Price Discovery in the U.S. Stock Options Market∗

YUSIF E. SIMAAN †

Graduate School of Business, Fordham University

L IUREN WU‡

Zicklin School of Business, Baruch College

This version: December 16, 2003

First draft: March 1, 2002

∗We thank Automated Trading Desk, LLC for providing data and computing assistance. We thank Robert Battalio, Joel Has-

brouck, Charles Jones, Maureen O’Hara, Martin Resch, Dan Weaver, David Whitcomb, and seminar participants at the 2003

Western Finance Association and the 2003 European Finance Associationfor insightful comments. We also thank Eric Crampton

for help on the data preprocessing, and Sandra Size Moore for copy editing. Part of the results in this paper were circulated in an

earlier draft titled “Price Discovery in the Equity Options Market: An Integrated Analysis of Trades and Quotes.”†113 West 60th Street, New York, NY 10023; tel: (212) 636-6116; fax:(212) 765-5573;simaan@attglobal.net.‡One Bernard Baruch Way, Box B10-225, New York, NY 10010; tel: (646) 312-3509; fax: (646) 312-3451;

Liuren Wu@baruch.cuny.edu; http://faculty.baruch.cuny.edu/lwu/.

1

Price Discovery in the U.S. Stock Options Market

ABSTRACT

Five U.S. exchanges compete to provide quotes and attract order flows on a common set of stock

options: the American Stock Exchange, the Chicago Board of Options Exchange, the International Se-

curities Exchange, the Pacific Stock Exchange, and the Philadelphia Stock Exchange. In this paper,

we investigate the price discovery in the U.S. stock optionsmarket. Our analysis shows that the newly

founded, fully electronic International Securities Exchange has become the leader in providing options

quotes that are the most informative, the most binding, and also the most executable.

Price Discovery in the U.S. Stock Options Market

The financial industry has come to recognize the importance of derivativesecurities in providing unique in-

vestment opportunities and risk-management vehicles. Trading volume on derivatives has increased tremen-

dously during the past few years. Accompanying this expansion in derivatives have been rapid expansions

and transformations in the options market. In the United States, in an effort to reduce trade-throughs and

other market segmentations, a series of regulatory changes have taken place since the late 1990s. Currently,

five options exchanges compete to provide quotes and attract order flowson a common set of stock op-

tions: the American Stock Exchange, the Chicago Board of Options Exchange, the International Securities

Exchange, the Pacific Stock Exchange, and the Philadelphia Stock Exchange.

Compared to the vast empirical market microstructure literature on the stock market, there has been

little research on the market microstructure, and price discovery in particular, of the options market. In this

paper, we investigate the price discovery process on the most actively traded options that are listed on all

five U.S. stock options exchanges. Based on the real-time feeds from the Options Price Reporting Authority

(OPRA) during January of 2002, we filter out both the quotes and tradeson the 50 most actively traded

stock options. We measure the Hasbrouck (1995) information share by using the second-by-second quote

book, and we analyze the link between price discovery and other market conditions. We also investigate the

general statistical properties of the bids and asks from each exchangeand analyze how they match with the

transactions.

Our analysis shows that the newly founded, fully electronic exchange, the ISE, has become the leader

in providing the most informative quotes. Comparing the information-share estimates with the trading

activities of each option contract shows that ISE’s leadership in price discovery is more pronounced when it

has a larger market share in the options contract and when the options contract has higher aggregate trading

activity.

We also find that the quotes from the ISE have the narrowest mean bid-askspread and represent the

national best bid or offer (NBBO) for most of time. In contrast, the other four traditional exchanges rarely

provide quotes that are at the NBBO alone. Furthermore, by comparing theoptions transaction price to the

quotes at different exchanges, we find that transactions on the ISE occur exactly at the bid or offer for 84.48

1

percent of the time, the highest among all five options exchanges. This observation shows that quotes from

the ISE are not only the most binding, but also the most executable.

The ISE leads in both price discovery and in providing the tightest and the most executable quotes.

The differences among the quotes from the other four traditional options exchanges are much smaller. The

relative ranking of the four traditional exchanges also differs when weexamine them by different measures.

The modified outcry systems on the CBOE and the PCX are better than the specialist systems on the AMEX

and the PHLX in providing more informative quotes. On the other hand, the average bid-ask spreads on the

AMEX and the CBOE are narrower than those on the PHLX and the PCX. We also find that quotes on the

PCX have the widest spread and are noncompetitive most of the time, but trades on the PCX are executed

inside its posted spread for over 70 percent of the time.

The microstructural design at the ISE differs from that at the traditional exchanges in several important

ways. On the traditional floor-based option exchanges, only one marketmaker drives the quotes for each

option contract, and the identity of this market maker is known to the public. This market maker is referred

to as a specialist at the AMEX and PHLX, a designated primary market maker (DPM) at the CBOE, and a

lead market maker (LMM) at the PCX. In contrast, on the ISE, 11 market makers, including one primary

market maker (PMM) and ten competitive market makers (CMMs), send in quotes electronically via their

respective terminals to a central quote-consolidating system. The market observes a consolidated quote

book from these 11 market makers. The identity of the specific market makerthat underlies each quote is

not known to the public.

Our empirical analysis shows that the microstructural design at the ISE leads to better quotes in terms

of price discovery, average spreads, and executability. Over time, the high-quality quotes at the ISE have

also attracted order flows. The ISE became the third largest options exchange in the United States in trading

volume 18 months after its launch. By mid 2003, the market share of the ISE hadbecome the largest in

options trading volume, excluding the S&P index options, which are solely licensed to the CBOE.

We believe that our paper is the first study to address the price discoveryissue among different options

exchanges. The few known empirical microstructure studies on the optionsmarket include Battalio, Hatch,

and Jennings (2003); Chakravarty, Gulen, and Mayhew (2003); deFountnouvelle, Fishe, and Harris (2003);

2

Mayhew (2002); Neal (1987, 1992); and Wang (2000). None of these studies directly address the issue of

price discovery among different options exchanges.

Battalio, Hatch, and Jennings (2003) use a data sample that is similar to ours, and investigate whether

the options market approaches that of a national market system. They findthat compared to the options

market in June 2000, the market much more closely resembles a national market during our sample period

in January 2002. Chakravarty, Gulen, and Mayhew (2003) analyze the price discovery between the stock

market and the stock options market. Their analysis is based on an old sample (1988-1992) and option quotes

from a single options exchange (the CBOE). The landscape of the optionsmarket has changed dramatically

since then. Since 1999, many option classes list on several exchanges (de Fountnouvelle, Fishe, and Harris

(2003)). The options quotes are now legally required to be firm. Most important of all, the new exchange,

the ISE, has become the leader over the traditional exchanges in providingthe most informative quotes.

In this paper, Section 1 provides some background information on the unique characteristics of the op-

tions market making. We also discuss the recent development in the U.S. options market, the microstructure

design of the new options exchange, the ISE, and its key structural differences from the four traditional

options exchanges. Section 2 describes the data source and the criteria for our sample selection. Section 3

describes the econometrics underlying our price discovery analysis. Section 4 discusses the results on the

price discovery in the options market. Section 5 provides further analysis of the trades and quotes on the

five options exchanges. Section 6 concludes.

