http://www.investmentnews.com/article/20120429/REG/304299962
I’ve done some work on these are can verify that the strategy tends to do very well relative to more naive strategies.
One of the critiques is that turnover will be high. In my experience, an optimization that chooses the portfolio that maximizes a utility function, like mean-0.5*lambda*variance, will tend to have much higher turnover than a constant risk strategy.
Agreed, it’s very useful.
The largest issue, imo, is that it’s a less efficient use of capital. By that I mean if you have two portfolios with the same sharpe the vol-targeting one will likely be in cash more often. Not the end of the world, but can limit absolute upside potential.
I’m not sure I follow the argument completely. Let’s say you have cash and a risky asset. Your mean-variance frontier is a line with an intercept through the risk-free rate and a slope equal to the Sharpe ratio of the risky asset. The volatility-targeting portfolio basically would be one place. The Sharpe ratio of every portfolio on that line would be the same (except when volatility equals zero since it is undefined).
Are you talking more about the dynamics? Like, for instance, if the risky portfolio’s volatility increases or decreases?
Needless to say, there are a lot of variations you could take on volatility targeting. For instance, if you switch to CVaR targeting, then you would presumably incorporate the different expected returns of different asset classes. Alternately, there is also the question of how to determine the volatility target. If your target was the volatility of the risky asset in the example above, then you would just always invest in that, so that isn’t very helpful. If I’m not mistaken, a guy who works at the Georgetown investment office fit an autoregressive model to his forward-looking CVaR estimates and used the long-run estimate of that as his. Either that’s what he did, or he did something similar and I thought this was better. Or some other form of time variation of the volatility target.
Yea you’re right - not quite sure how to articulate what I’m trying to get at (probably because it’s wrong, lol).
I’ve always liked the idea of targeting volatility. If you are comforatble expressing your risk tolerance in terms of volatility, then targeting volatility does a good job of maximizing your return for the amount of volatility you can stand, assuming that you’ve optimized your sharpe ratio (or other measure) appropriately.
Two issues put flies in the ointment, though. The first is that transaction costs may eat away at the benefits somewhat, because you may be rebalancing a lot more.
But the main challenge I see is that volatility can get very low just before a crisis, so you increase your exposure with falling volatility, and “BAM and then it hits you.” Your portfolio splattered all over the walls of the Brooklyn Museum…
That concern may not be quite so bad, however, if your rebalancing period is quick enough. It does seem that volatility often tends to tick up before a crash truly takes place.
I agree with some of that.
Volatility doesn’t always pick up before the crash takes place in a simple sort of manner. For instance, before the 2000s crash, volatility had been elevated for years and before the 87 crash volatility had been sort of mean-reverting (low the summer before the crash, but much stronger in the weeks leading up to it).
From a Garch perspective, if you’re in a period of low volatility, then projected volatility should be reverting to the mean and be expected to increase. Similarly periods of high volatility would likely be followed by lower volatility.
Nevertheless, turning points are critically important in these type of strategies. It’s not just switching from low volatility to high volatility, which would cuase you to be in too risky a portfolio, but also high volatility to low volatility. In that case, you’re in a conservative portfolio, when you should be willing to get more aggressive. In other words, you would be missing out on the upside.
I’ve taken a look at this SSgA fund in the past:
http://www.bloomberg.com/quote/SSGMVEI:LX
Max drawdown in Q3 2011 was about half that of MSCI World which is pretty good although ytd in 2012 it’s well behind the index so this fund seems to be acting like a low beta play. Concept described here:
http://www.ssga.com/library/capb/A_Case_for_Mgd_Volatility_Eq_Strategies_5.6.11CCRI1305725358.pdf
Yes, I agree with all of that. I guess a properly modeled GARCH process will help; it’s just that one shouldn’t expect too much out of it, because in a true crisis you will bolt way past the mean anyway. But if GARCH adds value, it should still be a little better than just a constant volatility model.
I think that’s an interesting link (the second, I mean). I have only looked at target volatility mainly in terms of broad asset allocation decisions, rather than at the individual equity level. They don’t really get into the nitty gritty, but I like research on low volatility and idiosyncratic volatility puzzles.