EFA

Eastern Financial
Finance
Association Review
The Financial Review 38 (2001) 123-152

Post-Reform Market-Order Execution

Quality: Multidimensional Comparisons
Across Market Centers
Brian Hatch*
University of Cincinnati
Robert Battalio
University of Notre Dame
Robert Jennings
Indiana University

Abstract
We use market-order data to determine execution quality on the NYSE, four regional
stock exchanges, and the Nasdaq InterMarket. We examine a sample period after the reduction
in the minimum price variation and after the SEC imposed new order-handling rules, and
analyze dimensions of execution quality in addition to trade prices. We find that in the post-
reform environment, the NYSE offers execution prices that are more favorable to the investor.
However, the regional exchanges and the InterMarket offer executions that are faster and
that more frequently allow investors to execute orders with sizes exceeding the quoted depth
at the quoted price.
Keywords: price improvement, liquidity premium, depth improvement, execution speed, re-
gional stock exchange
JEL Classification: G18

1. Introduction
The impact of market fragmentation on execution quality is a critical market
structure issue. To partially address this issue, on July 28, 2000, the Securities and

*Corresponding author: Department of Finance, University of Cincinnati, PO Box 210195, Cincinnati,

OH 45221-0195; Phone: (5 13) 556-7076; Fax: (5 13) 556-4891; Email: brian.hatch@uc.edu
We thank two anonymous referees for suggestions that have clearly improved the paper. Earlier versions
of this paper were entitled The Quality of Trade Execution on Regional Exchanges and were produced
for presentation at the Chicago Stock Exchange on May 20, 1997, and for the Conference on the Search
for the Best Price sponsored by the New York Stock Exchange on March 15, 1996. All views expressed
in the paper are those of the authors.

123
124 B. Hatch, R. Battalio, R. JenningdThe Financial Review 38 (2001) 123-152

Exchange Commission (SEC) proposed new rules that required brokers to disclose
order routing practices and market centers (defined as any exchange market maker,
over-the-counter market maker, alternative trading system, national securities ex-
change, or national securities association) to report their execution quality.’
Since execution quality can vary greatly across market centers (Macey and
O’Hara, 1997), and because of the increase in the costs of execution relative to
brokerage commissions, the order routing decision is more critical than ever. To
make an informed routing decision, investors and brokers must know the execution
quality of each market center.
Although many academic studies have examined execution quality across ex-
changes, the SEC indicates in the 34-43084 rule proposal that many of these analyses
are “of somewhat limited utility for evaluating order executions because of the
limited nature of their data sources.” These analyses usually rely on trade and quote
data and do not give such information as whether the order is to buy or sell or if
the order is a limit or market order. The SEC believes that complete reporting of
execution quality will further the objectives of the national market system, fostering
competition among market centers and leading to more efficient securities transac-
tions.
In this paper we address the lack of useful execution quality reports by using
current order data to show the relative execution quality across six market centers.
In 1997, the SEC implemented new order handling rules and the minimum tick size
was reduced to $0.0625. We study execution quality issues in this new environment,
extend the literature to jointly examine the NYSE, four regional exchanges, and the
Nasdaq InterMarket (on June 13,2000, Nasdaq officially renamed their third market
the Nasdaq InterMarket), and include other dimensions of execution quality beyond
trade price. We use a proprietary data set of market orders in a common set of
NYSE-listed securities directed to the NYSE, the Boston Stock Exchange (BSE),
the Chicago Stock Exchange (CHX), the Pacific Exchange (PCX), the Philadelphia
Stock Exchange (PHLX), and the Nasdaq InterMarket during the last two weeks
of June 2000.
We measure execution quality on the price dimension by computing the rate
at which execution prices better a benchmark quote (price improvement rates) and
how close execution prices come to the midpoint of a benchmark spread (the liquidity
premium). Using our proprietary data, we can judge execution quality in markets
where the width of the bid-ask spread is at minimum variation ($0.0625 for our
sample). We can accurately identify the trade initiator, identify the order submission
time, and identify the order type (market or limit).
In addition to price, many investors, particularly day traders, prefer quick
executions. Because we use order data rather than trade data, we can measure how
quickly market orders are filled. Even the NYSE has publicly recognized that “some

’ SEC Release No. 34-43084,“Disclosure of Order Routing and Execution Practices,” which is available
at http:llwww.sec.gov/mleslproposed/34-430.
 B. Hatch, R. Battalio, R. JenningdThe Financial Review 38 (2001) 123-152 125

individual investors seek certainty and speed of execution - even at the expense of
possible price improvement.’’’
Finally, by using order data, we can measure if an order was completely filled
at one price even if the order size exceeded the available depth at that price. We
call this depth improvement.A large order can be filled at the posted price schedule,
and still appear to be disimproved. If the order size exceeds the quoted depth, it
might fill at multiple prices and thus a portion of the order executes at a price
outside the order receipt time quotes. To the extent that an exchange is willing to
completely fill such large orders at the order receipt time quoted prices, that exchange
offers a superior execution to venues unwilling to deviate from the price schedule.
Like other researchers, we find that overall, the NYSE offers the best execution
prices. Some market centers offer competitive prices for certain order types: the
BSE for small orders in $0.06250 spread markets, the PCX for orders of at least
2,000 shares, and the Nasdaq InterMarket for small orders in $0.1250 spread markets.
Other market centers are seldom competitive: the CHX consistently offers inferior
executions for orders of at least 500 shares.
In other areas of execution quality, the non-NYSE market centers are more
competitive. All market centers except the CHX fill orders for greater than the
quoted depth at quoted prices or better (depth improvement) more frequently than
does the NYSE. When the quoted depth can accommodate an order size, all market
centers except the BSE tend to fill the market orders faster than the NYSE. This is
to be expected, since non-NYSE venues have automatic execution systems and the
NYSE uses an auction mechanism.
The paper proceeds in Section 2 with a further discussion of fragmentation
and execution quality and a review of the literature. Section 3 discusses our measures
of execution quality. Section 4 describes the data used in the study. We discuss the
results in section 5. Section 6 concludes.

2. Background
Regional stock exchanges have a long h i ~ t o r yThe
. ~ Philadelphia Stock Ex-
change, for example, predates the NYSE. Many of these regional exchanges devel-
oped in response to the need for capital formation in a particular geographic market.
Today there are five regional stock exchanges: the Boston Stock Exchange, the
Chicago Stock Exchange (formerly the Midwest), the Cincinnati Stock Exchange
(CSE), the Pacific Exchange (formerly the Pacific Stock Exchange), and the Philadel-
phia Stock Exchange. However, the nature of the regional exchanges’ business has
changed dramatically. Today, most of their trading volume is in stocks listed on

‘‘Reportto NYSE Board on Market-StructureRecommendations,” The Exchange,April 2000. Bacidore,

