Invesmen Managemen and Financial Innovaions, Volume 8, Issue 4, 2011 Juo-Lien Wang (Taiwan), Her-Jiun Sheu (Taiwan), Huimin Chung (Taiwan) Corporae governance reform and earnings managemen Absrac This paper explores wheher he Sarbanes-Oxley Ac (he SOX Ac) of 2002 is associaed wih he incidence of earnings managemen in he US. The resuls reveal significan reducions in abnormal accruals afer he implemenaion of he SOX Ac. Furhermore, following he implemenaion of he SOX Ac, he auhors find an associaion beween firms wih high pre-managed earnings and fewer incidences of income-reducing earnings managemen behavior. In conras, here is no evidence o sugges ha he SOX Ac has succeeded in resraining income-increasing manipulaion by firms wih poor pre-managemen earnings. Our findings sugges ha he SOX Ac has conribued significanly o he inegriy of financial saemens; however, for hose firms wih high incenives o achieve earnings benchmarks, he effec is limied. Keywords: earnings managemen, Sarbanes-Oxley Ac, discreionary accruals. JEL Classificaion: G01, G30, M40. Inroducion High-profile failures in he US corporae financial reporing have raised concerns regarding he inegriy of public financial informaion, promping he inroducion of he Sarbanes-Oxley Ac (he SOX Ac) of 2002 as a direc resul of he erosion of invesor confidence (Jain e al., 2008). These corporae scandals have demonsraed ha aggressive earnings managemen, indicaed by lower qualiy accouning informaion, is accompanied by serious shareholder losses. Consequenly, earnings managemen can provide an imporan signal showing ha, in pursuing privae benefis, managers are sacrificing shareholder wealh. The SOX Ac was designed o reform corporae governance, increase he accuracy and reliabiliy of corporae disclosure and reduce he likelihood of missaemens in financial reporing. For example, o reinforce he responsibiliies of Chief Execuive Officers (CEOs) and Chief Financial Officers (CFOs), he SEC adoped Secion 302 of he SOX Ac, which mandaes ha CEOs and CFOs of companies reporing o he SEC should provide personal cerificaions in each of heir quarerly and annual repors. These cerificaions should affirm ha he signing officer has reviewed he repor, and i is fair and free of maerial missaemens. The SOX Ac hus is expeced o aler managerial behavior in accouning ransparency and earnings managemen. Li e al. (2008) sugges ha invesors anicipaed ha he more firms had previously managed heir earnings, he more he Ac would limi earnings managemen and increase he qualiy of financial saemen informaion. However, Li e al. (2008) don esimae firm earnings managemen for he years afer he SOX Ac. This sudy explores one of he fundamenal goals of he Ac: wheher he en- Juo-Lien Wang, Her-Jiun Sheu, Huimin Chung, 2011. forcemen of he SOX Ac is associaed wih a decline in firm earnings manipulaion, paricularly for firms wih a high incenive o manage earnings. If he SOX Ac have improved financial reporing accuracy and reliabiliy, considerably less earnings managemen would be observed following is implemenaion. This sudy concenraes on he effec of he announcemen of he SOX Ac on discreionary accruals. To direcly capure he exen o which discreionary accruals relaing o firm prior performance are managed, his sudy employs a imeseries modified-jones model (Jones, 1991) o esimae he degree of earnings managemen, doing so by comparing he abnormal accruals beween differen periods wihin individual firms. Cohen e al. (2008) find ha firm managemen of earnings peaked around he passage of he SOX Ac, followed by a significan decline; however, heir sudy differs from he presen sudy in erms of boh is focus and he mehodology adoped for measuring earnings managemen. The mehodology adoped in his sudy for measuring earnings managemen (he ime-series modified-jones model) focuses more on deecing manipulaion variaions wihin an individual company. Besides, in conras o Cohen e al. (2008) sudy, his sudy explores variaions in he effec of earnings managemen across firm size, and focuses on firms wih high earnings manipulaion incenives by examining upward and downward manipulaions. When pre-managemen earnings are low, firms end o manage earnings upwards for psychological perspecive and o avoid high cos of capial. Furhermore, firms wih exremely high pre-managed earnings also have incenives o manage earnings downwards (Degeorge e al., 1999). This sudy hus invesigaes he robusness of he resuls by considering wo of he mos frequenly considered objecives: 109
