Does the Fama-French five-factor model of stock returns (employing market, size, book-to-market, investment and profitability factors) explain the outperformance of low-volatility stocks. In their July 2016 paper entitled “The Profitability of Low Volatility”, David Blitz and Milan Vidojevic examine whether: (1) any of several models expose a conventional return-for-risk market beta effect for stocks; and, (2) the low-volatility effect is distinct from a low-beta effect. They calculate volatilities for stocks and the market using daily or monthly returns over the past year. They calculate stock betas using these volatilities and daily or monthly stock-versus-market return correlations over the past five years, with shrinkage by 1/3 toward a value of one. They include momentum (return from 12 months ago to one month ago) as an explanatory factor, even though the five-factor model does not. Using data for a broad sample of U.S. common stocks and model factors (excluding extreme outliers) during July 1963 through December 2015, they find that:
- Based on monthly regressions between a stock’s beta and its next-month return in excess of the risk-free rate over the entire sample period:
- The Capital Asset Pricing Model (CAPM) does not hold. There is no relationship between a stock’s beta and its future return.
- Controlling for size, value and momentum factors, there is still no relationship between a stock’s beta and its future return.
- Further controlling for profitability and investment factors, there is still no relationship between a stock’s beta and its future return.
- All intercepts are significantly positive, suggesting that every stock earns the equity risk premium regardless of its beta (over long periods).
- Findings are robust to a different way of calculating beta and for two size groups.
- Based on a second round of regressions between a stock’s beta or volatility and its next-month return in excess of the risk-free rate and its beta-predicted return:
- Contrary to CAPM, there is a negative relationship between beta and future return.
- Controlling for other factors, the low-volatility effect dominates the low-beta effect.
In summary, evidence indicates that the five-factor model does not explain the low-risk effect, and this effect is stronger when measuring risk via volatility rather than market beta.
Cautions regarding findings include:
- While findings may help an investor decide whether and how to exploit the low-risk effect for stocks, they do not incorporate tests of any realistic strategies.
- The volatility estimation interval (past year) and the correlation estimate interval used in beta calculation (five years) are different, raising the possibility that different look-back intervals introduce differences in volatility and beta regressions. However, one of the authors reports that findings are the same when the intervals match.
See also:
- “Low Risk and High Return?”
- “Systematic Overpricing of High-beta Assets?”
- “Pervasive Outperformance of Low-volatility Stocks”
- “Global Low-volatility Stock Portfolio Performance Predictors”
- “Achieving a Low-volatility Stock Portfolio Efficiently”
Frequently Asked Questions
What is the low-volatility effect in stock investing?
The low-volatility effect is the tendency of stocks with lower return volatility to earn higher risk-adjusted (or even higher absolute) returns than conventional risk-return theory predicts. In the Blitz and Vidojevic study of U.S. common stocks from July 1963 through December 2015, volatility was measured from daily or monthly returns over the past year, and the effect persisted after controlling for the Fama-French five-factor model.
Does the Fama-French five-factor model explain the outperformance of low-volatility stocks?
The Fama-French five-factor model does not explain the outperformance of low-volatility stocks. In monthly regressions on U.S. stocks from July 1963 through December 2015, the low-risk effect remained after controlling for market, size, value, profitability and investment factors (plus momentum), and the effect was stronger when risk was measured by volatility rather than beta.
How are stock volatility and beta calculated when testing the low-volatility effect?
In the Blitz and Vidojevic study, stock and market volatility are calculated from daily or monthly returns over the past year. Beta is calculated from those volatilities and the stock-versus-market return correlation over the past five years, with shrinkage of 1/3 toward a value of one. The study then runs monthly regressions of next-month excess returns on beta or volatility, with controls for size, value, momentum, profitability and investment.
Is the low-volatility effect the same as the low-beta effect?
The low-volatility effect is not the same as the low-beta effect; in this study the low-volatility effect dominates the low-beta effect after controlling for other factors. Regressions on U.S. stocks from July 1963 through December 2015 found no relationship between beta and future return, contrary to the Capital Asset Pricing Model (CAPM), while a negative relationship appeared in a second round of regressions. The low-risk effect was stronger when risk was measured by volatility rather than market beta.
What are the limitations of research on the low-volatility effect?
The Blitz and Vidojevic findings do not incorporate tests of any realistic trading strategies, so they cannot show how well an investor could exploit the low-risk effect in practice. Also, volatility is estimated over the past year while the correlation used for beta spans five years, which could introduce differences between volatility and beta results. However, one of the authors reports the findings are the same when the intervals match.