Can investors earn a reliable premium from stocks with high downside risk? In their January 2012 paper entitle “Sorting Out Downside Beta”, Thierry Post, Pim Van Vliet and Simon Lansdorp measure in four ways (including regular beta) the premium associated with stock sensitivity to market movements. They estimate excess market returns based on total returns of a broad capitalization-weighted U.S. stock market index relative to one-month U.S. Treasury bills. They use rolling historical windows of 60 months to calculate beta and three alternative measures of downside beta. Using monthly total returns and firm characteristics for a broad sample of U.S. common stocks during 1926 through 2010, they find that:
- Excess market returns are negative for about 40% of months during the sample period.
- Correlations for the three measures of downside beta with regular beta range from 0.56 to 0.94.
- The downside (regular) beta premium is about 4.0% to 5.8% (0.60% to 2.4%) per year based on regressions, depending on the method of beta measurement. The downside semi-variance beta (variance only when the market falls) generates the largest premium. In other words, portfolios concentrated in stocks with large downside semi-variances tend to earn significantly positive gross returns.
- Correcting for firm size, book-to-market ratio and one-year momentum (with skip-month) reduces beta premiums by 1.3% to 1.8% per year.
- Correcting for idiosyncratic volatility, skewness, one-month reversal and liquidity has little effect on results.
- Excluding the smallest 20% of market capitalizations and excluding the early subperiod through June 1963 have little effect on results.
In summary, evidence from individual U.S. stocks indicates that downside beta is more effective than regular beta in identifying risk, and the reward that tends to accompany risk.
Cautions regarding findings include:
- Reported returns are gross, not net. Including reasonable trading frictions would reduce beta premiums and perhaps render them unexploitable.
- Investors with relatively small portfolios may not be able to hold enough stocks to capture the reward-for-risk identified by downside beta.
- The 60-month window for beta calculation is arbitrary and may impound data snooping bias.
