ETF Transaction Costs Are Often Higher Than Investors Realize

Article (PDF Available)inThe Journal of Portfolio Management 42(3):65-75 · April 2016with 638 Reads 
How we measure 'reads'
A 'read' is counted each time someone views a publication summary (such as the title, abstract, and list of authors), clicks on a figure, or views or downloads the full-text. Learn more
DOI: 10.3905/jpm.2016.42.3.065
Cite this publication
Exchange-traded funds (ETFs) and other exchange-traded products (ETPs) provide simple and efficient diversification and exposure to a wide range of investment asset classes. However, the apparently small bid-ask spreads for many ETFs can mask much higher transaction costs when ETF market prices deviate significantly from contemporaneous net asset values (NAVs). The deviations from NAV are often much greater than the bid-ask spreads suggest. The authors report that ETF closing prices are more than 90 basis points away from the NAV approximately 10% of the time. Investors' market-on-close (MOC) orders are likely to be executed at prices below the NAV.
The Voices of Influence |
JPM-COVER.indd 1 13/04/16 10:53 AM
ETF Transaction Costs
re Often Higher Tha
nvestors Realize
L. G
ETF Transaction Costs
Are Often Higher Than
Investors Realize
is an associate professor in
the McDonough School
of Business at Georgetown
University in Washington,
is the chief executive
off icer of Broms Asset
Management LLC in New
York, N Y.
is the president of ETF, Inc.
in Bonita Springs, FL.
Exchange-traded funds (ETFs) and
other exchange-traded products
(ETPs) provide simple and efficient
diversification and exposure to a
wide range of asset classes. The equity mar-
kets where ETF shares trade are usually much
more investor-friendly than the underlying
market structures for other asset classes. Low
bid–ask spreads combined with generally low
fund expenses have made ETFs an extremely
popular investment product—now accounting
for approximately 28% of the value of U.S.
exchange trading, according to Credit Suisse.1
However, ETF transaction costs are not
as low as they might seem. ETF prices can and
do deviate from their Net Asset Values (NAVs),
and an investor’s true transaction cost is the
amount of the deviation from the contempo-
raneous NAV. Narrow bid–ask spreads can be
a very misleading indicator of the true cost of
trading ETFs. This paper examines how and
why ETF prices can deviate from the NAV,
along with some of the implications for inves-
tors. Investors need to be aware of these poten-
tial deviations when they place their orders.
Market-on-close (MOC) orders, for example,
are often much more costly than they appear.
ETF trading prices are set by supply
and demand in the marketplace; there is no
regulator y requirement that they be tied in any
direct way to the underlying NAVs. Investors
rely upon an arbitrage mechanism to keep ETF
prices in line with the underlying portfolio
values. When the price of an ETF is below the
underlying portfolio value, arbitrageurs step in
to buy the cheap ETF. They then hedge their
risk in buying ETF shares, usually by selling
the basket of underlying securities or a corre-
lated proxy portfolio or index. This arbitrage
activity normally pushes the price of an ETF
into alignment with its underlying securities.
However, this arbitrage activity only takes
place when the ETF price has deviated enough
from the underlying portfolio value to make
arbitrage worthwhile. ETF prices will f luc-
tuate within a band determined by the cost of
arbitrage and the balance of supply and demand
for the ETF’s shares among investors.
To unwind an ETF arbitrage position,
an arbitrageur may reverse the original trade
if the price discrepancy has reversed, or close
out the position by creating or redeeming ETF
shares in transactions with the issuer at NAV.
As most investors in ETFs are aware, new
shares of an ETF are created when investor
demand for the shares is greater than the
number of shares outstanding. In a creation,
an Authorized Participant (AP) a brokerage
firm usually acting on behalf of a market
maker–typically exchanges a specified basket
of the fund’s portfolio securities for the ETF’s
shares. If demand for the fund shares slackens
and investors are selling shares on balance, an
JPM-ANGEL.indd 65 4/13/16 11:34:26 AM
AP will redeem ETF shares in kind by turning in shares
of the ETF and receiving the basket of portfolio securities
specif ied by the fund.2
Exhibit 1 illustrates in a schematic way the nature
and relationship of the various costs and prices associated
with common stock and ETF trading and ETF share
creations and redemptions. Point D denotes the current
value of the ETF’s underlying stock portfolio, and Point
E represents the value of the portfolio at the best ask
price. However, the ask price is usually only good for
a small number of shares. Because of market impact, a
market maker’s cost of acquiring the ETF’s underlying
holdings in creation unit size is higher (Point F). Fur-
thermore, the creation process is not free; transaction
fees push the market maker’s cost of actually creating
shares up to Point G. For the market maker to make
money by selling ETF shares in the market and then
transacting with the ETF to create shares, the average
price of the fund shares sold must exceed G.
Similar to creation fees, market makers pay trans-
action fees in connection with redemptions, which push
the market maker’s proceeds from redeeming shares
down to Point A. For a market maker to make money
by buying ETF shares in the market and then transacting
with the ETF to redeem the shares, the average price
of the fund shares purchased by the market maker must
be less than A.
An ETF’s price can f luctuate between A and G–and
beyond in either direction. The actual location of an ETF’s
trading price at any point in time will ref lect the current
balance of supply and demand for the ETF shares.
