Corporate Governance and Equity Liquidity: An Analysis of S&P Transparency and Disclosure Ranking

Similar documents
MgtOp 215 Chapter 13 Dr. Ahn

NYSE Specialists Participation in the Posted Quotes

THE VOLATILITY OF EQUITY MUTUAL FUND RETURNS

R Square Measure of Stock Synchronicity

Clearing Notice SIX x-clear Ltd

Money, Banking, and Financial Markets (Econ 353) Midterm Examination I June 27, Name Univ. Id #

Final Exam. 7. (10 points) Please state whether each of the following statements is true or false. No explanation needed.

Evaluating Performance

FORD MOTOR CREDIT COMPANY SUGGESTED ANSWERS. Richard M. Levich. New York University Stern School of Business. Revised, February 1999

Economic Design of Short-Run CSP-1 Plan Under Linear Inspection Cost

Price and Quantity Competition Revisited. Abstract

ASSET LIQUIDITY, STOCK LIQUIDITY, AND OWNERSHIP CONCENTRATION: EVIDENCE FROM THE ASE

Chapter 10 Making Choices: The Method, MARR, and Multiple Attributes

CHAPTER 9 FUNCTIONAL FORMS OF REGRESSION MODELS

Highlights of the Macroprudential Report for June 2018

Real Exchange Rate Fluctuations, Wage Stickiness and Markup Adjustments

Construction Rules for Morningstar Canada Dividend Target 30 Index TM

An Application of Alternative Weighting Matrix Collapsing Approaches for Improving Sample Estimates

Domestic Savings and International Capital Flows

Market Opening and Stock Market Behavior: Taiwan s Experience

SYSTEMATIC LIQUIDITY, CHARACTERISTIC LIQUIDITY AND ASSET PRICING. Duong Nguyen* Tribhuvan N. Puri*

Tests for Two Correlations

Spurious Seasonal Patterns and Excess Smoothness in the BLS Local Area Unemployment Statistics

An Empirical Study on Stock Price Responses to the Release of the Environmental Management Ranking in Japan. Abstract

TRADING RULES IN HOUSING MARKETS WHAT CAN WE LEARN? GREG COSTELLO Curtin University of Technology

Construction Rules for Morningstar Canada Dividend Target 30 Index TM

A MODEL OF COMPETITION AMONG TELECOMMUNICATION SERVICE PROVIDERS BASED ON REPEATED GAME

Construction Rules for Morningstar Canada Momentum Index SM

Teaching Note on Factor Model with a View --- A tutorial. This version: May 15, Prepared by Zhi Da *

FM303. CHAPTERS COVERED : CHAPTERS 5, 8 and 9. LEARNER GUIDE : UNITS 1, 2 and 3.1 to 3.3. DUE DATE : 3:00 p.m. 19 MARCH 2013

Which of the following provides the most reasonable approximation to the least squares regression line? (a) y=50+10x (b) Y=50+x (d) Y=1+50x

REFINITIV INDICES PRIVATE EQUITY BUYOUT INDEX METHODOLOGY

Module Contact: Dr P Moffatt, ECO Copyright of the University of East Anglia Version 2

Spatial Variations in Covariates on Marriage and Marital Fertility: Geographically Weighted Regression Analyses in Japan

A copy can be downloaded for personal non-commercial research or study, without prior permission or charge

ECONOMETRICS - FINAL EXAM, 3rd YEAR (GECO & GADE)

Limits of arbitrage and corporate financial policy

GROWTH STRATEGIES AND CAPITAL STRUCTURES OF SMALL AND MEDIUM-SIZED ENTERPRISES *

Mutual Funds and Management Styles. Active Portfolio Management

3/3/2014. CDS M Phil Econometrics. Vijayamohanan Pillai N. Truncated standard normal distribution for a = 0.5, 0, and 0.5. CDS Mphil Econometrics

Tests for Two Ordered Categorical Variables

Finance 402: Problem Set 1 Solutions

AN ANALYSIS OF LIQUIDITY ACROSS MARKETS: EXECUTION COSTS ON THE NYSE VERSUS ELECTRONIC MARKETS

Risk, return and stock performance measures

Quiz on Deterministic part of course October 22, 2002

The Impact of Governance on IFRS Restatement Quality

Wenjin Kang and Wee Yong Yeo. Department of Finance and Accounting National University of Singapore. This version: June 2007.

>1 indicates country i has a comparative advantage in production of j; the greater the index, the stronger the advantage. RCA 1 ij

Raising Food Prices and Welfare Change: A Simple Calibration. Xiaohua Yu

Prospect Theory and Asset Prices

On the Style Switching Behavior of Mutual Fund Managers

The Integration of the Israel Labour Force Survey with the National Insurance File

Earnings Management and Stock Exposure to Exchange Rate Risk

MODELING THE BID/ASK SPREAD: On the Effects of Hedging Costs and Competition

ISE High Income Index Methodology

Measurement and Management of Exchange Rate Exposure: New Approach and Evidence

Consumption Based Asset Pricing

Analysis of Moody s Bottom Rung Firms

Corporate Governance and Liquidity #

Elements of Economic Analysis II Lecture VI: Industry Supply

Foreign Exchange Exposures, Financial and Operational Hedge Strategies of Taiwan Firms

UNDERPRICING AND EX ANTE UNCERTAINTY IN IPOS: EVIDENCE FROM THE TUNISIAN STOCK MARKET

Chapter 5 Bonds, Bond Prices and the Determination of Interest Rates

2) In the medium-run/long-run, a decrease in the budget deficit will produce:

Risk and Return: The Security Markets Line

Linear Combinations of Random Variables and Sampling (100 points)

Asset Management. Country Allocation and Mutual Fund Returns

Informational Content of Option Trading on Acquirer Announcement Return * National Chengchi University. The University of Hong Kong.

The study of relationship between shareholder values added (sva) and different criteria of the risk adjusted return

Managing EPS Through Accelerated Share Repurchases: Compensation Versus Capital Market Incentives

Risk Reduction and Real Estate Portfolio Size

Conditional Beta Capital Asset Pricing Model (CAPM) and Duration Dependence Tests

Networks in Finance and Marketing I

Lecture 10: Valuation Models (with an Introduction to Capital Budgeting).

Network Analytics in Finance

Financial Risk Management in Portfolio Optimization with Lower Partial Moment

Labor Market Transitions in Peru

Kent Academic Repository

ISE Cloud Computing Index Methodology

A Meta Analysis of Real Estate Fund Performance

Accounting Information, Disclosure, and the Cost of Capital

Evaluation of the Factors Affecting Initial Public offering Underpricing by Newly-accepted Companies into Tehran Stock Exchange

Investment Management Active Portfolio Management

OCR Statistics 1 Working with data. Section 2: Measures of location

Hybrid Tail Risk and Expected Stock Returns: When Does the Tail Wag the Dog?

