How To Minimize Your Cost Of Trading ETFs

June 22, 2009

IllustrationOne of the features that long-term investors like most about exchange-traded funds is that each shareholder usually pays only the cost of his or her own fund share transactions and is protected by the ETF structure from the cost of other investors’ purchases and sales of fund shares. In spite of this feature, there is a great deal of misinformation in circulation about the cost of trading ETFs and how to trade them efficiently. This article tries to help you evaluate the quality of ETF markets and keep your ETF share trading costs low.

Trading ETFs Is Different Than Trading Stocks

A natural reaction to the title of this article is, “Why bother to write about trading ETFs? They trade “just like a stock.”

In fact, the reason for writing this article is that ETFs don’t trade like stocks. ETF markets behave differently than stock markets, and ETF shares trade differently than stocks in a number of ways.

Figure 1 lists some of the similarities and differences. To illustrate one difference, ETF investors have learned that ETF bids and offers usually change much more frequently than stock bids and offers. These more frequent changes can make ETF investors uneasy, as they suggest that individuals do not have as much information as professional traders have about what the ETF price should be or what the ETF bid or offer will do next.

Even if your only interest is in day-trading one of the several dozen major benchmark index ETFs that trade more than 10 million shares on an average day, the market in ETFs is different than the market in common stocks in ways that can affect your trading results. If you trade ETFs that are not based on major benchmark indexes or that trade fewer than 10 million shares a day, ignoring the significant differences between ETF and stock trading can be very costly.

This article will help investors find level stretches of the ETF playing field where professional and amateur traders have equal footing—and avoid the slippery slopes. Once you understand how the ETF market works, and how trading ETFs differs from trading stocks, you will be able to trade ETF shares confidently and efficiently. In fact, with the introduction of net-asset-value-based trading in ETFs, trading ETFs can be much simpler and less stressful than trading stocks.1

While the focus of these comments is on U.S.-listed ETFs holding U.S. common stocks, many of the observations apply to European and Asian ETF markets and to U.S. ETFs with other portfolio holdings.

The entry in Figure 1 that generates the most questions from investors is the trading hours for ETFs. Trading after 4:00 p.m. presents both opportunities and hazards to ETF traders. Trading in ETFs is active until well after 4:00 p.m., partly because the major benchmark index ETFs (such as the S&P 500 SPDRs, the iShares Russell 2000 Index Fund and ETFs tracking a number of other popular benchmark indexes) are part of major arbitrage complexes. These arbitrage complexes typically consist of an index that serves as a template for index mutual funds and other indexed portfolios and for a variety of index derivative financial instruments, including index futures contracts, index options and, of course, index ETFs. The arbitrage complexes also include derivatives on these derivatives, such as options on futures, options on ETFs and securities futures products (single “stock” futures) on ETFs. There are also exchange-traded and over-the-counter structured products and risk management contracts linked to many of these indexes.

Figure 1


Until the NYSE acquired the Amex in 2008, ETFs—like most of the other tradable components of the index arbitrage complexes—traded in a regular session that lasted until 4:15 p.m. to provide a structured ETF market that was fully contemporaneous with the futures markets. Since the change to a formal close at 4:00 p.m., ETF volume between 4:00 p.m. and 4:15 p.m. has not changed materially. On May 15, 2009, the day before this article was prepared for submission, ETFs were four of the five most active issues in after-hours trading. One of that day’s most active after-hours ETFs is based on an index that does not have an active futures contract.



ETF Intraday Net Asset Value Proxies

When trading began in the first ETF introduced in the United States (the S&P 500 SPDR launched in 1993), the Securities & Exchange Commission (SEC) required that the sponsors of the SPDR arrange for dissemination of an intraday share value proxy for the SPDR at 15-second intervals. These proxies are usually called indicative optimized portfolio values (IOPVs). The requirement for publishing these values was extended to every domestic equity ETF launched since 1993 and, with modifications, to ETFs holding foreign equities, fixed-income instruments and other financial instruments.

