Market liquidity around earnings announcements

Tilburg University

Market liquidity around earnings announcements

Pronk, M.

Publication date:

2002

Document Version

Publisher's PDF, also known as Version of record

Link to publication in Tilburg University Research Portal

Citation for published version (APA):

Pronk, M. (2002). Market liquidity around earnings announcements. CentER, Center for Economic Research.

General rights

Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain

• You may freely distribute the URL identifying the publication in the public portal Take down policy

..

.t.

_aw;

UNIVERSITEIT * <0 VAN TILBURG

_*

BIBLIOTHEEK TILBURG

Market Liquidity

around

Market Liquidity

around

Earnings

Announcements

Proefschrift

ter verkrijging van

de graad van

doctor aan de

Universiteit van

_Tilburg

_{op gezag van} de rector

magnificus, prof.dr. F.A. van der Duyn Schouten, in

het openbaar te verdedigen ten overstaan van een

door het _college voor _{promoties aangewezen}

commissie in de aula vandeUniversiteitopvrijdag 8

november 2002 om 14.15 uur door

Maarten

Pronk

Promotores: Prof.dr.D.V. _DeJong

'Unless the LORD builds the house, they labor in vain who build it, unless the LORD

guards the city, the watchman stays awake in vain. ' (Psalm 127:1)

PREFACE

One

ofthe

things I learned as a Ph.D. student is that research isaninteractiveprocess.

Feedback by

fellow

academics is necessary to makeprogress. I benefited from many

advices, comments and suggestions. Therefore, I would liketo express my gratitude

to thosewhocontributed tomy dissertation.

First of all, I would like

to thank my supervisors Doug DeJong and Gerard

Mertens. Doug explainedtheimportance ofPh.D. courses, stimulated metovisit the

University of

_Iowa,stressedtheimportance ofaninternational academicnetwork and

commented

quickly

andthoroughly on all versions of mypapers.Inother words, he is

largely responsible for myacademic education over the last

five

years. On top of that,

I am grateful for allpleasant non-academic meetings andtalks wehad. _{Gerard's major}

contribution lies in the

field

of

motivation. Heasked me to become a Ph.D. student,

createdapleasantworkingatmosphereatTilburg Universityandstimulated me all the

way long. I

also enjoyed our participation in the research

project of

the Limperg Institute intothequality

of

financial reporting intheNetherlands.

Part ofthe research was undertaken at the_{University of Iowa. I}am_{grateful to}

the accounting

faculty and

Ph.D. students

there for

the great time and enriching

experience. Our neighbours and friends in Iowa

City

changed this challenging time

into an _exciting

_{period. I}

_{gratefully acknowledge} financial

_{support from The}

Netherlands_Organization

for

ScientificResearchforthesevisits

_(NWO

R46-404 and

R46-435).

My colleagues at Tilburg University contributed to my dissertation too, by theiradvice and bycreatingastimulatingresearchenvironment. I amalsoindebted to

Peter Easton, Terry Warfield and MortPincus

for

their comments on myresearch as

well as the participants ofthe seminars atErasmus University Rotterdam, Lancaster

University, London Business School, Nijenrode University, University

of

Antwerp, University of Iowa,and participants ofthe PWC2000Doctoral Colloquium.Finally, I

would like

to thank the members of my committee, in addition to those already

mentioned:

Willem

Buijink,

Theo Nijmanand

Arthur

van Soest, forthe approval of

this thesis.

This dissertation closes a

period of

five years as a Ph.D. student. However,

downs. _{A clear up was mymarriage to}

Karin in July

1999. _{Every day 1 enjoy her}

decisionto share her life with me. A big loss was the sudden death of my father in

November 1999, as he was an important person in my life. He was companionable

and he wasalways interested in my activities. I have greatmemories of the time we

shared together. Asanexpression of mygratitude I dedicatethis thesis to him.

CONTENTS

1 Introduction 1

1.1 Introduction 1

1.2 Market

liquidity

around earnings

announcements 1

1.3 The_impact

of

_intraday

_timing

of

_{earnings announcements on market}

liquidity 5

1.4 Theimpact

ofthe

information environment on the drop inmarket

liquidity

before earningsannouncements 6

1.5 _Positioning

ofthe

_essays

within

financeandfinancialaccounting research 7

1.6 Outline

ofthe

thesis 8

2 The impact

of intraday timing of

earnings announcementson

market

liquidity 9

2.1 Introduction 9

2.2 Literatureoverview_{and hypothesis development 11}

2.2.1 Bid-ask_{spread anddepth 11}

2.2.2 Increasedprobability

of

informed tradingaround earnings announcements 14

2.2.3 Impact

of

intraday

timing

of

earnings announcements on increased_probability

of

informedtrading 15

2.3 Research

design 16

2.3.1 _{Earnings announcement}

times 16

2.3.2 Calculation

of

market

_liquidity

variables 17

2.4 Results 19

2.4.1

After

earnings announcements 19

2.4.2 Before_earningsannouncements 21

2.4.3 _Sensitivitytests 24

3 The impact of the

information environment on the drop

in

market

liquidity

before earnings

announcements 31

3.1 Introduction 31

3.2 Literature overviewand_{hypothesisdevelopment 32}

3.2.1 Bid-ask_{spread, adverse selectioncomponent and}

_{depth 32}

3.2.2 Increased_probability

of

_{informed tradingbefore earnings} announcements 35

3.2.3 Theimpact

ofthe

information environment on the drop in market

_{liquidity 35}

3.3 Research

_{design 37}

3.3.1 _{Earnings announcement}times 38

3.3.2 Variable

definitions 38

3.4 Results 45

3.4.1 Descriptive

statistics 45

3.4.2 _Regression

results 47

3.5 Discussion 52

3.6 Summaryand

conclusions 53

Appendix 3.A Approach by LSBto detect the adverse selection

component 54

4 Summary

and

conclusions 59

4.1 Introduction 59

4.2 Theimpact

of

intraday

timing

of

earnings announcements onmarket

liquidity 59

4.3 The_impact

ofthe

_{information environment on the drop in}market

liquidity

before earningsannouncements 61

References 63

CHAPTER 1

INTRODUCTION

1.1 Introduction

Market

liquidity

drops around earnings announcements. Lee,

Mucklow

and _Ready

(1993) (hereafter

LMR)

present evidencethatbid-ask spreads are higherand depths

are lower during trading hours around earnings announcements than during

non-announcement periods. Investorsdislike this drop, becausemarket

liquidity is one of

the most important characteristics that

investors look for in

an organized financial

market.Market

liquidity

relates to the

ability to buy or

sell significant quantities of a

security quickly, anonymously, and with relatively

little

price impact.

Low

market

liquidity has a negative impact on the demand

for

shares and the share price. This

reducedsharepricerelates toahigher cost

of

equitycapital. _{Therefore, it}isimportant

to explore the dropinmarket

liquidity

around earnings announcements.

