« Rocky Road to Economic Capital Modeling |
Main
| Risk Management and Economic Development »
September 16, 2008
The Forgotten Tail of Risk Management: lessons from the financial turmoil of 2007-?
it is evident that Risk Management will no longer be the same once the financial markets risk above the debris of the credit crisis of 2007- ?. It also appears that risk regulation will also undergo a major change if not a complete overhaul. the press conference statement by Paulson a few hours ago reflecting weaknesses in regulation reaffirmed the understanding that lot of soul searching is going on in the regulatory community.In this turbulent time, we are faced with many questions. Why were so many brilliant guys sleeping when the problem was piling up day by day? Why did tons of statistics churned out by agencies, literally on a tick-by-tick basis, failed to capture the early warning signals? Or is it that the sane voices got submerged in what Greenspan termed " creative destruction brought about by financial innovation"?
In this milieu, time has come to question or at least reexamine one accepted focus of risk management, the concept of downside risk. The reliance on the left tail showed things to be in great shape, when looked at through the prism of risk models. In reality, they were not. And when risk models deviate from reality, it is time to rethink the core assumptions. Time to dwell upon whether the left tail contains adequate information to control the risks and whether the downside risk numbers tells the story early enough to react in time.
Is it time for financial market participants to view risks from an alternate perspective?
An alternative, to my mind, and a useful one, would be to use information from both the tails of a distribution of a financial variable. A very standard approach to risk management that has cemented its place in risk management is to look at downside risk. Losses matter, profits don't, seems to be the key messege of risk management. The tail on the left hand side of the distribution is far more relevant to the right tail, that is the enveloping idea. Time is ripe to challange this notion.
Although not ex-post, Risk managers raise an alarm one the threshold toward the left tail is breached. Sometimes such an alarm is too weak and too late in the day to save the firm from the brink of disaster. If you notice, the subprime crisis of 2007 was preceded by a heavenly cocktail of low interest rates and low default rates, an ideal situation for unbridled credit expansion. The key point is, the downside risk was low, but some of the numbers clearly breached the upper tail, and went unnoticed. Had an alarm was raised just because of the fact that things look too good to be true, perhaps financial markets could have done some adjustments and restructuring ahead of the adverse shock brought about by erosion of portfolio value. Unfortunately, risk models were ill prepared to pick up the signals of destruction contained in opulence.
One way risk models could have done better was to use both the upper tail and the lower tail to evaluate risks. A foray into the upper tail would have given important information to risk managers about behaviour of the financial indicators of market, credit and other forms of risk. Had the model capacity was well developed, then early signals could have been traced in the upper tail. While we all know about bubbles ( and subsequent tumbles), risk modeling was deficient in factoring the upside risk in.
In retrospect, the information loss arising from ignoring the upper tail is something risk models must look at. The key messege of financial crisis of 2007 is that risk modeling is in need for a level shift. We can feel the need , but can't exactly articulate. My view is the risk models will be more useful once they accept the assumption that the upper tail is no less important that the lower tail. And when they become aware of the fact that the legacy of the good old Mean is as relevant as the heroics of the Standard Deviation!
Posted by sunandoroy at September 16, 2008 01:09 AM
As I saw the markets tumble along with Lehman and Merryl, I could not help myself but to think about that fool who said back in 1997 (this is 10 years ago) that our attempt to measure risk was intrinsically flaw! These "smart" CEO and risk managers had 10 years to consider the upper tail and they did not. I am not going to blame the risk models or any statistics since we, humans, built them based on our conception of our world of possibilities. These models like any mathematical model gave us what we wanted to see based on our assumptions. We wanted to believe that the future could be predicted from the past, that globalization, the enfant of the free market, was not going to increase correlation and make it more difficult for us to see and assess dependence, that we had became too smart to beat the economic cycle (A. Greenspan even envisionned endless economic growth and expansion). I mean, we are talking about the best and brightest here, forget the risk models, the mean or the standard deviation. Wall Street should become more humble after this crisis and these CEO or risk takers should be fired! Not because it would make the taxpayers or the investors or even Lehman employees feel better or relieve BUT because this is the ONLY way they will learn that risk taking is not about the certainty of risk.
As an novice in the risk field, I consider myself lucky to witness these events unfold because as an engineer by training, I understand now that managing risk is not about calculus or statistics, it is about understanding the world we live in and not our own fantasy world. Maybe these Lehman and Merryl were simply part of another world and sorry for the people who would wake tomorrow morning and face the reality of the another world: the VERY and PAINTFULL one.
Posted by: Yannick at September 16, 2008 04:36 AM
You mentioned about the ton of statistics churned out by agencies failed to capture the warning signals. This is not surprising. Statistics, in essence, are produced to support or confirm initial perceptions or hypothesis. This is our natural bias. We have a natural predisposition to look for observations to confirm our own view. Statistics are just tools to achieve this. And there are so many ways to confirm our view based on past or historical observations. However, as this sub-prime crisis shows us, it only takes one event to prove our view wrong.
What risk managers should instead do is to try to look for information and/or observations that would prove their initial view(s) wrong. Risk managers should have more discipline and put more attention to that negative skew. Only then will we be able to see warning signals as they arise (instead of being busy looking for past observations that everything is alright using bell-curve analysis).
Posted by: Kerry at September 17, 2008 04:01 AM
The heck man. You are correct all the way but I dont think any concrete lesson will be drawn from this because we live in market economy were making quick dollar is the name of the game. The worst is often we all jump in the wagon and follow those leaders with little or no objection. The only time we open our eyes to real world is when we fill threaten. For all what is happening out there we are all responsible. I have a question for all outrage people out there: who will accept to invest in a fund that wont perform? As long as capitalism exist in its current form, we will see crisis after crisis.
Posted by: olivier at September 17, 2008 05:08 PM
Simple - over reliance on models. We forget that models rely on data, and if the data input is flawed, then the model is flawed. we can talk about frequency and severity of loss until we are blue in the face, but the simplest probelm is one of data. The industry thought that it could model for default under the assumption that the underlying asset for the most part is liquid and will never decrease in value more than my liquidation costs. All the rocket scientists in the world never assumed that the "housing bubble" was just that - a bubble, and when bubbles burst, chaos is created.
Posted by: Al at October 10, 2008 08:56 PM
Post a comment