Unfortunately, too many investors assume that the
midpoint between an ETF’s bid and offer prices in intraday
trading is centered on or near the contemporaneous price
(last sale or midpoint value of the ETF’s portfolio) –point
D in Exhibit 1. In fact, transactions in ETF shares will
persistently unbalance the midpoint of the quoted ETF
bid–ask spread, and thus the midpoint need not straddle the
underlying value. For example, if a wave of selling senti-
ment hits the market for the ETF’s shares, both the bid and
ask prices could be pushed well below the fund’s current
portfolio value. This was particularly evident on August
24, 2015, when many ETFs traded at extreme discounts
to the value of their underlying assets.3
Exhibit 2 illustrates a case in which selling senti-
ment pushes an ETF’s bid and ask prices below the fund’s
current value almost to the level at which arbitrage will
Note that the midpoint of the quoted spread (A to
B) is not close to the current value of the ETF’s port-
folio. Obviously, the scale of A through G will var y with
the liquidity of the underlying instruments. An ETF
whose constituents are actively traded domestic stocks
will have a narrower range from A to G than an ETF
whose constituents are illiquid foreign stocks that don’t
trade during U.S. market hours.
The nominal spread between A and G is very large
for some ETFs, such as the iShares MSCI Emerging Mar-
kets ETF (EEM). By simplifying only slightly, we can see
that the maximum charge for market impact trading costs
over NAV (generally calculated using last sale prices) on a
cash creation of EEM shares is 3%, according to the EEM
prospectus.4 The comparable maximum on a redemption
is about 2%. The administrative fee is nominal relative
to these amounts, which suggests a possible 5% round-
trip cost, or $2 on a $40 share price. This maximum
round-trip trading cost is obviously inconsistent with
Parameters of ETF Trading Costs
Quoted Spread Need Not Straddle the Portfolio NAV
JPM-ANGEL.indd 66 4/13/16 11:34:26 AM
any expectation of a one or two basis point spread in the
intraday ETF exchange market.5
ETF Trading Costs Are Best Measured by
the Difference between the Price and the
Contemporaneous ETF Portfolio Value
Investors traditionally view equity trading costs
as a combination of bid–ask spread, market impact, and
commission costs. That may well be the most useful
measure of the cost of trading a stock, because the price
of a stock is the best determinant of its current value.
But ETFs are different from stocks. Their value is
derived from, and can be measured against, the value
of the fund’s underlying holdings. Accordingly, ETF
transaction costs are more accurately measured as the
difference between the net purchase or sale price and
the current value of the fund portfolio at the time the
transaction is priced.6 That is the real cost to the investor
of trading an ETF.
One might be tempted to dismiss f luctuations
between ETF transaction prices and cur-
rent portfolio values as random noises with
an expectation of zero. And one might
think that a buyer is just as likely to pay
a price below the current value as a price
above it. Such is not the case.
It is well known that investors tend to
move in herds, trading overwhelmingly in
the same direction. Traders acting as market
makers take the other side of investor trades.
The market makers, of course, charge
accordingly, because they have costs to
bear and expect to be paid for providing
liquidity. When investors are net buyers of
an ETF’s shares, they will push the price
of the ETF above the current portfolio
value. And when investors are net sellers
of an ETF’s shares, they will push the price
below current value. Thus, the average ETF
investor can expect to pay more than an
ETF is worth when buying—and receive
less than the ETF is worth when selling.
This more correct view of ETF trans-
action costs has a startling implication: ETF
bid–ask spreads tell investors nothing about the
costs associated with buying or selling ETF
The costs of buying or selling ETF shares are driven
by the costs of trading the underlying securities. ETFs can
have narrow spreads even when they are costly to trade.
Let us examine what we know about the relation-
ship between bid–ask spreads in the underlying securities
and the bid–ask spreads of ETFs. A 2013 publication
from BlackRock (Golub et al. [2013]), the sponsors of
the iShares family of ETFs, provides some comments
on ETF trading costs and the information on trading in
selected ETF holdings and the ETF markets illustrated
in Exhibits 3 and 4 below. The exhibits show some very
interesting price and volume relationships from Golub
et al. for some of the iShares domestic and international
ETFs, and their underlying indexes.
Exhibit 3 shows that the average daily U.S. dollar
market value of the stocks in the f ive indexes that trade
each day (the taller of each of the paired bars) is much
greater than the dollar value of the ETF index shares that
ETF vs. Underlying Average Daily Volume
Source: Blackrock, Trace, and Bloomberg via Golub (2013)
JPM-ANGEL.indd 67 4/13/16 11:34:26 AM
trade on the market each day (the shorter bars). Exhibit 4
compares the weighted bid–ask spreads for the baskets of
underlying securities in each of the funds listed (the taller
bars) to the bid–ask spreads in the market for the fund
shares (the shorter bars) measured in basis points (one basis
point = .01 of 1% = .0001). Regardless of how the bid–ask
spread for the basket of portfolio securities is weighted, the
spread for the average basket component is much wider
than the bid–ask spread reported for the shares of the
fund in every case, and the average bid–ask spreads on the
shares of the iShares funds that Golub et al. selected for
this examination are all less than three basis points.7
The iShares MSCI Emerging Markets Index Fund
(EEM), which has a large fraction of portfolio secu-
rities that trade on stock markets largely outside U.S.
stock exchange trading hours,8 has a bid–ask spread
on the fund shares of less than the equivalent of $0.02
on recent share prices. With such a narrow average
bid–ask spread, it is tempting to assume that the cost
of trading EEM shares is so small that it is not worth
worrying about. Perhaps some investors would believe
it is better to place a market order or a marketable limit
order than to allocate valuable time to
learning more about how ETF trading
costs work—unless this bid–ask spread
is too good to be true.
