and combines a perception of potential competencies (knowledge and skills) with credibility (Goldman 2001;Grundmann 2017, 28). Credibility underpins the coordinative function of expertise: it relieves other actors of the need to engage in time-consuming and costly verification of information (Chou and Riddervold 2015, 66).
As outlined above, we are also interested in whether the combination of expertise and objectivity leads ministries to consider international bureaucracies' policy advice. Objectivity was measured by asking the respondents about the degree to which they judged a bureaucracy's work in a given thematic area as (not) promoting certain ideologies or values.
For the alternative explanations, we gathered data as follows. For congruence, we asked respondents how similar an international bureaucracy's policy advice was to their country's  Boxplot showing the extent to which international bureaucracies' policy advice is taken into consideration, in descending order from high (= 7) to low (= 1). Only international bureaucracies with at least ten observations included; for a list of abbreviations, see table 1. own policy preferences in a given thematic area. For pressure from other actors, we asked respondents to what extent other actors had tried to pressure them to act on the policy advice of an international bureaucracy. Given that other actors could be either state or non-state actors, we explicitly asked respondents to consider actors of all types. For coercion, we asked respondents how they rated an IO's overall ability to use conditionality and/or enforcement to ensure that certain policies were implemented in the respective thematic area. Given that international bureaucracies typically do not have these powers, we explicitly referred to the IO as a whole.

Control Variables and Statistical Models
We controlled for various country-level factors in all models. First, we controlled for a country's demand for external advice, as this factor could affect both the dependent and the independent variables. For instance, a ministry in a country with sufficient financial and human resources to generate specialized knowledge could consider less of the policy advice provided by international bureaucracies than a ministry in a country with fewer resources (and, thus, fewer alternative sources of advice). Likewise, well-staffed ministries could regard international bureaucracies as having less expertise. Therefore, we included a proxy for a country's demand for external advice using the natural logarithm of a country's GDP. 8 This indicator is commonly used as a proxy for the human resources within a ministry and the financial means available to staff ministries so that they can develop their own expertise or commission outside expertise (Panke 2012, 316; see also Gray 2018). 9 Second, ministry officials from different regions could differ in their openness toward international bureaucracies and external advice in general. To account for such differences, we included dummy variables for Africa, the Americas, Asia and Oceania, and Europe. 10 We included fixed effects for each international bureaucracy in each thematic area in all models, to account for variations in IO's institutional designs, bureaucratic styles, and resources in different thematic areas (Knill et al. 2019;Patz and Goetz 2019). Since we are interested in determining the effects of our main explanatory variables on the consideration of policy advice, the potentially unobserved heterogeneity between, for example, the FAO in agribusiness and the IMF in debt management policy or the BIS and the IMF in banking regulation is beyond the scope of this article.
We ran multiple linear regressions with the independent and control variables specified above. 11 We used robust standard errors clustered on the respondent level (see Primo, Jacobsmeier, and Milyo 2007). 12 Models 1-5 include the full dataset, i.e., both policy fields (table 2). Model 1 reports the findings for the hypothesis on expertise (H1), while Model 2 tests the hypothesis on the objectivity of expertise (H2). More specifically, we used an interaction term of the variables "expertise" and "objectivity" to test whether the effect of expertise on the consideration of policy advice is conditional on the extent to which ministry officials regard international bureaucracies' work to be unbiased.
