Introduction

The International Monetary Fund’s (IMF, the Fund) 2010 Stand-By Arrangement with Greece (the SBA, Greece 2010) as part of the ‘Troika’ with the European Commission and the European Central Bank was unprecedented and a pivotal action in euro crisis decision-making. We find problematic gaps in the existing explanations for why the IMF chose to participate in the lending programme and what ideas guided the programme design. Hence, we reopen the case, as the SBA is interesting for multiple reasons. First, the Fund controversially overturned its own rules to provide “exceptional access” to Greece without requiring a restructuring of Greek public debt (Wyplosz and Sgherri 2017: 273–278): understanding this manoeuvre is crucial to understanding IMF behaviour during the crisis and the Fund’s relationship with eurozone. Second, Greece 2010 as the first manifestation of the eurozone crisis’ acute phase affected further crisis responses by IMF and eurozone actors and shaped the IMF’s global role. Third, the Greek economy suffered during the programme, directing interest at the austerity ideas affecting programme design. Given the sudden changes in IMF’s behaviour compared to its institutional development prior to the crisis and the considerable impact of the Greek economic deterioration to the eurozone institutional development, the first Greek programme may have been an especially critical juncture for both the Fund and eurozone. Thus, we contribute to scholarship on international monetary relations and on ideational drivers behind economic policymaking.

To better explain the Fund’s puzzling design of the SBA, we operationalise Ben Clift’s (2018) theory of IMF behaviour in a process-tracing exercise (Beach and Pedersen 2012; Bennett 2015) where we analyse primary and secondary text sources of IMF programme design. We use our interpretation of Clift’s theory as it appears promising for explaining the SBA and because it has evoked broad interest among scholars studying the IMF within the eurozone crisis (Ban 2020; Hodson 2020; Johnston 2020; Moschella 2020). Our process-tracing exercise within the field of international political economy is a novel exercise, as Clift does not aim to explain the detailed designs of individual programmes. Our causal mechanism stipulates that the IMF used two key analytical lenses, fiscal space and nonlinear multiple equilibria processes (NLMEP) to analyse the SBA situation and formulate the programme design. Further, we expect that these lenses were formed by key IMF bricoleurs (ibid.).

Our process-tracing exercise finds some limited support for our mechanism, but we also find that the mechanism is too narrow to fully explain SBA design. We therefore combine the results from our process-tracing with broader qualitative interpretations. Given the lack of other simple and coherent explanations, if a mid-range theory designed to explain IMF behaviour in the eurozone crisis has limited explanatory power, it is not useful to assume a narrow, coherent driver behind the SBA’s controversial design. Instead, IMF programme behaviour seems more accurately understood as driven by ad hoc judgement of crisis responders forced to make policy choices where incomplete ideational motivators produce action inconsistent with any single coherent framework or actor preference.

The (albeit limited) supportive process-tracing results we find of the analytical lenses of NLMEP and fiscal space being present in IMF programme analysis supports that financial contagion was an important motivator for IMF participation with significant monetary sums (NLMEP perspective). It also illuminates why, despite contagion fears, the IMF adhered to strict austerity due to limited Greek fiscal space even if this was detrimental to debt stabilisation crucial to reducing contagion effects. Thus, the programme design appears a compromise between motivating conditions (contagion) and procedural form conditions (fiscal space).

Finally, we find limited evidence of bricoleurs affecting the programme within our data. We also note for further research that the SBA could be potentially explained by a multi- veto player model. Such dynamics between the IMF and other Troika partners was referred to in the evidence, yet such a mechanism has not been decisively tested.

First, we discuss associated literature, the case context and the existing explanations for SBA design and their explanatory gaps. Second, we present Clift’s theory of IMF behaviour and our hypothesis. Third, we elaborate the process-tracing methods, the causal mechanism and the data. Then, we present the process-tracing results and discuss our broader interpretative evaluation of the material. To finish, we discuss what these considerations mean for our understanding of the SBA and the IMF’s role in the eurozone crisis, as well as for further research.

Literature review

A key area of IMF studies is its operations during the Global Financial Crisis and eurozone crisis, which helps understanding the Fund operating in future crises and promoting economic policy ideas. The SBA is especially interesting in this context. Understanding the rule change giving Greece “exceptional access” to funds without debt restructuring (Wyplosz and Sgherri 2017: 273–278) enables explaining the Fund’s broader relationship to crisis time programme design and its relationship with eurozone institutions. As Greece constituted the first manifestation of the eurozone’s existential crisis (Pisani-Ferry 2014), to which the IMF responded (Clift 2018; IEO 2016), it shaped both the IMF’s practices and the European policy environment regarding austerity policies implemented during the crisis. Further, austerity ideas implemented in the SBA are a key area of ideational economic policy research (see, e.g., Blyth 2013).