In spite of improvements in trading data calculation technology and the introduction of ETF portfolios that hold infrequently traded securities, calculations of these intraday value proxies are still based on the most recent trade of each portfolio component. Matt Hougan discusses the inadequacy of last-sale value proxies in an article scheduled to appear in the July 2009 issue of ETFR. These intraday value proxy calculations are made and/or disseminated by the National Securities Clearing Corporation (NSCC) and other service providers to “support” intraday trading of ETFs.

Professional ETF traders and market makers do not use the “official” every-15-second proxy value calculations to help them determine their ETF bids and offers. Professionals develop their own valuations or subscribe to real-time ETF value calculations based on contemporary bids and offers rather than last sales. The fact that they do not use the free IOPV does not mean that these professionals lack faith in the NSCC’s ability to calculate correct values. The simple facts are that the last sale is not a reliable indicator of contemporary values in most market situations, and the 15-second interval between valuations is too long for the values to be useful to a trader.

Because an ETF is a derivative security, its current value changes every time the value of any component of the ETF portfolio changes. The ETF value proxies used by professional traders are calculated from the midpoint of the bid and offer for each position in the ETF portfolio. Sometimes the value calculations made by and for professionals use the size of bids and offers and the pattern of “changes” to forecast short-term trends. Figure 2 illustrates how a naive investor might attempt to use the “free” intraday proxy information to develop a bid or offer for ETF shares.

In Figure 2, the latest IOPV is represented by a dot at the beginning of each of the five 15-second intervals illustrated. Investors might place limit orders at prices close to the most recent per-share value proxy. Columns A through E illustrate how bids and offers entered at equal distances, respectively, below and above a sequence of these every-15-second net asset value (NAV) proxy calculations might become transactions—but not always the transaction that the investor entering the order hoped to achieve. An offer to sell the ETF’s shares slightly above the proxy value posted at the beginning of time interval A would probably be lifted as the fund portfolio value rose during intervals A and B. That offer was below the changing per-share proxy value by the time that value was updated at the beginning of interval C—less than 30 seconds after the order was entered. The offer was also below the likely bid in the market at that time. Of course, some of the last-sale prices used to calculate each 15-second proxy might be more than a few minutes old at the time the calculation was made. Using the free 15-second values can be costly, but even if investors had access to proxy values based on every-15-second midpoints of bids and offers and could enter orders as soon as the value was published, a lot can change before or shortly after the next proxy value is published.

Figure 2


In my experience, many investors are aware of the existence of the every-15-second last-sale NAV proxy value, but few know how to find it for a particular ETF and even fewer think about how, if at all, to use it. It is probably a good thing that these proxies are not widely used. Any attempt to use them to manage an order is more likely to lead to disappointment than to a good execution. The information on share values, transaction prices, bids and offers summarized in Figure 1 indicates that ETF bid and offer updates are published more frequently than every 15 seconds (every time portfolio component bids or offers change materially), making the ETF share bids and offers a much better indication of the current market for an actively traded ETF than the every-15-second last-sale proxy calculation could possibly be.

The posted ETF bids and offers also have the advantage of being something you can trade with. You cannot trade with the every-15-second NAV proxy because it does not represent a bid or offer for the ETF shares. An investor can be confident that, even if his market data vendor is a bit slower and updates quotes less promptly than the best data vendors, he will not be seriously disadvantaged relative to other retail market participants. Bids and offers for the most actively traded ETFs tend both to be tighter and to change more frequently than stock quotes during active trading periods. In active trading periods, spreads are frequently as narrow as a penny per share between the bid and offer for some of the most actively traded ETFs. These periods of active trading are the best time to trade ETFs. However, if trading is not active, the quotes in the market tend to reflect a wide spread between the bid and the offer.

