Flawed Research on Factor Investing

An extensive discussion on this article is available on APViewpoint in this conversation.

The popularity of smart-beta products has raised concerns that certain factors have been “overgrazed” – that their expected return has been driven down due to popularity-driven demand. A new research paper purports to refute this concern. But its logic is flawed and practitioners should be highly skeptical of its conclusions.

The paper in question is Are Exchange-Traded Funds Harvesting Factor Premiums?. It was written by David Blitz, who is with Robeco Asset Management, a Netherlands-based asset manager and producer of quantitative, factor-based investment products.

Blitz’ paper was published on the SSRN platform and has not appeared in a peer-reviewed publication.

I shared this article with Blitz for his feedback and comments. I also had an email exchange with him to be sure I understood his research correctly.

I have no reason to believe that Blitz is intentionally misleading the investing public. I hope that he believes he is acting in the best interest of Robeco’s clients. I do not have any reason to believe that Robeco’s products are inferior to similar products from other fund companies.

Blitz attempts to refute the “overgrazing” hypothesis by showing that the aggregate exposure to factors among ETFs is close to zero. For example, he shows that the amount of assets invested in ETFs that offer a positive value exposure is roughly equal to the amount invested in ETFs with a negative value exposure.

He concludes that, “On aggregate, all factor exposures turn out to be close to zero, and plain market exposure is all that remains. This finding argues against the notion that factor premiums are rapidly being arbitraged away by ETF investors, and also against the related concern that factor strategies are becoming ‘overcrowded trades.’”

But, as I will explain, his methodology is insufficient to prove his hypothesis. There can be significant overgrazing even though there is net-zero factor exposure.

Before I offer my proof, consider this very rough, imperfect analogy. Let's say that there are a bunch of cars on the highway and that, through an omnipotent police force, an average (median) speed of 55 mph is always enforced. Then someone comes along and observes that there are as many cars traveling over 55 as there are under 55, and concludes that none of the cars going over 55 are accelerating. That is an obviously fallacious and unjustified conclusion. But it is the rough equivalent of Blitz’ findings.

To refute Blitz’ paper, one needs to show that expected returns from factor-based strategies can be driven down even when there is a net-zero factor exposure in the ETF market. Let’s assume that investors have driven up the prices of value ETFs. At the same time, other investors have increased their investment in growth ETFs proportionately to the increase in price of value ETFs[1]. In this case, the net exposure to the value factor among ETFs could be zero, but the expected returns of value ETFs has declined.