This thesis investigates two important aspects of this drop in market

liquidity,

namely (a)the

ability

of

management to mitigate the drop in market

liquidity

around

earnings announcements by using their discretion to announce the earnings news

during non-trading instead

of

trading hours and (b) the conjecture that the drop in

market

liquidity

before earnings announcements relates to the

richness of the

information environment. Before discussingthesetopics inmoredetail insection 1.3

and 1.4, I explain the notion

of

market liquidity and the behavior ofthe (adverse

selection

_{component of}

_the) bid-ask _spread and _depth around _earnings

announcements. Section 1.5 _positions the thesis

within

finance and financial

accountingresearchandsection 1.6_providestheoutline

ofthis

thesis.

1.2 Market liquidity around

earnings announcements

Tomaintain

_liquidity,

_{manyorganizedexchanges}usemarket makers,individuals who

stand

ready to buy or

sell whenever

investors wish to sell or buy.

In

return for

providing

liquidity,

market makers are _{granted monopolyrights by} the _{exchange to} postdifferent prices forpurchases and sales: they buy at the bid price and sell at a higher ask price. As figure 1.1 shows, the quoted bid-ask spread is the difference

between the quoted bid and ask

price. It

is possible that traders negotiate with the

a price that ishigher than the quoted bidprice and sell ataprice thatislower than the

quotedaskprice. Theeffective bid-askspread takesthis possibility intoaccount and is

calculated as two times the absolute

value of

the difference between the quote

midpoint (average of the bid andaskprice) and the bid orask price forapurchase or

sale, respectively.

Figure

1.1

Bid-ask

spreads

Price

QA _

lA 7 Effective spread 2*2 for a sale

Quoted I QM 3

spread _{'i-*1 Efective spread}

IB -) for a purchase

QB

Time

This figure demonstrates the quotedandeffectivebid-ask spread.QArepresents the quoted ask price,

IAtheinsideask_{price, QMthe quotemidpoint, IB}theinsidebid_{price and QB the quoted bid price.}

Extant market microstructure literature shows that thebid-ask spreadconsists

of

three components: order processing costs, inventory holding costs, and adverse

selection costs. The order processing cost component represents a fee charged by

market makers

for

standing ready to match buy and sell orders

(Tinic 1972). The

inventory holding cost component compensates dealers

for

managing the inventory (Stoll 1978, Hoand Stoll 1981).Finally,the adverse selection component represents a

reward to market makers

for taking on the risk

of

dealing

with

traders who may

possesssuperior information(Copeland andGalai 1983, Glosten and Milgrom 1985). When, for example, an investor knows before an earnings announcement that the

earnings are far lowerthan expected, (s)he willsell shares to themarketmaker.

After

the information arrival both the bid and ask price drop. As figure 1.2 indicates, the

a low

ask _{price (QA i).} Therefore, the market maker

earns only part of

the bid-ask

spread. Similarly, whenan investor_{knows before the announcement that the earnings} are higherthanexpected (s)he will buyshares.

After

the information arrival the bid

and ask price go up and the market maker misses the return on the shares because

(s)he already sold

the shares. In other words, when investors possess private

information, market makers are

missing part of

the

spread or part of

the return

because they have theobligationto trade.To compensate themselves

for

thesemissed

revenues, market makers increase thebid-askspread. Asaresult,marketmakers earn

more when they trade

with

traders without private information. These additional

revenues compensate for the missed spreads and missed returns

while

trading with

informed investors. The higherthe probability

of

informed trading is, the higher the

adverseselection component and, thus, the bid-askspread.

Figure

1.2_{Adverse selection component}

Price QAO

QAl

QMo

-QM' QBO _{J US} QB' Time

This figure illustrates the unrealized spread due to informedtrading. QArepresents the quoted ask

price, QMthe quotemidpoint, QBthe quotedbidprice and US the unrealized spread due toinformed

trading.

LMR state thatthebid-ask spread is onlyonedimension

of

market liquidity.

Theotherdimension isthe depth: the number

of

shares market makersare

willing to

trade at the quoted bid and ask prices, respectively.Market makers may declare the

maker quotes a large depth, an informed investorcantrade many shares at one time

againstthequoted bid oraskprice.However, whenthemarket maker declaresasmall

depth, theinformed investor canonlytradesmallportions at a time.

After

eachtrade,

themarket maker has theopportunityto change the bid and ask prices. Therefore, the

market

maker may lose

less money before discovering the

implications of new

information forthefundamental shareprice. Thus, whentheprobability

of

informed trading increases, the spread and adverse selection component are expected to

increase andthe depthis_{expected to}decrease.

Days around earnings announcements constitute a _period in which market

makers may experience an increasedprobability

of

informed trading.

LMR

mention

three reasons why the information asymmetry between market makers and informed

investors may

increase just bgore

an _{earnings release,}

_{namely: (a)}

a _higher

probability

of

leakage

of

value relevant information when earningsare known to the company, (b) the possibility that the

officially

filed information reaches investors

earlierthan _{market makers, and (c)}the expectation

of

imminent earnings news may stimulate some traders to search for information immediately

prior to the

announcement. Kim and _{Verrecchia (1991, 1994)} and McNichols and Trueman

(1994) provide analytical evidence on this

issue. LMR show

that market makers

indeed increase_{spreads and decrease}depths beforedaytimeearnings announcements.

Additionally,

Krinsky and Lee

(1996) (hereafter

KL)

present evidence that the

adverse selection component ofthe spread is significantly higher before daytime

earnings announcements thanduringthenon-announcement period. This suggests that

market makersare_facinganincreased_probability

of

informed trading.

Market makers may also face an increased probability

of

informed trading

after earnings announcements. Kim and Verrecchia (1994) and Livne (2000) model

this situation. Their

basic idea is that

some investors gather private information

through their superiorcapacities to interpret earnings news. Another argument is that earnings news is revealed in pieces and informed trading takes place when this information reaches investors earlierthan market makers. _{For example,} the_earnings

number may hit the news wire first,

followed by

theincome statement and balance

sheet, while detailed information is revealed later

during the day. LMR and KL

present empirical evidence on this

issue and show that

the spread and adverse

selection component are higher during trading hours after daytime earnings

1.3 The

impact

of

intraday

timing

of

earnings announcements on

market

liquidity

As discussed in

_{the previous sections,} empirical literature indicates that market

liquidity

drops around earnings announcements. Investors dislike this drop, because

market liquidity is one of themostimportantcharacteristics thatinvestors look for in

an _organized financial market. Even _though the period around earnings

announcements is a short interval, the

_{impact of}

the market

_{liquidity drop may be}

severe, because this period is

known for its

high _{trading volume (Morse (1981),}

Bamber (1986)). Therefore, itis_{important formanagement}

of

listedfirms tobeaware

of actions

that might reduce the decline in market

liquidity

around earnings

announcements.

