1 Introduction

All over the world, governments subsidize the consumption and the production of fossil fuels. Notwithstanding the environmental consequences of these subsidies, the international push for fossil fuel subsidy reform has come from economic international organizations (IOs), particularly the Organisation for Economic Co-operation and Development (OECD) and the International Monetary Fund (IMF), rather than environmental institutions.Footnote 1 While other institutions such as the International Energy Agency (IEA) and the World Bank have also been heavily involved in fossil fuel subsidies, the present article focuses on the IMF and the OECD since they are the two institutions that—in spite of similarities—differ most in their approach to fossil fuel subsidies. The IMF has adopted an approach that constitutes a radical break with previous definitions and defines any non-pricing of externalities as a subsidy, while the OECD uses a more conventional definition. The IMF approach led to an estimate of global fossil fuel subsidies of US$5300 billion, compared to other estimates of approximately US$550 billion (IEA 2014; Coady et al. 2015). This difference is puzzling considering the similarities between the institutions, notably their shared liberal economic worldview and the fact that the same set of industrialized countries prevail within the collective principal (Bernstein 2001; Vreeland 2007). Moreover, unlike the IMF, the OECD has a record of addressing environmental problems (Lehtonen 2007). So, if anything, one would expect the OECD to be more inclined to incorporate environmental concerns rather than the IMF.

Notwithstanding agreement that fossil fuel subsidies should be reformed, little agreement exists on the exact definition of such subsidies (Koplow 2009). An important aspect of the definition of fossil fuel subsidies is the emphasis on, respectively, the environmental, fiscal, macroeconomic and distributive consequences. If the IO’s engagement was motivated by the fiscal and economic consequences, why have they only addressed fossil fuel subsidies as a major issue since the late 2000s, when the policies today characterized as fossil fuel subsidies have in fact existed for decades? Environmental concerns constitute another possible factor, but if it was the main driver, environmental institutions should all things equal be more involved than economic IOs, which is not the case.

The growing body of the literature on fossil fuel subsidies and their reform has predominantly focused on domestic factors (Victor 2009; Bazilian and Onyeji 2012; Cheon et al. 2013). The issue of international economic institutions in fossil fuel subsidy reform has been overlooked—except for treating them as one factor among many causing domestic fossil fuel subsidy reform (Beaton and Lontoh 2010).

To address the above-mentioned puzzles and gaps, this article will seek to answer the following research question: Which factors explain why and how the IMF and the OECD have addressed fossil fuel subsidies? The article aims at identifying the factors that induced the economic IOs to address fossil fuel subsidies and influenced how they defined these subsidies. The factors studied include a hitherto overlooked factor in the IO literature: interaction with other international institutions, as well as the theoretically derived factors of IO bureaucratic culture and policy entrepreneurs, principal-agent autonomy, and decision-making procedures.

The article proceeds with a discussion of the various definitions of fossil fuel subsidies. Subsequently, the article develops a theoretical framework for studying the economic IOs’ efforts to address fossil fuel subsidies, a framework combining theories of institutional interaction and theories of IO behaviour. The theoretical framework is applied to the cases of the OECD and the IMF efforts to address fossil fuel subsidies.

2 Fossil fuel subsidies and their definition

Although fossil fuel subsidiesFootnote 2 are widely recognized as problematic, there is no agreement on their definition (OECD Secretariat 2010). While policies lowering the fuel price paid by consumers below the market price undisputedly constitute a fossil fuel subsidy, several other types of policies may be defined as fossil fuel subsidies, depending on the definition used (van Asselt and Skovgaard 2016).

There are two important distinctions for classifying the definitions of fossil fuel subsidies: first, the distinction between the definitions that define a fossil fuel subsidy in terms of the benefits conferred on a specific group and the definitions that define it in terms of a price-gap between the actual price and a benchmark price (Koplow 2009; OECD Secretariat 2010). The conferred-benefits approaches focus on individual policies, while the price-gap approaches focus on fuel prices as the unit of analysis. The price-gap approach identifies fossil fuel subsidies in terms of impact on consumer prices, more precisely whether prices are below a benchmark price. The benchmark price is based on the international price of a given fossil fuel, often including distribution, value-added tax and/or a tax corresponding to the externalities stemming from fuel use (van Asselt and Skovgaard 2016). Several countries have fossil fuel subsidies according to both definitions, but numerous countries will only be included under conferred-benefits definitions and not under price-gap definitions (Koplow 2009) and vice versa. The definition used also determines the estimated size of estimate of global fossil fuel subsidies.

