Copyright Undertaking

Copyrigh Underaking This hesis is proeced by copyrigh, wih all righs reserved. By reading and using he hesis, he reader undersands and agrees o he following erms: 1. The reader will abide by he rules and legal ordinances governing copyrigh regarding he use of he hesis. 2. The reader will use he hesis for he purpose of research or privae sudy only and no for disribuion or furher reproducion or any oher purpose. 3. The reader agrees o indemnify and hold he Universiy harmless from and agains any loss, damage, cos, liabiliy or expenses arising from copyrigh infringemen or unauhorized usage. IMPORTANT If you have reasons o believe ha any maerials in his hesis are deemed no suiable o be disribued in his form, or a copyrigh owner having difficuly wih he maerial being included in our daabase, please conac lbsys@polyu.edu.hk providing deails. The Library will look ino your claim and consider aking remedial acion upon receip of he wrien requess. Pao Yue-kong Library, The Hong Kong Polyechnic Universiy, Hung Hom, Kowloon, Hong Kong hp://www.lib.polyu.edu.hk

ESSAYS ON FINANCIAL REPORTING QUALITY: EVIDENCES FROM SEASONED EQUITY OFFERING AND PRODUCT MARKET COMPETITION WANG YUEQUAN Ph.D The Hong Kong Polyechnic Universiy 2012

The Hong Kong Polyechnic Universiy School of Accouning and Finance Essays on Financial Reporing Qualiy: Evidences from Seasoned Equiy Offering and Produc Marke Compeiion Wang Yuequan A hesis submied in parial fulfillmen of he requiremens for he degree of Docor of Philosophy Sepember 2011 2

CERTIFICATE OF ORIGINALITY I hereby declare ha his hesis is my own work and ha, o he bes of my knowledge and belief, i reproduces no maerial previously published or wrien, nor maerial ha has been acceped for he award of any oher degree or diploma, excep where due acknowledgemen has been made in he ex. ----------------------------- (Signed) Yuequan Wang (Name of Suden) 3

ABSTRACT This disseraion focuses on financial reporing qualiy. I is comprised of hree essays. The firs essay documens he imporance of financial reporing qualiy; he second essay records marke power as an imporan deerminan of financial reporing qualiy; he hird essay shows ha financial reporing qualiy is no he prevailing channel hrough which produc marke compeiion affecs audi fees. Essay I, Earnings imeliness and seasoned equiy offering announcemen effec demonsraes he imporance of financial reporing qualiy by examining he effecs of earnings imeliness on he Seasoned Equiy Offering (SEO) announcemen effec. Invesors view an SEO announcemen as a negaive signal ha reveals managers percepions regarding a firm s curren sock price. Invesors usually respond o his negaive signal by reducing he sock price significanly. This condiion can be miigaed, however, hrough a descripion of a firm s abiliy o capure curren value-relevan informaion hrough a measure of financial reporing qualiy, namely earnings imeliness. This is especially rue since earnings are imporan o invesors in assessing firm performance. Presening curren valuerelevan informaion wih earnings in a greaer efficien and imely way can reduce informaion asymmery beween managers and invesors. I predic and find, hen, ha firms wih greaer earnings imeliness have less negaive SEO announcemenperiod reurns. 4

Because of he imporance of financial reporing qualiy in capial marke, I explore he deerminans o financial reporing qualiy in my second essay, Marke power and accrual managemen. I examine wheher a firm s compeiion saus in produc markes affecs is financial reporing qualiy, measured as discreionary accrual. I argue ha because firms wih greaer marke power have a greaer abiliy o se prices for heir producs, hey have comparaively fewer incenives o manipulae earnings hrough accrual managemen. I use he Lerner index o measure produc marke power and asse-deflaed absolue discreionary accruals o proxy he magniude of accrual managemen. Using a large sample of firm-year observaions from 1997 o 2007, I find ha, as hypohesized, firms wih greaer marke power end o have lower levels of accrual managemen. The final essay, Produc marke compeiion and audi fees, goes one-sep furher han he second. As noed in he second essay, produc marke compeiion affecs a firm s financial reporing qualiy. However, financial reporing qualiy may no be he only facor audiors ake ino accoun when hey decide wha fees o charge a clien. The las essay, herefore, empirically explores he iner- and inra- indusry effec of produc marke compeiion on audi fees. Prior lieraure posis wo conradicory predicions on he relaion beween produc marke compeiion and audi fees. On he one hand, firms in a compeiive marke are expeced o face higher liquidiy risk, disress risk, and liquidaion risk, hus increasing audiors assessmens of a clien s business risk. So, audi fees are expeced o increase wih indusry compeiiveness. On he oher hand, i is ofen argued in prior lieraure ha produc marke compeiion decreases informaion 5

asymmery and miigaes agency problems beween shareholders and managers and increases he accuracy of financial reporing, hus decreasing audiors assessmens of a clien s audi risk resuling in necessary audis. So audiors end o charge lower fees on firms in a more compeiive indusry. The sudy, hen, empirically ess he relaion beween produc marke compeiion and audi fees and finds ha audiors charge higher fees on firms in a more compeiive indusry. I also finds ha audiors charge lower fees on firms wih greaer marke power wihin he same indusry. 6

ACKNOWLEDGEMENTS I would like o hank he admission commiee in School of Accouning and Finance a he Hong Kong Polyechnic Universiy. Their decision gave me a chance o move o Hong Kong. I had a pleasan ime during my wo years Ph.D. sudies in Hong Kong (I sared Ph.D. sudies in Sepember 2009, compleed Ph.D. hesis in May 2011, passed he inernal defense in Augus 2011, and passed he exernal defense wihou revision requiremens in December 2011.). I would like o express my graiude o my advisor, Professor Andy Chu for his paience, encouragemen and guidance. Professor Chui no only supervises me in research, bu also ses up an excellen example of how a decen person behaves. Alhough I sayed in he Ph.D. program for a comparaively shor period, his personaliy, working ehics, and commimen o academia and sudens moivae me a lo. I is one of he mos hankful hings in my life o have had Professor Andy Chui as my Ph.D. advisor. I am indebed o him more han he knows. I look forward o working wih my advisor in fuure. I would also acknowledge Professor Simon Fung, Nancy Su, and Seven Wei for heir helpful discussions wih me. 7

TABLE OF CONTENS ABSTRACT (p.4) ACKNOWLEDGEMENTS (p.7) CHAPTER 1. INTRODUCTION (p.9) 2. ESSAY 1: EARNINGS TIMELINESS AND SEASONED EQUIY OFFERING ANNOUNCEMENT EFFECT (p13) 3. ESSAY 2: MARKET POWER AND ACCRUAL MANAGEMENT (p48) 4. ESSAY 3: PRODUCT MARKET COMPETITION AND AUDIT FEES (p93) 5. CONCLUSIONS AND FUTURE WORK (p131) 8

Chaper 1 Inroducion This disseraion includes hree essays. The firs essay, Earnings imeliness and seasoned equiy offering announcemen effec, documens he imporance of financial reporing qualiy by showing he significan negaive relaion beween he Seasoned Equiy Offering (SEO) announcemen effec and earnings imeliness. I predic and es he hypohesis ha firms wih greaer earnings imeliness have less negaive SEO announcemen-period reurns in he firs essay. My hypohesis builds on he heory ha he sock price drop a an equiy issue announcemen is caused by informaion asymmery beween managers and invesors, and ha firms wih greaer earnings imeliness end o have less informaion asymmery. I regress he SEO announcemen-period reurn on earnings imeliness by using a sample of SEO evens from 1984 o 2006 and find ha he firms wih greaer earnings imeliness experience less negaive SEO announcemen reurns. In an addiional es, I also explore wheher he impac of earnings imeliness on he SEO announcemen effec would be subsumed by oher earnings aribues. I hen reexamine he relaion beween SEO announcemen effec and earnings imeliness wih oher earnings aribues in he regression. I find ha he significan influence of earnings imeliness on he SEO announcemen effec sill holds when all oher earnings aribues are considered. This suggess ha earnings imeliness has is unique and disinguishable impac on he sock reurn a he ime of he SEO announcemen. 9

Given he imporance of financial reporing qualiy in capial markes, he second essay, Marke power and accrual managemen, explores how financial reporing qualiy is deermined. Specifically, I examine wheher managers adjus heir accrual managemen policies based on heir produc marke power. I argue ha advanageous compeiion saus in a produc marke offers firms an alernaive way o increase or reduce earnings so ha managers in such firms have less incenive o manage earnings via accrual managemen. I use he Lerner index or price-cos margin (PCM) as he measure of marke power. Following Gaspar and Massa (2005) and Peress (2010), PCM is calculaed as he raio of operaing profi o sales. I use a modified cross-secional Jones model in Dechow, Sloan, and Sweeney (1995) o calculae he level of asse-deflaed discreionary accruals, a proxy of accrual managemen. Using a sample of 35,745 firm-year observaions from 1997-2007, I run a fixed-effec regression on he panel daa and find ha even conrolling facors ha describe a firm s operaing environmen, growh opporuniies, profiabiliy and regulaion environmen, here is a significan negaive relaion beween produc marke power and accrual managemen. Essay II documens he significan effec of produc marke compeiion on a firm s financial reporing qualiy. However, oher facors are also aken ino accoun in addiion o financial reporing qualiy when audiors decide wha fees o charge cliens. In oher words, financial reporing qualiy may no be he only channel hrough which produc marke compeiion affecs audi fees. The hird essay empirically examines he issue described above and explores he iner- and inra-indusry effec of produc marke compeiion on audi fees. Prior lieraure 10

shows ha audiors charge firms higher fees eiher because of cliens greaer audi risk or because of cliens greaer business risk. Audi risk describes he likelihood of maerial errors in cliens financial saemens, while business risk refers o circumsances ha are ou of audiors conrol and canno be eliminaed. Furher, I describe how indusry level produc marke compeiion affecs audi risk and business risk in wo differen direcions: on he one hand, compeiion plays a governance role and miigaes agency problems. Srenghened governance and miigaed agency problem hrough compeiion improve he accuracy of financial reporing and also reduce invesors demand on audi services. So, audi risks end o be less for firms in an indusry wih greaer compeiion inensiy. In oher words, here is a negaive relaion beween audi risk and indusry-level compeiion. On he oher hand, here should be a posiive relaion beween business risk and indusry-level compeiion because firms in compeiive indusries involve more business risks han hose in less compeiive indusries. The laer argumen is based on he prior lieraure ha documens more operaion risk, innovaion risk, liquidaion risk, and liigaion risk for firms in a more compeiive indusry. Therefore, I leave he iner-indusry effec of produc marke compeiion on audi fees as an empirical quesion. The inra-indusry effec of compeiion on audi fees looks ino he compeiion saus, i.e., produc marke power, on he audi fees wihin an indusry. The second essay ells ha firms wih greaer marke power end o have less accrual managemen. So, I argue ha firms having advanageous compeiion saus have less audi risk. Also, such firms end o have more sable cash flows 11

and earnings. They also have less disress risk and liquidaion risk. Thus, I hypohesize ha firms wih beer compeiion saus also have less business risk. Boh audi and business risk channels predic a negaive relaion beween marke power and audi risk. Using he Herfindahl-Hirschman index as a measure of indusry-level compeiion inensiy and PCM as a measure of firm-level compeiion saus, I empirically examine he relaion beween hese and audi fees, respecively in he U.S. manufacuring indusry. I find ha audiors charge higher fees in a more compeiive indusry, and hey charge lower fees on firms wih greaer marke power wihin an indusry. 12

Chaper 2 Essay I: Earnings Timeliness and Seasoned Equiy Offering Announcemen Effec 2.1 Inroducion This essay examines he relaion beween he Seasoned Equiy Offering (SEO) announcemen effec and earnings imeliness. I predic and es he hypohesis ha firms wih greaer earnings imeliness have less negaive SEO announcemen-period reurns. My hypohesis builds on he heory ha he sock price drop a an equiy issue announcemen is caused by informaion asymmery beween managers and invesors (Myers and Majluf 1984) and on he empirical sudies abou earnings imeliness by Bushman e al. (2004) and Ball e al. (2008). In he world of informaion asymmery, raional firm managers will no issue new socks when prices are low relaive o managers privae informaion abou firm value. Knowing his, invesors view an SEO announcemen as a negaive signal ha reveals managers percepions on a firm s curren sock price. Invesors respond o his negaive signal by reducing he sock price significanly. Measured as he adjused 2 R from a regression of annual earnings on conemporaneous sock reurns, earnings imeliness describes he abiliy of earnings numbers o capure curren value-relevan informaion. Because earnings 13

are imporan o invesors in assessing firm performance and earnings wih greaer imeliness capure firms informaion in a more efficien way, greaer earnings imeliness can reduce informaion asymmery beween managers and invesors. The above analysis suggess ha earnings wih greaer imeliness can reduce informaion asymmery beween managers and invesors and ha less informaion asymmery implies a less negaive SEO announcemen effec. Thus, I hypohesize ha he marke responds less negaively o SEO announcemens from firms wih greaer earnings imeliness. I es he above hypoheses on a sample of SEO evens from 1984 o 2006, requiring ha he sample firms have enough ime-series daa o compue a firmspecific measure of earnings imeliness. In order o calculae he announcemenperiod reurn, I also require ha he sample firms have CRSP daily sock reurns during he SEO announcemen period. Finally, I regress he SEO announcemenperiod reurn on he earnings imeliness and find ha he firms wih greaer earnings imeliness experience less negaive SEO announcemen reurn. In an addiional es, I explore wheher he impac of earnings imeliness on he SEO announcemen effec would be subsumed by oher earnings aribues. I reexamine he relaion beween he SEO announcemen effec and earnings imeliness wih oher earnings aribues in he regression. As in Lee and Masulis (2009), I find ha accrual qualiy is negaively correlaed wih he magniude of sock price drop a he SEO announcemen. Bu he significan influence of earnings imeliness on he SEO announcemen effec sill holds when all oher 14

earnings aribues are considered. This suggess ha earnings imeliness has a unique and irreplaceable impac on he sock reurn a he SEO announcemen This sudy makes several conribuions. Firs, accouning researchers have long been ineresed in he causes and consequences of financial reporing qualiies. This sudy conribues o he lieraure abou he consequences of financial reporing qualiies by examining he effecs of earnings imeliness in a financing even. Thus, in a broader sense, aken ogeher wih oher sudies ha address he roles of financial reporing qualiies on invesmen efficiency, deb conracing efficiency or sock price synchroniciy (e.g., Biddle e al. 2009; Zhang 2008; Huon e al. 2009), my essay enriches he lieraure and fills a need by examining he effec of earnings imeliness in he financing even. Second, my paper provides empirical evidence on he relaion beween earnings aribues and firms informaion environmens. Francis e al. (2004) sudy he relaion beween earnings aribues and informaion risk. However, informaion risk is difficul o measure and he causal chain beween hese aribues and informaion risk involves many links and assumpions. This sudy proposes ha informaion asymmery can serve as one link beween earnings aribues and informaion risk because informaion asymmery increases invesors uncerainy regarding firms and informaion risk is, hus, posiively correlaed wih informaion asymmery. Sudying he link beween earnings aribues and measures of informaion asymmery can enhance our confidence ha hese qualiy measures are causally linked o characerisics of firms informaion environmens. 15

Third, his paper enriches he curren, limied lieraure abou earnings imeliness. By measuring he exen o which curren earnings numbers capure value-relevan informaion, earnings imeliness plays imporan roles in capial marke. However, only a few papers conduc research on earnings imeliness. Ball e al (2008) explore he deb conracing value of earnings imeliness. Bushman e al. (2004) invesigae how earnings imeliness affecs corporae governance facors such as board srucure. Francis e al. (2004) examine he relaion beween he cos of equiy capial and earnings aribues, including earnings imeliness. My paper is he firs o explore he impac of earnings imeliness on he ransacion coss of an equiy offering. Finally, his paper also conribues o he lieraure seeking o explain he cross-secional variaion in announcemen-period reurn and is among he firs o examine wheher earnings aribues can be used o proxy he informaion asymmery and describe he informaion environmen. The remainder of his paper is organized as follows: Secion 2.2 reviews prior research concerning he SEO announcemen effec; Secion 2.3 discusses he role of earnings imeliness in influencing informaion asymmery and develops he hypoheses; Secion 2.4 inroduces he sample and research design; Secion 2.5 presens empirical resuls; Secion 2.6 supplies a summary and conclusion. 2.2 Lieraure review 16

Seasoned Equiy Offerings (SEO) refer o he even during which a publicly raded firm issues addiional sock. The SEO is a kind of primary offering because he firm issues new shares and he proceeds go o he firm. This is as opposed o a secondary offering, during which corporae insiders and block-shareholders sell shares while he number of shares ousanding remains he same afer he offering. I is well-documened ha he announcemen of a common sock offering engenders a significan sock-price drop in he magniude of beween -2% and -3% (Masulis and Kowar 1986; Asquih and Mullins 1986; Mikkelson and Parch 1986). Such a negaive SEO announcemen effec reflecs he large ransacion coss of he new issues. Researchers sugges differen heories o explain his phenomenon. Leland and Pyle (1977) show ha, in markes wih asymmeric informaion, he equiy fracion in he projec reained by he self-ineresed enrepreneur has a posiive associaion wih a fuure projec s qualiy. Well-informed managers would only sell heir sock shares when hey believe ha he shares are overvalued. So, share sales by managers serve as a negaive signal abou a firm s inrinsic value. The Leland and Pyle signaling heory applies o pure primary offerings and o a combinaion of secondary offerings and primary offerings because secondary offerings decrease he insiders or block-shareholders shares. Myers and Majluf (1984) ake heir findings beyond hose of Leland and Pyle. In heir adverse selecion model, hey assume ha managers always work for he ineress of exising shareholders and will no issue socks when he firm is 17

undervalued, because doing so would dilue he fracional ownership of exising shareholders. Thus, even when managers do no sell heir own shareholdings, he mere ac of equiy offering conveys a negaive signal ha he curren sock price is oo high. Knowing his, raional invesors adjus heir valuaion of a firm and he sock price drops as a consequence. The Myers and Majluf adverse selecion model applies o all kinds of offerings: primary offerings, secondary offerings and a combinaion. Jung, Kim and Sulz (1996) propose a heory based on agency problems, claiming ha when managemen has misaligned ineress wih he shareholders, raional invesors respond o he equiy offering announcemen negaively because hey are afraid of poenial misuse of proceeds. A subsanial volume of lieraure also ries, wih carrying findings, o explain he cross-secional variaion in he SEO announcemen-period reurns. For example, some researchers examine wheher equiy characerisics conribue o cross-secional variaion and find mixed resuls in he relaion beween he relaive size of he offering and he subsequen drop. Asquih and Mullins (1986) documen ha announcemen-period reurn is negaively relaed o he relaive size of he issue, compued as he raio of he planned proceeds o a firm s equiy value before he announcemen. Dierkens (1991), however, does no find a significan relaion beween he price drop and he relaive size of he issue, measured as he raio of he number of new shares o he number of shares ousanding before he announcemen. Mikkelson and Parch (1986) also do no find a relaion beween he sock price effecs and he amoun of new financing or he size of offering. 18

Use of proceeds is anoher characerisic of an offering ha can explain cross-secional variaion in he announcemen reurns o some exen. Mikkelson and Parch (1986) documen a less negaive announcemen effec when i is saed ha he proceeds are o be used for capial expendiures, raher han for deb refinance. Asquih and Mullins (1986) examine wheher a firm s pre-issue performance can be a facor used o explain cross-secion variaion in an SEO announcemen reurn. They find ha such announcemen-period reurn is posiively relaed o he previous eleven-monh cumulaive excess reurn. Masulis and Korwar (1986) documen a negaive relaion of sock announcemen reurn o he previous womonh firm reurn and a posiive relaion of sock announcemen reurn o he previous wo-monh marke reurn. In erms of he iming of an announcemen, Choe, Masulis, and Nanda (1993) documen a less negaive SEO announcemen effec when he economy is in an expansionary period of he business cycle, which implies less adverse selecion risk. Dierkens (1991) documens a significanly posiive relaion beween he announcemen-period reurn and he firm s growh opporuniies, he raio of he marke value of he equiy o he book value of he equiy for one fiscal year before he announcemen. Lang and Lundholm (2000) find by examining firms behavior paerns ha issuing firms end o reduce he informaion asymmery by making opimisic disclosures more frequenly, saring six monhs before he regisraion dae, and 19

ha he announcemen-period reurn increases wih such changes in firms disclosure behavior. Korajczyk,Lucas, and McDonald (1991) repor ha he negaive announcemen effec is less pronounced wih a decrease in he ime difference beween he offering announcemen and he preceding earnings announcemen. They argue ha a decrease in he informaion asymmery resuling from he earnings announcemen reduces he magniude of he price drop a he offering announcemen. Wih regard o CEO compensaion srucure, Brazel and Webb (2006) documen ha when he proporion of CEO equiy-based compensaion is large, invesors end o view he equiy offering as a las-resor source of capial and respond o he SEO announcemen effec more negaively. In his paper, I examine wheher firms wih earnings imeliness of differen magniude experience differen price drops during he SEO announcemen period. My sudy will no only enrich he lieraure regarding he consequences of financial reporing qualiies, bu will also provide a poenial link, informaion asymmery, o he argued relaion beween earnings aribues and informaion risk. Of course, he paper will also conribue o lieraure ha explains he crosssecional variaion in he SEO announcemen-period reurn and lieraure abou earnings imeliness. 2.3 Hypohesis developmen 20

