Wed, 4 Sep 2013
Mean reversion is a compelling concept in hindsight, but not as straightforward as a stand-alone signal in the here and now, says Morningstar's John Rekenthaler.
Jason Stipp: I'm Jason Stipp for Morningstar. Reversion to mean is a powerfully compelling concept, especially as some long-term metrics, such as the Shiller P/E, look to be above long-term averages, but there's more to reversion to the mean than it appears. Here to talk about that is John Rekenthaler. He is vice president of research for Morningstar and Morningstar.com columnist.
Thanks for joining me, John.
John Rekenthaler: Yes, Jason, glad to be here.
Stipp: The idea of reversion to mean is talked about a lot by a lot of investing greats, including [GMO's Jeremy] Grantham and many others, and folks are looking at certain long-term metrics such as Shiller P/E, which looks elevated right now and they're saying that we're probably in for a stock market correction because this powerful concept we've seen throughout history. You say reversion to the mean is compelling, but it's also susceptible to some biases. Can you explain that?
Rekenthaler: Yes, I can. Let me step back for a second because you talked about the Shiller P/E ratio, and there is an ongoing discussion that it's happening right now between Shiller and Jeremy Siegel of [the Wharton School], who often find themselves on the opposite end of a debate because Jeremy tends to be relatively bullish on stocks. He is the author of Stocks for the Long Run, after all, 20 years ago, and the Shiller ratio for the last 20 years, which looks at sort of cyclically adjusted or 10-year average price/earnings ratios, for the last 20 years has given off a relatively bearish signal.
The discussion of mean reversion is a little bit broader than that. That's a debate that's going on right now, and you can pick up The Financial Times this week and The Wall Street Journal last month and see Shiller versus Siegel: Who is right? But I want to step back and just think a little bit more broadly about the issue of mean reversion because what's the point of using these ratios, such as the Shiller ratio? There is a notion of mean reversion, when you get above some sort of trend line with the ratio, that's supposed to indicate that the market is too expensive, and when you get below, that indicates that it's cheaper and maybe you should buy.
I read an interesting paper recently by three London Business School professors. They tried to look at this concept of mean reversion more broadly. The basic point of the argument is, it looks very convincing when you go backwards and you fit a line to the data that you have. You have a full set of data and you look and say, "Well, in 1992, with this set of data that goes past 1992, here's what I would have done." It doesn't seem to work so well in practice.
It looks crystal clear right now; it looks crystal clear going backward. It doesn't seem to work so well in 1992 when you didn't have the benefit of the next 21 years' worth of data to help you create that mean line or that trend line. It's a pretty thorough work. The professors look at 20 different stock markets going back as far as they can, and the U.S. has the longest track records--113 years, a long time. But the other markets are multiple decades. The [professors] found it's very difficult to implement mean-reversion strategies in practice, that they're not able in general to find statistically significant results in improvement of portfolios--basically better returns on portfolios--by moving from stocks to cash or cash back to stocks based on market valuations.
Stipp: One example you pointed to is the Shiller P/E at one point in the 1990s looked like it was higher than it had been just about anytime historically, but that we saw stocks continue to go up for several more years after that signal would have said to you maybe at the time that it's time to get out of stocks.
Rekenthaler: Again, I'm not pointing a finger specifically at the Shiller P/E. The authors look at Shiller P/E, but they also look at price/dividend ratio. Even a noncyclically adjusted P/E ratio often tends to tell you the same thing. Market valuation, some measures are a little better than others, but they tend to move together. But in 1995, for the specific example of Shiller P/E, it moved past any level that it had ever been except for mid-1987 and 1929. If there were ever a sell signal to go higher than the two levels that had been previous famous stock market crashes, that was it. You would have been out of stocks, and you wouldn't have gotten back in until the end of 2008, early 2009. You would have been out for about 14 years because even after the market fell in the early 2000s that level remained historically high.
Well, it was a volatile period, but over that 14-year period stocks made about 5.75% per year, quite a bit better than cash. Now, you can argue with risk-adjusted return maybe you are better off being out of the market for 14 years because you had the 2000 technology decline and the 2008 crash. But actually you made less money, and this is the big signal.
Stipp: Is part of the difficulty, in being able to use reversion to the mean precisely, is part of the problem there because the mean also changes over time?
Rekenthaler: The problem is the mean changes over time, yes, and it's also the problem with any market-timing strategies. It's still a general guideline. You get [former Fed chairman] Alan Greenspan in 1996 famously, I think he had a little reversion to the mean in his mind when he used the term "irrational exuberance." That was in early 1996. There was a lot of money to be made from when Alan Greenspan said the markets were perhaps irrationally exuberant in 1996 over the next four years to 2000. There is a problem of getting out too early and just the standard issues associated with timing of the market that are not just with mean reversion, but with other signals, as well.
Stipp: Practically speaking, it seems like reversion to the mean is something that could be useful in a toolkit versus something that could be useful as gospel as always paying attention to this and only this. How would you suggest the investors think about reversion to the mean as they're looking at different metrics today and how they look compared with history and getting signals from that, but what does that mean for their portfolios? It sounds like if I just use that, I could miss out on certain continued gains, for a while anyway. So, how do I think about it?
Rekenthaler: I think it's best as a measure of extremity or how extreme the market conditions might be. There is a little hindsight because I am doing the same thing you always do: You go back and look at the data, the whole series of data that are available now. But when the signals do seem to be valuable, you're getting a result that is much, much lower than it has been in decades or much higher than it's been in decades. For example, in the late 1990s, you might have--using the signal--been out of the market too early. But there was something valuable to know at the time when people were saying this is a new era, things have changed, markets have changed, conditions have changed, to know that we were just far above any place that we've been before on valuations.
It was a check on enthusiasm of the time and a very valuable check. It wasn't really as helpful unfortunately coming after the 2008 crash because even then the ratio was only about middle of normal levels. I would have said you were OK to be in stocks, but it didn't give any kind of indication that if you get in now, in March 2009, you are going to be part of a historic market rally.
But actually I'd hope we don't reach a point where the cyclically adjusted P/E ratios are so low as to be something of a screaming buy, and, say, no matter how scared you are of stocks, you should get back in because that means we've been through one heck of a downturn. But at least on the reverse side, if you get an extreme number, I'd be concerned. We're not there now. We're on the higher end of the long long-term range, but really the middle of a range over the last 20 years or so. As a stock investor, I'd prefer to see the number lower. I am very valuation-conscious. The better prices in the security that you get in, the more chance you have to make money on it, but it's not a number that's telling me to take the extreme step of cutting back on stock exposure and moving into where you can move, such as cash, bonds, much lower-yielding investments.
Stipp: It's not just that in isolation; it's also what are your other options?
Rekenthaler: It's also what are your other options, right. If somebody could convince me there's another option that I should be looking at, that's fairly reasonably risky and has higher return potential than cash or high-quality bonds, yes, it's a signal I pay attention to and I think about. But it's not first or second or third on the list. Not at these levels, not where it's sending off. There is no red light flashing.
Stipp: All right, John. Thanks for giving us some levelheaded advice on how to think about this important concept of reversion to the mean and maybe use it with a little bit more utility. Thanks for joining me today.
Rekenthaler: Jason. Thank you.
Stipp: For Morningstar, I am Jason Stipp, Thanks for watching.