1. The Market Microstructure of the Options Exchanges

Option prices provide information about the underlying security that is not readily available from the primary

security market. The price quote of a stock represents a mean valuation of the stock, but the prices of options

underlying this stock at the whole spectrum of strike prices and maturities present a complete picture on

the conditional distribution of the stock value at different possible realizations and conditioning horizons.

Therefore, it is important to understand the information flow in the options market.

Market making in the options market has its own unique characteristics. An important characteristics

of the options markets is the different risk profiles faced by the market makers. In the options market,

3

underlying one stock are hundreds of options at different strike prices and maturities. The values of these

options are inherently linked by many no-arbitrage relations. When a marketmaker adjusts her quote on

one option contract, she should also consider the quotes on other options underlying the same stock, so that

she will not be locked in an arbitrage trading program.

The options market maker must also pay constant attention to the underlying stock market. Any stock

quote update necessitates the updates of hundreds of options underlyingthis stock. For example, an informed

trader who knows that a stock price will go up soon can buy the stock up to the ask size of the stock market

maker. On the options market, this informed trader can simultaneously buy all thecalls and sell all the puts

underlying this stock. Therefore, if we think of the risk exposure of the stock market maker as her current

ask size, then the risk exposure of the options market maker is the summation ofher ask sizes on all call

options and her bid sizes on all put options underlying this stock.

The unique characteristics of the options market dictate that the options market maker needs to have not

only the intricate knowledge of many different no-arbitrage relations, butalso the technology to update her

quotes rapidly across all options underlying the same stock as each stock quote updates. Furthermore, the

increased risk exposure due to the highly correlated moves among all options underlying the same stock also

necessitates the options exchanges to create a market microstructure to protect themselves.

1.1. The recent development of the options market

In July 2000, to reduce trade-throughs and other market segmentations,the Securities and Exchange Com-

mission (SEC) approved a plan to electronically link the various market centers.1 The SEC has also adopted

more stringent quoting and disclosure rules on the options market.

Another important development in the options market was the launch of a new exchange, the ISE. The

ISE started its first day of option trading on May 26th, 2000. By November 2001, the ISE had grown to

become the third-largest U.S. options exchange, trailing only the CBOE and the AMEX among the nation’s

1See Securities Exchange Act Release No. 43086 (July 28, 2000), 65 FR 48023 (August 4, 2000) (“Linkage Plan”).

4

five options exchanges.2 The traditional exchanges are facing significant pressure to either modify their

exchange structures to be more in line with the market trend, or face the risk of being forced out of the

competition. Thus, the need for microstructure research on the options market is more important and urgent

than ever.

1.2. The microstructure of the options exchanges

The four floor-based legacy exchanges provide market making for equity options under two slightly different

structures. The AMEX and the PHLX apply a specialist structure resemblingthat used in stock markets. The

CBOE trades options under a “Designated Primary Market Maker” (DPM), a modification of the original

open outcry structure used in the futures pit. The PCX also follows a similar modified outcry structure, and

trades options under a “Lead Market Maker” (LMM). The responsibilities of the DPM or LMM include

disseminating quotes, providing liquidity to thin markets by trading on his own account, and representing

public limit orders. The roles are similar to those of a specialist. The difference is that options traded under

a DPM may also be traded by other market makers. However, the DPM maintainsthe right to a certain

percentage of the public order flow (Mayhew (2002)).

In contrast to the four traditional exchanges, market making on the ISE is fully electronic and does

not have a floor. Market makers send in their quotes via their quoting engines. These quotes are then

consolidated via a centralized consolidating system. Both ISE’s central exchange system and trading stations

are provided by OM, part of the Swedish OM Group. APIs (Application Programmers Interface) on a

variety of platforms are provided to allow broker/dealers to link their order delivery systems to ISE’s order

management system. In this way, orders from participating broker/dealerscan be routed, executed, and

reported all electronically with no paper.

The ISE combines auction market principles with electronic trading. The ISE isnot an electronic com-

munications network (ECN) or alternative trading system (ATS), but an SEC-registered exchange. The ISE

has three classifications of members, all of whom are registered broker/dealers. They include ten Primary

Market Makers (PMMs), one hundred Competitive Market Makers (CMMs), and an unlimited number of

2The market share is reported in a news release from the ISE, based ontotal volume transacted in all listed options, equity as

well as index products, traded during the month of October, 2001.

5

broker/dealers functioning as Electronic Access Members (EAMs). In the ISE system, floor brokers are not

needed, since orders are input by broker/dealers directly to the point-of-sale (electronic order book).

PMMs are market makers with significant market-making responsibilities, including overseeing the

opening, providing continuous quotations in all of their assigned stock options, and ensuring that customer

orders are not automatically executed at prices inferior to those available at other options exchanges. Op-

tions traded on the ISE are divided into ten groups (“bins”), with one PMM assigned to each bin. PMMs

may also conduct a limited amount of trading (up to ten percent of their quarterly contract volume) in other

options traded on the ISE. A PMM must purchase or lease a PMM membership.PMMs are not permitted to

represent agency orders.

CMMs are market makers who quote independently and add depth and liquidityto the market. Each

CMM is required to provide continuous quotations in no less than 60 percent of the stock options in their

assigned group. There are up to ten CMMs appointed to each bin of options. CMMs may also conduct a

limited amount of trading (up to 25 percent of their quarterly contract volume)in other options traded on

the ISE. Like a PMM, a CMM must purchase or lease a CMM membership. Also,CMMs are not permitted

to represent agency orders.

EAMs are broker/dealers that represent agency and/or proprietaryorders on the exchange. An EAM

does not purchase a membership. Rather, once approved as an ISE member, an EAM pays an access fee that

permits the firm to place orders in all of the options traded on the exchange. EAMs cannot enter quotations

or otherwise engage in market making on the exchange.

An organization may obtain more than one membership. It is possible to be a PMM inone group of op-

tions, obtain several CMM memberships to provide markets in other groups, and have an EAM membership

to enter agency and proprietary orders in all groups. Firms that are bothmarket makers and EAMs must

conduct those activities separately. A member may not be both a PMM and CMMin the same group of

options.

The quotes and trades on the ISE proceed as follows: One PMM and ten CMMs provide quotes through

their own electronic quote engine on each options contract. These quotes are sent electronically to a central

quote consolidating machine provided by OM. The quote-consolidating machine generates a quote book

6

(bid, bid size, ask, ask size) that represents the best bid and ask fromthe 11 market makers. The market

(public) has access to the consolidated quote book, but not the market maker’s identity underlying each

quote.

On the other side, the unlimited number of EAMs enter their orders through theirelectronic terminals.

These orders are consolidated to generate an order book. Finally, the order book and the quote book meet in

the trading engine, also provided by OM, to generate transactions. The flow chart in Figure 1 summarizes

the quote and trade procedures at the ISE.

1.3. Key structural differences between the options exchanges

The key difference between the ISE and the four traditional exchangesis the number of market makers

who can drive quotes. On the ISE, 11 market makers can send in two-sided quotes into each option class.

On the traditional exchanges, only one market maker can drive quotes. As a result, on the four traditional

exchanges, one market maker must be on both sides of the market, bid and offer, but on the ISE it is normal

for the bid to be from one market maker and the offer from another marketmaker. For some actively traded

options, the firms that are the specialists (LMMs, DPMs) on the traditional exchanges are often also CMMs

on the ISE. Most of the time, these specialists provide the same quote to both exchanges. Therefore, for

these options, the traditional exchanges will almost never be the best quoteby itself.