In early 2010, rising Greek spreads prohibited sovereign market access, and since eurozone membership prevented Greece from monetary financing, the government, facing default, requested eurozone and IMF assistance (IEO 2016: 10–11; Pisani-Ferry 2012: 5–6; 2014: 77–78; Xafa 2014: 7). Despite Greece’s request, the IMF involvement was not immediately certain, as Greece was not politically free to accept IMF assistance without paying heed to other monetary union members, especially since it did not have monetary sovereignty (Blustein 2015). The IMF involvement was seen by some eurozone members such as France as detrimental to European monetary integration, while by others, especially Germany, as useful due to the Fund’s track record in implementing sovereign lending programme conditionality arrangements, as well as for maintaining the status quo of current institutional arrangements as opposed to radical reform (Bastasin 2012; Blustein 2015).

After heated eurozone political debate, the IMF involvement was confirmed, but only by the IMF joining on junior partner terms: this, and the fact that the European Commission (EC), European Central Bank (ECB) and IMF negotiated the Greek assistance programme as the ‘Troika’, was unusual for the IMF as it ordinarily designs programme conditionality single-handedly (Blustein 2015; IEO 2016: 10–11, 33). Further, the IMF could not directly agree the programme conditions with the Greek authorities as Greece’s monetary policy authority had been moved into the ECB, and thus, ECB had a veto position for reforms requiring monetary policy action (Blustein 2015; Moschella 2016). The €30 billion SBA (the eurozone provided €80 billion to the joint programme) was then the largest IMF SBA when measured against an IMF member’s access quota (IMF 2010b: 1, 128–129; Wyplosz and Sgherri 2017: 257). Further, since the Fund’s Debt Sustainability Analysis projected the Greek public debt not to be sustainable with “high probability”, the Fund altered its own rules for providing “exceptional access” without requiring a restructuring: this was achieved with the “systemic exemption” clause that allows exceptional access in the face of “systemic spillovers” (Wyplosz and Sgherri 2017: 273–278).

The SBA was controversial. Greek output and employment decreased severely while neither macroeconomic conditions nor debt stabilised (IEO 2016: 30; IMF 2013: 11; Pénet 2018: 1047–1048; Pisani-Ferry et al 2013: 49). The decision not to restructure public debt during the SBA but only in 2012 (IEO 2016: 15) has been criticised. First, some argue that the debt was perceivably unsustainable, and late restructuring exacerbated Greece’s debt burden and crisis uncertainty (IMF 2013: 28). Second, the debt burden may have been unduly shifted from private creditors to taxpayers (Janssen 2010: 6; Pénet 2018: 1047–1048). Third, the rationales of eurozone-wide contagion justifying no restructuring (see, e.g., Confidential IMF memo, cited in Blustein 2015: 11; Wyplosz and Sgherri 2017: 275) have been considered vague and unconvincing (Schadler 2017: 49–52). Fourth, overturning the IMF access rules designed to prevent undue political pressure was perceived as favouring the eurozone, key policymakers of which opposed restructuring (Blustein 2015: 2, 10–17; Henning 2017: 89; IEO 2016: 15, 17; Pénet 2018: 1036–1037, 1048; Xafa 2014: 15).

The IMF argued that the SBA could succeed due to Greece’s demonstrated fiscal adjustment capacity and that the risk that a default posed to global financial stability justified the unprecedented access to ensure the success of the programme (IMF 2010b: 25). Responding to the IMF’s Independent Evaluation Office’s (IEO) critical 2016 assessment of the SBA, IMF then-Managing Director Lagarde diverged from the original narrative, claiming that to prevent a systemic meltdown the IMF lent exceptionally to “[buy] time to mobilize political support among Euro Area members to build firewalls and a crisis management framework” (IEO 2016: 57–58). This admission finds support elsewhere (IEO 2016: 30; IMF 2013: 28).

However, it is difficult to explain the size of the programme using the narrative of the holding operation. For temporary relief, the programme was exceptionally large, and then-Managing Director Strauss-Kahn had already made assurances during the negotiations that the IMF would commit €250 billion to further eurozone programmes, a promise that was eventually fulfilled (Blustein 2015: 14; IEO 2016: 3; Kincaid 2017: 152–153). Such involvement suggests the IMF contributed to solving the crisis from the outset and did not only aim to gain time. Accounts of the Troika crisis management suggest that there was little understanding on what the European crisis solutions would be. Therefore, the IMF seems unlikely to have committed such a substantial (and politically explosive) blank cheque to eurozone if the IMF had no plan crisis solution of its own. This makes it unduly reductive to label the entire purpose of the IMF involvement as a holding operation. The above is also supported by the experiences of the SBA’s mission staff, who worked hard to prevent an immediate catastrophe (Blustein 2015).

Some argue that the IMF mainly protected European banks exposed to Greek bonds, shielding them from default (Janssen 2010: 6; Pénet 2018: 1031, 1047). However, France and Germany, key eurozone states with substantial domestic banking exposures that enabled IMF participation (Bastasin 2012; Blustein 2015; Henning 2017; Pénet 2018, 1039), did not act accordingly. Both German (initially) and French policymakers opposed IMF involvement, and Germany, the pivotal player for securing IMF involvement, was motivated by the Fund’s technical expertise and ability to take a tough position on Greece (Bastasin 2012: 140–145; Blustein 2015: 5, 11; Henning 2017: 93–6, 99).

