Governing the future: the European Central Bank’s expectation management during the Great Moderation

Abstract The experience of the global financial crisis has sparked renewed interest in the role of futurity in the capitalist economy in general and in the formation and coordination of expectations under uncertainty in particular. Economic sociologists have carefully studied the ‘defuturizing’ technologies devised by private actors, but have tended to neglect the increasingly pivotal part played by central banks. Political economists have had more to say on central banks, but have focused on institutional issues rather than on the concrete practices of central banking. Making an original contribution to both literatures, this paper traces the construction and subsequent evolution, up to 2007, of the European Central Bank’s communicative apparatus. Drawing on official documents and on interviews with both ECB staff and market participants, the paper shows how this apparatus created the conditions for the formation and coordination of private sector expectations. The insights from this empirical analysis into the performative dimensions of ‘credibility’ and ‘knowledge’ in monetary governance contribute directly to ongoing debates about the recent extension of the ECB’s communicative apparatus through forward guidance and quantitative easing.


Introduction
During the 1990s, monetary governance in the OECD world changed dramatically. After the failure of various monetarist experiments during the 1980s, the so-called Great Moderationa two-decade period of exceptionally stable inflation and GDP growth rates that lasted right up to 2007saw inflation-targeting emerge as the true successor to the post-war policy paradigm of 'Keynesian' demand management. Compared with the political earthquake that enabled monetarism, however, the revolution ushering in inflation targeting was a 'quiet' one (Blinder, 2004) that received little attention outside the discipline of macroeconomics. While the recent global financial crisis has further obscured the memory of this paradigm-forming moment, the expansion of 'forward guidance' policies ever further into the future has, at the same time, reinforced the overriding importance of private sector expectations. Even if the 'death of inflation targeting' has been pronounced many times, the recent crisis has, if anything, deepened the future-orientation of monetary governance. It is with a view towards the future of monetary policy, then, that this paper turns towards the expectation management experience of the European Central Bank (ECB) during the Great Moderation of the pre-crisis period (1999 to 2007).
By studying macroeconomic governance as the management of expectations, this paper answers recent calls by both economic sociologists (Beckert, 2013a(Beckert, , 2013bEsposito, 2013) and political economists (Nelson & Katzenstein, 2014;Nesvetailova, 2015;Palan, 2015;Soroka et al., 2015) to put the future-oriented nature of economic discourses and practices at the centre of the study of contemporary capitalism. At the heart of this literature is what could be called the 'futurity paradox'the inevitable tension between the future-oriented nature of economic action and the 'uncalculability' of the uncertain future (Keynes, 1937, p. 214;see also Commons, 1931;Knight, 1921). If economic activity is inherently forward-looking, how do economic actors avoid 'paralysis or randomness' in the face of an unknowable future (Beckert, 2013a, p. 226)? Looking for answers to this crucial question, economic sociologists have studied various social and socio-technical strategies to 'defuturize' (Luhmann, 1976, p. 141) the uncertain future, such as 'conviction narratives' (Chong & Tuckett, 2015), 'fictional expectations' (Beckert, 2013a), forecasting practices (Reichmann, 2013;Wansleben, 2013) and various risk-modelling technologies (Lockwood, 2014;MacKenzie, 2012). Yet where the political economy literature on central banks all too often abstracts from concrete governance and market practices, the economic sociology literature on expectations tends to describe a world in which the private sector is the monopoly producer of imagined futures. 1 This paper, by contrast, highlights the crucial part played by a bureaucratic authoritythe ECBin the socio-economic project of 'organizing to produce decidability and actionability' (Power, 2007, p. 5).
Detailed investigations of the communicative strategies of what Knafo (2013, p. 5) has called 'liberal financial governance' have mostly focused on the Federal Reserve (Abolafia, 2010;Hall, 2008;Krippner, 2007). Two recent studies have broadened the scope of this literature by investigating the communicative practices of the ECB (D.R. Velthuis, 2015). While Velthuis' primary concern is with the media, Holmes studies the communication between central banks and 'the public' in general. The current paper, by contrast, narrows the analytical focus to that part of the public that can be described as 'monetary insiders'. Comprising financial market participants as well as non-financial firms and social partner organizations, monetary insiders have both a greater stake in and a better understanding of monetary policy than the public at large. Adopting this narrower perspective brings into sharper relief the interactive nature of the 'signalling process' between the central bank and those economic actors whose activities propel the transmission mechanism of monetary policy, as opposed to the ECB just 'speaking to' an audience (Hall & Franzese, 1998, p. 505). Empirically, this emphasis is achieved by drawing not only on ECB documentation and on seven background interviews with ECB staff, but also on 18 interviews with German-based monetary insiders primarily employed in the financial sector. The Appendix contains an anonymized list of those interviews from which this text uses direct quotes. 2 This paper proceeds as follows. Elaborating on the theoretical framework that guides the analysis, the next section suggests treating 'governable' expectations as a performative effect of the interaction between the central bank and its insider audience. In this approach, the conditions for intersubjectively 'rational' expectations are created by a communicative apparatus that disentangles and frames private sector expectation formation. The third section gives a detailed account of the initial formulation of the monetary policy strategy of the Eurosystem. The evolution of the strategy during the pre-crisis period is then covered. Drawing on data from interviews with central bankers and private sector economists, the fifth section examines the communicative apparatus in action. Returning to a more conceptual level of discussion, the sixth section spells out two key implications for our understanding of the performative dimensions of credibility and knowledge in monetary governance. The final section concludes by emphasizing the relevance of these implications with regard to the most recent innovations in central bank practice.