The second distinction differentiates between producer subsidies (directed at the producers of fossil fuels) and consumer subsidies (directed at the users of fossil fuels). Consumer subsidies include reduced electricity prices, fuels sold at below-market prices, etc. Producer subsidies include direct financial transfers, tax rebates and loans to fossil fuel extraction, etc. Consumer subsidies are concentrated in the developing countries, while producer subsidies are also common in the industrialized countries.

Arguably, fossil fuel subsidies can be viewed as an ‘essentially contested concept’, whose criteria for application cannot be agreed upon, but which depends upon differing normative points of view (Connolly 1983). Defining a policy as a fossil fuel subsidy implies that it should be reformed, which is one reason why actors promote definitions that identify different policies as fossil fuel subsidies.

The contestation over fossil fuel subsidies may concern their environmental, fiscal, macroeconomic and distributive consequences (for a related typology, see Lockwood 2015). Regarding environmental consequences, the focus has been on climate change as well as local air pollution (e.g. Coady et al. 2015). Concerning fiscal consequences, most fossil fuel subsidies constitute public expenditure, either as direct spending or as revenue foregone (e.g. lower tax rates). Regarding the macroeconomic consequences, fossil fuel subsidies have been framed as distorting the optimal allocation of resources within society, for instance by encouraging over-consumption (see Clements et al. 2013, 15–19). While the previous consequences of fossil fuels are negative, its distributive consequences have been framed positively in terms of poverty reduction policies (in policy debates rather than academic literature) as well as negatively in terms of being regressive (for a discussion of the positive and negative distributional consequences, see Lockwood 2015, 479).

3 Theories of IO behaviour

A strand of the international relations literature focuses on IOs as actors in their own right, independent of state behaviour (Nielson and Tierney 2003; Barnett and Finnemore 2004; Park and Vetterlein 2010). Biermann et al. (2009b) distinguish between three kinds of influences on IOs: problem structure, extra-organizational (mainly member states), and the organization itself. Whereas the problem structure (fossil fuel subsidies and their reform) is constant between the IOs analysed here, both extra- and intra-organizational factors vary. Intra-organizational influences explain the role of IOs in terms of their organizational culture and policy entrepreneurs within the bureaucracies (Barnett and Finnemore 2004; Park and Vetterlein 2010). Extra-organizational influences mainly explain the role of IOs in terms of their status as agents contracted by principals (the member states) to perform a function that will benefit the principals (Nielson and Tierney 2003; Hawkins et al. 2006).

The present article argues that most of the literature on IO output ignores the interaction between international institutions (Young 2011; but see Van de Graaf 2016).Footnote 3 The developing literature on the dyadic interaction between institutions (Oberthür and Stokke 2011) and on the fragmentation and coupling of institutions into ‘regime complexes’ (Biermann et al. 2009a; Keohane and Victor 2011) rests on the assumption that international institutions cannot be understood without including their relationships to other institutions. Thus, interaction with other international institutions influences when and how an IO addresses fossil fuel subsidies. The two IOs differ in the OECD being requested by the G20 to study the scope of global fossil fuel subsidies and has become something akin to the secretariat of the G20, while the IMF was not commissioned by any institution.

Proposition 1

Interaction with international institutions explains the difference in how the two IOs addressed fossil fuel subsidies.

Beyond institutional interaction, explanations based on existing theories will be explored. Regarding intra-bureaucratic factors, the sociological institutionalist IO literature focuses on the differing perspectives of IO bureaucracies (Barnett and Finnemore 2004; Park and Vetterlein 2010). The OECD has a track record in dealing with environmental issues going back to the 1970s (see below) and has been crucial in promoting the paradigm of ‘liberal environmentalism’, stressing economic instruments and compatibility between economic growth and environmental protection (Bernstein 2001). The IMF has little experience regarding environmental issues and is more strongly influenced by neoclassical economics than the OECD (Chwieroth 2008; Howarth and Sadeh 2011). Thus, we should expect the OECD to frame fossil fuel subsidies in terms of environmental consequences to a larger degree than the IMF.