Another difference between these exchanges is the anonymity of the market makers who send the quote.

On the traditional exchanges, it is public information on which firm is the specialist that provides the quotes.

But on the ISE, the quotes can come from any of the 11 firms. The identity of the market maker underlying

each quote on the consolidated quote book is essentially anonymous. Simaan,Weaver, and Whitcomb

(2003) find that market makers often quote more aggressively under anonymity.

The methods that can be used to drive quotes on the five exchanges are also different. The specialists at

the AMEX and PHLX can only enter quotes through their respective exchange’s trading system. The PCX

and the CBOE allow the firms to drive quotes two ways, either through the exchange’s trading system or

remotely through the firm’s own quote engine. In contrast, the ISE does nothave a floor. Market makers

enter quotes only through their quote engine. Without a floor and floor brokers, transactions at the ISE are all

7

executed and reported electronically via user terminals. Orders from participating broker/dealers are routed,

executed, and reported electronically with a turnaround time of under one second. Such an automated system

can reduce the lags between trades and reports. It also provides quotes that are firmer and more reliable. In

essence, unless there are unusual conditions, any posted quotes arefirm and can be executed on until the

next update.

2. Data Source and Sample Selection

We extract the quotes and trades data on stock options from the electronic message feeds of S&P ComStock

XpressFeed. The feeds contain updates on both quotes and transactions, including both the quote or trade

prices and sizes.

The options data provider is the Options Price Reporting Authority. OPRA communicates to the public

all of the transactions and quotations from each options exchange through the facilities of its processor,

the Securities Industry Automation Corporation (SIAC). Each transaction and quote update is reported to

OPRA as a “message.” The options markets generate such messages for asubstantial number of products.

Currently, there are approximately 3,900 equity securities and indexes underlying listed option products,

and more than 178,000 individual option series. Trade and quote data aregenerated continuously for each

options product listed on each options exchange during the hours that markets are open.

Quote message traffic comprises most of the options message traffic. For example, in February 2000,

the average number of quotes per day was 37.5 million, while the average number of trades per day was

183,000.3 Quotes are usually generated automatically for individual options series based on changes in the

underlying stock price or index value: Every time a price changes for a particular equity security, the quotes

for all of the options on that security or an index in which that security is represented can be automatically

updated on each exchange that trade those options.

We select option contracts that are traded on all five exchanges and across all 20 business days in January

2002. During that month, 70,946 option contracts are traded on the five options exchanges. Five percent

3SEC Release No. 34-43621; release date: November 27, 2000.

8

(3,397 contracts) of them are traded simultaneously in all five exchanges at least for one day. Of the 3,397

contracts, 835 of them are traded on all business days in January. Theother contracts are either not as active

or expire in the middle of January.

Trading activities vary significantly among these 835 options contracts. Themost actively traded option

in both trades and volume is CYQBD, a February 2002 (maturity) call option onCSCO with a strike price at

$20.00. The option has 3,608 trades (239,100 contracts) over the 20 days in January 2002, an average of 180

trades per day. The least active in number of trades is VMFAT, a January 2003 call option on Microsoft with

a strike price at $100.00. This option only trades 78 times (7,688 contracts) during January 2002, averaging

less than four trades a day. The least active in trading volume is EMCDV with only 120 trades and 709

contracts. It is an April 2002 call option on EMC with strike price at $12.50.

Although all the 835 options contracts are traded on all five exchanges, for 290 of these contracts one

exchange takes up more than half of the market share in number of trades.For our empirical analysis, we

focus on the contracts where no exchange has a market share higher than 50 percent. Then, we choose the

50 most active contracts in number of trades from among them.

The 50th contract, the least active, is CYQND, a February 2002 put option on CSCO with a strike

price of $20.00. This contract has 833 trades in January 2002, averaging about 42 trades per day. The quote

updates are sufficiently frequent to span the trading activities. For the selected 50 option contracts in January

2002, the median quote updating frequency is about seven seconds.

Table 1 reports the summary statistics of the 50 options contracts. Panel A reports the total number of

trades and the market share for each exchange in January 2002 for the 50 options. The number of trades

ranges from 833 for CYQND to 3,608 for CYQBD, with an average of 1,431 trades over 20 days. On

average, 20.8 percent of the trades occur on the AMEX, 29.6 percenton the CBOE, 16.2 percent on the ISE,

25.2 percent on the PCX, and 8.2 percent on the PHLX.

Panel B reports market shares based on the trading volume. Compared to the market share based on

number of trades, the market share based on trading volume increases for the AMEX and the ISE, but

decrease for the PCX. This difference in market share implies that on average, transactions on the AMEX

9

and the ISE have a larger size than transactions on the PCX. As a result, thePCX reports more trades but

less volume.

3. Econometric Specifications

3.1. Data processing

We filter out the quotes and trades on the 50 active option contracts secondby second. We bundle trades that

happen at the same exchange at the same time (second) and transaction price. For the quote updates, if we

observe more than one quote update on a certain contract from a certain exchange within the same second,

we pick the last one in the sequence of the electronic message. Then, we expand the quote updates into a

second-by-second book. In this expanded book, quotes remain the same until the next update.

Using quotes from the five options exchanges, we also construct a series of national best bids and offers.

We define the national best bid as the maximum of the available bid quotes from the five options exchanges

and the national best ask as the minimum of the available ask quotes from the five options exchanges.

Therefore, for each options contract, we have six second-by-second series on the bid and six second-by-

second series on the ask. For each exchange, we also have a time-stamped series of transactions that include

both the transaction prices and transaction sizes.

3.2. Price discovery among the five options exchanges

When different exchanges provide competitive quotes on the same security, these quotes are cointegrated,

because all these quotes are noisy representations of the same fundamental value. Although each series

of bid or ask quotes may be nonstationary, the differences between the quotes from different exchanges

are stationary, representing temporary deviations from the long-run equilibrium relation. Engle and Granger

(1987) propose an vector error correction model (VECM) accommodating the cointegrating relation. Further

assuming that the fundamental value of the security price follows a random walk, Hasbrouck (1995) uses

this VECM specification to derive an information share measure that captures the percentage of the variance

10

of the permanent component explained by each quote series.4 Here, we use this information share measure

to gauge the price discovery in the five options exchanges.

We use a price vectorpt = [p1t , p2t , · · · , pnt]⊤ to denote the quotes from different exchanges. We treat

bids and asks separately and usen = 5 to denote the five options exchanges. We assume that underlying all

these price quotes is a random walk component that represents the true value of the security, or the efficient

price. The difference between the quotes and the true value constitutes transient noise. Therefore, quotes

from each exchange are nonstationary, but quotes across exchanges are cointegrated. We write out the vector

error correction model of orderk as follows:

∆pt = A1∆pt−1 +A2∆pt−2 + · · ·+Ak∆pt−k + γ(zt−1−µz)+ut , (1)

where the coefficient matricesAi , i = 1,2, · · · ,k, are square matrices of ordern, ut denotes the innovation

vector with covariance matrixΩ, and(zt−1−µz) denotes the error correction term with

zt = [p1t − p2t , p1t − p3t , · · · , p1t − pnt]⊤ , (2)

andµz being the mean vector ofzt . The vector moving average representation of the model is

∆pt = B0ut +B1ut−1 +B2ut−2 + · · · , B0 = I . (3)

We calculate the moving average coefficientsB by forecasting the system subsequent to a unit perturbation.