Expectation formation: performative effect of a communicative apparatus
Without denying the relevance of affective future-orientations, such as emotions (Pixley, 2004), this paper accords analytical priority to the cognitive category of expectations, through which 'meaning is objectivized and institutionalized' in financial markets (Langenohl, 2010, p. 25). If capitalism constitutes a 'system of expectations', it can be governed only through expectations (Beckert, 2013b, p. 326). In the context of monetary policy, this challenge is augmented by the nature of the transmission mechanism of monetary policy, which links the short-term interbank rate controlled by the central bank to those interest rates at the long end of the yield curvei.e. long-term bond yieldsthat actually influence investment, and thus aggregate demand. These long-term interest rates depend on two types of expectations private lenders hold about the future. On the one hand, they 'reflect expectations of the future evolution of short-term interest rates'that is, of the central bank's future rate-setting decisions (ECB, 2004b, p. 44). On the other hand, long-term interest rates 'depend to a large extent on market expectations for long-term growth and inflation trends in the economy' (ECB, 2004b, p. 45). A lender who expects the inflation rate to rise will demand a higher nominal interest rate in order to achieve a given real return. The central bank's control over long-term interest rates thus depends on its ability to manage expectations of the future path of the official short-term rate as well as of future growth and inflation rates. This 'expectationalist' (Morris & Shin, 2008, p. 88) consensus has been firmly established among monetary theorists and policymakers at least since the early 2000s (Woodford, 2003, p. 15). From this perspective, a central bank's decision to change its current policy rate appears less important than its communicative guidance about the future path either of the policy rate or of the macroeconomic developments that will determine future rate decisions (Gürkaynak et al., 2005). When the ECB took over at the helm of monetary policy in the euro area, the question was therefore no longer whether expectations mattered, but how they were formed and how they could be manipulated.
Although expectations have long played a central role in macroeconomic theory, the question of their formation and coordination has been muted by the notion of 'rational expectations', understood as information-efficient predictions derived from the 'true' model of the economy (Lucas, 1972;Muth, 1961). Enormous methodological success notwithstanding, however, the influence of rational expectations on policymakers should not be exaggerated. While certainly shifting the focal point of macroeconomic governance into the future, the rational expectations view never served as a direct guide to central bankers, who have always been aware of the uncertainties surrounding forecasts of the future, the measurement of present data, and even the interpretation of past events. 3 The first consolidated academic statement of the Eurosystem's monetary policy strategy explicitly acknowledged the crucial importance of uncertainty, noting that in 'a hypothetical world in which the "true" economic model of the world is known' to both policymakers and market actors there would be no need for a communication strategy at all: 'The announcement of, and credible commitment to, the monetary policy's final goal would be sufficient to characterize the entire central bank strategy' (Issing et al., 2001, p. 99). In the real world, a communication strategy is needed precisely because uncertainty renders the probabilistic calculation of future outcomes, and thus rational expectations, impossible.
Yet, rather than reject the notion of rational expectations altogether, this paper embraces Geoffrey Ingham's (2004, p. 149) Weberian conception of central bank practice as geared towards establishing 'the highest level of formal rationality of inflation expectations'. Ingham's view of central bank communication as a performative project in the sense of Callon's (1998bCallon's ( , 2007 pioneering work, recently elaborated by D.R. , implies that central banks perform the very institutional and cognitive traits that make the systematic formation of 'intersubjectively shared expectations' (Hall, 2008, p. 188) possible in the first place. This involves what Callon calls the 'formatting' or 'performation of calculative agencies' (Callon, 1998b, pp. 23, 30) agencies that react to changes in the economic environment in a way that is consistent with the central bank's model of the economy. In other words, expectations are performative effects of carefully crafted arrangements that enable monetary insiders to form expectations 'despite the incalculability of outcomes' (Beckert, 2013b, p. 325, original emphasis).
The social studies of finance generally describes these arrangements as 'agencements', a term that Callon (2007, p. 320) borrowed from Deleuze and Guattari. However, this concept usually refers to market agencies rather than to governmental structures (e.g. Hardie & MacKenzie, 2007). Following Langley (2014), this paper therefore opts for the Foucauldian notion of 'apparatus' (dispositif). The two concepts and the associated theoretical perspectives are mutually compatibleif agencements enact 'the economy', apparatuses govern it. First, Foucault's understanding of 'apparatus' is consistent with the idea of performative governance in the sense that the processes that are being governed are not 'natural' but created by the apparatus itself (Foucault, 2007, pp. 65-66, 71-72). Second, like agencements, apparatuses of government include 'the operations and outputs of material and scientific technologies of knowledge' (Langley, 2014, p. 20). In the case at hand, these technologies comprise, above all, the models and methods of macroeconomics and econometrics (Braun, 2014;Fligstein et al., 2014;D.R. Holmes, 2014). Moreover, the communicative apparatus of the ECB also includes elements that are performative in Goffman's (1959) sense of the term acts performed on the front stage for the benefit of specific audiences that may or may not be congruent with what goes on backstage. An important methodological implication of this is that, to the extent that monetary policymakers 'put on an act', students of central bank communication cannot always take at face value the 'front' of official central bank discourse, that is, 'the expressive equipment … intentionally or unwittingly employed by the individual during his performance' (Goffman, 1959, p. 13). Thus, in the same way in which the political significance of central bank independence becomes palpable when re-conceptualized as a 'rational fiction' (McNamara, 2002), the study of central bank communication calls for a certain conceptual distance. It is to establish and maintain that distance that this paper opts for the notion of a 'communicative apparatus', which includes, but is not identical to, the ECB's official 'communication strategy '. 4 The necessity of such a critical analytical stance is most clearly illustrated by the mantra of post-1990 central bank communication. If we were to consult the monetary economics literature or the ECB's public statements, we would inevitably come away with the impression that its communicative apparatus begins and ends with 'transparency'. A closer look at the underlying definition of transparency, however, casts doubt on that narrative. In economics, 'central bank transparency' is commonly defined as 'the absence of asymmetric information between monetary policy makers and other economic agents' (Geraats, 2002, p. 533). Similarly, the ECB defines transparency as 'an environment in which the central bank provides in an open, clear and timely manner all relevant information on its mandate, strategy, assessments and policy decisions as well as its procedures to the general public and the markets' (ECB, 2002, p. 59). Yet, such standard definitions of transparency are fundamentally at odds with the uncertainty of the information exchanged 'between monetary policy makers and other economic agents'. Under conditions of uncertainty, avoiding asymmetric information to achieve 'transparency' is not a straightforward task at all (Dow, 2013, p. 183). As one senior ECB economist put it, 'a central bank cannot be too transparentsimply because you don't know what will happen tomorrow' (Interview 2). Ultimately, transparency cannot simply be treated as a benchmark against which actual central bank practice can be measured. Uncertainty imposes limits on transparency (Best, 2005, p. 31), so that 'full transparency' is not an option available to central banks.
