Why ETFs Don’t Always Follow Their Underlying Market

The Study in a Nutshell

Exchange-traded funds (ETFs) are one of the most popular investment vehicles. However, there have been persistent concerns around their liquidity, especially outside of the most widely followed ETFs. A recently published Journal of Financial Economics paper sheds light on secondary market liquidity issues. Specifically, it examines how ETF prices differ to their benchmark indexes or net asset values (NAVs).

The study finds that many small ETFs do have significant deviations from their underlying asset values (tracking errors). And these tracking errors are directly linked to the ETF’s illiquidity. This impact is larger if illiquid ETFs invest in less liquid assets, even if they hold the same asset classes of the same market.

Moreover, the estimated results from the paper show that illiquid ETFs tend to reveal large absolute liquidity betas and have a positive liquidity risk premium. The annualized return due to liquidity risk is approximately 0.14%.

The results highlight that a lack of ETF liquidity is related to its expected return and variance, as well as ETF tracking errors. Also, they imply that investors investing in illiquid ETFs are more susceptible to market and liquidity risks. Instead, investors in such ETFs may want to consider investing directly in underlying portfolios or similar mutual funds.

When ETFs Lack Liquidity

In general, ETFs (for a definition, see Appendix) are traded on major stock exchanges, but their shares are created and redeemed in the primary market. This unique structure results in the existence of two prices for a single asset: (i) ETF market prices determined on stock exchanges, and (ii) net asset values (NAVs) calculated based on the value of underlying securities.

Intuitively, a no-arbitrage condition implies that the daily ETF returns and the NAV returns must be identical. Authorised participants (APs), who create or redeem ETF shares or construct the underlying ETF portfolios, are responsible for eliminating deviations. However, the arbitrage mechanism can be limited if either ETFs or the underlying assets are less liquid.

Less popular ETFs suffer from illiquidity. At the end of 2012, the top ten ETFs accounted for 61.5% of the total ETF dollar trading volume. In addition, the assets under management (AUM) of those ETFs accounted for 36% of the total AUM in the US ETF market.

For those ETFs at the other end of the distribution, a lack of liquidity means APs could find it difficult trading at desired prices at the time of setting arbitrage positions or profit realisation through unwinding. In this situation, APs in low liquidity ETFs could be reluctant to actively engage in arbitrage trading, or they could require additional returns even when available for arbitrage trading.

As a result, APs can strategically wait for a tracking error (i.e., a large arbitrage opportunity) to widen or increase the bid-ask spread to meet the additional required return on risk. This translates into higher transaction costs for investors and a delayed elimination of tracking errors. Moreover, this mispricing problem is worse if the underlying securities are also illiquid.

Data and Methodology

The paper uses all index-based passive US ETFs listed on major US stock exchanges from 2002-2012, including those that have delisted. The total sample consists of 1,307 ETFs, for which historical data on ETF prices, NAVs, index prices, shares outstanding, and trading volumes are extracted from Bloomberg. Other sources, such as the Centre for Research in Security Prices (CRSP), Optionmetrics, CFTC and Thomson 12D are used to build ETF characteristics.

Daily individual ETF liquidity is measured using the daily relative effective spread calculated from the NYSE Trade and Quote (TAQ) database. It measures the dollar cost per dollar invested. Furthermore, to examine the effect of ETF liquidity on tracking errors, the researchers analyse the return differences between ETF and NAV, NAV and index, and ETF and index. In a perfectly liquid world, all three would equal each other.

Finally, to explain the causal effects of liquidity on tracking errors, they use a threshold stock listing as an instrumental variable. A stock is classified as a threshold stock if it has an aggregate failures-to-deliver (FTD) position over five consecutive settlement days with a registered clearing agency, with trading totalling 10,000 shares or more and amounting to at least 0.5% of the total outstanding shares of the issuer (see Appendix). FTD data is collected from the Securities and Exchange Commission (SEC).

Illiquidity Increases Investor Transaction Costs

No-arbitrage conditions imply that three daily return series (ETF, NAV, and index returns) must be identical owing to arbitrage activities of APs in a frictionless market. Chart 1 shows that tracking errors exist among these return series, especially between ETFs and their NAVs (Panel A). Note that the 45-degree green line represents perfect tracking, while the red line is the actual tracking performance.

Source: Bae and Kim (2020) page 231

When checking these results against two of the most liquid US ETFs – SPY and EEM – the authors conclude that tracking errors in ETF market prices and their NAVs tend to be more severe than those in the NAV-index returns. The key, however, is that return differences (tracking errors) and illiquidity co-move over time (Chart 2).

Source: Bae and Kim (2020) page 233

Beyond simple correlations, the researchers find liquidity to be an important factor in determining the tracking error between the ETF and its NAV or its index at both the daily and the yearly level. As they write, ‘if an ETF’s average relative effective spread (illiquidity) increases by 1%, then the tracking error in ETF-index return increases by about 13%’ (p. 233).

Furthermore, they find that the lower the liquidity, the higher the transaction cost for investors – the expected transaction cost for the top 10% most liquid ETFs is 0.024%, while this rises to 0.397% for the bottom 10%. The turnover rate, which measures how an ETF is actively traded in the market, is also shown to be lower in the low liquidity funds than in high liquidity funds.