The

first

essayinvestigates intradaytiming

of

earnings releases as a way of mitigating the dropinmarket

liquidity

aroundtheseannouncements.Thereareseveral

reasons to expect that intraday

timing

of

earnings announcements influences the

probability

of

informed trading and, therefore, the change in bid-ask spreads and

depths around these announcements. The probability

of

informed trading increases

before earnings announcements during trading hours ('daytime' announcements)

because investors are more motivated tosearch _{for private information}and because

thepublicannouncement mayreachinvestorsearlierthanmarket makers.Thissecond

threat, however, does not exist

for

announcements during non-trading hours

('overnight'

_{announcements). In} addition, Livne (2001) shows

analytically that

investors

with

private information trade less aggressively before overnight than

daytimeannouncements.This_{suggests that the increase in theprobability}

of

informed tradingissmallerbeforeovernightreleases.

The intraday

timing may

also influence the probability

of

informed trading after earnings announcements. When the earnings are released overnight, investor do

not needto reactimmediately on the news andcan research the databefore making a

trading decision. In addition, when the information is revealed in pieces before

opening of

_{the market,} investors can also use this additional information in their decision

_{process. On the NYSE and AMEX}

investors can submit orders before

opening of

the market. When the submitted orders reflect the earnings news, the

openingprice

_will

capture the releasedinformation and theprobabilityissmaller that

market makers

willloose

money to investors

with

private information after opening of

makerscanstill setthe_{openingprice conditional on}therelease whentheycheck news wiresbefore theopening ofthe market.Basedon thesearguments, I hypothesize that

the percentage

deviation of

the spread from the median spread

during the

non-announcement period is _{larger andthe percentage}

deviation of

_{the depth} is smaller

before andafterdaytime announcements relative to overnight announcements.

The sample consists of 1000 daytime and 1802 overnight earnings

announcementsdeclared between 1993 and 1996 by 336firms traded at the NYSE or

AMEX.

The results show that percentagedeviations

of

quoted and effective spreads

are significantly larger and percentage deviations of depths are smaller after 'daytime'

announcements compared to 'overnight' announcements. This suggests that intraday

timing

is

possibly a way

to

mitigate the drop

in market

liquidity

after earnings announcements.

1.4 The

_{impact of}

the

information

environment on the drop

in market

liquidity

before earnings announcements

The second essay investigates the drop in market

liquidity

before earnings

announcements. So far,no evidence existswhethermarketmakers adjust spreads and

depths forall firms to thesameextentorconditionthemarket

liquidity drop on

firm-specific factors. I investigatetheconjecture that this drop inmarket

liquidity

relates to

the richness ofthe information environment. A rich information environment means

thatmoreinformation is

publicly

availableand captured in theshare_{price. Therefore,}

a smaller price reaction is expected at the earnings announcement and this reduces

investor's incentives to search

for

private information. In

_{addition, even when}

investors_{search for and find private information, the loss to}the marketmaker_might

be smaller, becausethe expected pricereaction to new information is smaller. I use

three

proxies for

the information environment, namely disclosure

quality, the

occurrence

of

management earnings forecasts and thenumber

of

analysts forecasting

the earnings news.

The dataset contains 2802 earnings announcements made between 1993 and

1996 by 336firms thataretraded at the NYSE or AMEX.The analysis provides some

evidence on the

impact of

the information environment on the market

liquidity drop

before earnings announcements by showing that the occurrence

of

management

earnings forecasts affects the change in depths.

However, I do not find

a robust

adverse selection component

_of)

spreads. One explanation for this lack

of

results is

that market makers do not believe that the information environment affects the probability

of

informed trading_{before earnings announcements and,}therefore, do not conditionon thesevariables.Another explanation is that the drop in market

liquidity

istoo small toshow cross-sectional differences.

1.5 Positioning oftheessays

within

finance and

financial

accounting research

The essays can be positioned in both finance and financial accounting research.

Within

finance, myresearchrelates to marketmicrostructure literature. This area of

research investigates trading systems and price formation processes on capital

markets.

My

thesis contributes to this

literature in two

ways. First, I investigate the

impact

of

information releases during trading and non-trading hours and show that announcing news _during a _{non-trading period improves} market _{liquidity. Second, I}

explore theinformation setthatmarket makers use toinfertheprobability

of

informed tradingbefore earnings announcements.

Alternatively, my

thesis can

be

viewed from

a _{financial accounting}

perspective, because

it

contributes to disclosure-related capital market research. This

research areausescapital markets toprovideevidence on the relevance

of

supplying

information toinvestors.Overall, two sets

of

capitalmarketvariablesareinvestigated,

namely (a) share_pricesandreturns, and (b) bid-askspreads and depths. Anexample

of

the

first

category isthe study byHealy, Huttonand Palepu (1999). They find that firms that expand disclosure experience significant contemporaneous increases in

stock prices thatareunrelatedto currentearningsperformance. Otherpapers focus on

bid-ask spreads and depths because, as discussed in section 1.2, these variables are

related to the information asymmetry on capital markets. While the

reduction of

information asymmetry is an aim

of

disclosure, this research design provides direct evidence on the impact

of

information supply on uncertainty on capital markets.

Welker (1995) documents,

for

example, a _{significant negative} relation between

analysts' ratings

of

firms'

disclosures and bid-ask spreads, while

Coller and Yohn

(1997) provide evidence on theimpact

of

management earnings forecasts on spreads

and depths.

My

thesiscontributes to this areabyinvestigating whetherdisclosure also

affects the changeinbid-ask spreadsand depths before earnings announcements and

by documenting that the

timing

of

announcements impacts uncertainty on capital

1.6 Outline of

thethesis

Chapter 2 presents the

investigation of

the impact

of

intraday

timing

of

earnings

announcements onmarket

liquidity,

while chapter3 contains the essay on theimpact

of

the information

environment on the drop

in market

liquidity

before earnings

announcements. Chapter 4 summarizes the thesis and proposes some themes for

futureresearch. Note that theessays in chapter 2 and 3 areself-contained. Therefore,

CHAPTER 2

THE IMPACTOFINTRADAYTIMINGOFEARNINGSANNOUNCEMENTS ON

MARKET LIQUIDITY

2.1 Introduction

Market

_liquidity

_{drops around earnings announcements. Lee,} Mucklow and _Ready

(1993) (hereafter

LMR)

present evidence thatbid-ask spreads are higherand depths

are lower during trading hours around earnings announcements than during

non-announcement periods. Investorsdislike thisdrop, becausemarket

liquidity is one of

the most important characteristics that

investors look for in

an organized financial

market. Market

liquidity

relates to the

ability to buy or

sell significant quantities of a

security quickly, anonymously, and

with

relatively

little

price impact. Low market

liquidity

impacts the demand

for

shares andthe shareprice negatively. Thisreduced

share price relates to ahigher cost

of

equitycapital. Even though the periodaround

earnings announcements is a short interval, the

impact of

the market

liquidity drop

maybe severe, becausethis_{period is}known for itshightrading volume (Morse 1981,

Bamber 1986). It is

of

great importance

for

management

of

listedfirmsto understand

therelevance

ofthe

marketliquidity drop and tobeaware

of

actions thatmightreduce

this decline.

