I am a fan in principle of the idea of smart beta funds. These funds eke out a bit of extra return by moving beyond capitalization weighted indexes. Many of these funds use some sort of alternate factor to weigh their exposures such as dividends, growth or value. I have discussed some pluses and minuses of specific funds at ETF.com here and here.
However, I wanted to make readers aware of a bit of academic research which we have been reading lately which makes me want to be a bit cautious when using factors. The first is by Duke professor Campbell Harvey (who I heard speak at the Society for Quantitative Analysts last week) and Texas A&M’s Yan Liu who argue in “And the cross section of stock returns…” most factors discovered are likely false. They contend that the advent of cheap computers and the strong pressure to find new ideas has led to so many factors being researched that, in effect, people have pronounced factors as statistically proven while, in fact, the relationship is just random. The details of this may have you looking back to your college statistics book but Professor Harvey motivated a recent talk with this XKCD cartoon. Harvey sees a few of the most important ideas such as value and momentum as significant, but many others such are earnings price ratio or small cap versus large cap are more suspect.
Another interesting piece of research was published in by Rob Arnott and his colleagues at Research Affiliates titled “The surprising alpha from Malkiel’s monkey and upside down strategies“. These researchers find several potential smart beta strategies get a good deal of their returns from value and size tilts, even those which do not explicitly take value into account. Imagine a stock trading at a given price to book (PB) ratio. If a random shock moves the price of the stock up with no change in the book value, then the stock has a bigger place in the cap weighted portfolio and a smaller weight in any other portfolio. Should the price of the stock decline, the opposite holds. In some of my own research which informed the ETF.com notes, I found several smart beta indexes which do not have the value or size tilts so you can find them if you look.
Arnott also finds that random portfolios outperform cap weighted portfolios, though I think it’s best to think of the random portfolios as noisy versions of equal weight portfolios, who also have tended to outperform cap weighted indexes and also show size and value exposures.
A more challenging idea in the Arnott paper is that some inverse smart beta portfolios also outperform the market. That is instead of buying stocks that are cheap from a PB perspective, buy those which are expensive. For some thoughts against this notion see this presentation by Felix Glotz who fails to replicate the Arnott findings.
I think both Harvey and Arnott would agree that it is important to make a deliberate decision about what sort of index you want to have and think carefully about what the economic sources of return are supposed to be when moving away from the default of capitulation weighted.