Retail Mortgage Backed Securities, Commercial Asset Backed Securities and Corporate Bonds: a Credit Spread Comparison +

4. Greek Letters, Value-at-Risk

Testing the weak efficient market hypothesis using Bangladeshi panel data

Random Variables. b 2.

Family control and dilution in mergers

Firm fundamentals, short selling, and stock returns. Abstract

Stochastic ALM models - General Methodology

occurrence of a larger storm than our culvert or bridge is barely capable of handling? (what is The main question is: What is the possibility of

Monetary Tightening Cycles and the Predictability of Economic Activity. by Tobias Adrian and Arturo Estrella * October 2006.

Notes are not permitted in this examination. Do not turn over until you are told to do so by the Invigilator.

15-451/651: Design & Analysis of Algorithms January 22, 2019 Lecture #3: Amortized Analysis last changed: January 18, 2019

The pricing discount for limited liquidity: Evidence from SWX Swiss Exchange and the Nasdaq

Forecasts in Times of Crises

Price Formation on Agricultural Land Markets A Microstructure Analysis

Anatomy of a Government Intervention in Index Stocks

Transcription:

Corporate Governance and Equty Lqudty: An Analyss of S&P Transparency and Dsclosure Rankng We-Peng Chen Humn Chung Cheng-few Lee We-L Lao ABSTRACT Ths paper nvestgates the effects of dsclosure and other corporate governance mechansm on equty lqudty. We post that companes wth poor nformaton dsclosure and transparency practces ncur more serous nformaton asymmetry problem. Snce poor corporate governance leads to greater nformaton asymmetry, lqudty provders wll ncur relatvely hgher adverse nformaton rsks and wll therefore offer hgher adverse nformaton components of the effectve bd-ask spreads. The S&P T&D rankngs on the ndvdual stocks of S&P 500 ndex are employed to examne whether frms wth greater T&D rankngs have lower the effectve bd-ask spread of ther stocks. Our results reveal that companes wth poor nformaton dsclosure and transparency practces have larger economc costs of equty lqudty,.e., the effectve spread. JEL classfcaton: G10; G30; G34 Keywords: Corporate Governance, Transparency and Dsclosure, Asymmetrc nformaton costs, Lqudty Correspondng author: We-Peng Chen. All authors are from Graduate Insttute of Fnance, Natonal Chao Tung Unversty, 1001 Ta-Hsueh Road, Hsnchu 30050, Tawan. Tel: +886-3-57111 ext.57070; Fax: +886-3-573360; E-mal: wpchen.ms9g@nctu.edu.tw (We-Peng Chen); chunghu@mal.nctu.edu.tw (Humn Chung); rahn7001@yahoo.com.tw (We-L Lao) The fnancal support of the Natonal Scence Councl of Tawan s also gratefully acknowledged. 1

1. Introducton Fnancal transparency and nformaton dsclosure are mportant elements of corporate governance. Better fnancal transparency and nformaton dsclosure can help shareholders to understand more amply about frm s management, and reduce the nformaton asymmetres faced by nvestors. Reflectng on the equty market, nvestors are wllng to pay hgher prce to get stocks of companes that have better nformaton dsclosure. Furthermore, there are more nvestors who are wllng to trade the stocks wth better nformaton dsclosure, so these stocks wll have better market lqudty. Recently, the ssue on frm s fnancal transparency and nformaton dsclosure has ganed much attenton by market regulators and nvestors. Rankng nsttutons such as Standard & Poor s and Moody tend to use fnancal transparency and nformaton dsclosure as one of ther crtera of assessng frm s managng ablty and reputaton. On October 16, 00, Standard & Poor s publsh the results of ther Transparency and Dsclosure Study (T&D Study) 1. Accordng to each frm s T&D practces, ths study provdes frm s Transparency and Dsclosure rankngs (T&D rankngs) n three dsclosure categores and then calculatng a fnal rankng. These rankngs provde a reference that enables nvestors to assess frm s transparency and dsclosure practce. Ths paper nvestgates the relatonshp between corporate governance and equty lqudty. We conjecture that companes wth poor corporate governance ncur hgher nformaton asymmetrc costs. Lqudty supplers wll take ther prce protecton acton and thus broaden the effectve bd-ask spread of the equty of frms wth poor corporate governance. The S&P T&D rankng s used as a proxy varable for corporate governance, and s employed to examne whether frms wth greater rankngs have better lqudty for ther stocks. Lqudty s usually defned as the ablty that an asset can be trade quckly wth the least cost of searchng counterpart and the least prce concesson. Stoll (000) ndcates that mmedate sales are usually made at the bd prce, and mmedate purchase are usually made at ask prce. On the one hand, the spread s the prce concesson needed for an mmedate transacton to lqudty demanders; on the other 1 The T&D study focus on several questons such as: whch companes provde the most extensve dsclosure n ther basc corporate flngs? Whch companes dsclose above and beyond what the law requres? See Patel and Dallas (00) for a detal descrpton.

hand, t s the revenue earned by lqudty supplers such as market makers or dealers. Thus, the quoted bd-ask spread s often used as a measure of market lqudty. Furthermore, from ther emprcal result, Ln, Sanger, and Booth (1995) ndcate that demanders of mmedacy servces rarely receved prces whch were less favorable than prevalng quotes on the NYSE. Therefore, another better measure, effectve spread, whch s defned as the absolute value of the dfference between the trade prce and the quote mdpont just pror the trade, s vewed as a more precse measure of frm s market lqudty. We use the effectve spread as the proxy for frm s market lqudty and examne whether t s nfluenced by T&D rankng. When a company s nformaton dsclosure and transparency practces become worse, the market wll expect to face more serous nformaton asymmetry problem. In order to offset ths possble adverse selecton problem, the lqudty supplers wll take ther prce protecton acton,.e. to broaden the spread of the frm s equty. Consequently, an ncrease n nvestors demand for trade mmedacy mght ncur hgher transacton costs due to hgher spread and therefore reducng market lqudty of the stock. Thus, we predct that there should be a negatve relaton between frm s S&P T&D rankng and the effectve spread of ts stock. From the vew of the lqudty supplers, bd-ask spread s prmarly composed of three components: the order processng cost, nventory cost, and adverse selecton cost of nformaton asymmetry (Ln, Sanger, and Booth, 1995). The order processng and nventory cost can be vewed as the real cost of lqudty supplers. These two costs come from the use of real economc resources to provde mmedate trades. Adverse nformaton component s a compensaton that arses from nformaton asymmetrc rsk faced by lqudty supplers. Because t s dffcult to tell who the nformed trader s, the lqudty supplers can t prevent the loss when they actually trade wth an nformed trader. Approprate adverse nformaton component of the effectve spread must be exsted to compensate ths rsk of loss, and the lqudty provders therefore could mantan ther operaton aganst nformed tradng actvtes. Intutvely, T&D rankng should be drectly correlated wth adverse nformaton component. Ths s because that the worse T&D rankng mples worse dsclosure practce and thus nduces hgher nformaton asymmetrc rsk faced by lqudty supplers. To compensate ths hgher rsk, lqudty supplers wll ncrease adverse nformaton component of the effectve spread n response. For the reason above-mentoned, we follow the model suggested by Huang and Stoll (1994), Ln (199), and Ln, Sanger, and Booth (1995) to calculate the adverse nformaton component of the effectve spread, and use t as a measure of mmedate transacton 3

