"either an idiot... or possessed of a dangerous agenda and very cunning"
Monday 5 January 2009
Risk mismanagement
There's an interesting NYT article here about the role of Value at Risk in the run up to the crisis. Taleb crops up quite a bit understandably, but it's more sympathetic to the role of VaR than he is (not difficult really).
It's a good, balanced article and the people quoted (with one notable exception !) make good & balanced statements
Taleb has just been silly since Black Swan: by contrast, Fooled by Randomness very properly lauds Monte-Carlo analysis. Now VaR can be approached through 3 methods: the original Gaussian (assuming log-normal distributions, not normal BTW); Historical; and Monte Carlo.
As eny fule kno, Monte Carlo is the way to go (as the pre-Swan Taleb says): the problem is that it takes an age to compute, versus seconds to do the other methods. All this talk of the '4:15' report relates to a Gaussian VaR done to give a quick-n-often-a-bit-dirty answer for immediate action. Its limitations are very well-known
Incidentally, nowhere mentioned in the article is the fact that VaR is itself a subset of a more general concept called CaR (C = Capital), which removes one of the most limiting aspects of VaR, namely that it assumes perfect liquidity over the relevant 'holding period' (which itself should only be 1 or 2 weeks at max for the maths to work, but is better when it is properly measured only in days)
CaR assumes an 'infinite' holding period (i.e. doesn't assume you can dump a position quickly if it's moving against you, a fundamental premise of VaR); is more subtle & complex; & can't be done at the press of a button at 4:15 each day, but rather should be chewed over in much the way GS reportedly did in the anecdote.
Overall I would say gaming, and wilful (compulsory!) institutional optimisim, lie at the heart of the 'analytic' problems. I rarely meet a senior risk manager who doesn't have a very shrewd idea of what the real risks are, but they are 'strongly urged' to keep their own counsel: the dealmakers rule the universe and they just don't wanna know: it has negative consequences for dealflow and bonus !
the risks that VaR measured did not include the biggest risk of all: the possibility of a financial meltdown: OK, so "now we know this" (of course we always did), what we gonna do about it ??? Beyond the 95% (or 99%, as will now become the very costly norm) lies (a) diversification and (b) sovereign support. Just as no industrial organisation can economically afford to create a "100%" physically safe workplace, no financial system can economically afford cover against all the ultra-extremes.
You end up taking the 17th / 18th C approach of actually banning all derivatives, which is just childish. This said, there are some extreme instruments that should be effectively banned - by making the conventional / mandatory calculation of the risk involved so costly (in terms of Capital Adequacy) that no-one will enter them
sorry for taking ages to reply, I don't know what to add though as you obviously know this area a lot better than I do! agree about Taleb becoming a bit silly. I thought Fooled By Randomness was the better of the two books.
3 comments:
It's a good, balanced article and the people quoted (with one notable exception !) make good & balanced statements
Taleb has just been silly since Black Swan: by contrast, Fooled by Randomness very properly lauds Monte-Carlo analysis. Now VaR can be approached through 3 methods: the original Gaussian (assuming log-normal distributions, not normal BTW); Historical; and Monte Carlo.
As eny fule kno, Monte Carlo is the way to go (as the pre-Swan Taleb says): the problem is that it takes an age to compute, versus seconds to do the other methods. All this talk of the '4:15' report relates to a Gaussian VaR done to give a quick-n-often-a-bit-dirty answer for immediate action. Its limitations are very well-known
Incidentally, nowhere mentioned in the article is the fact that VaR is itself a subset of a more general concept called CaR (C = Capital), which removes one of the most limiting aspects of VaR, namely that it assumes perfect liquidity over the relevant 'holding period' (which itself should only be 1 or 2 weeks at max for the maths to work, but is better when it is properly measured only in days)
CaR assumes an 'infinite' holding period (i.e. doesn't assume you can dump a position quickly if it's moving against you, a fundamental premise of VaR); is more subtle & complex; & can't be done at the press of a button at 4:15 each day, but rather should be chewed over in much the way GS reportedly did in the anecdote.
Overall I would say gaming, and wilful (compulsory!) institutional optimisim, lie at the heart of the 'analytic' problems. I rarely meet a senior risk manager who doesn't have a very shrewd idea of what the real risks are, but they are 'strongly urged' to keep their own counsel: the dealmakers rule the universe and they just don't wanna know: it has negative consequences for dealflow and bonus !
the risks that VaR measured did not include the biggest risk of all: the possibility of a financial meltdown: OK, so "now we know this" (of course we always did), what we gonna do about it ??? Beyond the 95% (or 99%, as will now become the very costly norm) lies (a) diversification and (b) sovereign support. Just as no industrial organisation can economically afford to create a "100%" physically safe workplace, no financial system can economically afford cover against all the ultra-extremes.
You end up taking the 17th / 18th C approach of actually banning all derivatives, which is just childish. This said, there are some extreme instruments that should be effectively banned - by making the conventional / mandatory calculation of the risk involved so costly (in terms of Capital Adequacy) that no-one will enter them
or admit they've entered them ... which brings us back to locking a few more of these bastards up! Pragmatic risk management takes several forms ...
Hi Nick
sorry for taking ages to reply, I don't know what to add though as you obviously know this area a lot better than I do! agree about Taleb becoming a bit silly. I thought Fooled By Randomness was the better of the two books.
Post a Comment