PsyFi Search

Thursday 1 December 2011

Losing Momentum: Is It Time to Exploit Mean Reversion?

One Last Free Lunch

There are a range of so-called anomalies that have preoccupied investors for many years, largely because they seem to offer a free lunch, which is a rare thing in investment markets. So the momentum effect and various value-related effects have spawned a whole host of exciting but not entirely convincing ventures.

A range of recent research now threatens to actually shed some light onto these anomalies suggesting that the momentum effect has vanished and that value effects are real but caused by idiosyncratic factors. It also suggests that mean reversion, upon which many investing careers is based, generally works but sometimes only if you have an investing lifetime to wait. On a positive note, NOW might be a really good time to try and exploit it.

Losing Momentum

The momentum effect was first documented by Jegadeesh and Titman, who established that a strategy of holding a portfolio of winning stocks and shorting losing stocks yielded a positive and abnormal return over timescales of up to a year. This finding generated a whole host of related research dedicated to explaining the effect, most of which was mutually contradictory and which makes you go cross-eyed from reading it.

One of the odd things we’ve noted in the past about this type of anomaly is that actually pointing them out often has the affect of causing them to disappear. The explanation for this, at least, is relatively clear cut: once the effect becomes generally known then investors will move to exploit it, and improve the efficiency of the markets in that respect. We’ve seen a similar occurrence in the Death of the Accrual Anomaly.

This seems to be what’s happened to the momentum effect as Bhattacharya, Kumar and Sonaer document in Momentum Losing Its Momentum: Implications for Market Efficiency. Since 1999, roughly when Jegadeesh and Titman published a follow-up paper showing that the momentum effect was still current, it’s disappeared. M. Scott Wilson in Are Momentum Strategies Still Profitable Work for U.S. Equity? records a similar result, and actually shows that momentum effect returns have been negative in the last five years.

Valuing Ignorance

So if momentum’s stopped working what about its polar opposite, the various value effects? The size effect – where small companies outperform large ones – and the book-to market effect – where companies trading at or below net asset value outperform the market – are well documented. There are, again, a wide range of explanations for this but a lot of them focus on the idea that these stocks outperform because they’re more risky.

In some more recent research Kevin and Marc Aretz manage to narrow down the range of potential causes by focussing on the small subset of firms that actually drive these anomalies, rather than by market wide analysis. What they show is that the main factors behind the size and book-to-value effects appear to be idiosyncratic risk and distress risk. Or, to put this in English, risks that are specific to individual stocks, and not market wide.

What this implies is that a value strategy that looks to generally exploit small firms or those trading at a low book-to-market value that doesn’t involve detailed analysis of the individual companies involved is flawed. Any form of strategic ignorance is based upon ignorance of the strategy: the market is not generally mispricing such firms based on normal risk factors, but is specifically mispricing some firms, which require some hard research to uncover. Which is a positive thing for those investors who spend time digging the dirt on small value stocks.

Mythical Mean Reversion

Of course, the underlying theme behind most value strategies is the concept of mean reversion, as in Value in Mean Reversion?, the idea that over the long-term most company valuations will revert to some norm after they’ve undergone a shock of some kind. Mean reversion is almost revered in some circles and, like the concept of just-enough-diversification we looked at in Diworsification is Good For You, appears to have attained mythical status amongst investors, and is accepted, unquestioned by all.  Which suggests it's exactly the kind of target we should want to throw a few bricks at, just to see what happens.

There are a lot of studies documenting mean reversion effects in markets – including the Jegadeesh and Titman study quoted earlier and the famous and seminal papers by Werner De Bondt and Richard Thaler in 1985 and 1987. Despite these and many subsequent pieces of research the jury is still out, as recorded in Mean Reversion in International Stock Markets: An Empirical Analysis of the 20th Century, a wide ranging paper by Laura Spierdijk, Jacob Bikker and Pieter van den Hook.

It’s well worth a read, but the main finding is that mean reversion does occur. So you can all breathe a deep sigh of relief. Although you should do this very, very slowly, because in the worst case it can take up to 23.8 years for a stock to return to half its previous value, which the researchers call a ‘half-life’. Which is more or less most people’s entire active investing lifetime. Over some time periods there is no mean reversion discernable at all.