This essay investigates the intraday

timing

of

earnings announcements as a

way

of

mitigating the drop

in market

liquidity

around earnings announcements.

Several analytical papers show that market

liquidity

drops before andafter earnings

announcements because market makers are concerned about an

increase in the

probability

of

informed

trading (Kim

and Verrecchia 1991, 1994, Livne 2000 and

McNichols andTrueman 1994). There are severalreasons to expect that the intraday

timing

of

earnings announcements influencesthisprobability

of

informedtrading and,

therefore, the change in the bid-ask spread and depth. The probability

of

informed

trading increases before earnings announcements during trading hours ('daytime'

announcements) because investors are more motivated to search

for

private

information and because the public announcement may reach investors earlier than marketmakers.Thissecond threat, however, doesnotexistforannouncementsduring non-trading hours

('overnight'

announcements). In addition, Livne (2001) shows

overnight than daytime announcements. This suggests that the

increase in the

probability

of

informed tradingissmallerbefore_overnightreleases.

The intraday

timing may

also influence the probability

of

informed trading

afterearnings announcements. When the earnings are released overnight, investors do

not need to reactimmediately on the news andcan research the databeforemaking a

trading decision. In addition, when the information is revealed in pieces before

opening of

_{the market,} investors can also use this additional information in their

decision _{process. On the New}

York

Stock _{Exchange (NYSE)} and American Stock

Exchange

(AMEX)

investors can submit orders before_{opening of}the market. When

the submitted orders reflect the earnings news, the opening price

will

capture the

released information and the probability is smaller that market makers

will

loose

moneyto investorswith privateinformation after opening ofthe market. Even when

the order flow does

not reflect _{the earnings news, market makers} can

still set the

opening price conditional on the release when they check news wires before the

opening of

the market. Based on these arguments I hypothesize thatthe percentage

deviation of

the spread from the median spreadduringthe non-announcementperiod

is larger and the percentage deviation of the depth issmaller_{before andafter daytime}

announcementsrelativeto_overnightannouncements.

The sample consists of 1000 daytime and 1802 overnight earnings

announcements declaredbetween 1993 and 1996 by 336firms traded at the NYSE or

AMEX. Relative

to overnight releases,the results showthat daytimereleases relate,

on average, to significantly larger percentage deviations ofthe quoted and effective

spread fromthe median during the non-announcement period and lower percentage

deviations of the depth after earnings announcements. This suggests that intraday

timing is possibly a way to mitigate the drop in market liquidity

after earnings

announcements.Sensitivity testsrevealthatthese findings arerobustto firm-specific

factorsand cross-listings, andthatthere is nodifference intheinformation content of daytimeandovernightannouncements in the sample. In fact, I find noevidence of a drop in market liquidity after overnight earnings announcements at all. LMR state that

a

valid

_{conjecture about}adecrease_{in market liquidity can only be madewhen spreads}

increase and _{depths do not rise orwhen depths} decrease and_{spreads do} notdecline,

simultaneously. However, after overnightearnings announcements the quoted spread

and depth are both significantly larger than during the non-announcement period.

liquidity,

whiletheother dimension_{(depth) shows}anincreasein market

_{liquidity. In}

addition,the effectivespread is_{significantly smaller after overnight} announcements

than during

the non-announcement period. Finally, the analysis provides some

evidence that intraday timing affects effective

spreads _{before earnings}

announcements. However, sensitivity tests reveal that this result may be

driven by

firm-specificfactors.

Theessayshows that thecall auction attheopening ofthe market and/or the opportunity forthemarket maker to interpretthe earnings news beforetradingstarts,

reducethe_probability

of

_{informed trading}and_improvemarket

_liquidity

after_earnings

announcements. In this waytheessay shows a benefit

of

announcing earnings news

outside trading hours. However,

firms that

are cross-listed on a non-US exchange

should be aware that the drop in market

liquidity

mayoccur on the other exchange

when that

exchange is open at

the announcement. Additionally, the

decision to

announce earningsduring trading or non-tradinghours depends on more factors than

just

the impact on market

liquidity.

The remainder ofthischapterisorganizedasfollows.In section 2.2, I discuss

the relevantliterature and developthe hypotheses. Section2.3 discussesthe research

design,whilesection 2.4 presents theempiricalresults. Section2.5concludes.

2.2

Literature

_{overview and hypothesis development}

2.2.1 Bid-ask spread and depth

To maintain

liquidity,

many organized exchangesusemarket makers,individuals who

stand

ready to buy or

sell whenever

investors wish to sell or buy.

In

return for

providing

liquidity,

market makers are granted monopoly rights bythe exchange to

post different prices for purchases and sales: They buy at the bid price and sell at a higher ask price. As figure 2.1 _{shows, the quoted} bid-ask spread is the difference

between the quoted bid andask

price. It

is possible that traders negotiate with the

marketmakerabout the bid oraskprice._Therefore,market makerssometimes buy at a price thatishigher thanthe quoted bid price or sell ataprice that islower than the

quoted ask price. Theeffective bid-askspread takesthis possibility intoaccount and is

calculated as two times the absolute

value of

the difference between the quote

midpoint (average of the bid andaskprice) and the bid oraskprice forapurchase or

Figure

2.1

Bid-ask

spreads Price

QA

IA 7 Effective spread F*2 for a sale Quoted K QM -H spread -*2 E(pctive spread IB -3 _{for a purchase} QB Time

This figuredemonstrates the quotedandeffectivebid-ask spread.QArepresents the quotedaskprice,

IAthe inside askprice, QMthe quotemidpoint, IBthe inside bid price and QB the quotedbid_price.

Extant market microstructure literature shows that the quotedbid-ask spread

consists

of

three components: order _processing costs, inventory holding costs, and

adverseselectioncosts.Theorder_{processing costcomponent represents a}fee_charged

by market makers forstanding ready to match buy and sell orders (Tinic 1972). The

inventory holding cost component compensates dealers for managing the inventory (Stoll 1978, HoandStoll 1981). Finally,the adverse selection component represents a

reward to market makers for

taking on the risk

of

dealing

with

traders who may

possesssuperior information(Copeland andGalai 1983, Glostenand Milgrom 1985). When, for example, an investor knows before an earnings announcement that the

earnings are far lowerthanexpected, (s)he will sell shares to the market maker.

After

the information arrival the bid and askprice drop. Asfigure 2.2 indicates, themarket maker bought the shares at a highbid price (QBo) and has to sell the shares at a low

ask price

_(QAi).