Aware of this problem, monetary policymakers have long taken a proactive approach towards the guidance of monetary insiders' perception of the macroeconomic outlook, thus fulfilling what has been described as the 'epistemic' (Hall, 2008, p. 193) or 'sensegiving' (Abolafia, 2010, p. 349) function of central banks. In order to shed light on how precisely this is achieved in the case of the ECB, Callon's twin concepts of disentanglement and framing are particularly helpful (Callon, 1998a, p. 249). It is through the process of disentangling a small number of variables from the otherwise infinite amount of economic information that the central bank is able to delimitate, or frame, 'an expectations environment that can be managed' (Morgan, 2013, p. 744;cf. Dow, 2013, pp. 182-183). In this framing process, monetary economics plays a performative role, as 'consistency [in central bank communication] can be accomplished only in relation to a consensus on the meaning of the monetary situation as described by economic theory and econometric models' (Ingham, 2004, p. 146). Only within such a limited (and limiting) frame can the central bank be 'transparent' and 'credible'. The empirical sections below are therefore concerned not so much with the specific content of 'conventions' (Nelson & Katzenstein, 2014), 'fictions' (Beckert, 2013a), or 'conviction narratives' (Chong & Tuckett, 2015), but with the communicative apparatus that disentangles and frames a space within which the formation and coordination of monetary expectations becomes possible in the first place.

The monetary policy strategy of the ECB
When the European Monetary Institute (EMI) 5 began its work in 1994, the ECB's future price stability mandate was already enshrined in the Maastricht Treaty (Art. 105). The Treaty left open, however, how the end (the inflation target) should be achieved through the available means (the instruments of monetary policy). The 'traditional' approach had been for central banks to aim for an intermediate target in pursuit of the ultimate but impalpable goal of price stability. Having already ruled out exchange rate targeting, interest rate pegging and nominal income targeting, the EMI left it to the Governing Council of the ECB to choose between the remaining options of monetary targeting and direct inflation targeting. The latter is different in that it does away with intermediate targets altogether. Instead, inflation-targeting central banks announce their target rate of inflationusing inflation forecasts rather than the current rateand then deploy monetary policy instruments so as to achieve that target (Svensson, 1997).
While monetary targeting had been the Bundesbank approach since 1975 (von Hagen, 1999), several OECD countries had successfully switched to inflation targeting between 1990(Freedman & Laxton, 2009). But inflation targeting had a powerful opponent in the ECB's first chief economist, Otmar Issing, who was entrusted with the task of formulating the Bank's communication strategy. He warned against the dangers of adopting a forecast-based strategy at a time when the available data were patchy and when the validity of existing models stood in question, arguing that '[f]orecast uncertainty is likely to be relatively large, possibly rendering the whole inflation targeting strategy ineffective' (Issing, 1998, p. 5). At the same time, however, the European Monetary Institute (1997, p. 9) cited a different type of uncertainty as an argument against the Bundesbank approach: 'Uncertainty concerning the empirical properties of money demand in the euro area in Stage Three is the main argument against a monetary targeting strategy'an argument that Issing says convinced him that a pure monetary targeting approach was not an option (Interview 1).
In the end, the Council opted for a monetary policy strategy whose numerous intentional ambiguities reflected the uncertainties of the transition. The centrepiece of the strategy was the price stability goal, now specified as 'a year-on-year increase in the Harmonised Index of Consumer Prices (HICP) for the euro area of below 2%', which was to be maintained 'over the medium term' (ECB, 1998b). In pursuit of this goal, monetary policy would be guided by both a monetary analysis and an economic analysis. The latter was the uncontroversial 'second pillar' of the strategythe analysis of 'a wide range of economic and financial variables as indicators for future price developments', such as wages, exchange rates, measures of real economic activity and fiscal policy indicators (ECB, 1998b). On the other hand, M3 was chosen as the monetary aggregate that would guide the analysis under the controversial 'first pillar'. A broad monetary aggregate, M3 includes (among others) currency in circulation, repos, short term deposits, debt securities (with a maximum maturity of two years) and shares of money market funds (ECB, 1998a). The ECB emphasized that its 'reference value' for the annual growth rate of M3 would not constitute a target for policymakers (ECB, 1998b), but a 'benchmark for assessing risks to price stability' (ECB, 2000b, p. 41), deviations from which would be treated as one factor among others in the considerations of the Governing Council. In technical terms, the reference value was defined as the sum of the envisaged inflation rate of 'below two per cent', a GDP growth trend of 2-2.5 per cent, and a velocity of circulation declining at a trend rate of 0.5-1 per cent per annum. This formula yielded a reference value for the annual growth rate of M3 of 4.5 per cent (ECB, 1998a).