Invesmen Managemen and Financial Innovaions, Volume 8, Issue 4, 2011 avoiding losses, and meeing prior period earnings (Degeorge e al., 1999; and Barov e al., 2002). This sudy examines no only he link beween he inroducion of he SOX Ac and income-increasing earnings managemen when pre-managed earnings are less han he hreshold arge, bu also ess wheher he SOX Ac influenced downwards manipulaion of earnings under circumsances of high pre-managed earnings. The findings reveal a significan reducion in US corporae earnings managemen following he SOX Ac, consisen wih he widely-held view ha he Ac conribued o improvemens in he qualiy of accouning informaion. The effec of he Ac on improving financial ransparency is boh for small firms and large ones, essenially because he SOX Ac comes ino force for all lised firms. This sudy also idenifies an associaion beween firms wih high pre-managed earnings and fewer incidences of income-reducing earnings managemen behavior; however, here is no evidence ha he SOX Ac has successfully limied income-increasing manipulaion by firms wih poor pre-managemen earnings. Several sudies have examined he impac of and marke responses o he SOX Ac in specific areas, wih some idenifying a variey of posiive effecs (Li e al., 2008; Jain e al., 2008; Kalelkar and Nwaeze, 2011), whils ohers have revealed several negaive effecs (Leuz e al., 2008; Chhaochharia and Grinsein, 2007). The sudy hus poenially conribues o he policy implicaions of corporae governance regulaions. The remainder of his paper is organized as follows. Secion 1 discusses he exan lieraure on invesor proecion and earnings managemen, followed in secion 2 by a descripion of he daa used, an explanaion of he research design and presenaion of he mehods used o idenify earnings managemen. The empirical resuls are presened in secion 3. The final secion presens he conclusions. 1. Relaed lieraure and hypohesis Earnings managemen involves he aleraion, or manipulaion, of firm repored economic performance by insiders, eiher o mislead cerain sakeholders or o influence conracual oucomes (Healy and Wahlen, 1999). Prior sudies have suggesed ha aggressive earnings managemen increases informaion asymmery beween insiders and ousiders, has he poenial o reduce shareholder wealh, and demonsraes lower accouning qualiy (Teoh e al., 1998). The evidence of Dechow and Dichev (2002) also show ha high earnings managemen signified lower qualiy and less persisen earnings. Previous sudies have suggesed ha whils insiders are likely o engage in aggressive earnings managemen o diver firm resources o hemselves, effecive laws and srong enforcemen may reduce such insider incenives and miigae such behavior (Leuz e al., 2003; Burgsahler e al., 2006). The SOX Ac aims o proec invesors by reinforcing corporae governance and improving he accuracy and reliabiliy of corporae disclosure. Li e al. (2008) sugges ha invesors anicipaed ha he SOX Ac would limi earnings managemen and enhance financial saemen informaion qualiy. This work focuses on he role of he Ac in consraining earnings managemen and hypohesizes ha earnings managemen should prove o be far less pervasive as a resul of he implemenaion of he SOX Ac. If he SOX Ac really improved he financial disclosure accuracy, his sudy predics ha earnings managemen would reduce following he inroducion of he Ac. Burgsahler and Dichev (1997) find ha when firms face sligh decrease or negaive pre-managemen earnings, execuives end o manage earnings upwards o avoid earnings decreases and losses. This work hus furher explores wheher he SOX Ac has inroduced processes ha can effecively reduce he incidence of earnings managemen in cases where firms have undesirable performance and he incenives for earnings manipulaion are high. Earnings managemen is no resriced solely o income-increasing behavior; for example, managers may be unwilling o repor subsanial gains in earnings because hey insincively know ha his will increase heir fuure performance arges. Consequenly, firms wih eiher exremely high or unwillingly low pre-managed earnings may have incenives o manage earnings downward. On he basis of he above discussion, his sudy predics ha execuives end o manage earnings upward (downward) when facing exremely low (high) pre-managed earnings and performing earnings manipulaion. However, manage earnings upward or downward is wha he SOX Ac wans o resric o. The SOX Ac aims o reduce firm incenives o conceal heir real operaing performance and may reduce earnings managemen for firms wih undesirable pre-managed earnings. This sudy ess he incenives for earnings managemen by comparing pre-managed earnings wih arge earnings levels and employs wo objecive benchmarks: (1) zero and (2) earnings repored in he previous year. If he SOX Ac does have an associaion wih curren earnings managemen, he abiliy o deec such a relaionship should be he greaes a close proximiy o he hreshold poins. The research hypohesizes ha he SOX Ac really achieves he purpose of improving financial disclosure accuracy, even for firms wih high incenives o manage earnings, and predics ha upward and downward 110
