When “Data Mining” Dates to the Eisenhower Administration
This is part three of a four-part blog series addressing the attacks on smart beta and ETFs. Today’s topic is supposed data mining in smart beta.
Going Way Back
We often hear assertions that smart beta strategies mine data based on factors that have performed handsomely in recent years. The problem with this argument is that many of the classic smart beta concepts, “value” being the leader of the pack, have been around for so long that many investors have forgotten just how many years it has been since their discovery came into being.
Of the main smart beta factors (value, quality, size, momentum and low volatility), none has been around as a concept quite as long as value, which goes back to Benjamin Graham and David Dodd in 1930.1 Additionally, Wharton Professor Jeremy Siegel’s data on value stocks dates to 1957,2 while Dartmouth’s Ken French has a data library going back to 1926 that is strongly supportive of the value and size factors in particular.3
Furthermore, a well-known index like the S&P 500 Equal Weight Index, which dates to December 29, 1989, outperformed the market capitalization-weighted S&P 500 Index by 123 basis points (bps) per year through August 16, 2017, a compounding that caused the former to multiply 17-fold, while “beta” (the S&P 500) rose 12-fold over the nearly three-decade horizon.
Additionally, we need to carefully consider the nature of the attack against “data mining.” What exactly does it mean, who defines it and what is the alternative? Take for example, WisdomTree’s 2006 vintage of dividend-weighted ETFs. Rather than weighting by market capitalization, we simply chose to have our original ETFs focus on a metric that made more logical sense than cap weighting. How is something like that “data mining”? It isn’t.
Furthermore, the charge of data mining is usually lobbed in tandem with an argument that “no one ever shows a bad back test.” That’s right, and that’s because the industry is and should be striving for nothing short of excellence. Does this charge mean that we should, instead of seeking out stocks with positive fundamentals, try to put our investors in companies that are expensive or have deteriorating fundamentals? Of course not; we have a responsibility to investors, as do our competitors in the smart beta business. Both smart beta and cap-weighted indexing align the money manager’s best interests with those of clients. We all put together indexes that we think will help our clients, because we want to keep our clients for generations. That’s just common sense.
A Mining Assignment
Finally, we would be remiss if we didn’t mention the one kind of mining that many in this industry desperately hope clients won’t engage in—and that is mining to page 96 of the Investment Company Institute’s 2017 Fact Book. It reveals a dirty truth: actively managed equity and bond mutual funds have average expense ratios of 82 bps and 58 bps, respectively.4 That is a problem … for mutual funds.
1Source: Benjamin Graham and David Dodd, “Security Analysis,” 1930.
2Source: Jeremy Siegel, “The Future for Investors” (2005), with updates to 2016. Data from 12/31/1957 to 12/31/2016, using the S&P 500 universe as of 12/31/16.
3French’s work on the value factor, for example, defines value stocks as those in the highest 30% by the book-to-market ratio, with higher values indicating a lower market price compared to this fundamental factor, within the large-cap universe. Broad market refers to a market capitalization-weighted measure of the returns of all firms captured by the Center for Research in Security Prices and listed on the New York Stock Exchange, American Stock Exchange or NASDAQ.
4Expense ratios are measured as asset-weighted averages. Data excludes mutual funds available as investment choices in variable annuities and mutual funds that invest primarily in other mutual funds.