“Beauty is a form of Genius--is higher, indeed, than Genius, as it needs no explanation…It cannot be questioned. It has divine right of sovereignty. It makes princes of those who have it.” - Oscar Wilde.
One does not usually associate supermodels with those in the risk management profession. While the term ‘models and bottles’ has long been associated with high-flying investment bankers, their middle-office brethren, the risk managers, have always played the nagging house-wife to the bread-winners of high finance. Yet those in the middle office, while far removed from the limelight of the catwalk, have taken to supermodels of a different kind; models whose beauty is defined by mathematically elegant expressions of risk.
The financial supermodels of risk managers are numerous, and their creators often achieve a form of celebrity status amongst fellow academics and industry practitioners. Markowitz’s Modern Portfolio Theory, the ensuing Capital Asset Pricing Model and the Black-Scholes-Merton Options Pricing model are all prime examples where the model’s creators were deemed (perhaps foolishly?) worthy of Nobel Prizes. A notable exception to this honour-roll, however, is the widely used (and Basel regulated) Value at Risk (VaR) metric.
Perhaps because of its practical development from within industry, as opposed to the white-glove environment of academia, VaR is seen as more of an every-day risk management tool, rather than a feat of ground-breaking intellectual rigor. Essentially, it summarises risk to a single number; the dollar amount of mark-to-market losses on a portfolio that should not be exceeded within a certain time frame at certain high level of confidence. But it is this supposedly simple nature of VaR, combined with the exposure of risk managers to it every day, which makes this unassuming supermodel potentially the most dangerous.
Nassim Taleb, author of Fooled by Randomness (2001) and The Black Swan (2007), has been one of the fiercest critics of VaR as a risk management tool. One of Taleb’s primary criticisms is that models of uncertainty are too precise, given the condensation of complex factors necessary to produce a single-figure risk measure for a large portfolio. He argues that such undue precision enables investors and managers to be lulled into a false sense of security, breeding contempt for risk and a false sense of control.
Yves Smith is similarly scathing of VaR as an effective risk management tool in her book Econned (2010). Aside from arguing that the degree of abstraction necessary for the production of a VaR discards important information about the behaviour of the underlying systems, she directly devalues what a VaR estimate is really worth. By adopting confidence levels of 95-99%, which are relatively good at predicting day-to-day risk, she argues that VaR does not focus on what is really important to risk managers, namely what lies in the remaining 1-5% of the loss tails.
Indeed, VaR is silent with regard to what lies beyond the chosen confidence level, and this silence can prove deadly. Hull (2012), details how undesirable risk-taking can be inadvertently encouraged when VaR is used to try and limit the risks taken by a trader. In essence, traders who like taking high risks in the hope of realising high returns (and fat bonuses) are able to structure trades that satisfy VaR limits imposed by a bank while simultaneously exposing the bank to massive losses in the silent tail, utilising a return distribution that is anything but normal.
Defenders of VaR, such as industry practitioners and authors Eric Falkenstein and Suna Reyent, point to more recent methods of estimating the tail risks, such as Extreme Value Theory, Expected Shortfall and Conditional VaR. Importantly, they note that most firms don’t rely solely on VaR, but supplement this measure with other methods such as stress-testing and Monte Carlo simulation. Yet detractors such as Smith and Taleb argue that all of these approaches send broadly similar signals, and do nothing much to solve the problem of reliance on historical data.
The very nature of historically-based volatility estimation methods upon which VaR is based, such as GARCH modelling, certainly seem to imply the adoption of pro-cyclical capital policies by financial institutions. By encouraging banks to hold less capital in good times and dictating they hold more capital in times of market turbulence, they encourage build-ups of leverage followed by a rush to the exits and the freezing of credit markets.
Sadly, given the cut-throat nature of the global capital markets, management’s hands are often effectively tied when deciding which risk metrics to push. It doesn’t take much imagination to envision competitive pressures forcing the hand of management to adopt capital-light risk metrics during times of low volatility, of which Basel regulated VaR classifies as a prime candidate (the upcoming Basel III is an attempt to rectify this). While the risk managers may have their stress-tests and methods for estimating tail risks, anecdotal evidence from the GFC overwhelmingly points to the sacrifice of long-term risk management on the altar of next quarter’s results; the nagging wife losing out again to the breadwinner of the household.
This race to the bottom of minimally regulated capital requirements in order to remain competitive and satisfy shareholders has the effect of focusing the attention of risk management on the numbers. Similarly, where profitable risk-taking by utilising high levels of leverage can be justified by focusing on the numbers and ‘black boxes’, as with Long Term Capital Management and other so called ‘scientific’ hedge funds, the focus inevitably remains on the justifiable metrics, rather the nature of the assets under management.
This approach is diametrically at odds with the value-based investment strategy of arguably the world’s most successful investor, Warren Buffett (a long time critic of VaR and Modern Portfolio Theory). Rather than focusing on statistically definitive volatility as a measure of risk, Buffett’s value investment approach relies instead on a hard-headed analysis of an investment’s prospects rather than its price movements. Unfortunately, these types of business judgements and valuation skills are not easily taught, and are ugly and vague when compared to mathematically beautiful models and claims of scientific justification.
Thus, while VaR and other quantitative risk measures are known to have significant shortfalls, management seeking justification for myopic risk taking and regulators seeking a hard, enforceable rule, are seemingly destined to focus on the simplified, backwards-looking numbers, which are worth almost nothing when it comes to estimating anything other than day-to-day risks; the very risks which least require sophisticated risk-estimation tools.