Proposition 2a

The organizational culture of the IOs explains the difference in how they addressed fossil fuel subsidies.

Policy entrepreneurs (understood as actors investing resources in order to change policy; Kingdon 2003, 122–124) operating within the organizations may promote and frame issues in ways that fit with existing ideas and lead to particular policy responses (Chwieroth 2008), in this case how the organization addresses fossil fuel subsidies.

Proposition 2b

Differences in the degree of policy entrepreneurship among the staff of the IOs explain the difference in how they addressed fossil fuel subsidies.

Turning to extra-bureaucratic factors, principal-agent theory focuses on the degree of IO autonomy from principals as the main explanatory factor of the ways in which the IOs have addressed fossil fuel subsidies. Since the IMF controls its own resources and operates more independently of the member states than the OECD (Dreher 2004; Busch 2009), one would expect the IMF staff to be more able to influence IO behaviour on fossil fuel subsidies than the OECD bureaucracy.

Proposition 3a

The degree of autonomy from principals explains the difference in how the IOs addressed fossil fuel subsidies.

Finally, membership and voting rules within the collective principal of an IO may influence how it addresses fossil fuel subsidies. The OECD covers only industrialized countries, and while the IMF has a broader set of principals covering most countries, its voting rules grant the major industrialized countries a position close to a combined veto power.Footnote 4 Therefore, the IMF would be expected to reflect the preferences of the largest industrialized countries to a larger degree than the OECD.

Proposition 3b

Membership and voting rules within the collective principal of an IO explain the difference in how they addressed fossil fuel subsidies.

4 Operationalization

The OECD and the IMF treatment of fossil fuel subsidies have been analysed on the basis of official documents and 20 key informant interviews. The key informants are civil servants from the OECD, the IMF and member states characterized by strong engagement with fossil fuel subsidy reform (the USA, the UK, India, and Sweden). At both IOs, a limited number of officials have fossil fuel subsidies as their main responsibility, and in both cases I have interviewed a significant share of these officials (a total of 10), as well as officials having fossil fuel subsidies as a smaller but nevertheless important part of their responsibilities, such as officials working with IMF country programs. They were interviewed during the period 2014–2016 at the headquarters of the two organizations, in the national capitals, via phone or email, and at the national representations to the OECD. The interviews were semi-structured, with the informants being asked the same general questions as well as more specific questions regarding their individual responsibilities. The analysis of the documents goes back to the first documents addressing fossil fuel or energy subsidies, be it as a distinct issue or as one among many.

The analysis has uncovered how the institutions have defined fossil fuel subsidies as well as the processes leading to the institutions addressing fossil fuel subsidies. Regarding the former, the analysis focuses on the IOs’ definition of fossil fuel subsidies as well as how they have framed such subsidies in terms of environmental, fiscal, macroeconomic or distributive impact. The definitions and frames have mainly been identified on the basis of official documents, whereas the mapping of the policy processes has relied on the interviews.

5 International efforts to reform fossil fuel subsidies

Fossil fuel subsidies became a high-priority international issue only when the G20 leaders in September 2009 in Pittsburgh adopted the commitment ‘[t]o phase out and rationalize over the medium term inefficient fossil fuel subsidies while providing targeted support for the poorest’ (G20 2009). This commitment was driven by US leadership and adopted in a context of high oil prices (interview with former senior US Treasury official, 6 May 2014) and importantly did not provide a definition of fossil fuel subsidies. Yet policies now characterized as fossil fuel subsidies have been in place for decades. Multilateral development banks and the IMF have long advocated that countries phase out ‘poorly targeted’ subsidies of all kinds without using the concept of fossil fuel subsidies (e.g. IMF 2004).