Elaborating the communicative apparatus: the early years of EMU During the early phase of Stage 3 of Economic and Monetary Union (EMU), the ECB's approach to communication was met with considerable scepticism. Criticsincluding most leading monetary economistsaimed at three aspects in particular, all of which came down to the question of 'transparency'. First, regarding the economic analysis (the second pillar), critics singled out the secrecy that surrounded the internal staff forecasts of macroeconomic developments, which initially the ECB did not publish (Svensson, 1999, p. 80). The second criticism was directed against the vague and asymmetrical definition of price stability -HICP inflation of 'below 2%'and of the time horizon -'the medium term'over which the goal was to be achieved (Galí, 2002, pp. 42-46). Finally, critics bemoaned the very existence of the monetary analysis (the first pillar) on both theoretical and empirical grounds. As we shall see, all subsequent modifications to the communicative apparatus of the euro area were related to one or more of these criticisms.
Strategic ambiguity: the definition of price stability and the role of the monetary pillar Four years into Stage 3, a comprehensive evaluation of its monetary policy strategy led the Governing Council 'to clarify to the public some aspects' (ECB, 2003, p. 79). First, the ECB added three words to its price stability goal, announcing that it would henceforth aim for a year-on-year increase of the HICP of 'below, but close to, 2%'. This slight modification reduced the asymmetry of the previous definition, which would have ruled out a 4 per cent annual price increase but would have allowed for zero inflation. The second part of the evaluation concerned the role of the controversial monetary analysis in the strategy of the Eurosystem. As indicated above, the two-pillar strategy was best understood as the introduction of an element of 'constructive ambiguity' (Best, 2005) that reflected policymakers' preference for flexibility in the face of significant uncertainty regarding the workings of the euro area economy and the transmission mechanism of monetary policy (Issing et al., 2001, p. 108). The standard phrase at the time evoked a 'prominent role' for money (ECB, 1998b). Asked for clarification in the press conference during which he announced the strategy, President Duisenberg (1998) replied ominously that 'one pillar is thicker than the other is [sic], or stronger than the other, but how much I couldn't tell you'.
The precise relative measurements of the pillars notwithstanding, monetary economists soon questioned both the theoretical and the empirical justification of the money-growth indicator to which the ECB was so eager to ascribe a 'prominent role', as 'essentially irrelevant' (Svensson, 1999, p. 80;Galí, 2002). 6 Again, the ECB met its critics halfway. It retained the first pillar, yet relegated itsomewhat confusinglyto a subordinate position behind the 'second pillar', 'as a means of cross-checking, from a medium to long-term perspective, the short to medium-term indications coming from economic analysis' (ECB, 2003, p. 87). In the President's introductory statements at the monthly press conference, the economic analysis would henceforth precede the monetary analysis. The Governing Council also announced that in order to underscore the long-term nature of the monetary analysis, the reference value for money growth would no longer be reviewed on an annual basis (ECB, 2003, p. 87).
When asked about his experience of the ECB's first attempts at communication with monetary insiders, Issing recalled the German Council of Economic Advisers alleging 'that we had a combination of an inflation target and a money target, which was complete nonsense' (Interview 1). The changes to the ECB's communicative apparatus described in this section made the reference value less likely to be mistaken for an intermediate target. In fact, the reduced emphasis on the first pillar and the specification of the inflation target moved monetary policy in the euro area closer towards formal inflation targeting. This tendency was reinforced by the ECB's gradual move towards greater openness with regard to its macroeconomic forecasts.

The question of forecast transparency
The first increase in forecast transparency occurred in December 2000, when the Eurosystem began to publish the results from the bi-annual 'staff macroeconomic projection exercise', which offered conditional forecasts of euro area GDP growth, inflation rates and the major expenditure items both for the current and the following year (ECB, 2000a). Such projections not only provide a key input for the rate decisions of the central bank, but also serve as a communicative tool explaining these decisions to monetary insiders (Goodhart, 2009, p. 86). The ECB's initial non-publication of these projections therefore fell short of what, by that time, had become best practice among central banks, and was criticized because the projections were thought to be 'the decisive input in policy decisions' that should be available to the public (Svensson, 1999, p. 80). This view seemed to be vindicated when Issing declared that the decision to publish the projections reflected 'the Governing Council's firm conviction that the information and analysis underlying its monetary policy decisions should be shared with the public to the greatest extent possible' (Issing, 2004, p. 725). 7 Yet, the ECB's decision in 2000 to publish the projections did not satisfy the critics. This was because, firstly, a bi-annual projection was too infrequent 'to be useful input in the monetary policy decision-making process', which is why Galí (2002, p. 60) expected that '[a]lmost surely, the Governing Council must have access to more frequent internal projections'. This was addressed in 2004, when the ECB switched from a bi-annual to a quarterly publication schedule for its macroeconomic projections. 8 Secondly, the uncertainty surrounding the projections was represented through 'ranges' for each variable. The ECB fell short of the practices of other central banks by not assigning probabilities (through confidence intervals) to its ranges, which significantly reduced their informational content (Galí, 2002, p. 60). This criticism was partly addressed in 2008, when the ECB began to publish ranges that implied a 75 per cent confidence interval (ECB, 2008).
