We were reviewing Merrill Lynch’s September Quantitative Review last week. The 65-page book published by Savita Subramanian reviews numerous investment styles’ performance over particular timeframes. Given our interest in investment strategies, their composition and respective performance, it is always of particular interest to us as to the parameters in these underlying strategies. We would like to highlight two particular strategies – not because they stood out in any particular fashion, but rather they were the first two in the book. The two strategies are value strategies based on Earnings Yield. One is based on trailing earnings, and the other on forward estimates.
Merrill takes the top 50 S&P companies ranked by Earnings Yield: Trailing 12-month EPS divided by month-end price. It has been a portfolio they have been actively rebalancing on a monthly basis since December 1988. They benchmark the returns to that of the full S&P 500 to look for outperformance.
With June 1989 as the equal mark, the strategy has outperformed the S&P cumulatively by about double. There are periods of time when value strategies underperform and times of outperformance (particularly look at the post-Internet bubble and post-financial crisis periods).
The next strategy was constructed the same way, but for the fact that they used Forward Earnings Yield as the ranking measure.
The graphs suggest greater volatility with slight cumulative underperformance compared to those of the Trailing ranking mechanism. Given the somewhat unreliability to forward looking estimates in general, and the community’s lack of vision in regards to economic tipping points, that seems intuitively correct. However, in our opinion, more granularity was needed to reach any substantial conclusions.
The above figures are comparative to the S&P as a whole. Merrill does not provide the absolute performance of those individual strategies beyond a five year time horizon.
Presuming these are 2-, 3- and 5-year rolling figures, they would date back no earlier than immediately prior to the Lehman collapse. The Forward Twelve Months (FTM) is handily outperforming the Trailing Twelve Months (TTM) during these periods. Still, we wanted to dig further.
The biggest attribute of The Bloodhound System is the ability to craft and create your own strategies and test them against the most accurate and comprehensive point-in-time database. We didn’t want to limit ourselves to monthly data, but rather build a strategy based on daily information and have it managed that way. Additionally, we didn’t want to limit ourselves to a subset of the S&P 500, but rather a wider universe of public traded companies.
We built two strategies based on the 50 most attractive earnings yields, and expanded the universe of stocks to include all U.S.-domiciled, NYSE and NASDAQ listed equities with stocks above $5 per share and market capitalizations of $250 million or more. Because we wanted to compare trailing twelve month numbers with forward estimates, we also added the limitation of stocks covered in Value Line’s 1700 securities for which they provide forward estimates. In the end, our universe numbered around 1,200 qualifying companies.
Comparatively to the absolute returns presented by Merrill over the last five years, the results are quite similar. They get notably better over the five year period, but some of that may be the smaller cap effect.
Although the two strategies are highly correlated (93%) to each other and both produce significant excess returns, our analysis suggests that using Forward Earnings Yield regularly outperforms Trailing Earnings.
Despite the intuitive leap discussed at the beginning of this post, the volatility of returns was not worse for the projected forward earnings yield compared to the trailing (although both were more volatile than the S&P). Additionally, in most years, the Sharpe and Sortino ratios are greater for Forward than Trailing.
Over the last ten years, the standard deviation of the Forward strategy is 23.3% while the Trailing Earnings strategy produced an almost equal standard deviation of 23.2%. The Forward Earnings strategy yielded significant improvement in Sharpe ratio over the same period (1.33x vs. 1.03x) thanks to an almost 500 basis point average annual outperformance. The Sharpe ratios diverge even further over a 20-year period.