Part I
Heuristics
1 Does foreign exchange economics âperformâ the currency markets?
Overflowings occur when goods act unpredictably ⌠transgressing the frames set for them and the passivity imposed on them.
(Callon 2007a: 144)
Michel Callonâs notion of the performativity of economics provides one of the authoritative approaches discussed among sociologists and anthropologists1 to conceptualizing the role of knowledge in markets. Borrowing its core idea from the Austinian theory of speech acts (e.g. Austin 1962), the performativity thesis posits that economics, much like a performative speech act, âperforms, shapes and formats the economyâ (Callon 1998: 2) and thereby âcontribute[s] toward enacting the realities [it] describesâ (Callon 2007b: 315).2 Economics, according to Callon, includes the academic discipline as well as various forms of practitioner expertise (bookkeeping, accounting, marketing, etc.). In his 1998 essay, he specifically describes how economics enacts the economy: it produces frames and thereby enables economic actors, including participants of a particular market, to calculate. Furthermore, frames are described as devices that âdemarcate, in the regards of networks of relationships, those which are taken into account and those which are ignoredâ (1998: 15). Thus,
It is owing to this framing that the market can exist and that distinct agents and distinct goods can be brought into play. Without this framing the states of the world can not be described and listed and, consequently, the effects of the different conceivable actions can not be anticipated.
(Ibid., 17)
By framing actorsâ relationships with the world, economics produces âcalculative agenciesâ (ibid., 23). Callon goes so far as to argue that, without any of these framings, the anthropological basis of all calculation, namely the profit-maximizing homo oeconomicus, would not be possible (ibid., 51). Arguing against the sociological enrichment of markets, Callon arrives at the proposition that âthe economy is not embedded in society but in economicsâ (ibid., 30). He is ready to admit, however, that the success of this constitutive process is far from certain (Callon 2010). Rather, in contrast to Mertonâs theory of self-fulfilling prophecies, his performativity theory emphasizes that any economic knowledge can only become effective in combination with âsociotechnical agencementsâ:
To move a statement from one spatiotemporal frame to another and for it to remain operational (that is, for it to be capable of describing situations and providing affordances for them), the sociotechnical agencement that goes with it has to be transported as well.
(Callon 2007b: 331)
Thus, for theories, concepts and models to move from the academic world to real-life markets, these calculative tools have to both materialize as market devices, and stabilize within the larger institutional arrangements and infrastructures that put, for instance, the ceteris paribus assumptions of a theory to reality (ibid., 332):
A formula that previously functioned in a paper world, which was perfectly real (for what could be more real than paper or equations), subsequently functions, after many investments, in a world of computers and silicon, algorithms, professional skills, and cleverly adjusted institutions. We could say that the formula has become true, but it is preferable to say that the world it supposes has become actual.
(Ibid., 320)
Hence, according to Callonâs approach, markets emanate from collective performances of calculations, made possible by economics; they are, in his words, âcalculative collective devicesâ (Callon and Muniesa 2005: 1230).
How can Callonâs concept be explored empirically? Donald MacKenzie (2008) shows that, from the 1960s onwards, financial economics has produced a whole set of mathematical models, including modern portfolio theory (MPT), the efficient market hypothesis (EMH), the capital asset pricing model (CAPM) and the BlackâScholesâMerton formula (BSM), which all can be used (thanks to their mathematical formulation) as devices of financial calculation. In order to test the potentially performative use of these devices in financial markets, MacKenzie distinguishes three different types of performativity. On the most basic level, financial economics becomes used in practice, for instance by speaking about finance in terms of models; MacKenzie calls this general use of economics âgeneric performativityâ. Financial economics becomes performative in a more narrow sense, though, only if the use actually makes a difference in terms of calculations, transactions and prices (ibid., 18). Unfortunately, this difference is difficult to qualify, as the empirical world does not readily offer experiments with control cases. Therefore, MacKenzie adopts a historical approach. He argues that economic processes must, through time, be âaltered in ways that bear on their conformity to the aspect of economics in questionâ (ibid., 18â19). MacKenzie calls this âBarnesian performativityâ. The empirical test for Barnesian performativity is whether prices and market conditions actually move âtoward greater conformity to the theory or modelâ (ibid., 21). MacKenzie and Millo (2003) and MacKenzie (2008) have tried to test Barnesian performativity in the case of the BSM and its use on the Chicago Board Options Exchange. While acknowledging the complementary importance of classic concepts of the new economic sociology, they find support for Callonâs thesis: for a limited period of time, prices comply with the theoretical prices of the model as a performative effect of the modelâs use.