Others suggest that the programme was dictated by eurozone policymakers, who were reluctant to enact politically costly crisis-tackling eurozone reforms such as public monetary backstops, and who pressured the IMF with the European governments’ joint veto in the IMF’s Executive Board (EB) and control over the managing directorship (Blustein 2015: 2, 17–18; Pénet 2018: 1039, 1047–1048). However, little evidence suggests that eurozone policymakers had a complete rescue plan to impose on the IMF, given the conflict regarding restructuring which continued even after the SBA was agreed in May 2010 (Bastasin 2012: 145; Blustein 2015: 13, 17–18; Pisani-Ferry 2014: 77–78, 91–93; Spiegel 2014a; b; c). The EC had strong views on Greek structural reforms and likely mainly produced the related programme conditionality terms (Nordström and Teivainen forthcoming), while, in addition to some member states, both the EC and especially the ECB stringently opposed restructuring as damaging to eurozone institutions (Blustein 2015; Moschella 2016). However, no accounts suggest that either the EC or the ECB were integrated into the IMF’s crucial discussions regarding the final judgements on whether to finance Greece and how to make the rule change enabling exceptional access.

Blustein (2015: 3, 16–17) posits that the SBA was driven by the opinion of its then -Managing Director Strauss-Kahn that the IMF had undergone a relevance crisis, requiring participation to maintain global significance. The European reluctance to engage in political reform of the eurozone could have provided the Fund with an opening for involvement. According to Strauss-Kahn, the IMF had to pledge considerable resources and forgo restructuring to secure participation (interviewed in Blustein 2015: 6, 13–14). However, ex-post evaluations suggest the IMF’s invitation into, or relevance in, the Troika was not significantly affected by the size of its financial contribution (Kincaid 2017: 148, 178–184).

Finally, some argue that the Fund, facing opposition to restructuring, decided that the best option available was lending substantially and gambling on the success of an uncertain SBA as it could not refuse to act when faced with a Lehman-like crisis (IEO 2016: 15; IMF 2013: 32; Pisani-Ferry et al 2013: 75; Wyplosz and Sgherri 2017: 272–282). However, this argument seems unnecessarily dismissive of the Fund’s active crisis management and provides few leads for further research.

We suspect the first Greek lending programme to be an especially critical juncture for the development of the IMF and the eurozone crisis due to the unprecedented break it marks in the IMF institutional development and its potential path-dependency effects on the following eurozone crisis programmes. To explain the SBA, we have formulated a hypothesis based on our interpretation of Clift’s (2018) account of IMF behaviour. The theory is promising for explaining the Greek programme due to its specificity regarding Fund action in the eurozone. However, Clift (ibid.) aims to explain the broad institutional developments of the IMF during the eurozone crisis and does not focus on explaining the specific designs of individual programmes. Therefore, our exercise is novel since we test the capability of Clift’s theory’s main components to explain an individual programme. Testing Clift’s theory is also interesting to scholars studying the IMF’s participation in the eurozone crisis (Ban 2020; Hodson 2020; Johnston 2020; Moschella 2020).

Within scholarship on the IMF, explanatory variables driving Fund behaviour include powerful shareholders like the US or European governments (see, e.g., Bird 2007: 702–703), IMF staffers’ ideas and internal bureaucratic dynamics (Ban 2015; Broome 2015; Chwieroth 2010; Gabor 2015; Grabel 2019) and interaction with the eurozone monetary policy authorities (Hodson 2015; Lütz and Hilgers 2019; Moschella 2016). Clift (2018) focuses on bricoleurs to detail how the Fund mobilizes its internal ideational environment into programme design. Bricoleur staffers combine major economic ideas and gain purchase for their synthesis by knowledgeably navigating internal channels of influence, creating new ideational packages without departing dramatically from previous paradigms (ibid.).

Besides bricolage, the conceptual tool of “fiscal space” is, per Clift, key to understanding how the Fund analysed the eurozone crisis (ibid.). According to fiscal space, fiscal policy effectiveness is not uniform: states with well-developed fiscal buffers were advised by the IMF to use fiscal policy to stimulate demand to smoothen the impact of an economic downturn, while countries with contrasting public budget situations were pushed to cut public expenditure and implement structural reforms to restore confidence within debt markets and facilitate a private sector-led recovery (ibid.). The Fund used fiscal space to draw flexibly on multiple, contradictory beliefs regarding fiscal policy when navigating the austerity debate surrounding lending programmes and to tailor its advice contextually (ibid.).

Another key concept the IMF used to analyse the eurozone crisis was that of endogenous financial market instability through nonlinear multiple equilibria processes (NLMEP) (ibid.). Here NLMEP mean that expectations of either future policy effects on a sovereign’s debt burden or the future behaviour of financial market participants may cause rapid and significant changes in the current costs of sovereign borrowing without equally qualitative changes in currently observed economic conditions, or ‘fundamentals’ (ibid.). Further, Clift argues that accepting the NLMEP perspective meant that the Fund eagerly considered backstops of public monetary firewalls to curb self-fulfilling adverse sovereign yield spirals due to investor sentiment within the interconnected and thus vulnerable European financial markets (Clift 2018; De Grauwe and Ji 2013; Pisani-Ferry 2012).

