Pablo Triana, a professor at ESADE Business School (Spain), is a man on a mission: to rid modern finance of complex mathematical models. In The Number That Killed Us: A Story of Modern Banking, Flawed Mathematics, and a Big Financial Crisis (Wiley, 2012) his target is one of the most widely used models, VaR (Value at Risk).
Triana’s thesis is fairly straightforward, although he spends over 200 pages fleshing it out. VaR is fundamentally flawed because it relies on historical data, viewing the past as prologue; it assumes a normal probability distribution; and it doesn’t differentiate among kinds of assets. Moreover, since VaR is the key metric invoked to determine leverage, traders can use it to game the system. They can put together portfolios with ostensibly low risk profiles and hence be eligible to use more leverage, an exercise that sometimes masks reckless behavior.
VaR, Triana argues, was the “top miscreant” in the financial crisis. “Without those unrealistically insignificant risk estimates, the securities that sank the banks and unleashed the crisis would most likely not have been accumulated in such a vicious fashion, as the gambles would not have been internally authorized and, most critically, would have been impossibly expensive capital-wise.” (p. 3)
Triana wants to get rid of VaR. What should replace it? “Going forward, let’s do less mathematical financial risk analysis, please. Softer sapience based on traders’ war scars, experience-honed intuition, historical lessons, and networking with other players will not only typically beat quant sapience when it comes to understanding and deciphering exposures (we humans can’t be that bad!), but most crucially should be far more effective in preventing obviously lethal, chaos-igniting practices.” (pp. 44-45)
Financial risk, he contends, “is a simple discipline. Or rather, a discipline that ought to be based on fairly simple tenets: Financial risk is not measurable or forecastable, the past is not prologue, battle-scarred experience-honed intuitive wisdom should be accorded utmost notoriety, certain assets are intrinsically riskier than others, too much leverage should be avoided, and too much toxic leverage should be banned.” (p. 213)
The book closes with a brief Q&A with Nassim Taleb and an essay by Aaron Brown that presents a more balanced view of VaR.
I’m certainly no expert on VaR or on the risk management practices of investment banks, but from the little I know Triana does justice to neither. Considering that he doesn’t want to reform quantitative risk management but either to abandon it or to keep a simplified version of it on a tight leash, I suppose fine points are irrelevant. Triana paints with broad, impassioned brushstrokes.
In my opinion supplementing VaR in particular and quantitative models in general with a large dose of human wisdom is a laudable goal. Maintaining a healthy margin of safety is important even if it means diminished profits during good times. Keeping models as simple as possible is undoubtedly good practice. On the other hand, chucking math and replacing it with so-called “battle-scarred experience-honed intuitive wisdom” is not. For one thing, behavioral finance has taught us that, despite our best intentions, we can be terrible bunglers. We humans really are that bad! For another, what masquerades as wisdom often turns out to be an oversized ego. Just think of …. Well, I’m sure you can easily fill in the blanks.