The first U.S. exchange-traded fund (ETF)—the SPY based on the S&P500—began trading in 1993. Since then, the number of such funds has grown dramatically, so that by mid-2016 there were more than 1,600 ETFs on U.S. exchanges valued at roughly $2.2 trillion. This means that ETFs are now roughly one-sixth the size of open-end mutual funds. And, with this ETF growth has come a broadening in their scope and character. Today, there are ETFs that include less liquid assets such as corporate bonds and emerging market equities, and there are funds that provide inverse or leveraged exposure to the underlying assets.
Given these trends, it is no surprise that ETFs have attracted regulators’ attention (see, for example, here and here). Should they be concerned? Is this a consumer protection issue? Do ETFs contribute to systemic risk? Or, is their design stabilizing? Might financial stability even be served by the conversion of all open-end mutual funds into ETFs?
In this post, we provide a basic description of ETFs. On balance, we see large benefits from the availability of the most popular unleveraged ETFs that pool relatively liquid instruments, but note a range of concerns. To date, academic inquiry into ETFs has been limited, so we hope that future research will help identify potential opportunities for useful expansion as well as fragilities that merit greater regulatory scrutiny.
Let’s start with the basics. The following charts show the value of ETFs and mutual funds since 2000. Over the past five years alone, ETFs outstanding have more than tripled, while comparable mutual funds grew by 38%. As a share of global ETFs, the U.S. industry accounts for more than 70% (see Figure 3.1 in the 2016 Investment Company Fact Book).
Exchange-traded funds (Billions of U.S. dollars), 2000-1Q 2016
Source: Financial Accounts of the United States 1Q 2016, table L.123, Federal Reserve.
Open-end Mutual funds (Billions of U.S. dollars), 2000-1Q 2016
Source: Financial Accounts of the United States 1Q 2016, table L.122, Federal Reserve.
In terms of trading volume, activity in U.S. stock ETFs recently has accounted for about 30% of the value of stocks traded (see chart here). And the five most popular ETFs all traded more actively over the past three months than any individual stock. While these popular funds all had an equity focus, they tracked very different asset categories, including broad indexes of the U.S. stock market and of emerging markets, gold miners, and the VIX futures. Notably, options on ETFs (especially the largest equity and volatility measures) have come to dominate CBOE options trading, accounting for 70% of the total in June 2016 (see here).
So, what are these funds? ETFs are pooled investment vehicles with properties of both open-end and closed-end funds. As the label indicates, ETFs trade like other securities on an exchange, so that—in normal circumstances—they can be purchased or sold through a broker whenever the exchange is open. Similarly, like other securities, traders can go short ETFs; they can leverage their positions (subject to the standard margin requirements); and they can use options on ETFs to alter their risk exposure. Most (but not all) ETFs are transparent, disclosing their holdings daily. Also for the most part, ETF holdings are taxed only when they are sold. Finally, ETFs on large U.S. indexes often trade more frequently than their underlying instruments, supported by low transactions costs. For example, a typical bid-ask spread on SPY was only 0.01%. However, not all ETFs have low expense ratios (see here). And, one must also add in broker commissions and ETF price deviations from the value of underlying assets. The latter can be large for assets that do not trade during U.S. market hours or for U.S. assets during periods of market stress (see, for example, here).
What distinguishes ETFs from open-end mutual funds and from closed-end funds? First, mutual funds do not trade on an exchange. They can only be redeemed from the vendor, with the redemption occurring once a day at the estimated net-asset value (NAV) of the fund as of 4pm. When there is a significant imbalance of sales and purchases in a mutual fund, the vendor can purchase or sell the underlying securities in order to meet demand for additional investment or for redemption of existing holdings. That makes the supply of a mutual fund highly elastic.
But, as we emphasized in an earlier post, there is a downside to the open-end character of certain types of mutual funds. For those holding illiquid assets, like corporate bond funds, following a negative shock, investors who exit early benefit when the fund manager’s subsequent portfolio readjustment imposes illiquidity costs on those that remain. This first-mover advantage means that an entire class of mutual funds could be subject to runs.
Closed-end funds issue a fixed number of shares backed by a portfolio of assets that typically is actively managed (in contrast to the passive management of many ETFs). Like ETFs and other securities, their shares trade on exchanges, with the fund’s price determined by supply and demand. As a result, in contrast with mutual funds, the price of closed-end funds can, and frequently does, vary significantly from the value of its constituent assets. And, because there is no redemption promise, there is no run risk.
ETFs constitute a hybrid between these two ends of the pooled investment spectrum. Like closed-end funds, shares trade on an exchange (even though they are legally organized as open-end funds). But ETFs typically have a fixed underlying basket of assets. They also have a special feature that—in contrast to closed-end funds—makes their supply elastic and usually keeps their market price reasonably close to their NAV. A special class of investors—called “authorized participants” (APs)—has the right to create new ETF shares or to redeem old ones by delivering to or receiving from the ETF vendor the specified mix of underlying assets that constitute the ETF in exchange for shares. The arbitrage activity of the APs in the “primary market” adds to or subtracts from ETF supply when the “secondary market” price (at which most ETF market participants are trading) deviates substantially from the net value of the assets that it represents.
