This is the first post in a mini-series on how factor returns from academia can be different to those from smart beta investing.
Perhaps the most obvious difference between factors and smart beta is that factors are long/short. Given that the vast majority of smart beta ETFs are long-only, it’s only sensible to assume that you wouldn’t get the same results investing in a smart beta fund as you would if you invested in factors. Researchers at AQR have written a fantastic paper on the topic, which does a far better job getting into the nitty gritty than I ever could, so I’ll go through their findings below. The paper broadly seeks to answer 2 questions:
- How much of the alpha from a long-short factor strategy can be captured with a long-only strategy?
- How does market cap affect the ability to capture long-only factor alpha?
To begin with, the authors compared the returns of long-short portfolios to long only portfolios for the size, value, and momentum factors. The graphs below show that whilst momentum and value produce statistically significant alpha for both long-short and long only portfolios (size doesn’t), the returns for long-only portfolios are lower than in the long-short portfolios. But it’s interesting to note that long only momentum and value still provide statistically significant outperformance.
Source: AQR
Source: AQR
Most of our factor exposure is from large cap
If we focus on large cap equities, as that’s where most people will be getting their factor exposure from, it’s useful to understand how much alpha comes purely from the long side of a large cap factor strategy, as this is what most investors will be receiving. Given that the size factor doesn’t produce statistically significant alpha either long/short or long-only, the authors focus on the value and momentum factors.
The table below forms factor portfolios by sorting them into size quintiles, to gauge the effect the size has on the percentage of alpha coming from each the long and short side. Size 1 is the portfolio formed of the smallest companies, and Size 5 is the portfolio with the largest companies.
Looking at the momentum factor first, the research shows that most of the alpha for a large cap momentum strategy is on the short side. 38% of the profits are on the long side, meaning 62% are from the short side. Whilst not apparently good for large cap momentum strategies, the research also shows that despite most of the profits coming from the short side, the long-only strategy still produces statistically significant alpha (t-stat of 3.83).
The opposite is true for value. The percentage of profits from the long side increases as size increases, which sounds good for a large cap long-only value strategy. However, the statistical significance of the alpha reduces as size increases, to such a point that the alpha in large cap value is not statistically significant from zero (t-stat of 1.21), unlike that of momentum (3.83).
Source: AQR
Graphically, we can see the result in the chart below. The alpha from the long side is in grey, and the alpha from the short side is in red:
Source: AQR
In AQR’s words, “there is no discernible pattern in momentum returns across size quintiles… For value strategies, shorting comprises a little more than half of the premium in small caps for value weighted portfolios and about a third for equal weighted portfolios, with the importance of shorting declining as firm size increases. However, since the premium for value also declines substantially as firm size increases, the decline in the importance of shorting for a value strategy coincides with a much lower value premium.”
Does this result hold outside the US?
The authors show that the split of profits on each the long and short side is similar across geographies and asset classes. They look at stocks in the US, UK, Europe, Japan, and globally, as well as equity indices, currencies, fixed income, commodities and ‘global other’, finding that “the contribution from the long side for value is almost exactly 50%, and the contribution from the long side for momentum is about 54%. Overall, across all markets and asset classes, the contribution to profits comes equally from the long and short sides for both value and momentum.”
Source: AQR
Overall
In other words, for long-only size factor investors, the factor doesn’t have any statistical significance when constructed using a long-only portfolio. This is bad news for smart beta funds which invest in small cap, but don’t employ any shorting (most of them).
Focussing on the value and momentum factors, strategies operating in the top quintile of size, which is where most smart beta funds operate, momentum looks like it does have the potential to produce statistically significant alpha, but much less than would be advertised from a long-short portfolio. This is good news for momentum investors.
For value investors, large-cap value strategies, whilst having a large proportion of the alpha deriving from the long side, don’t appear to generate alpha that’s statistically distinguishable from zero.
Overall, long-only factor portfolios are far removed from their ancestral long/short academic factors. Their differing methods of portfolio construction make it unlikely that they’d perform in similar a manner, and the evidence suggests that those investing in long-only factor portfolios should temper their expectations when comparing to the investing theory.
The next post in the series will focus on how factors might work better in small caps, and what that means for smart beta investors.