Mean reversion
Momentum bets that a move will continue. Mean reversion bets that it will reverse: what was extremely strong or weak recently tends to return to its average. This shows up over very short and very long periods, hardly in between.
Mean reversion is the bet on the opposite of momentum: that a security that has moved far from its usual level tends to go back there. What has fallen a lot, you buy. What has risen a lot, you sell.
The long term version became known through the economists Werner De Bondt and Richard Thaler. In 1985 they showed that the stocks that had done worst over three to five years clearly beat the former leaders over the following three to five years.
There's also a very short term version: stocks that fall sharply on one day often recover a little in the following days. Because you have to trade very often for this, trading costs eat up a particularly large amount here.
At first glance it seems contradictory: over three to twelve months momentum tends to hold, over very short and very long periods mean reversion. But the two patterns don't exclude each other. They simply apply to different periods.
De Bondt and Thaler (1985) sorted stocks by how they had done over the previous three to five years and formed a portfolio of losers and one of winners. Over the following three to five years, the losers' portfolio was ahead by several dozen percentage points in total. The reversal was unevenly spread: it was much stronger for the losers than for the winners.
For the short term version, Nagel (2012) shows that individual stocks tend to move against themselves from day to day and month to month, their correlation with themselves is negative. This gets stronger when market makers charge more for providing liquidity because funds are tight. So this version is often seen as payment for making trading possible for others, less as an investor thinking error.
There are two explanations for the long term effect. De Bondt and Thaler think investors overreact and correct that later. The other view says past losers are often struggling companies with higher risk, and that brings more return. The dispute isn't settled. What matters in practice: the short term version loses much of its force after realistic trading costs in several studies. The long term version needs far less trading and is therefore less vulnerable to costs.
Data as text
Over days to weeks, mean reversion dominates. Over 3 to 12 months, momentum dominates. Over 3 to 5 years, mean reversion dominates again.
Summary
- Mean reversion bets on a return to the average, momentum on continuation.
- The effect shows up over very short and very long periods, hardly over medium ones.
- The short term version is especially vulnerable to costs, because you have to trade very often.
Did you get it?
What did De Bondt and Thaler find in 1985?
That the stocks that did worst over three to five years clearly beat the former leaders over the following three to five years.
How can the short-term reversal effect be alternatively explained?
As payment for providing liquidity, when market makers charge more for it because funds are tight.
On which time scales do momentum and mean reversion seem to contradict each other without actually doing so?
Momentum tends to hold over three to twelve months, mean reversion over very short and very long periods.
Sources and further reading
- De Bondt, W. F. M. and Thaler, R. H. (1985), Does the Stock Market Overreact?, Journal of Finance
- De Bondt, W. F. M. and Thaler, R. H. (1989), Anomalies: A Mean-Reverting Walk Down Wall Street, Journal of Economic Perspectives View source ↗
- Nagel, S. (2012), Evaporating Liquidity, Review of Financial Studies
Related
- Momentum and trend followingStrategy Types
- Why active funds usually underperformStage 1
- Grid-trading and DCA botsStrategy Types