But a shortish, informal review on Amazon is all we´ve got, so let´s face reality and delve into the analysis. The review is a mix of positive and negative assessments. Such apparently unenthusiastic response may throw some authors back, but not me. I am too busy glowing in the fact that Brown finds good stuff buried within all those bothersome inverted pluperfect subjunctives (he says I remind him of Joda, of Star Wars fame). I can proclaim that I wrote a book that Aaron Brown found (somewhat) useful. That´s a very good thing. For not only is Brown a very prominent risk professional, or a very gifted and thoughtful "financial intellectual". He is also a renowned defender of the quanty stuff. So when someone like that is willing to compliment a book that goes heavily at the quanty stuff, you know that at least something has been done right (of course, I personally believe that a lot in the book is right).
Let´s stop basking in short-lived, modestly-served glory, and turn our attention now to Brown´s criticisms. He claims that I lose out to BSM, VaR, and other theoretical "straw men", that my views are deeply misinformed. I beg to differ. I think here I have the advantage of not being a quanty type and thus of having taken the approach of looking at the historical record and the opinion of other, much better informed, experts when analyzing the validity and wholesomeness of the best-known models, rather than try to engage them on mathematical grounds (though, naturally, I do attempt to dissect the essential technical deficiencies). So when I declare VaR a failure, I do so based on its behaviour during the credit crisis or during the Asian-LTCM meltdowns. I do it based on the monstrous shortcomings of the tool during crunch time out there in the real world. I think my book is the first ever to actually collect and display the actual data: what figures VaR registered, the number of exceptions incurred, how actual trading losses deviated from VaR. All that data comes straight from banks´ regulatory filings, a pretty solid source. I wouldn´t exactly call that misinformedness.
Same with BSM. My analysis of its role in the 1987 crash comes from official reports and other well-informed sources. My interpretation of what the volatility smile means in terms of the model´s reliability has been espoused by many through the years. And my summary of the Taleb&Haug paper is exactly that, a summary of a pretty trustworthy reference. So the foundations on which my lambasting of BSM rests also have the smell of solidity. I detect little misinformedness here too.
Brown says that risk management is not about predicting, but rather about preventing disasters. Fine, but then why use VaR? VaR is a predictor, with a degree of confidence. And VaR can´t, almost by definition, capture disasters. Many VaR lovers have, serendipitously, ceased to endow VaR with the predictor label following the (VaR-fueled, VaR-exposing) credit crisis. Upon seeing all those exceptions (80 real versus 5 theoretical in the case of UBS for 2007-2008, I seem to recall) it is only natural that they would want to deviate out attention from hitherto familiar arguments a la "99% VaR will not be exceeded more than twice a year". But that does not mean that I am wrong when, upon studying the hard evidence, I dare to proclaim that VaR failed monstrously.
Finally, Brown asserts that institutions with bad risk management fail only once. By that definition, VaR is clearly bad risk management. Anybody remembers Bear Stearns, Lehman Brothers, or Merrill Lynch? They used VaR for risk guidance and capital-charge setting. They had never failed in their entire combined history (three centuries?). All it took for such holly tradition to be broken was the concurrent abidance by a statistical tool that encouraged, allowed, and afforded the vast accumulation of the most toxic positions ever witnessed on Wall Street. VaR made sure that those legendary, erstwhile indestructible institutions would not fail a second time. When VaR kills you, it kills you for good.