![Sample our Politics & International Relations journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5947/TOC-Subject-Sample-2021-Politics-International-Relations.jpg"})
ABSTRACT
Starting from an observation about the high-profile predictive failures of Value-at-Risk (VaR), an internationally instituted financial risk model, this article has attempted to make sense of its continued use by analyzing its productive, rather than predictive, power. This line of inquiry leads me to identify VaR's (counter)performative effects and the way in which it produces banks as authoritative, responsible managers of an uncertain financial future. Viewing financial markets through the lens of Keynesian uncertainty and model performativity helps explain VaR's failures by revealing VaR to be an inherently limited and potentially destabilizing practice. Its use participates in the construction of a financial system that is only temporarily stable and controllable. At the same time, VaR is an important source of authority for banks vis-à-vis regulators and the public because it represents the future as statistically calculable and expert prediction as the optimal, objective mode of preparing for that future. This, in turn, makes less thinkable other responses to uncertainty – ones that might be better suited to contend with the possibility of devastating losses unforeseeable – and perhaps produced – by the widespread use of VaR.
ACKNOWLEDGEMENTS
The author wishes to thank Stephen C. Nelson for his very generous advising, from this article's inception through the review process; the International Relations Student Working Group at Northwestern University for its constructive feedback on an earlier draft; an excellent panel at the 2014 ISA Annual Convention in Toronto; and the detailed and constructive comments of three anonymous reviewers whose feedback was vital to strengthening the article.
Notes
1 For example, if a bank says the daily VaR of its portfolio is $40 million at the 99 per cent confidence level, that means there is a 1 in 100 chance that a loss greater than $40 million will occur.
2 Büthe and Mattli (Citation2011: 22) note the effects of the Basel Committee's recommendations extend well beyond the regulators that participated in setting them: ‘Numerous public regulators who had no voice in setting these capital adequacy standards thus ended up adopting them’
3 My position in this article is not that markets ‘really are’ governed by uncertainty rather than risk, but that it is analytically useful to view them as partially characterized by Keynesian uncertainty as this helps us see the way in which calculative tools like VaR construct them in contingent ways. I do not think risk and uncertainty are mutually exclusive, though I do think that the islands of predictability that do emerge in markets are not inherently so, but exist as social accomplishments.
4 Mandelbrot and Taleb (Citation2010: 51) point to the 1987 stock market crash, the 1992 crisis in the EU exchange rate mechanism, and the 2007–8 financial crisis as events that, according to extant risk models, should only happen one in a googol (one, followed by a hundred zeros) times.
5 While these extreme cases fall outside the scope of what the models claim to be able to predict, the magnitude of losses speaks powerfully to why we should be concerned with the limitations of risk modeling.
6 VaR models can also be based on a Monte Carlo simulation which generates a distribution of possible outcomes by running multiple random hypothetical trials. In this case, financial gains and losses are assumed to be stochastic and thus amenable to probabilistic analysis, in contrast to the non-ergodic view of the financial system that underlies contemporary understandings of Keynesian uncertainty (Holton, Citation2003: 193–198).
7 Brown (Citation2008: 20) nonetheless defends VaR in a subsequent article, though he acknowledges its limitations, noting that ‘A 99% one-day VaR has to operate for about three years before you can trust it. A 99.97% one-year VaR, which some people use for economic capital, requires 26,000 years for the same level of confidence. That makes deep tail VaR a matter of faith and assumptions, not something you can observe with reasonable statistical certainty over a moderate time interval’.
8 Other examples of financial models that incorporate statistical methods include the capital asset pricing model and the Black-Scholes-Merton option pricing formula.
9 Although MacKenzie uses the term ‘Barnesian performativity’ to distinguish this phenomenon from a more general sense of performativity, in which economic theories are used in economic practice, in this paper ‘performativity’ refers exclusively to MacKenzie's Barnesian variety, in which the practical use of models make economic processes more like their theoretical depiction. For an example of how MacKenzie's other forms of performativity can also be mobilized in IPE see Henriksen (Citation2013).
10 NatWest Associate Director for Trading Risk David Palmer said of negotiations, ‘Throughout the process of preparing the new rules, the Basel Committee have shown their willingness to listen to the industry's comments and take action based upon them’ (Richardson, Citation1995/96).
11 The qualitative standards for internal risk-models are: an independent risk management unit within the bank; back-testing of the model; senior management involvement in risk management; that the model be used in conjunction with the bank's trading and risk exposure limits; regular stress-testing; and independent external review of the model (BCBS, Citation1996b: 41–42). Quantitative standards include that the VaR be computed daily; a one-tailed 99 per cent confidence interval be used; the historical observation period of past price data be at least a year; data sets be updated every three months; and the model capture the non-linear price movements of options (BCBS, Citation1996b: 44–47).
12 One method that does attempt to account for the magnitude of losses in the extreme tails of the VaR distribution is expected shortfall or conditional VaR, which approximates the expected loss during a given period, conditional on that loss being greater than the Xth percentile of the loss distribution (Hull, Citation2007).
13 Richardson (Citation1995/96) writes, ‘In theory, the multiplication factor compensates for many of the nonquantifiable factors that can influence the estimation of risk such as flawed distribution assumptions, the inadequacy of past events as a guide to future ones, extreme market movements, and other factors that may limit the accuracy of a VAR approach but its ability to accomplish this seems doubtful.’
14 For one detailed proposal for alternative forms of financial regulation see: The Warwick Commission on International Financial Reform: In Praise of Level Playing Fields (Citation2009). The introduction of leverage-based regulation in the Basel III framework represents another such shift. Nonetheless, as Andrew Haldane (Citation2012: 19–20) observes, risk-weighted capital ratios are still favored over leverage ratios in the current framework.
Additional information
Notes on contributors
Erin Lockwood
Erin Lockwood is a PhD candidate in the Department of Political Science at Northwestern University, Evanston, Illinois, USA, studying International Relations and Political Theory. Her research interests include the politics of global financial markets, risk modeling, authority, derivatives, Keynesian uncertainty, and pirates.