The most interesting and important question, however, concerned the nature of the interest rate assumption underpinning the ECB's projections. Central banks face a choice between three alternative scenarios when predicting the future path of their own policy rate. Under the constant interest rate assumption, the current rate remains unchanged over the forecast horizon. This had been the approach of the ECB and of the Swedish Riksbank until 2006. The second option is to assume a future interest rate path that is 'consistent with current market expectations'that is, to use the forward rates that are implicit in the yield curve (Galí, 2011, p. 537). The ECB switched to this approach in 2006, thereby joining the Bank of Japan and the Bank of England. 9 Both the first and the second alternatives yield conditional forecasts (hence the term 'projections'). Third, unconditional forecasts are based on the assumption that the policy rate 'will follow whatever path the central bank expects it to follow' (Galí, 2011, pp. 537-538). This approach is currently followed by the central banks of the United States, Sweden and Norway.
The advocates of central bank transparency were critical of the ECB's constant interest rate assumption because the resulting conditional forecasts were based on an interest rate that did not correspond to the likely path of the policy rate as anticipated by the central bank at the time of the projection exercise (Leitemo, 2003;Woodford, 2005). While such forecasts will 'not actually [be] believed, even in the central bank itself', they avoid precommitment to a particular policy (Woodford, 2005, p. 44). The ECB was fairly explicit about this, emphasizing that its projections were 'not meant to be, and should certainly not be misperceived to represent, predictions of the most likely macroeconomic outcome', but as 'a counterfactual scenario' for policymakers to take into account (Issing, 2004, pp. 725-726). Nevertheless, the ECB met its critics halfway when, in 2006, it began to publish projections based on market expectations of the future path of the policy rate. While this introduces an element of realism regarding the interest rate assumption, it still left it to the ECB's insider audience to divine the Bank's stance in the light of a forecast of variables other than the policy rate (Rudebusch & Williams, 2008, p. 248)a forecasting exercise described as 'incestuous' by Charles Goodhart (2009, pp. 94-95) in a passage on the Monetary Policy Committee of the Bank of England: 'The market is trying to guess what the authorities will do, and their guess is then incorporated as the conditioning assumption to the initial forecast on which, in part, the MPC bases its decision'.

The communicative apparatus in action
So far, the analysis has been limited to the initial design and subsequent evolution of the ECB's official communication strategy. It is only when the audience of the central bank is taken into account that the performative dimension of the communicative apparatus becomes analytically accessible. Drawing on interviews with monetary insiders, this section shows that there is indeed more to ECB communication than meets the eye.

Pavlov's dog and the inflation Taliban
From the beginning, the ECB adopted a 'hawkish' stance with regard to inflation. While this is usually attributed to its narrow price stability mandate, my interview data suggest that the performance of 'hawkishness' also plays a crucial role in the communicative apparatus of expectation management. One senior ECB economist described Jean-Claude Trichet's approach to communication as follows: He insisted so much on our mantra … the close to two but below two percent. He repeated this concept again and again … He was accused by those who were less focused on inflation that he was like a Taliban, an inflation Taliban. But the main purpose of his insistence was really to educate, or to make really very clear … the whole point of the monetary policy of the ECB. (Interview 3) According to Benjamin Friedman (2002, p. 16), central bankers' (over-) emphasis on inflation is indeed pedagogically motivated in that it aims at conditioning the audience to disregard other variables when forming expectations about the future path of the key interest rates. Market actors, of course, are acutely aware of this. A former bank chief economist offered the following interpretation: When you have the inflation rate as a target, you can go about all this in a nicely mechanical way…. You have fixed relationships everywhere, inflation and output gap, short-term interest rates … So you're training the dog to react mechanically. It understands exactly how you set interest rates, so you're a pure mechanic. (Interview 9) Arguably, the preference of both the ECB and its insider audience for such quasi-mechanical predictability is rooted in a sense of mutual dependence and vulnerability. On the one hand, a bank that is wrong-footed by an interest rate decision may suffer significant losses on both its trading book and its loan portfolio. On the other hand, the operation of the transmission mechanism of monetary policy depends on the 'cooperative' behaviour of financial markets. A pension fund manager put this succinctly (Interview 6): 'Central banks want to manipulate [the interest rate]. This manipulation works only if the actorsthat is, we, and institutions like uslet ourselves be manipulated'. It is this mutual dependence that underpins what this interview partner called the 'unwritten agreement between central banks and capital markets that central banks tell markets in advance what they are going to do'. 10 Indeed, the middle of the first decade of monetary union was characterized by a remarkable unanimity between the ECB and its insider audience. According to a money market trader, Jean-Claude Trichet's statements and speeches were scrutinized by ECB watchers for 'signal words' that were regarded as giving clear indications of the direction of interest rate decisions (Interview 5). This view from the market side corresponds with the experience of a senior ECB economist, according to whom from 2003 onwards, 'the situation became almost boring': 'There was nothing really to discuss and to do … The impression from here was that [the banks] understood more or less fully the way it worked' (Interview 3). Between June 2003 and December 2005, the ECB's main refinancing rate stood unchanged at 2 per cent for the unprecedented period of 30 months.