Methods and data

Method: process-tracing

We hypothesise, based on Clift’s (2018) account of IMF behaviour, that the SBA’s design where IMF gave Greece exceptional access without requiring restructuring is best explained by focusing on IMF staffers viewing the Greek predicament via the conceptual tools of fiscal space and NLMEP, with these tools originating from bricolage. We chose a combination of “theory-testing process-tracing” and “explaining-outcome process-tracing” (Beach and Pedersen 2012: 14–21) to engage in a within-case study where we both test a theory and attempt to explain a certain case. Our design evaluates whether found observations, turned into “evidence” through contextual evaluation and interpretation (ibid.: 120), increase or decrease our confidence in the causal mechanism predicted by the theory being tested (ibid.). A mechanism needs to be “minimally sufficient”, fully explaining the outcome without redundant parts (ibid.: 14–21) while each of its parts must be a necessary condition for the mechanism (ibid.: 30–31).

To test our hypothesis, we formulate a causal mechanism with four steps:

  1. 1.

    The conceptual tools of fiscal space and NLMEP developed within the IMF via bricolage.

  2. 2.

    The IMF used the conceptual tools of fiscal space and NLMEP to analyse the economic conditions that defined the Greek programme and the policy options.

  3. 3.

    The IMF’s programme design process was driven mainly by fiscal space and NLMEP considerations.

  4. 4.

    Fiscal space and NLMEP considerations led to IMF pursuing an exceptionally large programme without restructuring (Fig. 1).

Fig. 1
figure 1

Causal mechanism

Full size image

Process-tracing tests utilise Bayesian inference to evaluate our confidence in the presence of each part of the causal mechanism (Beach and Pedersen 2012; Bennett 2015). First, we need a priori probabilities for the ‘prior’, ‘true positives’ and ‘false positives’ for each test–mechanism part combination. We convert our qualitative judgements into numeric probabilities (see Table 1) to unify these judgements and to facilitate replication. Second, we need a test, and third, evidence determining test passage (Beach and Pedersen 2012). Through testing, we obtain the posterior probability (our confidence in) for the presence of a part of our causal mechanism using the Bayesian Eq. (2015: 281–282).

Table 1 Probabilities assigned in the present study
Full size table

The core of our analysis, the process-tracing exercise systematises our judgements by contrasting them to a clear analytical benchmark for evaluating rigorously its explanatory power. However, we find unlikely that our original formulation of the mechanism captures all interesting aspects revealed in our extensive sources. Therefore, we will also review our data through a broader interpretative analysis.

We present the IMF as a unitary actor in our causal mechanism, but this requires qualifications. Chwieroth (2010) and Grabel (2019) showcase the differences between management planning and the bureaucratic processes, ideas and implementation work of staffers. Ban (2015), Blustein (2015) and Clift (2018) discuss the differences in ideas and policy preferences across IMF departments. When discussing the Fund, we mean the ‘project organisation’ consisting of the mission team mainly responsible for analysis and implementation of the programme, the IMF management determining the acceptability of programme options, and the involvement of IMF departments providing analytical and bureaucratic support for the mission and management (IMF 2013; Clift 2018). The key departments are Strategy, Policy and Review department responsible for due diligence regarding Fund’s internal rules (Blustein 2015; Clift 2018), and The Fiscal Affairs Department, Research Department, Monetary and Capital Markets Department, and European Department, which provided analytical support (Clift 2018).

However, in process-tracing we treat the IMF as a unitary actor since a multi-dimensional causal mechanism with different levels of IMF actors would be excessively heavy for our rigorous study of all available evidence. Further, we believe our approximation describes reasonably accurately the project organisation. The mission team exercises heavy influence on the programme design given its access to other Troika actors and has the capacity (or is forced) to perform creative solutions and policy compromises, while the programme’s political sensitivity kept the management involved in the design (Blustein 2015). Thus, in our model the IMF is primarily a tigh-knit team of professionals from the mission and the management. However, the Strategy, Policy and Review department exercises a veto-ish influence on management, and the programme’s heavy intellectual efforts and political consequences necessitated extensive analytical support (Blustein 2015; Clift 2018). Thus, the mission team’s and the managements understanding of the crisis and the available solutions were influenced by, and partially originated from, the supporting departments Clift (2018). Therefore, we expand on our process-tracing with qualitative interpretations discussing the importance of reconciling positions between various Fund actors.

Data

We use textual data as it is the most direct source available to us for examining the ideas and motivations of programme designers. Selected content discussed the programme explicitly.Footnote 1 Our primary sources include all publicly available IMF-created documents discussing the SBA design. The Interim Review and the First Review were examined because having been created within few a months of programme approval they could have traces of the original programme motivations. As secondary sources, we use all the IMF’s official evaluations of the SBA, as well as paraphrases and quotations of interviews conducted by other authors, and confidential programme documents intercepted by them.