Ross, and Sofianos (1999) present an analysis of execution speed on the NYSE.
Amold, Hersch, Mulherin, and Netter (1999) present a more complete discussion of the history of
regional stock exchanges.
126 B. Hatch, R. Battalio, R. Jennings/The Financial Review 38 (2001) 123-152

the national exchanges such as the NYSE and the Amex rather than the regional
exchanges. The regionals trade these securities by using unlisted trading privileges,
which allow trading on the regional after one day of seasoning on the stock’s primary
exchange. Because many securities trade at multiple locations, we can compare the
execution quality for the same securities across market centers.
Academics, regulators, traders, and government officials have vigorously de-
bated the ramifications of fragmentation for market qua lit^.^ Prior to the rule proposal
34-43084, the SEC issued a concept release on February 23, 2000 concerning
fragmentation and sought comment on six proposals to deal with this p r ~ b l e m . ~
The Senate Banking Committee convened a special meeting with industry leaders on
February 29,2000 to discuss market fragmentation. The proponents of concentration
argued that funneling trades to a single location can provide a deep, liquid market
in which investors can execute trades at more favorable prices than they could if
order flow is divided across market centers. Those who argued against concentration
voiced concerns about granting one market center too much market power. They
fear that the suppression of the competition that has led to innovation in the equity
markets could push some market participants to foreign markets.
Knowing the quality of executions across market centers has practical impor-
tance for brokers and traders. The SEC indicates that the broker has a fiduciary
responsibility to seek the best price reasonably available under the circumstances.6
Arthur Levitt, the SEC chairman, says that “. . . brokers have a duty to evaluate
all available options for routing customer orders, periodically assessing the quality
of competing markets to assure that their customers are receiving the best available
prices.”’ Furthermore, the SEC has requested that exchanges review their member
organizations’ compliance with these fiduciary responsibilities.’ The courts have
also weighed in on the broker’s responsibility for achieving the best execution. In
the case of Newton v. Merrill Lynch, Merrill Lynch was charged with breaching
their fiduciary responsibility of best execution by filling customer orders at the
“national best bid or offer” (NBBO) when they knew that there may have been
opportunities to fill those orders inside the NBBO. The U.S. Court of Appeals
claims that Merrill Lynch misrepresented by stating that they would execute customer
orders to maximize the customers’ economic benefit when they knew that they

See the Securities and Exchange Commission’s Market 2000 Report: Study 111, Market Fragmentation,
Competition, and Regulation and mock (1999).
See Release No. 34-42450, Commission Request for Comment on Issues Relating to Market Fragmenta-
tion, which is available at http://www.sec.gov/rules/sros/ny9948n.htm.
See the SEC’s Market 2000 Report: Study 111, Market Fragmentation, Competition, and Regulation.
See Investment Dealer’s Digest, “Smith Barney and A.G. Edwards Study Merrill’s NYSE Program,”
November 6, 1995.
See Securities Week, “SEC Warns Big Board to be ‘Fair’ in Best Execution Probe of Member Firms,”
March 31, 1997.
 B. Hatch, R. Battalio, R. JenningdThe Financial Review 38 (2001) 123-152 127

would execute these orders at the NBBO even if better prices were reasonably
available. Macey and O’Hara (1997) provide further discussion of this case and a
detailed analysis of the duty of best execution.
Price improvement on the NYSE and Amex has been thoroughly examined.’
However, we know less about the practice of the regional stock exchanges and the
Nasdaq InterMarket. Although non-NYSE venues are an important part of the equity
markets (the NYSE Fact Book indicates that in 1999 16% of all trades in NYSE-
listed stocks were executed on the regional exchanges and another 12% in the
Nasdaq InterMarket), there is less research focusing on their execution quality.
Some financial media accounts have implied that execution prices on the regional
exchanges are poor relative to those available on the NYSE. For example, in
announcing Merrill Lynch’s decision to move trades that had been automatically
executed on regionals (BSE and PCX) to the NYSE, a Merrill representative was
quoted as saying, “Our singular goal is to give our clients’ orders in these exchange-
listed equities in the U.S. the best opportunity to receive price improvement.” ’’
The NYSE does not discourage the idea that its executions are superior in
quality to regional exchange executions. The NYSE states that order flow routing
arrangements on the regionals “prevent customer from meeting customer and pre-
clude the opportunity for price improvement.”” The regional exchanges and some
broker-dealer firms disagree with these claims and emphasize other dimensions of
execution quality. The Pershing Division of Donaldson, Lufkin & Jenrette Securities
Corporation serves over 600 broker-dealers. Pershing claims to execute a majority
of their NYSE-listed order flow on regional exchanges because these exchanges
“offer faster executions at a lower cost than the primary exchange.”
Our work is most closely related to that of Lightfoot, Martin, Peterson, and
Sirri (1998), which originated as a part of the SEC’s “Report on the Practice of
Preferencing” (1997). In their study, Lightfoot, Martin, Peterson, and Sirri use order
data from 1996 to show an advantage for the NYSE over the regional exchanges
in quoted spreads, effective spreads, and price improvement rates. They find that
among the regional exchanges, the CSE has the smallest quoted and effective spreads
and offers the greatest amount of price improvement. They also examine the quality
of executions provided for limit orders routed to various exchanges. As we indicated
earlier, since their report the structure of the market has changed dramatically.
Our work is also closely tied to that of Battalio, Hatch, and Jennings (2000),
who examine orders in S&P 100 stocks that are directed to either the NYSE or
Nasdaq InterMarket broker-dealer Trimark Securities, Inc. They calculate net price

Angel, 1994; Petersen and Fiakowski, 1994; Harris and Hasbrouck, 1996; Knez and Ready, 1996;
ROSS,Shapiro, and Smith, 1996; Handa, Schwartz, and Tiwari, 1999; and Bacidore, Battalio, and Jennings,
2000 present studies of execution price quality on the NYSE and Amex.
lo“Memll Shifts on Small Investors’ Orders,” Wall Street JoloumaE, November 10, 1995.
I’ “The Auction Market versus Referencing,” The Exchange,June 1996. Battalio, Greene, and Jennings
(1997 and 1998) discuss preferencing and internalization.
128 B. Hatch, R. Battalio, R. JenningdThe Financial Review 38 (2001) 123-152

improvement rates, liquidity premiums, execution speeds, and depth improvement

for both venues. Battalio, Hatch, and Jennings find that the NYSE generally provides
better trade prices, but Trimark provides more depth improvement and faster execu-
tions. Using an overall measure of market maker trading revenue that includes order
flow payments, they find no significant difference between the NYSE and Trimark.
The SEC rule proposal indicates that the availability of order data is crucial
to any useful analysis of execution quality. To accurately measure price improvement
rates, we need to know if the order is a buy or sell, when the order is submitted,
and if the order is a market or limit order. With the exception of the Trades, Orders,
Reports, and Quotes (TORQ) database, commercially available data such as the
Trade and Quote data (TAQ) and data from the Institute for the Study of Security
Markets (ISSM), do not identify the trade initiator or order type. The TORQ data
identify the initiator, but cover only NYSE SuperDot transactions between November
1990 and January 1991 for 144 stocks. Thus, current research that uses commercially
available databases cannot calculate effective spreads, price improvement, or liquid-
ity premiums in minimum-variation spread markets. These studies must infer the
initiator in wider spread markets, cannot determine the spread at order receipt time,
cannot distinguish market and limit order executions, cannot measure execution
speed, and cannot accurately identify depth improvement.’2
Extant work on regional trade execution is subject to these limitations. Lee
(1993) uses the 1988 and 1989 ISSM data to compare execution prices across market
centers after controlling for the security, trade size, and execution time. Lee finds
that the NYSE appears to offer the most favorable prices for “simultaneously”
executed trades. Bessembinder and Kaufman (1996), using TAQ data from 1994,
conclude that the NYSE and the CSE have the smallest effective spreads.
Several papers address the specific impact of the reduction in tick size and the
implementation of the SEC’s new order-handling rules on trading costs. Barclay,
Christie, Harris, Kandel, and Schultz (1999) show a 30% reduction in spreads after
the implementation of the order handling rules. Bessembinder (1999) and Stoll
(2000) show that although trade execution costs on Nasdaq have fallen since these
changes in the market, significant differences still remain relative to the NYSE.
Although they do not directly measure regional exchange execution quality,
Van Ness, Van Ness, and Pruitt (1999) do look at the cost to trade a set of Nasdaq
stocks that were also traded on the Chicago Stock Exchange. The CHX was unable
to attract significant order flow, even with better quotes, but the dual listing did
lower quoted spreads. Van Ness, Van Ness, and Hsieh (1999) detail trading costs
and price improvement for 97 stocks in the CHX-Nasdaq dual trading experiment.
They find that the CHX offers lower quoted and effective spreads and more price
improvement.