The G20 commitment was most influential in the processes it set in motion. The IEA, the OECD, OPEC, and the World Bank were tasked with measuring the magnitude and the consequences of such subsidies, a request which moved the issue up their respective agendas. The G20 member states committed themselves to submitting strategies and timetables for phasing out their fossil fuel subsidies (G20 2009). Furthermore, in November 2009, the governments of the Asia Pacific Economic Co-operation countries adopted a commitment similar to the G20’s. Although environmental institutions addressed fossil fuel subsidy reform indirectly and directly in the 1997 Kyoto Protocol and the 2015 Sustainable Development Goals, but only as a possible means of implementation (van Asselt and Skovgaard 2016; see also van Asselt and Kulovesi, this issue).

6 The IMF

The IMF has traditionally not emphasized environmental protection. The considerable autonomy and power of the IMF is—according to Barnett and Finnemore (2004)—based on its independence from state funding and its authority on economic matters. The economic training of the IMF officials is fundamental to its bureaucratic culture and how the institution perceives and acts upon the world (Chwieroth 2008).

Roughly speaking, IMF policy has addressed fossil fuel along two strands, both increasing in importance. The first strand focuses on the lack of a carbon price (and environmental taxes generally) and on solving this problem from the perspective of an economist, i.e. ‘getting the price right’ (interview with IMF senior official, 17 February 2015). Prior to 2008, the IMF only occasionally addressed energy subsidies (rather than fossil fuel subsidies) in policy reportsFootnote 5 (Gupta et al. 2000; Baig et al. 2007) and in country-specific policy recommendations (e.g. IMF 2004). Energy subsidies were framed in terms of fiscal and macroeconomic impact, without referring to environmental impacts. Thus, subsidizing fossil fuels was framed as similar to subsidizing any other product. The IMF used price-gap approaches for measuring all kinds of subsidies and did not include externalities in its fuel benchmark prices (e.g. Said and Leigh 2006). Internally, from 2005 fossil fuel subsidies gained increasing attention, particularly driven by the fiscal impact of high oil prices and the Poverty and Social Impact Analysis Group within the Fiscal Affairs Department, and consequently the IMF stressed the fiscal impact of fossil fuel subsidies while stressing the importance of mitigating measures to protect the poor (interview with IMF official, 25 February 2015). Thus, the oil prices provided policy entrepreneurs within the IMF with a window of opportunity.

In 2008, the IMF published its first official publicationsFootnote 6 that addressed fuel subsidies as a distinct concept and included environmental externalities (priced at US$0.50 per litre gasoline and diesel; IMF 2008a, b). After 2008, the IMF increasingly addressed fossil fuel subsidies and their environmental impact, while maintaining the emphasis on fiscal and macroeconomic consequences. This increase was due to policy entrepreneurship within the organization. The notion of including environmental externalities in measures of efficient fuel prices had been floating around for some time among a circle of economists working for the IMF, the World Bank, the US government and various environmental think tanks in Washington, DC (interview with senior IMF economist, 24 April 2014). The IMF economists within this group also promoted the notion of including undercharging for environmental costs in a broad definition of fossil fuel subsidies, and this issue has received considerable attention from IMF management since 2011Footnote 7 under Christine Lagarde (interview with senior IMF economist, 24 April 2014). Lagarde had a more active interest in climate politics than her predecessors, which opened a window of opportunity for the above-mentioned economists (including members of the Poverty and Social Impact Analysis Group). In 2013, they published the report ‘Energy subsidy reform: Lessons and implications’ (Clements et al. 2013), which raised the IMF’s engagement with fossil fuel subsidies to a new level. Crucially, the report used a price-gap approach based on a benchmark price including both value-added tax and the social cost of externalities, particularly climate change and other environmental externalities. This approach was adopted on the basis of the IMF’s work on fiscal instruments (within the Fiscal Affairs Department) and the emphasis on—as the title of a key publication on fossil fuel subsidies says—‘Getting energy prices right’ (Parry et al. 2014).Footnote 8

The findings of the report were updated in a 2015 IMF working paper (Coady et al. 2015), whose estimate of global subsidies at US$5300 billionFootnote 9 in 2015—compared to estimates of US$1900 billion in the 2013 report (Clements et al. 2013) and US$550 billion in the IEA’s 2014 World Energy Outlook (IEA 2014)—received significant attention. This increase was due to revising the assessments of externalities—especially air pollution—upwards. Local air pollution accounted for three quarters of the externality and climate change for one quarter (Coady et al. 2015). Regarding climate change, the social cost of emitting a ton of carbon dioxide was estimated at US$35 (based on Parry et al. 2014), a figure originating from the US government’s Interagency Working Group on Social Cost of Carbon (2013). The US$35 ton figure is considered a low estimate of the social costs (see Tol 2011), yet it is much higher than actual levels of carbon taxation.

