Articles like this one in Investment News just confuse me. Apparently the latest trend among pension plan sponsors is to target volatility. I guess it is a human desire to eliminate volatility, but at the end of the day, you have to pay your pension benefits from your returns. Not risk-adjusted returns. Not volatility or standard deviation. Focusing primarily on volatility is completely missing the boat. From the article:
“There’s a big shift in terms of how plan sponsors are defining risk,” said Michael Thomas, chief investment officer for the institutional business in the Americas at Russell. “During the last 10 years, our industry has developed an unhealthy obsession with tracking error, but managing tracking error isn’t managing risk.”
He’s right—tracking error is not the same thing as risk. Nor is volatility the same thing as risk, I might add. Volatility management is just another unhealthy obsession. Besides, the source of all of the evil volatility is readily apparent.
So far, most of the target volatility asset allocation strategies focus on equity exposure, which is, “by far, the biggest contributor of [portfolio] volatility,” Russell’s Mr. Thomas said.
Equity exposure = volatility. To reduce it, just add some Treasury bills or bonds to the portfolio. Duh. That seems like a simpler solution if you really are concerned about reducing volatility.
I don’t think that investors are going to be any more successful targeting volatility than they are trying to target returns. We have no idea year to year what returns are going to be, even though we know exactly what they have been historically. We can’t forecast it or target it-we just put up with whatever returns we get. I don’t think volatility is going to be any more tractable.
Mike, I’m really surprised by your aversion to active volatility management. We obviously agree with the dominance of momentum as a long-term return factor, but it has been evident for some time that absolute and risk-adjusted returns to almost any portfolio can be improved via volatility management. For example, I would be keenly interested in your response to the following study, summarized here by CXO: Avoiding Momentum Strategy Crashes .
Further, in the pension and retirement space volatility management is even more critical due to the actuarial sensitivity of sustainable funding rates to portfolio volatility and negative skew. Volatility management improves the Sharpe ratio in every portfolio we’ve tested, but perhaps more importantly it normalizes the return distribution, making actuarial funding estimates more robust.
I have the deepest respect for your work, but your apparent bias against volatility management is confusing. I look forward to your thoughts.
I’m averse to the belief that volatility can be easily forecast and targeted in a simple fashion. Too many things can go wrong, much as they did for the portfolio insurance advocates in 1987. Various return factors have inherent levels of volatility and mucking with that usually ends badly, and suddenly.
Tigers are wild animals, even if they are “tamed.” Siegfried and Roy are a good case study.
Return factors exist in the wild too. We may think we have tamed them, but we often discover, too late, that we really haven’t. Let’s just say I have a healthy respect for history.
I’m not averse to managing portfolio volatility. Diversification, for example, is a good way to do that. Adding Treasury bills or cash to reduce equity exposure is another way to reduce volatility, one that I mentioned in the post.
(I’m familiar with the Barroso and Santa-Clara paper you referenced. In that paper, they used a long-short portfolio and targeted a constant 12% volatility by continuously adjusting the long-short ratio. Most retail investors and many retirement accounts cannot or will not use shorts. In addition, the transaction and tax costs of continuous adjustment are ignored in the paper. And why 12%? The authors pulled it out of the air. I don’t know if targeting 8% or 16% would have performed in the same way.)
Constant volatility portfolios are nothing new. If the volatility is held very low, the returns are usually low. Then you are taking equity risk for a very low payoff-and hoping nothing goes wrong with your forecast.
Some other ways of managing volatility without changing the inherent volatility in the factor exposure: 1) diversification 2) cash or 3) combining with other factors where the excess returns are negatively correlated.
The only way to truly eliminate the volatility is to actually eliminate the exposure. Otherwise, you’ve just changed the form of the risk.
At least that’s my take on the subject. I could be wrong and I’m always open to persuasion by persuasive evidence. I really appreciate that you read the blog and are willing to make thoughtful comments! Thanks.
Hi Mike,
Thanks for your thoughtful reply.
At root, I guess I’m surprised by your belief that returns can be forecast (via momentum), but volatility and correlation forecasts are elusive.
We just published an article on Adaptive Asset Allocation which demonstrates the effectiveness of integrating return (momentum), volatility and correlation forecasts across 10 major global asset classes into an integrated portfolio management framework. By the way, we used very standard parameters for the article - there are _much_ better ways to estimate volatility and correlation than 60 or 120 day rolling averages, and which don’t use complex algos like GARCH.
We are currently working on a piece that researches the relative ‘forecastability’ of each of the parameters - I will send it along to you. However, Eric Falkenstein did some interesting preliminary analysis at his site a while back on volatility forecasting:
http://falkenblog.blogspot.ca/2012/02/market-timing-rule-that-works.html
This and other research I’ve seen suggests that volatility may be more easily forecast than returns (though momentum based return forecasts are not dealt with in Falkenstein’s short article).
Lots to think about, and thanks again for all your excellent work.
I’m not sure I believe anything can be forecast! I don’t have much faith in forecasting generally. I certainly don’t think returns can be forecast. Relative strength is a return factor, like value, and by exposing a portfolio to it, you can reap whatever is available. I just don’t think anyone can say with any certainty what that return is likely to be any given year.
Regarding volatility forecasting specifically, I would note that VAR is based on a volatility forecast. If it were really that simple to do, Long Term Capital Management probably wouldn’t have hit the ground at terminal velocity, nor would JP Morgan have blown away a few billion dollars.
Traditional asset allocation requires returns, standard deviations, and correlations. All of these change over time, but returns change the most-and mean variance optimization is most sensitive to returns.
I’m looking forward with optimism to your forthcoming paper researching the relative forecastability of those factors. It could be a big contribution to the literature.