When asked about interest rate decisions that had caught them by surprise, several interview partners recalled two episodes that fell into the post-2007 period. The first such episode was July 2008, when it was perfectly visible and transparent that the economy was in a massive downturn, yet the ECB raised the interest rate to, I think, 4.25 per cent, even though everybodyeverybodyin the market knew that it was the wrong decision. (Interview 6) The second episode was an oil-price fuelled surge in the inflation rate to 3 per cent in mid-2011, to which the ECB responded by raising the main refinancing rate from 1 per cent to 1.5 per cent in two successive steps. Citing stable core inflation and the fact that the euro area had only just begun to recover from the 2009-2010 recession, a bank economist said that the ECB 'had done something that came as a complete surprise to me, in the sense that their [the ECB's] fear of inflation was much stronger than mine' (Interview 4). Tellingly, both episodes took place after 2007, when the ECB's 'inflation-only mantra' was gradually weakened by expectations that the central bank would increasingly have to give precedence to financial stability and/or economic growth concerns. Although they do not fall into the time period studied here, these premature rate hikesboth were quickly reversedare relevant to the present argument in that they constituted (ill-fated) attempts to re-establish the precrisis framing of expectation management.

The curious case of ultra-stable inflation expectations
The analysis of the ECB's communication of its macroeconomic projections pointed towards the 'incestuousness' (Goodhart, 2009, pp. 94-95) of the informational flow between the central bank and monetary insiders. The example of the Survey of Professional Forecasters (SPF) provides a particularly vivid illustration of this circularity. 11 Eliciting an average of 59 responses from forecasters primarily at financial or research institutions, the SPF questionnaire collects forecasts of HICP inflation, real GDP growth and unemployment for different time horizons (ECB, 2007, p. 10;Garcia, 2003, pp. 8-9). Most importantly, respondents give point estimates of euro area inflation rates for several future dates, including an estimate for the rate five years ahead. Given this long time horizon, one would expect the survey data to show considerable variation. However, this has never been the case. At any point between 2003 and 2006 the inflation expectations of at least 50 per cent of SPF forecasters fell into the very narrow range of 1.9-2 per cent, neatly coinciding with the ECB's inflation target of 'below but close to two per cent'. Asked about their views of this rather puzzling stability of long-term inflation expectations, two ECB watchers at different banks gave very similar responses: Forecasts that reach that far into the future are subject to very large uncertainty. As a consequence, there is a tendency to move towards the target rate, or, simply speaking, 'I don't know, I'll just put down a number slightly beneath 2 per cent'. (Interview 8) It is of course slightly tautological -'as long as we trust you, we are going to write down a number close to 2 per cent'. (Interview 7) These statements suggest that the primary function of the SPF is not to provide information about private sector inflation expectations to the ECB, but to the survey respondents themselves. 12 From this perspective, the SPF appears not to function as an analytical tool but a vital part of the communicative apparatus of expectation management, which helps anchor the very expectations it ostensibly measures. The stability of long-term inflation expectations as measured by the SPF has a performative effect on the way in which monetary insiders form their macroeconomic expectations. Not only do the forecasters surveyed for the SPF 'fail to systematically update their forecasts', as one recent study has found (Andrade & Le Bihan, 2013, p. 967), they also tend to shift their attention from future inflation rates to future interest rates. One ECB watcher explained this as follows: When today inflation rates fluctuate, sometimes even by a full percentage pointfor instance because of oil price shockspeople do not get very excited about that anymore … because they believe relatively firmly in the inflation targets. One then rather adapts expectations in other areas. No one says: 'Because of these inflation shocks I need to raise my inflation expectations for the next ten years' … Instead one draws the conclusion that monetary policy is going to be more restrictive. (Interview 4) The pension fund manager quoted above expressed a similar view: Twenty years ago the central bank would raise interest rates when inflation accelerated, and would lower them when inflation came down. So the game was about how the economy and inflation would go, that was the economic analysis. Today the question is: What does the central bank want? (Interview 6) These quotes point towards an important feature of monetary expectation management: forecasts of macroeconomic variables are valuable to monetary insiders not so much in their own right, but indirectly, as predictors of future interest rate decisions by the central bank. This is further illustrated by the role in the ECB's communicative apparatus of the concept of the 'output gap'.

Does the ECB know the value of the output gap? Does it matter if it doesn't?
The output gap is defined as the difference between actual and potential GDP. It is one of two key variables in the standard formulation of the Taylor rule, which describes the (putative) reaction function of the central bank as a quadratic loss function that contains the deviation of the inflation rate from the central bank's target rate, as well as the deviation of current output from its 'natural' level. Formulated over two decades ago by John Taylor (1993), the Taylor rule soon emerged as the single most important heuristic device for central bank watchers across the world (Koenig et al., 2012). The primary reason for this popularity was the robustness of the rule as a predictor of central banks' rate setting behaviour. When asked about his experience of the transition in 1999 from the Bundesbank to the ECB, a veteran ECB watcher at a German bank responded by pointing to the continuing applicability of the Taylor rule: For my actual work of forecasting, the differences were not particularly large in my view. I had a kind of Taylor rule for the Bundesbank, and I subsequentlywhen I had some dataput one together for the ECB, and it has done the job ever since. (Interview 8) The most important point, however, is that the considerable theoretical and empirical uncertainties surrounding the output gap (Orphanides, 2001) have not prevented it from featuring prominently in the Governing Council's explanations of its decisions and thus assuming a key role in the communicative apparatus of the euro area. Another ECB watcher explained his focus on the output gap as follows: We also look at output gaps. After all, in normal times the Taylor rule worked fairly well as a proxy. Of course, in retrospect we know that these output gaps were anything but exact science. But ok, so whatif the ECB lets itself be guided by them, then that's how it is. (Interview 7) Both this statement and the example of the SPF support the argument that central bank information plays a dual role 'of conveying fundamental information as well as serving as a focal point for beliefs' (Morris & Shin, 2002, p. 1522.