_{Therefore, the market maker} earns only part ofthebid-ask spread.

Similarly, when an investor knows before the announcement that the earnings are

higherthan expected (s)he will buyshares.

After

the information arrival the bid and ask price go up and the market maker missesthe return on the sharesbecause (s)he

the obligation to trade. Tocompensate themselves forthese missed revenues, market

makers increase the bid-ask spread. Asa_{result, market}makers earn more when they

trade

with

traders_{without private information.}These additional_{revenues compensate}

them forthemissed spreadsandmissedreturnswhile trading

with

informedinvestors.

The higher the probability

of

informed trading is, the higher the adverse selection

component and, thus, the bid-askspread.

Figure

2.2_{Adverse selection component}

Price

QAo

-QA1 _

QMo

-QMt QBO

QB, Jus

Time

This figure illustrates the unrealized spread due to informedtrading. QA represents thequoted ask

price, QMthe quotemidpoint, QBthe_{quoted bidprice and US the unrealized spread due to}informed

trading.

LMR state that thebid-ask spread is onlyonedimension

of

market

liquidity.

The otherdimension isthe depth: the number

of

sharesmarket makers are

willing to

trade at the quoted bid andask prices, respectively. Market makers may declare the

depthstrategically toprotect againstinformed trading. Whenamarketmaker quotes a

largedepth,aninformed investorcan_{trade many shares at one}time_{against the quoted}

bid or

ask price. _{However, when} the market maker declares a small depth, the informedinvestor canonlytradesmallportions at atime.

After

each_{trade, the market}

maker has the opportunity to change the bid and ask _{prices. Therefore, the}market

for

the fundamental share price. Thus, when the probability

of

informed trading

increases, the spreadisexpected to increase and the depthisexpected todecrease.

2.2.2 Increased probability of informed trading around earnings announcements

Days around earnings announcements form a _{period in which market makers may}

experiencean increasedprobability

of

informed trading.

LMR

mentionthree reasons

whythe informationasymmetry between market makers and informedinvestors may

increase

_just

_beforean earnings release, namely: (a) a higher probability of leakage of

value relevant information when earnings are

known to

the

company, (b) the

possibility that the

officially

filed information reaches investors earlier than market

makers, and (c) the expectation

of

imminent earnings news may stimulate some

traders to search

for

information immediately

prior to

the announcement. Kim and

Verrecchia (1991, 1994) and McNichols and Trueman _{(1994) provide analytical}

evidence on this issue. LMR show thatmarket makers indeed increase spreads and

decrease _{depths before daytime earnings announcements.}

Additionally,

Krinsky and

Lee (1996) (hereafter

KL)

present evidence thatthe adverse selectioncomponent of

the spreadissignificantly higherbeforedaytimeearnings announcements thanduring

the non-announcement period. This suggests that market makers are

facing an

increased_probability

of

informed trading.

Market makers may also face an increased probability

of

informed trading

after earnings announcements. Kim and Verrecchia (1994) and Livne (2000) model

this situation. Their

basic idea is that

some investors gather private information

through their superiorcapacitiestointerpret earnings news. Another argument is that earnings news is revealed in pieces and informed trading takes place when this information reaches investors earlierthan market makers. For_{example, the earnings} number may hit the news wire first,

followed by

the income statement and balance

sheet, while detailed information is revealed later

_{during the day. LMR and KL}

present empirical evidence on this

issue and show that

the spread and adverse

selection component are _{higher during trading} hours _{after daytime earnings}

2.2.3 Impact of the intraday timing of earnings announcements on the increased

probability of informed trading

The probability of informed tradingincreases_{before daytime earnings announcements}

because investors are moremotivated tosearch for private information and because

the public announcement may reach investors earlierthan market makers. This last argument does not hold, however, for overnight announcements. In addition, Livne (2001) models the intraday timing

of

earnings announcements. He shows that

investors

with

private information trade less agressively before overnight than

daytimeannouncements.Hisargument is thatintraday

timing

determineswhether the

(post-announcement) unwinding

of

short-term investors' prior positions takes place

before, or at the same time as, long-term investors trade on theirsupposedsuperior

assessment of

firm

value. Unwinding at the same time is lessadvantageous to short

term traders, because market makers

will

interpret their orders as informed trading

and adjust thesharepricequickly. This diminishes short term traders'_{profits. Overall,}

thesearguments suggest that the probability

of

informed tradingincreases lessbefore

overnight releases. Therefore, I propose the following hypothesis: The_percentage

deviation of the spread from the median spread during the non-announcement period

is larger and the percentage deviation of the depth from the median depth during the

non-announcement period is smaller before daytime earnings announcements than before overnight announcements.

The intraday

timing

of

earningsreleases mayalsoinfluencetheprobability of

informed trading after earnings announcements. When earnings news is announced during tradinghours investors mayfeel compelled toactquickly onthe information

theyhaveinstead o

f

researchingandanalyzing the data. When the newsisrevealed in

pieces, investors

will

continue to react on new information and market makers

experience a continuous threat

of

trading

with

investor with private information.

However, when the earnings are released overnight, investors do not need to react

immediately on the news andcan research the data before_makinga_tradingdecision.

In addition, whenthe informationisrevealed in pieces beforeopening of_{the market,} investors can also use this additional information in theirdecision process. On the NYSE and

AMEX

investors can _{submit buy and} sell orders before _{opening of the}

market. At theopen, market makers examine thissupply anddemand

for

shares and

may offset any or all of

the imbalance by trading on their own account. In other

the opening priceconditional on this information. Whenthe submitted ordersreflect the earnings news, the opening price

will

capture the released information and the probabilityis smaller that market makers

will

loose moneyto investors

with

private informationafter opening

ofthe

market. Even when theorder flow doesnotreflect the

earnings news, market makerscanstill setthe opening priceconditional onthe release

whenthey check newswiresbeforetheopening ofthemarket. Two empiricalpapers

investigate whether earnings news is captured in the opening price. Unfortunately,

these papers present conflicting evidence. Francis, Pagach and Stephan (1992)

(hereafter FPS) find

no

evidence that

the opening price reflects overnight

announcementinformation. However,Greene andWatts (1996) show thatthe opening

trade on the NYSE impounds most ofthe price response. _Therefore,

it

remains an empiricalquestionwhetherthe intraday

timing

of

earnings announcements affects the

probability

of

informed trading andthebid-ask spread and depthafterthesereleases.

The second hypothesis runs: The percentage deviation of the spread from the median

spread during the non-announcement period is larger and the percentage deviation of the depthfrom the median depth during the non-announcement period is smaller after

daytime earnings announcements than after overnight announcements.

2.3 _{Research design}

2.3.1 Earnings announcement times

The dataset consists

of

firms with

adisclosure_{quality rating in}

'An

Annual Review

of

Corporate ReportingPractices' prepared by the CorporateInformation Committee

of

the Association

for

_{Investment Management} and _{Research (AIMR)}

for

1993/94,

1994/95 and 1995/96.