Earnings are imporan sources for invesors o assess firm performance. Measured as he adjused 2 R of he firm-specific regression of annual earnings on annual reurns (Equaion 1), earnings imeliness is one measure of financial reporing qualiy and one of he hree marke-based earnings aribues in Francis e al. (2004). E j, MKTCAP j, 1 b j,0 b j,1 NEG j, b j,2 RET j, b j,3 NEG j, RET j, j, (1) In Equaion 1, E j, is he earnings before exraordinary iems, disconinued operaions and special iems for a given firm in fiscal year ; MKTCAP j, 1 is he marke capializaion a he end of fiscal year -1; RET j, is he sock reurn of firm j from nine monhs before he end of fiscal year o hree monhs afer he end of fiscal year ; NEG j, is a dummy variable equal o 1 if RET j, is negaive and 0 oherwise. Earnings imeliness, TL, is equal o he adjused TL correspond o greaer earnings imeliness. 2 R. Larger values of Sock prices aggregae all publicly available informaion abou firm value. Accouning numbers provide more deailed informaion abou he sources of firmvalue changes by gahering, classifying and summarizing he financial effecs of firms invesmen, operaing and financing aciviies (Bushman e al., 2004). Timely and precise accouning numbers, including earnings, can help even less sophisicaed invesors exrac he underlying informaion from sock prices and help hem o undersand equiy values changes beer. Therefore, imely and efficien accouning numbers provide cleaner and less noisy informaion, enabling 21

ouside invesors o monior firm performance, hus, improving he ransparency of he operaions and aciviies of he firm o ouside invesors. Earnings imeliness records he inheren abiliy of curren earnings o capure value-relevan informaion in a imely fashion. The greaer imeliness (higher adjused 2 R ) implies ha he earnings have he abiliy o capure new informaion in a more efficien manner. The presenaion of earnings numbers is, herefore, more informaive and highly qualiaive o ouside invesors and will decrease he informaion asymmery beween managers and invesors. Among he few papers ha alk abou earnings imeliness, Ball e al. (2008) explore he deb conracing value of earnings imeliness. Bushman e al. (2004) invesigae how earnings imeliness affecs corporae governance facors, such as board srucure. Francis e al. (2004) examine he relaion beween he cos of equiy capial and earnings aribues, including earnings imeliness. I argue in his sudy ha greaer imeliness also has implicaions in capial raising evens because i miigaes poenial adverse selecion problems in SEO evens and lead o a less negaive SEO announcemen effec. Hypohesis: Ceeris paribus, negaive SEO announcemen effec is less for firms wih greaer earnings imeliness. 2.4 Daa and research design 2.4.1 Sample selecion 22

I collec he iniial SEO samples from 1984 o 2006 from he Securiies Daa Company s (SDC s) New Issue Daabase. The offerings consis of pure primary offerings or a combinaion of primary and secondary offerings. I require he samples o be common socks lised on NYSE, NASDAQ, or AMEX. I exclude: 1) limied parnership; 2) righ s issue; 3) uni issues; 4) closed-end fund; 5) SEOs lacking informaion abou filing dae, issue dae, offer price, shares filed, filing amoun; 6) SEOs wih offer prices less han $5; 7) SEOs wih more han one issue for he same filing; 8) SEOs wih a lag in issue dae as compared o he filing dae < 5 days or > 60 days. I use his resricion because shor ime differences beween he filing dae and issue dae imply mixed sock responses during boh announcemen period and issue period. Also, if he filing dae is much earlier han he issue dae, hen his may no mean ha managers hink ha he sock price (on he filing dae) is overvalued; 9) SEOs lacking CRSP daily sock reurns/prices around he SEO filing dae; and 10) SEOs wihou a one-o-one correspondence beween CUSIP in SDC and idenifier in COMPUSTAT/CRSP. 2.4.2 Filing dae and announcemen dae Because of daa availabiliy, I use he filing daes in he SDC new issue daabase for he announcemen daes. This reamen is consisen wih some of he previous sudies (Clarke e al., 2001; Denis, 1994). My argumen is ha he rue SEO announcemen releases informaion abou fuure issuances and laer SEO filing furher confirms he fuure issuance so he sock price also drops a he filing 23

dae. Considering he fac ha informaion abou equiy offerings would likely have leaked o some exen prior o he announcemen dae, using a filing dae o proxy he announcemen dae may underesimae he adverse relaion beween earnings imeliness and SEO announcemen effec. 2.4.3 Dependen variable Referring o he filing dae as day 0, I define rading days -1, 0, and 1 as an even period and compue he cumulaive abnormal reurn (CAR) in his period as he dependen variable (Brown and Warner, 1985). I firs use OLS o esimae he marke model in order o compue he cumulaive abnormal reurns in he even period. R R =-180, -179,, -10 (2) j, 0, j 1, j M, j, AR j, R j, ˆ 0, j ˆ 1, j RM, =-1, 0, 1 (3) Model _ CAR j ( R j, RM, ) (4) 1,0,1 The esimaion period is from rading day -180 o rading day -10. The CRSP equally weighed index is used as he marke reurn R M,. R j, is firm j s raw reurn on day. ˆ 0, j and ˆ 1, j are esimaed coefficiens from he esimaion period. AR j, is he abnormal reurn of firm j on day. Dependen variable, cumulaive abnormal reurn Model _ CAR j, is he sum of abnormal reurns in he even period. 24

In he sensiiviy es, I also use he marke-adjused reurns in he even period o obain he cumulaive abnormal reurn. AR j, R j, RM, =-1, 0, 1 (5) Adjused _ CAR j AR j, (6) 1,0,1 Adjused _ CAR is he sum of marke adjused reurns in even period. j 2.4.4 Conrol variables Conrol variables I consider are as follows: Offer size (REL_SIZE, ABS_SIZE): The size of he equiy offering measures he size of he negaive informaive signal. The relaive size of he issue, REL_SIZE, is compued as he number of filing shares o he number of shares ousanding (Compusa #25) before SEO announcemen. The absolue size of he issue, ABS_SIZE, is he log of filing proceeds. Alhough heories of informaion asymmery and he alernaive models based on he opimal capial srucure predic ha an increase in he size of he issue will increase he magniude of he price drop (Smih, 1986), prior empirical work finds mixed resuls (Asquih and Mullins, 1986; Mikkelson and Parch, 1985). I ry boh relaive offer size and absolue offer size in my empirical ess and find ha absolue offer size has a significanly negaive relaion wih SEO announcemen-period reurn. 25

Firm size (FIRM_SIZE): FIRM_SIZE is compued as he naural logarihm of oal asses (Compusa #6). A larger firm is usually followed by more analyss and has more media aenion. Therefore, he informaion asymmery beween managers and invesors is less for a larger firm han for a smaller firm (Lee and Masulis, 2007). So, I predic ha large firm experiences less price drop a SEO announcemen. Invesmen opporuniies (MTB): I use marke-o-book raio, MTB, as he proxy of invesmen opporuniies (Jung e al., 1996). MTB is compued as ( [Compusa #199 * Compusa #54 +Compusa #6 - Compusa #60] / [Compusa #6] ). A higher MTB implies ha he firm has more inangible asses and greaer informaion asymmery and ends o have more profiable invesmen opporuniies. Invesors end o inerpre he announcemen of equiy issues from firms wih higher MTB as reflecing he need o fund fuure promising projecs and he reducion of he informaion asymmery abou fuure invesmen opporuniies. Therefore, I predic a significan posiive coefficien on his variable. 2.4.5 Tess of hypohesis I hypohesize a posiive relaion beween he SEO announcemen-period reurn and earnings imeliness (i.e., a negaive relaion beween he SEO announcemen effec and earnings imeliness). I es he hypohesis by esimaing he following model: 26

CAR j TL FIRM _ SIZE 3 0 1 j j OFFER _ SIZE MTB 4 2 j j j (7) TLj is he value of firm j s earnings imeliness. OFFER_ SIZE j is he issue s size. I es boh relaive offer size and absolue offer size in he regression. FIRM _ SIZE j is he naural log of he oal asses of he SEO firm j. Marke-obook raio, MTB j, describes he growh opporuniies of firm j. I close his secion by addressing wo poenial selecion bias issues wih my empirical design. Firs, as discussed above, I sample only firms ha issued new socks. Bias may resul if firms ha decide o issue SEOs insead of deb are hose wih greaer or less earnings imeliness. As an illusraion, I compare he earnings imeliness of he sample firms wih ha of firms in Francis e al. (2004) and find ha SEO firms end o have significanly smaller earnings imeliness values han hose repored by Francis e al. (2004). SEO firms, hen, end o be more opaque han non-seo firms. Second, selecion bias may resul because ime-series calculaions require he use of firms wih a minimum number of survival years before SEO evens. Thus, he sample firms end o be large and successful firms. Large and successful firms end o be more ransparen han small and young firms. 2.5 Empirical resuls 2.5.1 Descripive saisics 27

Table 2.1, Panel A presens he descripive saisics of he cumulaive abnormal reurn in SEO announcemen period. The average sock reurn a 3045 SEO announcemen is from around -2% o -3%, depending on he mehod used o calculae he CAR. To obain earnings imeliness, I furher require a leas 6 yearly daa poins from 8 years prior o he SEO filing dae o 1 year prior o he SEO filing dae. Using OLS, I obain earnings imeliness (TL) from Equaion 1 in Secion 3. In order o reduce he effecs of ouliers, I exclude he observaions in he op or boom 1% of dependen and independen variables in each equaion. Descripive saisics of earnings imeliness can be found in Panel B, Table 2.1. The earnings imeliness measure, TL, has a mean (median) value of 0.109 (0.093). 2.5.2 Earnings imeliness vs. SEO announcemen-period reurn In order o reduce he effecs of ouliers, I exclude he observaions in he op or boom 1% of all available earnings imeliness and SEO announcemen-period reurn. The SEO sample consiss of 947 SEOs by 723 firms. Table 2.2 presens he frequency disribuion of SEOs by filing year and he number of offerings per firm. Panel A shows ha SEOs were more frequen a he beginning of he 1990s. Panel B shows ha abou 80 percen of firms issue SEO only once. 28

Table 2.3 provides evidence on wheher negaive SEO announcemen effec is less severe for firms wih greaer earnings imeliness. I run he regressions using marke-model-based CAR and marke-adjused CAR as dependen variables, respecively. I firs sar from he base model ha includes only offer size, firm size and growh opporuniies. Columns 1 and 4 show ha here is no significan relaion beween he SEO announcemen-period reurn and he relaive offer size (REL_SIZE). Columns 2 and 5 show a significan, negaive relaion beween SEO announcemen-period reurn and he absolue issue size (ABS_SIZE). Therefore, I use he absolue issue size, log of filing proceeds, as he proxy of offer size, hereafer. Columns 3 and 6 show ha earnings imeliness has a significan, posiive relaion wih SEO announcemen-period reurn. Using a differen calculaion of cumulaive abnormal reurns a he announcemen, he regressed coefficien is from 0.6% o 0.7%, wih a 10% or 5% significance level. This means ha firms reporing earnings in a more imely fashion experience less price-drops a he SEO announcemen. This resul is consisen wih he hypohesis. 2.5.3 Addiional ess In his secion, I examine wheher he impac of earnings imeliness on he SEO announcemen effec is unique and would no be subsumed by oher earnings aribues. To be specific, hese earnings aribues are accrual qualiy, persisence, predicabiliy, smoohness and value relevance. I do no compue firm-specific 29

conservaism because many SEO firms do no have enough negaive annual reurns in he esimaion period o calculae conservaism. Accrual qualiy describes he effeciveness of curren accruals o map ino cash flows in prior, curren and subsequen periods (Dechow and Dichev, 2002; Francis e al., 2004). One measure of accrual qualiy is he negaive of he sandard deviaion of he residuals in he firm-specific regression of accruals on lagged, curren and fuure cash flows: TCA CFO CFO CFO (8a) Asses Asses Asses Asses j, j, 1 j, j, 1 0, j 1, j 2, j 3, j j, j, j, j, j, where TCA j, is firm j s oal curren accruals in year ; CFO j, is firm j s cash flow in year ; Asse j, is he average of firm j s oal asses beween fiscal year and -1. Accrual qualiy, AQ1, is equal o ( ˆ j, ). McNichols (2002) improves he model in Equaion 8a. He finds ha he explanaory power in cross-secional regressions is grealy improved by including deflaed changes in sales and deflaed propery, plan and equipmen. He proposes he model below: TCA CFO CFO CFO ' ' ' ' Asses Asses Asses Asses j, j, 1 j, j, 1 0, j 1, j 2, j 3, j j, j, j, j, SALES PPE ' ' e j, j, 4, j 5, j j, Asses j, Asses j, (8b) 30

Accrual qualiy as defined in his model, AQ2, is equal o ( e j, ). Larger values of AQ1 (AQ2) correspond o beer accrual qualiy. I include AQ2 as he measure of accrual qualiy in he regression o examine wheher earnings imeliness influence on he SEO announcemen period reurn would be suppressed by accrual qualiy. Greaer accrual qualiy, i.e., reduced variaion in he residual, indicaes ha earnings informaion provided by he firm is a more reliable measure of a firm s cash flow and performance, so higher accrual qualiy can increase informaion qualiy provided by earnings and decrease he informaion asymmery beween managers and invesors. Therefore, greaer accrual qualiy miigaes poenial adverse selecion problems in SEO evens and leads o a less negaive SEO announcemen effec (Lee and Masulis, 2009). Value relevance measures he abiliy of earnings level and earnings change o explain he reurns (Francis e al., 2004). One measure of value relevance is he adjused 2 R of he regression of annual reurns on conemporaneous earnings and changes in earnings. RET E E j, j, j, 0, j 1, j 2, j j (9) MKTCAPj, 1 MKTCAPj, 1 In Equaion 9, E j,, MKTCAPj, 1 and RET j, are defined as Equaion 1 and, equals E j, minus E j, 1. Value relevance, VR, is equal o he adjused R 2. Larger values of VR correspond o greaer value relevance. E j as 31

Francis e al. (2004) define earnings smoohness (SMTH) as he negaive of he sandard deviaion of he deflaed earnings divided by he sandard deviaion of he deflaed cash flows. SMTH j ( E / Asses _ end ) j, j, 1 (10) ( CFO / Asses _ end ) j, j, 1 where E j, and CFO j, are he earnings and cash flows for firm j in fiscal year ; Asses _ end is he oal asses a he end of fiscal year -1. Larger values of j, 1 SMTH correspond o more earnings smoohness. The ime-series persisence of earnings describes he auocorrelaion beween pas earnings and fuure earnings, while he predicabiliy of earnings reflecs he abiliy of pas earnings o predic fuure earnings (Lipe, 1990; Francis e al., 2004). One measure of earnings persisence is he auocorrelaion coefficien in AR(1) model for adjused earnings and one measure of earnings predicabiliy is he sandard deviaion of he negaive of he sandard deviaion of he residuals in he AR(1) model for adjused earnings. X X (11) j, 0, j 1, j j, 1 j, where X j, is firm j s spli-adjused earnings per share in fiscal year ; earnings persisence, PER, is equal o ˆ 1, j and earnings predicabiliy, PRED, is equal o ( ˆ j, ). Larger values of PER and PRED correspond o beer earnings persisence and greaer predicabiliy, respecively. 32

In calculaing for each earnings aribue, I require a leas 6 yearly daa poins from 8 years prior o he SEO filing dae o 1 year prior o he SEO filing dae. Using OLS, I obain accrual qualiies (AQ2), value relevance (VR) and smoohness (SMTH) from Equaions 8 hrough 10. Using a maximum likelihood mehod, I obain earnings persisence (PER) and predicabiliy (PRED) from he AR(1) model in Equaion 11. As in he calculaion of earnings imeliness, I exclude he observaions in he op or boom 1% of dependen and independen variables in my calculaion of each earnings aribue o reduce he effecs of ouliers. Table 2.1, Panel B presens descripive saisics of all earnings aribues. As discussed previously, earnings imeliness measure, TL, has a mean (median) value of 0.109 (0.093). In comparison, Francis e al. (2004) repor a mean (median) value of 0.466 (0.465). The measure of value relevance, VR, has a mean (median) value of 0.142 (0.116), while Francis e al. (2004) give a mean (median) value for value relevance as 0.423 (0.416). My resuls on accrual qualiy are comparable o hose repored by Francis e al. (2004). This paern may arise from he self-selecion problems inheren in my research design. To review, all sample firms cied in my work are hose ha issue new shares. I is possible, hough, ha firms deciding o issue hese shares are also hose ha do no include value-relevan informaion in a imely manner or are hose wih earnings daa ha do no accoun for heir reurns well. 33

In order o examine wheher oher aribues subsume earnings imeliness, I run he regression and examine wheher he coefficien and significance level of earning imeliness sill holds in he presence of he ohers: CAR j PRED 5 TL FIRM _ SIZE 8 0 j 1 SMTH 6 j j AQ MTB 9 2 j j j VR j j OFFER _ SIZE (12) 7 3 EP 4 j j I also exclude he observaions in he op or boom 1% of each earnings aribue and SEO announcemen-period reurn. The final sample consiss of 495 SEOs by 379 firms. Table 2.4 presens he frequency disribuion of SEOs by filing year and number of offerings per firm. Panel A shows ha SEOs are more frequen a he beginning of 1990s. Panel B shows ha abou 80% firms issue SEO only once. As in Lee and Masulis (2009), Table 2.5 shows ha negaive SEO announcemen effec is less for firms wih beer accrual qualiy a he 5% level ( saisics: 2.34). No significan relaions are found beween he SEO announcemen effec and earnings persisence, earnings predicabiliy, value relevance or earnings smoohness. More imporanly, i also shows ha he effec of earnings imeliness on he SEO announcemen period reurn sill holds a he 5% level (coefficien: 1.53%; saisics: 2.45) when oher earnings aribues, including accrual qualiy, are considered, suggesing ha earnings imeliness capures is own dimension in he cos of financing aciviies. Table 2.5 shows ha boh accouning-based and marke-based financial reporing qualiies may affec SEO announcemen effec. However, i is necessary 34

o poin ou ha, compared wih accouning-based earnings aribues, all markebased measures have heir inheren limiaions in even sudies. Marke-based accouning aribues, such as earnings imeliness, value relevance and conservaism, are calculaed in he way ha he sock marke is regarded as being efficien. However, even sudies in SEO, earnings announcemens or merger and acquisiions, usually deal wih siuaions in which socks are overvalued or undervalued and invesors regard hese evens as signals ha reveal he rue value of sock prices. 2.6 Conclusions This paper examines wheher he fac ha a firm repors is earnings in a imely way affecs invesors responses a he ime of a firm s announcemen of is SEO financing decision. I find ha firms wih greaer earnings imeliness end o experience less price drops a SEO announcemens. The resuls presen evidence ha imely financial reporing can help invesors o assess firm performance by reducing he informaion asymmery beween managers and invesors. So, his paper conribues o lieraure abou he consequences of financial reporing qualiy. This sudy provides empirical evidence on he relaion beween earnings aribues and a firm s informaion environmen and proposes ha informaion asymmery can be one link beween earnings aribues and informaion risk. This paper also enriches he scarce lieraure abou earnings imeliness. 35

This sudy suggess wo poenial avenues for fuure research. Firs, i may be inriguing o examine wheher SEO firms end o be less ransparen firms and, hus, se up a link beween corporae governance and corporae invesmen decisions. Second, disinguishing primary offerings and secondary offerings may yield some ineresing opics. 36

References Asquih, P., and D. W. Mullins Jr. 1986. Seasoned equiy offerings. Journal of Financial Economics 15: 61 89. Ball, R., R. Bushman, and F. Vasvari. 2008. The Deb-Conracing Value of Accouning Informaion and Loan Syndicae Srucure. Journal of Accouning Research 46: 247-287. Basu, A. 1997. The conservaism principle and he asymmeric imeliness of earnings. Journal of Accouning and Economics 24: 3 37. Biddle, G., G. Hillary, and R. Verdi. 2009. How does financial reporing qualiy relae o invesmen efficiency? Journal of Accouning and Economics 48: 112-131. Brazel J., and E. Webb. 2006. CEO compensaion and he seasoned equiy offering decision. Managerial and Decision Economics 27: 363 378. Brown, S., and J. Warner. 1985. Using daily sock reurns: The case of even sudies. Journal of Financial Economics 14: 3-31. Bushman, R., Q. Chen, E. Engel, and A. Smih. 2004. Financial accouning informaion, organizaional complexiy and corporae governance sysems. Journal of Accouning and Economics 37: 167 201. Choe, H., R. Masulis, and V. Nanda. 1993. Common sock offerings across he business cycle: Theory and evidence. Journal of Empirical Finance 1: 3 31. Clarke, J., C. Dunbar, and K. Kahle. 2001. Long-run performance and insider rading in compleed vs. canceled seasoned equiy offerings. Journal of Financial and Quaniaive Analysis 36: 415-430. Dechow, P., and I. Dichev. 2002. The qualiy of accruals and earnings: The role of accrual esimaion errors. The Accouning Review 77 (supplemen): 35 59. Denis, D. 1994. Invesmen opporuniies and he marke reacion o equiy offerings. Journal of Financial and Quaniaive Analysis 19: 159 177. Dierkens, Nahalie. 1991. Informaion asymmery and equiy issues. Journal of Financial and Quaniaive Analysis 26: 181 99. Francis, J., R. Lafond, P. M. Olsson, and K. Schipper. 2004. Coss of equiy and earnings aribues. The Accouning Review 79: 967 1010. 37

Huon, A.P., A.J. Marcus, and H. Tehranian. 2009. Opaque financial repors, R2, and crash risk. Journal of Financial Economics 94: 67-86. Jung, K., Y. Kim, and R. Sulz. 1996. Timing, invesmen opporuniies, managerial discreion, and he securiy issue decision. Journal of Financial Economics 42: 159 185. Korajczyk, R., D. Lucas, and R. McDonald. 1991. The effec of informaion releases on he pricing and iming of equiy issues. Review of Financial Sudies 4 (4): 685 708. Lang, M., and R. Lundholm. 2000. Volunary disclosure and equiy offerings: reducing informaion asymmery or hyping he sock? Conemporary Accouning Research 17: 623-662. Lee, G., and R. Masulis, 2009. Seasoned equiy offerings: qualiy of accouning informaion and expeced floaion coss. Journal of Financial Economics 92: 443-469. Leland, H., and D. Pyle, 1977. Informaion asymmeries, financial srucure, and financial inermediaion. Journal of Finance 32: 371 387. Lipe, R. 1990. The relaion beween sock reurns and accouning earnings given alernaive informaion. The Accouning Review 65: 49 71. Masulis, R. W., and A. N. Korwar. 1986. Seasoned equiy offerings: An empirical invesigaion. Journal of Financial Economics 15: 91 117. McNichols, M. 2002. Discussion of he qualiy of accruals and earnings: The role of accrual esimaion errors. The Accouning Review 77 (supplemen): 61 69. Mikkelson, W., and M. Parch. 1986. Valuaion effecs of securiy offerings and he issuance process. Journal of Financial Economics 15: 31 60. Myers, S., and N. Majluf. 1984. Corporae financing and invesmen decisions when firms have informaion ha invesors do no have. Journal of Financial Economics 13: 187 221. Smih, C. 1986. Invesmen banking and he capial acquisiion process. Journal of Financial Economics 15: 3 29. Zhang, J. 2008. The conracing benefis of accouning conservaism o lenders and borrowers. Journal of Accouning and Economics 45: 27-54. 38