cost due to nformaton asymmetrc rsk of the frm. We predct that there should be stronger negatve relaton between the frm s T&D rankng and the adverse nformaton component of ts effectve spread. Our study may also provde an analyss on the qualty and accuracy of T&D rankngs by testng f they have sgnfcant relaton to the market lqudty of frms stocks. Accordng to many prevous studes such as Copeland and Gala (1983), Glosten and Mlgrom (1985), Kyle (1985), Welker (1995), and Stoll (000), the spread ncludes the value of the adverse selecton cost due to nformaton asymmetry, so we expect the frm havng greater dsclosure practce wll have less nformaton asymmetrc cost, and thus, smaller effectve bd-ask spread. If S&P T&D rankng s adequate proxy of frm s dsclosure practce, we predct that the frm wth greater T&D rankng wll have both less effectve bd-ask spread and adverse nformaton component of ts equty, mplyng better market lqudty; nversely, f the frm has lower rankng, whch mplyng worse dsclosure practce, we predct both wder effectve spread and adverse nformaton component of ts stock, whch representng worse market lqudty. Several past researches, ncludng theoretcal and emprcal works, have ndcated that smultanety may exst n the determnaton of bd-ask spread and frm s dsclosure polcy (Dye, 1985; Lang and Lundholm, 1993; Welker, 1995). When a manager determnes frm s dsclosure polcy, he or she s lkely to consder present market lqudty of frm s stock. Besdes, when lqudty supplers quote the bd and ask prce of a stock, they necessarly take ths frm s dsclosure practce as mportant reference of the degree of nformaton asymmetry. Accordngly, a test for endogenety s needed. Hence, we apply the Hausman test to examne whether effectve spread and dsclosure polcy are smultaneously determned. If the result of Hausman test ndcates that there are no smultanety n the determnaton of effectve spread and frm s dsclosure polcy, an OLS procedure s approprate to examne the relaton of effectve spread and T&D rankng. If the result exhbts exstng of smultanety, we must use SLS procedure to remove the estmated bas nduced by endogenety of explanatory varable and get the consstent estmates of the parameters n our model. The same procedure s also employed before we analyss the partal effect of T&D rankng on adverse nformaton component of the effectve spread. Our emprcal results reveal that smultanety does not exst n the determnaton of effectve spread and frm s T&D rankng. Thus, we don t have to employ SLS procedure to estmate and test the relaton between the effectve spread and T&D 4

rankng. The same scenaro s found n the model of T&D rankng and adverse nformaton component of the effectve spread. After controllng frm s tradng characterstcs, the result of OLS procedure exhbts that the negatve relaton between effectve spread and composte bass T&D rankng s statstcally sgnfcant whle the coeffcent estmate of annual bass T&D rankngs s not statstcally sgnfcant. Although the sgn of parameter estmate s consstent wth predcton, the OLS result for adverse nformaton component and composte bass T&D rankngs s not statstcally sgnfcant. Unfortunately, the drecton of the relaton between adverse nformaton component and annual bass T&D rankngs s nconsstent wth our predcton and s not statstcally sgnfcant. The parameters estmates of all control varables except average daly dollar volume are n the drecton that we predcted and statstcally sgnfcant. The remander of ths paper s organzed as follows. A revew of the related lterature s undertaken n the next secton, followed, n the subsequent secton, by a descrpton of the data and the research methodology adopted for ths study. The penultmate secton presents the emprcal results of our research, wth the fnal secton provdng some concludng remarks drawn from ths study.. Lterature revew.1 Dsclosure practce, corporate governance, and nformaton asymmetry Many prevous studes had ndcated the relatonshp between dsclosure practce and corporate governance. Lowensten (1996) argues that good dsclosure has been a most effcent and effectve mechansm for nducng managers to manage better. Ths mples that frms wth better nformaton dsclosure may have better corporate governance. Patel and Dalas (00) argues, n ther report of S&P Transparency and Dsclosure methodology and study results, that good corporate governance ncludes a vglant board of drectors, tmely and adequate dsclosure of fnancal nformaton, meanngful dsclosure about the board and management process, and a transparent ownershp structure dentfyng any conflcts of nterests between managers, drectors, shareholders, and other related partes. Therefore, transparency and dsclosure are very mportant and basc element of corporate governance, ndcatng agan that good dsclosure practce could be vewed as good corporate governance. The extent of dsclosure practce wll affect the nformaton asymmetrc rsk of a 5

frm. Hgher levels of dsclosure should lead to lower cost of captal by reducng the nformaton rsk and the transacton costs (Lang and Lundholm 1999). Patel and Dallas (00) also show that both the composte and the annual bass T&D rankngs have negatve relatonshp wth market rsk. Botosan (1997) also argues that frms ncreased dsclosures can reduce the nformaton asymmetry between managers and nvestors and thus reduce frms cost of equty captal. Accordngly, our study predcts that f S&P T&D rankngs can descrbe frms dsclosure practces well, the frms wth hgher T&D rankngs wll have better dsclosure practces and thus ncrease ther market lqudty by reducng nformaton asymmetrc rsk.. Market lqudty and dsclosure practce Market lqudty could be measured by how long t takes optmally to trade a gven amount of an asset, or be measured by the prce concesson for an mmedate transacton (Lppman and Mccall, 1986; Demsetz, 1968). Under ths vew, the market lqudty s vewed as the prce of mmedacy, and the spread whch determned by dealer s order processng cost, nventory holdng cost, and nformaton asymmetrc cost s one measure of market lqudty. Stoll (1978) models the source of that spread n the sprt of Demsetz by analyzng cross-sectonal relaton of the stock s proportonal quoted half-spreads to frm s tradng characterstcs and fnds that ths relaton s strong and has changed a lttle over tme. Later vews of market lqudty reled on nformaton arguments as n Copeland and Gala (1983), Glosten and Mlgrom (1985), Kyle (1985), and Ln, Sanger, and Booth (1995). Under ths vew, the spread s the value of nformaton lost to tmeler or better nformed traders. Welker (1995) consdered that quoted bd-ask spread set by market specalsts are an ncreasng functon of the adverse selecton rsk perceved by specalsts, and perceved adverse selecton rsk s a functon of frm dsclosure practces. He used smultaneous equatons n whch both spreads and dsclosure practce rankngs appear as endogenous varables to conduct tests for cross-sectonal dfferences n the relaton between dsclosure polcy and bd-ask spreads. After controllng for return volatlty, tradng volume, and share prce, the emprcal results reveal predcted negatve relaton between dsclosure practce rankngs and proportonal quoted bd-ask spreads. However, the results are not sgnfcant due to the ncrease n standard errors accompanyng the two-stage least squares procedure and parttonng of the sample. Extendng the research of Welker, ths study uses S&P T&D rankng as a proxy 6