One of the most important things
our parents taught us is that if something
seems too good to be true, we should
take a very close look to see what is
going on. We have written this paper
because we believe that if one wants
to understand and control the costs of
buying and selling shares in these and
other ETFs, there are things going on
here that deserve attention.
Apparently tight nominal bid–ask
spreads for trading ETF shares create an
environment in which transaction costs
are higher than they appear. A fund’s offi-
cial end-of-day NAV is the best available
contemporaneous measure of the value of
the ETF portfolio at the market close. And
we will see that these ETFs often trade
much farther from the closing NAV than
indicated by the average bid–ask spread.
Exhibit 5 compares average
nominal (small transaction) ETF share
bid–ask spreads to average underlying stock basket nom-
inal bid–ask spreads for the funds highlighted in the
BlackRock (Golub et al. [2013]) article. Larger purchases
will often be at significantly higher prices and larger sales
will often be at significantly lower prices; thus the Best
Bid and Offer may not be a useful indicator of where
any trade will be priced during the trading day. The first
ETF vs. Underlying Bid –Ask Spread
Comparison of ETF Bid–Ask Spreads to Underlying
Basket Bid–Ask Spreads (in basis points)
Source: Golub et al. [2013, p. 9]
Source: Blackrock, Trace, and Bloomberg via Golub (2013)
JPM-ANGEL.indd 68 4/13/16 11:34:27 AM
data column in Exhibit 5 lists the nominal ETF spreads
for the iShares funds in Exhibit 4, and the second data
column shows the index basket spreads for those funds.
The bid–ask spread data for other large, actively
traded ETFs suggests that large, actively traded ETFs,
particularly those based on equity portfolios, trade in a
narrow nominal bid–ask range when, and only when, trading
in the fund shares is extremely active (with volume measured
in tens of millions or occasionally hundreds of millions
of shares per day). Less actively traded ETFs often have
much wider spreads. Unfortunately, even for the actively
traded ETFs, time-coordinated measures of the depth and
location of these fund bids and offers relative to current
markets in the portfolio securities or proxies are not
available to most investors. Consequently, the narrow
posted fund spreads often mislead investors into mistak-
enly assuming that the contemporaneous value of the
ETF share is “inside” or— at worst—at the other side
of the posted spread, and they often rely on fund share
spreads to estimate the cost of trading a specific ETF.
The fact that posted ETF average spreads can mislead
investors who want to control their cost of trading is
worth every ETF investor’s careful attention.
If an ETF holds only stocks traded in the U.S. and
Canada and a professional trader wants to keep track of
the current market for that ETF’s portfolio stocks, the
trader can obtain data to calculate a share value based
on the sizes and prices of bids and offers on all the stocks
in the ETF portfolio. This is not a useful calculation if a
significant fraction of the positions in the portfolio don’t
trade during U.S. trading hours. The professional traders
(whom we will refer to from here on as market makers
because they are trading in the ETF market in an attempt
to make a prof it by taking the other side of investor
trades) will manage the market risk of their firm’s trading
book with positions in various futures, options, stock
baskets, and ETF shares. A market maker will position
the market-making f irm’s bids and offers for an ETF’s
shares to interact with the overall balance of public order flow in
a way that will be profitable for the market-making firm.
ETF market makers are not indifferent to the pattern
of ETF investor trading. When public investors are sellers,
on balance, the market maker will usually try to bid for
and buy shares in the ETF at or below the bid-side value
of the securities in the portfolio creation/redemption basket.
On the offer side of the quote, the market maker may sell
ETF shares at a spread to its bid that is similar to the tight
spreads we see in the first data column 1 of Exhibit 5;
but the spread will only be this narrow if there are not
many buyers around. The posted spreads are tight, but as
long as the preponderance of customer trades are sales to
the market maker at a low bid, the market maker will be
buying shares below the cost of redeeming them. A small
buy order will be accommodated at the nominal offer
price, but if the direction of order f low changes and buyers
want to buy in size, the market maker will soon move the
offer up to ref lect the cost of creating more ETF shares.
Order f low in any ETF market is often predomi-
nantly in one direction or the other for hours or days at a
time. When a large new order or other new information
changes the direction of order f low, the market maker
will typically adjust the posted quote quickly to provide
liquidity to public buyers at a price where the market
maker can fully cover the cost of accommodating these
customers—plus a prof it. The appropriate trading cost
measurement for an investor is the difference between
the transaction price and the contemporaneous value of
the fund portfolio when the trade is priced. The true
round-trip trading spread will be closer to (and probably
greater than) the 24.8 basis points in the taller bar in the
MSCI Emerging Market ETF in Exhibit 4 than to the
2.4 basis points in the shorter bar. To clarify these rela-
tionships, we will examine this issue from other perspec-
tives (in both the text of this article and in the Appendix)
and see why the costs of buying and selling EEM shares
are often far higher than they initially appear.
Transaction Costs Are Often Much Higher
Than the Bid–Ask Spread for ETFs
One common misconception is that all ETFs are
inexpensive instruments to hold and trade. Although
broad market U.S.-based ETFs such as the SPDR S&P
500 ETF Trust (SPY) and Vanguard Total Stock Market
ETF (VTI) have rock-bottom expense ratios and usually
relatively low trading costs, this is not true for all ETFs.