What this means is that, so long as liquidity in the underlying assets backing the ETF is sufficient, the APs will execute an arbitrage that keeps the market price of the ETF close to the NAV. Importantly, this requirement is not met for all assets at all times. For example, emerging market assets may not trade when U.S. markets are open. More importantly, during periods of market stress, it may become difficult to trade underlying assets that are usually liquid. When this happens, investors wanting to sell their shares may have to settle for prices below the NAV. That is, they may have to sell at discount.
So, what are the concerns about ETFs? They include a range of issues. We’ll focus here on three: (1) induced volatility; (2) the risks associated with ETFs based on illiquid assets; and (3) leverage and the use of derivatives.
Ben-David, Franzoni and Moussawi argue that highly liquid U.S. stock index ETFs are a source of greater non-fundamental volatility in the underlying stocks because they attract less-informed retail customers. The idea is that, by moving the price of the ETF, uninformed trading induces arbitrage activity in the underlying stocks that boosts volatility and can lead to mispricing.
Bhattacharya and O’Hara argue that—when it is not possible to synchronize ETF prices with the value of the underlying assets—ETFs can cause herding in a way that amplifies aggregate disturbances. This problem applies generally to ETFs with illiquid assets and can become acute when trading in the underlying assets is halted in one location (think of Greek or Chinese stocks in 2015) while ETF trading continues elsewhere. The concern is that the tail (the pricing of the actively traded ETF) wags the dog (the pricing of the underlying securities).
Of course, this picture is not one-sided. Not only do ETFs make illiquid assets more accessible to investors, they also may provide a means to manage exposure in periods of stress. For example, when domestic trading in Greek and many Chinese stocks halted in 2015, investors could still buy or sell ETFs based on these assets. Even if its price only reflects aggregate, rather than idiosyncratic shocks affecting the underlying assets, the ETF offers a potentially useful, if imperfect, insurance mechanism. (There is some evidence that ETFs do facilitate price discovery in periods when the underlying instruments become illiquid.)
Finally, there is leverage and, closely related, the use of derivatives. Leverage is an obvious source of systemic risk. Fortunately, U.S. investment companies, including both mutual funds and ETFs, are limited by the 1940 Act (Sec. 18 (f)) to borrowing that is no more than one third of their assets. This is far less than the leverage of typical financial institutions (U.S. commercial banks’ liabilities are about eight times book equity). And many ETFs, especially the most widely held and traded ones that focus on U.S. broad equity indexes, are unlevered. Perhaps more important is that ETFs may create leverage by using derivatives. This has attracted regulators’ attention, and the SEC has proposed a rule that would tighten leverage limits from three times to 1.5 times. (Synthetic ETFs, which are more common in Europe than the United States, raise a host of additional issues. See the discussion here.)
Of course, there are still other key questions that face investors and policymakers alike. At the top of our list is how ETFs perform in periods of stress. Experience suggests that there is considerable heterogeneity. For example, according to an SEC study of the August 24, 2015 U.S. stock market disruption (when the S&P500 index temporarily fell by more than 5%), exchange-traded products (ETPs—an asset category that includes ETFs as the largest subset) were relatively prone to a surge in trading and to extreme volatility: 19.2% of the funds experienced a price drop of more than 20%—large enough to trigger one or more trading pauses. Even so, among the ETP group most affected by the event (those focused on U.S. equities), the majority of funds did not experience a trading halt.
The uncertainty about ETF performance and the potential to behave like closed-end funds when arbitrage becomes costly or difficult, raises questions of suitability, disclosure, and investor education. While ETF holders may be relatively wealthy and willing to accept risk (see figures 3.12 and 3.13 here), it remains unclear whether retail investors understand how far ETF prices can deviate from NAV in periods of stress. If, for example, they place a typical market sell order just prior to a period of stress, they end up selling at a large discount. Moreover, it is doubtful that most retail investors need the intra-day liquidity of ETFs—the daily liquidity provided by open-end mutual funds is almost surely enough. Indeed, one leading advocate for retail investors, John Bogle—the founder of Vanguard—has expressed serious doubt about the suitability of ETFs (other than those based on broad indexes) for traditional buy-and-hold investors.
Where does this all leave us? We see a range of benefits from ETFs, especially those focused on broad U.S. equity indexes. During normal times, they offer liquidity (at a low cost for the broad equity indexes). In stress times, when the underlying assets cannot be traded, they operate like closed-end funds, which are not prone to runs.
This leads us to ask the following intriguing question: would the financial system be more stable if all open-end mutual funds were to mimic the structure of ETFs so that they would trade on exchanges as closed-end funds when the underlying assets become illiquid? Put differently, to the extent that some ETFs pose systemic risk, might it be less than the risks posed by open-end mutual funds holding illiquid assets that are close substitutes? We don’t know, but it is surely something worth thinking about.
Note: This version corrects the original description of closed-end funds, making clear that these usually are actively managed.