Tuesday, January 25, 2011

How many risk measures are enough?

I was teaching our Fundamentals of Performance Measurement course yesterday and came upon a metaphor to use to justify the need for multiple views on risk: John Godfrey Saxe's poem about the blind men and the elephant. You are no doubt familiar with the story, how the blind men go to the elephant, each approaching it from a different part. One thinks that it's like a wall while another thinks it's like a snake. Since none are able to see the "big picture," their observations are very limited.

Well, to me, risk is like that elephant, and if we only use one or two measures, we only see a limited amount of what lies before us. The more measures, the better perspectives.

I love Yale Endowment Fund CIO David Swensen's statement, from his book Pioneering Portfolio Management, that “Quantitative measures of risk for individual portfolios leave much to be desired." Yes, I suspect that most people would agree that they are limited and have a much to be desired. Our only choice is to employ multiple, though limited, measures. A basket that includes Sharpe ratio, M-squared, tracking error, information ratio, value at risk, liquidity risk, extreme risk analysis, and, if you really want to, standard deviation, is a good start, I believe.


  1. Dave,

    I love it - the comparison is very apt.

    And I totally agree with the sentiment too – it really bugs me when people ask questions such as “What’s the best risk measure?” or “If you had to chose, which risk measure would you use?” I would, however, make 2 comments (you know me!): Firstly, I believe what you have is an analogy, rather than a metaphor (a great one definitely!). Also, I would add 1 word to your recommendation (as some people – other than your good self – so often ignore), I would say that we should employ multiple, appropriate measures. You very neatly allude to the limitations of standard deviation and that, clearly, is one measure which is not appropriate to many of today’s funds and can, of course, be misleading.

    Always a pleasure to read your notes

  2. Anthony,thanks for taking the time to chime in. In spite of its weaknesses, standard deviation remains, and the GIPS EC's decision to make it a mandatory disclosure only adds credibility to it (as I mentioned in another post). Oh, well...we'll get there eventually :-)

  3. "Risk" is a complex issue because it is neither fully objective nor fully subjective: risk depends on both characteristics of the assets involved and investor risk preferences.
    Obviously, a "universal" risk measure cannot exist in such a setting. But the "best" risk measure is the one providing the best fit with both asset and investor characteristics.
    Since a couple of years, there exist risk measures which capture very well both empirical characteristics of real-world assets (like fat tails, skewness etc.) and realistic investor preferences (relative thinking, loss aversion etc.)

  4. Andreas, thanks for sharing your thoughts. The subjective part is not easily (or possibly?) assessable, thus leaving us with objective. While there may be a "best" measure, I would think that it would still exist alongside of several others, in order to provide a broad perspective of what risks exist.

  5. Dave, the subjective part can be modelled well since Kahneman/Tversky, I am not concerned about this. The real issues in my opinion are related to risk governance and taking actions based on risk information: "Risk managers produce numbers which tell people who don’t want to know what they could lose if what they don’t believe can happen does happen."

  6. Andreas, thanks for your additional insights and thoughts.

  7. I love this analogy/metaphor? (who cares) - but it is a perfect illustration of my frustration with Financial Risk "Management".
    The REAL question is: WHY should you WANT to measure and/or model risk?
    The answer is: To MANAGE (i.e. CONTROL) the risk!
    Blind men simply cannot CONTROL an elephant.
    But little boys, all over India, have been sitting astride elephants for literally thousands of years, and getting them to do work that humans could not manage!
    Personally I remain frustrated with your Financial Risk "Management", and will stick to "Project Risk Management" where we manage individual risks - quite specifically - one at a time.
    Ned Robins.

  8. Dave
    Great point oft overlooked. Thanks to you I have a mental picture of a fractal elephant! If I may introduce a slightly different - complexity - slant on the same issue in this blog item.

    Eric Berlow: How complexity leads to simplicity


  9. Ned, interesting points. Yes, "project risk management" is much easier to handle and manage than "financial risk management." But the reality is that investors still want to see risk measures, thus the need to be aware that one or two alone simply won't do it.


Note: Only a member of this blog may post a comment.