I2Ellis, Michaely, and O’Hara (2000) and Finucane (2000) present tests on the accuracy of trade
classification rules.
 B. Hatch, R. Battalio, R. JenningdThe Financial Review 38 (2001) 123-152 129

In research that more directly addresses the fragmentation issue, McInish and
Wood (1996) construct five stock portfolios that are equal in various characteristics
that affect spreads, volatility, and execution quality but maximize the dispersion in
fragmentation.McInish and Wood show that competitionappears to benefit investors
as spreads tend to move inversely with market fragmentation.

3. Measures of execution quality

As measures of execution quality we use net price improvement, liquidity
premium, depth improvement,and execution speed. We say a trade is price improved
if it executes at a price better than the NBBO at order submission. We calculate
the net price improvement rate (NPIR) as the number of shares that receive price
improvement, less those that are disimproved (execute outside the NBBO), divided
by the total number of shares executed. Some orders fill with multiple executions
at different prices. The execution price that we compare to the NBBO is the volume-
weighted average price (VWAP). If this price betters the NBBO at order submission,
then we define the order as price improved.
In computing liquidity premiums, we have the advantage of knowing the
initiator of the trade. Therefore, we can compute the liquidity premium for a given
trade as:

LP = (Transaction price - midpoint of the benchmark bid/ask spread) x I (1)

In equation (l),I is an indicator variable takmg a value of one if the trade is buyer-
initiated and -1 if the trade is seller-initiated. The benchmark spread is the NBBO.
Once again, we use the VWAP as the transaction price used to calculate the
liquidity premium. We assume that the midpoint of the bidask spread is an estimate
of the true value of the security, thus the liquidity premium measures the price of
immediacy. Positive values for the liquidity premium indicate the trader is paying
for liquidity. A negative liquidity premium indicates that the trader is transacting
at prices better than the true price. Thus, the liquidity premium captures both the
existence and degree of price improvement and price disimprovement.
For example, in $0.1250 spread markets, the price improvementrate notes only
if trades are executed at prices better or worse than the quotes. However, the liquidity
premium notes whether the trade occurred at the midpoint of the spread (an LP of
zero) or at the opposite side (buying at the bid or selling at the ask) of the spread
(an LP = -$0.0625) and so captures the magnitude of the price improvement.
Often, an order arrives whose size exceeds the quoted depth available in the
market at the best quoted prices. If the liquidity provider fills the entire order at
the best quoted price or better, we say the dealer provides depth improvement.
However, the dealer is not obligated to fill the order at this price. The dealer may
fill the order to the size of the quote at the best price and fill the remainder at a
price inferior to the best quoted price. Clearly, a trader who receives depth improve-
130 B. Hatch, R. Battalio, R. JenningdThe Financial Review 38 (2001) 123-152

ment receives a higher quality execution than one who does not. To the degree that
one market center offers greater depth improvement than another, we say that they
offer a higher quality execution on this dimension However, we note that we are
not the first to suggest this as a dimension of execution quality. Battalio, Hatch,
and Jennings (2000) and Bacidore, Battalio, and Jennings (2000) suggest similar
measures of depth improvement.
We define an order as having received depth improvement if its size is greater
than the available depth at the NBBO, and if the VWAP for the order is at least as
good as the quoted price. The number of shares eligible for depth improvement is
the excess of order size over quoted depth on the appropriate side of the market.
To determine the depth improvement rate for each market center, we divide the
total number of shares improved by the shares eligible for improvement at that
market center.
Finally, the speed with which an order executes is very important to some
investors. To the degree that one market center can execute orders faster than others,
this market center provides superior execution quality to these traders, at least on
speed. The SEC notes the importance of execution speed to investors and has
suggested in its proposed rules on execution quality reporting that market centers
report the distribution of execution times.
We calculate the percentage of orders in different speed categories. We also
present the cumulative distribution function of execution speeds for each exchange,
conditional on whether the order size exceeded the available depth at the NBBO.

4. Data
As we noted earlier, commercially available databases do not contain the detail
needed to accurately measure execution quality. To obtain the necessary data for a
common set of securities over a common time period, we discussed the project with
colleagues at Market Systems with which we had a previous relationship, and asked
if they would be willing to participate. Although data distribution is not their line
of business and they do not routinely provide data to researchers, they agreed to
provide a data set for the last two weeks of June 2000.
Market Systems provides execution quality reports to their client brokerage
firms, which include online brokers, clearing firms, and full-service broker-dealers.
Our data set comprises the orders placed by the clients of these brokerage firms.
Thus, the order flow is representative of general retail order flow.
Market Systems screens all orders for less than 100 shares and more than
10,000 shares and any orders with special conditions (special conditions apply to
orders that require manual handling). Almost all of the order flow provided by
Market Systems meets the automatic execution criteria of, and is electronically
submitted to, the market centers. Our sample of order flow is comparable with that
used by Ross, Shapiro, and Smith (1996) who examine price improvement for
SuperDOT market orders on the NYSE and Lightfoot, Martin, Peterson, and Sirri
 B. Hatch, R. Battalio, R. JenningdThe Financial Review 38 (2001)123-152 131

(1998) who focus on orders submitted electronically to each of the exchanges. We

obtain the order identification number, date and time of order submission to the
market center, security symbol, origin of order (buyer or seller), order size, order
routing (market center), quote (NBBO) at time the order was submitted, date and
time of execution(s), and execution price and quantity. Market Systems also provided
time series of the NBBO for each stock in the sample.
The order flow in our sample of NYSE-listed securities is routed to the NYSE,
the BSE, the CHX, the PHLX, the PCX, and broker-dealer firms in the Nasdaq
InterMarket. Because Market Systems has no orders directed to it, the only regional
exchange not represented is the Cincinnati Stock Exchange.
To make valid comparisons across venues, we need to be certain that each
market center has a similar sample of stocks and order flow. Therefore, to be
included in our sample, we require that the stock be listed on the NYSE, trade on
at least three exchanges, and have at least 200 orders executed on the regional
exchanges in our sample period.
Table 1 summarizes the data provided by Market Systems. In Panel A, we
eliminate orders in preferred stocks and special share classes, orders received or
executed outside of regular trading hours, orders with missing or erroneous data,
and orders for which we could not establish a valid order submission time NBBO.
The screen that eliminates the largest part of the sample is one that requires sufficient
trading activity across several market centers.
Our final sample comprises 292,631 market orders in 164 stocks that trade
actively across multiple market centers. Forty-five of these stocks are listed in the
S&P 100.
Panel B of Table 1 shows the number of orders in our final sample by market
center, and shows that during our sample period, each market center has at least
49 stocks with at least 50 orders on that market center. We include all 164 stocks
in our execution quality measures.

5. Results
During the two-week sample period, 145 million shares traded on the various
market centers. The greatest number of shares (over 63 million) trades on the NYSE.
Throughout this paper, our statistical tests of relative execution quality consist
of two-tailed t-tests on across-stock averages. We choose each individual stock as
a unit of observation because broker routing decisions are typically made on a stock-
by-stock basis. We always compare venue i to the NYSE to determine the extent
of cross-sectionalvariation in execution quality. If we condition the execution quality
measure only on the exchange, then we require, to be included in the statistical
tests, that each stock have 75 orders on both venue i and on the NYSE. If we
condition on exchange and a variable such as order submission time spread, then
we require that the stock have 75 orders on venue i and on the NYSE and that there
be at least ten orders within a spread sub-category on both venue i and the NYSE.
132 B. Hatch, R. Battalio, R. JenningdThe Financial Review 38 (2001) 123-152

Table 1
Construction of market-order sample across the regional exchanges, the NYSE, and the Nasdaq
InterMarket
We construct a sample of 562,636 market order executions in 3,975 stocks during the last two weeks of
June 2000. These executions occur on the Boston (BSE), Chicago (CHX), New York (NYSE), Philadelphia
(PHLX), Pacific (PCX), and the Nasdaq InterMarket (NASD). The sample excludes odd lot orders and
orders for more than 10,000shares. The table shows the process by which we screen the sample to 292,63 1
market orders in 164 stocks that are traded actively across at least three of our market centers.
~ ~~~ ~ ~~

