VAR goes BYE

[rffrydr]

"Value at Risk" modeling is casualty of credit crunch. How say you, Bill?


http://www.bloomberg.com/apps/news?pid=20601109&sid=axo1oswvqx4s&refer=home
_________________
Today is the Tomorrow you worried about Yesterday!

[rffrydr]

Speaking of "normal VAR".... The golden era's killer app:

http://www.investopedia.com/articles/07/sharpe_ratio.asp#axzz1mx713V5i
_________________
Today is the Tomorrow you worried about Yesterday!

[rffrydr]

Clive Capital embracing "5 sigma deviation"....all blow'd up:

http://www.cnbc.com/id/42951408

[nodoodahs]

[rffrydr]

Forget "Black Swans," we're back to Chaos theory:

Front-month Brent crude futures sank more than $12 per barrel (well over 9 percent) in a series of vertiginous declines that took the market down from over $120 to under $110.

The price change was more than 4 standard deviations — which is something that should be seen on average only once in every 63 years — if the market was well-behaved and changes followed a normal or Gaussian distribution.

At times the move has approached 5 standard deviations — which should only occur once every 7,000 years.

So it may be time to throw overboard all those elegant value-at-risk (VaR) models or at least admit they mislead more than they reveal. It also spells near-fatal trouble for many option-pricing models.


Some analysts and risk professionals have tried to rescue VaR and option models by tweaking them to include fat tails (“kurtosis”) that give more weight to extreme outcomes. But no amount of tweaking can cope with the frequency of 4 and 5 standard deviation moves exhibited in the oil market over the last two decades. It is more realistic to accept that price movements are not really normally distributed at all.

Such large price movements confirm the brilliant mathematician Benoit Mandelbrot was right in arguing modern financial theory was founded on shaky foundations and massively underestimated the real amount of risk lurking in markets.

Crucially, Mandelbrot discovered the average amount of volatility in prices was not constant but varied over time. He noted that markets appeared to move through distinct phases of low and high volatility. Days with unusually big price moves tended to cluster together, and so did days with unusually small variability. He described markets shifting from a mild state to a turbulent one and back again.
NH
Mandelbrot’s description of price movements in commodity markets is much more realistic than the conventional theories that populate textbooks and lurk behind much academic research — which should come as no surprise because he originally discovered the non-normality of price movements in a long time series of cotton prices.


The second myth that deserves a decent burial is that commodity prices are driven almost exclusively by fundamental forces of supply and demand and are unaffected by the build up and liquidation of speculative positions.

This argument has never seemed very plausible. But its adherents have clung on with surprising tenacity. It remains the official line for much of the research establishment as well as regulators, who continue to argue price movements can be largely or entirely explained by fundamental factors and that there is “no evidence” that speculation or investment flows have an impact.

It would be hard to find ANY fundamental factor that could explain how Brent crude prices could be worth $120 at the start of Thursday but less than $110 under twelve hours later.
_________________
Today is the Tomorrow you worried about Yesterday!

[rffrydr]

Great combination here: the first quant, Ed Thorp and Scott Patterson, much in the news with his (rightly) quant-bashing book:

http://www.npr.org/templates/rundowns/rundown.php?prgId=13&prgDate=2-1-2010

Interesting that Thorp is a paranoid and never had a loosing year--never trusted his models. --But, put it all in other quants hands in the end. Suffered the unkindest cut along with the rest of us.
_________________
Today is the Tomorrow you worried about Yesterday!

[nodoodahs]

There is a "portfolio modified VaR" that takes into account the higher moments.

http://www.andreassteiner.net/performanceanalysis/?External_Performance_Analysis:Risk-Adjusted_Performance_Measures:Modified_Sharpe_Ratio

I have no doubt it's better than a measure that assumes normality, but I'm also sure it's an approximation; volatility has a whiplash smile.

As an aside, most T-bond classes (think IEI, holding a maturity range) have monthly and annual vol with positive skew, as opposed to equities with their negative skew.
_________________
I haven’t seen a beatin’ like that since somebody stuck a banana in my pants and turned a monkey loose.

[nodoodahs]

[rffrydr]

BIS rejigger:

http://ftalphaville.ft.com/blog/2009/10/20/78736/would-you-like-a-massage-for-your-stressed-var/

Of couse the "tail" was primarily a "rated" surprise--is this not, in the end, an across-the-board rerating of the ratings agencies?
_________________
Today is the Tomorrow you worried about Yesterday!

[rffrydr]

What you can't count is built into the numbers:

http://online.wsj.com/article/SB124940692698405243.html
_________________
Today is the Tomorrow you worried about Yesterday!

[rffrydr]

Stress Tests only make for more stress:

http://ftalphaville.ft.com/blog/2009/02/17/52569/stress-tests-debunked/
_________________
Today is the Tomorrow you worried about Yesterday!

[rffrydr]

Oops. Taleb exaggerates to make a point. I don't feel at all beholden to the equations. If I was an investment bank I don't think I could help it.

Yes, under the "Goldman" post you'll see by the graphic that they are the real traders of the bunch, holding positions longer and deeper. They've got themselves into a pickle with the huge ABX profit marked but un-coverable.

Rumoured to be GS heavy on the bid for MBIA stock; as well as scouring for subprime portfolios it can buy up on the cheap and thus eventually realize what has already been booked.

I kind of like betting MER against GS here.
_________________
Today is the Tomorrow you worried about Yesterday!

[nodoodahs]

I deleted the duplicate posts.
_________________
I haven’t seen a beatin’ like that since somebody stuck a banana in my pants and turned a monkey loose.

[rffrydr]

Taleb interview on the limits of math in the markets--or what we know about the tails (nothing):

http://media.bloomberg.com/bb/avfile/BBRECON/v6qwNDHZOHK4.mp3
_________________
Today is the Tomorrow you worried about Yesterday!

[nodoodahs]

VAR is SOL? Good riddance to bad rubbish.

Apologies if you've heard this from me before.

"Assuming normality means never having to say you don’t have enough data."

The real truth of market return distributions is somewhere closer to Chebyshev than most people think. Take your “normal” estimate of a three-plus sigma event, times 3 or 4; take your “normal” estimate of a less-than-half sigma event, times 1.5. A 95% confidence interval based on the wrong distribution is a vote of no confidence. Using a timeframe that excludes outliers just exacerbates the problem.

http://www.billakanodoodahs.com/2006/11/on-golden-swans/

I think of Taleb as “Captain Obvious.” Unfortunately too much mathematical work is based on standardized distributions while ignoring the inconvenient non-robustness of the underlying assumptions.