Haim Mozes and John Launny Steffens, authors of the study “Getting More Value Out of the Value Factor,” which was published in The Journal of Investing’s Winter 2015 issue, have attempted to create a model that can accurately predict the performance of the value premium.
The factors in their model include analysts’ long-term earnings growth forecasts, overall equity valuations and volatility. They found that value stocks tend to outperform growth stocks when: equity markets are performing well (although this certainly wasn’t the case during the late 1990s); analysts are more optimistic about long-term earnings growth; volatility is low; and when equity valuations are expensive (as they were in March 2000).
None of these findings should come as a surprise, especially if you believe in a risk-based explanation for the value premium. The authors also note that it is difficult to forecast the return to the value premium as well as its correlation to equity returns (which is time-varying and averages slightly above zero).
Following are some of their other findings:
- From 1930 through 1989, the value premium was 0.47% per month, with a much higher standard deviation of 3.72% per month. From 1990 through June 2013, the premium was about half the size of the previous period, at 0.24% per month, with a standard deviation of 3.18% per month. (Note that because returns are roughly linear in time, while the standard deviation increases with the square root of time, volatility will not be as large relative to the premium when looked at on an annual time scale. Thus, the monthly time frame makes volatility look particularly large relative to the size of the premium.)
- The value premium exhibits very high kurtosis (5.13). Kurtosis measures how fat the tails are in the distribution.
- The value premium was positive in just 55% of months. It was greater than 5% in about 6% of months. It was smaller than -1% in close to 30% of months. It was smaller than -5% in about 4% of months.
- Excluding months where the value premium exceeded 5%, the average monthly return turns negative, at -0.16% per month. Thus, more than 100% of the value premium is earned in just 6% of months. On the other hand, if you could have avoided the months when the value premium was worse than -5%, the value premium jumps to 0.7% per month.
You can see how difficult it is to capture the value premium if you are trying to time the premium, or if you aren’t a very disciplined buy-and-hold investor.
Another interesting finding was that, while the correlation of the value premium with the equity premium was positive for each of the three decades of the 1930s, 1940s and 1950s with correlations between 0.70 and 0.29, the correlations were declining. In each decade since then, the correlation has been negative, ranging from between -0.16 and -0.56.
Mozes and Steffens did find that there was some predictive power in past premiums. Correlations, while low, were significant at the 2% level, and predicted extreme future returns.
For example, they found that when the value premium exceeded 7%, the probability that the following month’s premium will be above 5% more than doubles, from 6% to almost 15%. However, the probability of a return of less than -5% also increases significantly. Thus, a highly positive premium really isn’t actionable.
On the other hand, they also found that when the value premium was highly negative, the odds that the next month’s premium would be highly negative increased by about three times, while the odds of a very high premium increased only about 50%. Thus, a highly negative premium might be actionable.
The explanation for this difference in outcomes is that momentum is strongest when liquidity is weakest, and liquidity dries up when volatility increases. Value tends to perform poorly when volatility rises.
Caution, however, is warranted because the risk in taking action on a highly negative premium is that when value stocks bounce back, their recovery can be quite strong. We saw this after the tech bubble in 2000, and again in March 2009, when the markets finally turned. In such periods, beaten-down value stocks recover strongly.
Mozes and Steffens further found that value tends to perform better when stock returns are strong, when corporate earnings growth is expected to be stronger and when equity valuations are high. And as mentioned above, they also found that value tends to perform poorly when volatility increases. All of these are consistent with a risk-based explanation for the value premium.
I’m personally skeptical that investors might be able to improve on performance based on the authors’ findings. And you can take advantage of the information on how value tends to perform when valuations are high, simply by rebalancing your portfolio.
The most compelling finding was that, just as is the case with equities, much of the returns to value stocks occur over a few short and unpredictable periods. For example, the study “Black Swans and Market Timing: How Not to Generate Alpha,” which covers the 107-year period ending 2006, found that the best 100 days (out of more than 29,000) accounted for virtually all (99.7%) of returns.
Another study, covering the period 1926 through 1993, found that just 7% of all months were responsible for basically all of the returns (the other months provided an average return of just 0.01%).
Unfortunately, the vast majority of investors don’t have the patience or discipline to wait for those short bursts of very high returns. They often fall victim to recency bias, allowing more recent returns to dominate their decision-making.
This is why the research shows so many investors in stocks end up with bondlike returns while living with all the volatility of stocks. Thus, they get the worst of both worlds: the lower returns of bonds and the higher volatility of stocks. It’s also why so few investors are able to earn the value premium that the market has provided over the long term.
The unpredictable nature of the value premium, as well as findings that show there are good risk-based explanations, are why we should expect the value premium to persist in the future.
In addition, while there are also good behavioral-based explanations for the value premium (perhaps leading one to think that having now been uncovered, the behavior would be arbitraged away), the research shows there are limits to arbitrage that prevent sophisticated investors from actually fully correcting mispricings.
Larry Swedroe is the director of research for The BAM Alliance, a community of more than 140 independent registered investment advisors throughout the country.