Models 3-5 include interaction terms, which combine expertise with the three alternative means of influence: congruence, pressure, and coercion. 13 First, if we find that expertise matters when one of the alternative factors is absent (or relatively weak), we can take this as further validation of the centrality of expertise for international bureaucracies' influence. Second, if we find that expertise and another factor reinforce each other, expertise can still be regarded as contributing to an international bureaucracy's influence by strengthening an alternative pathway. 8 Data taken from World Bank Open Data website, available at http://data.worldbank.org, accessed May 3, 2018. 9 Another conceivable indicator is GDP per capita, but it does not consider the possibility that countries with high GDP per capita but a small population size may still have few human resources to produce knowledge and therefore have greater demand for external knowledge. Nevertheless, we have included GDP per capita in the robustness checks below. 10 We pooled the observations of Asia and Oceania into a single regional dummy given the comparatively small number of countries in the Oceania region. 11 Since we used ordinal data (seven-point response scales) in a linear regression, we additionally verified our results by using ordered logistic regressions (see robustness checks below). As we used the same scale units in the dependent and independent variables, we can directly compare the effect size of all survey-based variables in the multiple linear regressions. 12 Since we asked each respondent to evaluate several international bureaucracies in one questionnaire, these evaluations may be correlated. To adjust for this possibility, we clustered the standard errors at the respondent level. Although we assume that our survey respondents are familiar with the work of international bureaucracies working in the same thematic areas as themselves, we offered the response category "not aware of any advice" or "work not known" (as explained in the online appendix) and we encouraged respondents to think of particular encounters with international bureaucracies. 13 To ease interpretation of regression tables and interaction terms, we rescaled the dependent and independent variables to a scale of 0-6, so that 0 refers to the response category 1.
In order to ensure that potential differences across the two broader policy fields do not distort the results, we additionally ran the same regression models for each of the two policy fields separately (see Models 6-15 in online appendix E).
This research design has a few limitations. Since our focus is on agriculture and finance, we can neither say whether nor explain why international bureaucracies influence the policy-making of national policy-makers in other fields. While our country sample is representative of all countries globally, our selection of international bureaucracies and thematic areas is not representative of the universe of international bureaucracies or thematic areas, but was made following the criteria outlined above. Thus, it is conceivable that one might find different associations in other policy fields. Future studies might extend the comparative analysis to include additional international bureaucracies, thematic areas, and policy fields. They might also expand the survey-based assessment to other domestic stakeholders. Moreover, we acknowledge that we have not examined either the process by which international bureaucracies generate their knowledge or the different kinds of expertise they generate, or the roles international bureaucracies adopt when advising decision-makers. Many of these aspects are difficult to study empirically, at least in a large-N setting, given a lack of proper operationalization and clear expectations on their impact (see also Christensen 2020, 6).

Empirical Analysis
Our findings reinforce the importance ascribed to expertise. As Model 1 indicates, expertise (H1) has a substantial and statistically significant relationship to the consideration of policy advice. If expertise increases by one unit, the consideration of policy advice increases by 0.346 (table 2). This finding is robust when examining the policy-specific subsamples separately (see Models 6 and 11 in online appendix E). A one-unit increase in expertise is associated with an increase in consideration by 0.363 (agriculture) or 0.365 (finance). Furthermore, we find that expertise has a stronger association with consideration than any of the alternative means of influence. It is even slightly more important than congruence, i.e., whether ministry officials deem the advice provided to be in their country's interest (0.332). It is also substantially more important than coercion and pressure. While both of these variables are statistically significant, the association is weaker. The main difference between the policy fields is that pressure and coercion have no significant effect in agriculture, whereas in finance the effect of congruence is slightly higher than that of expertise.
While the open-ended survey responses we received are not representative, some respondents used the opportunity to explain why they did or did not consider international bureaucracies' advice. Some noted that "there are several good papers/books we refer to from time to time […] and discuss with peers"; others commented that only the studies of two particular international bureaucracies were of relevance for them, as only these were "scientifically sound." To test Hypothesis 2 on objective expertise, we included an interaction term between expertise and objectivity in Model 2. The regression results do not reveal any statistically significant interaction, and we can therefore reject Hypothesis 2. Expertise seems to be related to the consideration of policy advice, independent of the level of objectivity. As the visualization of the marginal effects shows, expertise also has a statistically significant effect when objectivity is judged as very low, i.e., when national bureaucrats regard the  bureaucracy's work as strongly biased toward particular values and ideologies (figure 3, see also online appendix F). This finding should come as a surprise to those who claim that international bureaucracies' influence depends on their appearance of objectivity.