The Brave New World Of High-Frequency Electronic Trading

Few individual investors have the kind of information professional traders use or the capability to change their bids and offers as fast as the quotes on an ETF share change in common market situations. Professional traders and market makers not only calculate intraday bid/offer values for ETF shares continuously throughout the trading day, they use automated quote management systems that can change their ETF bids and offers in a millisecond or so (a millisecond is 0.001 of a second) every time the bid or offer changes for any security in the ETF’s portfolio. The fact that regular-session ETF trading volume usually exceeds 2 billion shares per day is partly due to the speed of order entry and execution to capture small changes in value. The quest for speed of execution has led the Nasdaq market to boast “peak [round-trip] trading speeds of 250 microseconds” (a microsecond is 0.001 of a millisecond). In this high-speed environment, time lags associated with information traveling even a few hundred miles at approximately the speed of light confer a premium value on computer centers located within a few yards of an exchange order-matching system.

If you have access to an institutional equity trading desk, ask to see an active montage monitor of the best bids and offers from all the markets trading SPY or QQQQ during regular trading hours. I have not included a static example of the montage monitor because a snapshot cannot convey a sense of the frequent and rapid changes in bid and offer prices and sizes. I have not seen comparable displays from any other activity except, perhaps, the last digit in an electronic sign showing an update of the estimated U.S. population. The census sign has only one or two high-frequency flashing digits. The montage monitor looks like a time-lapse video of a beehive. An individual investor can benefit from high-speed trading developments to the extent that they compress bid/offer spreads in actively traded ETFs. Investors using actively traded benchmark index ETFs trade in a very efficient market thanks to high-speed trading.

Auto-quote systems eliminate most of the need for hands-on attention by a market maker’s or professional trader’s staff. Consequently, even though the bid/offer spread is usually wider in a less actively traded ETF, the quote updates in the less active shares will still be more frequent than quote updates in a similarly active common stock. The frequency of quote changes is linked to the number and price volatility of positions in the ETF portfolio as well as to trading activity in the ETF shares. While the quality of the quotation services available to investors varies, it is usually possible to get a current bid and offer as well as the quantity bid for or offered for an ETF of interest. Because the best bid and offer are more useful than the every-15-second NAV proxy (that you cannot trade with anyway), the size of the best bid and offer and the spread between them are the best indicators of how many shares you can trade easily and at what price you might expect to complete a transaction. This bid and offer information is key to effective trading in the conventional ETF market.

Figure 3 shows ETF share volume by 15-minute intervals for a recent week of trading. The relatively heavy trading volume in ETFs in the first half-hour of trading both attracts and reflects retail ETF orders. A similar volume pattern in stock trading is usually attributed to institutional trading.2 There is substantial evidence that stock prices are, on average, very slightly lower in the first half-hour of trading and very slightly higher in the last half-hour of trading than over the balance of the trading day. This daily price pattern is statistically significant and is usually attributed to an artifact of institutional equity trading practices. It may not be economically significant for ETF traders, however.3 Traders who are concerned about their trading costs, especially the bid/ask spread and the ability to trade in size with minimal market impact, will usually wait until the markets in all the ETF portfolio components are open and updated quotes are available. Spreads on ETF shares tend to be relatively wide right after the opening, when the sizes of market-maker bids and offers are small. Trading is usually less costly later in the day. As illustrated in Figure 3, ETF trading volume is highest in the first hour or so of trading and in the last hour of trading.

A number of organizations and Web sites publish information on “average” bid/offer spreads for specific ETFs. Most of this information is based on data provided by NYSE Arca.4 Depending on what time of day you check the bid and offer prices and sizes for a specific ETF, you will probably find that the published “average” spread is narrower than the spread you see in actual quotes. The spread you observe is likely to be wider than the “average” spread because the published “average” spreads are weighted by the size of the bids and offers available at various times during the day. This weighting scheme means that a heavier weight is assigned to the spread at times when bid and offer sizes are larger. Larger bid and offer sizes usually coincide with times when trading volume is highest and spreads are tightest—times near the market close.

Trading in many ETFs is active between 4:00 p.m. and 4:15 p.m., but much of this trading is linked to futures markets, and bids and offers are often erratic. This trading is apparently not included in the data used to calculate the average spread. The period between 3:00 p.m. and 4:00 p.m. is generally the period when the cost of a conventional intraday trade in an ETF is lowest. This period of high volume and large bid and offer sizes largely determines the published “average” spreads. This is when the average spread on the S&P 500 SPDR (NYSE Arca: SPY) will be less than a penny—temporarily “locked” markets in the most actively traded ETF shares are common. Even if you are planning to trade shares in one of the most actively traded benchmark index funds that appear regularly on the most active list, the end-of-day period is almost certainly the best time to enter your ETF order, assuming that trading cost minimization is a significant objective of your trading plan and you decide to use the conventional ETF trading process.

