« Risks in the Rupee Dollar Rate |
Main
| Subprime Woes- VaR Models in need of Generational shift? »
August 13, 2007
Black Swan and the Bell Curve- An Emerging Market Perspective
Nassim Taleb's book Black Swan is a powerful one. The book is thought provoking as it throws a serious challege to the whole science and art of prediction, looking into the kernel of uncertainty in human existence.
I choose to discuss a single idea from Taleb - the Normal/Gaussian distribution. Even though its decent performance in normal times, it fails to deliver good results in situations with extreme variations. He devotes a complete chapter on " The Bell Curve, that great intellectual fraud" followed by two more discussing similar ideas. His key argument is that statistics that bases analysis on the centrality of data around the average and too few outliers leads to predictions out of touch with reality. To put it mildly, Taleb is not impressed with the estimates of risk analysts precisely due to their love for the bell shaped curve. Surprisingly, he did not mention Value at Risk models in the discussion, and I have no clue as to whether his opinion changed somewhat a decade after the famous Jorion- Taleb debate on VaR.
I read the chapter, however, through the lens of VaR models. Ever since the adption by BIS for market risk measurement in 1996, the central banks throughout the word has propagated this model and the model has been used by majority of financial market practitioners. Tremendous refinements have taken place, including the mandatory Stress testing practice, precisely to catch the black swans. It would be interesting if Taleb agrees to enter into another round of debate with Jorion on this. I, for one, will eagerly await the prospect!
Now coming to real life problems in emerging markets:
My more immediate problem is that in emerging markets, the bell curve not only fails us in abnormal times, it fails us in normal times as well. This part of the world is full of financial markets which are in transition, not deep enough and rarely demonstrate the beauty and certainty promised by the "bell curve". The 'bell curve" specialists from the Western markets routinely fail to produce good models of risk in India, China and others. As a result, even though tried out by MNC banks, risk modeling could not be successfully outsourced to the Tokyo or NY desk. such attempts, which were quite visible in late 1990s and early part of the present decade , slowly died their natural death in competition. The fact remains that the risk manager will have to predict good models that withstands backtesting.The struggle and the search for better models continues and escape routes are being developed from the intellectual dominance and practical failure of the normal distribution in such markets.
To get a glimpse of the ongoing efforts, I give links to few of my papers. I am sure there are many others who face similar problems with the Gaussian distribution and links to the works or comments will be very useful.
Fat tails in Indian Debt Market
http://www.gloriamundi.org/detailpopup.asp?ID=453057606
VaR measure- Comparison of methods and how to overcome non-normality
http://www.gloriamundi.org/detailpopup.asp?ID=453057546
On incorporating Liquidity Risk in Emeging Market VaR estimation
http://www.gloriamundi.org/detailpopup.asp?ID=453057507
Posted by sunandoroy at August 13, 2007 05:57 AM
no too au fait with the original Taleb-Jorion debate but it seems (from the paper referred in the below link which was published in May 2007) that Taleb has not changed his view:
http://www.lse.ac.uk/resources/riskAndRegulationMagazine/magazine/summer2007/epistemologyAndRiskManagement.htm
Posted by: david papert at August 17, 2007 01:01 PM
Post a comment