Our primary interest is the cumulative impulse response function:

Ck =k

∑i=0

Bk. (4)

4Recent microstructure applications of the Hasbrouck information shareinclude Chakravarty, Gulen, and Mayhew (2003),

Hasbrouck (2003), and Huang (2002).

11

The first columns of theCk matrix describe the prices subsequent to a shock in the first price, etc. Of

particular importance is the response to the permanent component by taking the limit,

C = limk→∞

Ck. (5)

The rows ofC are identical. If we usec to denote any row ofC, the variance of the common random walk

component of the quotes is

σ2w = cΩc⊤. (6)

If Ω is diagonal, we can define the information share of thei-th market as

ISi =c2

i Ω2ii

σ2w

. (7)

WhenΩ is not diagonal, the information share is not uniquely defined. We compute thelower and upper

bound of the information share by considering the Cholesky factorization of the all the rotations of the

disturbances.

We estimate the VECM model in equation (1) each day on the five ask series andthe five bid series

from each options contract separately. We specify a lag of ten minutes (600 seconds). To reduce the size

of the parameter set, we apply polynomial distributed lags (Greene, 1993) over lags 1-10 (seconds), 11-20,

and 21-30, and then apply moving averages on lags 31-60, 61-120, 121-300, and 301-600. To compute the

impact of the permanent component in (5), we letk = 10,800 (three hours). When we experiment with

different lags and different averages, the results are qualitatively similar.

4. Price Discovery in the Stock Options Market

We estimate the VECM model in equation (1) each day on each options contract.We estimate the model

using the five bid series and the five ask series separately. From the estimated model, we compare the lower

bound and upper bound estimates of the information share from each options exchange.

12

4.1. Average information share

Table 2 reports the sample averages of the information-share estimates for the five option exchanges. Panels

A and B are lower (Min) and upper (Max) bound estimates on the bids and Panels C and D are the lower

and upper bound estimates on the asks. We obtain the average across 1,000 sample estimates from the 50

options contracts multiplied by the 20 business days in January 2002. In each panel, we report the maximum

(MAX), mean (MEAN), minimum (MIN), and standard deviation of the sample estimates.

Of the five options exchanges, the ISE has the largest average information share (bolded numbers),

regardless of whether the comparison is based on the lower or upper bounds of the information share, or

whether we estimate the model using bid or ask quotes.

Differences among the other four exchanges are much smaller. The average information share by bid

quotes shows the following ranking, from the highest to the lowest: ISE, CBOE, PCX, AMEX, and PHLX.

The ranking by the ask quotes is: ISE, PCX, CBOE, AMEX, and PHLX. The ranking of PCX and CBOE

is different depending on whether we measure the information share usingbids or asks. However, in either

case, the difference between the information-share estimates of these two exchanges is very small.

The average ranking separates the three major microstructure designs ofthe five exchanges. On top

of the ranking is the ISE. The second group includes the CBOE and the PCX. The least informative quotes

come from the specialist systems adopted at the AMEX and the PHLX. As discussed in Section 1, these three

groups have three different types of microstructure design for optionsmarket making. Our information-share

analysis shows that the design at the ISE leads to the most price discovery.

To determine the statistical significance of the leadership of the ISE over other exchanges in terms of

price discovery, we investigate the statistical properties of the information-share difference between the ISE

and the other four options exchanges.

L i ≡ ISISE− ISi , i = AMEX,CBOE,PCX,PHLX. (8)

We label the differenceL i as the leadership of the ISE over the options exchangei. A positive estimate for

the leadership measure indicates that the ISE leads the other exchange in providing more informative quotes.

13

We construct at-statistic on the leadership over the four options exchanges based on the mean and standard

deviation estimates of the 1,000 samples,

t-statistici =AVERAGE(L i)

STDEV(L i)/√

N, (9)

whereN = 1000 denotes the number of sample points, AVERAGE denotes the sample average, and STDEV

denotes the sample standard deviation estimate. Table 3 reports thet-statistic estimates, which are all

strongly significant, regardless of whether we use the lower or upper bounds, bids or asks.

Thet-statistics also reveal another interesting feature of the options quotes. The statistics are larger for

asks than for the corresponding bids. The differentt-statistics imply that the price leadership of the ISE is

more significant in its ask quotes than in its bid quotes.

Also informative is the cumulative impulse response functions of the VECM model from different ex-

changes, as plotted in Figure 2. The left panels are based on estimates using the bid quotes and the right

panels are based on the ask quotes. These plots show the cumulative priceimpacts implied by an initial

unit shock to the quote in one exchange. For 50 option contracts over 20 business days, we have 1,000

estimates for each point in the response function. The lines in Figure 2 represent the sample averages of

these estimates.

By construction, att = 0, the impact is unity on one exchange and zero on others. The five figures

under each panel represent a unit initial shock on each of the five options exchanges. In the long run, the

impact for each initial shock is essentially identical for all exchanges. Thisconvergence is a consequence of

cointegration. However, the convergence speed and level differ fordifferent options exchanges. The most

obvious difference is between the behavior of the ISE and the behavior of the other four options exchanges.

For a unit initial shock to any other exchanges, the ISE always responds the fastest and converges to the

stationary state the earliest. On the other hand, when the unit shock is on the ISE, this unit shock has a larger

permanent impact on the market than all other cases.

The conclusions from bids and asks are approximately the same. The permanent impact of a unit shock

from the ISE is about 23 percent, and that from the other four exchanges is about 20 percent or less.

14

Therefore, based on the estimated VECM model and Hasbrouck (1995)’s information-share measure,

we find that out of the five options exchanges, the newly launched ISE stands out in providing the most

informative quotes. The mean information-share estimate is the largest for theISE quotes. A unit shock

on the ISE quotes has a larger permanent impact on the market than a unit shock on any other options

exchanges. The difference among the four floor-based options exchanges is much smaller.

4.2. Information share and the market share

Although all the selected option contracts trade at all five exchanges, the relative activity across different

exchanges varies significantly. For example, when we measure the marketshare in number of trades, we

find that for CYQBD, the most active multiple-listed option contract, CBOE accounts for 43 percent of the

trading activity but PCX accounts for a meager 16 percent. In contrast, for QQQBN, the second most active

contract in our sample, CBOE only accounts for 12 percent of the tradingactivity but PCX accounts for 41

percent.

When the trading activities of a certain contract concentrate on one exchange, we expect that this ex-

change has more incentive to keep its quotes updated and lead in price discovery. From another perspective,

the trading of a certain contract might be concentrated at one exchange because that exchange provides

the most informative quotes. Both arguments imply a positive correlation between market share and price

leadership.

To investigate how the concentration of the trading activity of a certain optionscontract is related to

the price leadership of that contract, we compute the average price leadership of ISE over the other four

exchanges on that contract,

L =180

20

∑t=1

4

∑i=1

L t,i , i = AMEX,CBOE,PCX,PHLX. (10)

Then, we compute the correlation between this ISE average price leadership and the market share of ISE on

that contract, both by the number of trades and by the trading volume.

In Table 4, in the columns titled “Market Share of ISE,” we report the correlation estimates between

ISE’s price leadership as defined in equation (10) and its market share by both the number of trades and the

15

trading volume. We observe positive correlations between ISE’s price leadership and its market share. The

positivity of the correlation estimate is robust whether we measure the information share by using the lower

or upper bounds, from the bids or asks, or whether we measure the market share by the number of trades or

trading volume. Therefore, ISE’s leadership in price discovery is more pronounced when the ISE possesses

a larger market share in trading activities.