ETF Trading Volume Is Huge, Growing And Highly Concentrated

The growth of trading volume in the most actively traded ETFs has been nothing short of phenomenal, as Figure 4 shows. ETFs now account for more than 20 percent of U.S. equity trading volume by shares, typically more than 2 billion shares per day; weighted by dollar volume, the number is significantly higher, and has been reported as high as 35 to 40 percent on certain days.

Most of the increase in ETF volume has been in the most actively traded funds. Typically, half of the most active “stocks” each day are ETFs. The most actively traded funds are often not the most attractive investments, however.

In any event, you do not necessarily have to pay a wide spread to trade a less active ETF. Beginning with some useful “rules of thumb” for intraday ETF trading, let’s look at how ETF investors can reduce their transaction costs, particularly on less actively traded ETFs.

How To Trade ETFs Efficiently

If (1) you are trading one of the major benchmark index ETFs that trades more than 10 million shares a day; (2) the current price of the shares is consistent with your objectives; and (3) the quote spread is close to the minimum of $0.01 per share, entering a market order is generally a safe choice. However, you may want to compare the size of your order with the quoted size on the other side of the market before you push the button to execute a market order.

If the thought of entering a market order in a volatile market environment is unsettling, you can enter a marketable limit order; that is, an order to buy at the offer price or sell at the bid price currently posted in the market. This order will usually be executed in full as long as the quote has not moved away from the limit on your order by the time your order reaches the market. Given the rapid changes that are often characteristic of ETF bids and offers and the heavy volume characteristic of the last hour of trading, there is always a risk that your limit order will not be a marketable order when it reaches the market and, consequently, it will not be executed. You should compare the opportunity cost of failing to execute to the possibility of a worse price with a market order.

If an ETF trades less than a million shares per day, take a close look at the bid/offer spread, at the size of the contemporary bid and offer, and at recent changes in the bid and offer; then consider a marketable limit order—or read on for more analysis and more options.

Most commentary on the cost of trading securities suggests that the appropriate way to measure the cost of the bid/offer spread in a purchase or sale is to assume that your cost of trading will include half of the spread on the purchase and half of the spread on the sale. That is a reasonable rule of thumb when you are trading common stocks in small size. However, it is not safe to assume that an inactively traded ETF’s current value is between the bid and offer in the intraday market. Most investor orders to buy or sell shares of an ETF on a given day will be on the same side of the market. If a fund has just been introduced, has enjoyed favorable commentary in the financial press or is being actively purchased by advisers for their clients’ accounts, most investor orders will probably be on the buy side for days or weeks at a time. In contrast, if a particular market segment is out of favor or a fund has under-performed its peers, the predominance of investor orders for a fund will probably be on the sell side for long periods.



Figure 3

For very actively traded benchmark index ETFs (where the spread during the last hour of trading will typically be a penny, with large quantities available on both sides of the market), the location of the midpoint of the bid and offer will nearly always be close to the fair value of the ETF. Arbitrage forces and heavy trading will ensure their closeness. In the case of less actively traded ETFs where cross-market arbitrage forces do not provide much pricing discipline, the midpoint of the spread will reflect the supply/demand pressures of investor purchases and sales of the ETF shares much more than the prices in the underlying portfolio securities markets. If you are an ETF investor trying to make the same trade as other investors, you should expect to pay more than half of the posted spread on most of your trades in less actively traded ETFs.

Figure 4

If an ETF trades less than 100,000 shares a day, investor supply or demand may move the bid/ask range so that it does not even encompass the contemporary share value. In other words, the bid may be above a contemporary NAV calculation, and the spread to the NAV for a purchaser of the shares may be greater than the posted bid/offer spread indicates. Arbitrage forces are undependable when the potential for aggregate arbitrage profit is small due to lack of volume.