“Observations” for process-tracing (Beach and Pedersen 2012: 120) were extracted from sources through exploratory thematic analysis (Guest et al 2012) using Taguette software (Rampin et al 2019). We identified all content relating to issues of bricolage, fiscal space and NLMEP and sorted it to “themes” (Guest et al 2012). Using themes, we systematised the second round of reading. We then contextually evaluated all resulting observations for their communicative purposes as well as for the agency and conventions underpinning their creation, gauging the reliability, or non-systematic measurement error, and source bias, or systematic error, in observations regarding their capacity to accurately depict the SBA (Beach and Pedersen 2012: 122–129). Both verdicts were made on the level of each source, separately for each mechanism part.

The testing process

Figure 2 illustrates the preparation of the tests, the required information and the judgements made at each stage of test preparation. Since we use textual data, we identified empirical occurrences which are necessary conditions for our causal mechanism to function and observable through textual data, leading to following two test categories:

  1. 1.

    Finding textual content where key IMF programme designers explicitly mention the combining of ideas (bricolage) to develop fiscal space and NLMEP (mechanism part 1), or fiscal space and NLMEP as being the basis of analysis for (part 2), or motivating (parts 3 and 4), the programme design.

  2. 2.

    Finding implicit communication of the processes/ideas above functioning per our mechanism.

Fig. 2
figure 2

Preparation of tests for the process-tracing exercise

Full size image

We gave each mechanism part a base rate for the probabilities of true and false positives regarding the existence of both explicit and implicit material. Further, the reliability/bias of our different sources affects our estimates of the probabilities for true and false positives for a given test. To accurately assign these probabilities, we test finding explicit and implicit references to our causal mechanism separately for each group of textual sources that, within group, have the same classifications of reliability and bias across all sources.

The resulting evidence was classified into 3 reliability categories: reliable, mostly reliable and limitedly reliable. For ‘reliable’ sources, we judged it highly likely that the document content matched the output of IMF staffers working with the SBA design, either due to being made by these staffers, being discovered in a confidential internal document or being found in a reliable interview. For reliable material, we made no correction to the probabilities of true/false positives. ‘Mostly reliable’ evidence came from sources where we could not ascertain the link of the observation to original IMF content production, but the context suggested that the observations were faithful to IMF considerations. To account for the increased measurement error of finding ‘true’ IMF considerations, we decreased the true positive probability for all mostly reliable evidence by 0.1. Finally, ‘limitedly reliable’ evidence originated from observations where the proximity to the original IMF design process was unclear (e.g. anonymous interviewees), or remote (e.g. interviewed non-Fund policymakers). For such evidence, the probability of true positives was decreased by 0.1 and the probability of false positives increased by 0.1.

The processed observations were also classified along their presumed bias towards the programme design, affecting the respective true and false positive probabilities (discussed in results). These biases ranged from observations being likely against the IMF or European policymakers, favouring senior IMF officials, or towards hiding key considerations about the programme design to guard the IMF’s reputation. These effects were assigned separately for each mechanism part as biases could affect the probability of finding evidence of certain behaviours or ideas, instead of the entire programme design.

Results

Results overview

Table 2 summarises our process-tracing results, which somewhat weakened our confidence in our causal mechanism. Test power was limited given the low reliability of the textual sources. Also, bias reduced test power for mechanism parts 3 and 4. That said, most explicit tests failed across mechanism parts, and only for parts 2 and 3 were most implicit tests passed. Assessing the explanatory power, we took a conservative approach by weighting lower-quality evidence (which reduces explanatory power through measurement error and bias) equally to higher-quality evidence. This prevents cherry-picking sources, but due to broad variation in reliability and bias, focusing on higher-quality evidence would likely have increased our design’s explanatory power.

Table 2 Key results
Full size table

Mechanism part 1: bricolage

Mechanism part 1 that IMF bricoleurs established fiscal space and NLMEP as key perspectives for IMF was given an average (0.5) prior probability. Clift (2018) documents IMF bricolage extensively and key potential bricoleurs were highlighted in multiple accounts of programme design (see, e.g., Blustein 2015), but other accounts besides Clift discuss bricolage only limitedly. We set the base probability of true positives for finding explicit confirmation as low (0.25), since the Funds’ key documents do not single out individual policy formulators, and external reviews focus more on substance than personnel input. Interviews may focus on key actors but more likely discuss policy negotiation than internal ideational work. Similarly, implicit true positive probability was set to average (0.5) with the addition that generic allusions to high-ranking officials as key drivers of ideational work could be made. The false positive probability for explicit testing was judged to be extremely low (0.05), as it seems extremely unlikely that personnel not responsible for a given synthesis of economic ideas would be falsely attributed. The false positive for finding implicit material was set to very low (0.1), as IMF documents and reviews unlikely discuss key bricoleurs without reason. Finally, we did not judge source bias to affect the quality of evidence regarding the role of key personnel as ideational originators within the Fund.

The testing results decreased our confidence in mechanism part 1 (average posterior 0.43 | minimum 0.36 | maximum 0.50), as all tests were failed for this part. We found little discussion of the ideational origins of key Fund conceptual tools and thus could not confirm bricoleurs facilitating the importance of either fiscal space or NLMEP. However, the update downwards was modest due to the test’s low true positive rates.