Much more important than fund expense ratios in many
instances is the largely “hidden” cost of trading ETFs—and
that cost is the difference between the market price paid or
received and the contemporaneous value of the ETF assets.
Arbitrage activity minimizes this discrepancy for ETFs
holding liquid securities that trade during U.S. market
hours. However, many ETFs hold securities that do not
trade during U.S. market hours, and this makes traditional
arbitrage diff icult. Investors and even market makers are
often f lying blind when they trade these funds.
JPM-ANGEL.indd 69 4/13/16 11:34:29 AM
To better understand the relationship between
ETF valuation and ETF trading costs, let’s look more
closely at the limited measurable ETF trading cost data
available to us. Most investors are surprised to see how
difficult it is to find useful trading cost data for ETFs
and to see how large some ETF trading costs are when we
can measure them in a useful way. The only method we can
find to measure one-way investor trading costs for large
numbers of shares is to calculate the difference between
the closing price of an ETF’s shares and the fund’s daily
NAV, which represents the fair value of its net assets at
the U.S. market close—and the value at which creations
and redemptions of fund shares are made.
Many investors inadvertently pay high transaction
costs when they buy into an ETF at prices above its cur-
rent value or when they sell below current value. For
most ETFs, this difference is much larger than the bid–ask
spread. Exhibit 6 illustrates the absolute percentage devia-
tions in price between the closing price and NAV of the
iShares MSCI Emerging Markets Fund (EEM) during
2014.9 The Triangles ( ) indicate the absolute value of the
percentage deviation in price between the closing price
and the NAV for each day. The lower (darker) line indi-
cates the percentage amount of the ETF’s bid–ask spread
(the ETF column from Exhibit 4 or 5), and the upper
(lighter) line indicates the average bid–ask spread for the
fund’s underlying portfolio holdings (from Exhibit 4
or 5). This fund is actively traded and generally shows a
one or two cent bid–ask spread, or about 3 basis points.
This tiny spread looks like quite a bargain, considering
that the bid–ask spread on the constituents in the under-
lying index is 24.8 basis points, as reported by Golub et
al. ([2013, p. 9]). However, the closing price of EEM
often deviates significantly from the NAV.
As one can see from Exhibit 6, the closing price
often differs from the NAV by an amount far larger than
the EEM’s average bid–ask spread or even the bid–ask
spread on EEM’s underlying portfolio holdings. During
2014, the closing price deviated from the NAV by an
average of 48 basis points, nearly one half of one per-
cent. And that is just an average; on some days, it was
much higher than that. Intraday deviations may be much
higher. These deviations imply that many investors are
unwittingly paying much higher transaction costs than
they realize. The Appendix provides an analysis of the
deviations between market closing prices and NAVs for
nearly all of EEM’s history.
For U.S. equity-based large-cap ETFs, the arbitrage
mechanism works much better. Exhibit 7 displays the
absolute values of the differences between the closing
price and the NAV for the day for the venerable SPY.
One can see from this chart that the closing price is
iShares Emerging Markets ETF % Difference between Closing Price and NAV, 2014
JPM-ANGEL.indd 70 4/13/16 11:34:29 AM
usually quite close to the NAV. The cost of trading SPY
is indeed low! The triangles and the two spread lines are
not distinguishable at the same scale used in Exhibit 6.
Deviations such as those seen in EEM are not an
anomaly. Once one gets out of the easily arbitraged
and liquid U.S. equity-based ETFs linked to popular
domestic stock indexes with exchange-traded futures
and options, substantial deviations from NAV are more
the norm than the exception. Exhibit 8 displays the
absolute percentage difference between the NAV and the
closing price for approximately 1.5 million daily obser-
vations.10 While the median is a mere 15 basis points, on
5% of the obser vations (nearly 75,000 observations), the
deviation is 136 basis points (1.36%) or more.
While the 4: 00 p.m. calculations of ETF NAVs do
not always meet the standards of mutual fund NAVs, ETF
investors are often attracted to buying and selling ETFs
at the market close. Placing an MOC order seems like an
easy way of minimizing transactions costs by ensuring
that one gets the price from the closing auction.
The most appropriate measure of trading cost for
MOC transactions is the difference between the closing
price of the ETF and its NAV calculated at 4:00 p.m.
The closing price is published daily but the histor y
of closing prices is not usually available on the fund
website. Closing premiums and discounts available on
fund websites are nearly always based on the difference
between the midpoint quote and the NAV at the market
close. But in making the spread a midpoint versus NAV
comparison, we usually understate the premium or dis-
count to NAV reflected in the closing price. Let’s look
at some exemplary data.
A Case Study: iShares MSCI Emerging
Markets ETF
On September 12, 2014, the consolidated trading
volume in iShares MSCI Emerging Markets ETF (symbol
EEM) was 83,094,024 shares. At the closing cross on its
S&P 500 SPDR ETF % Difference between Closing Price and NAV, 2014
Distribution of Difference between ETP Closing
Price and NAV for all U.S. ETPs
JPM-ANGEL.indd 71 4/13/16 11:34:29 AM
principal market (NYSE Arca), 2,754,208 shares traded
at $43.79. The NAV calculated at 4:00 p.m.11 was $44.06,
and the midpoint of the market maker’s bid–ask spread at
4:00 p.m. was $43.81. The iShares website reported that
the EEM discount from NAV for the midpoint of the
spread was 56 basis points. It is more meaningful to use
the difference between the NAV and the closing price
(which is 61 basis points in this instance). Considering
that nearly 3 million shares traded at the close on the
principal market, the closing price was the price for a
signif icant fraction of the day’s volume in EEM shares.12
The closing price discount of 61 basis points is far from
the 2.4 basis point trading spread indicated in Exhibits
4 and 5.
