Can investors beat a typical active U.S. equity mutual fund via a small portfolio of periodically re-weighted equity exchange-traded funds (ETF)? In their February 2019 paper entitled “Are Passive Funds Really Superior Investments: An Investor Perspective”, flagged by a subscriber, Edwin Elton, Martin Gruber and Andre de Souza:
- Determine via cluster analysis a small set of ETFs that captures most of the variation in 69 broad U.S. equity indexes.
- Explore use of this set to mimic past performances of many active U.S. equity mutual funds via 24-month linear regressions with ETF coefficients scaled to sum to one.
- Compare next-year (close of first trading day of the year after coefficient calculation to close of first trading day next year) returns of mimicking ETF portfolios and active mutual fund counterparts.
Their target set of 883 active U.S. equity mutual funds are those with at least: three years of data as of January 2003; $15 million in assets; and, 90% of assets allocated broadly to stocks. Using monthly returns for 69 U.S. equity indexes, the small set of passive equity ETFs that capture variation in these indexes and 883 active U.S. equity mutual funds during January 2003 through December 2018, they find that:
- Cluster analysis over the full sample period finds that the 69 indexes consist essentially of four groups: large value, large growth, small growth and midcap value. ETFs tracking these four groups plus an ETF that tracks the overall market capture most of the variation in all available U.S. equity ETFs.
- Mimicking portfolios of the five ETFs explain 91%-94% (89%-92%) of the variation in active mutual fund styles when shorting is (is not) allowed. However, ETF weights when shorting is allowed can be extreme (e.g., -175%) and may require leverage.
- Regarding next-year performance versus individual mutual funds:
- On a gross basis, mimicking ETF portfolios outperform active mutual fund counterparts 77% (78%) of the time when shorting is (is not) allowed.
- On average, mimicking portfolios outperform active mutual fund counterparts by a gross 1.44% (1.37%) when shorting is (is not) allowed, and they have lower average standard deviations of returns.
- On average, mimicking portfolios outperform active mutual funds during 12 (13) of 15 years when shorting is (is not) allowed.
- Annual mimicking ETF portfolio trading frictions would reduce returns, but four of five ETFs have very small bid-ask spreads. Also, some active mutual funds have loads and transaction costs.
- With shorting allowed (not allowed), mimicking ETF portfolios significantly beat active mutual fund counterparts on average for nine of nine (eight of nine) Morningstar style categories. And, again, average standard deviations of returns by style are lower for mimicking ETF portfolios than active mutual fund counterparts.
- Adding momentum, midcap growth or smallcap value ETFs to the base set of five has little effect on results.
- Screening mutual funds for the lowest expense ratios and highest Morningstar ratings boosts mutual fund performance, but they still lose on average to their mimicking ETF portfolios.
- Simply investing in one ETF that matches an active mutual fund prospectus benchmark beats the active fund 72% of the time by an average 1.01% per year, just modestly less than the mimicking ETF portfolios.
In summary, evidence indicates that investors can reliably outperform a typical active U.S. equity mutual fund via annual mimicking reallocations to a set of five broad equity ETFs.
Cautions regarding findings include:
- As noted, reported results are gross, not net. However, trading frictions for annual reformation of mimicking ETF portfolios should be low. When shorting is allowed, there may be additional costs for shorting and leverage.
- The authors acknowledge that “…costs of buying and selling the ETF might well have been higher in prior years. However, what is relevant to investors is their size going forward and current costs are the best estimate.” However, changes in trading frictions may drive changes in investor behaviors.
- The authors employ full-sample cluster analysis to select the set of five core ETFs, thereby incorporating look-ahead bias. An investor operating in real time may have selected different core sets at different times. Said differently, the selected core set may not work as well after 2018 as before. Moreover, cluster analysis for selecting a new core set in the future is beyond the reach of most investors, who would bear fees for delegating such an update to an expert.
- Annual scaled regressions of core set ETF returns versus selected active mutual fund returns to determine next-year ETF weights is beyond the reach of many investors, who would bear fees for delegating the process to an advisor/fund manager.
- The study addresses U.S. equities only. Findings for other asset classes may differ.