Nevertheless, we must interpret our results with caution, because thet-statistics show that many of the

correlation estimates are not statistically significant. Thet-statistics are greater than two only when we

measure the market share of the ISE by trading volume and the information share by using the upper bounds

and the ask quotes. The lowt-statistics suggest that a positive correlation, if it exists, is a mild one.

To visually inspect the relation, we average ISE’s price leadership further across bids and asks, and lower

and upper bounds. Then, in Figure 3 we plot the ISE’s price leadershipon each option contract against ISE’s

market share in that contract. We observe a weak positive correlation between ISE’s market share and its

leadership in price discovery.

From the scatter plots, we also see that ISE’s average leadership in pricediscovery is positive for 45 of

the 50 option contracts. Only five of the contracts generate slightly negativeresults.

4.3. Information share and trading activity

The 50 options contracts also differ significantly in aggregate trading activities. The most active contract

averages 180 trades per day, but the least active contract averages only 42 trades per day. When a contract

is very active and generates large order flows, the exchanges have the incentive to compete and provide the

best quotes on this contract to attract order flow. On the other hand, foran inactive contract, the incentive is

less, since the total order flow on that contract is small. Therefore, stronger competition in the more liquid

contract may more vividly reveal the structural differences among different options exchanges.5

5An options market maker uses the options on QQQ as an example, saying,“We may not actively compete in quoting on many

other options contracts, but we cannot afford not competing strongly for options on QQQ because they constitute the largest order

flow.”

16

The columns in Table 4 titled “Aggregate Trading Activity” report the correlation estimates between the

price leadership of ISE for each contract and the aggregate trading activity of that contract on all five options

exchanges. The correlation estimates are larger than the correlations with market shares. Furthermore,

when we measure the aggregate trading activity by the number of trades, allthe correlation estimates are

statistically significant at 95 percent confidence level. The statistical significance declines when we measure

the aggregate trading activity by trading volume.

The two scatter plots in Figure 4 show the relation between the average price leadership of ISE on each

options contract and that option’s trading activity. We observe a generally positive trend in the relation. We

also see that the ISE’s leadership is strongly positive for the most active options contracts.

If more active contracts lead to stronger competition, which in turn show evenmore clearly the structural

differences between different option exchanges, then the results in Figure 4 show that the leadership of the

ISE in price discovery is due to its microstructural design, not to some other factors such as historical client

relations. The latter factor is also unlikely for the ISE because the ISE has only been operating for a short

time and therefore has less of a legacy.

5. Further Analysis of Quotes and Trades

Our information-share analysis shows that ISE’s quotes are the most informative. However, an informative

quote does not necessary lead to a transaction if the bid and ask spread isvery wide. Furthermore, an

exchange can avoid the duty of market making by providing wide bid-ask spreads but agreeing to execute

trades at the best bid offer available. Hence, it is also important to understand how wide the quoted spread

is and whether the posted quotes are readily executable at each options exchange.

5.1. Quote spreads

The magnitude of the bid-ask spreads is a simple and direct measure of transaction cost when all transactions

happen at the quotes. When trades do not always happen on the quotes, the bid-ask spread reveals the

aggressiveness of a specific exchange in providing binding quotes. Since market liquidity often varies

17

across different times of a day, we divide each day into half-hour periods and compute the average bid-ask

spreads for each exchange under each half-hour period. The average is a simple time average based on the

expanded second-by-second quote book.

Figure 5 plots the average bid-ask spreads, across both time and the 50 options contracts, for quotes

from the five options exchanges. Of the five option exchanges, the quotes from the ISE have the narrowest

average spreads. The average spread from the ISE is about five cents narrower than the next best, quotes

from the AMEX and the CBOE.

From all exchanges, we also observe wider mean bid-ask spreads in themorning when the market has

just opened. The spread declines as trades proceed, flattening out bynoon.6 Nevertheless, this intraday

pattern does not in any way obscure the predominant leadership of the ISE in providing the narrowest bid-

ask spreads.

In an integrated market, all transactions should happen at the NBBO. Quotes outside the NBBO become

nonbinding. To understand the percentage of times when an exchange provides binding quotes, we compute

the percentages of quotes from a certain exchange that are on the NBBOand on the NBBO alone. We

construct the NBBO by defining the highest bid as the national best bid andlowest ask as the national best

ask.

Figure 6 plots the percentage of quotes from each options exchange thatare on NBBO and on NBBO

alone at different times of the day. About 70 percent of time, ISE’s quotes, both bids and asks, are the NBBO.

The next best (AMEX and CBOE) only have 50 percent of their quotes on NBBO. About 15 percent of time,

ISE’s quotes are the only ones that are at the NBBO. In contract, the percentages of times that quotes from

any other exchanges are on the NBBO alone are less than ten percent.

We can trace the narrow bid-ask spread of the ISE quotes to its unique microstructural design. Quotes

from the ISE are consolidated quotes from 11 market makers, but quotesfrom the other four exchanges are

from one specialist only. Many times, the specialist on the four traditional exchanges is also a competitive

market maker on the ISE. Therefore, if the specialist provides the same quotes to both exchanges, the con-

solidated quotes from the ISE are going to dominate the single-source quote from the other four exchanges

6The information story in Easley and O’Hara (1992) can generate such intraday patterns

18

by default. If the specialist provides different quotes to different exchanges, empirical evidence such as in

Simaan, Weaver, and Whitcomb (2003) shows that a specialist is likely to provide more aggressive quotes

when his or her identity is not known to the public. The specialist’s identity is public information on the four

traditional exchanges, but unknown to the public when he or she sends quotes to the ISE. Both arguments

imply that ISE should generate narrower quotes than the other four traditional exchanges do.

Similar to the price discovery result, ISE’s relative aggressiveness in providing binding quotes is also

more obvious in providing asks than in providing bids. The percentage of ISE asks that are on the NBBO

alone is around 16 percent and that for the bids is about 14 percent.

Although the leading position of the ISE is unequivocal in both price discovery and in providing binding

quotes, the ranking of the four traditional exchanges in quoted spreadsis different from the ranking in price

discovery. For example, the quotes from the PCX are more informative thanthose from the AMEX and the

PHLX, but the quoted spreads from the PCX are the widest. The mean bid-ask spread at the PCX is about

ten cents wider than the mean spread at the ISE.

5.2. Quote executability

We investigate whether the options transactions at different exchanges occur at exactly the posted quotes. If

transactions often occur outside the quotes, either the quotes are not firmor the trades are reported with a

delay. On the other hand, if transactions often happen inside the quoted spread at an exchange, either this

exchange is giving preferential treatment to a certain group of clients or itis posting noncompetitive quotes

but agreeing to execute trades at a better price than the posted spreads.Therefore, comparing the percentage

of the trades inside and outside the quoted spreads provides important information on the executability of

the quotes.

We compute the percentage of the transactions at an exchange that occuron, inside, or outside the bid-

ask quotes. Panel A in Table 5 summarizes the results. The most striking observation from this panel is

that the ISE has the highest percentage (84.48 percent) of trades that happen exactly on the bid or ask. This

high percentages shows that the quotes posted at the ISE are not only themost binding, but also the most

executable. In contrast, the PCX has the largest percentages of transactions (70.34 percent) that happen

19

inside their bid-ask quotes. Although the PCX mostly posts noncompetitive quotes, the exchange executes

trades that are mostly inside their posted spread.