The closing price is an extremely important piece
of data: Approximately 2.75 million shares were traded
at the close on the principal market on this day, and other
markets trading an ETF typically price closing trades at
the principal market’s close. Unfortunately, some ETF
regulatory documents require the use of the midpoint
quote before the closing price auction in calculating
premiums and discounts at the end of the trading day.
The midpoint quote before the close does not reflect
an actual trade, but it is usually the only time series of
premium and discount data found on fund websites.
There is no use for this data item that would not be
better performed by closing price data.
Using the discount from NAV reflected in the
EEM’s September 12, 2014 closing price, we are ready
to calculate the major element of trading cost for the
EEM shares traded at the close. Nearly 3 million shares
were sold at a discount of 61 basis points to NAV (cal-
culated at the 4:00 pm close)—the best measure we
have of the contemporaneous value of the portfolio; the
same number of shares was purchased at this discount,
providing the purchasers a negative transaction cost.
Clearly, the premium or discount is not randomly
distributed. Market makers buy shares from investors
when ETFs (or any shares) trade at a discount and sell
shares at a premium when customers want to buy more
shares than other customers want to sell. The location
and size of the market makers’ bids and offers depend
on the market makers’ business model, but it should be
clear that a market-making firm could not sur vive if its
average trading spread to NAV approached 2 basis points
on this emerging markets ETF. Market makers profit
on most of the trades they participate in throughout the
day—and there is no reason to assume that the average
transaction price over the course of the day is particu-
larly close to the contemporaneous value of the fund
Market makers usually carry inventories and they
also have costs associated with the creation and redemp-
tion of ETF shares. They would not survive if they did
not have negative transaction costs nearly all the time.
The information we have at the close is not necessarily
similar to the trading costs investors incur throughout
the trading session, but the size of the average spread
between the closing price and the NAV is very clear.
In the market for this ETF on September 12, 2014,
one or more market makers provided liquidity at the
close and bought the shares that investors wanted to sell
at a discount of 61 basis points from NAV. There is abso-
lutely nothing wrong with this. Market makers provide
valuable services and deserve to be compensated in order
to make a living. However, their living becomes part of
an investor’s trading costs when market makers provide
liquidity, and it is important that investors understand
the cost of market makers’ participation.
The MOC to NAV comparison is the only system-
atic estimate of ETF trading costs we can make in the con-
ventional just-like-a-stock ETF market. While intraday
values are disseminated for each ETF every 15 seconds,
they are generally based only on the ETF portfolio securi-
ties’ last sale prices, which can be quite stale. Most sentient
market professionals create their own estimates of current
portfolio values using adjusted real-time data.
Using MOC Orders Routinely to Buy or Sell
ETFs Looks Like a Terrible Idea
The majority of investor ETF market-on-close
orders are not likely to be executed at or better than
NAV, and many MOC executions will not even be close
to NAV. MOC orders can be executed at a much greater
distance from the NAV than the reported premium and
discount data indicate to investors. Anyone trading ETFs
should understand how these orders work and how ETF
MOC orders differ from stock MOC orders in important
ways. If you compare a fund’s NAVs and closing prices
over a few days, you will probably conclude that you do
not want to use an MOC order—unless your compari-
sons suggest that, for some reason, other investors are
predominantly selling when you buy or buying when
you sell. At the least, you will conclude that the cost to
trade most index ETFs is more than a few basis points.
JPM-ANGEL.indd 72 4/13/16 11:34:30 AM
One more question that we need to address is,
“How good are ETFs’ NAV calculations? ” The answer
to this question is that the official closing NAV calcula-
tions for ETFs are reasonably good most of the time, but
they are not up to mutual fund standards.13 They will be
getting a lot more attention in the future, and most of the
new ETFs started from now on will meet mutual fund
standards. When NextShares (actively managed ETPs
that will trade at or relative to the NAV determined at
4:00 pm) begin trading, the NAV calculations for these
funds will fully meet the highest mutual fund standards.
It will be interesting to monitor trading costs when all
parties to the trade can enter their orders relative to the
NAV—and all will know and control their transaction
costs with a high degree of precision.
To the extent that an ETF traded in the U.S. holds
securities that are not actively traded at 4:00 p.m. in
New York, it is appropriate to take a very close look at
the NAV calculation. It does not require rocket science:
Mutual funds with all kinds of domestic and foreign
securities portfolios have taken investors’ cash invest-
ments and redeemed fund shares for cash at 4:00 p.m.
Eastern Time for several generations. Although most
ETFs do not use the sophisticated fair value NAV
pricing methodology that conscientious U.S. mutual
funds embrace, the introduction of NextShares—with
portfolios managed and NAVs calculated like actively
managed mutual funds—will focus more attention on
ETF trading costs and the reliability of the NAV calcu-
lation. The NextShares funds should prompt investors
and fund managers to improve NAV calculations for
all exchange-traded products—as both the breadth of
ETP offerings and the focus on NAV will, appropriately,
continue to grow.