If you are interested in an ETF that trades fewer than 500,000 shares a day, don’t consider anything other than marketable limit orders when you are trading in the intraday ETF marketplace. If your order is larger than the number of shares quoted at your limit, expect to spend some time working the order—at best. If you are an adviser trading ETFs for a number of accounts, your broker may give you access to an algorithmic trading model that manages bids and offers relative to changes in the bids and offers for the securities in an ETF’s portfolio, much like a market maker would use.5 Even better, your broker may arrange a dialogue with a market maker in the ETF’s shares. Working with a market maker on a large transaction is usually a very good idea. The market maker can trade at lower risk if he is filling an order rather than guessing what an anonymous bid or offer might mean. The probability of repeat business with you may also favorably affect the terms of a trade.

After we take a close look at risks and costs of market-on-close (MOC) orders in ETFs, we will examine a new kind of trading in which orders can be entered for execution at or relative to the closing net asset value of the ETF. This new trading method will be particularly useful to ETF investors who want a good execution without spending a lot of time on order management and who have had unfavorable experiences with the intraday ETF market or with MOC orders for ETFs. This new trading method is important because ETF trading volume is increasingly concentrated in the most actively traded ETFs, while the most attractive investments are often in the less actively traded ETFs with wider bid/ask spreads. A trading mechanism that narrows spreads and reduces total trading costs on less actively traded ETFs can change the ETF landscape in many important ways.


Market-On-Close Transactions In ETFs Can Be Surprisingly Costly

There is a great deal of misunderstanding about how MOC transactions in ETFs work. This section probably contains a great deal more information on these transactions than most investors will want, but any investor who considers using ETF MOC orders will find a careful reading of this material worthwhile. The principal message of this section is that a MOC execution in an ETF will not necessarily be priced at or even very close to the midpoint of the indicative bid and offer published on the fund’s Web site and in its prospectus.

Reported ETF Premium And Discount Pricing

As a prelude to an examination of MOC orders, it is important to understand what the information on premiums and discounts published in fund prospectuses and on fund Web sites means. ETF issuers collect information on ETF share bids and offers on the listing exchange market each day at 4:00 p.m. and compare the midpoint of these quotes with that day’s NAV calculation for the ETF. Premium and discount tables or graphs reflecting these comparisons are published in ETF prospectuses and annual reports. They give investors and the SEC inappropriate comfort that end-of-day ETF transactions occur very close to NAV. Market makers in even the most thinly traded ETFs understand that the midpoint of their daily 4:00 p.m. quote will be preserved in prospectuses and on ETF Web sites for years. Market makers have a stake in drawing traders to the ETFs they trade. Consequently, they monitor their real-time bid/offer NAV calculations very closely as 4:00 p.m. approaches. Even if they have to widen their spread for a few seconds, they will work to get the midpoint of their bid and offer as close to the expected 4:00 p.m. NAV6 as possible. Their 4:00 p.m. quote is the most widely scrutinized and least useful bid/offer of the day.

Publication of premium and discount information based on 4:00 p.m. ETF share quotes has led to overuse of MOC orders, especially for some ETFs that are thinly traded. Most investors do not realize that MOC transactions in ETFs are not reflected in most ETF-reported premiums or discounts in any way. Nonetheless, MOC orders are often used by individuals and defined contribution retirement plan investors who are accustomed to mutual fund trades at net asset value and to MOC orders on stocks.

MOC Orders For Most ETFs Are Not Likely To Be Executed At Or Even Near NAV

ETF market-on-close orders often result in executions at a much greater distance from the fund’s end-of-day NAV than the reported premium and discount data leads investors to expect.