Panel A: Data set Construction

Category # of Orders

Total Orders 562,636

Preferred stock or special share class 5,530
Pre-open orders 44,176
Post-close orders 978
No order identification number 66
Special buylsell indicator (we eliminated any order not accompanied by an indicator 13.969
other than an unconditional buy or sell, - e.g., short sales)
Apparent data error (either an error in the dates, execution times, or execution prices) 1,955
No match between order ticker and NBBO tickers (34 ticker symbols that were in the 134
order data set were not present in the NBBO data set)
Remaining Orders 495,828
Orders in stocks not actively traded across multiple venues (the stock must trade 200,699
across at least three of the exchanges and have at least 200 orders on the regional
exchanges during our sample period)
Incomplete fills 487
No valid quote match at order submission time (no firm quote without special 1,970
conditions or data errors was available at order submission time)
NBBO crossed or locked at order submission time 41
Remaining orders 292.631
Number of stocks in the final sample 164
Number of S&P 100 stocks in final samule 45
~ ~~

Panel B: By Market Center Description

BSE CHX NYSE PHLX PCX NASD
No. of stocks with at least 50 orders 86 108 159 49 91 138
No. of S&P 100 stocks 38 40 45 18 34 45
Total Orders 41,169 51,805 72,971 17,422 37,445 71,819

When we increased our paired-sample size for the statistical test by reducing the
required number of orders in a stock from 75 to 50, we found virtually identical
results to those reported in the paper. We also calculated sign tests, which largely
confirm the results of our t-tests. Thus, we create a paired sample of actively traded
stocks whose orders have similar characteristics. We use the NYSE as the benchmark
for two reasons. It has the largest sample size and so is likely to provide the largest
paired-sample sizes when comparing across venues, and because current studies
show that the NYSE offers the best execution prices.
 B. Hatch, R. Battalio, R. JenningdThe Financial Review 38 (2001) 123-152 133

Table 2
Execution price quality by market center
The table shows the number of shares and orders included in the sample conditional on trading market
center for the 164 stocks that traded actively across at least three stock exchanges during the last two
weeks of June 2000. For these orders we show the net price improvement rate (NPIR), which we calculate
as the number of shares price improved less the number of shares disimproved divided by the total
number of shares executed; and the liquidity premium (LP), which measures the cost of executing the
order relative to the order submission time spread midpoint. The execution quality at a particular market
center is very sensitive to the order submission time spread, order size, and market conditions. Therefore,
conclusions about the execution quality offered on any specific market center must be properly
conditioned. We cannot draw conclusions based on the results in this table. We conduct an across-stock,
two-tailed t-test to determine whether the indicated mean differs statistically from the comparable
NYSE mean.

Avg.
Venue Orders Shares Order Size NPIR LP
BSE 41,169 15,882,805 386 16.04% $0.05134
CHX 51,805 18,684,802 361 8.83%* $0.061lo**
NYSE 72,971 63,084,129 865 17.46% $0.04721
PHLX 17,422 5,963,120 342 19.32% $0.05 188
PCX 31,445 14,222,411 380 18.80%** $0.04462
NASD 71,819 27,118,653 378 16.29% $0.07257

** Indicates statistical significance at the 0.01 level.

* Indicates statistical significance at the 0.05 level.

Table 2 provides the results of using all of the eligible orders. We find that,
unconditionally, all exchanges offer very similar net price improvement rates of
16% to 19%. The exception is the CHX, at 8.83%. The CHX and the InterMarket
stand out as having higher liquidity premiums, although only the CHX liquidity
premium is statistically different from the NYSE. However, in their rule proposal
on execution quality reporting, the SEC noted that overly general statistics mask
differences in order flow. Therefore, we must adequately condition our analysis to
be sure that we are comparing similar sets of orders across market centers.
For example, Table 2 shows that the average NYSE order size is more than
double the average order size on the other market centers. If it costs more to fill
larger orders, then the average NYSE liquidity premium may be larger than those
of the other venues even if the NYSE has a lower liquidty premium in every order
size category.

5.1. NPIR and LP conditional on order submission time spread

In Table 3 we condition on the NBBO at order submission. Of the orders in
our sample, 89% are entered when the spread is $0.1875 or less. Therefore, when
we condition on spread size, we report results only for submission time spreads up
to $0.1875.
134 B. Hatch, R. Battalio, R. JenningdThe Financial Review 38 (2001) 123-152

We first test whether the net price improvement rate in $0.0625 spread markets
is different from zero at each market center using an across-stock, two-tailed t-test.
Only the BSE and the PCX do not show a negative net price improvement rate.
This finding implies that all the other market centers trade through the NBBO more
often than they provide price improvement in $0.0625 spread markets. We also find
that the Pacific Exchange has a significantly higher net price improvement rate than
the NYSE in $0.0625 spread markets. We expect low price improvement rates since
the minimum tick of $0.0625 means that improvement will result in the customer
selling at the ask or buying at the bid. However, when the spread widens to $0.1250,
providing a tick between the bid and ask, price improvement increases dramatically.
Four of the exchanges provide net price improvement of 30% to 32%, with
the exception of the CHX (NPIR=21.77%) and the InterMarket. The InterMarket
provides a net price improvement rate of 37.24%, which is significantly greater
price improvement than the NYSE. The InterMarket can do so because some of its
member firms use an algorithm that automatically supplies one tick price improve-
ment to certain types of orders if the spread is greater than minimum variation
(Battalio, Hatch, and Jennings, 2000 describe such an algorithm used by Trimark
Securities, Inc.) Regardless of the order submission time spread or the venue, the
liquidity premiums are statisticallyidentical. However, the CHX’s liquidity premium
in $0.1250 spread markets is about $0.01 higher than the others. As the spread
reaches $0.1875, the dispersion in liquidity premiums across venues increases.

5.2. NPIR and LP conditional on order size

Ross, Shapiro, and Smith (1996) and Lightfoot, Martin, Peterson, and Sirri
(1998) show that market quality statistics vary by trade size. Because there are very
few large trades on the regional exchanges and in the InterMarket, our ability to
examine their execution quality is limited. Nevertheless, we can examine the net
price improvement rate and liquidity premium across several trade size categories.
We use the four size categories suggested in the SEC’s proposed rule on execution
quality disclosure: 100-499 shares, 500-1,999 shares, 2,000-4,999 shares, and at
least 5,000 shares. Table 4 presents the results of our analysis conditional on
trade size.
Because there is an inverse relation between the price improvement rate and
order size shown by Lightfoot, Martin, Peterson, and Sirri (1998) and the concept
that adverse selection increases with order size, as demonstrated in the models of
Admati and Pfleiderer (1988) and Easley and O’Hara (1987), we expect to find that
the cost of liquidity increases with order size. As we expected, at each market
center, the net price improvement rate drops with an increase in order size and the
liquidity premium rises. In other words, execution price quality decreases with
increases in order size.
For the smallest orders (100-499 shares), the NYSE offers a net price improve-
ment rate significantly better than the BSE, CHX, and PCX, and offers a liquidity
B. Hatch, R. Battalio, R. JenningdThe Financial Review 38 (2001) 123-152 135

3
2
w
r-
vl
5
0
(A

3
2
2
0
N
vl

1
0
(A

P
E
vl
%
8
m
 .p

3
-
&
R
Table 4
a
Execution price quality by market center and order size
The table shows the number of orders in each order size category by market center and the share-weighted execution price quality measures of the net price
e
3
improvement rate (NF'IR), which we calculate as the number of shares price improved less the number of shares disimproved divided by the total number of -5
shares executed; and the liquidity premium (LP), which measures the cost of executing the order relative to the order submission time spread midpoint. Our a
sample comprises 164 stocks that traded actively across at least three stock exchanges during the final two weeks of June 2000. We conduct an across-stock, 3
two-tailed t-test to determine whether the indicated mean differs statistically from the comparable NYSE mean. s
~

