The tax benefits, along with their low costs, has led to the dramatic growth of exchange-traded fund (ETF) assets. Assets under management (AUM) have increased more than tenfold, now totalling more than 5,000 ETFs and $4.5 trillion. With so many choices, it raises the question of whether individual investors are making optimal decisions when choosing ETFs. Highlighting the importance of this issue is that so many ETFs are highly correlated, yet may have significant differences in expense ratios and liquidity (trading costs).
David Brown, Scott Cederburg and Mitch Towner examined how investors allocate assets to ETFs in their September 2020 study (Sub)Optimal Asset Allocation to ETFs. Their database covered U.S. equity ETFs from January 2000 through June 2018. They classified ETFs into five mutually exclusive categories: Index ETFs that track well-known indexes, quasi-index ETFs that follow straightforward rule-based strategies (e.g., equal-weighted S&P 500), Active ETFs that have actively managed portfolios or use proprietary strategies, Sector ETFs that provide exposure to one of 11 broad industries, and smart beta ETFs that pursue factor strategies related to value, growth, small cap, momentum, profitability, quality and/or low volatility. They used 21 Vanguard ETFs to develop peer-based benchmarks. Following is a summary of their findings:
Sector ETFs accounted for 40% of ETFs, and smart beta ETFs accounted for 32%.
Consistent with intuition, ETFs with lower expense ratios, lower bid-ask spreads, and higher trading turnover attract more assets. A one-standard-deviation improvement in liquidity was associated with an increase in ETF size of 29% (t-statistic of 3.8) for average absolute premium, 33% (t-statistic of 3.0) for bid-ask spread, and 39% (t-statistic of 5.6) for trading turnover.
Investors make substantial allocations to ETFs that are similar to existing ETFs while incurring the costs of higher fees and lower liquidity.
There is little systematic relation between ETF benchmark-adjusted performance and size.
Index ETFs tend to be the largest, oldest, most liquid and cheapest. In contrast, Active ETFs are the smallest, on average, by an order of magnitude, have the highest fees and have the worst liquidity — ETF age, which may proxy for less investor-value-based reasons for investing, such as a first-mover advantage or investor familiarity, is strongly related to ETF size.
More than 20% of the ETFs failed before the end of the sample period in July 2018.
90% of assets were in ETFs that were at least 80% correlated with SPY, and more than 50% of assets were at least 95% correlated with SPY.
As of June 2018, 49 of the 449 non-leveraged U.S. equity ETFs, accounting for half of all ETF AUM, were more than 95% correlated with SPY—though they may have higher expenses and higher trading costs. Even within families, many ETFs are highly correlated. For example, Vanguard offered 43 ETFs in the sample, and 15 of those were more than 95% correlated with SPY, providing little in the way of diversification benefits.
Relative performance and ETFs’ abilities to deliver promised risk exposures fail to explain why high-fee, low-liquidity ETFs attract assets.
Smart beta ETFs were best at providing small-cap exposure, with an average beta of 0.76. However, there were large differences, with 10% of small-cap ETFs having betas of 0.49 or lower. Smart beta ETFs tended to provide much lower exposure to value, growth and low volatility, with average betas of just 0.25, 0.19 and 0.24, respectively.
Only a few smart beta ETFs were able to provide substantial exposure to momentum, profitability or quality. The average betas were 0.09, 0.05 and 0.00, respectively.
For smart beta ETFs, the variation in exposures within strategy was large, as the difference between the 10th and 90th percentiles of beta estimates was about 0.5 for each factor. For the growth, momentum, profitability, quality and low volatility factors, the 10th percentile beta estimates were negative, suggesting some smart beta ETFs are delivering the opposite of their promised factor exposures.
While, on average, 56% of the systematic factor risk that an ETF was exposed to was explained by its stated risk factors, the 10th percentile was only 9%, while the 90th percentile was 95% — some funds do very well at delivering on their stated purpose, while others do very poorly. This highlights the importance of fund selection.
Smart beta ETFs with greater risk in the dimensions of their stated factor strategies attract more assets, whereas idiosyncratic risk is negatively associated with ETF size — investors reward ETFs for providing the exposures they promise and penalise them for taking on unwanted risk.
The estimated aggregate cost to ETF investors from investing in dominated funds was $1.1 billion to $17.5 billion during the sample period, depending on the benchmarking approach. In 2017 annual costs were $129 million to $1.5 billion, depending on the method. Surprisingly, the results were similar for institutional and retail investors alike.
Excess costs were driven by higher expense ratios in most years. However, because of the heightened volatility during the financial crisis, additional trading costs in dominated funds were an order of magnitude larger than additional fees.
To measure excess costs, the authors used two methods. The first was from the perspective of a short-term investor who demands liquidity and market exposure (e.g., for cash management) and compared each ETF to SPY, which was the most liquid ETF throughout their sample. The second method identified dominated ETFs, defined as those that were highly correlated (at least 95%, 97.5% or 99%) with an ETF that had lower fees and higher liquidity.
Using the second and stricter approach, the authors found that “16% to 64% of ETFs are dominated and 7% to 35% of total AUM are in these dominated funds (depending on the correlation threshold and quarter). Dominated ETFs tend to be sector or smart beta ETFs, whereas index ETFs tend to be dominant. Dominated ETFs do not outperform dominant ETFs nor do they provide better risk exposures, such that fund performance does not explain the substantial assets invested in dominated ETFs.”
An interesting finding was that investors prefer pure smart beta strategies that only have exposure to one factor as opposed to ETFs that claim to simultaneously target several factor strategies. This is an inefficient way to access multiple factors. For example, if an investor wants exposure to both the value and momentum factors, it is more efficient to access both factors in one fund because, otherwise, the value fund might be buying a security that has now become cheap while the momentum fund would be selling the same security because of the recent poor performance. The investor would have paid two expense ratios and had two funds incurring trading costs, with no change in their net holdings.
Conclusion
Their findings led Brown, Cederburg and Towner to conclude: “Our results indicate that, despite ETFs as a whole being known for low fees, ETF investors may be overpaying because of their investment choices.” They added: “ETF investors may be better off, as a whole, by concentrating on a relatively small set of low-fee, high-liquidity funds.”
These findings highlight the importance of fund selection. Before investing, you should carefully consider an ETF’s fund construction rules and what exposures the fund provides to the desired factors (a good tool is available atPortfolio Visualizer) as well as the expense ratio, AUM and bid-offer spreads. You don’t want to pay excess fees for what is really just market beta, which can be obtained more cheaply in total market funds. Currently, the lowest cost one for U.S. investors is the Fidelity® Total Market Index (FSKAX) with an expense ratio of just 1.5 basis points. Another low-cost alternative is Vanguard’s Total Stock Market ETF (VTI), which has an expense ratio of three basis points.