While not

a direct requirement for this study, this selection

criterion should have no impact onthe

findings.' In fact, it may be

more

difficult to

detect the items

of

interest withsuchfirms.

AIMR

ratedfirms tend tobe larger and to

have higher stock prices_{(Welker 1995). In} the _{literature, firm size}isoften used as a

proxy for information

availability in

themarket. Whenmoreinformation isavailable

about firms before earnings announcements, the price reaction and

probability of

informed tradingatannouncementsareexpected tobesmaller.Therefore, the drop in

market

liquidity may

be smaller and

it might be

more

difficult

to detect any

1

differences between the market

liquidity

reaction to overnight and daytime

announcements.

TheDow Jones Interactivedatabase is usedtodetermine the date and time of

the earningsannouncements.Forfirms with an

AIMR

rating inthe 1993/94volume, I

look for quarterlyearnings announcements between July 1,1993 and June 30,1994.

The othertwo years' reports arematched

with

earnings announcements in asimilar

way. I take the

time

stamp of

the

first publication of

the quarterly earnings

announcement byBusiness Wire,

Dow

Jones_{News Service,}DowJonesInternational

News orPRNewswire. Announcements on _tradingdaysbetween 9.30 AM and 4.00

PM EST areconsidered daytime announcementswhile all other announcements are

regardedas_overnightannouncements.

2.3.2 Calculation of market liquidity variables

The Trades and Quotes

(TAQ)

database is usedtocalculate thevariables

of

interest,

namely the percentage

deviations of

the quoted spread, effective spread, depth and volumebeforeandafterearnings announcements from the related mediansduring the

non-announcement period.

Following LMR, I exclude

all

_thinly

traded _{stocks (on}

average less than ten trades a day), and stocks with an average price

below $5 or

above $100 duringthe month

of

January that isclosest to the announcement. I focus

onfirms traded at the NYSE or

AMEX,

becausethe_{NASDAQ has}adifferent trading

system.

All

quotes coded

differently

from opening or normal trading quotes are

deleted. Quotes that aresetbefore 9.30 AM

or

after 4.00 PMand tradesthatoccurred

before a bid andask_{price were quoted, are also removed.}

In accordance to LMR and KL, eachtrading day isdivided into thirteen half-hour trading intervals. Foreach half-hour interval, the time weighted quoted bid-ask

spreadanddepth, the averageeffective bid-askspread, and thevolumeare calculated

for each

firm,

separately. In thesecomputations,theeffectivespread isdefined as two

times the absolute

value of

the difference between the trade price and the quote

midpoint (average of the bid and ask price).2 The depth equalsthenumber

of

shares

the market maker is

willing to buy plus

the shares (s)he is

willing to sell,

simultaneously.

Next, the periods before and after earnings announcements are defined. For

firms with

a _daytime announcement, the half-hour trading interval containing the

earnings announcementisdetermined according to the time stamp of the news wire.

The period before the earnings announcement consists

of6.5

tradinghoursbefore this

half-hour interval. The

period _{after the earnings release consists of the half-hour}

trading interval containingthe announcement plus thesix tradinghoursthereafter. For

firms with

an_{overnightannouncement, the period before the earnings announcement}

consists of6.5trading hoursbefore the

closing of

the market

_(full

tradingday),while

the period afterthe announcement consists of 6.5 trading hours after the opening of

the market.Forearnings announcements between July 1, 1993 and June 30, 1994, the

non-announcement periodconsists ofalltradingdays betweenthese datesexcept two

tradingdays_{before, the day of and two trading}daysafterearnings announcements as

well as days with management earnings forecasts or dividend announcements: For

announcementsbetween July 1994 and June 1996, comparable periods apply.4

The percentage

deviation of

the quoted spread before an earnings

announcement from the medianduringthe non-announcement periodis calculated as

follows.First, I average thetime weightedquotedspreadsduringtheperiodbefore the

earnings announcement. Next, I divide the non-announcement _{period into similar}

periods of 13 half-hour trading intervals. Thus, whenthe period before the earnings

announcementranges from 11.00 AM theday before the announcement to 11.00 AM

the day

ofthe

announcement, the non-announcementperiodisdividedinperiods from

11.00 AM day 1 to 11.00 AM day 2, 11.00 AM day 2 to 11.00 AM day3etcetera. For

each period, I average the time weighted quoted spreads during the 13 half-hour trading intervals during that period. This approach results foreach announcement in

one average quoted spread before the announcement and approximately 220 average

quoted spreads during the non-announcement

period. I determine for each

announcement separately the

median of

the approximately 220 average quoted

spreadsduringthenon-announcement period.

Finally, I divide

thedifferencebetween

3 TheDowJonesInteractivedatabase is usedto investigate the Business Wire,DowJones News Service,DowJonesInternational News and PRNewswirefor_{management earnings forecasts and}

dividendannouncements.

4 The accuracyofthetimestampisimportant. Therelativeprecisionofthesetimestampsisdifficult to

gauge. LMR,whofollowasimilarapproach, show that nosignificantincrease in trading volume occurs until thehalf-hourintervalcontainingdaytime announcements This findingstronglysuggests

the quoted spreadbefore the announcement and the median quoted spread during the

non-announcement period by the median quotedspread andmultiply it with 100. In

these calculations the median is used

instead of the mean of

the average quoted

spreads during the periods of 13 half-hour trading intervals during the

non-announcementperiodbecausethedistribution o

f

these_quotedspreads isskewed. The

percentage

deviations of

the effective spread, depth and volume before and after

earnings announcements are calculated ina _{comparable way.} However, to compute

the percentage deviations ofthe volume I sum the volume during the 13 half-hour

tradingintervals instead

of

takingthe average.

The final sample consists of 1000 daytime and 1802 overnight earnings

announcements made by 336 firms. Table 2.1 _{summarizes the sample selection}

process.

Table 2.1 Sample selection

Quantitativedisclosure_{quality rating AIMR}in 1993/94, 1994/95, 1995/96 854

Expectedannouncements per firm per year x 4

Expected announcements 3416

Date or timeofearnings announcement not available (54)

No listing at NYSE orAMEX (320)

Thinlytradedor extremely priced stocks (168)

Dataforcalculations not_{available from TAQ for specific}announcement (72)

Earnings announcements 2802

Daytimeannouncements 1000

Overnightannouncements 1802

2.4 Results

2.4.1 After earnings announcements

Table 2.2 shows the cross-sectional medians

of

percentage

deviations of

the quoted

spread, effective spread, depth and volume during the 6.5 tradinghours

_after

_daytime and overnight earnings announcements from the medians

during the

non-announcement period, separately. The first

observation is that

the percentage

deviations ofthe quoted spread, effectivespread andvolume aresignificantly larger than zero after _daytime earnings announcements. These findings are in line with the

distributions of the percentage deviations are skewed. After

_{overnight earnings}

announcements, the percentage deviations

ofthe

quoted spread,volume andthe depth

are significantly larger than _zero, while the percentage

deviation of

the effective

spread is smaller than zero. This observation is interesting because the effective

spread _{(depth) is not} expected to be smaller (larger) after earnings announcements.