Mechanism part 2: fiscal space and NLMEP in analysis

For mechanism part 2, that fiscal space and NLMEP were at the core of SBA economic analysis, we gave an average prior (0.5). Clift (2018) amply showcases these analytical frames in use and multiple sources discuss both the importance of Greece's difficult fiscal position limiting its fiscal options from the Fund’s perspective, and the contagion perils faced by the eurozone (see, e.g., Blustein 2015; Wyplosz and Sgherri 2017). The true positive probability for finding explicit material to support part 2 was set at high (0.75). Fiscal space would likely fit well with previous Fund work (Clift 2018) and should thus be explicitly communicated. NLMEP being a newer analytical avenue (ibid.), the Fund’s staffers might hesitate to discuss it openly for credibility reasons. However, there would be limited reason not to discuss key analytical frames actually used in programme design. The true positive probability for finding implicit material was judged to be very high (0.9), as allusions to key analytical frames used would be difficult to avoid in interviews or even in 'sanitised' Fund documents. The probability of false positives of finding explicit support was judged to be low (0.25). While explicit false flags to a combination of fiscal space and NLMEP seem improbable, discussion of both fiscal policy in a supply-side focused form and some vague form of contagion could be pervasive. Thus, recognising fiscal space and NLMEP from more generic analysis could prove difficult. For similar reasons, the probability of implicit false positives was judged to be average (0.5), as disentangling implicit mentions to fiscal space and NLMEP might be even harder. The impact of source bias on evidence pertaining to Fund analysis, not recommendations, was negligible and did not lead to further probability adjustments.

Our confidence in mechanism part 2 was updated slightly downwards (average posterior 0.44). While most implicit tests were passed and some resulted in high posteriors for individual tests (max. 0.75), their confirmatory power was weak and most explicit tests were failed. Further, the failed tests had strong disproving power (min. 0.17). Much of the discussion across sources was consistent with fiscal space and NLMEP. The lack of any alternatives for fiscal consolidation, structural reforms, and privatisation, combined with the emphasis on the perilous Greek public debt position, supported the fiscal space narrative. Meanwhile, discussion on contagion risks, and different kinds of contagion models abound. However, limited explicit evidence for a coherent fiscal space and NLMEP analysis, such as discussing the merits of consolidation vis-à-vis expansion through the lens of fiscal space, or elaborating sentiment-based spillovers due to NLMEP, made it difficult to increase our confidence in the mechanism part.

Mechanism part 3: fiscal space and NLMEP in programme design

Our prior for mechanism part 3, that fiscal space and NLMEP generally motivated programme design, was set at average (0.5), with similar motivations as part 2 prior. As IMF needs to build a case for a programme, explicitly communicated motivation is essential. Therefore, we set the explicit true positive rate at high (0.75). Further, while finding explicit evidence of coherent ideational frameworks like fiscal space or NLMEP might be difficult, implicit mentions of central elements of such frameworks should be found if these enabled the IMF to ‘sell’ the programme. Hence, we set the implicit true positive rate at very high (0.9). Regarding the explicit false positive rate, discussion of both fiscal policy in a supply-side focused form and generic contagion would likely be pervasive regardless of fiscal space and NLMEP. Thus, disentangling fiscal space from general fiscal policy analysis and NLMEP from generic contagion discussion as programme motivators could be difficult, leading to a low (0.25) but noticeable probability. The pervasiveness of such generic discussion meant that the implicit false positive rate would also be average (0.5). Further, due to the Fund’s and other key sources’ potential bias towards hiding the messiness of programme design, we judged testing group material with such bias to have a decreased true positive rate (by -0.1) for mechanism part 3.

The probability for mechanism part 3 was updated slightly downwards (average posterior 0.42), although there was noticeable variation in the testing outcomes (min. 0.17 | max. 0.72). The reasons were similar as with mechanism part 2. Implicit tests were broadly passed: the lack of alternatives to consolidation and structural reforms (supporting fiscal space) and the broad discussion on sentiment contagion (supporting NLMEP) were present throughout material. However, there were few explicit and coherent arguments where these analytical frames were observed to shape the programme’s form. Therefore, all but 1 explicit test failed, dragging downwards the posterior.