The particular case of the option pricing formula and MacKenzieâs work on financial economics in general have put the performativity thesis at the centre of debate on financial market knowledge. Before introducing an alternative conception, therefore, let me first discuss how far the concept of Barnesian performativity can usefully be applied to the foreign exchange markets. The more specific empirical question is whether actual practices on currency markets have come to âperformâ the models developed in foreign exchange economics; in particular, conformity between theoretical and empirical prices would be a preliminary indication that economics in fact delivers the models according to which traders and analysts construct devices of calculation used in these markets. I explore this question in two steps. I first reconstruct how the field of contemporary foreign exchange economics has emerged and how it has formulated models for pricing fluctuating exchange rates. I will restrict myself to a brief, non-technical discussion of foreign exchange models after 1973; this may leave the economic expert dissatisfed but allows me to keep focussing on the broader argument. I then turn to the processes that have led to the failure of these models, challenging the performativity concept (at least in its narrow, Barnesian sense) as an adequate theoretical approach to my particular case.
General equilibrium, currency speculation and rational expectations
A distinct field of foreign exchange economics, focussed on the mathematical modelling of theoretical prices, only fully developed after the effective abandonment of the global regime of exchange rate pegs called Bretton Woods.3 A key event, in this respect, was a conference in Stockholm, held from 26 to 27 August 1975, where leading economists met to discuss âflexible exchange rates and stabilization policyâ. During this conference, economistsâ primary concern was to âdevelop a theory suggestive of the observed large fluctuations in exchange ratesâ (Dornbusch 1976: 1161).4 For after the âfloatâ, ânominal and real exchange rates proved to be more volatile than when currencies were pegged and than predicted by academic proponents of floating. Nominal rates frequently moved by 2 or 3 percent a monthâ (Eichengreen 2008: 128).
At the conference, foreign exchange rates were explicitly conceptualized as market-determined prices for currencies. In modelling this market-pricing, economists tried to combine two theoretical aspects. On one hand, they saw exchange rates as international equilibrium prices corresponding to purchasing power parity.5 On the other hand, currencies were understood as financial assets that were driven by expectations for future profits.6 This ambivalence led Rudiger Dornbusch to the following statement during the closing discussion of the Stockholm-conference:
(1) In principle the exchange rate is determined by the general equilibrium ⌠of the relevant model.
(2) Specifically, asset markets are important.
(3) Among the asset markets it is unobjectionable to single out the money market as the centre of analysis.
(4) Independent of school of thought, capital mobility and expectations are of critical importance in the determination of exchange rates.
(Calmfors and Wihlborg 1976: 389)
How would the theorists bring âgeneral equilibriumâ (equilibrium in the goods and money markets) and âcapital mobility and expectationsâ together? The most important solution to this question was dubbed the âmonetaryâ or âasset approachâ,7 promoted by, among others, Jacob Frenkel and Rudiger Dornbusch. Frenkel, representing the Chicago variant of the asset approach, argued:
The demand for domestic and foreign moneys depends, like the demand for any other asset, on the expected rate of return. Thus it may be expected that current values of exchange rates incorporate the expectations of market participants concerning the future course of eventsâŚ. If the foreign exchange market is efficient â as many other asset markets appear to be â then current prices should reflect all available information.
(Frenkel 1976: 204)
Thus, Frenkelâs solution consisted of incorporating Eugene Famaâs theory of efficient markets into foreign exchange economics. Frenkel posited that purchasing power parity holds because the subject of expectation formations is inflation, that is, change in the value of a currency in terms of goods. Likewise, Rudiger Dornbuschâs more Keynesian variant aimed at integrating macroeconomics with financial economics. Dornbusch stated, in correspondence with Frenkel, that exchange rates depend on the demand for one money relative other monies. He saw this demand as primarily driven by the interest rate differential, the economiesâ outputs and the relative price levels of domestic versus foreign goods. Since interest rates were seen as internationally arbitraged (producing one real world rate) and output was supposed to be fixed, the Dornbusch model stated that, in the long run, exchange rates developed according to purchasing power parity, i.e. the change in inflation of the domestic as compared to the foreign economy. Dornbuschâs model further stated that in the short run, asset markets, i.e. international markets for financial instruments, drive exchange rates. He wrote: âIt is quite obvious from the preceding explanation that the short-run effects of monetary expansion, in this model, are entirely dominated by asset markets and, more specifically, by capital mobility and expectationsâ (Dornbusch 1976: 1169).
Muthâs theory of rational expectations,8 then, allowed the integration of the macroeconomic assumption of purchasing power parity with the short-term impact of speculation. Dornbusch assumed that speculators would use his foreign exchange rate model in order to form their expectations. Hence, they would formulate expectations that are perfectly consistent with the laws of the macro-economic adjustment of exchange rates to relative prices, including the fact that the monetary adjustment is slower than the reallocation of financial assets in response to unexpected interest rate changes: âThe expectations coefficient ⌠corresponds to perfect foresight, or, equivalently, ⌠is consistent with the modelâ (ibid., 1167). Thus, by employing rational expectations theory, Dornbusch and others argued that foreign exchange markets are in equilibrium as long as market participants form their expectations according to the models through which exchange rates are related to economic fundamentals. Arguably, in saying that expectations correspond to a commonly held model explicated by economists, the monetary approach entailed what can be called an âimplicitâ performativity thesis.