for frm s dsclosure practces, and we conjecture that the rankng could be a good measure of adverse selecton rsk perceved by dealers or market makers. Furthermore, rather than usng quoted bd-ask spread, we use the effectve spread, a more precse measure of frm s lqudty, and adverse nformaton component of the effectve spread to examnng the relaton between frm s dsclosure practce and ts market lqudty. If S&P T&D rankng s ndeed a good proxy for frm s dsclosure practces, we expect that the frm wth hgher T&D rankng wll have both smaller effectve spread and adverse nformaton component, ndcatng better market lqudty..3 S&P Transparency and Dsclosure Rankngs The proxy for frm s dsclosure practces n our study s the T&D rankng provded by S&P Transparency and Dsclosure study. The study dentfes 98 dsclosure tems, classfed nto three broad categores (Patel and Dallas, 00): (1) Ownershp structure and nvestor rghts, () Fnancal transparency and nformaton dsclosure, and (3) Board and management structure and process. The study ndcates whether these ndvdual tems are dsclosed, focusng prmarly on annual reports as the prmary source of corporate dsclosure. In addton, ths study also consders about addtonal forms of regulatory flngs for another source of corporate dsclosure. Therefore, the study evaluates dsclosure patterns both on annual report alone, whch s called annual bass, and on a composte bass, whch ncorporates annual reports, 10-Ks, and other proxy statements. Each rankng of the three categores s evaluated on both two bases and then the fnal rankngs of these two bases are calculated. We use these two fnal rankngs respectvely as the proxy for frm s dsclosure practces, and want to examne f the rankng can be a good proxy for frm s dsclosure practces by testng whether the rankng s sgnfcant negatve to the effectve spread and adverse nformaton component. Although Patel and Dallas clam that whle transparency and dsclosure are key components of corporate governance, T&D rankngs are not proxes for corporate governance, they stll fnd that the rankngs reveal some nterestng relatons to frm s market rsk, prce to book rato, and captalzaton. Several recent studes also provde evdences that T&D rankngs could be good proxes for corporate governance. 7

Durnev and Km (00) show that the S&P T&D rankngs are postvely correlated wth the strength of corporate governance n emergng countres. Cheng, Collns, and Huang (003) use T&D rankngs as proxes for corporate governance to nvestgate the effects of the level of these rankngs and the dfferental rankngs between composte and annual report rankngs on three market metrcs: market beta, rsk-adjusted abnormal returns and earnngs response coeffcents surroundng the announcement date. The results reveal that the release of the S&P T&D rankngs brought new nformaton to the market and that the rankngs affect shareholder wealth n a manner that s consstent wth the rankngs measurng the strength of corporate governance. In ths study, we also vew S&P T&D rankng as a possble proxy for corporate governance, and want to fnd whether T&D rakngs can correctly measure the dfferences among dfferent frms nformaton dsclosure practces. We do ths by testng f these rankngs have good explanatory ablty to frms market lqudty. If T&D rankngs are good measures of frms nformaton dsclosure practces, we predct that there wll be negatve relatons between T&D rankngs and effectve spreads (and adverse nformaton components), whch ndcate better dsclosure practce accompanyng wth better market lqudty..4 Smultanety of bd-ask spread and dsclosure practce Several past researches, ncludng theoretcal and emprcal works, have ndcated that smultanety may exst n the determnaton of bd-ask spread and frm s dsclosure polcy. Lang and Lundholm (1993) analyzes the determnants of voluntary dsclosure polcy and argues that there s smultanety n the determnaton of bd-ask spread and dsclosure practce. Welker (1995) suggests that dsclosure polcy choce may be nfluenced by the level of nformaton asymmetry between management and unnformed nvestors as well as other determnants of bd-ask spreads. If the smultanety exsts ndeed, the OLS procedure wll have bas estmates, and our emprcal tests wll also have ncorrect results. Accordngly, we utlze the determnants of dsclosure practce and bd-ask spread to construct Hausman test, whch s used to detect whether an explanatory varable n a regresson model s endogenous. The determnants of dsclosure practce usng n the Hausman test are from Lang and Lundholm (1993) and Welker (1995). Lang and Lundholm fnd that both the market adjusted return and frm sze are postvely related to dsclosure polcy, and that the dsclosure polcy s negatvely related to return standard devaton and return-earnngs correlaton. Besdes, dsclosure levels ncrease f the frm wll ssue 8

new equtes or new debts n the followng two years. Welker follows the fndngs of Lang and Lundholm and uses share prce, securty offerng, return, and return standard devaton as the determnants of dsclosure practce. The determnants of dsclosure practce used n our study are closng prce, return, market value, and, return standard devaton. We wll use all these varables and the other determnants of bd-ask spread as exogenous varables n the Hausman Test. If the result of test reveals sgnfcant evdence that T&D rankng s endogenous, we then have to use two-stage least square procedure nstead of OLS to estmate and test our emprcal models. Otherwse, the OLS procedure s adequate and approprate. 3. Data and Research Methodology 3.1 Data A jont hypothess examned n ths study s whether S&P T&D rankng s a good proxy for frm s nformaton dsclosure practce. For ths purpose we choose the consttuent stocks of S&P 500 ndex for our emprcal test, and the S&P T&D study has T&D rankngs for these stocks. Because the S&P T&D study s publshed on October 16, 00, we pck October 17, 00 to December 31, 00, a perod of 5 tradng days, as our studyng perod. We expect that the mpact of T&D rankngs on frm s market lqudty mght be the most sgnfcant durng ths perod. Stoll (000) ndcated that several prevous emprcal studes have shown that market desgn appears to have an effect on spread. Especally, there s large emprcal evdence comparng dealer and aucton markets, such as NASDAQ and NYSE (Huang and Stoll, 1996; Barclay et al, 1999). In partcular, the spreads n dealer markets are wder than those n aucton markets because dealers may have more market power n dealer markets. The dealers or market makers wth stronger market power are expected to ncrease ther revenues by wdenng bd-ask spreads. In order to elmnate ths dfference among the consttuent stocks of S&P 500 ndex, we only use the stocks lsted n NYSE. Under ths condton, our sample sze becomes 405 stocks. In order to calculate the proxy of frm s market lqudty and other varables used n our cross-sectonal model, we need daly ntraday transactons data and quote data for these 405 stocks. Our data are obtaned prmarly from the Trade and Quote (TAQ) database whch contans ntraday transactons data (trades and quotes) for all securtes lsted on the New York Stock Exchange (NYSE) and Amercan Stock Exchange (AMEX), as well as Nasdaq Natonal Market System (NMS) and SmallCap 9

