"What matters in thinking about risk is the quality of the decisions we make in the face of uncertainty."
Peter Bernstein is an elder statesman of finance. He served time in the rough-and-tumble world of portfolio management. He was the first editor of the Journal of Portfolio Management, and his diverse writings have made him somewhat the chronicler of twentieth century finance. He is best known for his book Capital Ideas: The Improbable Origins of Modern Wall Street. It is a wonderful book that introduces readers to modern finance theory. Many books do the same thing through weighty mathematics. Bernstein did it with fascinating historical narratives.
A problem with being a bestselling author is the fact that you have an editor who wants you to do it again. Bernstein, with some coaxing, has published several more books that have never matched the brilliance of Capital Ideas. One of these was Against the Gods: The Remarkable Story of Risk. Tacitly identifying risk with risk measurement, this was largely a non-technical history of the development of probability theory. There are better histories of probability theory, but this didn’t matter. Bernstein was a star, and the book sold healthily.
I am increasingly concerned with how the risk management business today focuses so intently on the tools of risk measurement—probability, normal curve, sampling, regression to the mean, mean-variance. To me, risk management is not about measurement at all. It is about how we make decisions and only incidentally about the math we use in making those decisions. If we stare at just the models and equations, we lose sight of the mystery of life—we lose sight of the unknown. There would be no such thing as risk if everything were known. If only a finite number of things could happen, risk would not exist. Even the most brilliant mathematical genius will never be able to tell us what the future holds. What matters in thinking about risk is the quality of the decisions we make in the face of uncertainty.
I have long held a similar view. In my 2004 paper Defining Risk, I concluded that it is impossible to define risk. We speak intuitively of risk, but there is no “true risk.” At best, we can define perceived risk—a far less compelling notion. Risk metrics such as volatility, delta or value-at-risk, are really metrics of perceived risk. In my paper, I elaborated:
So why discuss Bernstein now? He has a new book out called Capital Ideas Evolving. This is a follow-up to Capital Ideas. Picking up where the earlier book leaves off, it focuses on subsequent developments in both finance theory and practice. Bernstein’s new thinking on risk is front-and-center. Indeed, I now perceive his comments at the 2006 symposium as just a preview.
If he thinks there would be no risk if there were only a finite number of outcomes, he’s obviously never played poker or craps. I don’t think I agree with him about the math. I’ve certainly met lots of people who think they understand what is going on simply by looking at equations, when they don’t have a clue. But I think the math gives us more information than Bernstein lets on. Truth lies somewhere between these two methods. The search for the proportion is the key to understanding.
I am trying to figure out Bernstein’s comment “If only a finite number of things could happen, risk would not exist.” At first I thought he might have meant that there was no risk because we could build a replicating portfolio to hedge against all the possibilities, but that can’t be right. Just because there are finitely many possible outcomes doesn’t mean we know what they are. Also, the market may not be complete, so replicating portfolios won’t always be constructable. I think he misspoke.
It is bordering on crazy to review a book based upon one paragraph. Taken in context he may be referring to the different definitions of risk and narrowly defining it his own way. For example, the great philosopher Vice President Dick Cheney (while paraphrasing another) said that the risk is not of what we know, but of what we don’t know that we don’t know. I assumed the view from Bernstein’s paragraph followed that kind of buildup.
Agreed — my comment was just a thought on it, not a critique of Bernstein or his work.
Isn’t Bernstein’s point that in a deterministic world, where all causation linkages would be known, there would be no risk? This goes back to Glyn’s point (in his paper) about whether you believe that risk is an objective or subjective notion. Bernstein seems to be taking the subjectivist view: probabilities do not exist, they are just a reflection of our ignorance. “We may regard the present state of the universe as the effect of its past and the cause of its future. An intellect which at a certain moment would know all forces that set nature in motion, and all positions of all items of which nature is composed, if this intellect were also vast enough to submit these data to analysis, it would embrace in a single formula the movements of the greatest bodies of the universe and those of the tiniest atom; for such an intellect nothing would be uncertain and the future just like the past would be present before its eyes.” (Laplace 1809) or course determinism itself is a belief that one does not have to have.
What I found appealing about Bernstein’s comment was the last sentence: “What matters in thinking about risk is the quality of the decisions we make in the face of uncertainty.” This parallels so closely my own comment from two years earlier: “A more practical question is whether a risk metric is useful. Used in a given application, will it promote behavior that management considers desirable?” I think neither comment discredits the value of risk measurement. They recharacterize that value. So often, questions come up about whether a risk metric accurately captures risk (note the whole debate about coherent risk metrics) or whether a risk measure is accurate (the whole business of backtesting). The two quotes suggest that what really matters is whether the risk metric and/or measure contribute to sound decision making. Bernstein and I offer sightly different criteria. He calls for “quality decisions.” I think that is a difficult criteria to test for. I call for the slightly less relevant criteria of whether use of a particular risk metric (or risk measure or other tool of risk assessment) promotes “behavior that management considers desirable.” That is more practical and certainly easier to test for.
I think making good decisions is a desirable, if obvious goal. But we have to get there somehow. Isn’t implementing good (accurate, appropriate, whatever) risk measures the way to achieve that?
My point is that “accurate” is not meaningful. At best, a risk measure can address perceived risk. How should we act knowing there are possible outcomes we cannot perceive — and hence cannot measure? Risk measurement is not — cannot– be the answer to this particular question. That doesn’t mean it is useless. It is very good for helping us address perceived risks, which are very real.