All investors considering buying ETFs should ask
themselves, “Can I trade ETFs efficiently?” With respect
to most ETFs available today, our answer to most inves-
tors has to be an emphatic “no.” We can only answer
“yes” for some of the major domestic equity benchmark
index ETFs that sometimes trade hundreds of millions
of shares each day. The just-like-a-stock trading method
adopted for the first ETF, SPY, works well for the most
actively traded, large-cap U.S. domestic equity index
ETFs—and less well, in varying degrees, for other ETFs.
A good trading solution for many ETFs could be NAV-
based trading, soon to be introduced for NextShares. A
buy or sell order in either conventional or NAV-based
ETF trading may lead to other orders or transactions that
have an effect on the fund’s NAV determination, just as
bids or offers in any market affect related markets.
NAV-based trading is the only way to achieve
transparency in trading costs—a far more important goal
for protecting investors than daily portfolio composi-
tion transparency, which often increases fund or investor
trading costs without delivering anything useful to inves-
tors.14 Just as most investors will be able to measure their
trading costs only with NAV-based trading, most market
makers will find it both easier and more economical to
manage their risks with NAV-based trading.
NAV-based trading was conceived and developed
to facilitate secondary market (exchange) trading of less-
than-fully-transparent, actively managed ETPs. One of
its features is that it enables an investor to know the cost
of a trade at the time the NAV-based order is entered.
However, given the extremely high apparent costs of
MOC trading in EEM, it appears that investors’ average
trading costs could be sharply reduced by introducing
NAV-based trading for many index ETFs in which the
underlying portfolio instruments do not trade during
U.S. trading hours.15
Case Study: Discounts and Premiums
on EEM
Exhibit A1 shows for the iShares Emerging Markets
ETF (EEM) the average (absolute) daily differences between
the closing prices on the NYSE Arca market and that day’s
NAV from the beginning of EEM’s second week of trading
in 2003 through March 31, 2015.
The NAV data behind Exhibit A1 is from Morningstar.
The Yahoo EEM closing price data series is raw (unadjusted)
data, but the supplementary text with the data and two very
distinct share price changes make it easy to adjust for the pair
of three-for-one stock splits. Al l these percentages represent
the average daily premium or discount at the close, measured by
calculating the average (absolute) daily percentage difference
between NAV and the closing price for each year or shorter
period since a week after the fund was launched. The f igures
are expressed as percentages of NAV, so .57% can be read as
57 basis points, if a basis point comparison is preferred.
JPM-ANGEL.indd 73 4/13/16 11:34:30 AM
Not surprisingly, many of the largest differences occurred
during the f inancial crisis in 2008. More to the point, the average
spread between NAV and closing price has been more than half a per-
cent in most years since the fund was launched. The fact that the
apparent transaction cost for the public side of this trade is this
high suggests that the average investor’s trading cost for this fund
is likely to be over 50 basis points. Daily spreads in excess of 100
basis points are common. Given the practical impossibilit y of
determining their trading costs at any time other than the close,
investors must resign themselves to high MOC trading costs in
this fund if they use conventional ETF trading.16
Interestingly, we made some other calculations that
devotees of MOC trades should find interesting. The average
change in NAV from day to day over the life of EEM was
just under 1% per day. It was this high only because of some
very large jumps in and around 2008. For the period from the
beginning of 2014 through March 31, 2015, the average daily
change in NAV was 0.61% (61 basis points), a little larger than
the average difference between NAV and the closing price
during this period. It is worth considering how to reduce
transaction costs with a more patient trading policy. Spe-
cif ically, the patient trader should be able to benefit by not
trading on days when his order at the close would put the
closing price on the “wrong” side of the expected NAV. If
the trader is selling and the balance of orders at the close that
day seems likely to put the close below NAV, it might pay to
wait for a better order pattern tomorrow.
2For many ETFs, part or all of a creation or redemption
transaction may be made for cash with a correspondingly
higher creation/redemption fee to cover the fund’s transac-
tion costs in buying or selling portfolio securities. The last or
closing sale prices of portfolio securities in each of their pri-
mar y trading markets are the dominant element in most ETFs’
NAV calculations; unlike most mutual funds, ETFs typica lly
do not adjust their NAVs to ref lect changes in value of foreign
holdings between local market closes and the time of the NAV
calculation. Fair value price adjustments are a mutual fund
industry best practice made to treat all shareholders fairly.
3See Driebusch et al. (2015).
4EEM Prospectus, p. 38.
5Although a 5% round-trip spread is onerous, unusua lly
strong or weak demand for an ETP’s shares might be ref lected
in occasional spreads even wider than this analysis indicates.
6As is common with most discussions of transaction
costs, we stress the costs directly associated with different types
of trades and neglect commission charges for simplicity.
7This is consistent with market microstructure theory.
Because an individual stock has a large amount of idiosyncratic
risk, along with the risk of informed trading, market makers
quote a wider spread. An index has less idiosyncratic risk and
a lower risk of informed trading, so market makers can usu-
ally safely quote narrower spreads on index products, but not
necessarily as much narrower as these charts suggest.
8Some of the larger markets represented in EEM are
China (including Taiwan), Korea, South Africa, Brazil, and
9The NAVs for the ETFs are from the Center for
Research in Security Prices (CRSP) mutual fund database
and the closing prices are from the CRSP daily stock f ile.
10This is based on data for all ETPs from the CRSP
mutual fund and daily stock price f iles for which matching
data were available.
