Published in Euromoney’s Corporate Finance 1996 by Ralph McKay & T Erle Keefer
A means of representing risk as a single statistic is very appealing- and highly flawed, argue Ralph McKay and T Erle Keefer of consultancy Risk Technology. The value-at-risk method can conceal large risk holes, is insensitive to the possibility of catastrophic market events and ignores risk reshaping difficulties.
Value-at-risk and its new incarnations represent a form of risk appeasement. It ignores the real market risk problems. It is fashioned to impart a feeling of security to the people that it is supposed to give comfort to – central bankers, directors of banks and financial institutions, regulators and politicians. Objective scrutiny exposes this as unwarranted.
Given the notorious billion-dollar derivatives losses and absence of senior supervision, the clamour for a simplified risk measure such as value-at-risk is understandable. And the arrival of value-at-risk has served positively to focus attention on the need for risk measurement.
What is wrong with value-at-risk? Popular, broadly accepted definitions eschew too much essential information. They represent market risks as a single statistic. This single statistic is appealing and easy to explain. Unfortunately, it is even easier to use it to hide large risk holes. Managers consequently oblivious to the actual risk can increase systemic risk of derivatives bungles.
The manager presented with a value-at-risk number is like a fruit and vegetable trader who has a basket of fruit to sell but knows only the weight of the basket. He, along with the rest of the market-place, does not know from this single parameter whether the basket contains potatoes, mangoes, a mixed bag or bad apples. Thus he cannot determine what price to ask – he cannot assess the risk. To the fruit and veg trader, weight is not a market risk measure. Similarly, on a detailed analysis conventional values-at-risk are barely worthy of being treated as a measure of market risk.
Beware Tunguska Events
Even as a simple measure of risk, value-at-risk is highly exposed to criticism. Values-at-risk assume well-behaved markets but ignore the impact of real worry, the financial markets’ Tunguska events , such as the stock market crash of October 1987.
Derivative markets are very young. There is little data from which to predict the frequency of market crashes – that is the size and frequency of so-called outliers. But the limited data that exists is alarming and the only prudent approach is to assume that the big crashes have not happened yet. A prudent risk measurement must highlight exposure to the totally unexpected – because, in truth, it should be expected.
The value-at-risk concept of measuring risk in terms of a 99% probability of market movements is dangerous and misleading. For example, portfolios with deep out-of-the-money short option positions frequently show little sensitivity to the 99% movement. However, for even larger movements they may incur accelerating losses. Values-at-risk can be blind to this form of market jump risk.
Value-at-risk relies heavily on past statistics as a guide to the future. Another example of this is the actual or implied use of correlations between different markets and risk diversification. Negative correlations produce cancelling risks in the value-at—risk. Most of the time, diversification helps to minimize losses. However, when the security of diversification distracts attention from absolute protection risk management practice – as value-at-risk appears to be doing now its effect can be destructive in the medium to long term. This is because correlation and diversification assumptions can prove inadequate when the risk Tunguskas hit.
An absolute risk measure seeks to protect wealth in a defined manner regardless of the probability of adverse market movement. For example, the maximum wealth loss may be limited to an acceptable level. Compared with the value-at-risk 99% probability-type protection, outcomes between OK and destruction may be different when a Tunguska hits.
Risk reshaping
Quality risk management is only half the story; risk reshaping is its complement. For example, two portfolios may have identical value-at-risk numbers. However, the risk may be eliminated easily in one of the portfolios, whereas the other may be difficult to modify because of the bumps-to-the-risk-carpet problem. That is, attempts to control one aspect of the market risk may worsen other aspects.
The value-at-risk statistic ignores these risk-reshaping difficulties. Frequently, such complexities of risk reshaping may be the real risk problem, even though value-at-risk numbers may appear benign. Value-at-risk hides all information required to assess reshaping risk. Using the value-at-risk, it is impossible to determine which transactions are required to reshape market risks.
A single risk number ignores market asymmetry. A portfolio may be more exposed to loss than profit, but this cannot be known from the value-at-risk number. Its limitations extend to overall trading philosophy.
Value-at-risk tends to give validity to the popular notion that financial institutions make money by incurring market risks. The more risk, the higher the potential returns. This is a myth. The target value—at-risk number should be zero. All risks should be fully hedged — particularly in zero-sum markets. The objective should be to eliminate market risk, leaving only counterparty and operational risks to worry about.
This is the case when a bank sees itself as a manufacturer of financial instruments. Such a manufacturer may not be able to achieve the ideal zero-risk position at all times, but how many senior bankers hold the misguided belief that a zero value-at-risk means zero profit?
The underlying reality is that market risk management is a complex problem. It is dangerous to oversimplify the issues. An effective set of risk numbers should indicate the present value change in the portfolio in response to all economic parameters affecting the portfolio, whether correlated on not, and indicate exposure to both small and and very large moves. If the portfolio is properly hedged, the exposure to a very large market jump may be the same as the exposure to smaller moves — zero.■
The ‘Tunguska event was a large mysterious aerial explosion in Siberia at 7.40am on June 30 1908. It devastated 500,000 acres. It is thought to have been caused by a comet. June 30 remains the financial day of reckoning for many companies.
Published version – Euromoney’s Corporate Finance – 1996
Ralph McKay 