Table 2.1 Descripive saisics The iniial samples consiss of 3045 SEOs from 1984 o 2006 lised on NYSE, NASDAQ, or AMEX and excludes: 1) limied parnership, 2) righ s issue, 3) uni issue, 4) closedend fund, 5) SEOs lacking informaion abou filing dae, issue dae, offer price, shares filed, filing amoun, 6) SEOs wih offer prices less han $5, 7) SEOs wih more han one issue for he same filing, 8) SEOs whose lag of issue dae compared o filing dae is smaller han 5, or larger han 60 9) SEOs lacking CRSP daily sock reurns for he hree rading days around SEO filings or from he prior 180 rading days o he prior 10 rading days, 10) SEOs wihou one-o-one correspondence beween CUSIP in SDC and idenifier in COMPUSTAT/CRSP. Model_CAR is he cumulaive abnormal reurn in he even period using he OLS marke model. Adjused_CAR is he cumulaive marke-adjused reurn in he even period. (Filing dae: day 0; even period: rading day -1, 0, 1; esimaion period: period from rading day -180 o rading day -10). For each earnings aribue s calculaion, a leas 6 years necessary financial saemens daa are required wihin 8 years prior o he SEO filing dae. Accrual qualiy has wo measures: AQ1 and AQ2. AQ1 is equal o ( ˆ j, ) in TCAj, CFO j, 1 CFOj, CFO j, 1 0, j 1, j 2, j 3, j j,. Asses Asses Asses Asses j, j, j, j, AQ2 is equal o ( eˆ j, ) in TCAj, CFO j, 1 CFOj, CFOj, 1 SALES j, PPE j, ' ' ' ' ' ' e Asses Asses Asses Asses Asses Asses. 0, j 1, j 2, j 3, j 4, j 5, j j, j, j, j, j, j, j, Earnings persisence (EP) and predicabiliy (PRED) are measured as ˆ 1, j and ( ˆ j, ) in X X, respecively. j, 0, j 1, j j, 1 j, 39

Earnings smoohness (SMTH) is measured as ( E / Asses _ end ) j, j, 1. ( CFO / Asses _ end ) j, j, 1 Value relevance (VR) and earnings imeliness (TL) are measured as he adjused Ej, Ej, RETj, 0, j 1, j 2, j j and SHARES P SHARES P E SHARES j, P j, 1 j, 1 j, 1 j, 1 j, 1 j, 1 b b NEG b RET b NEG RET j,0 j,1 j, j,2 j, j,3 j, j, j,. 2 R in TCA Toal curren accruals = CA - CL - Cash + STDEBT ( : change beween year -1 o year ); CA Curren asse (Compusa #4); CL Curren liabiliies (Compusa #5); Cash Cash and shor-erm invesmens (Compusa #1); STDEBT Deb in curren liabiliies (Compusa #34); CFO Asses Cash flow from operaions = E TCA + depreciaion amorizaion (Compusa #14); Average oal asses (Compusa #6) in year and year -1; Sales Sales (Compusa #12); Assess_end Toal asses a he end of fiscal year; PPE Propery, plan and equipmen (Compusa #7); X Spli-adjused earnings per share (Compusa #58); RET E Twelve-monh raw reurn ending hree monhs afer he end of fiscal year ; Earnings before exraordinary iems, disconinued operaions, and special iems (Compusa #18); SHARES Common shares ousanding (Compusa #25); P Sock price fiscal year close (Compusa #199). 40

Mean Sd. Dev. 10% 25% Median 75% 90% Model_CAR -0.0284 0.0666-0.1072-0.0627-0.0259 0.0054 0.0404 Adjused_CAR -0.0194 0.0646-0.0924-0.0529-0.0200 0.0113 0.0494 Table 2.1 Panel A: Descripive saisics of SEO announcemen period reurns Panel B: Descripive saisics of earnings aribues N Mean Sd. Dev. 10% 25% Median 75% 90% AQ1 598-0.0317 0.0273-0.0676-0.0453-0.0240-0.0109-0.0053 AQ2 573-0.0236 0.0236-0.0530-0.0312-0.0152-0.0072-0.0037 EP 1106 0.2529 0.3744-0.2485-0.0106 0.2827 0.5412 0.7422 PRED 1106-0.9950 1.0399-2.2110-1.1936-0.6370-0.3603-0.2166 SMTH 944-0.7438 0.4423-1.2521-0.9773-0.6956-0.4333-0.2472 VR 950 0.1424 0.3642-0.3302-0.1503 0.1164 0.4069 0.6540 TL 958 0.1086 0.4403-0.4437-0.2205 0.0933 0.4450 0.7284 41

Table 2.2 Frequency disribuion The samples consiss of 947 SEOs by 723 firms from 1984 o 2006 lised on NYSE, NASDAQ, or AMEX and excludes: 1) limied parnership, 2) righ s issue, 3) uni issue, 4) closed-end fund, 5) SEOs lacking informaion abou filing dae, issue dae, offer price, filing shares, filing amoun, 6) SEOs wih an offer prices less han $5, 7) SEOs wih more han one issue for he same filing, 8) SEOs whose lag of issue dae compared o filing dae is smaller han 5, or larger han 60 9) SEOs lacking CRSP daily sock reurns for he hree rading days around SEO filings or from he prior 180 rading days o he prior 10 rading days, 10) SEOs wihou one-o-one correspondence beween CUSIP in SDC and idenifier in COMPUSTAT/CRSP. Panel A: Frequency disribuion of SEOs by filing year Cumulaive Cumulaive SEO_year Frequency Percen Frequency Percen 1984 26 2.75 26 2.75 1985 51 5.39 77 8.13 1986 58 6.12 135 14.26 1987 41 4.33 176 18.59 1988 16 1.69 192 20.27 1989 29 3.06 221 23.34 1990 24 2.53 245 25.87 1991 73 7.71 318 33.58 1992 69 7.29 387 40.87 1993 75 7.92 462 48.79 1994 37 3.91 499 52.69 1995 51 5.39 550 58.08 1996 52 5.49 602 63.57 1997 31 3.27 633 66.84 1998 30 3.17 663 70.01 1999 30 3.17 693 73.18 2000 33 3.48 726 76.66 2001 41 4.33 767 80.99 2002 44 4.65 811 85.64 2003 41 4.33 852 89.97 2004 40 4.22 892 94.19 2005 27 2.85 919 97.04 2006 28 2.96 947 100.00 42

Panel B: Frequency disribuion of SEOs by number of offerings Cumulaive Cumulaive N Frequency Percen Frequency Percen 1 573 79.25 573 79.25 2 107 14.80 680 94.05 3 24 3.32 704 97.37 4 11 1.52 715 98.89 5 5 0.69 720 99.59 6 2 0.28 722 99.86 7 1 0.14 723 100.00 43

Table 2.3 Regression of SEO announcemen effec on earnings imeliness This able presens OLS regression esimaes of SEO announcemen-period reurn on earnings imeliness (TL). The SEO sample consiss of 947 filings by 723 firms over he period from 1984 o 2006. In he firs 3 columns, I use model-based cumulaive abnormal reurn as he dependen variable. In he las 3 columns, I use marke-adjused cumulaive abnormal reurn as he dependen variable. The absolue value of saisics is in brackes. ***,**, and * represen 1%, 5%, and 10% significance respecively. Model_CAR Adjused_CAR 1 2 3 4 5 6 REL_SIZE -0.0005-0.0009 [1.35] [0.24] ABS_SIZE -0.0073-0.0071-0.0040-0.0039 [5.17]*** [3.74]*** [2.16]** [2.09]** MTB 0.0015 0.0022 0.0022 0.0028 0.0032 0.0032 [2.52]*** [3.58]*** [3.51]*** [4.88]*** [5.35]*** [5.29]*** Firm_Size 0.0034 0.0061 0.0062 0.0019 0.0035 0.0035 [3.89]*** [5.54]*** [5.58]*** [2.28]** [3.22]*** [3.25]*** TL 0.0075 0.0062 [2.01]** [1.71]* Inercep -0.0456-0.0340-0.0356-0.0326-0.0266-0.028 [7.35]*** [5.17]*** [5.38]*** [5.43]*** [4.15]*** [4.32]*** N 947 947 947 947 947 947 Adj_R2 0.0162 0.0310 0.0342 0.0234 0.0281 0.0301 44

Table 2.4 Frequency disribuion The iniial samples consiss of 495 SEOs by 379 firms from he period of 1984 o 2006 lised on NYSE, NASDAQ, or AMEX and excludes: 1) limied parnership, 2) righ s issue, 3) uni issue, 4) closed-end fund, 5) SEOs lacking informaion abou filing dae, issue dae, offer price, filing shares, filing amoun, 6) SEOs wih an offer prices less han $5, 7) SEOs wih more han one issue for he same filing, 8) SEOs whose lag of issue dae compared o filing dae is smaller han 5, or larger han 60 9) SEOs lacking CRSP daily sock reurns for he hree rading days around SEO filings or from he prior 180 rading days o he prior 10 rading days, and 10) SEOs wihou one-o-one correspondence beween CUSIP in SDC and idenifier in COMPUSTAT/CRSP. I also require a leas 6 daa poins wihin 8 years prior o he SEO filing dae in calculaing for 6 earnings aribues: imeliness, value relevance, accrual qualiy, earnings persisence, predicabiliy, and smoohness. Panel A: Frequency disribuion of SEOs by filing year Cumulaive Cumulaive SEO_year Frequency Percen Frequency Percen 1984 20 4.04 20 4.04 1985 20 4.04 40 8.08 1986 28 5.66 68 13.74 1987 27 5.45 95 19.19 1988 9 1.82 104 21.01 1989 16 3.23 120 24.24 1990 17 3.43 137 27.68 1991 33 6.67 170 34.34 1992 40 8.08 210 42.42 1993 39 7.88 249 50.30 1994 26 5.25 275 55.56 1995 34 6.87 309 62.42 1996 29 5.86 338 68.28 1997 23 4.65 361 72.93 1998 16 3.23 377 76.16 1999 14 2.83 391 78.99 2000 16 3.23 407 82.22 2001 19 3.84 426 86.06 2002 23 4.65 449 90.71 2003 17 3.43 466 94.14 2004 14 2.83 480 96.97 2005 15 3.03 495 100.00 45

Panel B: Frequency disribuion of SEOs by number of offerings Cumulaive Cumulaive N Frequency Percen Frequency Percen 1 304 80.21 304 80.21 2 49 12.93 353 93.14 3 15 3.96 368 97.10 4 9 2.37 377 99.47 6 2 0.53 379 100.00 46

Table 2.5 Relaion beween SEO announcemen-period reurn and earnings aribues This able presens OLS regression esimaes of SEO announcemen-period reurn on earnings aribues. The SEO sample consiss of 495 filings by 379 firms over 1984 o 2006. The dependen variable is he model-based cumulaive abnormal reurn in SEO even period. The absolue value of saisics is in brackes. ***,**, and * represen 1%, 5%, and 10% significance respecively. 1 2 3 4 5 6 7 8 ABS_SIZE -0.0124-0.0123-0.0107-0.0124-0.0125-0.0124-0.0116-0.0102 [4.33]*** [4.27]*** [3.65]*** [4.34]*** [4.33]*** [4.33]*** [3.97]*** [3.44]*** MTB 0.0057 0.0056 0.0061 0.0057 0.0058 0.0058 0.0059 0.0064 [3.66]*** [3.62]*** [3.92]*** [3.66]*** [3.66]*** [3.68]*** [3.75]*** [4.08]*** Firm_Size 0.0112 0.0113 0.0093 0.0111 0.0112 0.0109 0.0104 0.0081 [5.73]*** [5.82]*** [4.43]*** [5.69]*** [5.71]*** [5.46]*** [5.17]*** [3.60]*** TL 0.0113 0.0153 [2.09]** [2.45]** AQ2 0.2702 0.2798 [2.47]** [2.34]** VR -0.0017-0.0109 [0.25] [1.41] EP -0.0016-0.0026 [0.24] [0.41] PRED -0.0013-0.0040 [0.52] [1.47] SMTH 0.0090 0.0047 [1.42] [0.68] Inercep -0.0501-0.0528-0.0401-0.0496-0.0500-0.0501-0.0431-0.0355 [5.31]*** [5.56]*** [3.92]*** [5.12]*** [5.21]*** [5.31]*** [4.05]*** [3.13]*** N 495 495 495 495 495 495 495 495 Adj_R2 0.0594 0.0657 0.0691 0.0576 0.0576 0.0580 0.0613 0.0758 47

Chaper 3 Essay II: Marke Power and Accrual Managemen 3.1 Inroducion Earnings include cash flows and accruals. Accrual managemen refers o aciviies underaken by managers o inflae or reduce repored earnings via accruals, bu no o change curren cash flows. The pas 20 years have seen an enormous increase in accrual managemen (Bergsresser and Philippon, 2006). This essay examines wheher a firm s accrual managemen is affeced by is produc marke power. My hypohesis builds on Peress (2010). Peress ses up a heoreical model in which here is perfec compeiion in he sock marke bu imperfec compeiion in he produc marke. Imperfec compeiion in he produc marke provides each firm some abiliy o se prices for is produc. A firm wih greaer marke power has greaer abiliy o pass on produciviy shocks o is cusomers by seing prices (Kale and Loon, 2011). Wihin his framework, I hypohesize ha firms wih greaer marke power have less incenives o manipulae heir earnings hrough accrual managemen ha pu he managers a more aud liigaion or regulaion risk, because hey could mee he earnings expecaions by seing prices accordingly. 48

In he empirical analysis, I use he Lerner index o measure produc marke power and he asse-deflaed absolue discreionary accruals o proxy he magniude of accrual managemen. Using a sample of 35,745 firm-year observaions over he 1997 o 2007 period, I find supporing evidence ha firms wih greaer marke power end o have lower levels of asse-deflaed discreionary accruals, even afer conrolling for oher well-known variables ha affec he pracice of accrual managemen. This essay is relaed o Marciuauye and Park (2009) in he sense ha we boh work on he relaion beween produc marke and earnings managemen. However, hese wo sudies are differen in several aspecs. Firs, Marciuauye and Park (2009) employs an indusry-level measure, he Herfindahl-Hirschman index o proxy he overall compeiion inensiy of an indusry. My paper uses he firm-level Lerner index o characerize a firm s compeiion saus in is indusry. Second, my firm-level sudy explores he effec of a firm s compeiion saus in indusry on is earnings managemen aciviies, while Marciuauye and Park examine he effec of indusry-level compeiion inensiy on earnings managemen. In oher words, Marciuauye and Park examine he compeiion s iner-indusry effec on accrual managemen, and I examine he compeiion s inra-indusry effec on accrual managemen. Third, he wo sudies are based on differen heories. My argumen considers he effecs of produc marke power on boh he needs o manage earnings due o differen agency problems and he abiliies o manage earnings specifically hrough accrual managemen due o differen price- 49

seing abiliies, while heir argumen akes ino accoun agency conflic heory only. This essay has several conribuions. Firs, i exends he earnings managemen lieraure by documening he imporance of a firm s marke power for driving managers accrual managemen decisions. Prior lieraure suggess facors ha affec he degree o which a firm engages in accrual managemen are deermined by he need and abiliy o manage earnings. Healy and Wahlen (1999) classifies accrual managemen reasons ino hree groups: capial marke, conrac and regulaory reasons. For example, avoiding losses increases managers need o manipulae earnings; ransparen operaing environmens resric managers abiliy o manage earnings. This essay adds o he exising lieraure by showing ha a firm s compeiion level in produc marke, a facor ha has never been documened in prior lieraure, is also a significan deerminan of managers incenives o manage earnings hrough accruals. Second, my sudy complemens recen sudies on he relaion beween produc marke and capial marke by providing empirical evidence on wheher a firm s produc marke power affecs is financial reporing pracices. Due o increased globalizaion and inensiy of impor peneraion, relaxaion of barriers o enry and rade and he speed of echnological change, compeiion in produc markes is increasingly inense (Peress, 2010; Gaspar and Massa, 2005). Researchers have become more ineresed in he effecs of he compeiion on various aspecs of finance or accouning, such as managers invesmen decisions, managers disclosure decisions, analyss forecass properies and asse pricing 50

(Fee and Thomas 2004; Booson and Sanford, 2005; Ali e al., 2010; Marciukaiye and Park, 2009; Hou and Robinson, 2006). This paper adds o exising sudies by documening he effec of a firm s compeiion saus on a manager s accrual managemen pracices. Peress predics several possible effecs of marke power on capial marke, such as sock liquidiy, analyss forecass, asse allocaion, and informaion efficiency. My paper adds o his work by demonsraing he impac of marke power on he financial reporing qualiy. Third, he demonsraed resuls also have direc implicaions for regulaors and audiors by offering empirical evidence ha he use of discreionary accruals o manipulae repored earnings is more pronounced a firms wih weak marke power. For example, in indusries where many firms are in a disadvanageous compeiion saus, regulaors may refine exising accouning sandards o enhance heir financial reporing qualiy. Audiors can also benefi from his sudy by knowing how o adjus audi fees based on heir cliens compeiion levels accordingly, which I will discuss in Essay III. The remainder of he paper is organized as follows. The nex secion reviews prior lieraure on accrual managemen and marke power and develops he hypohesis. Secion 3.3 inroduces he daa and he empirical approach. Secion 3.4 presens empirical resuls. The final secion conains a conclusion. 3.2 Hypohesis developmen 51

Earnings managemen has been a opic of enormous ineres for many years. I occurs for a variey of reasons. Healy and Wahlen (1999) classify hese incenives ino hree groups: capial marke moivaions, conracing moivaions, and regulaory moivaions. Accouning daa are imporan for equiy holders o value a firm. Researchers have provided evidences on earnings managemen due o high sock marke expecaions in periods prior o specific evens such as equiy offering and sock-financed acquisiions (Teoh e al., 1998a; Teoh e al., 1998b; Erickson and Wasn, 1998). Burgsahler and Dichev (1997) presen evidence ha managers may manipulae earnings o avoid negaive earnings by showing abnormal disconinuiies in he disribuion of repored earnings. There is also evidence showing ha some managers manipulae earnings o avoid reporing a loss, earnings declines or falling shor of marke expecaions (Burgsahler and Eames, 1998; Abarbanell and Lehavy, 2003; Degeorge e al., 1999). Accouning daa are no only used by equiy holders o value a firm, bu also o help oher sakeholders monior and regulae conracs such as compensaion conracs and lending conracs. Many sudies also presen evidence demonsraing earnings managemen aciviies for conracing reasons. For example, Healy (1985) finds ha managers end o use heir discreion in accrual judgmens o increase heir earnings-based bonuses. DeFond and Jiambalvo (1994) show ha some firms accelerae earnings o reduce he likelihood of violaing lending covenans. In addiion, a number of oher sudies have examined he impac of anirus regulaion or indusry-specific regulaion on managers propensiies o manipulae earnings. 52

Managers also have differen mehods of managing earnings. Earnings are composed of cash flows and accruals. Accruals reflec changes in firm value ha are no refleced in curren cash flows (e.g., in accouns receivable and/or accoun payable). Accruals are relaively hard o measure and involve a grea deal of managemen discreion. Accrual managemen refers o wihin-gaap (Generally Acceped Accouning Principles) opporunisic aciviies o obscure rue economic performance (Dechow and Skinner, 2000). Accrual managemen involves risks and coss. Demers and Wang (2009) find, for insance, ha he reversing naure of accruals causes younger managers o handle accruals less in he early sages of heir careers because of concern for heir own career pahs. Zang (2006) and Cohen e al. (2008) show ha aggressive accrual managemen pus firms a higher risk of regulaory scruiny and liigaion. Furhermore, financial reporing choices mus mee he requiremens of audiors and hus have limied accouning flexibiliy. So, managers end o consider accrual managemen less if hey have an alernaive o boos or decrease earnings under less pressure from regulaory scruiny and audiing. The purpose of his paper is, herefore, o invesigae wheher a firm s superior compeiion saus in produc marke provides i an alernaive o boos or decrease earnings in addiion o accrual managemen. The discussion above clearly shows ha when managers make decisions on heir earnings managemen pracice, hey need o consider he need o manage earnings and heir abiliy o do so hrough accrual managemen. Therefore, I examine wheher a firm wih advanageous compeiion saus has more need o 53

manage earnings. If i does, will is compeiion saus affec is abiliy o achieve his goal? On he one hand, marke power may increase a firm s needs o manage earnings. Firms wih greaer marke power usually enjoy he underlying profis from heir superior compeiion saus, so hey may engage in more earnings managemen o decrease heir earnings in order o deer any new enries or avoid governmen inervenion. Also, firms wih greaer marke power end o have greaer profiabiliy han hose wih less marke power, hus, leading o more managerial slacks, greaer agency conflics and, consequenly, more need o manage earnings. On he oher hand, marke power provides firms wih superior compeiion saus alernaive ways o boos or reduce heir earnings. Even if firms wih superior compeiion saus have more need o reduce earnings o deer new enries or avoid governmen inervenion, hey do no have o realize his hrough accrual managemen because heir compeiion saus provides hem a legal and safer way o mee heir requiremens on earnings. Wih a number of alernae ways of boosing or decreasing earnings a is disposal, firms wih greaer marke power may rely less on accrual managemen. Peress (2010) ses up a heoreical model in which here is perfec compeiion in he sock marke bu imperfec compeiion in he produc marke. He poins ou ha imperfec compeiion in he produc marke provides each firm wih some abiliy o se prices for is produc. A firm wih greaer marke power has a greaer abiliy o pass on produciviy shocks o is cusomers by seing prices (Kale and Loon, 2011). For example, a firm can raise 54

is produc prices o boos sales when hey mee negaive produciviy shocks o heir oupu. Hence, I argue ha firms wih greaer marke power rely less on accrual managemen o mee heir earnings goals because hey could mee he earnings expecaions by seing prices accordingly. To summarize, agency heory and price-seing heory provides wo conflic predicions as o he relaion beween marke power and accrual managemen. However, superior compeiion saus also provides firms wih alernaive ways o mee heir earnings expecaions and, hus, rely less on accrual managemen. I herefore have he hypohesis below: Hypohesis: Ceeris paribus, firms wih greaer marke power end o have less absolue value of discreionary accruals. Before I conclude his secion, i is necessary o clarify he difference beween price-seing aciviies and real managemen. Managers may boos or reduce earnings eiher via changing accruals or via changing cash flows. The price-seing aciviies aken by firms wih marke power are a differen mechanism from accrual managemen because he former has direc cash flow effecs. They are also differen from real managemen even if boh mehods have impacs on cash flows in curren period. Roychowdhury (2006) defines real managemen aciviies as deparures from normal operaional pracices, moivaed by managers desires o mislead a leas some sakeholders ino believing cerain financial reporing goals have been me in he normal course of operaions. He finds evidence suggesing price discouns, overproducion and reducion of 55

discreionary expendiures o improve repored earnings. Thus, real managemen aciviies could be long-erm, value-desroying and resul in he reducion of fuure revenue generaing capabiliy, while price-seing aciviies aken by firms wih greaer power would no affec a firm s normal operaional pracices. Therefore, price-seing aciviies are also significanly differen from real managemen aciviies. Price-seing aciviies are he privileged righs enjoyed by firms wih superior marke power in an imperfecly compeiive produc marke. They may involve coss resuling from a loss in marke power or repuaion. Bu, sricly speaking, hey do no fall wihin he scope of earnings managemen. They provide a possible choice for firms in need of avoiding accrual managemen or real managemen. Fuure research may explore he rade-off beween real managemen aciviies and price-seing aciviies. 3.3 Research design This secion documens he consrucion of he dependen variable, independen variable and conrol variables. I also discuss he esimaed coefficiens on conrol variables and he regression model. 3.3.1 Dependen variable: discreionary accruals 56

