Factor performance, as conceived by Fama and French and refined by others, is based on adding the returns of a “long” portfolio of securities that most embody the factors to a “short” portfolio that least represent the factors. But it is common practice for mutual funds and ETFs to use only the long portfolio. New research show that this approach does effectively capture the returns of the underlying factors.
In finance, a factor can be defined as a property or set of properties common across a broad set of securities that provides exposure to a unique risk that has delivered a premium return. The factors with the most support in the literature are market beta, size, value, momentum, profitability/quality, investment and low beta/volatility. Thus, a factor is a quantitative way of expressing a qualitative theme.
By design, factors are long-short portfolios. For example, a value factor portfolio might go long the top 30% of the cheapest stocks by some metric (such as price-to-book or price-to-earnings) and go short the 30% most expensive stocks (the growth or glamour portfolio). In this way, the portfolio captures not only the outperformance of value stocks but also the underperformance of growth stocks. A benefit of a long-short approach is that it basically eliminates exposure to market beta, making factors uncorrelated to the market portfolio.
An issue for investors, especially retail investors, is that many are constrained to investing in long-only portfolios, which raises the question of whether long-only factor portfolios generate expected benefits. David Blitz, Guido Baltussen and Pim van Vliet contribute to the literature with their November 2019 study “When Equity Factors Drop Their Shorts.” To determine the source of factor premiums, they broke down common equity factor strategies into their long and short “legs.” They began by noting: “The issues involved with shorting individual stocks can be solved effectively by hedging the market beta of a long-only factor portfolio with (liquid derivatives on) broad market indices. With this approach one captures the performance of the long legs of factor premiums. The performance of the short legs can be isolated in a similar fashion, i.e. by considering the short portfolio in combination with an offsetting long position in broad market indices that brings the market beta to zero.” Their data covers the period 1963 through 2018; the value, momentum, profitability, investment and low volatility factors; and four regions (North America, Europe, Japan and Asia ex Japan). Following is a summary of their findings:
- Most added value tends to come from the long legs. The individual Sharpe ratios range from 0.40 to 0.61. For the value, momentum and investment factors, the long legs have a higher Sharpe ratio than the short legs. For the profitability and low-risk factors, the short side is a bit stronger – the individual Sharpe ratios vary between 0.37 and 0.54.
- The long and short sides of a factor tend to be highly correlated, with correlations ranging between 0.59 and 0.85 for individual factors. This number increases to 0.87 for the multifactor combination. However, the alphas of the long legs tend to be stronger than the alphas of the corresponding short legs. The longs have positive and mostly statistically significant alphas compared to the shorts, while short-leg factor exposures do not add significant value when controlling for the long legs.
- The long legs of factors offer more diversification than the short legs. The average correlation among all long legs is negative (-0.04), while the corresponding number for the short legs is positive (0.31). For an equally weighted portfolio of the five long legs, the Sharpe ratio increases to 1.10 (due to diversification effects) versus 0.69 for the short legs, and a combined Sharpe ratio of 0.86.
- The results hold across large and small caps, are robust over time, carry over to international equity markets (for the global sample, the long legs have a Sharpe ratio of 1.19 versus 0.66 for the short legs, though for Japan the short legs appear to be slightly better), and cannot be attributed to differences in tail risk, as the shorts exhibit more negative skewness and higher excess kurtosis than the longs.
- Moreover, this does not account for the substantially higher implementation costs involved with the shorts compared to the longs. Those higher costs include not only borrowing costs but also higher trading costs (the securities in the short leg tend to be less liquid).
- As with small-cap stocks themselves, factor premiums tend to be larger in the small-cap space than in the large-cap space. The Sharpe ratios are about twice as large in small caps as they are in large caps.
- The poor performance of high-risk and growth stocks can be explained by their “junk” resemblance. However, value is different than quality. Similarly, high risk is junk, but low risk is not quality.
- Low risk and value are distinct factors on the long side – they are not subsumed by the profitability and investment factors (as had been found in prior research that considered long-short portfolios).
Blitz, Baltussen and van Vliet concluded: “We find that the long-minus-market approach has typically been more powerful than the full-fledged long-short approach for individual factors, and even more so for a multi-factor combination.” They added: “From a theoretical perspective these results imply that it is important to understand the long and short sides of factor premiums separately. The practical investment implication is that since there is no unique alpha in the short legs, an efficient approach to factor investing is to simply concentrate on the longs and hedge out the market beta with liquid market index derivatives. We note that the above findings are without considering shorting costs and the feasibility of shorting, and that incorporating such information would most probably lead to even stronger conclusions (as shorting costs tend to be substantial and shorting is difficult for smaller stocks).”
Their findings present some interesting challenges to financial theories. For example, why are the correlations low in the long legs but high in the short legs? In terms of the efficient market hypothesis, if the long side represents riskier assets, the short side should be relatively safer. However, the findings show that the poor performance of growth stocks is due to their resembling junk when they should be hedges of your labor capital as safer assets they should perform better during times when labor capital is at risk. In terms of behavioral finance, the short side should show large alphas due to mispricings, which are difficult to correct due to limits to arbitrage. However, the alphas are on the long side. The good news is that investors who don’t want to short the broad market can gain exposure to the long side of factors while simply avoiding the short side by investing in funds that screen out stocks with those characteristics. Doing so avoids the high costs and risks of shorting.
Another observation is that intuitively we would like to observe a monotonic relationship between the characteristic (such as P/E ratio) and returns (returns increase across deciles, quintiles or quartiles). This is the way many factors behave. A notable exception would be low-volatility-type factors, where the highest volatility stocks (for example, quintile) tend to have notably lower returns and the other four quintiles are all comparable. That said, Blitz, Baltussen and van Vliet provide us with important insights – insights that investors can use to construct portfolios such as the ones described in my book, Reducing the Risk of Black Swans.
Larry Swedroe is the director of research for The BAM Alliance, a community of more than 130 independent registered investment advisors throughout the country.
More Factor-Based Investing Topics >
