September 6, 2013
The Past-Performance-Is-Bunk Warning Isn’t Quite Right
The topic of past performance, and how to think about it, requires some clarification. There are a lot of folks who advise us that we should ignore historical performance. It’s worthless, they argue. Case closed. I understand the motivation to embrace this extreme form of skepticism since it springs from a crucial problem in the wider world. Indeed, many investors look at a hot hand for a mutual fund manager, for instance, and blindly assume that they can easily join the party and reap the rewards. All too often that’s an assumption that’s destined for disappointment. But that’s not the same thing as saying that historical returns provide no value for projecting, analyzing and modeling expected returns.
Mean reversion and momentum, to cut to the chase, offer quite a lot of insight for generating reasonable return forecasts. They’re hardly infallible, but nothing else rises to that standard either. Granted, looking at historical performance by way of these two factors isn’t typical for the average investor, especially in a quantitatively rigorous process. But the literature on both counts is clear: high (low) returns over trailing periods of one-to-five years imply low (high) returns in the future. On a shorter-term basis (typically 12 months or less), positive (negative) returns tend to persist.
Sure, the details matter and so it’s impossible to assume any definitive parameters in modeling these factors. Much depends on the asset class in question and the look-back period under scrutiny, for example. In other words, a fair amount of detail keeps everyone guessing in real time about exactly what reversion to the mean and momentum tell us about expected return.
Sometimes, however, the future is a bit less hazy than usual when you’re looking in the rear-view mirror. The implied return forecast is probably a bit more reliable when trailing performance is relatively extreme—a topic I discussed a few weeks ago. Momentum too makes irregular appearances on these pages, including this post from earlier in the year. Considering the two sides of this historical performance coin together, it’s not too far off to say that positive return momentum comes in waves, eventually giving way to negative momentum, and then reverses.
It’s debatable how, or if, you should incorporate reversion to the mean and momentum into your portfolio management process. Even if you decide that’s these factors are worthwhile, they’re hardly a silver bullet. They don't exist in a vacuum either.. As Antti Ilmanen explains in Expected Returns: An Investor's Guide to Harvesting Market Rewards , there are dozens of useful methodologies for estimating future asset class returns. All come with baggage, of course, albeit for different reasons, and so the future remains uncertain no matter what we do. Still, the hard work of developing expected returns is essential and history provides one facet in our toolbox.
No, we can’t mindlessly assume that a strong return over some recent period will bring more of the same. But be careful about dismissing history entirely. Like everything else in the money game, so-called iron rules that supposedly apply to all portfolios at all times under all conditions are in reality far more nuanced concepts than some folks recognize (or admit).
Posted by jp at September 6, 2013 6:17 AM
Agreed. But it's important to point out that most investors still need to rebalance, even in a world with efficient markets. Even if we're not interested in short term trading, historical return data can still provide useful information for informing a rebalancing strategy in a long-term context, particularly in those times when returns have been extreme. That doesn't mean you must "trade" in a big way or rebalance frequently. On the other hand, the idea that we're simply going to remain oblivious to market activity over, say, 20 years, isn't practical either. The real issue is finding a happy medium that allows for a bit of intelligent short-term risk management without sacrificing too much of the positive long-run expected return of betas. Easier said than done, of course, but a worthwhile pursuit just the same.
Posted by: JP at September 6, 2013 9:38 AM
That is true but to the average investor it doesn't matter. Sure there might be some sliver of useful info in past stock statistics but efficient markets means that only the few top analysts can make use of this information.
Right now, the best analysts are not humans. They are computers that rely on sophisticated algorithms that are able to take into account millions of data points for each trade they make including not just stock data but news, social network data, documents that are not publicly available, reports from people on the field, hidden correlations and indirect information, things like weather etc...
If you don't have that kind of analytical capability and knowledge reach, you are trading with a severe handicap and you will get owned by the machines. Indexing, by getting you mean performance allows you to capture part of the performance of the machines. Otherwise, in the long term, unless you have insider info that the machines don't have, you will do worse than average. Any other outcome is statistically impossible.
Posted by: Benoit Essiambre at September 6, 2013 9:24 AM