Mechanism part 4: fiscal space, NLMEP and the programme’s final form

Regarding mechanism part 4 (fiscal space and NLMEP being the main driver behind the IMF decision to lend extraordinarily without restructuring), no account focuses on the combination of these lenses behind this decision. Even Clift (2018) does not utilise these to unpack the SBA loan amount and restructuring arrangement. However, contagion and limited space for fiscal manoeuvring as motivations for the programme were discussed across sources, setting the prior at low (0.25) probability. If the IMF had solid analytical underpinnings based on fiscal space and NLMEP for not restructuring and still providing considerable lending, these would feature explicitly in relation to the restructuring and loan amount decisions. This is because a controversial programme would warrant communication of all good reasons behind its design. Therefore, we judge the explicit probability of true positives for mechanism part 4 to be average (0.5), and implicit material should be found (high probability, 0.75) if fiscal space and NLMEP motivated the financing and restructuring decision. While controversial programme decisions could be artificially dressed up with analytical reasoning, the likelihood that precisely fiscal space and NLMEP were used for dressing-up seems unlikely, and thus, the false positive rate for explicit findings supporting mechanism part 4 was set at low (0.25). However, general false allusions to contagion and the Greek fiscal situation might be difficult to untangle from allusions to fiscal space and NLMEP, increasing our implicit false positive probability to average (0.5). Further, due to the Fund and other key sources potentially being biased towards hiding the messiness of programme design, or some anti-IMF biased actors blaming undue political influence, we judged the testing groups of material with such biases to have a decreased true positive rate (by -0.1) for mechanism part 4. The false positive rate was also increased by 0.1, as some pro-IMF biased actors could emphasise solid analytical foundations for programme design despite other motivators to maintain Fund reputation.

Confidence in part 4 was not updated (average posterior 0.24 | min. 0.14 | max. 0.35). Although most implicit tests and all but one explicit test were failed, the evidence was unreliable (and thus testing power low) for this mechanism part so that only a high passage rate could have increased the posterior above the low prior. It was difficult to make any kind of connection in our evidence between the frames of fiscal space and NLMEP and the designed size and restructuring terms of the programme.

Discussion

In our process-tracing exercise, we find support that the analytical lenses of fiscal space and NLMEP were used in IMF analysis during the programme (mechanism parts 2 and 3). However, our evidence does not suggest that these analytical lenses drove the SBA decisions to lend exceptionally while simultaneously requiring no restructuring (part 4). Likewise, our evidence does not support bricolage shaping programme design (part 1), although our sources lack key information for studying bricolage, such as individuals’ personal impressions or detailed programme-time discussions. Hence the motivations of the exceptional access without restructuring remain inconclusive based on our causal mechanism only. A hasty crisis programme like the SBA could be subject to so much ad hoc judgement that even a context-specific coherent mid-range theory cannot provide sufficient explanatory power for its key elements. In addition, as Clift’s (2018) study focuses on the broad ideational developments of the Fund visible in its key analytical policy work, it could be that such bricolage is not visible in SBA programme documents in the same form. Further, the effects of bricolage may not work uniformly or rapidly enough across all the IMF project organisation’s sub-actors to produce noticeable effects on the outcome. Finally, our focus on the first Greek lending programme as a critical juncture for the IMF and eurozone limits the time-horizon for bricolage effects, which Clift (ibid.) studied over the entire crisis.

However, while fiscal space and NLMEP do not directly explain the combination of exceptional access and non-restructuring, the fact that our evidence yields some support for IMF staffers utilising them in programme design warrants a broader interpretation of their significance within IMF behaviour during the SBA. This is since we study the SBA to understand if it provided a critical juncture in the ideational development of the IMF or in the policy development of the eurozone crisis. The overall impression from our thoroughly reviewed evidence is that the IMF worries about contagion were genuine, although perhaps not analytically coherent. Even if NLMEP was not the only crucial analytical driver, the likely usage of this lens among staffers further reinforces the importance of contagion. Contagion worries seem the key reason why the Fund considered it had to act, even if the relatively recently developed analysis could not provide a clear plan. At least these contagion worries made justifying non-involvement difficult given the institution’s crisis-fighting essence. Further, given the ECB’s credible veto on restructuring, any action could only be done by accepting no restructuring, at least during the first programme (even if it could be possible later through some back-door agreement with other eurozone authorities and private debtors). Thus, even if contagion worries by themselves could not cause the peculiar design of the Greek programme, they prompted the IMF to tackle a macro-financial crisis.

Further, the programme form of lending conditional on fiscal consolidation and structural reforms is compatible not only with the IMF’s historical legacy, but also with the fiscal space perspective. With Greece’s limited fiscal buffers, the fiscal space view gave IMF staffers little room to not use harsh austerity measures in programme design, even if this risked the stabilisation of public debt and thus may have worked against the goals identified by contagion analysis. The fiscal space framework was designed to accommodate the austerity legacy of the IMF to gain credibility within IMF fiscal policy circles and the broader austerity policy debate, even if this credibility was designed to expand the Fund’s toolbox towards more stimulus when that was considered beneficial. The path dependency from earlier Fund austerity thinking was included into Greece, leaving little options to fiscal consolidation. Therefore, the programme seems like a compromise between the main motivating factor for involvement, contagion and NLMEP, and the form conditions of programme design, which had to account for fiscal space.

The above considerations suggest that the IMF was ready to act ad hoc without a coherent ideational framework which could have unified the IMF sub-actors in defending a large programme without restructuring. Further, the IMF considered the constraint on restructuring as binding due to eurozone’s incapacity of decision-making, but still was motivated by the contagion. Thus, the fund combined the incomplete, yet powerful motivating reasons with form conditions of the programme forced from elsewhere. These form conditions were austerity (from within the IMF institutional legacy) and no restructuring (from the ECB). Thus, instead of formulaic, regime-changing new guidelines, the Fund used broad analytical principles to fit varied considerations within a single programme. This is supportive of the analytical spirit of Clift’s (2018) account of IMF behaviour. Further, given that our data presents limitations for studying bricolage, it remains interesting whether potential key bricoleurs such as Strauss-Kahn could have facilitated the Fund’s ideationally flexible operation. Such bricolage, where ideational creativity enables navigating through unprecedented crises, would still have different implications for our understanding of IMF compared to simply regarding the organisation willing to pragmatically take any route available. Thus, analysing IMF bricolage warrants further research.

