Thoughts on the merits of selling winners and buying losers.
I was surprised to see Michael Edesess’ recent article, “Does Rebalancing Really Pay Off?” which questioned the value of William Bernstein’s oft-cited “rebalancing bonus”—the gain that accrues to investors from periodically selling their winning assets and putting the proceeds into their losers. As neither author is prone to mistakes and rebalancing is a straightforward topic, I didn’t see how there could be a serious disagreement.
There isn’t. Both parties agree on the salient point: Rebalancing works best if the expected returns on asset classes are similar. As they diverge, so do the benefits of rebalancing.
Consider a portfolio that consists of two assets, stocks and bonds. Assume that stocks and bonds each will gain 6% annualized over the next 10 years. There is no uncertainty about this projection; the two assets will finish the decade in an exact tie. After the first measurement period (one day, one week, one month, one year, whatever), bonds have gained 2% and stocks have lost 2%. At that point, rebalancing by selling stocks to buy bonds must be correct. Bonds are currently ahead of stocks, but the two assets will eventually finish at the same place. Thus, the future return on stocks is higher than that of bonds.
The better transaction, of course, would be to liquidate all bonds and hold only stocks. Which illustrates the artificiality of this example— the very reason that rebalancing exists is because investors hold multiple assets, and the very reason that they hold multiple assets is because there are no performance guarantees.
Nevertheless, the principle is instructive. We don’t know an asset’s true expected returns, but we can observe its realized risks, and we can reason that assets that are similarly risky should have roughly similar expected returns. Thus, when sifting through assets that have fairly equal levels of risk and return, the odds favor selling the winners to buy the losers. One asset might temporarily look to be higher-returning than another, but the results likely will converge over time.
Implicitly, I’ve described mean reversion. Which is indeed what researchers find when documenting the behavior of financial markets. Although winners often remain winners over the short term, meaning for a few weeks or months, they tend to slide back when the time period extends past one year. Meanwhile, losers rebound. Intermediate-term mean reversion was initially documented for U.S. stocks over five-year time periods by Werner De Bondt and Richard Thaler in 1985 and has since been expanded to include asset classes as well as individual securities and to use a variety of time horizons.
Thus, it’s sensible to rebalance assets that have similar risk levels. One example is different segments of the U.S. stock market, such as value versus growth. Another would be between the stocks of different geographic regions, for example, the United States, Europe, and Asia.
I would extend that advice to include developed- markets versus emerging-markets exposure. While emerging-markets stocks are certainly a riskier bunch, and thus should have superior long-term returns, it’s not clear to me that the gap between the two groups is large enough to eliminate the benefits of rebalancing.