ssues. We use the TAQ database to obtan ntraday tradng and quoted data such as closng prces, bd and ask prces, number of trades, and daly dollar volume. Addtonally, we also obtan stocks daly returns wthout dvdends and market values from CRSP and Compustat database. The data s arranged as follows. For ntraday tradng and quoted data, we delete the data outsde the ordnary tradng tme,.e. 9:30 AM to 4:00 PM. Besdes, we calculate the average value of each varable durng our sample perod. The frms that have at least one ncorrect or mssng value of varable used n ths study are deleted from our sample. Consequently, our sample sze reduces to 364. 3. Methodology 3..1 Effectve spread and adverse nformaton component Followng standard market mcrostructure lterature, the effectve spread s used as the proxy of equty lqudty. Ths varable s used by Huang and Stoll (1996), Ln, Sanger, and Booth (1995), and Stoll (000), to measure frm s market lqudty, and s defned as the absolute value of the dfference between the trade prce and the quote mdpont just pror the trade : ESP = P Q (1), where P s the trade prce and Q s the quote mdpont just pror to the trade. Lee and Ready (1991) conclude that prevalng quote may sometmes be recorded ahead of trades. Accordng to ther suggeston, we follow Ln, Sanger, and Booth (1995) to dentfy the prevalng quotes for each transacton as the quotes that are n the effect fve seconds, and then compute the quote mdpont as the average of the prevalng quoted bd and ask prces. In order to examne the relaton between the frm s dsclosure practce and the nformaton asymmetrc cost faced by lqudty supplers when tradng ts stocks, we addtonally estmate the adverse nformaton component of the effectve spread of each frm s stock. Followng Ln (199), Huang and Stoll (1994), and Ln, Sanger, and Booth (1995), we compute the adverse nformaton component of the effectve spread by followng procedure. Frst, let Q be the quote mdpont just pror to the t 10

trade, and z t = Pt Qt s the sgn effectve half-spread where t P s the trade prce at tme t. After computng the value of z t for each trade, we estmate followng models to get the adverse nformaton component as a fracton of the effectve spread: Q z t + 1 t + 1 Q = θ z t t = + λ z η t t + 1 +, e t + 1, (), (3), where λ s the adverse nformaton component as a fracton of the effectve spread, and the dsturbance terms e t + 1 and η t + 1 are assumed to be uncorrelated. The dea of ths model s straghtforward. The lqudty supplers such as market makers or dealers wll adjust ther quotes after each trade n response to the new nformaton conveyed by the pror trade. Therefore, the coeffcent λ can be explaned as the adverse nformaton component of effectve spread, and lqudty supplers use t to determne the approprate quote revson after a trade. We compute λ ˆ for each trade durng our samplng perod, and then calculate the average z t value of t for each frm durng the sample perod as our measure of the dollar adverse nformaton component. 3.. Control varables of the cross-sectonal regresson model Prevous cross-sectonal studes of spreads suggest a number of spread determnants other than dsclosure polcy that should be controlled for n the emprcal analyss (Welker, 1995). Stoll (000) models the source of the spread, and fnd that daly dollar volume, the return varance, the number of trades per day, and the stock s closng prce have sgnfcant relaton to proportonal quoted half-spread. These varables explan over 65 percent of the cross-sectonal varance n proportonal quoted half-spread. Followng Stoll (000), we use the stock s closng prce (CLP), daly dollar volume (DOLVOL), return standard devaton (RETSTD), number of trades (N), and market value (MKV) as control varables of the effectve spread (ESP) and adverse nformaton component (INF). Return standard devaton s calculated as the measure of return volatlty by usng the return data of each frm durng September 16 to October 15. Our control varables for effectve spread are defned respectvely as follows: 11

CLP = the average closng prce of stock durng our samplng perod, DOLVOL = the average daly dollar volume of stock durng our samplng perod. N = the average daly number of trades of stock durng our samplng perod RETSTD = the return standard devaton of stock durng Sep.16 to Oct.15, MKV = the frm s average market value durng our samplng perod. Accordng to prevous studes such as Welker (1995), Stoll (000), and Agrawal et al (004), we predct that the stocks lsted n NYSE wth large return volatlty wll have wder effectve bd-ask spreads and adverse nformaton components, and the stocks havng larger daly dollar volume, number of trades, and market value wll have smaller effectve spreads and adverse nformaton components. Futhermore, n the studes of Welker (1995), and Stoll (000), they use relatve spread whch s defned as the quoted spread dvded by quote mdpont or closng prce to be a measure of market lqudty, and fnd a sgnfcant negatve relaton between closng prce and relatve spread. Ths s not surprsng because the closng prce s the denomnator of relatve spread. Oppostely, we use the dollar value of effectve spread and adverse nformaton components as explaned varables n our model, so the negatve relaton between closng prce and our measure of lqudty, as well as prevous studes, may not exst. We expect that there should be postve relaton between closng prce and our two measures of lqudty Both T&D fnal rankngs on annual bass and composte bass are used n our emprcal model to see whch rankngs are better n representng frms nformaton asymmetrc rsk perceved by the market. We use the stock s closng prce (CLP), daly dollar volume (DOLVOL), number of trades (N), return volatlty (RETSTD), and market value (MKV) as control varables of effectve spread (ESP) and adverse nformaton component (INF) n our emprcal models n order to analyss the partal effect of T&D fnal rankngs. All varables except RETSTD are taken log value n order to smooth the data and to get elastcty coeffcents from the estmated equaton. Thus, our cross-sectonal emprcal regresson models are set up as follows: ln ESP = α 1 + α + α 7 ln CFR ln MKV + ε + α 3 ln CLP + α 4 ln DOLVOL + α 5 ln N + α 6 RETSTD (4), 1

ln INF = β 1 + β + β 7 ln CFR ln MKV + η + β 3 ln CLP + β 4 ln DOLVOL + β ln N 5 + β 6 RETSTD (5), ln ESP = γ 1 + γ + γ 7 ln ln MKV AFR + ν + γ 3 ln CLP + γ 4 ln DOLVOL + γ 5 ln N + γ 6 RETSTD (6), ln INF = δ 1 + δ + δ 7 ln AFR + δ ln MKV + u 3 lnclp + δ 4 ln DOLVOL + δ 5 ln N + δ 6 RETSTD (7), where CFR = composte bass T&D fnal rankng for frm, T&D fnal rankng for frm, and ε, η, ν, and AFR = annual bass u are dsturbance terms 3..3 Hausman test for endogenety Before we use OLS to estmate equaton (4) to equaton (7), we have to confrm that the proxy for dsclosure practce, T&D rankng, s not an endogenous varable n each regresson model. Thus, we use Hausman test to examne f T&D rankng s endogenous n each model. The procedure of Hausman test for endogenety has three steps. Frst, we must determne the possble endogenous explanatory varable n our model and sutable exogenous nstrument varables. As earler motoned, the possble endogenous explanatory varable n our model s T&D rankng. The problem s how to decde sutable exogenous nstrument varables. In general, we can pck the exogenous determnants of the possble endogenous explanatory varable and the other exogenous explanatory varables n orgnal model as sutable nstrument varables. Followng Lang and Lundholm (1993) and Welker (1995), the determnants of dsclosure practce used n our study are closng prce (CLP), return (RET), market value (MKV), and return volatlty (RETSTD). The exogenous explanatory varables except for T&D rankng n orgnal model are closng prce (CLP), daly dollar volume (DOLVOL), number of trades (N), return volatlty (RETSTD), and market value (MKV). Therefore, the approprate nstrument varables are CLP, DOLVOL, N, RETSTD, MKV, and RET. The next step of Hausman test s to perform OLS for T&D rankng on all nstrument varables. Thus, the frst-stage regresson model s set up as follows: 13