Thus, after overnight releases noconjecture can be made about a decline in market

liquidity,

because one dimension

of

market

_liquidity

(quoted spread) signals a

decreased

liquidity,

while the other dimension (depth) signals an increased market

liquidity.

Table

2.2 _{Percentage deviations}

of

_{quoted spread, effective spread,}

_{depth and}

volume

_after

_{earnings announcements}

Quoted spread El!ectivespread Depth Volume Daytime (N= 1 000) Expected sign + + -Median 3.85 1.14 -2.46 61.00 Significance level 0.01 0.01 0.07 0.01 Overnight (N= 1802) Expected sign + + -Median 3.04 -1.51 5.31 63.91 Significancelevel 0.01 0.01 0.01 0.01 Daytime¢*Overnight Expected sign + + -Difference median 0.81 2.65 -7.77 -2.91 Significancelevel 0.01 0.01 0.01 0.14

This table reports the cross-sectional medians ofthe percentage deviations ofthe quoted spread,

effective_{spread, depth and volumeduringthe period}afterthe earnings announcement from the median

during the _{non-announcement period. The one-tailed significance} levels for daytime andovernight

announcementsarebased on the sign test, while theone-tailedsignificance levels forthedifference between the median afterdaytimeandovernightannouncementsarebased on theWilcoxon rank sum test.

With regard to

the hypothesis, the differences between

medians of the

percentage deviations ofthe quoted spread, effective spread and depth afterdaytime

and _overnight

_{announcements are 0.81, 2.65 and -7.77, respectively, and all}

significant at0.01 one-tailed significancelevels accordance to theWilcoxon rank sum

liquidity

aftertheseannouncements. Thehigherspread andlowerdepth afterdaytime

earnings announcements affect the wealth

of

investors who trade

after the

announcement. However,thiseffect is not thatstrong that

it

prevents investors from

trading, because the percentage

deviation of

the trading volume after overnight

announcements is notsignificantlylarger than the percentagedeviation ofthevolume

after daytimeannouncements. In_{other words, despite the} higher tradingcosts due to

the relativelyhigh spreads andlowdepths, traderspreferto tradeduring the day after

daytimeannouncements instead

of

postponingtheirtrades.

2.4.2 Before earnings announcements

Table 2.3 shows results for the_{period before earnings announcements.}

Table 2.3 Percentage deviations

of

quoted spread, effective spread,

depth and

volume before earnings announcements

Quotedspread Effective spread Depth Volume Daytime (N= 1000) Expected sign + + Median 0.90 0.47 -2.08 -0.30 Significance level 0.02 0.10 0.09 0.47 Overnight (N= 1802) Expected sign + + -Median 0.67 0.28 -3.37 1.79 Significancelevel 0.01 0.23 0.01 0.17 Daytime** Overnight Expected sign + + -Differencemedian 0.23 0.19 1.29 -2.09 Significance level 0.40 0.10 0.42 0.11

This table reports the cross-sectional medians ofthe percentage deviations ofthe quoted spread,

effective spread, depth and volume during the period before the earnings announcement from the

median _{during the non-announcement period. The one-tailed significance} levels _{for daytime and}

overnightannouncementsare based on the sign test,while the one-tailed significancelevels for the difference _{between the median before daytime and overnight} announcements are based on the Wilcoxon rank sum test.

The quoted and effective spread are higher and the depth is lower before

daytimeannouncements compared to the non-announcement period. Before overnight

drop in market

liquidity

before earnings announcements.

With

_{respect to theimpact of}

intraday timing on this dropinmarket

liquidity

thetable showsone-tailed significance

levels of

the difference in the

_{median of 0.40, 0.10 and 0.42 for}

the percentage

deviations ofthe quoted spread, effective spread and depth, respectively. Thus, the

analysis presents only

evidence of

the impact

of

intraday

timing

of

earnings

announcements on theeffectivespread.

A possible

explanation for the lack

of

significant

findings for

the quoted

spread and depth is thattheperiod beforethe earningsannouncement is too long. If,

for instance, the market maker knows that an announcement will take place on

Wednesday the threat

of

informed tradingis higheron Wednesday than on Tuesday.

However, I analyze

6.5 trading hours before the

announcement. When the

announcement takes place at 11.10 AMthis period consists of 1.5 trading hours on

Wednesday and5 _tradinghours on Tuesday. Therefore,focusing on a_{shorter period}

before the announcement mayimprovethepower

ofthe

tests.

Unfortunately, the

investigation of

a shorter period before the earnings

announcementis_{problematic for two}reasons.First, when I focus onashorterinterval

the effect

of

possible

mismeasurement of the time

of

earnings announcements

becomes stronger. When for instance an _{earnings announcement} took place 31

minutes before the time stamp of the news wire and the spread increased strongly

after the announcement, this mismeasurement has a stronger effect on the average

spread duringonetrading hourbefore the announcement than on the average spread

during6.5_tradinghoursbefore the announcement.

The secondreason isthe

impact of

the intraday pattern

of

spreads and depths.

LMR showthat during trading dayswithoutearnings announcements the quoted and

effective spread have a U-shape pattern with relatively high spreads

during the half

hour after opening and the half hour before closing ofthe_{market. The depth shows an}

opposite pattern. When I focus onone trading hourbefore earnings announcements,

for

overnight announcementsthis trading hour is always the lasttrading hourbefore

closing of

themarket. Spreadsduring this hour duringthe non-announcementperiod

are relatively high while depths are

relatively low.

For daytime announcements the

trading hourbefore the announcementdepends on the time oftheannouncement and

also occursduringperiods

with

relatively lowspreads andhighdepths. Therefore, the

spread and depth during the non-announcement period

differ for daytime and

before announcements. This effect makes

it

impossible to compare the percentage

deviations of

the spread and depth during a short interval before daytime and

overnightearnings announcements.

Table 2.4 _{Percentage deviations}

of

_{quoted spread, effective spread,}

_{depth and}

volume before

earnings announcements for firms with

/ow turnover

Quoted spread E1Iective spread Depth Volume

Daytime (N=567) Expected sign + + -Median 0.82 0.98 -2.23 2.53 Significancelevel 0.16 0.05 0.09 0.29 Overnight (N=831) Expected sign + + -Median 1.44 0.90 -5.16 2.15 Significancelevel 0.01 0.08 0.03 0.27 Daytime** Overnight Expected sign + + -Differencemedian -0.62 0.08 2.93 0.38 Significancelevel 0.16 0.11 0.33 0.17

This table reports the cross-sectional medians ofthe percentage deviations ofthe quoted spread,

effective spread, depth and volume during the period before the earnings announcement from the

median _during the _{non-announcement period. The one-tailed significance levels for daytime and}

overnightannouncementsarebased on the sign test, whiletheone-tailed significance levels for the difference between the median before daytime and overnight announcements are based on the

Wilcoxon rank sum test.