Panel B of Table 5 reports the percentages of the inside trades that are outside, on, and inside the NBBO.

The most striking result is on the PCX, where 83.96 percent of the inside trades happen exactly on the

NBBO. This finding, together with its widest bid-ask spread and large proportion of noncompetitive quotes,

implies that the PCX attracts order flows by matching the best bid offer, but avoids providing competitive

quotes and hence is shunning part of their market-making duty.

We can also trace the findings on the PCX to its microstructure design. The PCXhas programmed an

“automatic step up” feature into its automatic execution system for small customer orders. An LMM at the

PCX is able to customize his aggressivesness based on the NBBO and the size of the order. There are six

different settings that vary from full, automatic NBBO execution regardless of the state of the market, to

protect from locked and crossed markets, to one tick step up, and to no step up.

Our analysis of the quotes shows a slight asymmetry in behaviors between theasks and bids. The ISE’s

leadership in price discovery is more pronounced in asks than in bids. Thepercentage of the ISE asks on the

NBBO alone is also two-percentage points higher than the percentage of theISE bids on the NBBO alone.

To understand the reason underlying this asymmetry, we investigate whetherthere is an asymmetry in

the buy and sell transaction. We label a transaction a buy from the customerside if the transaction occurs

on the ask side and a sell if the transaction occurs on the bid side. Panel C of Table 5 reports the ratio of

the number of buys to the number of sells on each options exchange. We classify buys and sells using two

criteria. In the first row in panel C, we classify as a buy transaction if the transaction occurs exactly on the

ask quote of this exchange and a sell if the transaction occurs exactly on the bid quote. Under this criterion,

we ignore trades that happen inside or outside the bid/ask quote. In the second row, we use a more inclusive

criterion by classifying any trades that happen above the mid-quote as buys and below the midquote as sells.

We ignore the trades that happen exactly at the midpoint of the quotes.

The two criteria generate similar results. The ratios of buy to sell transactionsare less than one on all

exchanges, indicating that there are more sell trades than buy trades during our one-month sample period.

The most balanced order flow is from the ISE, where the ratio is close to one.

20

A buy transaction from the customer generates a short position on the optioncontract for the market

maker. This short position generates much bigger risk exposure than a long position due to the unique

payoff structure of an option contract. We conjecture that some market makers on the traditional exchanges

may be reluctant to take short positions and are therefore not as aggressive in competing for providing ask

quotes as in competing for providing bid quotes. As a result, the ISE is alone on the national best ask more

often than it is alone on the national bid. This reluctance to compete on asks also makes ISE’s leadership in

price discovery more pronounced when we estimate it by using the ask quotes.

5.3. The quoting and trading behavior of the PCX

The quotes from the PCX have a unique feature. Although these quotes are rarely binding and hence rarely

useful, when we measure by information share we find that they are more informative than the quotes from

the AMEX or the PHLX. The reason for this different ranking for PCX mightbe that the PCX mimics a

source of quote that is more informative than that for the quotes at the AMEXor the PHLX.

If the PCX always mimics the quotes from a certain exchange with an-second lag and a fixed spread,

the difference between time-t quote from the PCX and the time-(t −n) quote from this leading exchange

should be a constant for allt,

QPCXt −QL

t−n = constant (11)

Furthermore, this constant should be negative for bids and positive forasks. To identify which exchange the

PCX is mimicking, we run the following set of regressions on both bids and asks,

QPCXt = a+bQj

t−n +ejt,n, j = NBBO,AMEX,CBOE, ISE,PHLX; n = 0,1,2, (12)

whereejt,n denotes the regression residuals.7 From the regression that generates the best fit (the highest

R-square), we can identify the exchange (j) as the one that the PCX is the most likely mimicking and the

number of seconds (n) that the mimicking lags the original source. Furthermore, the slope estimate should

be close to unity under the null hypothesis of a fixed spread.

7We also experiment with regressions on the logarithm of quotes, under thenull hypothesis that the added spread is proportional

to the price level. The average results are similar and thus are not reported.

21

We run the regression on each of the 50 option contracts on each of the 20business days. We use the

R-square estimates as a measure for the regression fit. Table 6 reports the sample averages of the regression

R-squares on bids (panel A) and asks (panel B). We observe that theaverageR-square of the regressions

is the largest when we run the regression on the NBBO with zero lags (the bolded numbers). This finding

implies that the PCX possibly mimics the national best bid offer, and does so almost synchronously.

Given a fixed lag, we find approximately the same ranking in terms of theR-square for both bids and

asks. The ranking is, from the largest averageR-square to the smallest NBBO, ISE, CBOE∼ AMEX, and

PHLX. Since quotes from the ISE represent the NBBO about 70 percent of the time (see Figure 6), following

the NBBO is equivalent to following the ISE most of the time. Therefore, it is not surprising to see that the

ISE follows the NBBO in generating the next bestR-square for the above regressions. Furthermore, given

a fixed exchange, the regression with zero-second lag (n = 0) generates the largestR-square, indicating that

the mimicking behavior, when it happens, can be done without visible lags (upto a second).

6. Conclusion

We perform price discovery analysis on the stock options market in the United States and find that the In-

ternational Securities Exchange has become the leader in providing the mostinformative, the most binding,

and the most executable quotes. Perhaps because of both the informativeness and the executability of its

quotes, the ISE has rapidly gained market share in options trading activities. The sample period in this study

shows that the ISE is the third largest in market share. By mid 2003, the ISE had claimed the largest market

share in the options trading activities, excluding the S&P 500 index options.

What makes the ISE quotes the most informative, binding, and executable liesin its key structural

difference from the four traditional options exchanges. In the traditional exchanges, only one market maker

makes the market for a certain class of options. At the ISE, one primary market maker and ten competitive

market makers compete to make market on the same class of options. In fact, most primary market makers

in the traditional exchanges are also competitive market makers on the ISE. When these market makers

provide the same quotes to both exchanges, the consolidated quotes from the ISE almost always dominate

the single-source quote from the other exchanges.

22

Most of the traditional options exchanges have realized the challenges facing them and have been

proposing new or reformed trading systems (such as CBOE direct, PCX plus) that are closer to the de-

sign of the ISE. The market is still rapidly evolving. It will be interesting to seehow these options markets

consolidate and develop.

23

References

Battalio, R., B. Hatch, and R. Jennings, 2003, “Toward a National Market System for U.S. Exchange-Listed

Equity Options,”Journal of Finance, forthcoming.

Chakravarty, S., H. Gulen, and S. Mayhew, 2003, “Informed Tradingin Stock and Option Markets,”Journal

of Finance, forthcoming.

de Fountnouvelle, P., R. P. H. Fishe, and J. H. Harris, 2003, “The Behavior of Bid-Ask Spreads and Volume

in Options Markets During the Competition for Listings in 1999,”Journal of Finance, 58(6), 2437–2463.

Easley, D., and M. O’Hara, 1992, “Time and the Process of Security Price Adjustment,”Journal of Finance,

47(2), 577–605.

Engle, R. F., and C. W. J. Granger, 1987, “Co-integration and Error Correction: Representation, Estimation,

and Testing,”Econometrica, 55, 251–276.

