The Independent Market Observer | Outlook. Opinion. Insight.

11/25/13 – What Does “Risk” Mean, and How Can We Match It to Clients?

Written by Brad McMillan, CFA®, CFP® | Nov 25, 2013 1:51:57 PM

A great comment came in recently from Commonwealth advisor John Smallwood. He writes: “My problem is that it is difficult to match a client’s risk profile with a model, since modern portfolio theory has been shown to be invalid and cannot measure risk, as we saw in 2008.”

I actually think there are at least two issues there, along with an assumption, so let’s unpack this a bit. To start, the fundamental question here seems to be “What is risk?” If we haven’t properly defined risk, at least in our own minds, then the rest of the investment process becomes guesswork.

The conventional definition of risk, that of variability of returns, has reigned more or less unchallenged for some time. Never mind that it doesn’t reflect how real people think of risk. Never mind that it’s not a good reflection of the ultimate goals of investment. The important thing is that it’s mathematically tractable, and theories based on it can be developed thoroughly.

This mathematical and unrealistic treatment of risk reflects a deeper flaw in modern portfolio theory (MPT). It is not that it failed in 2008; it is that it’s based on a series of unrealistic assumptions, which are known to be so, and that those assumptions necessarily result in a gap between market reality and portfolio performance. Even worse, since those assumptions don’t reflect how people really think, any attempt to match client risk tolerance with MPT-based models is bound to show some gaps.

The core idea behind MPT remains valid—that a truly diversified portfolio will show smoother performance, with less variability, than a nondiversified portfolio. Note, however, that it doesn’t say such a portfolio will actually make money. You can have a portfolio that loses money every year, and as long as it shows low variability—that is, it loses money on a consistent basis—MPT will be satisfied. MPT also doesn’t deal with extreme cases, or point events, but with trends over time. It simply is not possible to evaluate risk in an MPT sense over a period of less than, say, three to five years.

With this understanding, John, MPT did not fail in 2008; it simply wasn’t applicable. Looking at it five years later, MPT actually did pretty well over the appropriate time frame. Which brings us to your real point: If MPT works over longer time frames, what the heck happened from 2004 to 2009? Again, MPT is not guaranteed to prevent loss, and as long as a diversified portfolio had less variability than a nondiversified one—even when they both got hammered—then MPT holds.

Risk, therefore, in an MPT sense of variability of returns, is not the answer to the question you’re posing. I would argue that risk has to be redefined in order to provide a meaningful answer to that question. Clients don’t mind positive returns with high variability; they mind losing money. That gives us the key to a better definition of risk for investor purposes, which is “How much money can I lose in this portfolio without bailing out?”

This idea is behind the risk management tool known as “value at risk,” or VaR, which is widely used in the professional investment management industry. Using probabilistic models, it attempts to quantify how much will be lost over a given time frame. The problem is that, in extreme situations, this doesn’t work either. It can serve as a guide, but it’s no panacea. What it can do is provide guidance as to how much is at risk, and how closely that risk should be managed.

Getting back to John’s question, of how we match a client’s risk tolerance to the appropriate model, I would argue that we first have to better understand what “risk” means to the client—and here I most definitely include the client we see in the mirror every morning. For most people, it means the risk of loss. How much loss is he or she willing to tolerate? Regardless of the long-term trend, if people pull out because the loss exceeds their limits, any model will fail. This is another way of looking at the failure of MPT models in 2008.

The fact of the matter is no probabilistic model can limit the risk of loss through portfolio construction. If we define risk as the risk of capital loss at a point in time, then portfolio construction simply cannot guarantee that. The only thing that can make something close to a guarantee is a commitment to actively manage the risk—either by simply exiting the market and going to cash when risk levels appear high or when we get close to our loss limit, or using other strategies such as buying puts—but these have costs in the form of foregone gains and taxes. There really is no free lunch.

None of this is to minimize the role of diversification. Diversification is a necessary component of any portfolio, but it has to be real and not assumed, and it will not be enough to manage the real risks, as perceived by clients. This also happens to be a particularly timely question, as I’m working on a book proposal centered around constructing portfolios with minimum risk, and have been spending a great deal of time thinking about it.

The takeaway here is that diversification under MPT can provide reduced volatility over time and potentially mitigate point losses, but it is neither designed to nor intended to guarantee a profit or limit losses. Other techniques, which cost, are necessary for that, and managing expectations to account for this has to be a critical part of any investment process. To understand how much investors can risk is to ask how much they are willing to lose—and managing that risk is making sure they don’t lose it.