In the following, we move to the results of the models that include the interaction terms between expertise and congruence, pressure, or coercion. We are interested in the interplay of expertise and the other factors when explaining the consideration of policy advice, because it provides another important check on the centrality of expertise. First, interaction terms allow us to probe whether expertise matters most and, second, whether alternative factors matter only when expertise is attributed to an international bureaucracy: How is the effect of expertise altered by the absence or presence of the other factors? How do different levels of expertise alter the effect of alternative means of influence? Again, we are looking at the effects for the full sample (table 2, Models 3-5) as well as for each policy field separately (Models 8-10 and 13-15, online appendix E).
Although all interaction terms are statistically insignificant (table 2), interplays at certain levels of the respective variables are still conceivable (Brambor, Roberts Clark, and Golder 2006, 74). Therefore, we provide an additional visualization of the marginal effects.
Most importantly, international bureaucracies' expertise remains the central determinant of the extent to which their policy advice is taken into consideration. Throughout the reports on the marginal effects for the full sample and all but one report on the policy-specific subsamples, expertise is correlated with the consideration of policy advice, and the correlations are statistically significant (see online appendices E and F). The effect is independent of the level of the other variable of interaction, i.e., whether congruence, pressure, or coercion is absent, low, or high. Expertise is nearly always associated with the consideration of policy advice, irrespective of the levels of the other variables (online appendix F). There is one exception to this general trend: In the policy field of agriculture, an international bureaucracy's expertise has no statistically significant association with the consideration of policy advice when that advice diverges entirely from domestic policy preferences. 14 Conversely, we found that the associations of the alternative explanations with the consideration of policy advice often depend on expertise: The congruence of policy advice with national preferences only matters when the bureaucracy is assessed to have at least a small degree of expertise (figure 4). In other words, tailoring the policy advice to national preferences does not lead to the consideration of policy advice if the international bureaucracy is assessed to have no or little expertise (i.e., response categories 1 and 2).
Similarly, the correlation between pressure and consideration of policy advice is also conditional on the level of expertise in the policy field of finance (figure 5). Here, pressure only matters when international bureaucracies enjoy a high degree of expertise (i.e., response categories 6 and 7).
This means that, for instance, even if international bureaucracies in this policy field try to use third-party pressure to increase the extent to which their policy advice is taken into consideration, they still have to be seen as experts. This means that pressure has to be backed up by a clear impression that international bureaucracies possess relevant expertise in their field.
Finally, we find mixed results for the marginal effects of coercion and expertise (see online appendix F). In finance, the consideration of policy advice is always correlated with coercion, except when the bureaucracy is perceived to have no expertise at all. Conversely, in the policy field of agriculture, the association is not statistically significant, except when the international bureaucracy is perceived to have no or little expertise. Hence, the interplay between expertise and each of the means of manipulating cost-benefit calculations (i.e., pressure and coercion) appears to have opposing 14 This is the response category indicating the lowest degree of congruence.
The exact wording of the questionnaire was "advice is completely different." Figure 5. Average marginal effects of the variable "pressure" on the consideration of policy advice, conditional on the level of "expertise" in the policy field of finance. Notes: 95 percent confidence interval; the effect is indistinguishable from zero when the bars cross the zero line. effects in the two policy fields. Both increase the consideration of the advice on financial policies but not of the advice on agricultural policies, where pressure has no effect and coercion only has an effect when the bureaucracy has little or no expertise.