ln Rank = θ + θ + θ ln CLP + θ 1 3 4 5 (8), 6 ln MKV + θ 7 ln DOLVOL ln RET + ϕ + θ ln N + θ RETSTD where Rank can be CFR or AFR, and ϕ s the dsturbance term of the regresson e model. After gettng the estmated equaton, we can calculate the resdual seres { } of Eq. (8). Fnally, add the resduals as a new explanatory varable to orgnal regresson model and run OLS to estmate and test the parameter of ths new varable. For example, the second-stage regresson model of ESP on CFR and other control varables s then as follows: ln ESP = α 1 + α + α 7 ln CFR ln MKV + ρ e + α 1 3 ln CLP + α + ε 4 ln DOLVOL + α 5 ln N + α 6 RETSTD (9), If the parameter, ρ 1, s sgnfcant dfferent from zero, then the composte bass T&D fnal rankng s endogenous n our regresson model, and we have to use SLS procedure to remove the estmated bas nduced by endogenety of explanatory varable. If the result of Hausman test ndcates there s no smultanety n the determnaton of market lqudty and frm s dsclosure polcy, an OLS procedure for Eq.(4) to Eq. (7) s approprate to examne the relaton of frm s market lqudty and ts T&D rankng. 4. Emprcal results 4.1 Sample characterstcs Panel A of Table 1 shows the descrptve statstcs of all selected varables durng our sample perod (October 17 00 December 31 00, 5 tradng days). The mean of effected spread (ESP) s about 1.305 cents, and ts range s about.47 cents. The mean of dollar adverse nformaton component (INF) s around 0.737 cents, and s about 56% of the effected spread. The range of adverse nformaton component s between 0.7 cents and 1.686 cents. The average closng prce (CLP) for our sample s approxmately $34.64, and ranges are between $6.07 and $16.4. The mean dollar volume (DOLVOL) s around $6.30 mllon and the sample range s 14

between $.16 mllon and $507.69 mllon. The mean daly number of trades (N) for the sample s 1549, and the sample range s from 48 to 5573. The return volatlty (RETSTD) has the average value about 0.0378, and the sample range s from 0.01036 to 0.1554. The average market value s approxmately $16767 mllon and ts range s between $543.76 mllon and $54441 mllon. The returns (RET) for our sample have the mean around 0.14 %, and the range s about 3.81%. The medan of S&P T&D rankngs on composte bass (CFR) s 7.55, and the range s between 7 and 9. The medan of S&P T&D rankngs on annual bass (AFR) s 4.78, and the range s between 1 and 8. Takng notce of the dfference between these two rankngs, the annual bass rankngs have lower medan but larger range whle the composte bass rankngs have hgher medan but smaller range. Ths characterstc s consstent wth the argument of Pantel and Dallas (00). They suggest that the annual bass rankngs whch only focus on frms annual reports could be vewed as frms voluntary dsclosures. On the contrary, the composte bass rankngs whch nclude annual reports, 10-Ks, and other proxy statements mght be regarded as regulatory dsclosure practces. Thus, due to strct laws of nvestor protecton and severe dsclosure regulatons n U.S., the frms reveal consstently hgher rankngs on composte bass, and there are smaller dfferences between frms composte bass rankngs than ther annual bass rankngs. Panel B of Table1 presents Pearson correlaton coeffcents between selected varables. 4. Regresson results of effectve spread aganst control varables. We frst examne that durng our sample perod f the effectve spreads of the sample frms are related to the determnants of spreads found n earler studes, such as Stoll (1978), Welker (1995), Stoll (000), and Argrawal at el (00). Panel A of Table presents the regresson result of effectve spread aganst these determnants. All explanatory varables except the daly number of trades (N) are sgnfcant at the 0.05 confdence level. The p-value of the daly number of trades (N) s 0.0973. The drectons of the parameters except average daly dollar volume (DOLVOL) are consstent wth prevous studes that we mentoned above. The average daly dollar volume (DOLVOL) reveals a sgnfcant postve relaton wth effectve spread durng our samplng perod. Notce that the result shows that about 85 percent of cross-sectonal varaton can be explaned by these control varables, ndcatng that there s not much space for other varables to explan the effectve spreads. 4.3 Regresson results of adverse nformaton component aganst control varables. 15

Panel B of Table presents the regresson result of adverse nformaton component aganst these control varables. Smlar to the result of effectve spread, all explanatory varables except the number of trades (N) are sgnfcant at the 0.05 confdence level. The p-value of the number of trades (N) s 0.1869. The drectons of the parameters except average daly dollar volume (DOLVOL) are consstent wth prevous studes that we mentoned above. The average daly dollar volume (DOLVOL) also reveals a sgnfcant postve relaton wth adverse nformaton component durng our samplng perod. The adjusted R-square of the model s 0.8643 whch s slghtly hgher than the model of effectve spread. 4.4 Results of Hausman test To detect whether the T&D rankngs are endogenous varables n Eq. (4) to (7), we employ Hausman test to test the endogenety. The results of Hausman test for annual bass rankngs and composte bass rankngs are respectvely shown n Panel A to D of Table 3. Panel A present that the coeffcent ρ 1 n Eq. (9) for composte bass rankngs s not statstcally sgnfcant, ndcatng that composte bass rankng s not a endogenous varable n Eq. (1). The smlar results are found n the regresson of adverse nformaton component on composte bass rankng, the regresson of effectve spread on annual bass rankng, and the regresson of adverse nformaton component on annual bass rankng. These results are shown n Panel B to Panel D of Table 3 where the coeffcents, ρ, ρ 3, and ρ 4 are all nsgnfcant n the models. Accordng to these results, the smultanety n the determnaton of spread and frm s dsclosure polcy does not exst for our sample frms durng ths studyng perod. Thus, the OLS procedure s approprate for Eq.(4) to Eq.(7) to estmate and test the parameters of our nterests. 4.5 Regresson result of the effectve spread and adverse nformaton component on composte bass T&D rankng The OLS procedures for effectve spread and adverse nformaton component on composte bass T&D rankngs and other control varables s approprate because composte bass T&D rankngs do not reveal sgnfcant endogenety n Eq.(4) and Eq.(5). Thus we smply regress the effectve spread and adverse nformaton component respectvely on composte bass T&D rankng and other control varables. The results of estmatng and testng are shown n the Panel A and Panel B of Table 4. Panel A of Table 4 shows the regresson result of effectve spread on composte 16