The dependen variable is he absolue value of asse-deflaed discreionary accruals. I use a modified cross-secional Jones model described in Dechow e al., (1995) o calculae he deflaed discreionary accruals (Jones, 1991). Discreionary accruals are he unexplained porion of oal accruals. They are obained by subracing non-discreionary accruals from oal accruals, while oal accruals are measured as he difference beween repored earnings and cash flows from operaions. To deermine non-discreionary accruals, I firs run oal accruals on variables ha proxy normal accruals, i.e., changes in sales and gross propery, plan and equipmen. I use he cross-secional OLS regressions by he firs 2-digi SIC code o esimae 0, 1, and 2 in Equaion 1. A leas 10 consecuive firmyear observaions are required in each cross-secional regression. In order o conrol for heeroscedasiciy, all variables are deflaed by lagged oal asses. Asse-deflaed nondiscreionary and discreionary accruals are he fied values and residuals of he regression in Equaion 1. TA 1 0 1 Ai, 1 A REV 1 A 1 2 PPE A 1 (1) In Equaion 1, he subscrip i refers o firms, he subscrip refers o years. TA, oal accruals, equals earnings before exraordinary iems and disconinued operaions less operaing cash flows from coninuing operaions. To avoid he nonariculaion problem menioned in Collins and Hribar (2002), I collec operaing cash flows from firms cash flow saemens repored under he Saemen of 57

Financial Accouning Sandards no. 95 (SFAS no. 95, FASB 1987) insead of firms balance shees in successive years. REV represens he changes in revenues. PPE is gross propery, plan, and equipmen. A is he oal asses. I inroduce he esimaed coefficiens ˆ 0, 1 ˆ, and ˆ 2 o Equaion 2 o calculae he asse-deflaed nondiscreionary accruals. AR in Equaion 2 is he changes in accoun receivables. NonDis _ Accrual ˆ 1 ˆ REV AR 0 1 (2) 2 Ai, 1 Ai, 1 Ai, 1 ˆ PPE Finally, I derive he asse-deflaed discreionary accruals as TA DIS _ AC Accrual. The dependen variable is he absolue NonDis _ Ai, 1 value of asse-deflaed discreionary accruals. 3.3.2 Independen variable: produc marke power As suggesed in previous lieraure (Lerner, 1934; Carlon and Perloff, 2000; Kale and Loon, 2010), I use he Lerner index or price-cos margin (PCM) as he measure of produc marke compeiion saus. Following Peress (2010), I measure PCM as he raio of operaing profi o sales. Operaing profi is sales less cos of goods sold and selling, general and adminisraive expenses. As noed in Kale and Loon (2010) and McFalls (1997), cours and governmen agencies usually employ marke share as a measure of marke power. 58

So I use MKT_SH as an alernaive measure of marke power in he robusness es. MKT_SH is calculaed as he raio of he firm s sales o oal sales in he same 4- digi indusry sales. 3.3.3 Conrol Variables Based on exising research on accrual managemen, I consider firms operaing environmens, growh opporuniies, profiabiliy and regulaion environmens in he regression. Below are he descripions on he conrol variables and he prediced sign of heir coefficiens (Yu, 2008; Bergsresser and Philippon, 2006; Marciukaiye and Park, 2009): SIZE (-): firm size, measured as he naural log of marke value of equiy. The coefficien for he firm size is expeced o be negaive because large firm size implies a more ransparen informaion environmen. Managers have fewer opporuniies o manage earnings. MTB (+): marke-o-book raio, measured as he raio of marke value of a firm o oal asses. Marke value is oal asses plus marke value of common equiy minus book value of common equiy. The coefficien for MTB is expeced o be posiive because firms wih more growh opporuniies end o be less ransparen. Managers are more likely o engage in discreionary accrual managemen. GROWTH (+): growh rae of asses, measured as he change of asses scaled by lagged asses. The coefficien for GROWTH is expeced o be posiive because 59

growh rae of asses can also proxy a firm s growh opporuniies and volaile saus. ROA (-): reurn on asses, measured as he earnings before ineres and ax divided by oal asses. The coefficien for ROA is expeced o be negaive because firms wih low profiabiliy end o have volaile cash flow and hus firms end o engage in discreionary accrual managemen. VOLAT (+): sandard deviaion of annual asse-deflaed cash flow growh over las five years. The coefficien for VOLAT is expeced o be posiive because managers have more rooms o manage earnings when firms are in more volaile saes. LEVER (?): leverage, calculaed as he raio of long-erm deb o oal asses. On he one hand, deb holders play a monioring role on a firm s operaion; on he oher hand, high leverage may creae pressure on managers o manipulae earnings. Previous lieraure also finds mixed resuls on he coefficien sign for LEVER. So I do no make any predicions on he sign for he esimaed coefficien on LEVER. EXTER (+): exernal financing aciviies, measured as he sum of ne cash received from equiy and deb issuance scaled by oal asses. The coefficien for EXTER is expeced o be posiive because managers end o manipulae earnings via accrual managemen in periods of equiy offerings or deb issuance (Teoh e al., 1998a, 1998b). 60

BUS_SEG (+): number of business segmens. The coefficien for he number of business segmens is expeced o be posiive because more indusry diversificaion leaves more space for earnings manipulaion. GEO_SEG (+): number of geographic segmens. The coefficien for he number of geographic segmens is expeced o be posiive because more indusry diversificaion leaves more space for earnings manipulaion. SOX (-): SOX=1 if he daa year is laer han 2002, oherwise SOX=0. The coefficien for SOX is expeced o be negaive because Sarbanes-Oxley Ac (SOX) is proven effecive in lowering he level of accrual managemen (Cohen e al., 2008). 3.3.4 Regression The regression model is shown in he equaion below: DIS _ AC a a 5 10 ROA Year a VOLAT GEO _ SEG PCM 6 0 Indusry a 1 11 a a 7 SOX a Exchange 2 LEVER SIZE 12 a PCM a 8 3 MTB EXTER SOX a 4 GROWH a BUS _ SEG 9 (3) In Equaion 3, he subscrip i refers o firm he subscrip refers o ime in years. I consider year, indusry and exchange effecs in he model. I run he fixed-effec regression on he panel daa se from 1997 o 2007. Following Peersen (2009), I correc he unobserved firm effec in he calculaion of sandard errors. 61

3.4 Empirical ess 3.4.1 Sample selecion I collec he business segmens and geographical segmens daa from COMPUSTAT Indusrial Segmen daabase. I obain financial daa from he COMPUSTAT Annual Indusrial and Research Files o calculae discreionary accrual and produc marke power and oher conrol variables. I require ha necessary inpus be available o calculae he dependen, independen and conrol variables. I also require a leas 10 observaions in each 2-digi SIC grouping per year. I exclude firm-year observaions wih deflaed absolue value of oal accruals greaer han oal asses because i is likely ha such observaions are due o recording errors (Kohari e al., 2005). Because PCM is defined as he raio of operaing profi o sales, i canno be greaer han 1. So I resric he sample o observaions wih PCM less han 1. I furher discard observaions wih boom 1% values of PCM because hey have exreme values as negaive as several housand. For he similar reason, I eliminae he op 1% values of VOLAT because hese firm-year observaions have absolue values of VOLAT around 4000. Ulimaely, I have 35,745 firm-year observaions wih 6,841 firms spanning from 1997 o 2007, one year before he financial crisis. 3.4.2 Descripive saisics Table 3.1 presens he descripive saisics for he dependen variable, independen variable and conrol variables. The level of asse-deflaed discreional 62

accruals has he mean 0.0889 and median 0.0567. Price-cos margin, PCM, has he mean 0.0415 and median 0.0948. Marke share, MKT_SH, has he mean 0.0678 and median 0.0087. Table 3.2 displays he Pearson correlaion coefficiens beween variables. Boh price-cos margin and marke share exhibi a negaive correlaion wih he absolue value of asse-deflaed discreionary accruals. Table 3.3 presens he frequency disribuion of he samples. 3.4.3 Main resuls Table 3.4 shows he regression resuls. Column 1 of Table 3.4 presens resuls based on Equaion 3 wihou conrol variables. The coefficien on PCM suggess one percenage poin increase in PCM is associaed wih a 5 basis poin decrease in he absolue value of deflaed discreionary accruals. A movemen from he 25h percenile of PCM (0.0255) o he 75h percenile (0.1696) would be associaed wih a 72 (=(0.1696-0.0255)*100*5) basis poin decrease in he absolue value of deflaed discreionary accruals. Column 2 of Table 3.4 presens resuls based on Equaion 3 wih conrol variables only. Leverage and number of geographical segmens do no show significan relaion wih he level of discreionary accruals, alhough he correlaion of number of geographical segmens wih accrual managemen has he expeced sign. All oher esimaed coefficiens have he expeced signs, and mos of resuls are significan a 1% excep MTB ( saisics=1.75, significan 63

level=10%) and number of business segmens ( saisics=2.31, significance level=5%). Column 3 of Table 3.4 presens he regression resuls of Equaion 3. Adding conrol variables reduces he magniude bu does no affec he saisical significance of he esimaed coefficien on PCM. The esimaed coefficien on PCM is -0.0241 ( saisics: -4.79). The coefficiens for he conrol variables also remain qualiaively he same as hey are in Column 2. I have an ineracion iem beween PCM and SOX in he regression. I expec ha he impac of marke power on accrual managemen is less when liigaion risk is more severe because low PCM firms end o less engage in accrual managemen, while he accrual managemen pracice of high PCM firms does no change much afer he passage of SOX. As expeced, I find ha afer he passage of SOX, he differences in he accrual managemen pracice beween high and low PCM firms become less. The resuls sugges ha he negaive relaion beween marke power and he magniude of discreionary accruals is no driven by he operaing environmens or firms growh opporuniies. 3.4.4 Addiional resuls 3.4.4.1 Alernaive measure of marke power As a robusness check, I inroduce an alernaive measure of marke power ino he empirical es. Marke share, MKT_SH, is calculaed as he raio of he 64

firm s sales o oal sales in he same 4-digi indusry. Such a measure is widely used by cours and governmen agencies o proxy marke power. Is descripive saisics are also included in Table 1. MKT_SH has he mean 0.0678 and median 0.0087. I repea he same seps described above wih MKT_SH in replace of PCM in Equaion 3. Table 3.5 displays he regression resuls. Column 1 of Table 3.5 presens he regression resuls of absolue value of discreionary accruals on MKT_SH. I find ha a one percenage poin increase in MKT_SH is associaed wih 3 basis poin decrease in he absolue value of deflaed discreionary accruals. Column 2 of Table 3.5 presens he regression resuls on conrol variables only. Leverage and he number of geographical segmens do no show significan relaion wih he level of discreionary accruals. All oher esimaed coefficiens have he expeced signs, and mos of resuls are significan a 1% excep MTB ( saisics=1.75, significance level=10%) and he number of business segmens ( saisics=2.31, significance level=5%). Column 3 of Table 3.5 displays he regression resuls of Equaion 3, where PCM is replaced wih MKT_SH. Alhough adding conrol variables decreases he saisical significance of he esimaed coefficien for MKT_SH, he sign of he esimaed coefficien for MKT_SH is sill negaive. The esimaed coefficien for MKT_SH is -0.0055 ( saisics: -1.32). The coefficiens for he conrol variables also remain qualiaively he same as hey are in Column 2. Again, I have he ineracion iem beween MKT_SH and SOX in he regression. As expeced, he 65

impac of marke power on accrual managemen becomes less when liigaion risk is more severe because firms wih low marke shares end o engage in less accrual managemen afer SOX, while he passage of SOX does no affec he accrual managemen pracice of high PCM firms. I furher examine he join impac of price-cos margin and marke share on he level of discreionary accruals. Table 3.6 presens he esimaes of he regression wih boh PCM and MKT_SH as independen variables. I find ha he negaive relaion beween PCM and he magniude of discreionary accruals remains significan, while he correlaion of MKT_SH and DIS_AC is saisically insignifican. Thus, I conjecure ha price-cos margin is he economic linkage ha underlies he correlaion beween marke share and he level of discreionary accrual. In addiion, following he logic in Ali e al (2009), we can see ha MKT_SH is acually a biased measure o describe a firm s marke power because MKT_SH is calculaed using daa from COMPUSTAT, bu COMPUSTAT only includes informaion abou public firms. For some indusries, privae firms consiue a significan par of he whole indusry. Evidence shows ha he impac of marke power on he accrual managemen becomes less afer he passage of SOX. 3.4.4.2 Indusry-adjused measure of marke power Following Gaspar and Massa (2005), I subrac he indusry mean and median PCM o conrol for cross-indusry effecs. This allows me o ge rid of indusry-specific facors unrelaed o marke power. Table 3.7 repors he resuls 66

wih indusry-adjused marke power, PCM_ADJ, as an independen variable. When PCM_ADJ is calculaed by subracing he indusry median and mean from he firm s PCM respecively, he coefficiens on PCM_ADJ are -0.0235 and - 0.0228, boh wih a significance level of 1% ( saisics=-4.63 and -4.51). Coefficiens on he conrol variables are qualiaively he same as wha is found in he regression wihou indusry adjusmen 1. 3.4.4.3 Alernaive measures of discreionary Accrual Ball and Shivakumar (2006) improve he discreionary accrual measure by conrolling for asymmeric imeliness of accruals in recognizing gains and losses. Their model is described below: TA ~ DCFO 4 ~ 0 1 A 1 2 3 1 A 1 A 1 A 1 (4) ~ REV ~ CFO 5 A 1 AR * DCFO ~ PPE ~ CFO CFO represens cash flows from operaion. DCFO is defined as a dummy variable ha equals 1 if CFO is negaive and 0 oherwise. All oher variables are he same as previously defined in Equaions 1 and 2. Sill, I require a leas 10 observaions for each wo-digi SIC indusry and year. Thus, for he same sample in he previous secion, I obain DIS_AC2. DIS_AC2 denoes he absolue value of he difference beween acual asse-deflaed accruals and he fied values of he deflaed accruals in Equaion 4. I also use DIS_AC2 as alernaive measure of he 1 All ess in his paper have qualiaively similar resuls wih indusry-adjused marke power measure. 67

dependen variable in Equaion 3. Table 3.8 Panel A presens he descripive saisics for DIS_AC2. Table 3.8 Panel B repors he esimaes of he regression. The firs column shows he resuls wih only conrol variables. All coefficiens have he esimaed signs. Mos of hem are saisically significan excep hose for leverage and he number of business/geographical segmens. The second and hird columns show he resuls wih PCM and MKT_SH, respecively. The coefficien on PCM is significanly negaive a he 1 percen level wih he magniude of -0.0281 and =- 5.82. The coefficien on MKT_SH has he magniude of -0.0073 and =-1.81. The coefficiens on conrol variables remain qualiaively similar in heir magniudes and significance levels. 3.4.4.4 Sign of discreionary accruals I classify samples o POSITIVE and NEGATIVE groups based on he signs of firms discreionary accruals and run he regressions for hese wo groups independenly. Table 3.9 repors he regression resuls. The coefficiens on PCM are significanly negaive for boh groups. Esimaed coefficiens are -0.0121 ( saisics=-1.75) and -0.0289 ( saisics=-3.36) for Group POSITIVE and NEGATIVE, respecively. This indicaes ha low PCM firms end o manipulae earnings more han high PCM firms do no only o boos earnings, bu also o reduce earnings. This finding could clear he concern ha high PCM firms end o engage in less accrual managemen because heir profis are guaraneed. If guaraneed profis due o monopoly saus were he underlying reason, he negaive 68

relaion beween PCM and accrual managemen for POSITIVE group only should have been demonsraed. As Yu (2008) shows, he sign of esimaed coefficien on ROA is posiive for he POSITIVE group and is negaive for he NEGATIVE group. As for he leverage, for firms ha ry o boos heir earnings, more leverage means more monioring, hus deerring managers accrual manipulaion. No significan relaion beween leverage and accrual manipulaion is found for firms wih need o reduce heir earnings via accrual managemen. 3.4.4.5 Indusry-inheren liigaion risk In order o examine he effec of liigaion risk on he relaion beween produc marke power and accrual managemen, I run he regressions for indusries wih differen liigaion risk separaely. Following Francis e al. (1994), I classify samples o HIGH_RISK and LOW_RISK groups based on heir inheren liigaion risk. Bioechnology (SIC codes 2833-2836), compuers (SIC codes 3570-3577), elecronics (SIC codes 3600-3674), and reailing (SIC codes 5200-5961) are idenified as high liigaion risk indusries, while ohers are idenified as low liigaion risk indusries. I run he regressions for hese wo groups independenly. Table 3.10 repors he regression resuls. The coefficien on PCM in low liigaion risk indusries has a significanly negaive risk a 1% wih a magniude of -0.0220 and saisics of -4.32. The relaion beween PCM and accrual managemen in high liigaion risk indusries is less significan han ha found in low liigaion risk indusries. Is coefficien is -0.0093, and saisics is -0.81. 69

The ineracion erm of SOX and PCM is a ime-series es on he effec of he liigaion risk on he relaion beween PCM and accrual managemen. Table 3.10 shows a cross-secional version of he same es. Boh ime-series and crosssecional ess describe ha low PCM firms change heir accrual managemen pracices more wih he liigaion risk han high PCM firms. In oher words, managers in high PCM firms are no sensiive o liigaion risk in deciding heir accrual managemen policy. 3.4.4.6 Indusry-level compeiion I run regression wih an indusry-level compeiion measure in replace of firm-level compeiion level as anoher robusness es. Imagine an indusry wih inensive compeiion. Inensive compeiion implies less agency conflics beween managers and shareholders. Then managers in his indusry have less incenive o engage in earnings managemen. In he meanime, firms in an indusry wih greaer compeiion end o have less price-seing abiliies, in general, so hey have o rely on accrual managemen if hey have o mee or bea earnings expecaions. So, again, indusry-level compeiion also has mixed predicions on he effec of accrual managemen. I use he Herfindahl-Hirschman index from he U.S. Census of Manufacurers, which covers boh public and privae firms in an indusry, as he indusry-level measure of compeiiveness. Herfindahl-Hirschman index (HIndex) is defined as he sum of he square of percenage marke share. Greaer HIndex means less indusry-level compeiion inensiy. The U.S. Census Bureau repors 70

concenraion raios for hundreds of indusries in he manufacuring secor in heir Census of Manufacures Publicaions. A U.S. Census akes place every five years. The mos recen wo were in 1997 and 2002. Following prior lieraure (Ali e al., 2009), I assume he indusry concenraion level does no change rapidly, so I use he concenraion raio in he 1997 and 2002 Censuses as concenraion raios from 1995-1999 and from 2000-2004. Boh of hese imeframes are from wo years before a survey year o wo years afer such a survey year. Table 3.11 repors he regression resuls of accrual managemen on Herfindahl-Hirschman index. The daa samples are limied o hose in manufacuring indusry. There are 13,289 firm-year observaions in he final sample. Mos oher conrol variables sill have a qualiaively similar effec on accrual managemen, and HIndex has a saisically significan, posiive impac on accrual managemen. This shows ha if all indusries have averaged same need o mee or bea earnings expecaions, firms in a more compeiive indusry end o have less price-seing abiliy and less agency conflics, so firms in a more compeiive indusry end o manage heir accruals less. 3.4.4.7Audior indusry specializaion I consider he audi marke compeiion in my regression. I examine wheher naional indusry leadership and specializaion in audi marke affec he relaion beween marke power and discreionary accruals. Following Francis e al. (2005), I inroduce wo variables o calculae audior indusry specializaion a he naional level: IND_SPE1 and IND_SPE2. These wo measures are based on he 71

assumpion ha indusry experise increases in marke share. IND_SPE1 is a dummy variable. I equals o 1 if he audior is a naional indusry specialis and 0 oherwise. IND_SPE2 is he marke share wihin a wo-digi SIC indusry. Panel A in Table 3.12 shows he resuls wih IND_SPE1. We can see ha he relaion beween produc marke power and discreionary accruals sill holds wih he inclusion of dummy-version audior specializaion. Dummy version of audior indusry specializaion has insignificanly expeced negaive effec on he magniude of discreionary accruals. Panel B in Table 3.12 repors he resuls wih IND_SPE2. Sill, he relaion beween marke power and discreionary accruals holds wih he inclusion of coninuous-version audior specializaion. IND_SPE2 has significanly negaive effec on accrual managemen, bu his effec is subsumed by BIG_FIVE. This ells ha a firm ha hire a Big Five audior firm end o have less discreionary accruals no maer wheher his audior firm has specialized audiing skills in he clien s indusry. 3.5 Conclusions By idenifying marke power as a significan deerminan of he levels of discreionary accruals, his essay provides evidence ha firms wih greaer marke power end o manage heir earnings less via accrual managemen because hey are in a comparaively privileged marke saus enabling hem o ransfer heir produciviy shocks o cusomers by seing prices. An imporan message of his 72

paper is ha a firm s marke power provides one way o boos or reduce earnings in addiion o accrual managemen. This essay is he firs sudy ha examines he impac of a firm marke power on is financial reporing qualiy. There are many opporuniies for fuure research on marke power and earnings managemen. For example, I discuss he benefis of shock-ransfer over accrual managemen, such as less liigaion, scruiny and audi risk. However, I do no explore he disadvanage or limiaion of shock-ransfer o boos or decrease earnings. This begs a number of quesions, including wheher his ransfer will weaken a firm s compeiion saus, wha he comparaive coss or benefis o sakeholders beween price-ransfer o cusomers and earnings managemen are, or how he capial marke will respond o his behavior, ec. Recen sudies show ha managers have shifed from accrual o real managemen in his pos Sarbanes-Oxley Ac period. My sudy only considers he accrual managemen and produc marke power, so far. I is of imporance o examine he real managemen behaviors of firms wih differen marke powers. For example, does i sill hold for real managemen ha firms wih greaer power end o engage less in real managemen aciviies? Alhough shock-ransfer aciviies aken by firms wih greaer power do no affec heir normal operaional pracices as real manipulaions do, wha are he economic consequences of a firm s ransfer shocks o cusomers? Wha oher facors decide heir abiliies o ransfer produciviy shocks hrough price-seing? 73