As argued above, Michel Callon conceptualizes âperformativityâ as an act of translation where stakeholder groups, theories, concepts and empirical phenomena are enrolled in a network that is connected by a specific language and communal project (Callon 1986). The âfelicityâ of a framing, then, primarily depends on the capacity of rearticulating reality in its own terms. Indeed, this is what Dornbuschâs and othersâ models achieved successfully: they combined macroeconomics and financial economics in order to frame the phenomenon of volatile exchange rates in their own, general equilibrium terms. Accordingly, Kenneth Rogoff acknowledges that
the âasset approach to exchange ratesâ, pioneered by Dornbusch, Frenkel, Mussa and others, seemed to âhit the nail on the headââŚ. The logic was impeccable and the most attractive feature of the new theories was that they seemed to offer an explanation for why exchange rates are so volatile.
(Rogoff 2001: 2)
In fact, the plausibility of these models was so great that they were quickly dubbed âcanonicalâ. However, successful framing can only be maintained if the framing survives crucial trials of strength. Economics has its own trials: models become established truths if they ultimately enable the prediction of the dependent variable with which the model is concerned.
This approach, entitled positive economics, was first advocated by Milton Freedman (1953), and had considerable impact on foreign exchange economics. First, economists tested their models econometrically within the samples for which the models and their parameters had been developed. The real trial of strength, however, would be to test models âout-of-sampleâ, i.e. for other time periods and other exchange rates than originally used in the modelâs development. In the early 1980s, Richard Meese and Kenneth Rogoff, then staff members of the Federal Reserve Board and its foreign exchange forecasting group, were given the task of conducting such tests. They took the different variants of the widely accepted asset models and exposed them to new data, then compared the modelsâ forecasting performance with that of a random walk, i.e. a series of consecutive forecast values generated by chance. When running the tests, the economists were surprised to see that no model could predict exchange rate movements better than the random walk. Even when taking actual values instead of forecasts for the independent variables, the economists had to acknowledge that the models performed poorly. At first, they hardly believed the results, but as Rogoff reports:
a couple of thousand regressions later, we began to realize that this result was VERY robust to data set, model specification, error term specification, estimating technique, choice of theoretical model, etc. Though I admit we were more than a little amused by our failure, we were not sure how to convey our findings to our colleagues in the exchange rate forecasting group, especially after having set off so sure that our knowledge of the new theoretical literature on exchange rates would prove invaluable.
(Rogoff 2001: 4, emphasis in original)
In fact, Meese and Rogoff faced considerable difficulty publishing these results: the editor of the American Economic Review sent back their manuscript, saying that âthe results are obviously garbage and if we wish to remain in the economics profession, we had better develop a more positive attitudeâ (ibid., 4). The economists finally got the paper published in another periodical: the Journal of International Economics (Meese and Rogoff 1983).
However, this single paper, despite its resolute judgement on the modelsâ bad out-of-sample performance, should not destroy the asset approach, especially if the latter was supported by such a high degree of plausibility. After all, the BSM, as described by MacKenzie and Millo (2003), was not initially true either; it only became true through institutional changes, widespread adoption by market participants, development of new market infrastructures, etc. However, the failure of FX models proved more persistent than that of the BSM: subsequent tests showed that neither the specific models of Dornbusch nor in fact any other macroeconomic model could provide robust predictions. Indeed, within the regime of floating exchange rates since 1973, neither has the purchasing power parity condition held9 nor have macroeconomic variables correlated sufficiently with exchange rate movements to serve as possible explanatory variables (Meese 1990). In 2001, Rogoff noted: âTo make a long story short not only have a subsequent twenty years of data and research failed to overturn the Meese-Rogoff result, they have cemented itâ (Rogoff 2001: 1). Such theoretical problems have led to a serious crisis in FX economics:
To repeat a central fact of life, there is remarkably little evidence that macroeconomic variables have consistent strong effects on floating exchange ratesâŚ. Such negative findings have led the profession to a certain degree of pessimism vis-Ă -vis exchange rate research.
(Frankel and Rose 1995: 1709)
Overflowing
How to make sense of this constellation? Attempts at framing seem to generate another process that Callon describes as overflowing (Callon 1998: 18; 47â48). Overflowing occurs âwhen goods act unpredictably ⌠transgressing the frames set for them and the passivity imposed on themâ (Callon 2007a: 144). Callonâs view, then, is that overflowing occurs because any attempt at disentangling and demarcating entities from others creates unexpected new entanglements. Economists indeed recognize overflowing in the sense that market actorsâ expectation-building does not seem to correspond with modelsâ prescriptions. In fact, the fieldâs crisis means that its theoretical frames cannot capture the complexity of these market processes. Anecdotal field observations, surveys and, most importantly, a growing interest in the activities of important market participants (primarily dealers) indicate the inadequacy of most models.10
The consequence is a situation in which many conceptions of foreign exchange rates exist simultaneously (especially macroeconomic and microstructure approaches) without any possibility of refuting one theory in favour of another.11 ...