Order Size Market Center

c?
(shares) Y
BSE CHX NYSE PHLX PCX NASD z
100-499
Orders 31,887 40,254 41,423 13,932 28,754 55,125
2
?i
NPIR 22.41%** 28.82%* 34.51% 34.69% 23.47%** 28.94%
LP $0.04479 $0.04534** $0.03853 $0.03670 $0.04271** $0.06585** 2
500-1999 E.
2
Orders 8,003 10,174 2 1,864 3,070 7,617 14,562 cu
8
NPIR 15.63%** 2.06%** 26.07% 16.43%** 17.22%** 14.87%** -
LP $0.05224** $0.06468** $0.04312 $0.05346* $0.04546** $0.07201**
(continued)
z
h
",
,
cu
I
L,

3
Table 4 (continued)
Execution price quality by market center and order size

Order Size Market Center

(shares) BSE CHX NYSE PHLX PCX NASD
2000-4999
Orders 1,080 1,132 1,365 376 945 1,648
NPIR 11.03% -10.88%** 14.78% -7.26%* 14.02% 4.12%*
LP $0.05431* $0.08030** $0.04715 $0.08490* $0.05152 $0.08823**
>=5000
Orders 199 245 2,319 44 129 484
NPIR -1.71% -1 7.40%' -0.07% -5.46% 14.09% -4.19%*
LP $0.07221 $0.08448 $0.05779 $0.06210 $0.02935 $0.07420*
** Indicates statistical significance at the 0.01 level.
* Indicates statistical si&icance at the 0.05 level.
138 B. Hatch, R. Battalio, R. JenningdThe Financial Review 38 (2001) 123-152

premium significantly lower than the CHX, PCX, and the InterMarket. Among the
regional exchanges, the PHLX appears to do the best job on small orders with the
highest net price improvement rate and the lowest liquidity premium.
For orders of 500-1,999 shares, the NYSE has a significantly higher net price
improvement rate and a significantly lower liquidity premium than all other market
centers.
Because our sample consists of retail order flow, we do not have a sufficient
sample of paired orders to make reliable across-exchange comparisons for the largest
two order size categories (2,000-4,999 shares and at least 5,000 shares). However,
we can say that the magnitudes of the NYSE net price improvement rate and liquidity
premium are better than the other market centers in the 2,000-4,999 share group.
Only the PCX shows evidence of execution quality comparable to the NYSE for
the largest orders.
The distribution of order sizes and execution quality in Table 4 shows why it
is so important to properly condition execution quality measures. Each market center
except the NYSE has 77% to 80% of their order flow in the smallest size category.
Only 57% of the NYSE orders fall in this group. As we noted, the small orders
tend to have the highest NPIRs and the lowest LPs. Over 13% of NYSE orders are
for at least 2,000 shares. The other market centers have, at most, only 3% of their
order flow in such large orders.

5.3. NPIR and LP conditional on order submission time spread and or-
der size
Knowing that execution quality fluctuates with both spread and order size, we
condition on both of these variables. However, the more finely we condition the
sample, the more difficult it is for us to create a big enough matched sample across
exchanges to detect across-exchange differences in execution quality. Table 5 shows
the results of conditioning on both order size and submission time spread. Holding
spread constant, execution quality deteriorates with order size. The exception is the
PCX, on which large orders seem to receive excellent prices. Holding order size
constant, the net price improvement rate and the liquidity premium tend to increase
with spread size, except for large orders on the PCX. Surprisingly, orders for at
least 5,000 shares entered in $0.1875 spread markets on the PCX have the lowest
liquidity premium for any order group. The only statistical differences that we see
across market centers are that the NYSE offers a higher net price improvement rate
and a lower liquidity premium than any other market center in $0.1250 spread
markets for orders of 500-1,999 shares.
For large order sizes, the PCX offers excellent execution prices, Among the
regional exchanges, the BSE does exceptionally well with small orders in $0.0625
spread markets. Both Table 5 and Table 4 show that price improvement on the
CHX is very sensitive to order size. Most of the price improvement the CHX
 ?J
Table 5
Execution price quality by market center, order submission time spread, and order size
The table shows the share-weighted net price improvement rate (NPIR), which we calculate as the number of shares price improved less the number of shares
disimproved divided by the total number of shares executed; and the liquidity premium (LP), which measures the cost of executing the order relative to the
order submission time spread midpoint. We calculate the NPIR and LP conditional on executing venue, order size, and order submission time NBBO spread
for our sample of 164 stocks that traded actively across at least three stock exchanges during the final two weeks of June 2000. We include the number of
orders to demonstrate the sample size in each category. We conduct an across-stock, two-tailed t-test to determine whether the indicated mean differs statistically ?J
from the comparable NYSE mean. 3
~
s
3.
s
Venue $0.0625 spread $0.1250 spread $0,1875 spread 0s

Order Size Orders NPIR LP Orders NPIR LP Orders NPIR LP

B
BSE
2
100-499 17,323 6.05% $0.02766 9,778 38.05% $0.03721 2,190 53.54% $0.04995
500- 1999 4,403 3.95% $0.03035 2,423 28.14%* $0.04447* 533 36.81% $0.06582
2000-4999 616 -0.81% $0.03505 302 21.35% $0.05369 76 51.96% $0.05417
25000 124 -9.16% $0.04330 44 13.40% $0.04838 14 31.49% $0.04579
CHX
100-499 18,341 -1.06% $0.03405 13,623 52.54% $0.02994 3,945 60.84% $0.05 143
500-1999 4,928 -8.16% $0.04090 3,293 4.81% $0.06377 1,010 19.12% $0.08520
2000-4999 573 -19.74% $0.05370 358 -2.25% $0.07600 100 -10.69% $0.11830
25000 140 -25.55% $0.06145 59 -10.07% $0.08422 20 5.00% $0.09350
(continued)
 P
0

Table 5 (continued)
?J
Execution price quality by market center, order submission time spread, and order size
~~

Venue $0.0625 spread $0.1250 spread $0.1875 spread

Order Size Orders NPZR LP Orders NPIR LP Orders NPIR LP ?J

NYSE
100-499 18,103 9.71% $0.02586 13,669 49.59% $0.02571 4,374 62.02% $0.02727
500-1999 9,716 4.59% $0.03160 7,224 39.29% $0.03303 2,423 52.22% $0.04365
2000-4999 3,805 -0.34% $0.03346 2,376 29.50% $0.04811 639 36.15% $0.05057
25000 1,235 -10.10% $0.03960 737 13.18% $0.05533 183 9.11% $0.09129
PHLX
100-499 5,750 0.15% $0.031 11 5,201 50.90% $0.03067 1,698 73.79% $0.04416
500- 1999 1,395 -3.36% $0.03406 1,052 23.00%* $0.05061* 350 47.40% $0.06461
2000-4999 191 -16.04% $0.07645 109 -0.83% $0.07319 45 -1.41% $0.10784
25000 27 -19.09% $0.05568 9 0.00% $0.06250 6 16.67% $0.07375
PCX
100-499 13,520 1.76% $0.03126 9,920 37.05% $0.03918 2,867 54.97% $0.05330
500-1999 3,950 4.16% $0.02962 2,398 27.08%** $0.04627** 673 41.07% $0.05856 Rs.
2000-4999 487 5.03% $0.03163 311 23.54% $0.04928 16 27.14% $0.07179 m
25000 76 1.88% $0.03007 40 26.10% $0.03775 6 83.87% $0.02117 s
NASD
100-499 25,106 1.35% $0.03181 16,248 53.43% $0.02902 5,001 59.00% $0.05349
500-1999 7,327 -.33% $0.03445 4,004 34.92%** $0.04283** 1,237 34.16% $0.07 141
2000-4999 895 -5.59% $0.03669 437 18.46% $0.05730 125 15.74% $0.11661
25000 303 -14.15% $0.04684 116 9.76% $0.08506 27 2.13% $0.12305
** Indicates statistical significance at the 0.01 level.
* Indicates statistical significance at the 0.05 level.
 B. Hatch, R. Battalio, R. JenningdThe Financial Review 38 (2001) 123-152 141

offers is for orders of less than 500 shares. The InterMarket broker-dealers do not
distinguish themselves in any particular category.
It is difficult to compare our results with those of Lightfoot, Martin, Peterson,
and Sirri (1998). Because of the change in the minimum tick, there are no comparable
categories. Perhaps the only fair comparison is the relative ranking of market centers
that are common to each study. Compared to the Lightfoot, Martin, Peterson, and
Sirri sample, the regional exchanges are now more competitive with the NYSE on
execution price. The one common trait is that the CHX tends to offer the worst
execution prices in both samples.