Figure 5

Market-on-close orders both in stocks and ETFs are integrated with the limit order books for these securities. The hypothetical schedule of bids and offers (limit orders) for an ETF at the end of the trading day displayed in Figure 5 will help illustrate how this integration works. In this limit order book, the best bid is for 2,000 shares at $24.90, and the best offer is at $25.10 for 2,000 shares. The MOC book will operate alongside this limit order book with buyers and sellers entering market orders of various sizes for execution at the close. If the balance of the MOC orders is to buy 4,000 shares of the ETF at the market-on-close and the limit order book matches the table, all the MOC orders will be filled at $25.30, unless a market maker or a last-minute customer order improves on the $25.30 offer. The lowest price at which an order to buy 4,000 shares can be filled is $25.30, and market rules require that all 4,000 shares trade at that price.

Until the official trading close for ETFs was changed from 4:15 p.m. to 4:00 p.m., MOC orders for ETFs were not subject to the same rules as MOC orders on stocks. Now, all MOC orders are accepted until an exchange-specified cutoff time for such orders, usually 3:40 p.m. After that time, orders to trade at the market-on-close will be accepted only on the side of the market that will reduce any trade imbalance. Specifically, if the balance of market-on-close orders is to buy 4,000 shares of XYZ and the limit order book looks like Figure 5, additional MOC buy orders will not be accepted after 3:40 p.m. Market-on-close sell orders will be accepted to reduce the imbalance. Regular trades will interact with the limit order book until 4:00 p.m.

These MOC rules work for very actively traded ETFs because active trading in index instruments attracts arbitrageurs. Market-on-close orders for less active ETFs often lead to trades far from NAV because the order book is sparse and because market makers tend to widen spreads at 4:00 pm. Less actively traded ETFs are not subject to continuous monitoring by arbitrage-motivated traders.

In our numeric example of ETF market-on-close trading, the net asset value of the fund is $25.00. The midpoint between the bid and offer at 4:00 p.m. is also $25.00. Yet barring a last-minute MOC sell order or a new limit order to sell at least 2,000 shares at less than $25.30, market-on-close orders will be filled at $25.30, 1.2 percent above the net asset value. The ETF’s premium/discount calculation for the day will show the “market price” based on the bid and offer at 4:00 p.m. matching the net asset value at 4:00 p.m. Publication of a zero premium or discount based on 4:00 p.m. quotes relative to NAV encourages ETF investors who do not understand the transaction mechanism to use MOC orders incautiously.

While the premium/discount information published for ETFs is calculated as the prospectus says it is, this calculation has led to unanticipated results for many investors. Even if some professional investors find market-on-close ETF executions useful at times, I can think of no reason why a typical investor should use an MOC order to trade ETFs. For actively traded ETFs, the intraday market in the last hour of trading operates well, and spreads are among the tightest of any time during the day. For less actively traded ETFs and for investors who want a more easily managed trading option, the NAV-based market—described at length in the next section—will usually deliver executions closer to and more consistent with net asset value than either an intraday or an MOC execution. The availability of NAV-based secondary market trading will assure that an investor can lock in a price related to a specific net asset value calculation. Investors can access trading liquidity available during the entire period that the NAV-based secondary market is open—a full-day trading session, at minimum.

If defined contribution plans like 401(k)s and advisers accustomed to buying and selling mutual funds at net asset value use MOC orders for ETF transactions, the deviation of market-on-close executions from NAV may increase as these plans and advisers make greater use of ETFs, especially less actively traded ETFs.7

Introducing NAV-Based Trading In ETFs

As noted in the opening paragraph, one of the compelling advantages of most ETFs for long-term investors is that each shareholder pays only the cost of his or her own fund share transactions and is protected by the ETF structure from the cost of other investors’ purchases and sales of fund shares. Net asset value-based trading in ETF shares preserves this protection from other investors’ fund share trading costs and enables investors to buy and sell ETF shares at a price related to—and no further than a predetermined distance from—net asset value. We will see that, in contrast to trading cost uncertainty in the intraday trading of ETF shares “just like a stock,” NAV-based trading makes it possible for investors to know and control their ETF transaction costs with minimal order monitoring.

