Purifying Value

• Unless controlled, value portfolios usually have exposure to style factors other than value.
• Over the past decade, unintended exposures to nonvalue factors were a contributor to many value portfolios’ underperformance.
• Applying constraints to limit the nonvalue factor exposures in a value portfolio to be nonnegative, would have improved performance over the last decade.

We’ve previously demonstrated that the value factor’s underwhelming performance over the past 10+ years contributed to the underperformance of value portfolios, but that there were two other culprits: stock-specific returns and unintended exposure to nonvalue factors.

In this blog post, we focus on the impact of that unintended exposure. Using the MSCI World Index as the starting universe, we simulated a value portfolio that aims for investability, diversification and maximum exposure to the value factors, book-to-price ratio and earnings yield.¹

Our hypothetical portfolio had significant exposure to the value factors from 2011 to 2020, as seen in the exhibit below. Average exposure to book-to-price ratio and earnings yield were over 0.7 and 0.8, respectively, highlighting the effectiveness of the portfolio in achieving its main objective.

We discussed how large value exposures translated into returns earlier in this blog series. What about nonvalue factors? Exposures to such factors in our portfolio were not significant and mostly below 0.25, as our methodology ensures. Some of these small exposures, however, had significant negative impact on return, as with momentum, for example. Like value itself, momentum is a premia factor, and exposure to such factors has historically resulted in outperformance. To eliminate the negative impact of these factors on the portfolio’s performance, we need to eliminate or reduce their “negative”² exposures.

Hypothetical Value Portfolio’s Exposure to Nonvalue Factors and Their Contributions to Return

Controlling for Unintended Factors

By optimizing the portfolio, we can apply different and specific constraints to these factors to ensure zero exposure at the time of the rebalance. Adding constraints often comes at a cost; in this case, it may mean lower exposure to value factors as well.

To test the effects, we compared the average exposure to value and nonvalue factors for three approaches that are similar except for how nonvalue style factors are constrained versus the benchmark MSCI World Index.

• Base case: Exposures to nonvalue factors are controlled to be within 0.25 of the benchmark.
• Tight-symmetric-constraint case: Exposures³ are constrained to be zero.
• Asymmetric-constraint case: Exposures are constrained to be nonnegative.

We can clearly see in the exhibit below the trade-off between exposure to value factors and reducing exposure to others. Both the tight-symmetric and asymmetric constraints eliminated the negative exposure to nonvalue factors to a good degree, but did so at the cost of reduced exposure to the target value factors (using the less-restrictive, asymmetric constraints had a less-negative impact).

Impact of Constraints on Exposure to Value and Nonvalue Factors

Constraints are applied on the semi-annual rebalances. Between rebalances, exposures change. The numbers shown are monthly averages.

Accounting for Risk and Return when Value Underperformed

Overall, by eliminating the negative impact of nonvalue factors, the two constrained approaches had higher annual returns than the base case. Lower exposure to these factors also meant they contributed less to absolute and relative risk (tracking error).

While both constrained strategies eliminated the negative return contributions of nonvalue factors, the return contributions of the value factors were also impacted due to lower exposure, as seen in the exhibit below. The return contribution from book-to-price dropped from 0.6% to 0.4%. The earnings-yield factor’s contribution also fell in absolute terms, which actually helped performance, as this factor’s return over the period was negative.

Risk/Return Characteristics of Three Value Approaches (2011-2020)

What About When Value Did Well?

To better understand the differences between the two constraint approaches, we analyze them for the period of 2001 to 2010, as the behavior of the value factor and value portfolios over the past two decades was quite different. While they suffered from 2011 to 2020, they outperformed the broad market from 2001 to 2010.

From 2001 to 2010, the exposures of a hypothetical value portfolio to nonvalue factors were mostly positive and these factors contributed positively to returns, as highlighted in the exhibit below. Tight-symmetric constraints reduced positive exposure to some nonvalue premia factors and eliminated their positive return contributions. The asymmetric constraints, however, only controlled these factors to be nonnegative and as a result, the performance contribution from nonvalue factors increased from 1.4% to 1.7%.

The value factors, book-to-price and earnings yield performed well in this period. Their contributions to return were positive and large. As we saw for the 2011-2020 period, though, constraints on nonvalue factors also resulted in lower value-factor exposure and reduced their contributions to return. The tight-symmetric constraints had a much larger negative impact.

Risk/Return Characteristics of Three Value Approaches (2001-2010)

We can also see in the exhibit above that returns for the constrained approaches were lower than the base case. The reduction, however, was much smaller for the asymmetric constraint (0.2%) than the tight-symmetric constraint (2.1%). Absolute and relative risk were also reduced, and, again, much more so for the tight-symmetric constraint.

Looking over the full 20-year period, the asymmetric-constraint case fared better in most years.

Annual Return of Three Value Approaches Relative to MSCI World Index

Regardless of a value investor’s approach to portfolio construction, it remains critical to understand the nuances and dynamics of the value factor and how other factors have contributed to a value portfolio’s risk/return characteristics.

The authors thank Saurabh Katiyar for his contribution to this blog post.

¹We used MSCI’s Barra^® Open Optimizer and Global Equity Model for Long-Term Investors (GEMLT) to maximize exposure to the weighted average of book-to-price ratio and earnings yield. Sector, country and stock-level weights are constrained relative to their weights in the MSCI World Index. Exposure to nonvalue factors are constrained to be within 0.25 of the MSCI World Index. The portfolio is rebalanced semi-annually with a one-way turnover budget of 20% for each rebalance.

²We changed the sign of some GEMLT factors to consistently have positive exposure associated with historical premia: e.g., size changes to low size.

³The exposures presented in this blog are all exposures relative to the benchmark.

Further Reading

The Theory of (Value) Relativity

Value-Performance Anxiety

Bringing Value to the 21st Century

The growth-factor premium: Seeking a systematic approach for capturing it