5.4. Depth improvement rate

Table 6 shows the number of shares eligible for depth improvementby exchange
and by order size. The table also shows the percentage of those shares that are price
improved.
We focus first on all orders at each market center. We determine that all of
the venues except the CHX offer greater depth improvement rates than the NYSE.
The PCX offers the highest level by improving 67.39% of all eligible shares.
As we would expect, smaller orders are seldom greater than the quoted depth.
When they do exceed the quoted depth, it is by such a small amount that they are
usually filled at the quoted price and thus are improved. As order size increases,
the improvement rate drops. In each size category the CHX has the lowest depth
improvement rate.

5.5. NPIR and LP conditional on depth improvement eligibility

Orders whose size exceeds the quoted depth may execute at the quoted price
schedule and still be counted as disimproved. Therefore, we condition our measures
of execution price quality on whether the size of the order exceeds the quoted depth.
We do this in Table 7. For those orders whose size is accommodated by the quoted
depth, the NYSE offers the best net price improvement rate and the best liquidity
premium, both in each spread size category and overall.
Considering all orders, regardless of order submission spread size, the NYSE
net price improvement rates significantly better that of the BSE, CHX, or the PCX,
and its liquidity premium is significantly lower than the liquidity premiums of all
competing market centers. The NYSE provides these superior prices for orders
whose average size is twice that on any other market center.
Among the non-NYSE venues, the only conspicuous result is that the CHX
offers the worst net price improvement rate in all categories. The CHX gives the
highest liquidity premium in every category except in $0.1875 spread markets.
When the order is oversized relative to the quoted depth, the market center is
not obligated to accommodate the entire order at quoted prices. Therefore, when
we compare the VWAP for an order to the benchmark quoted prices, the execution
Table 6
Depth improvement rate by market center and order size
*
The table shows the number of shares that are eligible for depth improvement, conditional on order size and market center, for our sample of 164 stocks for
s
the final two weeks of June 2000. An order is eligible for depth improvement if its size is larger than the NBBO depth on the appropriate side of the market
at order submission. The number of shares eligible equals the maximum of order size less NNBO depth at order submission and zero. The improvement rate 3
equals the total number of eligible shares that received depth improvement (execution at quoted price or better) divided by the total number of shares eligible 3
for depth improvement. We conduct an across-stock, two-tailed t-test to determinewhether the indicated mean differs statisticallyfromthe comparableNYSE mean. 3.
3

Order Size Market Center P

(shares) BSE CHX NYSE PHLX PCX NASD ?t

3
100-499 s
Shares Eligible 24,871 32,213 60,020 15,605 29,542 57,356 3.
Percent Eligible 0.43% 0.46% 0.72% 0.63% 0.57% 0.61% f
Improvement Rate 85.56% 70.00% 78.65% 95.83% 89.68% 86.16%
500-1999 5.
Shares Eligible 325,077 420,171 1,394,349 146,281 333,957 720,922 $
Percent Eligible 5.19% 5.46% 7.20% 6.38% 5.64% 6.45% ,"
Imurovement Rate 73.87% 60.62% 71.88% 71.28% 74.11% 71.91% 2
b
(continued)
b
-
t
u
Table 6 (continued)
9
Depth improvement rate by market center and order size

Order Size Market Center

(shares) BSE CHX NYSE PHLX PCX NASD 9
2000-4999
Shares Eligible 500,957 545,19 1 3,673,912 196,347 414,589 753,019
Percent Eligible 18.44% 19.57% 17.53% 20.63% 17.50% 18.24%
Improvement Rate 61.73% 50.35% 62.94% 60.61% 62.24% 63.55%
>=5000
Shares Eligible 357,868 279,700 4,291821 62,312 202,891 622,440
Percent Eligible 31.13% 22.64% 29.75% 25.40% 27.15% 24.55%
Improvement Rate 53.01% 46.05% 5 1.40% 55.95% 63.64% 56.37%
All Orders B
Shares Eligible 1,208,773 1,277,275 9,420,212 420,605 980,979 2,153,797 f.
Percent Eligible 7.61% 6.84% 14.93% 7.05% 6.90% 7.92% 2
Improvement Rate 65.39%* 53.28%* 59.11% 64.94% 61.39%** 64.88%*
** Indicates statistical significance at the 0.01 level.
* Indicates statistical significance at the 0.05 level.
Table 7
Execution price quality by order size relative to NBBO depth
For those orders whose size is greater than the quoted depth, we can expect their execution quality to be inferior to orders whose size can be accommodated
by the quoted depth. This table provides the execution quality measures of the share-weighted net price improvement rate (NPIR), which we calculate as the
number of shares price improved less the number of shares disimproved divided by the total number of shares executed; the liquidity premium (LP), which
measures the cost of executing the order relative to the order submission time spread midpoint; and eligibility for depth improvement. Our sample comprises
164 stocks that traded actively across at least three stock exchanges during the final two weeks of June 2000. We conduct an across-stock, two-tailed r-test
to determine whether the indicated mean differs statistically from the comparable NYSE mean.

Order size I quoted depth Order size > quoted depth

Venue Avg. Order Avg. Order
spread Size Orders NPIR LP Size Orders NPIR LP
BSE
$0.0625 21,891 5.93% $0.02840 575 -22.19% $0.05422
$0.1250 12,241 33.33% $0.03988 306 3.86% $0.07103
$0.1875 2,692 50.49% $0.05332 121 22.77% $0.07024
All 345 39,800 19.32%** $0.04316** 1,570 1,369 -4.88%* $0.10358
CHX
$0.0625 23,362 -4.72% $0.03715 620 -29.07% $0.08674
$0.1250 16,900 27.79% $0.04699 433 -14.94% $0.10648
$0.1875 4,862 40.52% $0.06720 213 - 12.05% $0.12 128
All 326 50,127 13.76%** $0.05292** 1,389 1,678 -25.79%** $0.11856*
Table 7 (continued)
Execution price quality by order size relative to NBBO depth