Another_{possibleexplanation for the lack}

of

_{significant findings forthe quoted}

spread and

depth is

the sample

selection. One of

the selection

criteria is the

availability of

adisclosurequality rating

ofthe AIMR.

Asstatedbefore,this criterion

may bias

the sample against

finding

results during the period before earnings

announcements, because

AIMR

rated

firms tend to

be larger. To provide some

evidence onthe

impact of firm size I redo

the analysis for announcements

of

firms

with

anaveragedaily turnover onthe

NYSE/AMEX

_duringthemonth

of

January, that

is closest to the announcement, that issmaller thanthe mediandaily turnover of the

full

sampleduringthat period.Table_{2.4 presents the}results for thissampleconsisting

between thedeviations of_{the quoted spread and depth beforedaytime}and _overnight

announcements are not_significantatconventionallevels.

2.4.3 Sensitivity tests

Prior analysis reveals strong evidence that intraday

timing

affects market

liquidity

after earnings announcements and some evidence that intraday timing impacts market

liquidity

before announcements. In this section I investigate whether these findings are driven by firm-specific factors, cross-listings or differences in the information content ofthe announcements.

Firm-specific factors

In the previousanalysis I didnotcontrol

for

firm-specific issues. However, firms that

consequently announcetheirearningsovernight are only included inthe sample with

overnight _{announcements,} while firms that always release their earnings during tradinghours areonlypresent in the other sample. _{Therefore, it} is possible that some

firm-specific factors are driving the difference between market

liquidity

around

daytimeandovernightannouncements.Toensure that theresults are not due to firm-specific factors I redothe analysis foramatched sample

of

earnings announcements.

This ispossible because for each

firm

several earningsannouncementsareincluded in

the sample. The announcementsarematchedasfollows. For each firm I determine the

number

of

daytime andovernight announcements. When there areless daytime than overnight announcements, the daytime releases are matched with overnight

announcements thatarerandomly drawn fromtheavailable overnightreleases for that

firm.

When there are less overnight announcements, these announcements are

matched_{with randomlypickeddaytime}announcements. The final_sampleconsists of

407 daytime and 407overnight earningsannouncements made by 184 firms. Table

2.5 shows the results for this matched sample

with regard to the period

before

earnings announcements. The median percentage

deviation of

the effective spread

before daytime announcements

declines from 0.47 to 0.01, while

the median

percentage deviationbeforeovernightannouncements _{drops from 0.28 to -0.23. The}

one-tailed significance

_{level of}

the difference in the median percentage

deviation of

0.28. This

may

indicate that

the previous finding that intraday

timing of

announcementsaffectseffectivespreads isdrivenbyfirm-specificfactors.

Table 2.5 Percentage deviations

of

quoted spread, effective spread,

depth and

volume before earnings announcements

for

_{matched sample}

Quotedspread Effective spread Depth Volume

Daytime (N=407) Expected sign + + -Median 0.90 0.01 -2.23 0.73 Significancelevel 0.21 0.40 0.16 0.40 Overnight (N=407) Expected sign + + -Median 1.16 -0.23 -4.19 -3.86 Significance level 0.08 0.42 0.06 0.08 Daytime¢* Overnight Expected sign + + -Differencemedian -0.26 0.24 1.96 4.59 Significancelevel 0.24 0.28 0.34 0.21

This table reports the cross-sectional medians ofthe percentage deviations ofthe quoted spread,

effective spread, depth and volume duringthe period before the earnings announcement from the

median during the non-announcement period. The one-tailed significance levels for daytime and overnightannouncementsare based on the sign test,whilethe one-tailedsignificance levels for the

difference between the median before daytime and overnight announcements are based on the

Wilcoxon rank sum test.

Firm-specijic factors and cross-listings

In addition to theimpact

of

firm-specific_factors, cross-listingsmay affectthe results

for the

period _{after earnings announcements. When shares are traded on another}

exchange during

or

after the earnings announcement, US market makers may

conditiontheopening price not only ontheorder flowandearnings news, but also on the shareprice on the otherexchange. This additional information item may impact the probability

of

informed trading and may drive the difference between market

liquidity

after daytime and overnight announcements.

However, in that case, the

reduction in

the market

liquidity

drop after earnings

announcements on the

NYSE/AMEX

takes _{place when} a market

_{liquidity drop}

_may occur on another

earnings announcements, but may

just

shift the decline in

liquidity

to another

exchange.

Table 2.6 _{Percentage deviations}

of

_{quoted spread, effective spread,}

depth and

volume

after

earnings announcements

for

matched sample of

firms

without

cross-listingsat

London

Stock Exchange

or Tokyo

Stock Exchange

Quoted spread Effectke spread Depth Voiume

Daytime (N=297) Expected sign + + -Median 4.09 1.08 -3.59 74.27 Significance level 0.01 0.18 0.32 0.01 Overnight (N=297) Expected sign + + -Median 2.91 -0.07 5.06 49.66 Significance level 0.01 0.48 0.10 0.01 Daytime¢* Overnight Expected sign + + -Difference median 1.18 1.15 -8.65 24.61 Significance level 0.07 0.08 0.01 0.03

This table reports the cross-sectional medians ofthe percentage deviations ofthe quoted spread,

effective_{spread, depth}andvolumeduringthe periodafter_{the earnings announcement from the median} during the _{non-announcement period. The one-tailed} significance levels for daytimeand overnight announcements arebased on the sign test, whilethe one-tailedsignificance levels forthe difference between the median after daytime andovernightannouncementsarebased on theWilcoxon rank sum test.

To test for

this conjecture and the impact

of

firm-specific factors I redo the analysis for the period after earnings announcements for the matched sample after excluding firms

with

cross-listings at the London _{Stock Exchange or Tokyo} Stock

Exchange during 1993 to 1996: Table 2.6 revealsthatdifferences in the percentage

deviations in

_{the quoted spread,} effective spread and depth after daytime and

overnightannouncementsare

still

_{significant at 0.07,0.08 and} 0.01 one-tailed levels.

5 ACCordingto Roberts, WeetmanandGordon (1998)thesetwo exchanges were themajornon-US stock exchangesduring1996.Listing attheLondonStock ExchangeisbasedonTimbrelland Tweedie

(1994,1995 and 1996),while listing attheTokyoStock Exchangeisdetermined withanoverview on

listingand_delistingof_{foreigncompanies from 1973 up to} 1996_{provided by}the_TokyoStock