Entering an order to buy or sell ETF shares at or relative to the current day’s NAV is only superficially like entering a limit order to buy or sell shares in the traditional intraday ETF market. Buy and sell limit orders are entered and executed relative to a proxy of 100.00 for the per-share net asset value that will be calculated based on the value of the ETF’s portfolio at 4:00 p.m. A transaction at net asset value plus 1 cent per share would be recorded at 100.01. If the fund’s NAV for the day turns out to be $20 per share, the 100.01 would translate to a price of $20.01 because each .01 translates into $0.01 (1 cent) per share. (We omit $ signs on the proxies to avoid the implication that the transaction will occur near $100 per share; 100.00 is merely a reference point.)

Most investors will want to check current bids and offers in the NAV-based market before entering orders in the conventional intraday market. The NAV-based quotation might be stated as “10,000 Bid at 99.99, 20,000 Offered at 100.01” reflecting buy orders totaling 10,000 shares at one penny below the end-of-day net asset value, and sell orders totaling 20,000 shares at one penny above the end-of-day net asset value for the ETF. It may be possible to buy ETF shares below the NAV or to sell them above the NAV, depending on the bids and offers available in the market over the course of the trading day. Even if a standing limit order to sell below NAV is not available in the marketplace when an investor checks the quote, an investor can place his own bid below net asset value at the market close, and that bid may be hit by an order to sell during the course of daily trading activity. Limit orders can be cancelled and new limit orders or market orders can transact with the limit order book until shortly after the close of traditional ETF trading. Transactions will occur throughout the day at or relative to NAV and the dollar execution price will be determined when the NAV is published shortly after 4:00 p.m. After-hours trading relative to the current day’s and the next day’s NAV will also be possible.

The ability to create or redeem ETF shares each day should limit the size of any premium or discount on an ETF share as long as professional traders acting as market makers and arbitrageurs are reasonably attuned to the costs and opportunities of meeting demand for additional shares or redeeming existing shares of the ETF. Transactions in the NAV-based and traditional ETF markets are subject to similar fees and commissions. Information on these costs and trading choices will be available from your broker or adviser. Investors can compare differences in the spreads between contemporary bids and offers in the conventional intraday market and in the NAV-based market throughout the trading day. We have seen that locating the midpoint of the bid/offer range relative to an intraday ETF value is not easy for the average investor in the traditional ETF market. Even if an investor has ready access to intraday NAV proxies based on contemporary bids and offers for an ETF’s portfolio securities, the portfolio value of many ETFs can change by much more than the typical bid/offer spread in just a few seconds. In conventional ETF trading, most investors cannot be confident that their execution will be as close to the contemporary NAV as they intend.

The diagrams in Figure 6 compare limit orders in the NAV-based market (where an unchanged limit order may be appropriate for most or all of the trading day with little need for monitoring) with a sequence of limit orders in the intraday ETF market (where the appropriate level for a bid or offer can change more quickly than most investors can react). The right-hand side of Figure 6 illustrates the variability in the intraday value of the fund and highlights the limit order management problem that most investors face in the conventional ETF market as fund share values change. As noted, an investor’s order can trade only with bids and offers that are available in the market. Orders must be entered in the conventional ETF market to buy or sell at the market or at a specific dollar price, say $20.00 per share. Orders to buy or sell ETF shares relative to a contemporary or future ETF (portfolio) value are not accepted in the conventional ETF market. Orders to trade at or relative to the day’s official NAV are the essence of this new market that lets buyers and sellers express their bids and offers relative to net asset value. The transaction cost to buy or sell an ETF share in the NAV-based market is the sum of any fees and commissions plus (or minus) any difference between the execution price and the net asset value. With limit orders stated relative to NAV, an investor both knows and controls trading costs by this measure, whether the ETF is thinly or actively traded.

Figure 6

Professional traders and market makers understand that the economics of their business is based on volume and risk management. Most of them welcome the introduction of NAV-based trading because it will substantially increase trading volumes in ETFs that are not actively traded today. These market professionals will earn far more from small profits on a lot of trades than from much larger profits per share on very few shares traded. NAV-based trading also gives market makers an additional way to reduce their ETF inventory risk without incurring the costs of frequent creation or redemption transactions in lightly traded ETFs.