According to the IMF’s definition, practically all states subsidize fossil fuels, even countries with carbon taxes that do not fully price in externalities. Industrialized countries account for a quarter of energy subsidies, and emerging and developing Asia for half (Coady et al. 2015). Consequently, the 2013 and 2015 reports break with previous analyses which—using different definitions—identified fossil fuel subsidies as primarily a developing country phenomenon. The reports also contradicted the claims of influential IMF member states (including the UK and Japan) that they do not subsidize fossil fuels (G20 2012), something which made the report unpopular among the governments of these countries but popular among non-governmental organizations. This divergence underscores the limited member state influence on the reports (Proposition 3b). The influence of institutional interaction was also limited (Proposition 1), as the IMF was not requested by the G20 to analyse fossil fuel subsidies.

More important were the policy entrepreneurs within the IMF—particularly the Fiscal Affairs Department (Proposition 2a), who framed the environmental impact in macroeconomic terms, and thus successfully defined fossil fuel subsidies in a way resonating with the organization’s bureaucratic culture (Proposition 2b) as well as the professional background of IMF officials. It is difficult to distinguish the influence of the IMF’s bureaucratic culture from the professional background of its officials, since they lead to similar positions due to shared roots in neoclassical economics. The negative fiscal consequences were still emphasized. Regarding distributive consequences, the IMF framed fossil fuel subsidies as a highly ineffective way of supporting low-income households.

The second strand consists of bilateral interactions with countries having fiscal problems exacerbated by fossil fuel subsidies and includes policy recommendations and Extended Credit Facility Arrangements inter alia promoting subsidy reform to improve fiscal balances (interview with IMF senior official, 17 February 2015). Over the last 10 years, fossil fuel subsidies have been addressed in an increasing number of IMF recommendations to individual countries (see Table 1) and have increasingly been treated as distinct from other kinds of subsidies. The IMF suggested phasing out such subsidies as they were an inefficient, fiscally costly, and often economically distorting way of providing welfare benefits (interview with IMF officials, 9 April 2014). The 2013–2016 IMF Extended Credit Facility Arrangement for Burkina Faso constitutes an example of this, with detailed recommendations on how Burkina Faso should liberalize the government-fixed fuel prices (IMF 2015; interview with senior IMF official 22 June 2016). This strand mainly emphasizes the fiscal consequences of the subsidies, while also stressing the distributional (both positive and negative) and macroeconomic consequences. Environmental consequences (including local externalities) have been less accentuated. Regarding Burkina Faso, the IMF policy was developed by local IMF officials and officials from the Fiscal Affairs Department and focused on what the IMF refers to as pre-tax subsidies, namely the subsidies lowering consumer prices below the international market price plus distribution costs, not the absence of full pricing of the externalities (interview with senior IMF official 22 June 2016).

Table 1 IMF reports on fossil fuel subsidies

7 The OECD

The OECD’s membership consists of 35 industrialized countries. The term ‘OECD’ refers to the entirety of the OECD, including the OECD Council (consisting of member state representatives, committees, working groups, etc., which report to the Council), as well as the OECD Secretariat, an intergovernmental bureaucracy constituting an independent actor. The Secretariat staff drafts all OECD publications, which are then subject to peer review in OECD committees. Those publications that represent the opinion of the OECD as a whole pass through consensus-based approval by the member states, while those that represent the opinion of the OECD Secretariat need approval only from the Secretary General (Ruffing 2010).