bass T&D rankng. All varables except average daly dollar volume (DOLVOL) are n the drecton that we predcted, and all coeffcents except average daly number of trades (N) are sgnfcant at ordnary confdence level. Although the p-value of average daly number of trades s 0.0575, t s slghtly bgger than ordnary confdence level, 0.05. The estmated parameter of composte bass T&D rankng s -0.3070 and statstcally sgnfcant from zero. Ths result mples that frms wth hgher composte bass T&D rankngs wll have less effectve spreads, and thus have better market lqudty. Ths result s consstent wth our predcton. Panel B of Table 4 shows the regresson result of adverse nformaton component on composte bass T&D rankng. Lke the result of effectve spread, all varables except average daly dollar volume (DOLVOL) are n the drecton that we predcted, and all coeffcents except average daly number of trades are sgnfcant at ordnary confdence level. The estmated parameter of composte bass T&D rankng s -0.744 and s statstcally sgnfcant from zero. Consstent wth our expectaton ths result mples that frms wth hgher composte bass T&D rankngs wll have less nformaton asymmetrc problems, and thus have less adverse nformaton component. The results mentoned above also suggest that composte bass T&D rankngs are good proxy for frms dsclosure practces and nformaton asymmetrc rsk perceved by the market. Our emprcal results have some mportant meanng for corporate governance: the managers should endeavor to conform varous dsclosure regulatons and nvestor protecton codes by dsclosng frm s nformaton to the best of ther abltes. When a frm can provde better dsclosure and transparency, t wll get a hgher level composte bass T&D rankng, and ths wll lower frm s nformaton asymmetrc rsk perceved by market. Consequently, the frm wll has smaller adverse nformaton component and effectve spread, and therefore ncreasng market lqudty of ts stock. 4.6 Regresson result of the effectve spread and adverse nformaton component on annual bass T&D rankng The OLS procedures for effectve spread and adverse nformaton component on annual bass T&D rankng and other control varables s also approprate because annual bass T&D rankngs do not reveal sgnfcant endogenety n Eq.(6) and Eq.(7). Therefore we smply regress the effectve spread and adverse nformaton component respectvely on annual bass T&D rankng and other control varables. The results of estmatng and testng are shown n the Panel A and Panel B of Table 5. 17

Panel A of Table 5 shows the regresson result of effectve spread on annual bass T&D rankng. All control varables except average daly dollar volume (DOLVOL) are n the drecton that we predcted, and all of ther coeffcents except average daly number of trades (N) are sgnfcant at ordnary confdence level. The estmated coeffcent of annual bass T&D rankng s n the drecton that we predcted, but t reveals statstcally nsgnfcant. Smlar to the result for effectve spread on annual bass T&D rankng, the regresson result of adverse nformaton component on annual bass T&D rankng shows that all control varables except average daly dollar volume (DOLVOL) are n the drecton that we predcted, and the coeffcents except average daly number of trades (N) are sgnfcant at ordnary confdence level. Unfortunately the estmated coeffcent of annual bass T&D rankng s not only n the opposte drecton, but also reveals statstcally nsgnfcant. The result s shown n the Panel B of Table 5. Our emprcal results suggest that the annual bass T&D rankng s not a good explanatory varable of effectve spread and adverse nformaton component. We argue that there are least two possble reasons for ths suggeston. Frst, the dsclosure regulatons and nvestors protecton n U.S. are qute well. Even though the nvestors can t easly get the regulatory dsclosure documents such as 10-Ks, and other proxy statements, they may beleves that although the frm s annual report, whch represents frm s voluntary dsclosure, does not dsclose enough nformaton, these regulatory dsclosure documents wll make sure that the frm has done suffcent dsclosure practce. Second, when the market makers or specalsts quote the spreads of these frms, they mght use not only annual reports but also the regulatory dsclosure documents such as 10-Ks, and other proxy statements as reference materal for frm s dsclosure practces and correspondng nformaton asymmetrc rsks. So t s reasonable that our result reveal that the composte bass T&D rankng s a better explanatory varable of market lqudty than the annual bass T&D rankng. 5. Robustness check In our regresson analyss, there are two econometrc ssues whch need to be further explored. Frst of all, the estmated coeffcent of average daly dollar volume (DOLVOL) n each model s statstcally sgnfcant but reveals contrary drecton to our predcton. Ths result s nconsstent wth the fndngs documented n prevous lteratures. One possblty to cause ths result s that hghly multcollnearty mght exst amo ng the ndenpendent varables used n our models. To elmnate ths 18

possble problem, we calculate the Varance Inflaton Factors (VIFs) of all varables to measure the nflaton n the varances of the parameter estmates due to collneartes that exst among the ndependent varables. Panel A of Table 6 reports the varance nflaton factors (VIFs) and estmated parameters for all explanatory varables of our orgnal emprcal models. The VIFs of the average daly dollar volume (DOLVOL) and the average daly number of trades (N) are extremely hgher than the ordnary tolerance value, 10. For ths reason, we consder that both the average daly dollar volume and the average daly number of trades mght be hghly lnear dependent wth other ndependent varables. Thus, we exclude these two control varables from all of our models, and estmate these new models by OLS agan. We also calculate the VIFs of the remander explanatory varables for comparson wth orgnal models. These results are reported n Panel B of Table 6. We can see that after excludng average daly dollar volume and average daly number of trades from our orgnal regresson model the coeffcent estmates of remander explanatory varables are stll statstcally sgnfcant and the values of these estmates do not change a lot. More mportantly, all estmated coeffcents stll reveal the consstent sgn wth our predctons, and the adjusted R-square of each regresson model just decreases slghtly. The VIFs of the explanatory varables n these new regresson models, comparng to orgnal models, also decrease sgnfcantly, especally for average daly market value. These results seem to suggest that the three control varables, average daly closng prce (CLP), return standard devaton (RETSTD), and average daly market value (MKV), already have qute enough explanatory power and do not have the problem of multcollnearty for our sample frm durng the studyng perod. The next econometrc ssue s that there mght be heteroskedastcty problem n our regresson analyss. To correct for heteroskedastcty, we calculate the Whte s heteroskedastcty-consstent standard errors (HCSEs) for each coeffcent estmate, and use these robust standard errors to test whether coeffcent estmates are statstcally sgnfcant. The estmated results are reported n Table 7 and Table 8. Table 7 report the heteroskedastcty-robust regresson results of the effectve spread and adverse nformaton component on composte bass T&D rankngs whle Table 8 report the annual bass one. In these regresson analyses, we also exclude the average daly dollar volume (DOLVOL) and the average daly number of trades (N) from the regresson models. We fnd that the heteroskedastcty-robust standard errors do not have very bg dfferences between OLS standard errors and the t-value and p-value calculated by heteroskedastcty-robust standard errors do not change a lot ether. Therefore, we argue that the heteroskedastcty problem s slght n our emprcal models. All coeffcent estmates under heteroskedastcty-robust test are statstcally 19