11The 4:00 p.m. portfolio price is based on adjustments
to the most recent trade on home markets, which are not
open in most cases for the emerging markets represented in
this fund. Fair value NAV calculations and other features
designed to improve the usefulness and reliability of NAV
calculations are discussed in the text. The method of NAV
calculation descr ibed in the EEM prospectus differs from the
best mutual fund fair value calculations. This is an important
topic for further inquir y.
12The percentage of daily volume traded at the close varies
greatly among funds and over time. We have not attempted
a study of the pattern for EEM shares, but casual checks from
time to time indicate that the closing trades for EEM shares
typically account for a significant fraction of the total trading
volume. MOC trades away from the listing market are often
priced at the listing market close because that is usually the
only market with an organized closing auction for the fund.
Average Daily Premium or Discount in iShares
Emerging Markets ETF (EEM) at the Close, Mid-
April 2003 through March 31, 2015
JPM-ANGEL.indd 74 4/13/16 11:34:30 AM
13Many ETFs rely upon the last sale price to value
portfolio securities for determining fund NAVs, even if the
home market for that securit y closes before 4: 00 pm Eastern
time. Mutual funds generally use fair value pricing, adjusting
the valuation of foreign holdings from local market closing
prices to ref lect post-close trading in related securities and
14See Gastineau [2002, 2004, and 2008] for discussions
of some of the performance problems linked to index and
ETF transparency.
15Two of the authors, Broms and Gastineau, have a
financial interest in licensing of NAV-based trading patents.
16A sim ilar analysis to Exhibit 8 reveals that EEM’s
closest ETF competitor, the Vanguard Emerging Markets
Stock Index Fund (V WO), displays a difference between its
closing price and NAV that has been consistently less than
1/3 the percentages in Exhibit A1. Note that VWO is the
larger fund by nearly 50 % in terms of assets under manage-
ment and has lower operating, creation, and redemption costs
and smaller creation/redemption units. Yet EEM is by far the
more actively traded of the two, in spite of higher trading costs
when measured against NAV. Trading at the close volume is
relatively light in VWO. VWO is a share class of a mutual
fund that uses high quality fair value accounting in its NAV
calculations. While VWO is clearly the “eating sardine” in
this pair, it is puzzling that EEM is still the “trading sardine”
in spite of its much higher trading costs. (For a discussion of
trading sardines, see Gastineau [2010], pp. 245- 6.)
Driebusch, C., S. Vaishampayan, and L. Josephs. “Wild
Trading Exposes Flaws in ETFs.” Wall Street Journal, Sep-
tember 13, 2015.
exposed-f laws-in-etfs-1442174925. Analytical Material.
Gastineau, G.L. “Equity Index Funds Have Lost Their Way.”
The Journal of Portfolio Management, Vol. 28, No. 2 (2002),
pp. 55-64.
——. “The Benchmark Index ETF Performance Problem.”
The Journal of Portfolio Management, Vol. 30, No. 2 (2004),
pp. 96-103.
——. “The Cost of Trading Transparency: What We Know,
What We Don’t Know and How We Will Know.The Journal
of Portfolio Management, Vol. 30, No. 1 (2008), pp. 72-81.
——. The Exchange-Traded Funds Manual, Second Edition.
Hoboken, NJ: John Wiley & Sons, 2010, especially Chap-
ters 5 and 7.
Golub, B., B. Novick, A. Madhavan, I. Shapiro, K. Wal-
ters, and M. Ferconi. “Exchange Traded Products: Over-
view, Benefits and Myths,” 2013. https://www.blackrock.
iShares Emerging Markets ETF Prospectus, 2015. http://
emerging-markets-etf-8-31.pdf. December 31, 2015.
To order reprints of this article, please contact Dewey Palmieri
at dpalmieri@ or 212-224-3675.
JPM-ANGEL.indd 75 4/13/16 11:34:31 AM
  • ... Investors are able to obtain broad exposure to these markets at a rela-tively low cost in terms of management fees and expense ratios. However, Angel et al., (2016) [1] demonstrate that the true cost of ETF investing must incorporate deviations from net asset value (NAV). They show that, although bid-ask spreads for ETFs may be narrow even for emerging markets ETFs, the true cost of trading should incorporate a consideration of deviations from NAV, since it can be an order of magnitude higher than expense ratios due to the much larger bid-ask spreads of the securities held by the ETF. ...
    ... Investors are able to obtain broad exposure to these markets at a rela-tively low cost in terms of management fees and expense ratios. However, Angel et al., (2016) [1] demonstrate that the true cost of ETF investing must incorporate deviations from net asset value (NAV). They show that, although bid-ask spreads for ETFs may be narrow even for emerging markets ETFs, the true cost of trading should incorporate a consideration of deviations from NAV, since it can be an order of magnitude higher than expense ratios due to the much larger bid-ask spreads of the securities held by the ETF. ...
  • ... 5 Much of the research on the costs to investors of exploitation has been conducted in the context of adversarial trading (e.g., frontrunning). See, for example, Angel et al. [2016], Gastineau [2008], Saglam [2018], van Kervel and Menkveld [2019], Yang and Zhu [2019], and references therein. 6 For brevity, throughout the rest of the chapter, when referring to sustainability-related alt-data, we will use the term "sustainability data". ...