The OECD Environment Directorate has for four decades been a key actor in crafting environmental policy solutions based on environmental (predominantly neoclassical) economics (Ruffing 2010). The Environment Directorate was instrumental in developing the paradigm of liberal environmentalism, which is prevalent within global environmental policy and predicates international environmental protection on a liberal economic order (Bernstein 2001). Since the OECD does not possess hard, legal instruments to induce states to change policy, the OECD Secretariat and its Environment Directorate relies on softer methods such as learning, socialization into norms and peer review (Lehtonen 2007; Busch 2009). The OECD Secretariat is increasingly acting as a kind of secretariat to the G20, providing analyses of key issues, including fossil fuel subsidies (see also Lesage and Van de Graaf 2013).

Crucially, the OECD tends to talk about ‘support’ rather than ‘subsidies’, since the term subsidy is seen as referring to a smaller set of measures than support (supported fuels can still be priced above the world market price) and since it may be legally problematic in relation to World Trade Organization (WTO) disputes to define a measure as a subsidy (interview with OECD officials, 29 April 2015; see also Meyer, this issue). The OECD uses the term fossil fuel support in the sense that others use fossil fuel subsidies, which is evident in the joint reports to the G20 on fossil fuel subsidies authored by the OECD, World Bank, IEA, and OPEC (IEA et al. 2010, 2011). Hence, when analysing how the OECD addressed fossil fuel subsidies, this article will focus on its efforts regarding what it itself refers to as fossil fuel support.

The OECD addressed fossil fuel subsidies before the G20 commitment as part of the regular environmental performance reviews of individual member states, studies of pricing policies and more general studies.Footnote 10 These efforts tended to categorize fossil fuel subsidies together with other subsidies damaging to the environment, such as agricultural subsidies (OECD 2005).

In the summer of 2009—before the Pittsburgh summit—an OECD report modelled the climate change (and economic) impact of eliminating the subsidies measured by the IEA (OECD 2009), thus constituting a result of institutional interaction driven by the OECD. Furthermore, institutional interaction in terms of the G20 requesting the OECD Secretariat (among others) to analyse the scope of fossil fuel subsidies and suggestions for the implementation of this initiative lifted OECD involvement to a new level (interview with OECD officials, 29 April 2015). It was only following the G20 commitment that the member states gave the OECD Secretariat the mandate to scrutinize their national fossil fuel subsidies (interview with OECD officials, 29 April 2015). At a later stage, the decision by the G20 members that have so far committed to undergo a voluntary peer review of their fossil fuel subsidies (including China, Germany, Mexico, and the USA) to invite the OECD Secretariat to chair those peer reviews once again lifted the OECD Secretariat involvement to a new level. Since institutional interaction constitutes the most important factor increasing OECD interest in fossil fuel subsidies without much influence on how the OECD addressed the issue, Proposition 1 is more relevant regarding the timing of OECD involvement than the content of this involvement. In the following years, the OECD Secretariat arranged workshops for representative of member states and reported (individually and with IEA, OPEC, and the World Bank) to the G20 on fossil fuel subsidies (OECD Secretariat 2010, 2011; IEA et al. 2011).

Member states have not been directly involved (Proposition 3a), as the important OECD texts were published representing only the OECD Secretariat. Yet, member states played an indirect role by limiting how far the OECD staff could go (interview with Swedish Ministry of Foreign Affairs official, 30 April 2015). When it came to drafting the reports, especially the country-specific ones, the OECD staff collaborated with the World Bank and the IEA, which in this way influenced the OECD approach to the subject (interview with OECD officials, 29 April 2015). The Economics Department (the most influential OECD Directorate) and the Environment Directorate have also focused on fossil fuel support in, respectively, their ‘Economic Surveys’ and ‘Environmental Performance Reviews’, which analyse individual OECD member states.