Interestingly, when looking ETF characteristics, the extent to which ETF illiquidity affects tracking errors is smaller for in-kind ETFs, especially if the fund follows a full replication strategy (Appendix). Also, ETFs either replicating US-based indexes or equity-type ETFs are less likely to have large tracking errors. Meanwhile, ETFs are prone to large deviations in return if they are not actively traded in the market and/or if they construct portfolios when underlying indexes are volatile.

Finally, the authors construct ten liquidity portfolios and ten tracking error portfolios to investigate the effect of liquidity on the ETF return. They find that illiquid ETFs are more sensitive to the change in the market return (market risk) or the market liquidity (liquidity risk). Furthermore, both market and liquidity risks are positively related to the expected ETF return. Overall, the annualized return due to the liquidity risk is approximately 0.142%, i.e. liquidity risk explains approximately 0.14% of the annual ETF returns.

Liquidity and Volatility

Similar to return differences, a lack of liquidity creates differences in the ETF return variance and the NAV return variance. This difference is larger than that between the NAV return variance and the index return variance, and it is bigger in unleveraged products than in the leveraged products.

These results imply that the price movement of the ETF is more volatile than that of the index, suggesting substantial tracking errors in the ETF with respect to the index. Also, the return variance or volatility increases when the ETF is not actively traded in the market.

Bottom Line

One potential benefit of investing in ETFs is that investors can avoid high transaction costs when ETFs track inaccessible markets or assets that are difficult to track. The results from this paper, however, show that illiquid ETFs could be riskier investments than investing directly in underlying assets owing to the high tracking errors. Investing in illiquid ETFs with severe tracking errors could offset the benefits of ETF investments.

Appendix – Further Information on Concepts Discussed in the Paper

What are ETFs?

An exchange-traded fund (ETF) is designed to provide an indirect investment opportunity for particular markets, countries, or industries by replicating a specific representative index. Index-based ETFs allow investors to access foreign markets or a variety of asset classes at relatively low transaction costs.

US ETFs, introduced in 1993, have grown significantly in recent years. At the end of 2016, their market size was about $3 trillion, accounting for nearly 30% of the dollar trading volume and 23% of the share volume in US stock markets. ETFs have grown to become the most popular asset class for many institutional and individual investors.

ETFs are fundamentally similar to, but are not, mutual funds. They are structured, managed, and regulated just like traditional mutual funds. Different from conventional open-ended mutual funds, ETFs are traded continuously on regular exchanges, like regular stocks are.

In addition, ETFs are similar to closed-end funds (CEFs) in that they are traded on exchanges. Unlike CEFs, the total number of shares can be increased or decreased depending on market demand and supply. In other words, ETFs are designed to combine the creation and redemption process of open-end funds with the continuous trading of the CEFs.

Tracking Underlying Indexes And Securities

In the ETF market, APs play an important role in keeping the series of ETF, NAV, and index returns close to each other. The return differences are typically removed by the arbitrage activity of APs. APs or market makers keep ETF prices in line with the values of their underlying portfolios by trading both ETFs and underlying securities simultaneously, that is, the so-called creation-redemption process.

For instance, if an ETF price is lower (higher) than its NAV, then APs buy (sell) ETF shares and sell (buy) the basket of securities. If the current market price of an ETF becomes higher than its NAV, APs buy underlying securities to form a creation unit and deliver it to the ETF provider. After receiving the ETF shares from the ETF issuer, APs sell these ETF shares to the market.

How Do AP’s Get Into Trouble?

Arbitrage activity would be possible when APs are able to trade ETFs or underlying securities immediately and limitlessly. APs can get into trouble by constructing the basket of securities or by trading ETFs if ETFs or their underlying securities suffer from the lack of liquidity.

Depending on the market conditions, observed ETF prices are frequently different from their announced NAVs. In other words, the lack of liquidity or low trading volume in the ETF market could lead to a large price impact or the presence of stale prices, thereby causing a price gap between the ETF and its NAV.

As a result, APs could bear unwanted costs related to borrowing underlying securities or holding inventories to manage the ETF market, if ETFs or underlying securities are not fully liquid. This situation implies that the lack of liquidity in the secondary market causes an increase in trading costs and ETF tracking errors.

FTD’s

The FTD is a situation in which a market participant does not deliver the underlying security he or she sold at the time of settlement. Market makers or APs can strategically decide FTDs to reduce transaction costs when a demand shock occurs in the ETF security. These strategic FTDs are particularly effective for ETFs with high creation fees or illiquid underlying basket securities.

Fund Characteristics and Replication Strategy

Bloomberg categorizes ETFs into in-kind, cash, and cash and in-kind according to the create or redeem method and into full replication, optimized, and derivatives-used according to the replication strategy. In the case of in-kind creation or redemption, APs must construct and deliver all underlying assets of relevant ETFs for full replication, some underlying assets for optimized replication, and index-based derivatives for derivatives-used replication.

To view the full paper, please click here