sgnfcant and thus have the same fnancal mplcatons that we dscussed n secton 4.5 and 4.6. 6. Concluson Ths paper nvestgates the qualty and accuracy of the S&P T&D rankngs by testng f they have sgnfcant relaton to effectve spreads and adverse nformaton components of frms stocks. Besdes, we also examne whether frms dsclosure practces can affect ther market lqudty. In order to ncrease the robustness of our cross-sectonal emprcal analyss, we also ncorporate several determnants of bd-ask spreads suggested by prevous studes as explanatory varables n our models. Moreover, we consder about the smultanety that mght exst n the determnaton of the spread and frm s dsclosure polcy. We employ Hausmant test to detect whether the S&P T&D rankngs n our model reveal endogenety. The results show that composte and annual bass T&D rankngs do not have sgnfcant endogenety, and thus the OLS procedure s approprate to proceed our estmates and tests. Our emprcal results suggest that the annual bass T&D rankng s not a good explanatory varable of effectve spread and adverse nformaton component. Under OLS procedure, the coeffcent of annual bass T&D rankng s not statstcally sgnfcant n the model of effectve spread and n the model of adverse nformaton component. The results of OLS procedure for effectve spread and adverse nformaton component on composte bass T&D rankngs and other control varables are better. In each regresson model, the coeffcent of composte bass T&D rankng s n the same drecton that we predct, and reveal statstcally sgnfcant at ordnary confdence level n our study. These results mply that the frms wth hgher composte bass T&D rankngs wll have less effectve spread and adverse nformaton component, and thus have better market lqudty of ther equtes. We argue that the composte bass rankng s a better explanatory varable for frm s market lqudty perceved by the market, and therefore a better proxy for frm s dsclosure practce. There are two reasons for our suggeston. Frst, the dsclosure regulatons and nvestors protecton n U.S. are qute well. Even though the nvestors can t easly get the regulatory dsclosure documents such as 10-Ks, and other proxy statements, they may beleves that although the frm s annual report, whch represents frm s voluntary dsclosure, does not dsclose enough nformaton, these regulatory dsclosure documents wll make sure that the frm has done suffcent dsclosure practce. Second, when the market makers or specalsts quote the spreads 0

of these frms, they mght use not only annual reports but also the regulatory dsclosure documents such as 10-Ks, and other proxy statements as reference materal for frm s dsclosure practces and correspondng nformaton asymmetrc rsks. Therefore, we consder that the annual bass T&D rankngs have less predcton power n explanng frms market lqudty than composte bass T&D rankngs. The results of our study have some mportant meanng for corporate governance: the managers should endeavor to conform to varous dsclosure regulatons and nvestor protecton codes by dsclosng frm s nformaton to the best of ther abltes. When a frm can provde better dsclosure and transparency, t wll get a hgher level composte bass T&D rankng, and ths wll lower frm s nformaton asymmetrc rsk perceved by market. Consequently, the frm wll has smaller adverse nformaton component and effectve spread, and therefore ncreasng market lqudty of ts stock. We have to note that our study only provdes an ndrect way to examne whether the composte and annual bass T&D rankngs are good proxes for frms corporate governance by testng f they are sgnfcant related to frms market lqudty. Although the annual bass T&D rankng can not be a good explanatory varable for frm s market lqudty perceved by the market n our study, t stll mght be a good proxy for frm s corporate governance. Further researches are needed to examne the extent of the annual bass T&D rankng n measurng frm s corporate governance. 1

References Agrawal, V., Kothare, M., Rao, R. K. S., and Wadhwa, P., 004, Bd-ask spreads, nformed nvestors, and frm s fnancal condton, The Quarterly Revew of Economcs and Fnance 44, 58 76. Barclay, M. J., Chrste, W. G., Harrs, J. H., Kandel, E., and Schultz, P. H. 1999, Effects of market reform on the tradng costs and depths of Nasdaq stocks, Journal of Fnance 54, 1-34. Botosan, C. A., 1997, Dsclosure level and rs, and the the cost of equty captal, the Accountng Revew 7:3, 33-350. Cheng, A. C. S., Collns, D., Huang, H., 003, The effect of the S&P T&D rankngs on market beta, abnormal returns and earnngs response coeffcents n the perod surroundng the report release date, workng paper. Copeland, T. E., and Gala, D., 1983, Informaton effects of the bd-ask spread, Journal of Fnance 38, 1457-1469. Demsetz, H., 1968, The cost of transactng, Quarterly Journal of Economcs 8, 33-53. Durnev, A., and Km, E., 00, To steal or not to steal: Frm attrbutes, legal envronment, and valuaton, Workng Paper, Unv. of Mchgan Busness School. Dye, R. A., 1985, Dsclosure of nonpropretary nformaton, Journal of Accountng Research, 13-145. Glosten, L. R., and Mlgrom, P. R., 1985, Bd, ask and transacton prces n a specalst market wth heterogeneously nformed traders, Journal of Fnancal Economcs 14, 71-100. Huang, R. D., and Stoll, H. R., 1994, Market mcrostructure and stock return predctons, Revew of Fnancal Studes 7, 179-13. Huang, R. D., and Stoll, H. R., 1996, Dealer versus aucton markets: A pared comparson of executon costs on NASDAQ and the NYSE, Journal of Fnancal Economcs 41, 313-357. Kyle, A. S., 1985, Contnuous auctons and nsder tradng, Econometrca 53, 1315-1335. Lang, M. and Lundholm, R., 1993, Cross-sectonal determnants of analyst ratngs of corporate dsclosures, Journal of Accountng Research, 46-71. Lang, M. and Lundholm, R., 1999, Corporate dsclosure polcy and analyst behavor, The Accountng Revew 71, 467-493. Ln, J.-C., 199, Order persstence, adverse selecton, and gross profts earned by NYSE specalsts, Journal of Fnance 48, 1108. Ln, J.-C., Sanger, G.C., and Booth, G, 1995, Trade sze and components of the bd-ask spread, The Revew of Fnancal Studes 8,1153-1183.

Lppman, S. A., and McCall, J. J., 1986, An operatonal measure of lqudty, Amercan Economc Revew 76, 43-55. Lowensten, L., 1996, Fnancal transparency and corporate governance, Columba Law Revew, Vol 96, Aprl. Patel, S. A., and Dallas, G., 00, Transparency and Dsclosure: Overvew of Methodology and Study Results Unted States, http://governance.standardandpoors.com. Stoll, H. R., 1978a, The supply of dealer servces n securtes markets, Journal of Fnance 33, 11-1151. Stoll, H. R., 1978b, The prcng of securty dealer servces: An emprcal study of NASDAQ stocks, Journal of Fnance 33, 1153-117. Stoll, H. R., 000, Frcton, Journal of Fnance 55, 1479-1514. Welker, M., 1995, Dsclosure polcy, nformaton asymmetry, and lqudty n equty markets, Contemporary Accountng Research 11, 801-87. TABLE 1 3