At the other extreme a mean reversion half-life can be as short as 2.1 years, and this level of relatively rapid mean reversion seems to be linked to periods of extreme economic uncertainty:
“The highest speed of mean reversion is found for the period including the Great Depression and the start of World War II. Similarly, the early years of the Cold War and the period covering the Oil Crisis of 1973, the Energy Crisis of 1979 and Black Monday in 1987 also show relatively fast mean reversion.”
This finding suggests that relying on mean reversion to work its magic is probably not a generally safe bet. The implication is that when stocks are marked down in normal economic conditions this is the market doing its normal reasonably efficient job of accurately valuing the impact of events; and that the recovery times are unreliable, presumably because they’re dependent on the managers of the firms involved actually managing the problems properly and ensuring recovery.

Waiting For Uncertainty

On the positive side were we to ever find ourselves in a situation of great economic uncertainty, say where the world’s financial institutions had pretty much bankrupted everyone including sovereign nations, and the leaders of those nations were unable to introduce the changes needed due to partisan politics, then we should expect mean reversion to work quite well and quite quickly. This would presumably be because markets weren’t doing their efficient best anymore due to various factors, including a general dearth of anyone with any money to take any risks with.

So just as soon as those conditions come around we should be on the outlook for mean reverting bargains.  Can’t wait. Bound to happen some day.

So there you have it. Momentum is dead and mean reversion is under suspicion, but is probably offering decent returns for anyone with a surfeit of courage and a suitcase stuffed full with cash. Yesterday’s sure-fire winner is tomorrow’s dead-beat loser. That’s the fun thing about research into markets – you’re never quite sure if you’re measuring eternal market fundamentals or transient human behavior. It’s probably best to assume nothing works for ever.


  1. Waiting for uncertainty clearly sums up the current global macro situation.I couldn't agree more . I can not think of another time in history when the global economy has been this close to the edge . We live in interesting and scary times.

  2. "The explanation for this, at least, is relatively clear cut: once the effect becomes generally known then investors will move to exploit it, and improve the efficiency of the markets in that respect..."

    Can you clarify how this works for momentum again? Its actually more clear for mean reversion but limelight would benefit momentum over the short run (as well as make the asset more volatile in the medium run).

  3. Can you clarify how this works for momentum again? Its actually more clear for mean reversion but limelight would benefit momentum over the short run (as well as make the asset more volatile in the medium run).

    It's a general statement about anomalies: if they become well known they get arbitraged away. The point you're making, I think, is that highlighting momentum would tend to have the opposite effect: that investors seeking to benefit from the anomaly would exaggerate it. However, this isn't any different from any other anomaly, the question is simply how smart people figure out how to play the game. So hedge funds will find ways of both gaining profits early and then taking them quickly, the result of which would be the anomaly disappearing. At the simplest level, momentum based stocks are over-valued in the shortish-term so it makes sense to short them with a slightly less short-term option. As funds get ahead of the game and start this process earlier and earlier the anomaly disappears.

    That's the theory, in practice it'll be more difficult, but that's the point, really. The overall effect, of course, is to make markets more efficient ...

  4. Take a look at this for another point of view:

  5. Hi Gary

    Interesting paper, clickable link here: Is Momentum Really Dead?

    Couple of thoughts around this. The argument is that the disappearance of the momentum anomaly is related to its publication causing it to be arbitraged away - which would be consistent with other such findings related to the widespread awareness of such anomalies. The alternative argument is that this is a facet of the peculiar circumstances we've experienced over the past decade. Both arguments are credible and we don't have any way of deciding between those two positions. The "truth" will out only in the longer term, but longtitudinal performance data going back before the initial publication of the anomaly doesn't really have any relevance to the argument. On the other hand the paper quoted in the above article pointing to "momentum crashes " in 1932 and 2009 is interesting, because if your sample includes this kind of one-off event it can seriously skew the results. In addition it's potentially indicative that momentum continued to work for a number of years after it was first identified - although I suppose this could be because no one figured out how to exploit it for a few years: the accrual anomaly continued to work for the best part of a decade in some form.

    However, it's interesting that the authors find that momentum still holds over shorter periods. This would be consistent with an arbitrage policy as smart investors try to get ahead of the game. It's not proof, it's just interesting ...