To summarize, his essay documens he relaion beween a firm s marke power and is incenives o manage earnings hrough accrual managemen. I idenifies anoher channel hrough which firms wih grea marke power can alleviae managers pressure o manage earnings hrough accruals. I also complemens he exising lieraure regarding he impacs of compeiion on he capial marke. Regulaors and audiors will benefi from his sudy because i helps hem undersand where he accrual managemens are more pervasive. 74

References Abarbanell, J. and R. Lehavy. 2003. Can sock recommendaions predic earnings managemen and analyss earnings forecas errors? Journal of Accouning Research 41: 1-31. Al A., S. Klasa, and E. Yeung.2009. The limiaions of indusry concenraion measures consruced wih Compusa daa: implicaions for finance research. Review of Financial Sudies 22: 3439-3871. Al A., S. Klasa, and E. Yeung. 2010. Indusry concenraion and corporae disclosure policy. Working paper. Universiy of Texas a Dallas. Ball, R. and L. Shivakumar. 2006. The role of accruals in asymmerically imely gain and loss recogniion. Journal of Accouning Research 44: 207-242. Bergsresser, D. and T. Philippon. 2006. CEO incenives and earnings managemen. Journal of Financial Economics 80: 511-529. Boosan, C and M. Sanford. 2005. Managers moives o wihhold segmen disclosures and he effec of SFAS no. 131 on analyss informaion environmen. The Accouning Review 80: 751-771. Burgsahler, D. and I. Dichev. 1997. Earnings managemen o avoid earnings decreases and losses. Journal of Accouning and Economics 24: 99-126. Carlon, D. and J. Perloff. 2000. Modern Indusrial Organizaion. Addison-Wesley. Massachuses. Cohen, D., A. Dey, and T. Lys. 2008. Real and accrual-based earnings managemen in he pre- and pos-sarbanes-oxley periods. The Accouning Review 83: 757-787. Collins, D. and P. Hribar. 2002. Errors in esimaing accruals: implicaions for empirical research. Journal of Accouning Research 40: 105-134. Dechow, P. and D. Skinner. 2000. Earnings managemen: reconciling he views of accouning academics, praciioners, and regulaors. Accouning Horizons 14: 235-250. Dechow, P., R. Sloan, and A. Sweeney. 1995. Deecing earnings managemen. Accouning Review 70: 193-225. 75

Dechow, P., R. Sloan, and A. Sweeney. 1996. Causes and consequences of earnings manipulaion: An analysis of firms subjec o enforcemen acions by he SEC. Conemporary Accouning Research 13: 1-36. DeFond, M. and J. Jiambalvo. 1994. Deb covenan effecs and he manipulaion of accruals. Journal of Accouning and Economics 17: 145-176. Degeorge, F., J. Pael, and R. Zeckhauser. 1999. Earnings managemen o exceed hresholds. Journal of Business 72: 1-33. Demers, E. and C. Wang. 2009. The impac of career concerns on accruals based and real earnings managemen. Working paper. INSEAD. Fee, C. and S. Thomas. 2004. Sources of gains in horizonal akeovers: evidence from cusomer, supplier, and rival firms. Journal of Financial Economics 74: 423-460. Francis, J. D. Philbrick, and K. Schipper. 1994. Shareholder liigaion and corporae disclosure. Journal of Accouning Research 32: 137-164. Francis, J., K. Reichel, and D. Wang. 2005. The pricing of naional and ciyspecific repuaions for indusry experise in he U.S. audi marke. The Accouning Review 80: 113-136. Gaspar, J. and M. Massa. 2005. Idiosyncraic volailiy and produc marke compeiion. Journal of Business 79: 3125-3152. Healy, P. 1985. The effec of bonus schemes on accouning decisions. Journal of Accouning and Economics 7: 85-107. Healy, P. and J. Wahlen. 1999. A review of he earnings managemen lieraure and is implicaions for sandard seing. Accouning Horizons 13: 365-383. Hou, K. and D. Robinson. 2006. Indusry concenraion and average sock reurns. Journal of Finance 4: 1927-1956. Jones, J. 1991. Earnings managemen during impor relief invesigaions. Journal of Accouning Research 29: 193-228. Kale, J.R. and Y.C. Loon. 2011. Produc marke power and sock marke liquidiy. Journal of Financial Markes 14: 376-410. Kohar S.P., A. Leone, and C. Wasley. 2005. Performance mached discreionary accrual measures. Journal of Accouning and Economics 39: 163-197. 76

Lerner, A.P.. 1934. The concep of monopoly and he measuremen of monopoly power. The Review of Economic Sudies 1: 157-175. Marciukaiye, D. and J. Park. 2009. Marke compeiion and earnings managemen. Working paper. Drexel Universiy. McFalls, M. 1997. The role and assessmen of classical marke power in join venure analysis. Federal Trade Commission, Saff Discussion Draf. hp:www.fc.gov/opp/joinven/classic8.shm. Peersen, M. 2009. Esimaing sandard errors in finance panel daa ses: comparing approaches. The Review of Financial Sudies 22: 435-480. Peress, J. 2010. Produc marke compeiion, insider rading, and sock marke efficiency. Journal of Finance 1: 1-43. Roychowdhury, S. 2006. Earnings managemen hrough real aciviies manipulaion. Journal of Accouning and Economics 42: 335-370. Teoh, A., I. Welch, and T. Wong. 1998a. Earnings managemen and he pos-issue performance of seasoned equiy offerings. Journal of Financial Economics 50: 63-99. Teoh, A., I. Welch, and T. Wong. 1998b. Earnings managemen and he long-erm marke performance of iniial public offerings. Journal of Finance 53: 1935-1974. Yu, F. 2008. Analys coverage and earnings managemen. Journal of Financial Economics 88: 245-271. Zang, A. 2006. The rade-off beween earnings managemen and accruals managemen. Working paper. Duke Universiy 77

Appendix Variables Definiions DIS_AC = he absolue value of asse-deflaed discreionary accrual. PCM = price-cos margin, calculaed as he raio of operaing profi o sales. MKT_SH = marke share, calculaed as he raio of he firm s sales o oal sales in he same 4-digi SIC indusry. SIZE = firm size, esimaed as he naural log of marke value of equiy. MTB = marke-o-book raio, measured as he raio of marke value of a firm o oal asses, where marke value is oal asses plus marke value of common equiy minus book value of common equiy. GROWTH = growh rae of asses, measured as he change of asses scaled by lagged asses. ROA = reurn on asses, measured as he earnings before ineres and ax divided by oal asses. VOLAT = sandard deviaion of annual asse-deflaed cash flow growh over las five years. EXTER = exernal financing aciviies, measured as he sum of ne cash received from equiy and deb issuance scaled by oal asses. LEVER = leverage, calculaed as he raio of long-erm deb o oal asses. BUS_SEG = he number of business segmens. GEO_SEG = he number of geographical segmens. SOX = 1 if he daa year is laer han 2002, oherwise 0. 78

Table 3.1 Descripive saisics The sample consiss of 34414 firm-years over he period from 1997 o 2007. This able presens he descripive saisics of he dependen variable DIS_AC, he independen variable PCM, and conrol variables. Conrol variables include ROA (reurn on asses), EXTER (exernal financing aciviies), LEVER (leverage), GROWTH (growh rae of asses), MTB (marke-o-book raio), SIZE (firm size), and VOLAT (sandard deviaion of annual asse-deflaed cash flow growh over las five years). BUS_SEG and GEO_SEG are he number of business segmens and geographical segmens respecively. MKT_SH, marke share, is an alernaive proxy for marke power. N Mean Sd. Dev. 10% 25% 50% 75% 90% DIS_AC 35745 0.0889 0.1077 0.0102 0.0258 0.0567 0.1086 0.1964 PCM 35745 0.0415 0.3992-0.1418 0.0255 0.0948 0.1696 0.2799 MKT_SH 35745 0.0678 0.1497 0.0002 0.0087 0.0541 0.1996 ROA 35745-0.0460 0.8145-0.2693-0.0509 0.0288 0.0755 0.1257 EXTER 35745 0.0217 0.4171-0.1108-0.0482-0.0014 0.0538 0.1933 LEVER 35745 0.1834 0.2736 0.0000 0.0018 0.1117 0.2783 0.4539 GROWTH 35745 0.1360 0.6641-0.2046-0.0569 0.0504 0.1868 0.4433 MTB 35745 2.0634 4.5258 0.8823 1.0957 1.4718 2.2027 3.5064 SIZE 35745 5.1807 2.5449 1.9341 3.4099 5.2077 6.9085 8.4230 VOLAT 35745 0.1269 0.1337 0.0288 0.0486 0.0865 0.1533 0.2616 BUS_SEG 35745 2.3785 1.8902 1 1 1 3 5 GEO_SEG 35745 2.8399 2.2050 1 1 2 4 6 79

Table 3.2 Correlaion marix of key variables This able presens he correlaion marix of key variables in he paper. The sample consiss of 35745 firm-years over he period from 1997 o 2007. The variables include he dependen variable DIS_AC, he independen variable PCM, and conrol variables. Conrol variables include ROA (reurn on asses), EXTER (exernal financing aciviies), LEVER (leverage), GROWTH (growh rae of asses), MTB (marke-o-book raio), SIZE (firm size), and VOLAT (sandard deviaion of annual asse-deflaed cash flow growh over las five years). BUS_SEG and GEO_SEG are he number of business segmens and geographical segmens respecively. MKT_SH, marke share, is an alernaive proxy for marke power. DIS_AC PCM MKT_SH ROA EXTER LEVER GROWTH MTB SIZE VOLAT BUS_SEG DIS_AC 1.0000 PCM -0.2430 1.0000 MKT_SH -0.1229 0.1054 1.0000 ROA -0.1518 0.2353 0.0450 1.0000 EXTER 0.0936-0.1813-0.0399 0.5328 1.0000 LEVER 0.0036 0.0303 0.0556-0.0605 0.0504 1.0000 GROWTH 0.1107 0.0728-0.0035 0.0700 0.1317 0.0009 1.0000 MTB 0.0860-0.1033-0.0219 0.1053 0.1094 0.0588 0.0258 1.0000 SIZE -0.2501 0.2874 0.3587 0.1186-0.0448-0.0197 0.1261 0.0611 1.0000 VOLAT 0.3311-0.2957-0.1893-0.1116 0.1268-0.0605 0.0677 0.1274-0.3328 1.0000 BUS_SEG -0.0951 0.0961 0.2560 0.0349-0.0324 0.0322 0.0032-0.0446 0.3285-0.1804 1.0000 GEO_SEG -0.0572 0.0511 0.1127 0.0192-0.0102-0.0625 0.0019 0.0095 0.2925-0.0891 0.2106 80

Table 3.3 Frequency disribuion The sample consiss of 35745 firm-years over he period from 1997 o 2007. This able presens he frequency disribuion of he samples in he paper Year No. of Indusries No. of Firms 1997 63 3785 1998 63 3756 1999 62 3004 2000 63 3163 2001 63 3187 2002 64 3205 2003 63 3317 2004 63 3336 2005 63 3169 2006 64 2980 2007 63 2843

Table 3.4 The effec of price-cos margin on accrual managemen This able presens he effec of marke power on accrual managemen. Dependen variable is he absolue value of asse-deflaed discreionary accrual. Independen variable is price-cos margin. Conrol variables include firm size, marke-o-book raio, growh rae of asses, reurn on asses, sandard deviaion of annual asse-deflaed cash flow growh, leverage, exernal financing aciviies, number of business segmens, and number of geographical segmens. The Sarbanes-Oxley Ac is also considered in he regression. Sandard errors are adjused for firm clusering and saisics are repored in brackes. ***, **, and * represen 1%, 5%, and 10% significance respecively. Dependen variable: DIS_AC PCM ( ) 0.0503*** 0.0241*** [ 16.84] [ 4.79] SIZE ( ) 0.0058*** 0.0053*** [ 11.38] [ 11.17] MTB (+) 0.0016* 0.0014* [1.75] [1.73] GROWTH (+) 0.0179*** 0.0188*** [6.68] [6.73] ROA ( ) 0.0244*** 0.0202*** [ 3.97] [ 3.43] VOLAT (+) 0.1582*** 0.1503*** [15.86] [16.03] LEVER (?) 0.0038 0.0017 [ 1.10] [ 0.50] EXTER (+) 0.0337*** 0.0263*** [5.06] [4.37] BUS_SEG (+) 0.0006** 0.0006* [2.00] [1.83] GEO_SEG (+) 0.0003 0.0003 [1.09] [1.13] SOX ( ) 0.0067*** 0.0080*** [ 2.82] [ 3.25] SOX*PCM (+) 0.0073 [1.29] Consan 0.1292*** 0.1098*** 0.1100*** [10.20] [10.06] [10.04] EXCHANGE Included Included Included INDUSTRY Included Included Included YEAR Included Included Included N 35745 35745 35745 Adj. R^2 12.33% 19.66% 20.11% 82

Table 3.5 The effec of marke share on accrual managemen This able presens he effec of marke share on accrual managemen. Dependen variable is he absolue value of asse-deflaed discreionary accrual. Independen variable is marke share. Conrol variables include firm size, marke-o-book raio, growh rae of asses, reurn on asses, sandard deviaion of annual asse-deflaed cash flow growh, leverage, exernal financing aciviies, number of business segmens, and number of geographical segmens. The Sarbanes-Oxley Ac is also considered in he regression. Sandard errors are adjused for firm clusering and saisics are repored in brackes. ***, **, and * represen 1%, 5%, and 10% significance respecively. Dependen variable: DIS_AC MKT_SH ( ) 0.0367*** 0.0055 [ 9.25] [ 1.32] SIZE ( ) 0.0058*** 0.0058*** [ 11.38] [ 10.96] MTB (+) 0.0016* 0.0016* [1.75] [1.75] GROWTH (+) 0.0179*** 0.0179*** [6.68] [6.67] ROA ( ) 0.0244*** 0.0244*** [ 3.97] [ 3.97] VOLAT (+) 0.1582*** 0.1582*** [15.86] [15.87] LEVER (?) 0.0038 0.0037 [ 1.10] [ 1.09] EXTER (+) 0.0337*** 0.0337*** [5.06] [5.06] BUS_SEG (+) 0.0006 0.0006* [2.00]** [1.94] GEO_SEG (+) 0.0003 0.0003 [1.09] [1.08] SOX ( ) 0.0067*** 0.0078*** [ 2.82] [ 3.11] SOX*MKT_SH (+) 0.0138*** [2.69] Consan 0.1309*** 0.1098*** 0.1099*** [10.00] [10.06] [10.06] EXCHANGE Included Included Included INDUSTRY Included Included Included YEAR Included Included Included N 35745 35745 35745 Adj. R^2 9.43% 19.66% 19.67% 83

Table 3.6 The join effec of price-cos margin and marke share on accrual managemen This able presens he join effec of price-cos margin and marke share on accrual managemen. Conrol variables include firm size, marke-o-book raio, growh rae of asses, reurn on asses, sandard deviaion of annual asse-deflaed cash flow growh, leverage, exernal financing aciviies, number of business segmens, and number of geographical segmens. The Sarbanes-Oxley Ac is also considered in he regression. Sandard errors are adjused for firm clusering and saisics are repored in brackes. ***, **, and * represen 1%, 5%, and 10% significance respecively. Dependen variable: DIS_AC PCM ( ) 0.02096*** 0.0239*** [ 5.06] [ 4.75] MKT_SH ( ) 0.0004 0.0058 [ 0.10] [ 1.41] SIZE ( ) 0.0058*** 0.0053*** 0.0053*** [ 11.38] [ 10.72] [ 10.68] MTB (+) 0.0016* 0.0014* 0.0014* [1.75] [1.71] [1.73] GROWTH (+) 0.0179 0.0188*** 0.0188*** [6.68] [6.73] [6.72] ROA ( ) 0.0244*** 0.0203*** 0.0203*** [ 3.97] [ 3.43] [ 3.44] VOLAT (+) 0.1582*** 0.1502*** 0.1503*** [15.86] [16.02] [16.04] LEVER (?) 0.0038 0.0019 0.0017 [ 1.10] [ 0.54] [ 0.49] EXTER (+) 0.0337*** 0.0263*** 0.0263*** [5.06] [4.38] [4.37] BUS_SEG (+) 0.0006** 0.0006** 0.0006* [2.00] [1.82] [1.81] GEO_SEG (+) 0.0003 0.0003 0.0003 [1.09] [1.13] [1.12] SOX ( ) 0.0067*** 0.0076*** 0.0088*** [ 2.82] [ 3.17] [ 3.48] SOX*PCM (+) 0.0068 [1.20] SOX*MKT_SH (+) 0.0120** [2.36] Consan 0.1098*** 0.1100*** 0.1100*** [10.06] [10.03] [10.04] EXCHANGE Included Included Included INDUSTRY Included Included Included YEAR Included Included Included N 35745 35745 35745 Adj. R^2 19.66% 20.09% 20.11% 84

Table 3.7 The effec of indusry-adjused marke power on accrual managemen This able presens he effec of indusry-adjused marke power on accrual managemen Conrol variables include firm size, marke-o-book raio, growh rae of asses, reurn on asses, sandard deviaion of annual asse-deflaed cash flow growh, leverage, exernal financing aciviies, number of business segmens, and number of geographical segmens. The Sarbanes-Oxley Ac is also considered in he regression. Sandard errors are adjused for firm clusering and saisics are repored in brackes. ***, **, and * represen 1%, 5%, and 10% significance respecively. Dependen variable: DIS_AC Median Adjused Mean Adjused PCM_ADJ ( ) 0.0503*** 0.0235*** 0.0502*** 0.0228*** [ 16.74] [ 4.63] [ 16.67] [ 4.51] SIZE ( ) 0.0053*** 0.0053*** [ 11.20] [ 11.19] MTB (+) 0.0014* 0.0014* [1.72] [1.72] GROWTH (+) 0.0188*** 0.0187*** [6.74] [6.76] ROA ( ) 0.0203*** 0.0204*** [ 3.44] [ 3.45] VOLAT (+) 0.1503*** 0.1504*** [16.03] [16.03] LEVER (?) 0.0017 0.0018 [ 0.51] [ 0.52] EXTER (+) 0.0263*** 0.0266*** [4.38] [4.40] BUS_SEG (+) 0.0006* 0.0006* [1.85] [1.85] GEO_SEG (+) 0.0003 0.0004 [1.17] [1.19] SOX ( ) 0.0075*** 0.0079*** [ 3.18] [ 3.28] SOX*PCM (+) 0.0062 0.0051 [1.07] [0.88] Consan 0.1247*** 0.1078*** 0.1272*** 0.1090*** [9.84] [9.86] [10.03] [9.96] EXCHANGE Included Included Included Included INDUSTRY Included Included Included Included YEAR Included Included Included Included N 35745 35745 35745 35745 Adj. R^2 12.31% 20.09% 12.26% 20.08% 85

Table 3.8 Resuls wih alernaive measure of discreionary accrual Panel A: Descripive saisics This able presens he descripive saisics of he dependen variable DIS_AC2 suggesed in Ball and Shivakumar (2006). N Mean Sd. Dev. 10% 25% 50% 75% 90% 35745 0.0800 0.1050 0.0079 0.0204 0.0463 0.0955 0.1840

Table 3.8 Resuls wih alernaive measure of discreionary accrual Panel B: Reexaminaion on he effec of marke power on accrual managemen This able presens he effec of marke power on accrual managemen. Dependen variable is he absolue value of asse-deflaed discreionary accrual in Ball and Shivakumar (2006). Independen variable is price-cos margin or marke share. Conrol variables include firm size, marke-o-book raio, growh rae of asses, reurn on asses, sandard deviaion of annual asse-deflaed cash flow growh, leverage, exernal financing aciviies, number of business segmens, and number of geographical segmens. The Sarbanes-Oxley Ac is also considered in he regression. Sandard errors are adjused for firm clusering and saisics are repored in brackes. ***, **, and * represen 1%, 5%, and 10% significance respecively. Dependen variable: DIS_AC2 PCM MKT_SH PCM/MKT_SH ( ) 0.0281*** 0.0073* [ 5.82] [ 1.81] SIZE ( ) 0.0060*** 0.0054*** 0.0060*** [ 11.73] [ 11.41] [ 11.30] MTB (+) 0.0017* 0.0015* 0.0017* [1.80] [1.78] [1.80] GROWTH (+) 0.0146*** 0.0156*** 0.0146*** [6.27] [6.42] [6.27] ROA ( ) 0.0255*** 0.0209*** 0.0255*** [ 4.10] [ 3.56] [ 4.11] VOLAT (+) 0.1467*** 0.1380*** 0.1468*** [15.18] [15.31] [15.19] LEVER (?) 0.0044 0.0021 0.0044 [ 1.30] [ 0.61] [ 1.29] EXTER (+) 0.0353*** 0.0270*** 0.0353*** [5.61] [4.87] [5.61] BUS_SEG (+) 0.0001 0.0002 0.0001 [0.38] [0.16] [0.29] GEO_SEG (+) 0.0004 0.0004 0.0004 [1.42] [1.47] [1.41] SOX ( ) 0.0040* 0.0055** 0.0054** [ 1.72] [ 2.32] [ 2.25] SOX*PCM/MKT_SH (+) 0.0109** 0.0196*** [2.03] [3.99] Consan 0.1033*** 0.1034*** 0.1035*** [9.59] [9.57] [9.60] EXCHANGE Included Included Included INDUSTRY Included Included Included YEAR Included Included Included N 35745 35745 35745 Adj. R^2 20.41% 21.01% 20.43% 87