Order size I quoted depth Order size > quoted depth

Venue Avg. Order Avg. Order a
spread Size Orders NPR LP Size Orders NPIR LP
NYSE
$0.0625 30,214 8.12% $0.02709 2,645 -28.76% $0.05576
$0.1250 22,166 44.90% $0.02912 1,840 -10.14% $0.08249 a
$0.1875 6,876 58.83% $0.03515 743 -0.37% $0.09106
Au 706 66.35 1 27.84% $0.03357 2,454 6,620 -12.46% $0.08652
PHLX
$0.0625 7,137 -0.55% $0.03852 226 -41.91% $0.06745
$0.1250 6,217 36.28% $0.04078 154 -12.38% $0.08653
$0.1875 2,023 58.13% $0.05559 76 9.62% $0.09845
All 311 16,831 23.72% $0.04761* 1.242 591 -12.03% $0.08236
PCX
$0.0625 17,512 5.73% $0.02845 521 -16.76% $0.04814
$0.1250 12,310 34.13% $0.03982 359 0.82% $0.07262 R5
$0.1875 3,486 50.08% $0.05387 136 21.73% $0.07742 2
All 343 36,093 2 1.47%** $0.04048** 1,357 1,352 0.78%** $0.07259* LJ
03
NASD
$0.0625 32,739 -0.001% $0.03258 892 -30.85% $0.06152
$0.1250 20,286 43.81% $0.03660 519 -14.60% $0.09723
$0.1875 6,112 49.04% $0.05962 278 -5.05% $0.13722
All 340 69,037 20.68% $0.05806** 1,342 2.782 -11.29% $0.16369
** Indicates statistical significance at the 0.01 level.
* Indicates statistical significance at the 0.05 level.
CL

R
146 B. Hatch, R. Battalio, R. JenningdThe Financial Review 38 12001) 123-I52

100%

‘0 80% +BSE
-
.-
Y
0)
+CHX
Q)
0)
60% +NYSE
20)
40% j+PHLX
+PCX
2
2 20% +NASD
0%
Q %

Execution Speed (seconds)

Figure 1

Cumulative distribution function of execution speeds for small orders

quality statistics are far worse, because the order is likely to “walk the book”
before being filled. Among all market centers, the BSE and the PCX do the best
job of filling these orders at favorable prices. These two exchanges tend to offer
the most price improvement and the lowest liquidity premiums across all spread
size categories.

5.6. Execution speed

As we mentioned earlier, to some investors the speed of execution is an
important aspect of execution quality. Figures 1 and 2 display the cumulative
distribution functions for execution speed, conditional on whether the order size
exceeds the quoted depth. Obviously, orders whose size is less than or equal to the
quoted depth fill more quickly than oversized orders. Because the NYSE is an
auction market and does not provide automatic executions, we would expect their
order execution times to be slower than those of the other market centers that do
provide automatic execution opportunities. We look first at orders that can be
accommodated by the existing quote, and find that the distribution of execution
speeds for the BSE and the NYSE are very similar and clearly inferior to the other
market centers. In terms of speed, the PCX is the intermediate market. The CHX,
PHLX, and the InterMarket offer the quickest executions, with the PHLX filling
over 80% of its orders in two seconds or less. For oversized orders, no one market
offers clearly faster executions. The InterMarket fills 45% of oversized orders within
two seconds, but they also take more than one minute to fill 12% of these orders.
In Table 8 we categorize execution speed, using the SEC-recommended size
and speed categories. The table also provides the median execution speed for each
 B. Hatch, R. Battalio, R. JenningdThe Financial Review 38 (2001) 123-152 147

100%

80% +BSE
-0)
+CHX
ii 60%
p,
w +NYSE
40% ++PHLX
3 +PCX
2 20% +NASD
0%

Execution Speed (seconds)

Figure 2

Cumulative distribution function of execution speeds for oversized orders

order size. For the smallest orders, the PHLX, InterMarket, and the CHX are fastest
and the NYSE is the slowest. As order size grows to 2,000-4,999 shares and then
to 5000+ shares, the InterMarket is the market that fills these orders fastest.

6. Conclusions
In this paper, we provide useful execution quality reports of the type identified
by the SEC. Earlier studies suffer from a lack of order data from which to calculate
reliable execution quality measures. Very few studies other than those by Lightfoot,
Martin, Peterson, and Sirri (1998) and Battalio, Hatch, and Jennings (2000) have
used order data.
We also provide one of the few analyses of regional exchange and Nasdasq
InterMarket execution quality. Hence, we extend the Lightfoot, Martin, Peterson,
and Sirri analysis to the Nasdaq InterMarket. Our analysis examines a post-reform
sample period in which market dynamics have changed substantially, and measures
dimensions of execution quality beyond trade prices. We also extend the analysis
of Battalio, Hatch, and Jennings (2000) to the regional exchanges.
We find that the NYSE offers the best execution prices, and that the Chicago
Stock Exchange frequently offers the worst trade prices. However, non-NYSE venues
offer competitive execution prices for certain classes of orders. In depth improve-
ment, the BSE, PHLX, PCX, and the InterMarket offer superior improvement rates
to the NYSE. For orders for shares less than or equal to the quoted depth, the CHX,
PHLX, and the InterMarket have the quickest executions, with the PHLX filling
over 80% of such orders in two seconds or less. Because of its auction market
system, the NYSE tends to fill these relatively small orders more slowly than most
Table 8
Execution speed by market center and order size 3
c;
s
The table shows the distribution of order execution times based on the speed categories suggested by the SEC as a measure of execution quality. The speed
of execution is from the time the order is submitted by the broker to the time the order is reported as executed at the exchange. The execution speeds are for
our sample of 164 stocks that traded actively across at least three stock exchanges during the final two weeks of June 2000.

Order Size Market Center

Speed Bin BSE CHX NYSE PHLX PCX NASD

100-499
0-9 seconds 56.50% 82.01% 9.14% 92.38% 57.56% 84.05%
10-29 seconds 32.39% 9.37% 62.63% 2.14% 24.69% 7.47%
30-59 seconds 7.23% 3.27% 19.89% 4.06% 8.69% 3.21%
60-299 seconds 3.71% 4.94% 6.84% 1.08% 8.30% 5.11%
300-1800 seconds 0.13% 0.32% 0.45% 0.23% 0.58% 0.12%
> 1800 seconds 0.03% 0.10% 1.06% 0.10% 0.17% 0.05%
Median Speed 7 seconds 2 seconds 19 seconds 0 seconds 8 seconds 1 seconds
500- 1999
0-9 seconds 42.53% 61.87% 9.54% 76.12% 39.87% 72.86%
10-29 seconds 37.92% 22.01% 60.50% 9.25% 30.38% 12.42%
30-59 seconds 11.76% 5.57% 20.17% 8.60% 15.43% 6.06%
60-299 seconds 7.52% 9.68% 6.98% 5.54% 13.26% 8.31%
300-1800 seconds 0.17% 0.58% 0.84% 0.33% 0.72% 0.25%
z 1800 seconds 0.09% 0.29% 1.97% 0.16% 0.34% 0.10%
Median Speed 11 3 19 1 13 1
(continued)
Table 8 (continued)
Execution speed by market center and order size
ts
a
Order Size Market Center
Speed Bin BSE CHX NYSE PHLX PCX NASD
”%
2000-4999
0-9 seconds 21.20% 35.34% 7.75% 21.28% 21.80% 58.50%
10-29 seconds 39.44% 24.56% 57.75% 32.71% 28.04% 18.08%
30-59 seconds 21.57% 13.52% 21.41% 19.95% 25.19% 10.32%
60-299 seconds 17.31% 24.73% 6.88% 24.20% 23.70% 12.74%
300-1800 seconds 0.46% 1.24% 2.51% 1.86% 0.53% 0.24%
> 1800 seconds 0.00% 0.62% 3.69% 0.00% 0.74% 0.12%
Median Speed 21.5 17 22 24 30 2
>=5000
0-9 seconds 10.05% 18.37% 6.90% 6.82% 11.63% 51.45%
10-29 seconds 34.17% 15.92% 51.44% 34.09% 32.56% 15.50%
30-59 seconds 20.60% 20.41% 22.81% 29.55% 26.36% 11.98%
60-299 seconds 32.16% 42.86% 8.88% 29.55% 27.91% 20.45%
300-1800 seconds 2.01% 2.45% 4.70% 0.00% 0.00% 0.62%
> 1800 seconds 1.01% 0.00% 5.26% 0.00% 1.55% 0.00%
Median Speed 37 52 24 38 34 7.5
All Orders
0-9 seconds 52.64% 76.74% 9.04% 87.77% 52.90% 80.97%
10-29 seconds 33.66% 12.21% 61.14% 4.13% 25.96% 8.77%
30-59 seconds 8.55% 4.02% 20.22% 5.27% 10.54% 4.01%
60-299 seconds 4.95% 6.48% 6.95% 2.44% 9.77% 6.04%
300-1800 seconds 0.16% 0.40% 0.91% 0.28% 0.61% 0.15%
> 1800 seconds 0.04% 0.15% 1.73% 0.11% 0.23% 0.01%
Median Speed 9 L 20 0 9 1
150 B. Hatch, R. Battalio, R. JenningdThe Financial Review 38 (2001) 123-152