The performative dimensions of credibility and knowledge in monetary governance
In order to explore the broader implications of the empirical analysis, this final section returns to a more conceptual level of discussion. Taking Ingham's (2004) and D.R. Holmes' (2014) performative understanding of monetary governance one step further, it argues that in order to get to the performative dimensions of 'credibility' and 'knowledge' it is crucial to 'read between the lines'that is, to account for those elements of the ECB's communicative apparatus that transcend its official communication strategy.

Performing credibility: 'inflation nutters' and pretence of single-mindedness
The ultimate ECB target during the pre-crisis period was (and still is today) the inflation rate. But given that, in the short run, inflation is governed by factors over which they have no direct control, why do central banks choose inflation as their target at all? Monetary economist Benjamin Friedman argues that, above all, an exclusive focus on inflation conditions the way in which market actors form expectations by 'remov[ing] from explicit discussion whatever objectives the central bank may hold for output, employment, or other real outcomes, over less than the long run' (Friedman, 2002, p. 16). Yet if in reality, as noted by Mervyn King (1997), central bankers are not the 'inflation nutters' they pretend to be but care just as much about real outcomes such as output and employment, then this form of communication is not particularly transparent. Paradoxically, then, 'pretending' that only inflation matters is a crucial precondition for what is generally described as the credibility of the central bank. Only 'by keeping out of the discussion those considerations that would reveal that commitment to be qualified' is the central bank able to commit credibly to keeping inflation low (Friedman, 2002, p. 16).
During the Great Moderation, central banks became masters in the art of pretending that inflation was all they cared about. They thereby framed the communicative space in a way that allowed them to appear more consistent and predictable than would have been possible had equal consideration been given to the multitude of disparate or even contradictory economic trends that can be observed at any given time. The welfare-enhancing effects that accrue, in Rogoff's (1985Rogoff's ( , p. 1170) influential model, from 'appointing as head of the central bank an agent whose dislike for inflation relative to unemployment is known to be stronger than average', are the result of this individual's ability to perform credibly as an 'inflation nutter'. Somewhat counterintuitively, then, central bank credibility during the Great Moderation was based on a performance that was not 'sincere' (Goffman, 1959, p. 10).
That said, it is important to note that the radical expansion of the ECB's explicit and implicit responsibilities in the wake of the recent crisis has abruptly upended this 'pretence of single-mindedness' (Braun, 2015;Krampf, 2014). Although the same is true for most central banks, Mario Draghi epitomizes this rupture. Confronted with a proliferation of the ECB's tasks, Draghi made no attempt at re-enacting the 'inflation nutter' performance of his predecessors.
Pretence of knowledge: pathological anomaly or constitutive element of expectation management?
Friedrich Hayek used to warn tirelessly against the dangers of an exaggerated belief in the governability of the economy on the basis of theoretical knowledge: To act on the belief that we possess the knowledge and the power which enable us to shape the process of society entirely to our liking, knowledge which in fact we do not possess, is likely to make us do much harm. (Hayek, 1989, p. 7) The constant reiteration of variations of Hayek's pretence-of-knowledge argument notwithstanding, 13 however, the Great Moderation inspired a new bout of overconfidence in the ability of economists and policymakers to control the business cycle. Alan Greenspan was hailed as the 'maestro' conducting the orchestra that was the US economy (Woodward, 2000), while macroeconomists thought they had finally succeeded in the sense that their discipline's 'central problem of depression prevention has been solved, for all practical purposes' (Lucas, 2003, p. 1). Most observers tend to describe this phenomenon of technocratic overconfidence in terms of delusion or aberration a 'Great Complacence' (Engelen et al., 2011) or a 'pretense-of-knowledge syndrome' (Caballero, 2010). The current paper, by contrast, suggests a slightly different interpretation. 14 The above discussions of the output gap and of the Survey of Professional Forecasters have shown that the data provided by a 'transparent' central bank do not merely convey information about economic fundamentals but also serve as focal points around which private sector actors build and coordinate their expectations (Morris & Shin, 2002, p. 1522. Market actors are keenly aware of this dynamic. Explicitly mentioning Keynes' example of the beauty contest, a former bank chief economist put it thus: You have to believe that others believe that the central bank knows more than they do. Even if you believe that the central bank does not know more than the market, even that it knows significantly less than the market, then you still follow the central bank if you believe that the others believe that the central bank knows more than the market. (Interview 9) Jens Beckert recently argued that under conditions of uncertainty, expectations are necessarily 'pretended representations of a future state of affairs'actors are able to commit to specific actions only on the basis of the 'pretention that the fictional depictions were indeed true representations of the future' (Beckert, 2013a, p. 226). The present paper provides empirical evidence supporting this argument: the coordination of expectations around central bank forecasts is conditional on the willingness of monetary insiders to act as if they believed that the central bank knows more than it doesor can possiblyknow. Certainly, arguing that 'pretence of knowledge' has been a productive element of expectation management during the Great Moderation is not to deny its contribution to the 'Great Complacence' and the quite devastating consequences. The statements by ECB staff and market participants quoted above give a clear sense of the illusion of control inspired by the all too successful framing of expectation formation around consumer price inflation, which implied a benign view of asset price inflation and of the recently discovered 'risk-taking channel' of monetary policy transmission (Borio & Zhu, 2012). The evidence does, however, support the counterintuitive argument that 'pretence of knowledge', which has generally been dismissed as a pathological development in central banking, has actually beenand arguably still isan integral part of the communicative apparatus of expectation management, in the euro area and elsewhere.