The IMF’s unusual Troika environment could have modified the effects of NLMEP and fiscal space thinking on the programme. The ECB’s veto on restructuring is a fundamental setting for the programme, making the design less a process of the IMF creating a programme best fitting for Greece than the IMF fitting the best possible programme not including restructuring to the situation, with the other option being walking away entirely. In this environment the contagion fears (and some NLMEP thinking) is the reason why the IMF would not refuse acting, but the restructuring constraint is also a generative condition for the peculiar programme design. However, the NLMEP considerations internal to the IMF were essential for the exceptionally large size of the programme, as they reflect the crisis calling for large public funding to stave off rapid financial market flows due to collapses in confidence. In contrast, little evidence suggests that the exceptional programme size could have emanated from elsewhere in the Troika than the Fund itself.

Regarding fiscal space, the EC’s austerity tendencies could have accentuated the path-dependency in the Fund’s fiscal thinking. Greece had thin fiscal buffers and the EC had a stringent political interpretation of what fiscal prudence and competitiveness constituted by eurozone standards, and technical expertise to code this into the programme. Thus, when considering the IMF potentially sliding towards a default pattern of consolidation, the EC provided also another austere default option for the Fund to gravitate towards.

Operationalising the analytical lenses of fiscal space and NLMEP in further research could help us understand how much the IMF’s policy positions guided European governments in their economic policymaking during the eurozone crisis. For instance, Moschella (2020) questions whether the IMF had any policy-guiding effect on European governments. As fiscal space and NLMEP can be utilised to explain the IMF both recommending and not recommending austerity measures, these could provide nuance and operationalisable hypotheses for analysing the austerity policy debate (Clift 2020), for instance by studying whether European policymakers adopted these frameworks.

Additionally, a potential new explanatory method for understanding the SBA could be a multi- veto player model. Such dynamics between the IMF and other Troika partners were referred to in the evidence and highlighted by IEO auditors (IEO 2016: 34). Yet such an explanation has not been tested. Further, building the multi-veto model to include IMF sub-actors would better account for the effect of the complex set of policy preferences driving the Fund’s internal development, and the model could also consider the importance of European actors more broadly than our research.

Conclusions

We aim to explain the IMF’s (2010a, b, c, d) SBA with Greece, specifically the Fund’s decision to lend exceptionally without restructuring the Greek debt, by using process-tracing and Clift’s (2018) theory of IMF behaviour. Explaining this SBA is important, as the first Greek austerity programme was pivotal for the eurozone crisis, especially the battered Greek economy. Further, being the IMF’s entry into the crisis, it is crucial for studying the IMF’s behaviour in the era after the establishment of the eurozone. Meanwhile, there are important gaps in previous accounts of the programme.

We employed our interpretation of Clift’s (2018) account as a midrange theory for explaining the SBA case. In our causal mechanism, key IMF bricoleurs led to the development of fiscal space and NLMEP analytical lenses among the IMF programme designers. These analytical lenses would then be used to analyse the Greek programme’s economic situation, motivate programme' design and ultimately drive the call of the Fund to lend exceptionally without restructuring. To test our mechanism, we conducted a process-tracing exercise using textual evidence from key primary and secondary sources.

Our results slightly reduced our confidence in the presence of our mechanism. We found some supportive (albeit limited) evidence of fiscal space and NLMEP analytical frameworks being used by the Fund in programme design. While these frameworks do not solely explain an exceptionally large programme without requiring restructuring per our mechanism, contagion analysis seems a crucial motivator for IMF involvement. Meanwhile, the fiscal space considerations potentially explain why the IMF adhered to austerity-based form conditions, thus rendering the programme a compromise between motivation (contagion) and style (austerity). More broadly our results support the importance of examining partially coherent ideational frameworks as drivers to action, with the expectation that the outcome may be similarly incoherent. Finally, we find little evidence of bricolage affecting the key ideas behind programme design. This could be due to the difficulty of detecting bricolage in our textual evidence, prompting future research to test Clift’s bricolage theory with more appropriate evidence.

The potential importance of fiscal space and NLMEP to the IMF prompts further research into whether the Fund’s use of these ideas systematically altered the crisis responses and austerity policy debate of eurozone actors. Further, based on our evidence and on IEO’s (2016: 34) findings, a multi- veto player model seems a promising future avenue for explaining the SBA.

Finally, the study contributes methodologically to process-tracing, by developing a novel, systematic approach for generating the a priori probabilities required for testing. Our contribution is facilitated by all information required to replicate the study being provided freely on request from the author. We hope our workflow facilitates further rigorous process-tracing studies.