Markets Insight.
.
May 28, 2012 12:28 pm
‘Whale’ makes a big splash on risk models
.
By Pablo Triana
The whole world has by now heard of the London whale, widely dubbed Voldemort. Operating from JPMorgan Chase’s imperial tower in Canary Wharf, traders in the bank’s Chief Investment Office have managed to suffer a $2bn mark-to-market loss on what appear to be credit-related positions apparently linked to a bunch of US corporate names.
No journalist has shied away from the news, no blogger has refrained from commenting. Rival financiers have expressed morbid curiosity, regulators have started a probe, JPMorgan executives have been replaced. The implications for the banking industry could be severe, given how JPMorgan had been portrayed as the salutary exception in an ocean of malfunctioning and troublesome institutions.
This incident will add fuel to the perennial debates on traders’ pay and supervision, the policing of the proprietary activities of banks, and the dangers of credit default swaps – themes that are now almost as familiar to a general audience as to the expert elite.
But perhaps the biggest takeaway from the fall of the whale lies in a much more mysterious, more technical and decidedly much less general matter: that of value at risk, the hitherto glorified mathematical model used to estimate likely maximum portfolio losses based on statistical analysis of historical price trends and movements.
The JPMorgan case has served to highlight, once more, the inherent deficiencies of the model and in a particularly pugnacious fashion. Coming on the heels of a post-credit crisis anti-VaR barrage, culminating in the recent revolutionary decision by the Basel Committee on Banking Supervision to stop relying on the model for bank capital regulation purposes, the latest mishap may deliver the final blow to what for 20 years was most probably the most powerful metric ever in financial history. VaR was the tool that could dictate how much risk and how much leverage banks could take on; in essence, how likely it was for the banking industry to go down. The first two weeks of May 2012 may be forever remembered as the time when VaR was killed. The erstwhile unthinkable is now not only imaginable but almost certainly true.
Voldemort’s losses have provided a global reminder of how insultingly unreliable VaR can be and why the model should never have been given any significant power. It’s not only that the setbacks “predicted” by VaR turned out to be infinitesimal with regards to the actual ones, but that the calculation methodology itself had to be restated.
The VaR at JPMorgan’s CIO was one number using certain computations and it was a completely different one (twice as high actually) using other computations. This goes to the core of VaR’s flakiness. It shows how easily manipulable a thing like VaR is. The perceived risk of a firm should not depend on whether one technicality is used instead of another when designing a quantitative machine.
Or on the amount of historical data used for the calculation (VaR can swing wildly simply because you selected three rather than 10 years of past data in your Excel spreadsheet; how far down you arbitrarily drag your computer mouse can effectively dictate how risky a bank looks and how much leverage it can enjoy).
A group’s risk should, in a rather more sophisticated approach, depend on fundamental common sense analysis of what is, and is not acceptable, so as to be able to discriminately ban the latter from the premises. VaR, by being so blind to true risk, can enthusiastically welcome the unacceptable, as it did in the run-up to the summer of 2007 when it permitted leverage of 100-to-1 and even 1,000-to-1 on massive trading books populated by massively toxic stuff.
You know the world’s financial system is built on shaky foundations when bank risk depends on the dictates of a flawed model whose results can be later restated. Why trust a model that leads to utterly lethal outcomes and whose mathematical insides can be modified on a whim, controlled by and known only to those with a clear vested interest in reporting one type of number over another? Regulators should at least pretend to be preventing banks from unfettered manipulation of their risk numbers.
This is all highly ironic, for no other group did more to promote VaR in the crucial early days than JPMorgan. The very public 1994 distribution of its VaR methodology and knowledge was key in converting the industry and, in particular, the regulators to the new mathematical religion. In puzzlingly circular fashion, events within JPMorgan could now help bury VaR for good. That whale has certainly made a splash.
.
Pablo Triana is a professor at ESADE Business School and author of “The Number That Killed Us: A Story of Modern Banking, Flawed Mathematics, and a Big Financial Crisis”
.
No journalist has shied away from the news, no blogger has refrained from commenting. Rival financiers have expressed morbid curiosity, regulators have started a probe, JPMorgan executives have been replaced. The implications for the banking industry could be severe, given how JPMorgan had been portrayed as the salutary exception in an ocean of malfunctioning and troublesome institutions.
But perhaps the biggest takeaway from the fall of the whale lies in a much more mysterious, more technical and decidedly much less general matter: that of value at risk, the hitherto glorified mathematical model used to estimate likely maximum portfolio losses based on statistical analysis of historical price trends and movements.
The JPMorgan case has served to highlight, once more, the inherent deficiencies of the model and in a particularly pugnacious fashion. Coming on the heels of a post-credit crisis anti-VaR barrage, culminating in the recent revolutionary decision by the Basel Committee on Banking Supervision to stop relying on the model for bank capital regulation purposes, the latest mishap may deliver the final blow to what for 20 years was most probably the most powerful metric ever in financial history. VaR was the tool that could dictate how much risk and how much leverage banks could take on; in essence, how likely it was for the banking industry to go down. The first two weeks of May 2012 may be forever remembered as the time when VaR was killed. The erstwhile unthinkable is now not only imaginable but almost certainly true.
Voldemort’s losses have provided a global reminder of how insultingly unreliable VaR can be and why the model should never have been given any significant power. It’s not only that the setbacks “predicted” by VaR turned out to be infinitesimal with regards to the actual ones, but that the calculation methodology itself had to be restated.
The VaR at JPMorgan’s CIO was one number using certain computations and it was a completely different one (twice as high actually) using other computations. This goes to the core of VaR’s flakiness. It shows how easily manipulable a thing like VaR is. The perceived risk of a firm should not depend on whether one technicality is used instead of another when designing a quantitative machine.
Or on the amount of historical data used for the calculation (VaR can swing wildly simply because you selected three rather than 10 years of past data in your Excel spreadsheet; how far down you arbitrarily drag your computer mouse can effectively dictate how risky a bank looks and how much leverage it can enjoy).
A group’s risk should, in a rather more sophisticated approach, depend on fundamental common sense analysis of what is, and is not acceptable, so as to be able to discriminately ban the latter from the premises. VaR, by being so blind to true risk, can enthusiastically welcome the unacceptable, as it did in the run-up to the summer of 2007 when it permitted leverage of 100-to-1 and even 1,000-to-1 on massive trading books populated by massively toxic stuff.
You know the world’s financial system is built on shaky foundations when bank risk depends on the dictates of a flawed model whose results can be later restated. Why trust a model that leads to utterly lethal outcomes and whose mathematical insides can be modified on a whim, controlled by and known only to those with a clear vested interest in reporting one type of number over another? Regulators should at least pretend to be preventing banks from unfettered manipulation of their risk numbers.
This is all highly ironic, for no other group did more to promote VaR in the crucial early days than JPMorgan. The very public 1994 distribution of its VaR methodology and knowledge was key in converting the industry and, in particular, the regulators to the new mathematical religion. In puzzlingly circular fashion, events within JPMorgan could now help bury VaR for good. That whale has certainly made a splash.
.
Pablo Triana is a professor at ESADE Business School and author of “The Number That Killed Us: A Story of Modern Banking, Flawed Mathematics, and a Big Financial Crisis”
.
Copyright The Financial Times Limited 2012.
0 comments:
Publicar un comentario