Table 3.9 The effec of marke power on accrual managemen in groups wih posiive and negaive discreionary accruals This able presens he effec of marke power on accrual managemen in firms wih posiive and negaive discreionary accruals. Dependen variable is he absolue value of asse-deflaed discreionary accrual. Independen variable is price-cos margin. Conrol variables include firm size, marke-o-book raio, growh rae of asses, reurn on asses, sandard deviaion of annual asse-deflaed cash flow growh, leverage, exernal financing aciviies, number of business segmens, and number of geographical segmens. The Sarbanes-Oxley Ac is also considered in he regression. Sandard errors are adjused for firm clusering and saisics are repored in brackes. ***, **, and * represen 1%, 5%, and 10% significance respecively. Dependen variable: DIS_AC POSITIVE NEGATIVE PCM ( ) 0.0121* 0.02889*** [ 1.75] [ 3.36] SIZE ( ) 0.0047*** 0.0064*** [ 10.02] [ 7.24] MTB (+) 0.0011* 0.0017 [1.85] [1.48] GROWTH (+) 0.0308*** 0.0123*** [5.24] [4.31] ROA ( ) 0.0284* 0.0353*** [1.72] [ 2.93] VOLAT (+) 0.1446*** 0.1386*** [13.41] [9.71] LEVER (?) 0.0099*** 0.0026 [ 2.62] [0.45] EXTER (+) 0.0139 0.0486*** [1.21] [4.65] BUS_SEG (+) 0.0006* 0.0007 [1.87] [1.13] GEO_SEG (+) 0.0004 0.0000 [1.20] [ 0.08] SOX ( ) 0.0040 0.0125*** [ 1.39] [ 3.00] SOX*PCM (+) 0.0014 0.0094 [0.21] [1.21] Consan 0.0760*** 0.1474*** [6.05] [8.04] EXCHANGE Included Included INDUSTRY Included Included YEAR Included Included N 22065 13680 Adj. R^2 17.69% 28.00% 88

Table 3.10 The effec of marke power on accrual managemen in indusries wih differen liigaion risks This able presens he effec of marke power on accrual managemen in indusries wih differen liigaion risks. Conrol variables include firm size, marke-o-book raio, growh rae of asses, reurn on asses, sandard deviaion of annual asse-deflaed cash flow growh, leverage, exernal financing aciviies, number of business segmens, and number of geographical segmens. The Sarbanes-Oxley Ac is also considered in he regression. Sandard errors are adjused for firm clusering and saisics are repored in brackes. ***, **, and * represen 1%, 5%, and 10% significance respecively. Dependen variable: DIS_AC HIGH RISK LOW RISK PCM ( ) 0.0093 0.0220*** [ 0.81] [ 4.32] SIZE ( ) 0.0042*** 0.0063*** [ 5.18] [ 12.19] MTB (+) 0.0004** 0.0042*** [2.46] [3.96] GROWTH (+) 0.0459*** 0.0163*** [5.55] [6.06] ROA ( ) 0.0662*** 0.0218*** [ 2.72] [ 5.14] VOLAT (+) 0.1383*** 0.1405*** [8.41] [13.21] LEVER (?) 0.0006 0.0029 [0.09] [ 0.69] EXTER (+) 0.0342 0.0274*** [ 1.35] [5.57] BUS_SEG (+) 0.0008 0.0008** [1.20] [2.33] GEO_SEG (+) 0.0000 0.0006* [0.01] [1.66] SOX ( ) 0.0078 0.0078*** [ 1.46] [ 2.83] SOX*PCM (+) 0.0014 0.0062 [0.21] [0.95] Consan 0.4742*** 0.1090*** [15.93] [9.18] EXCHANGE Included Included INDUSTRY Included Included YEAR Included Included N 8463 27282 Adj. R^2 22.11% 20.98% 89

Table 3.11 The effec of indusry-level compeiion on accrual managemen This able presens he effec of indusry-level compeiion on accrual managemen. The sample consiss of 13289 firm-years. HIndex is Herfindahl-Hirschman index downloaded from he U.S. 1997 and 2002 Census of Manufacurers. Conrol variables include firm size, marke-o-book raio, growh rae of asses, reurn on asses, sandard deviaion of annual asse-deflaed cash flow growh, leverage, exernal financing aciviies, number of business segmens, number of geographical segmens, and passage of he Sarbanes-Oxley Ac. Sandard errors are adjused for firm clusering and saisics are repored in brackes. ***, **, and * represen 1%, 5%, and 10% significance respecively. Prediced Esimae Saisics HIndex (?) 0.0333* 1.93 SIZE ( ) 0.0038*** 4.47 MTB (+) 0.0016 1.33 GROWTH (+) 0.0405*** 5.93 ROA ( ) 0.0576*** 2.74 VOLAT (+) 0.1671*** 10.68 LEVER (?) 0.0147*** 3.07 EXTER (+) 0.0068 0.31 BUS_SEG (+) 0.0004 0.66 GEO_SEG (+) 0.0003 0.55 SOX ( ) 0.0053 1.55 Consan 0.0941*** 6.71 EXCHANGE Included YEAR Included N 13289 Adj. R^2 21.51% 90

Table 3.12 Panel A: Regression of discreionary accruals on marke power wih audior specializaion The able includes he regression resuls of accrual managemen on marke power wih dummy version of audior specializaion, IND_SPE1. Conrol variables are described as in previous ables. Sandard errors are adjused for firm clusering and saisics are repored in brackes. ***, **, and * represen 1%, 5%, and 10% significance respecively. Dependen variable: DIS_AC PCM ( ) 0.0515** 0.0515** 0.0529** [ 2.26] [ 2.26] [ 2.32] SIZE ( ) 0.0059*** 0.0059*** 0.0052*** [ 7.76] [ 7.69] [ 6.60] MTB (+) 0.0064*** 0.0064*** 0.0061*** [4.38] [4.38] [4.29] GROWTH (+) 0.0330*** 0.0330*** 0.0327*** [3.60] [3.60] [3.57] ROA ( ) 0.0384 0.0384 0.0388 [ 1.29] [ 1.29] [ 1.30] VOLAT (+) 0.0604*** 0.0604*** 0.0603*** [4.71] [4.71] [4.74] LEVER (?) 0.0022 0.0022 0.0028 [0.32] [0.32] [0.41] EXTER (+) 0.0094 0.0094 0.0095 [ 0.50] [ 0.50] [ 0.51] BUS_SEG (+) 0.0003 0.0004 0.0003 [0.79] [0.79] [0.70] GEO_SEG (+) 0.0001 0.0001 0.0001 [0.12] [0.12] [0.24] SOX ( ) 0.0084** 0.0085** 0.0111*** [ 2.09] [ 2.10] [ 2.78] SOX*PCM (+) 0.0023 0.0024 0.0039 [0.17] [0.17] [0.29] IND_SPE1 ( ) 0.0006 0.0011 [ 0.38] [0.68] BIG_FIVE ( ) 0.0101*** [ 3.68] Consan 0.0473 0.0477 0.0539 [1.29] [1.31] [1.46] EXCHANGE Included Included Included INDUSTRY Included Included Included YEAR Included Included Included N 14304 14304 14304 Adj. R^2 18.75% 18.75% 18.86% 91

Table 3.12 Panel B: Regression of discreionary accruals on marke power wih audior specializaion The able includes he regression resuls of accrual managemen on marke power wih coninuous version of audior specializaion, IND_SPE2. Conrol variables are described as in previous ables. Sandard errors are adjused for firm clusering and saisics are repored in brackes. ***, **, and * represen 1%, 5%, and 10% significance respecively. Dependen variable: DIS_AC PCM ( ) 0.0515** 0.0524** 0.0530** [ 2.26] [ 2.30] [ 2.32] SIZE ( ) 0.0059*** 0.0055*** 0.0051*** [ 7.76] [ 7.08] [ 6.54] MTB (+) 0.0064*** 0.0063*** 0.0061*** [4.38] [4.33] [4.28] GROWTH (+) 0.0330*** 0.0328*** 0.0327*** [3.60] [3.59] [3.57] ROA ( ) 0.0384 0.0386 0.0388 [ 1.29] [ 1.29] [ 1.30] VOLAT (+) 0.0604*** 0.0604*** 0.0604*** [4.71] [4.72] [4.73] LEVER (?) 0.0022 0.0025 0.0028 [0.32] [0.36] [0.41] EXTER (+) 0.0094 0.0095 0.0096 [ 0.50] [ 0.51] [ 0.51] BUS_SEG (+) 0.0003 0.0004 0.0003 [0.79] [0.80] [0.72] GEO_SEG (+) 0.0001 0.0001 0.0001 [0.12] [0.24] [0.27] SOX ( ) 0.0084** 0.0100** 0.0112*** [ 2.09] [ 2.47] [ 2.80] SOX*PCM (+) 0.0023 0.003 0.0039 [0.17] [0.21] [0.28] IND_SPE2 ( ) 0.0176*** 0.0029 [ 2.94] [ 0.45] BIG_FIVE ( ) 0.0091*** [ 2.96] Consan 0.0473 0.0523 0.0549 [1.29] [1.42] [1.49] EXCHANGE Included Included Included INDUSTRY Included Included Included YEAR Included Included Included N 14304 14304 14304 Adj. R^2 18.75% 18.80% 18.86% 92

Chaper 4 Essay III: Produc Marke Compeiion and Audi Fees 4.1 Inroducion Due o increased globalizaion and inensiy of impor peneraion, relaxaion of barriers o enry and rade, as well as he speed of echnological change, compeiion in produc marke is increasingly inense (Peress 2010; Gaspar and Massa 2006). Naurally, a quesion would arise regarding wha consequences of his change in produc marke on capial marke are. To answer his quesion, prior sudies have examined he effec of he increase in compeiion on he behaviors of some capial marke paricipans, such as managers, analyss, deb holders and equiy holders (Ali e al. 2009). This essay sheds new ligh on his quesion by performing analysis on he effecs of produc marke compeiion on audiors risk assessmen on cliens. I find ha audiors charge higher on firms in more compeiive indusries. Firms in indusries wih srong compeiion end o have differen behavior characerisics. This leads audiors o charge firms heerogeneously. On he one hand, produc marke compeiion plays a corporae governance role and miigaes agency problems beween managers and shareholders. Thus, he likelihood of earnings managemen decreases. Facing less audi risk, audiors could spend less resources and labor, hereby requiring lower audi fees. On he oher hand, firms in 93

compeiive indusries face more liquidiy, disress and liquidaion risk (Schumpeer 1912; Schmid 1997). The clien s business risk can, in urn, increase audiors business risk and, herefore, wha hey charge cliens. I leave as an empirical quesion he consideraion of conracing predicions. Employing he Herfindahl-Hirschman index downloaded from he U.S. Census of Manufacurers as a measure of an indusry s compeiion inensiy, I perform an indusry-level analysis on he 2201 firm-year observaions (760 firms) on he inersecion of he U.S. Census of Manufacurers, Audi Analyics daabase and Compusa from 2000 o 2004. I find ha firms in more compeiive indusries incur higher audi fees. This signifies ha, according o audiors viewpoins, he effec of produc marke compeiion on business risk prevails over ha on audi risk. In robusness ess, I es and find ha he relaion beween produc marke compeiion and audi fees sill holds afer conrolling for audi marke compeiion. In addiion o indusry-level analysis of produc marke compeiion on audi fees, I also perform firm-level analysis and invesigae wheher audiors change heir risk assessmen based on he cliens compeiion saus wihin he same indusry. Firms wih beer compeiion saus end o have earnings or cash flows wih less volailiy, hus inducing less business risk. They also have sronger abiliy o se prices so ha hey have less pressure o manipulae earnings hrough he channel of accrual managemen. So firms wih greaer marke power end o have less discreionary accruals and hus less audi risk. Therefore, boh business risk channel and audi risk channel have predicions of negaive relaion beween a 94

firm s compeiion saus and audi fees. As prediced, I find ha, wihin an indusry, audiors end o charge hose firms wih high produc marke power less. This essay makes several conribuions. Firs, his sudy is he firs one ha examines and explains he effec of indusry-level marke compeiion and firmlevel marke power on audi indusry. An increase in produc marke compeiion has caused researchers o become more ineresed in compeiion s effecs on he paricipans of capial markes. For example, managers invesmen decisions (Song and Waslking 2000; Fee and Thomas 2004), payou policy (Grullon and Michaely 2007), corporae disclosure decisions (Booson and Harris 2000; Booson and Sanford 2005; Rogers and Socker 2005; Verrecchia and Iber 2006), accouning choices (Zmijewski and Hagerman 1981), earnings qualiies (Dhaliwal e al. 2008; Marciukaiye and Park 2009; Wang 2011), analyss forecass characerisics (Ali e al. 2010), cos of deb (Vala 2010) and asse pricing (Hou and Robinson 2006) are all shown o be relaed wih marke compeiion in earlier work. As par of he broader lieraure ha links indusrial organizaion and issues in financial markes, his sudy provides evidence of he impac of produc marke compeiion on audi fees no well recognized in previous lieraure. By showing ha compeiion affecs audiors assessmen on firms risk, his essay enriches he lieraure abou he impac of produc marke compeiion on he capial markes by seing up he economic link beween compeiion and audi fees. Furhermore, i also shows ha audiors rea firms wih grea marke power differenly. Togeher wih prior research, his sudy proves ha he basic and inrinsic economic facor, compeiion in produc marke, conribues significanly o he capial marke. 95

Second, his sudy furher conribues o he audiing lieraure by clarifying confusions abou he exisence of indusry effecs in prior audiing lieraure and providing an economic explanaion and predicions for hem. There appears o be lile consensus in he audiing lieraure as o wheher here is an indusry effec on audi fees. Simunic (1980) claims ha: While loss exposure may well vary wih he indusry(ies) in which an audiee operaes, here is really no basis o hypohesize any specific indusry effecs. Some audiing papers follow his argumen and do no consider indusry-level deerminans of he audi fees (Griffin e al. 2010). Gul and Tsui (1997, 225) realize ha here are possible clien indusry effecs, bu hey jus use 1 o represen indusrials and 0 for ohers. They do no make any predicions on heir indusry dummy and do no find any significan resuls wih eiher. Francis (1984) finds a significan effec on financial insiuions only. Craswell e al. (1995) recognize he indusry-wide differences in audi fees resuling from specific knowledge in specialized accouning rules and reporing requiremens or frequency of complex conracs. Seeharaman e al. (2002) conrol for indusry effecs as a regular economeric procedure, bu fail o explore he underlying economic logic. This sudy no only provides an indusry-level deerminan of audi fees, bu also furher breaks he black box of his indusry-level difference and exposes he inside economic world o is readers. Third, his sudy is meaningful for firms, audiors and even regulaors. The resuls show ha firms in compeiive indusries induce audiors o charge hem 96

higher fees. Thus, wih an increase in produc marke compeiion, firm managers need o adjus heir budge for addiional audi fees accordingly. Wih an increase in compeiion in one cerain indusry, audiors also need o explain he reasons for charging heir cliens higher fees because an audi fee is usually he mos imporan deerminan in a firm s audior selecion (Eichenseher and Shields 1983). Furhermore, some changes in regulaions, such as he deregulaion of he airline indusry in he 1970s, he Bell Sysem divesiure in he 1980s or he European Union Single Marke Program in he 1990s changed he compeiion level. Wih an increase in produc marke compeiion, fund ransfer from firms o heir audiors is one non-negligible regulaion change effec. This sudy also helps regulaors by providing hem wih a horough predicion or assessmen on he consequences of regulaion changes. This essay is relaed o he recen paper by Levenis e al. (2011), bu differs from i in four respecs. Firs, my paper works on differen daa. Levenis e al. use proprieary daa for Greek firms, while he presen discussion uses a daase from he U.S. Compared wih he Ahen Sock Exchange (ASE), which was caegorized as an emerging marke unil May 2001, he U.S. marke is more maure and developed marke under a srong Anglo-American influence. Second, my paper employs a more reliable compeiion measure. Among four compeiion measures in Levenis e al. (2011), wo variables represening he percenage of indusry sales and concenraion raio in he conex of he ASE are significan. However, because of he limiaions poined ou in Ali e al. (2009), he concenraion compeiion measure consruced using daa for public lised firms is biased and 97

resuls based on such compeiion measure are quesionable. To avoid his problem, I use he Herfindahl-Hirschman index downloaded from he U.S. Census of Manufacurers o measure he compeiion level. Third, our papers consider differen heories. Audi risk from agency coss is he only channel hrough which produc marke compeiion affecs audi fees considered in Levenis e al. (2011), while my paper shows ha compared wih audi risk, business risk is a more prevailing channel hrough which compeiion affecs audi fees. Fourh, we find differen resuls due o he differences discussed above. Levenis e al. (2011) show a negaive relaion beween produc marke compeiion and audi fees, while my sudy finds ha audiors end o charge firms higher fees in compeiive indusries. The remainder of his essay is organized as follows. Nex secion reviews prior research concerning produc marke compeiion and audi fees and develops he hypohesis. Then I describe he sample and research design. I presen empirical resuls afer he research design. The las secion draws a conclusion. 4.2 Lieraure review and hypohesis developmen Audiors are paid for heir asserions in a clien s financial saemen (Bell e al. 2001). All else being equal, audi fees increase wih liigaion risk. Audiors are sued eiher because here are maerial misakes in heir audied financial repors or simply because heir clien is undergoing financial disress or bankrupcy. The likelihood ha an audior would face he former siuaion is called audi risk, while 98

he likelihood ha an audior would face he laer siuaion is called business risk. Thus, he amoun of audior fees is posiively relaed wih a clien s audi risk and business risk. A basic characerisic shared by firms in a compeiive indusry is ha hey have many peers. This fac induces such firms o behave differenly from hose in a less compeiive indusry. Their behavior paerns are likely o change he audi risk and business risk faced by heir audiors. Audiors are hus likely o charge hese firms differenly as a consequence. I will discuss how produc marke compeiion affecs audi fees hrough audi risk and business risk channels in he following subsecions. 4.2.1 Compeiion vs. Audi Risk Audi risk describes he likelihood of maerial errors in he clien s financial saemen (Gul and Tsui 1997). Audior fees reflec he effors ha audiors devoe o reduce he audi risk o an accepable level in order o assure ha he clien s financial repors are free of maerial missaemen (Lemon e al. 1993; O Keefe e al. 1994). As saed previously, a basic characerisic shared by firms in a compeiive indusry is ha hey have many peers. This characerisic has some implicaions. Firs, his allows for a more comparable performance comparison among firms in a compeiive indusry and also allows for more relaive performance evaluaion (RPE) (Har 1983). For insance, DeFond and Park (1999) find ha RPE-based 99