Conclusion
The question of how economic actors cope with the problem of a fundamentally uncertain future has recently emerged as a central concern in both political economy and economic sociology. Starting out from the observation that the literature has tended to focus on the defuturizing projects of the private sector, this paper emphasized the key role played in that context by a particular type of public actorcentral banks. Building on the work of Ingham and Callon, it conceptualized monetary insiders' expectations as the performative effect of a communicative apparatus that disentangles and frames expectation formation. On that theoretical basis, the paper showed how the ECB's initially contested communication strategy went through a series of modifications geared towards establishing a frame that would allow monetary insiders to form expectations in a predictable and thus 'governable' manner. The decisive analytical step, however, consisted of moving on from the description of the ECB's official communication strategy to the analysis of its communicative apparatus, which was achieved by explicitly accounting for the performative dimension of expectation management. Interviews with ECB staff as well as with members of the ECB's insider audience revealed three key performative elements of the communicative apparatusthe performance of hawkishness, the performative function of surveys in anchoring the very expectations they ostensibly measured, and the performative use of the empirically dubious concept of the output gap. The final section explored the implications of these empirical insights for the understanding of the performative dimension of central bank 'credibility' and 'knowledge' during the Great Moderation. It argued that credibility was, to a greater extent than commonly acknowledged, based on policymakers successfully pretending to be single-mindedly focused on inflation. As for the role of knowledge, my analysis suggests that the 'pretence of knowledge syndrome', widely bemoaned in the wake of the global financial crisis, cannot simply be discarded as a pathological aberration. Instead, it constitutes an essential performative element of the communicative apparatus of expectation management.
Given that the period under investigation in this paper is separated from the present by several years of innovation and transformation in central bank practice, what does this analysis contribute to our understanding of the brave new world of 'central bank-led capitalism' (Bowman et al., 2013)? The defining features of this world have been the twin-policies of forward guidance and quantitative easing, which aim at extending central bank control from shortterm money market rates to long-term interest rates, thus turning the latter into a policy variable (Turner, 2014). Central banks, including the ECB, have attempted to achieve this by reducing uncertainty regarding the interest rate levels that will prevail not just a month, but one or even two years from the present. My account of the gradual tendency of the ECB's communicative apparatus towards greater forecast transparency points towards an evolutionary interpretation of forward guidance as the latest step in a long-standing development. Nevertheless, if the recent financial upheaval is the crisis of a 'future that has become overcrowded' (Nesvetailova, 2015, p. 450) the question arises if the governance of the economy through expectations has reached a limit (Gabor & Jessop, 2015). Otmar Issing, a vocal critic of the ECB's forward guidance, has warned that financial market and media actors' demand for information on monetary policy was 'insatiable' (Issing, 2014, p. 10). Paradoxically, by accommodating this demand for the defuturization of ever more distant futures, the central bank undermines its own claims to knowledge and credibilitywithout which forward guidance cannot workas monetary insiders adopt a more cynical view of central bank forecasts as 'cheap talk' (Buiter, 2013, p. 2) or 'jawboning' (Goodhart, 2012). Such statements raise the prospect of a potential future communicative apparatus of expectation management that must do without pretence of knowledge and credibility. Will market participants become increasingly prone to seeing central bank forecasts as rhetorical rather than informational devices, deployed strategically by central banks trying to manipulate expectations? Will they, in other words, move closer to the understanding of expectation management developed in this paper? Looking forward, the production of knowledge about the future and the contestation between competing producers and users of such knowledge will provide fertile ground for further research in both economic sociology and political economy. 5 The EMI was set up by the Maastricht Treaty and was operative from 1994 until the Eurosystem took over on 1 June 1998. 6 In order to make sense of this academic hostility towards the first pillar it is important to recallas the ECB researchers conducting the background studies for the evaluation did (Masuch et al., 2002, p. 219)that the theoretical edifice of the new neoclassical synthesis in macroeconomics does not foresee a meaningful role for money. 7 Interestingly, given the debates surrounding the justification and relevance of the monetary analysis, the ECB continues to circulate its monetary projections only internally (Interview 8). 8 The new March and September editions were ECB staff projections, as opposed to the June and December editions, which were Eurosystem projection exercises that included the staff of the NCBs (ECB, 2004a). 9 Since 1998, the Bank of England has regularly published two forecastsone with a constant interest rate assumption, the other on the basis of market expectations. 10 Note that this is very different from how communication used to be approached by central banks. One interview partner recalled an episode during the 1990s when the Bundesbank actually postponed an interest rate change because it was too widely expected by market participants (Interview 9). 11 Initially, the ECB published only brief summaries of the results of these surveys in the Monthly Bulletin. Yet here, too, the pattern of increasing disclosure was repeated as the ECB began, in October 2003, to publish the complete aggregate results of the survey on its website (Garcia, 2003, p. 8). 12 This interpretation is further supported by the fact that market participants' expectations about future inflation can also easily be deduced from inflation-linked financial products. Specifically, the ECB computes so-called break-even inflation rates 'as the difference between yields on nominal bonds and yields on comparable inflationlinked bonds at the same maturity' (ECB, 2012, p. 71). 13 The argument has been used against Keynesians by Keynesians (Coddington, 1976), against Keynesians by monetarists (Friedman, 1968) and New Classical economists (Lucas, 1976), and against over-ambitious goals for monetary policy by central bankers (Greenspan, 2004). 14 For an interesting take, albeit in the context of a different argument, on 'rationalities of ignorance' in finance, see Davies and McGoey (2012).