One of the reasons for the failure of risk management is that even though the correct risk metric has been identified, the risk managers mismeasure the risks known to them.
This mismeasurement can happen in many ways. The two broad areas that provide scope for mismeasurement are the use of distribution and the use of historical data.
With respect to the distribution of returns/losses, a couple of things can go wrong.
Distribution specification are wrong. Let’s take an example. Suppose a manager has chosen binomial distribution to measure risk. Even if he is right in choosing this distribution, he might get the probabilities wrong, for example, he might assess a lower probability of loss. Alternatively he might assess the extent of loss incorrectly.
Correlations are mismeasured. The risk management may incorrectly measure the correlations between different positions. So, even if the distribution estimates for specific positions are accurate, the overall distribution will be misspecified.
Wrong distribution is used. The risk management may use a distributionwhich is not suited for their requirement. In our LTCM example, we had assumed that there is a 99% chance of 25% returns and 1% chance of 70% capital loss. In reality it might have been so that there was 1% chance of 70% loss and 9% chance of a 100% loss of capital. This would change the expectations.
Use of Data
Another area that could lead to mismeasurement of known risks is the use of data, especially the historical data. For example, in case of subprime crisis, there was no historical data that could be used to generate the distribution of losses. In the absence of historical data or a good sample, the risk measure will statistical measures which makes it quite subjective.
This article is based on the paper “Risk Management Failures: What are They and When Do They Happen?” by Rene M. Slutz, which is a part of the FRM syllabus.