(firm-specific) accouning measures are more closely associaed wih CEO urnover in a compeiive indusry. Therefore, an increase in compeiion generaes addiional informaion o miigae moral hazard problems. Second, more peers due o he increased compeiion reduce a firm s profiabiliy. If managers do no keep coss low in a compeiive indusry, hen a reducion in profis may pu he firm in an unprofiable posiion, so he probabiliy ha he firm would have o be liquidaed would go up (Schmid 1997). Therefore, in order for managers o keep heir jobs and avoid a firms liquidaion, managers in a compeiive indusry have o work harder, avoid wasing company resources on waseful projecs, and engage less in non-value-maximizing aciviies. Chrisie and Zimmerman (1994) indicae ha non-value-maximizing managers end o manipulae accouning informaion o hide heir non-opimal aciviies more han value-maximizing managers. Thus, compeiion helps o align he ineress of managers and shareholders and decreases he likelihood of earnings managemen. I plays a corporae governance role and works as an effecive mechanism o miigae agency problems beween shareholders and managers. Empirical sudies confirm he above argumen regarding produc marke compeiion, agency cos and, furhermore, earnings managemen. Marciukaiye and Park (2009) find ha firms in more compeiive indusries are less likely o engage in opporunisic earnings managemen as measured by he magniude of discreionary accruals. Using forced resaemen daa from he Financial Saemen Resaemen Daabase and he federal class acion securiies fraud lawsuis idenified from he Sanford Securiies Class Acion Clearing house, Marciukaiye 100

and Park also find ha forced earnings resaemens and securiy fraud lawsuis are less common in compeiive markes. Their resuls sugges ha produc marke compeiion decreases he informaion asymmery beween managers and he marke and improves he accuracy of financial reporing effecively. Firms in compeiive indusries are less likely o repor misleading earnings and are more likely o provide informaive financial repors. I go one-sep furher o he audi fees area. Chow (1982) provides evidence ha firms wih higher agency coss have more incenive o hire more exernal audiors. Was and Zimmerman (1986) claim ha demand for high-qualiy audis increases wih agency coss, wheher hey are volunarily underaken by managers as a bonding mechanism or are exernally imposed by sakeholders as a monioring mechanism. Empirical sudies provide suppor for his argumen on agency cos and audi fees. For example, firms wih severe agency problems due o free cash flow are shown o incur more audi services (Gul and Tsui 1997; Griffin e al. 2010). Following he same reasoning, audi fees should be lower if produc marke compeiion miigaes agency problems by discouraging managers from acing unwisely and no masking heir behaviors by manipulaing financial saemens. In summary, prior sudies sugges ha agency heory is one channel hrough which produc marke compeiion can impac audi fees. Compeiion plays a governance role and miigaes agency problems. Srenghened governance and miigaed agency problems hrough compeiion also improve he accuracy of financial reporing. This reduces invesors demand on audi services. Based on hese heoreical argumens and he empirical evidence in prior lieraure, I predic 101

ha compeiion will decrease audiors effors o reduce audi risk o an accepable level. Audiors are likely o assess firms in more compeiive indusries as being hose wih low levels of audi risk. Therefore, ceeris paribus, audi fees decrease wih he level of produc marke compeiion. 4.2.2 Compeiion vs. Business Risk Business risk describes circumsances ha are ou of he audior s conrol and canno be eliminaed. According o American Insiue of Cerified Public Accounans (AICPA 1992), business risk includes wo componens: clien s business risk and audior s business risk. Clien s business risk is associaed wih he clien s coninued survival and well-being, while audior s business risk is defined by SAS No. 47 as he risk of poenial legal coss and oher expendiures from a business associaion wih a clien, wheher or no an audior failure exiss (Bell e al. 2001; Morgan and Socken 1998). The clien s business risk and he audior s business risk are, herefore, closely correlaed. O Malley (1993) furher claims ha anyone who suffers a financial loss may sue audiors and demand compensaion from hem even if here are no missaemens in he audied repors. Thus, business risk is usually regarded as he risk faced by audiors even when an audi repor is flawless under regulaions or accouning rules. Because audiors canno eliminae business risk, hey end o charge cliens higher fees due o higher business risk. Prior lieraure suppors his argumen. For example, Palmrose (1987) find a relaion beween bankrupcies and lawsuis agains audiors. Bell e al. (2001) 102

empirically find ha audi fees are higher for high risk cliens. Morgan and Socken (1998) also show ha audi fees increase wih business risk. Firms in compeiive indusries run more business risks han hose in less compeiive indusries. In he aspec of firm operaion, firms in compeiive indusries end o engage in innovaive aciviies more han hose in less compeiive indusries (Schumpeer 1912; Hou and Robinson 2006), hereby incurring greaer innovaion risk. Also, wihou barriers o enry, firms in compeiive indusries face more hreas from new enries and from exising rivals, hus incurring liquidiy risk. Having a liquidiy problem is also predicaive of a firm s financial failure (Seeharaman e al. 2002). Schmid (1997) argues ha produc marke compeiion increases he probabiliy of liquidaion. Hou and Robinson (2006) empirically suppor heir predicion ha innovaion and disress risk represen wo ways for firms in compeiive indusries o require higher expeced reurns. A clien s business risk, in urn, affecs he audior s business risk (O Keefe e al. 1994). For example, Palmrose (1997) shows ha financially disressed firms are ofen involved in audior liigaion. Therefore, audiors end o charge higher fees on firms in a more compeiive indusry due o he associaed business risks. 4.2.3 Compeiion vs. Audi Fees To summarize, produc marke compeiion may affec audi fees in wo direcions. On he one hand, firms in a more compeiive indusry are less likely o 103

manipulae financial saemens due o he corporae governance role played by compeiion Thus, audi risk decreases wih produc marke compeiion. Audi fees hen decrease wih produc marke compeiion hrough he channel of audi risk. On he oher hand, firms in a more compeiive indusry face more disress risk and liquidaion risk. Audior liigaion risk increases wih produc marke compeiion due o disress risk and liquidaion risk. Thus, business risk increases wih produc marke compeiion. Audi fees hen increase wih produc marke compeiion hrough he channel of business risk. Since wo channels have conradicory predicions on he relaion beween produc marke compeiion and audi fees, he ne effec of compeiion on audi fees is ambiguous. I leave i as an empirical issue and explore which channel s effec prevails. Hypohesis: There is no significan relaion beween produc marke compeiion and audi fees. 4.3 Empirical ess 4.3.1 Dependen Variable: Audi Fees (LAF) The dependen variable is he naural log of oal audi fees (LAF). I obain he dependen variable from he Audi Analyics daabase. The Audi Analyics daabase sars from 2000. 104

4.3.2 Independen Variable: Herfindahl-Hirschman Index (HIndex) The independen variable is Herfindahl-Hirschman index (HIndex), which is defined as he sum of he square of percenage marke share (Equaion 1). In Equaion 1, here are N firms in he ineresed indusry. Firm i s revenue is s i. S represens he oal revenues in he indusry and is calculaed as N S i i 1. N si 2 Herfindahl _ Index ( ) (1) S i 1 The smaller he Herfindahl-Hirschman index, he more compeiive he indusry will be. A larger Herfindahl-Hirschman index means ha he indusry is concenraed in he hands of a few large firms (Hou and Robinson 2006). As poined ou in Ali e al. (2009), a Herfindahl-Hirschman index based on Compusa daa only considers public firms in an indusry and, herefore, is a biased measure of marke compeiion level. Ali e al. find he correlaion beween Compusabased concenraion measures and U.S. Census-based concenraion measures are only 13%, and he resuls in some imporan prior sudies do no hold when he U.S. Census measure is employed. So, I use he Herfindahl-Hirschman index from he U.S. Census of Manufacurers, which covers boh public and privae firms in an indusry, as he measure of compeiiveness. The U.S. Census Bureau repors concenraion raios for hundreds of indusries in he manufacuring secor in heir Census of Manufacurers Publicaions. A U.S. Census akes place every five years. The wo mos recen 105

were in 1997 and 2002. I only downloaded he concenraion raio daa for year 2002 because audi fee daa is available only afer year 2000, and year 2002 is he only inersecion year provided by he Audi Analyics daabase and he U.S. Census. Following prior lieraure (Aggarwal and Samwick 1999; MacKay and Phillips 2005; Campello 2006; Haushaler e al. 2007; Ali e al. 2009), I assume ha he indusry concenraion level does no change rapidly. I, herefore, use he concenraion raio in he 2002 Census for concenraion raios from 2000 o 2004, which is wo years before 2002 o wo years afer 2002. 4.3.3 Conrol Variables Prior lieraure shows ha he audi complexiy, clien size and clien financial condiion affec audi fees. Conrol variables I considered are hose usually used in audi fee models (Simunic 1980; Francis 1984; Craswell e al. 1995; Gul and Tsui 1997). Below are he descripions of he conrol variables and he prediced sign of heir coefficiens (Seeharaman e al. 2002): SIZE (+): clien s firm size, measured as he naural log of oal asses. The coefficien for he clien size is expeced o be posiive because large firms end o be more complex and also imply a larger poenial damage awards (Kellogg 1984). CURRENT (+): curren raio, measured as he raio of curren asses o oal asses. The coefficien is expeced o be posiive because he curren raio is a proxy for audi complexiy. 106

QUICK (-): quick raio, measured as he raio of curren asses, less invenory o curren liabiliies. The coefficien is expeced o be negaive because a greaer quick raio implies more liquidiy and less likelihood of financial disress. ROA (-): he reurn on asses, measured as he earnings before ineres and ax divided by oal asses. The coefficien is expeced o be negaive because a greaer ROA means more profiabiliy and less likelihood of financial disress. DE (+): leverage, measured as he long-erm deb o oal asses raio. The coefficien is expeced o be posiive because greaer leverage implies more risk of financial disress. LNAF (+): naural log of non-audi fees. The coefficien is expeced o be posiive for he repored posiive associaion beween audi and non-audi fees (Simunic 1984). NO_BUS_SEGMENTS (+): number of business segmens. The coefficien for he number of business segmens is expeced o be posiive because he more indusry diversificaion, he more audi complexiy. NO_GEO_SEGMENTS (+): number of geographic segmens. Geographic dispersion of operaions is one aspec of he clien s srucure (O Keefe e al. 1994). The coefficien for he number of geographic segmens is expeced o be posiive because he more geographic dispersion, he more audi complexiy here is. 107

LOSS (+): LOSS=1 if income before exraordinary iems in he audied year is negaive; oherwise i equals 0. The coefficien is expeced o be posiive because liigaion risk increases in loss years and audiors would hen require more effors. FISCAL (+): FISCAL=1 if he clien s fiscal year end is December 31; oherwise i equals 0. The coefficien for FISCAL is expeced o be posiive because audiors end o charge higher fees in a busy season. BIG_FIVE (+): BIG_FIVE=1 if he audior is Arhur Andersen LLP, Erns & Young LLP, Deloie & Touche LLP, KPMG LLP or PricewaerhouseCoopers LLP. The coefficien is expeced o be posiive because he Big Five end o represen greaer experise and have a greaer repuaion and, herefore, earn sysemaically higher audi fees. 4.3.4 Regression The regression model is shown in he equaion below: LAF j, LOSS HIndex j, j, NO _ BUS _ SEGMENTS 6 9 0 1 FISCAL 10 SIZE j, 2 j, j, LNAF DE j, j, NO _ GEO _ SEGMENTS 11 7 3 ROA 12 QUICK 4 j, j, j, CURRENT BIG _ FIVE Year _ Effec 5 8 j, j, j, (2) Where he subscrip i refers o firm he subscrip j refers o indusry j, and he subscrip refers o year. Year ranges from 2000 o 2004. 108

4.4 Empirical resuls 4.4.1 Descripive Saisics Our sample includes all acive firms ha are conained in he inersecion of he Audi Analyics daabase, he U.S. 2002 Census of Manufacurers, Compusa annual files and he Compusa segmen daabase beween 2000 and 2004. I obain audi fees and non-audi fees daa from he Audi Analyics daabase, Herfindahl- Hirschman index from he Census daabase; he number of segmens from he Compusa Segmen daabase and oher conrol variables from he Compusa annual files. Herfindahl-Hirschman index, represened as HIndex, is calculaed using he download value from he Census daabase divided by 10,000. I include all acive firms, excep hose wih 1) resaed financial repors; 2) audi fees in non- US dollars; 3) zero audi fees; 4) more han one repor on audi fees; and 5) zero repored oal asses. The final sample includes 2201 firm-year observaions ha represen 760 unique firms from 2000 o 2004. Table 4.1 shows he numbers of observaions and he number of loss firms in each year from 2000 o 2004. From 2000 o 2004, 35.56%, 47.50%, 42.57%, 38.74% and 31.51% of he sample firms have negaive incomes before exraordinary iems. NAICS ranges from 311230 o 339999, bu no coninuously. To provide some sense abou NAICS, 311230 represens he Breakfas Cereal Manufacuring indusry, 339995 represens he Burial Caske Manufacuring indusry and 339999 represens all oher miscellaneous manufacuring indusries no included in 311111 o 339998. 109

Table 4.2 shows he descripive saisics of he dependen variable, independen variable and conrol variables. The naural log of he audi fees ranges from 8.2965 o 17.8228, wih he mean 12.8556 and median 12.7156. Herfindahl- Hirschman index numbers range from 0.000065 o 0.2707, wih he mean 0.0791 and median 0.0635. Abou 58 percen of he sample firms have heir fiscal year ending on December 31. Abou 87 percen of he sample firms hire Big Five as heir audi firms. 4.4.2 Main Resuls Table 4.3 shows he associaion beween produc marke compeiion and audi fees. The regression formula is shown in Equaion 1. I consider he year fixed effecs in he regression. All conrol variables excep he curren raio have he coefficiens wih he expeced signs a he significance level of 1%. No significan relaion is found beween he curren raio and audi fees, bu he coefficien for CURRENT has he expeced posiive sign. The esimaed coefficien for HIndex is -0.5745 ( saisics: -2.59). This means ha here is a negaive relaion beween he Herfindahl-Hirschman index and audi fees, boh demonsraing saisical significance and economic significance. In oher words, audiors charge higher fees on firms in a more compeiive indusry. I recognize he possible exisence of he heeroskedasiciy and ime series dependence in he regression. So I use he Whie correcion and Fama-MacBeh mehods o conrol for hese wo concerns, respecively. Panel A in Table 4.4 shows he resuls wih he Whie correcion. All conrol variables have he 110

coefficiens wih he expeced signs and all of hem, excep he curren raio, are significan a 1%. In Panel A of Table 4.4, he coefficien for HIndex is -0.5565 ( saisics: -2.54). This means ha here is a negaive relaion beween Herfindahl- Hirschman index and audi fees, boh in saisical significance and economic significance. In oher words, audiors charge firms in more compeiive indusries higher fees. Panel B in Table 4.4 shows he regression resuls when a Fama- MacBeh regression is employed. The esimaed coefficien for HIndex is -0.5733 wih 1% significance. The resuls on he conrol variables remain qualiaively same, bu he significance levels on DE, FISCAL, and BIG_FIVE are weakened. To conrol for he skewed disribuion of a Herfindahl-Hirschman index, I also replace he original HIndex value wih he log of he value downloaded from he Census daabase. The unabulaed resuls sugges ha he relaion beween compeiion and audi fees are qualiaively idenical. To summarize, he evidence provides suppor for he argumen ha firms in a more compeiive indusry face more liquidiy risk, disress risk and liquidaion risk. Audior liigaion risk increases wih produc marke compeiion due o increased business risk, so audiors charge firms in more compeiive indusries higher fees. 4.4.3 Addiional ess 4.4.3.1 Audior Indusry Specializaion 111

I consider he audi marke compeiion in my regression. I examine wheher naional indusry leadership and specializaion in he audi marke affecs he relaion beween audi fees and clien compeiion. Following Francis e al. (2005), I inroduce wo variables o calculae audior indusry specializaion a he naional level: IND_SPE1 and IND_SPE2. These wo measures are based on he assumpion ha indusry experise increases wih audi marke share. IND_SPE1 is a dummy variable. I equals o 1 if he audior is a naional indusry specialis and 0, oherwise. IND_SPE2 is he audi marke share wihin a wo-digi SIC indusry. Panel A in Table 4.5 shows he resuls wih IND_SPE1. We can see ha he relaion beween produc marke compeiion and audi fees sill holds wih he inclusion of a dummy-version audior specializaion. Audior specializaion has an insignificanly posiive effec on audi fees. Panel B in Table 4.5 repors he resuls wih IND_SPE2. Sill, he relaion beween produc marke compeiion and audi fees holds wih he inclusion of a coninuous-version audior specializaion. IND_SPE2 has a significanly posiive effec on audi fees, bu his effec is subsumed by BIG_FIVE. This signifies ha a Big Five audior firm ends o charge clien s higher fees, regardless of indusry specializaion. 4.4.3.2 Firm-level Compeiion Measure For compleeness, I also examine wheher audiors charge firms wih differen compeiion saus in he same indusry heerogeneously. Peress (2010) poins ou ha imperfec compeiion in a produc marke affecs firms behaviors in equiy marke. Firms wih differen produc marke power end o behave differenly. Wih pricing power due o he marke power, firms wih good 112

compeiion saus can easily ransfer demand or supply shock o he cusomers, so hey end o have cash flows or earnings wih less volailiy. I invesigae wheher his difference changes he audiors risk assessmens by inroducing produc marke power ino he regression. I argue ha firms wih greaer produc marke power face less business risk due o heir superior compeiive saus. Due o sronger price-seing abiliies (Wang 2011), hey also have less incenive o manipulae heir earnings hrough accrual managemen, so audiors will view hem hese cliens as having less audi risk. As such, I hypohesize ha firms wih a high compeiion saus end o be charged less by audiors. Equaion 2 is he regression formula. LAF j, CURRENT 6 ROA HIndex j, j, BIG _ FIVE (2) 9 13 0 1 j, j, LOSS Year _ Effec PCM j, j, j, SIZE NO _ BUS _ SEGMENTS 7 10 2 FISCAL 11 3 j, j, j, DE LNAF j, j, QUICK j, NO _ GEO _ SEGMENTS 12 8 4 5 j, As suggesed in previous lieraure (Lerner 1934; Carlon and Perloff 2000; Kale and Loon 2010), I use he Lerner index or price-cos margin (PCM) as a measure of produc marke compeiion saus. Following Peress (2010), I measure PCM as he raio of operaing profi o sales. Operaing profi is sales less he cos of goods sold, as well as selling, general and adminisraive expenses. I discard he PCMs wih exreme absolue values of several housand and only keep hose wih absolue value no greaer han 1. Afer merging hese wih he oher daases, I have 1970 firm-year observaions. Panel A in Table 4.6 shows he descripive saisics of PCM. Panel B in Table 4.6 includes he regression resuls. 113

As hypohesized, I find a negaive significan relaion beween produc marke power and audi fees (esimaed coefficien =-0.3534; saisics=-3.11). This means ha wihin he same indusry, audiors charge firms wih beer compeiion saus less. The relaion beween audi fees and Herfindahl-Hirschman index and oher conrol variables sill holds. The sign of he relaion beween LOSS and audi fees is sill as expeced, bu no significan. On average, audiors charge firms higher fees in more compeiive indusries. However, audiors end o charge less on firms wih high produc marke power wihin an indusry. 4.4.3.3 Newey-Wes Tes I also consider he poenial bias problem in he empirical resuls resuling from an unobserved firm effec. The residuals of a given firm may be correlaed across years for a firm, so I apply he Newey-Wes mehod in he analysis. Table 4.7 shows he resuls of Newey-Wes when he parameer lag is se as 4 (oher lag values do no change he resul qualiaively). Table 4.7 shows ha he coefficiens on HIndex and PCM are sill negaive. The p-value of he esimaed coefficien on HIndex is 0.0443 and he p-value of he esimaed coefficien on PCM is 0.0162. Mos conrol variables excep LOSS have he prediced signs. Wih an unprediced sign, LOSS has he saisics of -0.60. Hence, Newey-Wes resuls sill suppor he argumen ha audiors end o charge higher fees on firms in more compeiive indusries, while hey end o charge less on hose wih high produc marke power wihin an indusry. 114

4.5 Conclusions This essay sudies he impac of produc marke on he audiing indusry. I examine wheher audiors charge firms in indusries wih differen compeiion levels heerogeneously. I also invesigae wheher audiors reques more fees from firms wih grea produc marke power wihin an indusry. Exising heories posi wo conradicory predicions on he associaion beween indusry concenraion and audi fees. On he one hand, produc marke compeiion miigaes agency problems beween shareholders and managers and increases he accuracy of financial reporing, hus, decreasing required audi effor and audi fees. On he oher hand, firms in compeiive markes are expeced o face higher liquidiy risk, disress risk and liquidaion risk, hus increasing audiors assessmens of business risk and audi fees. I empirically es he relaion beween indusry concenraion and audi fees and find ha he second heory prevails. In addiion o he indusry-level analysis of produc marke compeiion on audi fees, I also invesigae firm-level compeiion saus effecs on audi fees. I explore wheher audiors rea firms wih beer compeiion saus differenly. Firms wih grea marke power can easily ransfer demand or supply shock o cusomers, so hey end o have less volaile cash flows or earnings. They hen face less disress risk and liquidaion risk, and, consequenly, less business risk. Due o a small magniude of volailiy peraining o heir earnings and cash flows, hey also have less incenive o manage heir earnings. So, audiors will view hese 115

cliens as having less audi risk. Our finding suppors he reasoning ha firms possessing advanageous marke saus pay less in audi fees. This sudy no only complemens he exising lieraure regarding he impacs of compeiion on he capial marke, bu also confirms and explains he exisence of he economic indusry effec on audi fees. Finally, I need o poin ou ha, due o he daa availabiliy of indusry-level compeiion measures, his paper is based on daa in he U.S. manufacuring indusry only. I is necessary, hen, o examine wheher he conclusions sill hold for all indusries. 116

References Al A., S. Klasa, and E. Yeung. 2009. The limiaions of indusry concenraion measures consruced wih Compusa daa: implicaions for finance research. Review of Financial Sudies 22: 3439-3871. Al A., S. Klasa, and E. Yeung. 2010. Indusry concenraion and corporae disclosure policy. Working paper. Universiy of Texas a Dallas. Bell, T., W. Landsman, and D. Shackelford. 2001. Audiors perceived business risk and audi fees: analysis and evidence. Journal of Accouning Research 39: 35-43. Boonsan, C., and M. Sanford. 2005. Managers moives o wihhold segmen disclosures and he effec of SFAS no. 131 on analyss informaion environmen. The Accouning Review 80: 751-771. Campello, M. 2006. Deb financing: does i hur or boos firm performance in produc markes? Journal of Financial Economics 82: 135-172. Carlon, D., and J. Perloff. 2000. Modern Indusrial Organizaion. Addison-Wesley. Masschuses. Chow, C.W. 1982. The demand for exernal audiing: Size, deb and ownership influences. The Accouning Review 57: 272-291. Chrisie, A.A., and J. Zimmerman. 1994. Efficien and opporunisic choices of accouning procedures: corporae conrol coness. The Accouning Review 69: 539-566. Craswell, A., J.R. Francis, and S.L. Taylor. 1995. Audior brand name repuaions and indusry specializaions. Journal of Accouning and Economics 20: 297-322. DeFond M., and C. Park. 1999. The effec of compeiion on CEO urnover. Journal of Accouning and Economics 27: 35-56. Dhaliwal, D., S. Huang, I. Khurana, and R. Pereira. 2008. Produc marke compeiion and accouning conservaism. hp://ssrn.com/absrac=1266754. Eichenseher, M., and D. Shields. 1983. The correlaes of CPA-firm change for publicly held corporaion. Audiing: A Journal of Pracice & Theory 2: 23-37. 117

Ellis, J., E. Fee, and S. Thomas. 2009. Produc marke compeiion and he disclosure of informaion abou cusomers. Working Paper. Universiy of Pisburgh. Francis, J. 1984. The effec of audi firm size on audi prices: a sudy of he Ausralian marke. Journal of Accouning and Economics 6: 133-151. Francis, J., K. Reichel, and D. Wang. 2005. The pricing of naional and ciyspecific repuaions for indusry experise in he U.S. audi marke. The Accouning Review 80: 113-136. Fee, C., and S. Thomas. 2004. Sources of gains in horizonal akeovers: evidence from cusomer, supplier, and rival firms. Journal of Financial Economics 74: 423-460. Gaspar, J., and M. Massa. 2005. Idiosyncraic volailiy and produc marke compeiion. Journal of Business 79: 3125-3152. Giroud X., and H. Mueller. 2007. Does corporae governance maer in compeiive indusries? Working paper. New York Universiy. Griffin, P., D. Lon, and Y. Sun. 2010. Agency problems and audi fees: furher ess of he free cash flow hypohesis. Accouning and Finance 50: 321-350. Grullon, G., and R. Michaely. 2007. Corporae payou policy and produc marke compeiion. Working paper. Rice Universiy. Gul, F., and J. Tsui. 1997. A es of he free cash flow and deb monioring hypoheses: Evidence from audi pricing. Journal of Accouning and Economics 24: 219-237. Har, O. 1983. The marke mechanism as an incenive scheme. Bell Journal of Economics 14: 366-382. Haushaler, D., S. Klasa, and W. Maxwell. 2007. The influence of produc marke dynamics on a firm s cash holdings and hedging behavior. Journal of Financial Economics 84: 797-825. Hou, K., and D. Robinson. 2006. Indusry concenraion and average sock reurns. Journal of Finance 4: 1927-1956. Kale, J.R., and Y.C. Loon. 2011. Produc marke power and sock marke liquidiy. Journal of Financial Markes 14: 376-410. 118