Overall, these results highlight the importance of expertise attributed to international bureaucracies for the influence they are able to exert. An international bureaucracy's expertise seems to lead to their policy advice being taken into consideration, independent of other means of being heard. In fact, other determinants only matter when a certain level of expertise is ascribed. This proved to be the source of the only substantial difference between the two policy fields: In agriculture, the effect of congruence depends on a certain level of expertise. In finance, the effect of third-party pressure depends on a certain level of expertise. We can only suggest a possible explanation: As discussed above, existing research has reported that third-party pressure in finance is quite frequent. Assuming that countries face pressure from various actors, they might be more willing to give in to pressure when they recognize the source of the advice as having expertise.

Robustness of the Results
We took several precautions to guard against biased inferences. Regarding the reliability of our survey data, we aimed at increasing reporting accuracy by using a self-administered survey, by developing the questionnaire with extreme care, and by thoroughly pretesting it (see also online appendix C). The data we collected show high reporting accuracy. In some cases, the questionnaire was forwarded to other ministry officials who were also working in the respective thematic areas. In these cases, we checked the extent to which the answers of a given respondent differed from the additional respondent(s) in the same country and thematic area. 15 The variance of the means of the dependent and independent variables in these cases was much smaller compared to the whole dataset.
Regarding the robustness of the results of our empirical analysis, we undertook several tests (online appendix G). Given that interpretation of the regression coefficients in ordered logistic regressions is more challenging and the coefficients are not directly comparable between models with different variables (Mood 2010), our empirical analysis above only reported the coefficients from ordinary least squares (OLS) estimations. However, a critical reader might prefer using ordered logit models to account for the ordinal scale of the dependent variable. Therefore, we verified our OLS estimates by performing ordered logistic regressions for Models 1-5. The results are similar.
Additionally, and given the (slightly) diverging response rates in the different regions and thematic areas, we ran our main regression Models 1-5 with non-response-adjusted survey weights to check for potential biases induced by nonresponse. We used weights at the levels of international bureaucracy and region. The key estimates retain significance, and the coefficient values change only slightly.
Furthermore, we accounted for further heterogeneity across countries and corrected for a possible omitted variable bias by controlling for additional differences at the level of the respondents' countries. Control variables include the countries' level of development, measured by their GDP per capita (GDP/capita (log)), 16 countries' political rights and civil liberties (FHI), 17 an indicator for corruption (corruption (WG)), 18 and the priority of the respective thematic area on the government's policy agenda as indicated by the respondents in our survey (policy importance).
Finally, we controlled for exposure to international bureaucracies' advice. While all international bureaucracies in our sample publish advice for countries worldwide at either the global or the regional level, they also advise countries individually. We therefore cannot ensure that domestic stakeholders have been equally exposed to their advice. Therefore, we collected additional data to create three indicators that account for variations in country-specific exposure to advice. First, we checked whether an international bureaucracy's advice is targeted explicitly at particular countries. Based on all country-specific reports by each international bureaucracy in 2014 and 2015, we created a binary indicator for each of the eight thematic areas. This covers the period respondents were asked about in the survey. Second, we gathered data on whether there was a national office for each international bureaucracy and country in our sample. A country office fosters the direct exchange between the staff of an international bureaucracy and a national ministry and, thus, facilitates the provision of general and countryspecific advice. Third, we investigated the level at which respondents were unaware of the advice of each international bureaucracy in a subregion. We calculated the proportion of the answer "not aware of any advice" per international bureaucracy in every given policy area and region. The indicator informs us about the level at which neighboring states are exposed to the advice of a bureaucracy. We assume that international bureaucracies approach states with regional proximity in somewhat similar ways, and that awareness levels in a subregion are somewhat similar. 16

Conclusion
This article examined whether expertise plays a role in the ability of international bureaucracies to influence various countries' domestic policy-making. Many international bureaucracies provide advice, which they typically direct toward national policy-makers. Yet, the research to date has not provided systematic evidence on the extent to which these gatekeepers of advice at the national level actually listen to international bureaucracies and why they do so. We regard this as a serious limitation because the consideration of international bureaucracies' policy advice can be taken as an indicator of their influence. We presented a large-N effort to assess the consideration of international bureaucracies' policy advice across policy fields, bureaucracies, and countries. To do so, we surveyed a representative sample of ministerial units in more than a hundred countries across the globe. Our survey covers nine global and nine regional bureaucracies and two areas of global public policy: finance and agriculture. This sample is composed of many more than the "usual suspects" of IO research. It includes bureaucracies with varying resources, memberships, policy scopes, and bureaucratic styles.
