Many investors implicitly assume that the price of exchange-traded funds (ETFs), vehicles that provide passive exposure to a basket of securities and real-time liquidity, stays extremely close to their net asset value (NAV). But is it really the case? This question is of rising importance for investors given that the ETF sector is now responsible for more than $2 trillion of US assets.

Doubts about the relationship between ETF prices and NAV came to prominence in 2008 in the aftermath of Lehman Brothers’ collapse. Two US ETFs offered similar exposures to out-of-favour long-term municipal bonds, but the difference in the end-of-day premiums between the two ETFs varied from about +7% to –3% of NAV. Junk bond ETFs exhibited similar behaviour. These vehicles traded more like closed-end funds than passive ETFs created and redeemable at NAV.

So, although ETFs are popular for many reasons, investors have reason to be concerned about their major weakness — mispricing. The author sought to understand the magnitude of ETF mispricing and to what extent stale NAVs are responsible for the mispricing.

The author set out to investigate whether inefficiencies in the pricing of ETFs are limited to just a few funds and a few occasions in economic history or whether inefficiencies are widespread.

To help investors understand ETF pricing better, the author’s first task was to find out whether inefficiencies, as represented by ETF price premiums relative to NAV, are common in the ETF universe. Using data from 2007 to 2014, the author surveyed the entire US ETF sector, which accounts for around two-thirds of global ETF assets.

The next task was to examine to what extent premiums could simply be due to stale pricing. Although pricing for many stocks is usually current, pricing can be out of date, or “stale,” in some securities, including illiquid securities, such as high-yield bonds, and securities traded in international markets, such as Japan, where the trading day ends before it even begins in the United States.

Instead of simply comparing ETF prices with NAVs, the author compared ETF prices with the market prices of a peer group of similar ETFs. He considered that each ETF price deviation from its peer-group average was the premium for that ETF.

Finally, the author created an active trading strategy to verify and—possibly—exploit mispricings to test whether substantial profits could be made before transaction costs.

The average premium across the ETFs is only 6 bps. So, on average, ETFs are neither underpriced nor overpriced.

However, the prices of some ETFs can deviate significantly from their NAVs: The deviations are largest—about 190 bps—in funds holding non-US or illiquid securities. In contrast, funds holding liquid US securities are priced relatively efficiently.

Stale pricing only partly explains the mispricing. Sectors such as real estate, technology, health care, and financials contain mostly US-focused ETFs, which have liquid and transparent holdings, and yet have premiums of up to 40 bps. Even after stale-pricing adjustments, ETFs with international or illiquid holdings fluctuate within a pricing band of 100–200 bps.

Active trading strategies exploiting these inefficiencies produce substantial abnormal returns of about 7% before transaction costs. The profits tend to come from international ETFs and ETFs with illiquid underlying securities, which is consistent with the author’s first findings. Diversified US equities, US Treasury bonds, short-term bonds, and commodities tend to produce only very modest abnormal returns.

Although mispricing was found to peak during the financial crisis in late 2008 and is higher during periods of high volatility, material mispricing is seen throughout the period of the study.

For more sophisticated investors, the findings may present a valuable arbitrage opportunity. If the price of an ETF is below its NAV, an arbitrageur could purchase the ETF shares, redeem them for the underlying assets, sell the underlying assets at NAV, and pocket the difference. Likewise, if the ETF price is higher than the NAV, an arbitrageur can do the reverse. The premiums amount to over $40 billion a year, or $20 billion when adjusted for stale NAVs. By comparison, the US ETF industry is estimated to earn about $6 billion a year in management fees.

But executing arbitrage strategies comes with potential difficulty and cost. This study does not account for transaction costs, which may be substantial. In international markets, time zone differences mean underlying security prices quickly render ETF prices stale, whereas in some bond instruments, illiquidity enhances execution risk. The peer-group approach suggested by the author to counter this difficulty is not conveniently accessible for most retail investors.

For most investors, these same mispricings represent a cost. Trading ETFs of illiquid securities or international securities exposes investors to the risk of poor timing when ETF and NAV prices diverge. Retail investors are generally not aware of mispricing and assume that all ETF prices stay extremely close to NAVs. This may introduce errors in their efforts to diversify optimally.

In general, in the light of this study, small and large investors alike should probably examine the ETF prices they are offered more carefully, especially in non-US and illiquid assets. The low-fee attributes of ETFs are widely extolled, but they should not obscure mispricing issues, which can substantially increase the costs of investing in ETFs.