Standard Deviation as a deviation from reality part deux

Felix opens a can of worms when he picks on the Standard Deviants at Morgan Stanley, who, reliant on standard-deviation-based risk limiting measures, got their butts well and truly kicked in the recent Troubles which we choose to call “The Subprime Meltdown”; losses exceeded a 95% confident -$85m Value-at-Risk assessment 6 days in a row. More than that, they dropped $390m on the worst of those days, over 4 times the assessed VAR. Commenters jump on him, I think unfairly, accusing him of misunderstanding the er value of VAR measurements as a risk limiting tool. Tremble, commenters anonymous, jck, and Sandy, for it is YOU who are the hidebound know-nothings here. You have awakened the sleeping Baruch! His dislike of standard deviation models is legendary.

VAR for banks is like the sacred, magically fireproof, undergarment worn by Mormons, meant to protect them when the fiery apocalypse comes; it gives a false sense of security. With Mormons, this is fine, as they are generally unable to start the fiery apocalypse themselves. Let them continue in their harmless and mildly amusing superstition. Surely the point is that with large banks it is different; they themselves control the levels of risk they carry, and if they get it worng they can hurt a lot of people. Believing themselves protected when they are not can thus be very dangerously misleading. Felix is right to be incredulous when commenters point out that VAR is a “useful” measure when “employed properly”, except of course when conditions in the market change and the old certainties no longer hold true: that is precisely when you are most vulnerable to risk! To have a risk-limiting measure which stops working when it is most needed makes the top of my head fly off it is so stupid.

It is much as Taleb says; the business of banking — in this case, more properly prime-broking, liquidity provision and prop-desking — is one that is marked by gross misconceptions of the level of risk. It is a classic nickel and steamroller business model, one admittedly with a rather slow but very heavy steamroller. Blowups occur very rarely; but when they do they wipe out multiples of the profits of past periods. The events of August and the carnage in the financials are probably only mild occurrences of what can happen, and has happened in the past.

And the VAR fallacy is very widespread; my own beloved employers have a new multi-strategy project they are starting. They have hired multi-million-dollar-earning hotshots from “top players” from prop desks on Wall Street. One of the major criteria that will be used in daily risk-assessment, imported by said hotshots, is, you guessed it, VAR.

So excuse Baruch’s getting exercised about this; it’s just getting a bit close to home. The only point Felix is wrong about here is his opening statement that “Morgan Stanley has some of the most sophisticated risk-management systems on planet earth”. In fact they merely wear some of the frilliest fireproof underpants.