other market centers. The exception is the BSE. The BSE and the Nasdaq InterMarket
usually fill oversized orders faster than the other market centers.
One constant in execution quality is that the Chicago Stock Exchange consist-
ently ranks at or near the bottom. The CHX also ranked poorly in the Lightfoot,
Martin, Peterson, and Sirri (1998) study. Yet, according to the NYSE Fact Book,
since 1988 of all the regional exchanges, the CHX has annually attracted the most
volume in NYSE-listed securities. Clearly, the CHX does not attract order flow
because of its excellent execution quality. Perhaps this volume indicates that dealers
on the CHX offer more payment for order flow than other market centers, thus
making their execution quality competitive with other venues once these payments
are netted out.
Execution quality is a multidimensional concept. Maximizing execution quality
for an investor also depends on that particular investor's preferences for that particu-
lar order. Thus, to provide useful information to investors, we must produce execution
quality reports that condition on a sufficient number of parameters and measure
execution quality on a sufficient number of dimensions.
Finally, if we look only at trade prices or execution speeds, we cannot determine
whether brokers should send orders to market centers other than the primary ex-
change. Investors are interested in all-in-costs of trading. Any computation of total
trading costs includes brokerage commissions. If brokers can obtain some type of
cost advantage from trading away from the primary exchange and if they pass some
of that advantage on to their customers by charging lower commissions, then we
should consider this fact before concluding that customers would prefer that order
flow be automatically routed to the primary exchange.

References
Admati, A. and P. F'fleiderer, 1988. A theory of intraday patterns: Volume and price variability, Review
of Financial Studies 1, 3-40.
Angel, J., 1994. Who gets price improvement on the NYSE? Working paper FIN&-3377-12-694, George-
town University.
Arnold, T., P. Hersch, J. Mulherin, and J. Netter, 1999. Merging markets, Journal of Finance 54,
1083-1107.
Bacidore, J., R. Battalio, and R. Jennings, 2000. Depth improvement and adjusted price improvement
on the NYSE. Working paper, New York Stock Exchange.
Bacidore, J., K. Ross, and G. Sofianos, 1999. Quantifying best execution at the NYSE. Working paper,
New York Stock Exchange.
Barclay, M., W. Christie, J. Harris, E. Kandel, and P. Schnltz, 1999. The effects of market reform on
the trading costs and depths of Nasdaq stocks, Journal of Finance 54, 1-34.
Battalio, R., J. Greene, and R. Jennings, 1997. Do competing specialists and preferencing dealers affect
market quality? Review of Financial Studies 10, 969-994.
Battalio, R., J. Greene, and R. Jennings, 1998. Order flow distribution, bid-ask spreads, and liquidity
costs: Merrill Lynch's decision to cease routinely routing orders to regional stock exchanges, Journal
of Financial Intermediation 7 , 338-358.
Battalio, R., B. Hatch, and R. Jennings, 2000. All else equal? A multidimensional analysis of retail
market order execution quality. Working paper, Indiana University.
Bessembinder, H., 1999. Trade execution costs on Nasdaq and the NYSE: A post-reform comparison,
Journal of Financial and Quantitative Analysis 34, 387-407.
 B. Hatch, R. Battalio, R. JenningdThe Financial Review 38 (2001) 123-152 151

Bessembinder,H. and H. Kaufman, 1996.Quotations and trading costs on the domestic equity exchanges.
Working paper, Arizona State University.
Easley, D. and M. O’Hara, 1987. Price, trade size, and information in security markets, Journal of
Financial Economics 19, 69-90.
Ellis, K., R. Michaely, and M. O’Hara, 2000. The accuracy of trade classification rules: Evidence from
Nasdaq, Journal of Financial and Quantitative Analysis 35, 529-552.
The Exchange, 1996. The auction market versus preferencing, June, page 3.
The Exchange, 2000. Committee reports on market structure, April, page 2.
Finucane, T., 2000. A direct test of methods for inferring trade direction from intra-day data, Journal
of Financial and Quantitative Analysis 35, 553-576.
Handa, P., R. Schwartz, and A. Tiwari, 1999. Price improvement and price discovery on a primary
market, The Journal of Portfolio Management 25, 55-64.
Harris, L. and J. Hasbrouck, 1996. Market vs. limit orders: The SuperDOT evidence on order submission
strategy, Juurnal of Financial and Quantitative Analysis 3 1, 212-232.
Investment Dealer’s Digest, 1995. Smith Barney and A.B. Edwards study Memll’s NYSE Program,
November 6, page 8.
Klock, M., 1999. The SEC’s new Regulation ATS: Placing the myth of market fragmentation ahead of
economic theory and evidence, Florida Law Review 51, 753-797.
Knez, P. and M. Ready, 1996. Estimating the profits from trading strategies, Review ofFinanciaZ Studies
9, 1121-1163.
Lee, C., 1993. Market integration and price execution for NYSE-listed securities, Journal of Finance
48, 1009-1038.
Lightfoot, L., P. Martin, M. Peterson, and E. Sim, 1998. Preferencing and market quality on U.S. equity
exchanges. Working paper, Securities and Exchange Commission.
Macey, J. and M. OHara, 1997. The law and economics of best execution, Journal of Financial
Intermediation 6, 188-233.
McInish, T. and R. Wood, 1996. Competition, fragmentation, and market quality. In Andrew T. Lo,
Ed., The Industrial Organizationand Regulation of the Securities Industry. (The University of Chicago
Press, Chicago, IL).
New York Stock Exchange Fact Book, 1999. (New York Stock Exchange, Inc., New York, NY).
Petersen, M. and D. Fialkowski, 1994. Posted versus effective spreads: Good prices or bad quotes,
Journal of Financial Economics 35, 269-292.
Ross, K., J. Shapiro, and K. Smith, 1996. Price improvement of SuperDOT market orders on the NYSE.
Working paper, New York Stock Exchange.
Securities and Exchange Commission, 1994. Market 2000 Report: Study Ill, Market Fragmentation ,
Competition, and Regulation.
Securities and Exchange Commission, 1997. Report on the Practice of Preferencing, April 11.
Securities and Exchange Commission, 2000. Release No. 34-42450, Commission request for comment
on issues relating to market fragmentation, February 23.
Securities and Exchange Commission, 2000. Release No. 34-43084, Disclosure of order routing and
execution practices, July 28.
Securities Week, 1997. SEC warns Big Board to be ‘fair’ in best execution probe of member firms,
March 31, 1.
Stoll, H., 2000. Friction, Journal of Finance 55, 1479-1514.
US. Court of Appeals for the Third Circuit, 1998. Decision in re: Newton v. Memll Lynch, Pierce,
Fenner & Smith Inc., 135 F.3d 266.
Van Ness, B., R. Van Ness, and S. Mitt, 1999. An empirical examination of the NasdaqlCHX dual-
trading experiment, The Financial Review 34, 65-78.
Van Ness, B., R. Van Ness, and W. Hsieh, 1999. Nasdaq and the Chicago Stock Exchange: An analysis
of multiple market trading, The Financial Review 34, 145-158.
Wall Street Journal, 1995. Merrill shifts on small investors’ orders, November 10, Cl.