The lead Directorate was the Trade and Agriculture Directorate in close collaboration with the Environment Directorate and the Centre for Tax Policy. The Trade and Agriculture Directorate’s prominent role is due to their agricultural subsidy expertise. Because of this institutional legacy, the OECD definition of fossil fuel subsidies is derived from the OECD definition of agricultural subsidies, again derived from the World Trade Organization’s definition of subsidies (of all kinds) as direct transfers, fiscal incentives and provision of goods and services (OECD Secretariat 2005). The past experience of working with agricultural subsidies was entrenched in the organizational culture and hence influenced the choice of a conferred-benefits approach and the general framing of fossil fuel subsidies. Importantly, the OECD Secretariat placed strong emphasis on environmental (particularly climate) and fiscal consequences of fossil fuel subsidies, but less on macroeconomic and distributive consequences. The fiscal emphasis is evident in that subsidies are measured in terms of budgetary expenditure and tax expenditure, i.e. their impact on public budgets, an emphasis also present in the OECD’s work on agricultural subsidies. The light emphasis on macroeconomic consequences is notable considering the macroeconomic objectives of the OECD.

In sum, how the OECD addressed fossil fuel subsidies has been shaped by the organizational culture (Proposition 2a) and more indirectly by institutional interaction (Proposition 1) with the WTO (and the World Bank and the IEA in the cases of the joint reports to the G20) but not the G20. Given the commitment of the OECD staff—particularly chief OECD expert on fossil fuel subsidies Ron Steenblik—to the conferred-benefits approach (Steenblik 2003), it is unlikely that the OECD would have defined the issue differently if the organization had had more autonomy from its principal. OECD staff played an important role in seizing the opportunity and produced reports on their own initiative (e.g. OECD 2015), in this way acting as policy entrepreneurs (Proposition 2b). Although the bureaucratic culture constituted the space within which the policy entrepreneurs had to operate, they nonetheless had considerable freedom within this space to determine how much emphasis to place on, for example, the environmental consequences of fossil fuel subsidies.

8 Conclusions

The two organizations differ notably in their definitions of fossil fuel subsidies, with the IMF using a price-gap definition that constitutes a radical break with previous definitions, while the OECD follows a more conventional path. The IMF to a larger degree than the OECD framed fossil fuel subsidies in environmental terms (climate change and local air pollution). This is surprising, since, unlike the OECD, the IMF does not have a strong environmental profile.

In terms of the propositions explaining the differences in IO output, institutional interaction (Proposition 1)—more precisely the G20 commitment—lifted OECD involvement in fossil fuel subsidies to a new level, yet without influencing how the IO addressed the subsidies. This was to some degree influenced by interaction with the WTO, and to a lesser degree the World Bank and the IEA. Such interaction did not influence the IMF, which was not requested by the G20 to address fossil fuel subsidies, but did so on the initiative of IMF staff. IMF staff acting as policy entrepreneurs are also the main reason why the IMF to a larger degree than the OECD framed fossil fuel subsidies in environmental terms (thus supporting Proposition 2b): IMF staff linked the macroeconomic and environmental framing of such subsidies on the basis of the neoclassical notion of getting the price right. The differences in how the organizations addressed fossil fuel subsidies were also influenced by their bureaucratic cultures (supporting Proposition 2a): the IMF framed the environmental impact in the above-mentioned neoclassical way, and the OECD framed fossil fuel subsidies in a fashion reflecting how it had addressed other subsidies.

The degree of autonomy of the IOs proved to be an important scope condition for the influence of bureaucratic culture and policy entrepreneurs, in this way supporting Proposition 3a. This is evident in that the IMF, to a greater degree than the OECD, adopted positions running against the preferences of its member states, most notably the claim that industrialized countries have significant fossil fuel subsidies. The OECD had less autonomy and only got the mandate to scrutinize its members’ subsidies after the G20 commitment. The differences between the two organizations demonstrate that differences in membership and voting rules did not have an impact, as the IMF contradicted member states that are more influential within the IMF than within the OECD, thus undermining Proposition 3b.

On a theoretical level, the article contributed to the study of IOs by demonstrating the importance of including interaction with other institutions. Intra-bureaucratic factors may be the key factor shaping how an IO addresses an issue, but institutional interaction and principal-agent relations constitute key scope conditions for which actions are possible for these bureaucracies. In the case of the OECD and other IOs with limited autonomy from their principal, IO interaction may increase the room for manoeuver. Furthermore, the analysis underscores the added academic value of focusing on how fossil fuel subsidies are defined. Future research could build coherent knowledge concerning how institutional interaction influences IO behaviour, inter alia by looking at entire institutional complexes.