Kellogg, R. 1984. Accouning aciviies, securiies, securiy prices, and class acion lawsuis. Journal of Accouning and Economics 6: 185-204. Lerner, A.P. 1934. The concep of monopoly and he measuremen of monopoly power. The Review of Economic Sudies 1: 157-175. Levenis, S., P. Weeman, and C. Caramanis. 2011. Agency coss and produc marke compeiion: he case of audi pricing in Greece. The Briish Accouning Review 43: 112-119. Lemon, W., A. Arens, and J. Loebbecke. 1993. Audiing: an inegraed approach. Prenice-Hall, Englewood Cliffs, NJ. Mackay, P., and G. Phillips. 2005. How does indusry affec firm financial srucures? Review of Financial Sudies: 1433-1466. Marciukaiye, D., and J. Park. 2009. Marke compeiion and earnings managemen. Working paper. Drexel Universiy. Morgan, J., and P. Socken. 1998. The Effecs of Business Risk on Audi Pricing. Review of Accouning Sudies 3: 365-385. O Keefe T., D. Simunic, and M. Sein. 1994. The producion of audi services: evidence from a major public accouning firm. Journal of Accouning Research 32: 241-261. O Malley, S. 1993. Legal liabiliy is having a chilling effec on he audior s role. Accouning Horizons 7: 82-87. Palmrose, Z. 1987. Liigaion and independen audiors: he role of business failures and managemen fraud. Audiing: a Journal of Pracice & Theory (Spring): 90-102. Palmrose, Z. 1997. Who go sued? Journal of Accounancy 183: 67-69. Peress, J. 2010. Produc marke compeiion, insider rading, and sock marke efficiency. Journal of Finance 1: 1-43. Schmid, K. 1997 Managerial incenives and produc marke compeiion. Review of Economic Sudies 64: 191-213. Seeharaman, A., F. Gul, and S. Lynn. 2002. Liigaion risk and audi fees: evidence from UK firms cross-lised on US markes 33. Journal of Accouning and Economics: 91-115. 119

Simunic, D. 1980. The pricing of audi services: heory and evidence. Journal of Accouning Research 18:161-190. Simunic, D. 1984. Audiing, consuling, and audior independence. Journal of Accouning Research 22: 679-702. Vala, P. 2010. Compeiion and he cos of deb. Job marke paper. Swiss Finance Insiue. Wang, Y. 2011. Marke power and accrual managemen. Working paper. The Hong Kong Polyechnic Universiy. Was, R., and J. Zimmerman. 1986. Posiive accouning heory. Prenice-Hall. Englewood Cliffs. NJ. 120

Table 4.1 Sample firms frequency disribuion from 2000 o 2004 The sample includes all acive firms ha are conained in he inersecion of he Audi Analyics daabase, he U.S. 2002 Census of Manufacurers, Compusa annual files and Compusa segmen daabase beween 2000 and 2004. The sample includes all acive firms excep hose wih 1) resaed financial repors 2) audi fees in non-us dollars 3) zero audi fees 4) more han one repors on audi fees 5) zero repored oal asses. There are 2201 firm-years ha represen 760 unique firms wih six-digi Norh American Indusry Classificaion Sysems (NAICS) from 2000 o 2004. Loss firms represen hose wih negaive income before exraordinary iems in he audied year. 2000 2001 2002 2003 2004 No. of Firms 284 440 404 524 549 No. of Loss Firms 101 209 172 203 173 Loss Firms (%) 35.56% 47.50% 42.57% 38.74% 31.51% 121

Table 4.2 Descripive saisics The sample consiss of 2201 firm-years over he period from 2000 o 2004. This able presens he descripive saisics of he dependen variable LAF (he naural log of audi fees), he independen variable HIndex (Herfindahl-Hirschman index), and conrol variables. Conrol variables include audiee size (SIZE naural log of oal asses), curren raio (CURRENT raio of curren asses o oal asses), quick raio (QUICK raio of curren asses less invenory o curren liabiliies), reurn on asses (ROA- he earnings before ineres and ax divided by oal asses), leverage (DE deb o oal asses raio), non-audi fees (LNAF-log of non-audi fees), number of business segmens (NO_BUS_SEGMENTS), number of geographic segmens (NO_GEO_SEGMENTS), loss year (LOSS=1 if loss year, oherwise 0), audior s repuaion (BIG_FIVE=1 if he auhor belongs o Big Five, oherwise 0), and fiscal year end (FISCAL=1 if fiscal year end is Dec. 31, oherwise 0). Mean Sd Dev 10% 25% Median 75% 90% LAF 12.8556 1.3548 11.2226 11.8845 12.7156 13.7157 14.7480 HIndex 0.0791 0.0573 0.0189 0.0337 0.0635 0.1307 0.1453 SIZE 5.6523 2.2035 2.8489 4.1405 5.6411 7.1512 8.4081 CURRENT 0.5698 0.2121 0.3002 0.4015 0.5579 0.7305 0.8816 QUICK 2.3526 5.8645 0.2595 0.4429 0.8767 2.1545 4.9309 ROA -0.0233 0.4537-0.2550-0.0379 0.0583 0.1178 0.1823 DE 0.1753 0.2658 0 0.0004 0.0957 0.2609 0.4197 LNAF 12.1664 1.8958 9.7410 10.8435 12.2144 13.4225 14.5186 NO_BUS_SEGMENTS 2.4284 1.9722 1 1 1 4 5 NO_GEO_SEGMENTS 3.5488 2.4349 1 2 3 5 7

Table 4.3 Relaion beween produc marke compeiion and audi fees (OLS regression wih year fixed effecs) This able presens he regression resuls of audi fees on produc marke compeiion. The sample consiss of 2201 firm-years over he period from 2000 o 2004. Dependen variable is he naural log of audi fees. The independen variable, HIndex, is Herfindahl- Hirschman index. Conrol variables include audiee size (SIZE naural log of oal asses), curren raio (CURRENT raio of curren asses o oal asses), quick raio (QUICK raio of curren asses less invenory o curren liabiliies), reurn on asses (ROA- he earnings before ineres and ax divided by oal asses), leverage (DE deb o oal asses raio), non-audi fees (LNAF-log of non-audi fees), number of business segmens (NO_BUS_SEGMENTS), number of geographic segmens (NO_GEO_SEGMENTS), loss year (LOSS=1 if loss year, oherwise 0), audior s repuaion (BIG_FIVE=1 if he auhor belongs o Big Five, oherwise 0), and fiscal year end (FISCAL=1 if fiscal year end is Dec. 31, oherwise 0). ***, **, and * represen 1%, 5%, and 10% significance respecively. Prediced Sign Esimae Saisics HIndex (?) 0.5745*** 2.59 SIZE (+) 0.3504*** 30.79 CURRENT (+) 0.0287 0.40 QUICK ( ) 0.0120*** 5.36 ROA ( ) 0.1321*** 4.36 DE (+) 0.1364*** 2.81 LNAF (+) 0.1913*** 17.07 NO_BUS_SEGMENTS (+) 0.0681*** 9.77 NO_GEO_SEGMENTS (+) 0.0280*** 5.25 LOSS (+) 0.0828*** 2.95 FISCAL (+) 0.2410*** 9.48 BIG_FIVE (+) 0.1310*** 3.16 YEAR EFFECTS Included Included Adj. R^2 82.81%

Table 4.4 Panel A: Relaion beween produc marke compeiion and audi fees (OLS Regression wih year fixed effecs; Whie correced) This able presens he regression resuls of produc marke compeiion on audi fees. The sample consiss of 2201 firm-years over he period from 2000 o 2004. Dependen variable is he naural log of audi fees. The independen variable HIndex, is Herfindahl-Hirschman index downloaded from he U.S. 2002 Census of Manufacurers. Conrol variables include audiee size (SIZE naural log of oal asses), curren raio (CURRENT raio of curren asses o oal asses), quick raio (QUICK raio of curren asses less invenory o curren liabiliies), reurn on asses (ROA- he earnings before ineres and ax divided by oal asses), leverage (DE deb o oal asses raio), non-audi fees (LNAF-log of nonaudi fees), number of business segmens (NO_BUS_SEGMENTS), number of geographic segmens (NO_GEO_SEGMENTS), loss year (LOSS=1 if loss year, oherwise 0), audior s repuaion (BIG_FIVE=1 if he auhor belongs o Big Five, oherwise 0), and fiscal year end (FISCAL=1 if fiscal year end is Dec. 31, oherwise 0). ***, **, and * represen 1%, 5%, and 10% significance respecively. Prediced Sign Esimae Saisics HIndex (?) -0.5565** -2.54 SIZE (+) 0.3508*** 27.96 CURRENT (+) 0.0507 0.69 QUICK (-) -0.0120*** -3.64 ROA (-) -0.1397*** -3.33 DE (+) 0.1357*** 2.73 LNAF (+) 0.193*** 14.14 NO_BUS_SEGMENTS (+) 0.0678*** 10.22 NO_GEO_SEGMENTS (+) 0.0278*** 5.06 LOSS (+) 0.0692** 2.35 FISCAL (+) 0.2532*** 9.97 BIG_FIVE (+) 0.1236*** 3.09 YEAR EFFECTS Included Adj R^2 82.45% 124

Table 4.4 Panel B: Relaion beween produc marke compeiion and audi fees (Fama-Macbeh) This able presens he regression resuls of produc marke compeiion on audi fees. The sample consiss of 2201 firm-years over he period from 2000 o 2004. Dependen variable is he naural log of audi fees. The independen variable, HIndex, is Herfindahl- Hirschman index downloaded from he U.S. 2002 Census of Manufacurers. Conrol variables include audiee size (SIZE naural log of oal asses), curren raio (CURRENT raio of curren asses o oal asses), quick raio (QUICK raio of curren asses less invenory o curren liabiliies), reurn on asses (ROA- he earnings before ineres and ax divided by oal asses), leverage (DE deb o oal asses raio), non-audi fees (LNAFlog of non-audi fees), number of business segmens (NO_BUS_SEGMENTS), number of geographic segmens (NO_GEO_SEGMENTS), loss year (LOSS=1 if loss year, oherwise 0), audior s repuaion (BIG_FIVE=1 if he auhor belongs o Big Five, oherwise 0), and fiscal year end (FISCAL=1 if fiscal year end is Dec. 31, oherwise 0). ***, **, and * represen 1%, 5%, and 10% significance respecively. Prediced Sign Esimae Saisics HIndex (?) -0.5733*** -6.10 SIZE (+) 0.3401*** 12.94 CURRENT (+) -0.0210-0.17 QUICK (-) -0.0133*** -4.12 ROA (-) -0.2585*** -4.21 DE (+) 0.0840** 2.26 LNAF (+) 0.2047*** 12.7 NO_BUS_SEGMENTS (+) 0.06618*** 17.92 NO_GEO_SEGMENTS (+) 0.0275*** 6.19 LOSS (+) 0.0654*** 2.88 FISCAL (+) 0.1867* 1.95 BIG_FIVE (+) 0.0599 0.89 YEAR EFFECTS Included Adj. R^2 82.45% 125

Table 4.5 Panel A: Regression of audi fees on produc marke compeiion wih audior specializaion Table 4.5 includes he regression resuls of audi fees on indusry-level compeiion wih audior specializaion. The sample consiss of 2201 firm-years over he period from 2000 o 2004. Dependen variable is he naural log of audi fees. HIndex is Herfindahl- Hirschman index downloaded from he U.S. 2002 Census of Manufacurers. IND_SPE1 is he dummy variable for audior specializaion. Panel A shows he resuls wih IND_SPE1 in regression. Conrol variables are same as described in previous ables. ***, **, and * represen 1%, 5%, and 10% significance respecively. Prediced Esimaed HIndex (?) 0.5643** 0.5592** [ 2.53] [ 2.51] SIZE (+) 0.3579*** 0.3499 [32.23] [30.69]*** CURRENT (+) 0.0503 0.0295 [0.71] [0.41] QUICK ( ) 0.0116*** 0.0121*** [ 5.19] [ 5.38] ROA ( ) 0.1294*** 0.1314 [ 4.26] [ 4.33]*** DE (+) 0.1292*** 0.1375*** [2.66] [2.83] LNAF (+) 0.1934*** 0.1910*** [17.25] [17.03] NO_BUS_SEGMENTS (+) 0.0664*** 0.0679*** [9.54] [9.75] NO_GEO_SEGMENTS (+) 0.0287*** 0.0280*** [5.38] [5.25] LOSS (+) 0.0876*** 0.0820*** [3.12] [2.92] FISCAL (+) 0.2434*** 0.2413*** [9.56] [9.49] BIG_FIVE (+) 0.1265*** [3.02] IND_SPE1 (+) 0.0336 0.0209 [1.19] [0.74] YEAR EFFECTS Included Included Adj. R^2 82.75% 82.82% 126

Table 4.5 Panel B: Regression of audi fees on produc marke compeiion wih audior specializaion Table 4.5 includes he regression resuls of audi fees on indusry-level compeiion wih audior specializaion. The sample consiss of 2201 firm-years over he period from 2000 o 2004. Dependen variable is he naural log of audi fees. HIndex is Herfindahl- Hirschman index downloaded from he U.S. 2002 Census of Manufacurers. IND_SPE2 is he coninuous variable for audior specializaion. Panel B shows he resuls wih IND_SPE2 in regression. Conrol variables are same as described in previous ables. ***, **, and * represen 1%, 5%, and 10% significance respecively. Prediced Esimaed Esimaed HIndex (?) 0.5629** 0.5640** [ 2.53] [ 2.54] SIZE (+) 0.3533*** 0.3492*** [31.35] [30.58] CURRENT (+) 0.0437 0.0300 [0.61] [0.42] QUICK ( ) 0.0118*** 0.0121*** [ 5.29] [ 5.39] ROA ( ) 0.1294*** 0.1312*** [ 4.27] [ 4.33] DE (+) 0.1331*** 0.1375*** [2.74] [2.83] LNAF (+) 0.1927*** 0.1912*** [17.21] [17.06] NO_BUS_SEGMENTS (+) 0.0664*** 0.0677*** [9.56] [9.71] NO_GEO_SEGMENTS (+) 0.0288*** 0.0282*** [5.39] [5.28] LOSS (+) 0.0846*** 0.0818*** [3.01] [2.91] FISCAL (+) 0.2432*** 0.2415*** [9.56] [9.50] BIG_FIVE (+) 0.1023** [2.12] IND_SPE2 (+) 0.2375*** 0.1228 [2.62] [1.16] YEAR EFFECTS Included Included Adj. R^2 82.79% 82.83% 127

Table 4.6 Panel A: Descripive saisics of produc marke power Table 4.6 includes he resuls of produc marke compeiion and produc marke power on audi fees. The sample consiss of 1970 firm-years over he period from 2000 o 2004. Panel A presens he descripive saisics of produc marke power. Produc marke power is measured as he raio of operaing profi o sales. Panel B presens he regression resuls. Dependen variable is he naural log of audi fees. HIndex is Herfindahl-Hirschman index downloaded from he U.S. 2002 Census of Manufacurers. Conrol variables include audiee size (SIZE naural log of oal asses), curren raio (CURRENT raio of curren asses o oal asses), quick raio (QUICK raio of curren asses less invenory o curren liabiliies), reurn on asses (ROA- he earnings before ineres and ax divided by oal asses), leverage (DE deb o oal asses raio), non-audi fees (LNAF-log of nonaudi fees), number of business segmens (NO_BUS_SEGMENTS), number of geographic segmens (NO_GEO_SEGMENTS), loss year (LOSS=1 if loss year, oherwise 0), audior s repuaion (BIG_FIVE=1 if he auhor belongs o Big Five, oherwise 0), and fiscal year end (FISCAL=1 if fiscal year end is Dec. 31, oherwise 0). ***, **, and * represen 1%, 5%, and 10% significance respecively. N Mean Sd Dev 10% 25% Median 75% 90% PCM 1970 0.0901 0.1896 0.0863 0.0477 0.1101 0.1767 0.2699 128

Table 4.6 Panel B: Relaion beween produc marke power and audi fees Table 4.6 includes he resuls of produc marke compeiion and produc marke power on audi fees. The sample consiss of 1970 firm-years over he period from 2000 o 2004. Panel B presens he regression resuls. Dependen variable is he naural log of audi fees. HIndex is Herfindahl-Hirschman index downloaded from he U.S. 2002 Census of Manufacurers. Produc marke power is measured as he raio of operaing profi o sales. Conrol variables are same as described as in previous ables. Prediced Sign HIndex (?) 0.0346*** 0.0333*** [ 3.16] [ 3.04] PCM ( ) 0.3534*** [ 3.11] SIZE (+) 0.3753*** 0.3810*** 0.3863*** [30.84] [31.03] [31.24] CURRENT (+) 0.2263*** 0.2419*** 0.2222*** [2.82] [3.01] [2.76] QUICK ( ) 0.0311*** 0.0303*** 0.0298*** [ 6.82] [ 6.66] [ 6.54] ROA ( ) 0.3696*** 0.3919*** 0.1356 [ 3.92] [ 4.16] [ 1.08] DE (+) 0.1034* 0.1007* 0.1114** [1.83] [1.78] [1.98] LNAF (+) 0.1825*** 0.1815*** 0.1809*** [15.42] [15.37] [15.35] NO_BUS_SEGMENTS (+) 0.0620*** 0.0598*** 0.0588*** [8.91] [8.57] [8.44] NO_GEO_SEGMENTS (+) 0.0193*** 0.0208*** 0.0221*** [3.49] [3.76] [4.00] LOSS (+) 0.0630* 0.0688** 0.0469 [1.90] [2.08] [1.39] FISCAL (+) 0.2509*** 0.2445*** 0.2469*** [9.61] [9.37] [9.47] BIG_FIVE (+) 0.1423*** 0.1460*** 0.1338*** [3.25] [3.34] [3.06] YEAR EFFECTS Included Included Included Adj. R^2 82.99% 83.08% 83.17% 129

Table 4.7 Newey-Wes resuls: Regression of audi fees on indusry concenraion and firm marke Power Table 4.7 includes he Newey-Wes resuls of indusry-level compeiion and firm-level marke power on audi fees. The sample consiss of 1970 firm-years over he period from 2000 o 2004. Dependen variable is he naural log of audi fees. HIndex is Herfindahl- Hirschman index downloaded from he U.S. 2002 Census of Manufacurers. PCM is he firm-specific produc marke power, calculaed as he operaing profi o sales. Conrol variables include audiee size (SIZE naural log of oal asses), curren raio (CURRENT raio of curren asses o oal asses), quick raio (QUICK raio of curren asses less invenory o curren liabiliies), reurn on asses (ROA- he earnings before ineres and ax divided by oal asses), leverage (DE deb o oal asses raio), non-audi fees (LNAFlog of non-audi fees), number of business segmens (NO_BUS_SEGMENTS), number of geographic segmens (NO_GEO_SEGMENTS), loss year (LOSS=1 if loss year, oherwise 0), audior s repuaion (BIG_FIVE=1 if he auhor belongs o Big Five, oherwise 0), and fiscal year end (FISCAL=1 if fiscal year end is Dec. 31, oherwise 0). ***, **, and * represen 1%, 5%, and 10% significance respecively. Prediced Sign Esimae Saisics HIndex (?) 0.0274** 2.01 PCM ( ) 0.4219** 2.41 SIZE (+) 0.4433*** 26.04 CURRENT (+) 0.3612*** 3.44 QUICK ( ) 0.0308*** 3.43 ROA ( ) 0.1518 1.02 DE (+) 0.1187 1.54 LNAF (+) 0.1172*** 7.07 NO_BUS_SEGMENTS (+) 0.0583*** 6.74 NO_GEO_SEGMENTS (+) 0.0380*** 5.49 LOSS (+) 0.0239 0.60 FISCAL (+) 0.1979*** 6.00 BIG_FIVE (+) 0.0622 1.34 YEAR EFFECTS Included Adj. R^2 77.71% 130

Chaper 5 Conclusions and Fuure Works This disseraion focuses on financial reporing qualiy. I documens he imporance of financial reporing qualiy, conribues produc marke power o he deerminans of financial reporing qualiy and, furher, demonsraes ha financial reporing qualiy is one channel hrough which produc marke compeiion affecs audi fees. The firs essay, Earnings Timeliness and Seasoned Equiy Offering Announcemen Effec explores he role of financial reporing qualiy on he capial raising even. Specifically, i examines wheher he fac ha a firm repors is earnings in a imely way affecs invesors responses a he firm s announcemen of is SEO financing decision. I find ha firms wih greaer earnings imeliness end o have less informaion asymmery beween managers and shareholders. These firms will, herefore, experience less price drops a SEO announcemens. The second and hird essays are among he firs sudies ha posi an economic link beween produc marke compeiion and financial reporing qualiy. They show ha boh firm-level compeiion saus and indusry-level compleion inensiy affec financial reporing qualiy and, furhermore, audi fees. There are many opporuniies for fuure research in relaed areas. For example, disinguishing primary offerings and secondary offerings may yield some ineresing opics. I discuss he benefis of shock-ransfer over accrual managemen, 131

such as less liigaion, scruiny and audi risks. However, I do no explore he coss of price-seing o boos or decrease earnings. Will his ransfer weaken he firm s compeiion saus? Wha are he comparaive coss or benefis o sakeholders beween price-seing o cusomers and earnings managemen? How will he capial marke respond o his behavior? Recen sudies show ha managers have shifed from accrual o real managemen in pos Sarbanes-Oxley Ac (SOX) period. My sudy only considers he accrual managemen and produc marke power o dae. I is of imporance o examine he real managemen behaviors of firms wih differen marke powers. For example, i is necessary o ask wheher firms wih greaer power end o engage in real managemen aciviies less and, alhough price-seing aciviies aken by firms wih greaer power do no affec a firm s normal operaional pracices as real manipulaions do, wha he economic consequences of a firm s ransfer shocks o cusomers are. In addiion, i is necessary o ask wha oher facors help o decide heir price-seing abiliies. Anoher poin I need o make regarding he hird essay is ha, due o he availabiliy of indusry-level compeiion measures daa, he sudy on compeiion and audi fees is based on daa in he U.S. manufacuring indusry only. I is worhwhile o examine, hen, wheher he conclusion sill holds for all indusries so ha we can generalize he resuls. 132