In part 1 of this article, we looked at the relationship between valuations and future performance. In part 2 of this article, we will continue to explore if timing factor strategies will work.
The scatterplot below illustrates this point for the large-value factor. The X-axis shows the ratio of the valuation multiple of the Russell 1000 Growth Index to the Russell 1000 Value Index. The Y-axis shows the performance differential between the value index and the growth index during the ensuing five-year period. The positive relationship between relative valuations and performance is evident. But the strength of this relationship diminishes significantly after excluding the rolling five-year periods beginning in 1999 through 2000. These are highlighted in red in Exhibit 3. Excluding these periods, the explanatory power of the value and small size factor regressions falls considerably, as shown in Exhibit 4.
.png)
.png)
I also found there was no significant relationship between price/book valuation spreads and the subsequent performance of the minimum variance index, consistent with Research Affiliates’ findings. And while there was a positive relationship between the index’s price/earnings valuation discount to the market’s and its subsequent performance, it was weak, as the regression’s low R-squared value indicates. Similarly, I didn’t find a significant relationship between valuations and long-term performance for the momentum index, likely because of its high turnover.
The results for the quality index were mixed. It exhibited the expected positive relationship between its price/book valuation discount and future performance (though this relationship was weak). But surprisingly, it tended to underperform as it became cheaper relative to the market on price/earnings. This could be a fluke resulting from the fairly short 18-year sample period. Yet it demonstrates that mean-reversion can take a long time, which could make a contrarian factor-timing strategy difficult for many investors to stick with.
Factor Timing
Given the scraps of evidence I’ve uncovered here demonstrating that some factors tend to do better as they become cheaper relative to their history, a disciplined contrarian factor-timing strategy would seem to have the potential to be profitable. Arnott and his colleagues emphatically argue that is in fact the case in their article, “Timing ‘Smart Beta’ Strategies? Of Course! Buy Low, Sell High!”4 The authors set up a strategy that targets the three factor portfolios (out of eight) that were trading at the lowest valuations relative to their historical norm up until that point. They also ran a separate returns based strategy that held the three factor portfolios with the worst average performance during the past one-, three-, five-, and 10-year periods. They found that both contrarian strategies earned higher returns than a simple equal-weighted average of all the factor portfolios, running the simulations from 1977 through August 2016.
To check the robustness of this analysis, I ran a similar contrarian strategy with the factor indexes from my previous analysis, listed in Exhibit 5. I excluded the Russell 2000 Value Index because of its overlap with the Russell 2000 Index. This simulation tracked the performance of the three indexes with the worst returns during the previous five years and held them for a year before refreshing the portfolio. This contrarian strategy beat the equal-weighted factor portfolio by 1.1 percentage points annualized from November 2003 through November 2015, consistent with Research Affiliates’ findings. I did not have enough data to adequately test the valuation strategy.
.png)
Momentum is also a decent candidate to test as an input for timing factor exposures. In their work, the Research Affiliates authors found that chasing factor performance detracted from returns. My own findings were consistent with their conclusion. I ran a simulated strategy that targeted three of the five indexes with the best returns during the past year, refreshed monthly. From November 1999 through August 2016, this strategy trailed the equal-weighted factor portfolio by 33 basis points annually.
Despite the apparent profitability of the contrarian factor-timing strategy, investors would probably be better off sticking with a set factor allocation. The return benefits of factor-timing in a long-only context will likely be modest at best (before transaction costs), and it doesn’t always pay off. The relationship between factor valuations and subsequent performance is weak to moderate and is influenced by extreme events. Investors often have to wait a long time before valuations mean-revert, which can make it challenging to adhere to such a contrarian approach. Finally, factor timing leads to more- concentrated portfolios that can increase risk.
4 Arnott, R., Beck, N., & Kalesnik, V. 2016. “Timing ‘Smart Beta’ Strategies? Of Course! Buy Low, Sell High!” Research Affiliates. https://www.researchaffiliates.com/en_us/publications/articles/541_timing_smart_beta_strategies_of_course_buy_low_sell_high.html
