I think that the term “factor investing” has been diluted by the onslaught of new, poor quality, factor-based products in the investment product market place. There are countless ETFs and mutual funds employing the word “factor” in their names. This dilution makes appreciating the importance of factor research a challenge; it is easy to brush off the idea as a sales pitch when in many cases it is used as one.
Most of the attention on factors is given to their expected outperformance relative to the market. This makes it increasingly challenging to warm up to factors over certain time periods because they do not always outperform. It is possible, and probable, that factors will underperform the market for extended periods of time.
The benefits of factor investing are derived as much from diversification as from higher expected returns. Diversifying across factors is at least as important as diversifying across geographic regions, and it should be viewed that way. In practice, geographic diversification is generally accepted while factor diversification is often overlooked.
With geographic diversification we expect to see periods where one region dominates the others in a portfolio. At those times an investor would wish that they had all of their eggs in the top performing basket, but sensibility will (hopefully) keep them diversified, and even have them rebalancing into the underperforming asset classes.
The reason that global diversification has become generally accepted as a necessity in portfolio management is that global stocks tend to be imperfectly correlated with each other. For example, we see the imperfect correlation between Canadian, US, and International stocks going back to 1988 in the following table.
| Market Index | S&P/TSX Composite | Russell 3000 | MSCI EAFE + EM |
|---|---|---|---|
| S&P/TSX Composite | 1.00 | 0.61 | 0.58 |
| Russell 3000 | 0.61 | 1.00 | 0.65 |
| MSCI EAFE + EM | 0.58 | 0.65 | 1.00 |
It may come as a surprise to some that factors have been less correlated with each other than various stock markets, and they have been negatively correlated with the market. This highlights the importance of factor diversification. The following table shows the correlations between global factor portfolios going back to 1990.
| Global Factor | Market | Size | Value | Profitability |
|---|---|---|---|---|
| Market | 1.00 | -0.12 | -0.16 | -0.17 |
| Size | -0.12 | 1.00 | 0.54 | 0.58 |
| Value | -0.16 | 0.54 | 1.00 | 0.67 |
| Profitability | -0.17 | 0.58 | 0.67 | 1.00 |
A factor portfolio is long a long-short portfolio. For example, the size factor is long small stocks and short big stocks; the factor return, or premium, is the difference in average returns between small stocks and large stocks. We expect that difference to be positive over the long term, but the point of this table, and this post, is that the factors do not move in lock-step with each other. Finding an asset class with a negative correlation to add to a portfolio is the holy grail of investing, and here it is.
Correlations are notoriously unreliable. They can change quickly, and that usually happens at the worst possible times. For example, the following table shows the increasing correlations of Canadian, US, and International stocks through the worst 12-month period of the financial crisis.
| Market Index | S&P/TSX Composite | Russell 3000 | MSCI EAFE + EM |
|---|---|---|---|
| S&P/TSX Composite | 1.00 | 0.69 | 0.78 |
| Russell 3000 | 0.69 | 1.00 | 0.84 |
| MSCI EAFE + EM | 0.78 | 0.84 | 1.00 |
Correlations falling apart at the worst possible time has obvious implications. In many ways, it diminishes the benefits of diversification. In a 2012 paper in the Journal of Portfolio Management titled The Death of Diversification Has Been Greatly Exaggerated, Antti Ilmanen and Jared Kizer demonstrated that while global diversification may fall apart in a crisis, factor diversification does not. The following table demonstrates that while the correlations between size, value, and profitability did increase in the financial crisis, the correlations between those factors and the market became increasingly negative.
| Global Factor | Market | Size | Value | Profitability |
|---|---|---|---|---|
| Market | 1.00 | -0.38 | -0.22 | -0.37 |
| Size | -0.38 | 1.00 | 0.73 | 0.89 |
| Value | -0.22 | 0.73 | 1.00 | 0.79 |
| Profitability | -0.37 | 0.89 | 0.79 | 1.00 |
Combining higher expected returns with imperfect or even negative correlations is an attractive proposal.
In a 2017 paper published in the Journal of Portfolio Management, Louis Scott and Stefano Cavaglia looked at the effect of factor diversification on the odds of retirees not outliving their portfolios. Not only did they find that adding in factor exposure improved the expected outcomes for retirees, they also found that a factor diversified portfolio provided a smoother ride with smaller drawdowns. This favorable analysis even persisted when the study authors cut the expected factor premia in half.
In a 2018 paper in the Financial Analyst’s Journal, Eugene Fama and Kenneth French used statistical analysis to explain that even over long (10+ years) time periods the US market, market value, large value, small value, and small premiums each have a substantial probability of being negative. This means that it should come as no surprise that, over a given time period, a factor premium may be negative. We have seen this for US value in the past 10-year period. We have also seen this for US market from 2000 through 2010. However, as the paper demonstrates, market value and small cap value stocks are the least likely to lose money over most time periods due to the shape of their distribution.
The following table shows a sample of data from Fama and French’s paper. We see that over 10, 20, and 30-year time periods, there is a material probability of any premium being negative, but that probability is smallest for value stocks and small value stocks.
| Market over T Bills | Market Value over Market | Large Value over Market | Small Value over Market | Small over Market | |
|---|---|---|---|---|---|
| 10-Year | 17.79% | 10.90% | 22.44% | 6.09% | 24.37% |
| 20-Year | 11.22% | 5.22% | 16.03% | 2.13% | 18.33% |
| 30-Year | 8.02% | 3.11% | 12.64% | 0.97% | 14.97% |
Source: Adapted from Volatility Lessons, Eugene Fama and Kenneth French, Financial Analysts Journal
The implications of the data in this post are extremely important to investors. Factors have had imperfect correlation with each other and negative correlation with the market, even during periods of market stress. This means that a factor diversified portfolio has performed differently from the market, and it may continue to do so.
We have been living through a strong equity bull market where large cap growth stocks have stolen the show. We would not expect size, value, and profitability to exhibit a positive premium during this time. As we have seen from Fama and French’s paper, 10-year periods of factor underperformance should be expected.
Ilmanen and Kizer conclude their paper by asking why every investor is not investing in a factor diversified portfolio. They offer an insightful list of reasons:
Factor research is not new. Professional money managers and any serious investor should be at least aware of factors, the research behind them, and their potential benefits in portfolios.
A distrust in factor premia is reasonable as there is always the concern of data mining. Even when data are robust there is risk in drawing conclusions from the past. However, as we have seen from Fama and French, the distribution of outcomes for market value and small value stocks appear very attractive over long horizons. This is true even if we do see extended periods of underperformance.
A lack of consensus on the factors to include in a portfolio is a real concern. There have been at least 300 factors published in the academic literature, and it is estimated that 40 more are published each year. The important aspects to consider with any factor-based strategy are how the factor research was vetted, who did the vetting, how they interpreted the data, and their ability to understand of the limitation of factor models in live portfolio construction.
Finally, an aversion to shorting and leverage is important. Factors are long-short portfolios. Most investors are not comfortable shorting, nor should they be. Ilmanen and Kizer concede that although factor diversification is most effective when shorting and leverage are employed, it does not follow that long-only investors should ignore factor diversification – the benefits are meaningful even for long-only investors.
Despite some periods of underperformance, which should be expected, the diversification benefits of factors, and the statistical reliability of market value and small cap value stocks over long horizons, make ignoring factors in the portfolio construction process a potentially massive oversight.