We utilized these data to test whether expertise is the primary source of international bureaucracies' influence. Scholars in global governance, PA, and IO research have long argued that expertise is central to international bureaucracies' influence. However, the evidence used in previous studies stemmed mainly from case-study research. For the first time, we tested the external validity of existing claims about the centrality of international bureaucracies' expertise and, thus, the direct influence of international bureaucracies. Does expertise explain the consideration of policy advice? Does (only) objective expertise matter? How does our argument compare to other explanations? Is the effect of expertise dependent on such alternative means of influence, or are their effects actually dependent on expertise?
The findings of our empirical analysis provide strong support for our expertise-based explanation. First, the perception of international bureaucracies as experts opens the door to domestic policy-making and allows bureaucracies to exercise an influence at the national level. Second, we found an effect for expertise even when the international bureaucracy is not regarded as objective. Third, when comparing expertise with alternative means of influence (congruence, pressure, and coercion), we found that expertise has the most substantial effect. Fourth, when examining interaction terms, we also found that the effects of alternative means of exercising influence depend partly on expertise. Fifth, our findings even hold when accounting for countries' (varying) GDP, which can be seen as a proxy for the human resources within national ministries. All this confirms the centrality of expertise.
Our article contributes to the burgeoning literature on international bureaucracies' autonomy and policy influence, on the one hand, and on the role of expertise in global governance and policy-making, on the other hand. First, by examining the practices of key addressees of international bureaucracies' advice, we were able to trace their "hidden" influence. We have systematically shown that international bureaucracies' advice does not require either third-party pressure or IO coercion to be taken into consideration by this important group of domestic stakeholders. Hence, international bureaucracies are able to have an influence on national policy-making without assistance from other actors.
Second, our research shows how an analysis of stakeholder attributions, e.g., by means of surveys, can fruitfully complement our understanding of the role of international bureaucracies. Expertise is not naturally located in these types of actors, and they do not necessarily influence the policy-making of those at whom their advice is directed. Given the striking variation in the extent to which international bureaucracies' advice is taken into consideration, we should be cautious when attributing influence to the bureaucracies of certain IOs or to international bureaucracies per se.
Third, our findings can inform more recent efforts to study how international bureaucracies interact or are even intertwined with domestic stakeholders. Our research indicates that the consideration of international bureaucracies' advice depends on the expertise attributed to them, i.e., on the perceptions of what counts as specialized and credible knowledge by those at whom the advice is targeted. This ultimately leads to the question of why some international bureaucracies are regarded as better experts than others. In this context, it seems promising for future research to further explore both the links between realist and relational conceptions of expertise and the role of professional networks.
From a policy perspective, our findings suggest that international bureaucracies are even influential in times of fragmentation and significant budget cuts. National ministries are important gatekeepers that consider the advice of international bureaucracies when they regard them as experts. Even though many populist governments oppose IOs, national bureaucrats still allow them to "invisibly" influence domestic policy-making.
Although we have argued that national ministries are important interlocutors, we acknowledge that they are not the only domestic addressees of international bureaucracies' advice. Given the complexity of domestic policy-making processes, future research could expand the assessment to other domestic stakeholders, for example, parliamentarians. Our work also invites further examination of policy areas that have been less in the spotlight of IO research to date. Pursuing these avenues for future research would help to further specify the relationship between expertise and policymaking and also to further generalize the results.

Supplementary Information
Supplementary information is available at the International Studies Quarterly data archive.