Greene, W. H., 1993,Econometric Analysis. Macmillan Publishing Company, New York, second edn.

Hasbrouck, J., 1995, “One Security, many Markets: Determining the Contributions to Price Discovery,”

Journal of Finance, 50(4), 1175–1199.

, 2003, “Intraday Price Formation in U.S. Equity Index Markets,”Journal of Finance, 58(6), 2375–

2399.

Huang, R. D., 2002, “The Quality of ECN and Nasdaq Market Maker Quotes,” Journal of Finance, 57(3),

1285–1319.

Mayhew, S., 2002, “Competition, Market Structure and Bid-Ask Spreadsin Stock Option Markets,”Journal

of Finance, 57(2), 931–958.

Neal, R., 1987, “Potential Competition and Actual Competition in Equity Options,”Journal of Finance,

42(3), 511–531.

, 1992, “A Comparison of Transaction Costs Between Competitive Market Maker and Specialst

Market Structures,”Journal of Business, 65(3), 317–334.

Simaan, Y., D. G. Weaver, and D. K. Whitcomb, 2003, “Market Maker Quotation Behavior and Pretrade

Transparency,”Journal of Finance, 58(3), 1247–1268.

24

Wang, E., 2000, “Competition Among Exchanges: Does Multiple Listing AffectTrading Costs on Options

Market?,” Working paper, University of Chicago, Chicago.

25

Figure 1Quote and trade procedures at the ISE

26

0 1000 2000 3000 4000 5000 60000

0.05

0.1

0.15

0.2

0.25

Lags in SecondsP

rice

Impa

ct

Initial Unit Shock to AMEX Bids

0 1000 2000 3000 4000 5000 60000

0.05

0.1

0.15

0.2

0.25

Lags in Seconds

Pric

e Im

pact

Initial Unit Shock to AMEX Asks

0 1000 2000 3000 4000 5000 60000

0.05

0.1

0.15

0.2

0.25

Lags in Seconds

Pric

e Im

pact

Initial Unit Shock to CBOE Bids

0 1000 2000 3000 4000 5000 60000

0.05

0.1

0.15

0.2

0.25

Lags in Seconds

Pric

e Im

pact

Initial Unit Shock to CBOE Asks

0 1000 2000 3000 4000 5000 60000

0.05

0.1

0.15

0.2

0.25

Lags in Seconds

Pric

e Im

pact

Initial Unit Shock to ISE Bids

0 1000 2000 3000 4000 5000 60000

0.05

0.1

0.15

0.2

0.25

Lags in Seconds

Pric

e Im

pact

Initial Unit Shock to ISE Asks

0 1000 2000 3000 4000 5000 60000

0.05

0.1

0.15

0.2

0.25

Lags in Seconds

Pric

e Im

pact

Initial Unit Shock to PCX Bids

0 1000 2000 3000 4000 5000 60000

0.05

0.1

0.15

0.2

0.25

Lags in Seconds

Pric

e Im

pact

Initial Unit Shock to PCX Asks

0 1000 2000 3000 4000 5000 60000

0.05

0.1

0.15

0.2

0.25

Lags in Seconds

Pric

e Im

pact

Initial Unit Shock to PHLX Bids

0 1000 2000 3000 4000 5000 60000

0.05

0.1

0.15

0.2

0.25

Lags in Seconds

Pric

e Im

pact

Initial Unit Shock to PHLX Asks

Figure 2Cumulative impulse response functionWe compute the cumulative price impacts based on the estimates of the VECM model. We estimate themodels daily for each option contract with one-second resolution. The plotsare grand averages across alloption contracts and 20 business days in January 2002. The five optionsexchanges are represented bydashed lines (AMEX), dash-dotted lines (CBOE), solid lines (ISE), and dotted lines (PCX, and PHLX).

27

5 10 15 20 25 30 35 40−0.05

0

0.05

0.1

0.15

0.2

0.25

0.3

ISE Market Share in Number of Trades

ISE

Info

rmat

ion

Sha

re L

eade

rshi

p

0 10 20 30 40 50 60 70−0.05

0

0.05

0.1

0.15

0.2

0.25

0.3

ISE Market Share in Trading Volume

ISE

Info

rmat

ion

Sha

re L

eade

rshi

p

Figure 3Price leadership of the ISE and its market shareThe scatter plots depict the relation between the price leadership of ISE andits market share in number oftrades (left) and trading volume (right). The price leadership is the average difference between the informa-tion share of the ISE and the other four exchanges across dates, bids,asks, and lower and upper bounds.

500 1000 1500 2000 2500 3000 3500 4000−0.05

0

0.05

0.1

0.15

0.2

0.25

0.3

Total Number of Trades

ISE

Info

rmat

ion

Sha

re L

eade

rshi

p

0 0.5 1 1.5 2 2.5

x 105

−0.05

0

0.05

0.1

0.15

0.2

0.25

0.3

Total Trading Volume

ISE

Info

rmat

ion

Sha

re L

eade

rshi

p

Figure 4Price leadership of the ISE and the options trading activityThe scatter plots depict the relation between the price leadership of ISE on acertain options contract and theaggregate trading activity of the options contract. The aggregate trading activity is expressed through thenumber of trades in the left panel and trading volume in the right panel. The price leadership is the averagedifference between the information share of the ISE and the other four exchanges across dates, bids, asks,and lower and upper bounds.

28

9 10 11 12 13 14 15 1610

12

14

16

18

20

22

24

Time of the Day

Mea

n B

id−

Ask

Spr

ead

Figure 5Mean bid-ask spreads for quotes from the five options exchangesThe figure shows the average bid-ask spreads for quotes from the five options exchanges at half-hour in-tervals. The five options exchanges are represented by square-dashed lines (AMEX), diamond-dash-dottedlines (CBOE), circle-solid lines (ISE), star-dotted lines (PCX), and plus-dotted lines (PHLX).

29

9 10 11 12 13 14 15 160.2

0.3

0.4

0.5

0.6

0.7

0.8

Time of the Day

Per

cent

age

on N

BB

O

Bids

9 10 11 12 13 14 15 160.2

0.3

0.4

0.5

0.6

0.7

0.8

Time of the DayP

erce

ntag

e on

NB

BO

Asks

9 10 11 12 13 14 15 160.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0.18

0.2

0.22

Time of the Day

Per

cent

age

on N

BB

O A

lone

Bids

9 10 11 12 13 14 15 160.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0.18

0.2

0.22

Time of the Day

Per

cent

age

on N

BB

O A

lone

Asks

Figure 6Percentages of bids and asks on NBBO and NBBO aloneThe figure shows the percentages of quotes from each options exchange that are on the national best bidoffer (NBBO) and that are on the NBBO alone. The five options exchanges are represented by square-dashed lines (AMEX), diamond-dash-dotted lines (CBOE), circle-solid lines (ISE), star-dotted lines (PCX)and plus-dotted lines (PHLX).

30

Table 1Trade composition of the 50 multiple listed optionsEntries report the sample mean, minimum, maximum, and standard deviation of the totalnumber of trades(panel A) and total trading volume (panel B), as well as market shares for each options exchange by tradesand volume, respectively. We compute the summary statistics based on the 50 multiple listed options con-tracts we use for the empirical analysis.