While the conventional ETF market usually works well for investors who know how to use it to trade the largest and most actively traded ETFs, it does not serve investors in less actively traded funds very well. Trading costs are high and hard to measure in the conventional market for less actively traded ETFs. When an ETF’s trading volume is low in the conventional ETF market, its bid/offer spreads are often wide, and professional traders have little choice but to respond opportunistically to retail orders rather than to arbitrage pricing relationships.

NAV-based trading increases the opportunity for ETFs based on newly developed indexes to compete with ETFs based on established benchmark indexes. It makes trading easier for advisers who are used to buying mutual fund shares at net asset value. In an extension of trading services, we anticipate the availability of executions based on dollar amounts and fractional shares provided by a financial intermediary to reduce ETF trading costs for defined contribution plans and other accounts that make frequent small transactions. NAV-based trading will facilitate the development of nontransparent ETFs that offer active portfolio management and nontransparent indexes that are not plagued with high composition change costs—in an ETF structure that protects shareholders from the cost of other investors’ fund share purchases and sales.

In an earlier paragraph, I noted that there is statistically significant evidence that the average stock trades at a lower price near the market opening than near the close on the same day. A fund net asset value can be calculated from opening prices as well as from closing prices. If there is demand for it, NAV-based trading around opening or hourly portfolio values as well as the traditional end-of-day NAV calculation is possible.

The author gratefully acknowledges helpful suggestions from the editor, Matt Hougan; participants at the NASDAQ OMX Symposium on Exchange Traded Funds at New York University’s Stern School of Business; and the Nasdaq OMX for the preparation of Figures 3 and 4.



1 Net-asset-value-based trading is covered by two U.S. patents and a number of pending patent applications.

2 Individual investors probably follow similar trade patterns in stocks. The previous comments reflect conventional wisdom, not a systematic study of the underlying cause.

3For discussion of these patterns and references to other research that has found similar patterns, see Heston, Steven L., Robert A. Korajczyk, and Ronnie Sadka, “Intraday Patterns in the Cross-Section of Stock Returns,” March 18, 2009. We will return briefly to the issue of intraday price patterns at the end of the article.

4One of the best articles describing this data is by Hougan, Matt, “ETFs, Spreads and Liquidity,” Journal of Indexes, July/August 2008, pp. 30–33. See also one of Hougan’s blogs comparing spreads in October 2007 and 2008, The Index Universe website,, lists monthly ETF average spreads. Click on Section and then on Data.

5 Do not rely on the published IOPV for any purpose.

6 They will use their proprietary value calculations to estimate NAV, not the posted IOPV.

7The most important paper on ETF trading premiums and discounts is Robert Engle and Depojyoti Sarkar, “Premiums-Discounts in Exchange-Traded Funds,” Journal of Derivatives, Summer 2006, pp. 27–45. This paper was reprinted in Bruce, Brian R., A Guide to Exchange Traded Funds and Indexing Innovations, Fall 2006, pp. 36–54. Engle and Sarkar found that ETFs holding domestic stocks had an end-of-day average premium of the closing price over the reported NAV of just +1.1 basis points (a discount would be reported as a premium with a negative sign). However, this tiny average premium is misleading. The average standard deviation of the last trade premium was 42.1 bps with a range of 17.6 bps to 142 bps for various funds. A 42-bp standard deviation is more than four-tenths of a percent of the value of the fund share. Using this standard deviation as a rough indicator of the cost of an MOC execution neglects the effect of “last” transactions that occurred before or after 4:00 p.m. Nonetheless, the average bid-asked spread for these domestic stock ETFs at the close was 37.7 bps. Neither this spread nor the reported premiums and discounts provide a useful indication of the price an investor should expect on an MOC transaction today. The study was based on a set of data that ended in September 2000. The procedures for trading ETFs around the market close have changed in a number of ways since then. The domestic stock ETFs that Engle and Sarkar studied now trade tens of millions of shares daily, but today’s less actively traded ETFs display the same variability of premiums and discounts found in the earlier data. A similar study of contemporary trading in less active ETFs would make interesting reading.


Find your next ETF