Swedroe: Another Angle On Factor Diversification

Last week, we examined the data (from my new book, “Your Complete Guide to Factor-Based Investing,” which I co-authored with Andrew Berkin) on the odds that the premiums associated with some common investment factors would produce a negative return over various horizons.

We then examined how constructing a diversified factor portfolio might impact those odds of underperformance. Today we’ll tackle factor diversification from another angle, by looking at the annualized returns and annual standard deviation of two simple portfolios.

The ‘Typical’ Portfolio
In the table below, Portfolio A is a “typical” portfolio with an allocation of 60% stocks and 40% bonds. To implement this allocation, it invests 60% in Vanguard’s Total (U.S.) Stock Market Index Fund (VTSMX) and 40% in the Vanguard Intermediate-Term Treasury Index Fund (VFITX). Portfolio B invests its equity portion in small value stocks but keeps its bond allocation in VFITX.

Because the stocks it does hold have a higher expected return, Portfolio B is able to use a smaller equity allocation. In this case, it’s 40% instead of 60%. To implement this allocation, it invests in the DFA U.S. Small Cap Value Fund (DFSVX). (Full disclosure: My firm, Buckingham, recommends DFA funds in constructing client portfolios.)

The reason for choosing DFA’s fund instead of Vanguard’s fund is that it has higher loadings on some other factors. You can see that in the Morningstar data. Their portfolio tool shows that the average market capitalization of DFA’s small value fund is currently less than half that of the Vanguard small value fund (VISVX), and its price-to-earnings, price-to-book value and price-to-cash flow ratios are also much lower. In other words, DFSVX is more “valuey” as well.

The data covers the 24-year period from April 1993 (the inception date of DFSVX) through March 2016. The figures inside the parentheses are the loadings of the funds used on (or their exposure to) each of the factors. To obtain the portfolio’s factor loadings (the figure to the left of the parentheses), we multiply the loading on the factor by the percentage allocation of the fund.

A More Efficient Portfolio

Portfolio B was the more efficient portfolio, producing both higher returns and lower volatility. And because investors care about downside risk (more than upside returns), we’ll also look at returns in 2008. While Portfolio A lost 16.9%, Portfolio B lost just 6.7%. The lower downside risk is especially important to those in the withdrawal phase of their investment life cycle, when the order of returns can matter a great deal. What led to this result? The table provides a clue.

The first thing you should note is that all of Portfolio A’s equity risk is in the market-beta factor. That risk also dominates the risk of the overall portfolio, as the only other factor exposure the fund has is a 0.11 exposure to the term premium. Using volatility as the measure of risk, from a portfolio perspective, close to 90% of Portfolio A’s risk is concentrated in the sole factor of market beta. On the other hand, Portfolio B is far more diversified, with relatively more equal weightings on the other factors. It is more of what is referred to as a “risk parity” portfolio.

The benefits of diversification across these factors, each with premiums and also with low to negative correlations with other factors, led to the results you observe—a more efficient portfolio. It’s also what led to Portfolio B’s lower downside risk. VFITX gained 13.3% in 2008, and Portfolio B has a 60% allocation to it versus just 40% for Portfolio A.

The table below shows the correlations of these common investment factors for the period from 1964 through 2015.

Again, note the low-to-negative correlations each factor has with the others. The low correlations you can see here are what help to dampen the volatility of Portfolio B.

In addition, portfolios can be structured to load on the momentum factor, to which neither Portfolio A or Portfolio B has any exposure. And there are funds with higher loadings on the size and value factors than the DFA fund.

Conclusion

The academic research has given us a small number of equity factors that have provided premiums and added explanatory power to the cross section of returns while meeting all of my criteria for considering an allocation to a factor.

Specifically, it should be persistent across long periods of time; pervasive across sectors, countries, regions and asset classes; robust to various definitions (for example, you’d get a value premium of similar size no matter if your metric is P/E, P/B or P/CF); is implementable (survives transaction costs and fund expenses); and it has intuitive explanations (either risk-based or behavioral-based) why investors should expect the factor to continue to generate a premium. And with factors of this type, we have a new way to think about diversification.

Rather than viewing a portfolio as a collection of asset classes, we can view it as a collection of diversifying factors. Support for such factor-based investing strategies is provided by Antti Ilmanen and Jared Kizer in their 2012 paper, “The Death of Diversification Has Been Greatly Exaggerated.” The paper, which won the Journal of Portfolio Management’s prestigious Bernstein Fabozzi/Jacobs Levy Award for the best paper of the year, made the case that factor diversification has been more effective at reducing portfolio volatility and market directionality than asset class diversification. Other studies have come to the same conclusion.

You have now seen the evidence. And because we cannot know which factor will deliver the highest premiums, or even a premium, over even long periods of time, the prudent strategy whenever we don’t have a clear crystal ball is to diversify broadly across many factors.

Larry Swedroe is the director of research for The BAM Alliance, a community of more than 140 independent registered investment advisors throughout the country.