Exchanges A C I P X

Panel A Trades Market Share, %

MEAN 1431.0 20.8 29.6 16.2 25.2 8.2MIN 833 2.3 9.6 5.4 1.9 2.3

MAX 3608 49.8 49.9 35.8 48.8 30.2STD 706.3 11.2 13.0 5.8 12.9 6.2

Panel B Volume Market Share, %

MEAN 46318.7 27.4 30.1 22.0 11.3 9.1MIN 7837 4.0 8.8 5.1 0.5 0.7

MAX 239100 50.9 61.0 67.9 33.8 38.9STD 45813.4 12.7 13.6 11.8 7.2 9.0

31

Table 2Information shares of option quotesWe based our information share statistics on a vector error correction model of bids (panels A and B) andasks (panels C and D) from the five option exchanges AMEX, CBOE, ISE, PCX, and PHLX. We estimatethe models at each day for each of the 50 option contracts for January 2002. The table reports the summarystatistics of these daily estimates.

Exchanges AMEX CBOE ISE PCX PHLX

A. Max Information Share From Bids

MAX 0.836 0.881 0.942 0.845 0.801MEAN 0.196 0.226 0.266 0.215 0.143MIN 0.000 0.000 0.000 0.000 0.000STD 0.166 0.169 0.178 0.172 0.143

B. Min Information Share From Bids

MAX 0.814 0.880 0.941 0.843 0.780MEAN 0.178 0.203 0.247 0.196 0.132MIN 0.000 0.000 0.000 0.000 0.000STD 0.159 0.161 0.174 0.164 0.138

C. Max Information Share From Asks

MAX 0.810 0.943 0.973 0.984 0.835MEAN 0.189 0.210 0.280 0.215 0.147MIN 0.000 0.000 0.000 0.000 0.000STD 0.162 0.166 0.181 0.169 0.150

D. Min Information Share From Asks

MAX 0.810 0.943 0.966 0.981 0.834MEAN 0.173 0.191 0.261 0.197 0.137MIN 0.000 0.000 0.000 0.000 0.000STD 0.158 0.158 0.177 0.162 0.146

32

Table 3Price leadership of the ISE over other options exchangesEntries report thet-statistics on the difference between the information share of the ISE and theinformationshare of the other four exchanges. We define the statistics as the mean difference over the standard deviationof the difference, scaled by the square-root of the number of sample points. We compute thet-statisticsbased on the lower bound (MIN) and upper bound (MAX) estimates of the information share on bids andasks.

Exchanges AMEX CBOE PCX PHLX

MAX from Bids 7.70 4.30 5.65 14.60

MIN from Bids 7.88 4.93 5.87 14.06

MAX from Asks 9.92 7.44 7.03 14.89

MIN from Asks 9.94 7.80 7.13 14.35

33

Table 4Correlations between the price leadership of the ISE and the trading activitiesEntries report the correlation estimates between the price leadership of the ISE over other four optionsexchanges and trading activity variables. For each option contract, we define the price leadership of the ISEas the average difference between the information share of the ISE and the other four options exchanges,

L =180

20

∑t=1

4

∑i=1

I S t,ISE− I S t,i(τ), i = AMEX,CBOE,PCX,PHLX.

We measure the information share by both the lower and the upper bounds and from both the bids and theasks. The trading-activity variables include the market share of the ISE by both the number of trades andthe trading volume, and the aggregate trading activity of these 50 options by both the number of trades andthe trading volume. We also report thet-statistics of the correlation in parentheses.

Market Share of ISE in Aggregate Trading Activity inCorrelation Number of Trades Trading Volume Number of Trades Trading Volume

MAX from Bids 0.17 0.15 0.30 0.24( 1.17 ) ( 1.06 ) ( 2.15 ) ( 1.74 )

MIN from Bids 0.16 0.15 0.28 0.22( 1.12 ) ( 1.02 ) ( 2.02 ) ( 1.58 )

MAX from Asks 0.17 0.28 0.28 0.19( 1.22 ) ( 2.06 ) ( 2.03 ) ( 1.35 )

MIN from Asks 0.15 0.26 0.28 0.19( 1.03 ) ( 1.90 ) ( 1.99 ) ( 1.32 )

34

Table 5Percentage of trades outside (on, inside) the bid-ask QuotesEntries in panel A report the percentages of trades at each exchangethat are outside, on, and inside thatexchange’s bid-ask quote range. For trades that are inside the quote range, panel B reports the percentage ofthese inside quotes that are outside, on, and inside the NBBO. We averagethe estimates across all trades onthe selected 50 option contracts during January of 2002 at each options exchange. Panel C reports the ratioof the number of buy transactions over the number of sell transactions. Weclassify a transaction as a buywhen the transaction happens exactly on the ask in the first row in panel C and when the transaction happensabove the midquote in the second row in panel C. Accordingly, the first rowin panel C defines the sell as atransaction that happens exactly on the bid and the second row defines thesell as a transaction that happensbelow the midquote.

Exchanges AMEX CBOE ISE PCX PHLX

A. Percentages Out Of Total Trades

Outside BA 2.83 3.47 1.81 1.48 3.55On BA 59.35 43.00 84.48 28.19 51.93Inside BA 37.82 53.53 13.71 70.34 44.52

B. Percentages Out Of Inside Trades

Outside NBBO 5.67 8.85 7.63 6.96 5.54On NBBO 75.98 74.99 68.86 83.96 78.33Inside NBBO 18.35 16.16 23.51 9.09 16.13

C. Buy to Sell Ratio

On Ask/Bid 0.78 0.73 0.95 0.82 0.81Above/Below Mid 0.74 0.79 0.95 0.79 0.69

35

Table 6The mimicking behavior of PCXEntries report theR-square estimates of the following regression:

QPCXt = a+bQj

t−n +ejt,n, j = NBBO,AMEX,CBOE, ISE,PHLX; n = 0,1,2,

whereQ jt denotes the quote (bid in Panel A and ask in Panel B) from exchangej at timet (seconds). We

perform the regression daily on each option contract. Entries report thesample average (standard error inparentheses) of the estimates across the 50 option contracts and over the 20 business days in January, 2002.

( j,n) 0 1 2

A. Bids

NBBO 0.9275 ( 0.2354 ) 0.9198 ( 0.2486 ) 0.9196 ( 0.2486 )AMEX 0.8288 ( 0.2718 ) 0.8287 ( 0.2718 ) 0.8286 ( 0.2718 )CBOE 0.8326 ( 0.2599 ) 0.8326 ( 0.2599 ) 0.8325 ( 0.2599 )ISE 0.8750 ( 0.2433 ) 0.8750 ( 0.2432 ) 0.8749 ( 0.2432 )PHLX 0.8140 ( 0.2672 ) 0.8139 ( 0.2672 ) 0.8138 ( 0.2672 )

B. Asks

NBBO 0.8488 ( 0.2186 ) 0.8486 ( 0.2186 ) 0.8485 ( 0.2186 )AMEX 0.8057 ( 0.2411 ) 0.8057 ( 0.2411 ) 0.8056 ( 0.2411 )CBOE 0.8009 ( 0.2544 ) 0.8008 ( 0.2544 ) 0.8008 ( 0.2544 )ISE 0.8310 ( 0.2262 ) 0.8310 ( 0.2261 ) 0.8309 ( 0.2261 )PHLX 0.7885 ( 0.2918 ) 0.7884 ( 0.2917 ) 0.7883 ( 0.2917 )

36