  • ... However, this is also a source of hidden costs of ETFs which is basically determined by bid-ask spread and price deviations from Net Asset Values (NAV). Gastineau (Angel et al. 2016) coducted research to examine how ETF prices can deviate from the NAV. The authors made findings that ETFs full transparency of trading costs could be achieved only by NAV-based trading. ...
    Full-text available
    Growing acceptance of passive bond exchange traded funds (ETFs) and actively managed bond mutual funds has exposed the need to find a divide between these two comparatively similar types of instrument. This paper provides a comparative analysis of actively managed bond funds and passive bond ETFs in the context of multiple criteria. The research of risk-adjusted performance of a sampled group of bond funds and ETFs using the TOPSIS multi-criteria decision-making method revealed that actively managed bond funds have a modest advantage over passive bond ETFs. Moreover, the final findings indicate the funds’ performance dependability on portfolio composition by fixed income sector.
  • ... Ultimately, a big question is whether ETF returns are better than mutual fund returns (Chen et al. 2015;De Jong and Rhee 2008;Gastineau 2004;Toolson 2012). The ETFs tend to have lower expense ratios, but higher trading costs, so balancing these two against each other is a worthy consideration (Angel et al. 2016;Dellva 2001). Beyond this choice, ETFs can provide an easy route for buy and hold strategies and are easily accessible on any stock trading platform. ...
    Full-text available
    The buy and hold stock market strategy, which gained tremendous popularity in the 1970s, may no longer be such a profitable method for accumulating wealth for the average investor in the new millennium. This paper investigates the relationship between compound return and holding period length to see how long an Exchange Traded Fund (ETF) investment must be held before a positive return on principal is 100% likely. Because the ETF is a relatively new investment vehicle that could be considered particularly well-suited to the requirements of the buy and hold strategy, we begin our investigation here. We find that the compound returns earned over a rolling holding period are much more volatile than one might assume given historic rules of thumb for average return expectations. Using monthly return data for all listed NASDAQ ETFs between their date of inception and 2015, we find it takes ten years for the average probability of a gain on principal to be over 95 percent.
  • Chapter
    Full-text available
    Around the world, Index Arbitrage and ETF Arbitrage remains a major problem (this chapter does not cover the arbitrage of Mutual Funds). The US Congress and many governments have attempted to tackle these problems and have enacted various laws which have not been effective. Index Arbitrage and ETF Arbitrage don’t facilitate price discovery, they amplify volatility and financial instability and distort indices and the perceptions of the true values and risks of companies. This chapter contributes to the existing literature by: (i) introducing new adverse effects of Index Arbitrage and ETF Arbitrage, (ii) introducing new methods for eliminating Index Arbitrage and ETF Arbitrage, (iii) introducing new theories-of-liability and causes-of-action against perpetrators of Index Arbitrage and ETF Arbitrage.
  • Chapter
    Some of the problems inherent in the structure of Financial Indices, ETFs and Index Funds were discussed in earlier chapters in this book—and the large size of the Global Index Products Market amplifies these problems. Another dimension is that the Social Welfare problems of Indexing can have wide-ranging negative “Multiplier Effects” (on households, companies and government agencies) and which have not been addressed by index sponsors, fund sponsors or regulators. This chapter: (i) discusses the implications of ETFs, Indices and Index Funds for enforcement, Sustainability, Inequality and financial stability; (ii) discusses “path-dependence” and “Lock-ins” and proposes new models of government intervention; and (iii) proposes new sustainability measures that are designed to reduce the wide-ranging adverse effects of Indices, Index Funds and ETFs (such as Destructive Urbanization, Inequality, Pollution and Climate Change, harmful Arbitrage, and Costly Technological Change).
  • Wild Trading Exposes Flaws in ETFs
    • C Driebusch
    • S Vaishampayan
    • L Josephs
    Driebusch, C., S. Vaishampayan, and L. Josephs. "Wild Trading Exposes Flaws in ETFs." Wall Street Journal, September 13, 2015. laws-in-etfs-1442174925.
  • What We Don't Know and How We Will Know
    . "The Cost of Trading Transparency: What We Know, What We Don't Know and How We Will Know." The Journal of Portfolio Management, Vol. 30, No. 1 (2008), pp. 72-81.-. The Exchange-Traded Funds Manual, Second Edition. Hoboken, NJ: John Wiley & Sons, 2010, especially Chapters 5 and 7.
  • Exchange Traded Products: Overview, Benefits and Myths
    • B Golub
    • B Novick
    • A Madhavan
    • I Shapiro
    • K Walters
    • M Ferconi
    Golub, B., B. Novick, A. Madhavan, I. Shapiro, K. Walters, and M. Ferconi. "Exchange Traded Products: Overview, Benefits and Myths," 2013. https://www.blackrock. com/corporate/en-pl/literature/whitepaper/viewpoint-etpsoverview-benefits-myths-062013.pdf.
  • The Exchange-Traded Funds Manual, Second Edition
    --. "The Cost of Trading Transparency: What We Know, What We Don't Know and How We Will Know." The Journal of Portfolio Management, Vol. 30, No. 1 (2008), pp. 72-81. --. The Exchange-Traded Funds Manual, Second Edition. Hoboken, NJ: John Wiley & Sons, 2010, especially Chapters 5 and 7.
  • Article
    Full-text available
    Exchange-traded funds have generally underperformed competitive conventional mutual funds, apparently because the ETF structure and fund management process seems to discourage efforts to recapture transaction costs embedded in the benchmark index portfolio change process. An analysis of the problem produces a rather simple solution.