Excerpt from GARP.org
Overcoming the limitations of risk models with forward-looking observations
By David X Martin and Yan Shi
[Editor’s note: For an earlier treatment of this subject, see “Radar as Risk Management Paradigm”.]
Radar is an excellent example of a modern, forward-looking risk detection system – that is, a system built to detect emerging risks – and financial risk managers might greatly improve their results by co-opting some of the principles by which it operates. That said, radar uses historical performance not unlike the brain uses memory – to predict outcomes in the future – and with about the same accuracy. Put another way, radar readings are susceptible to the same major risks that plague any predictive model.
The first is unknown risks, for which radar has no past data. Say, for instance, you are using radar to track an approaching plane, and the plane is shot out of the sky by a missile. Nothing of the plane’s flight to that point can prepare you for this event.
The second is the risk of environmental change, e.g., the plane loses altitude and the radar readings are obscured by a mountain range. In this case, the earlier data lose almost all of their ability to predict the plane’s final flight path.
Lastly, there are the built-in risks of linear interpolation, i.e., the most recent readings are given greater weight in the formula.
Modern financial risk managers are familiar with these limitations, and they have developed various approaches to address these issues.