David Blitz has recently discussed the concern in some quarters that factor premiums may disappear as they are arbitraged away by exchange-traded funds.

Blitz, the head of quantitative equity research at Robeco Asset Management, says that such a concern is not justified, because “the exposures of ETFs that may be suitable for factor investors are almost perfectly offset by opposing exposures of other ETFs.”

Blitz’ study focused on the factor exposures of U.S. equity ETFs. His sample included every ETF listed in the U.S. that invests in U.S. equities and that had at least thirty-six months of return history by the close of 2015. There are 415 such funds, and their combined assets under management exceed $1.2 trillion, which is about 5% of the entire U.S. equity market.

Blitz worked from the market excess return of these ETFs, which he defines as the total return minus the risk-free return. He then regressed the time-series of monthly excess total returns for each ETF against the four mainstream factors: size, value, momentum, and low-versus-high volatility.

Four Factors

Of course an ETF can be either long or short on either of these factors. If ETFs as a whole were shorting a factor, it would have the effect of exacerbating the significance thereof, or increasing the excess return available for anyone going long. On the other hand, if the ETFs as a whole were long a factor, that would have the effect of reducing the excess return for anyone else. It would constitute (to some degree) a matter of arbing away the market inefficiency that the factor represents.

By way of review:

The size factor is the excess to be gained by buying stocks with relatively low market capitalization; the value factor is the excess to be gained from stocks that have a low price relative to their underlying value, whether the latter is defined by book value, earnings, dividends, or free cash flow; the momentum factor is the excess to be gained from stocks that have outperformed in the near-term past (three months to one year); and the volatility factor is the excess to be gained from stocks with, well … low vol.

For the most part, the smart-beta ETFs are long on these factors, but the conventional ETFs are short. The AUM weighted aggregate exposure of conventional ETFs on the value factor, for example, is -0.08. The same number for smart-beta ETFs on the same factor is, neatly enough, +0.08.

For the volatility factor, the aggregate exposure of the conventional funds is -0.03. The same number for the smart-beta ETFs on the same factor is 0.06. This is not quite so neat a matching up of short with long, but it suggests a situation where ‘arbing away’ the factor is very distant.

Arbing and Overcrowding

Some of the apparent concern about arbing, as expressed by example by Chincarini in The Crisis of Crowding, a book published by John Wiley & Sons in 2012, might actually be concern about the somewhat more nebulous concept of over-crowding.

The concepts overlap, so that the above summarized data also tends to show that worries about a crowding “crisis” are overstated. Still, one might still worry that (as Blitz puts it) “ETFs might collectively develop highly concentrated positions in certain stocks at a certain point in time.” That issue, Blitz says, is beyond the scope of his paper.

Getting More Granular

Blitz’ tables show for each factor the ten funds that have the largest exposures (positive and negative) toward each factor. On the value factor, the funds with the highest positive exposure are the following:

Source: Blitz, “Are ETFs Harvesting Factor Premiums,” Journal of Investment Consulting 18:1 (November 2017) at 28.

This top 10 is dominated by financial and energy ETFs, which indicates to Blitz that “certain sector indexes provide more value exposure than indexers that are specifically designed to provide value exposure.”Further, the valuation differences among sectors “tend to be persistent over longer periods of time, so sector ETFs might be the preferred choice for investors whop wish to harvest the value premium with ETFs.”