manipulaion decline following he SOX Ac. We invesigae ha wheher he SOX Ac limis incomeincreasing (income-decreasing) earnings managemen when pre-managed earnings undershoo (significanly exceeds) hese hreshold poins. Boh upward and downward manipulaions are explicily examined. Because he benefis associaed wih oversaing earnings exceed hose associaed wih undersaing hem, his sudy anicipaes asymmery wih regard o upward and downward manipulaions and hus hypohesizes ha managerial incenives o increase earnings exceed heir incenives o decrease hem. 2. Daa source and mehodology 2.1. Measuring earnings managemen. This sudy mainly examines wheher earnings managemen has declined following he implemenaion of he SOX Ac. Alhough Barov e al. (2000) and Shaw (2003) sugges ha he cross-secional version of he modified-jones model is superior o is ime-series counerpar, heir ess evaluae he abiliy of discreionary accrual models o idenify firms engaging in exreme forms of earnings managemen, under he cavea ha he resuls may no be generalized o firms engaging in moderae levels of earnings managemen wihin generally acceped accouning principles (GAAP). This sudy aims o deec ear- Accruals 1 Sales PPE i, s i, s i, s i, i, i, i, i, s1 i, s1 i, s1 i, s1 Invesmen Managemen and Financial Innovaions, Volume 8, Issue 4, 2011 nings managemen in general firms observing GAAP, raher han in hose engaging in excessive earnings fraud. Furhermore, he cross-secional version of he modified-jones model, which focuses on comparing discreionary accruals wihin he same indusry period, does no mach he overall objecives of his invesigaion. For example, large earnings managemen proxy of he cross-secional version indicaes ha a firm manages earnings more han oher firms wihin he same indusry period. If he proxy decreases during he nex period, he accruals declines compared o oher companies wihin he same indusry period, bu no compared o firm hisorical daa. The hypoheses of his sudy call for direc measuremen of managerial engagemen in discreionary accruals relaed o hisorical performance of he firm, as well as he use of a ime-series modified Jones model (Jones, 1991) o esimae he exen of earnings managemen by comparing abnormal accruals beween differen periods wihin individual firms. To esimae non-discreionary accruals, his sudy regresses he accruals on he changes in revenues and he levels of propery, plan and equipmen and esimaes he parameers of he following modified- Jones model, which is a ime-series ordinary leas squared (OLS) regression model: TA TA TA TA, (1) s = 1 o 9 for each firm i in year, where Accruals denoes he oal accruals for firm i in year, measured as he ne income before exraordinary iems minus cash flow from operaions; Sales represens he change in sales for firm i in year ; PPE is gross propery, plan and equipmen in year ; TA -1 denoes he book value of oal asses for firm i from he previous year; and, and are firm-specific parameers for sample year. The regression equaion NDA 1 ˆ ( SALES TR ) PPE is deflaed by lagged oal asses o reduce heeroskedasiciy. Following Dechow e al. (1995), firms wih fewer han nine observaions for parameer esimaion are excluded from he sample. The coefficien esimaes from equaion (1) are used o esimae he firm-specific non-discreionary accruals (NDA ) for each firm: ˆ ˆ, (2) TAi, 1 TAi, 1 TAi, 1 where ˆ, ˆ and are OLS esimaes for he i, regression parameers in equaion (1), and TR denoes he change in rade receivables, subraced o permi he possibiliy of credi sales managemen by he company (Dechow e al., 1995). Discreionary accruals (DA ) is hen he remaining porion of he oal accruals: Accruals DA NDAi,. (3) TAi, 1 Accruals reverse over ime, and he managemen of earnings eiher upwards or downwards are hypohesized o be earnings managemen. Following Leuz e al. (2003), he hypohesis of his sudy does no rely on he direcion of he discreionary accruals, bu raher on he magniude; hus, he es saisics are based on he value of he absolue discreionary accruals (ADA). In order o eliminae operaional variaion, which can cause unreasonable variaions in oal accruals, firms wih ADA > 1 are excluded from he sample 1. 1 ADA > 1 means he accouning discreionary accruals is greaer han firm s lagged oal asses and is supposed o be unreasonable. There are 22 firm-years in such case during he sample period. This sudy also considered he crieria 0.9 and 0.8; however, hese alernaive limis produced qualiaively similar resuls. 111
Invesmen Managemen and Financial Innovaions, Volume 8, Issue 4, 2011 Recen sudies focus on he effec of individual firm behavior on earnings managemen and compare accruals wihin a single indusry period. In conras, his work focuses on he general effec of he SOX Ac on he US business environmen. If he subsanive reforms associaed wih he Ac in 2002 have improved he reliabiliy of financial reporing and reduced discreionary accruals while mainaining non-discreionary accruals, he dependen variables (oal accruals) in equaion (1) will be decreased while he independen variables mainain heir usual level. The esimaed parameers in equaion (1) may auomaically be diminished and he esimaed NDA in equaion (2) migh also be underesimaed. This violaes he assumpion of his invesigaion ha non-discreionary accruals are holding ou. As a resul, discreionary accruals (DA ) may exhibi esimaion error. Any error in esimaing nondiscreionary accruals will lead o equal error in esimaing discreionary accruals, possibly causing an assumed relaionship of earnings managemen beween he pre- and pos-sox Ac periods. In he unabulaed sensiiviy es, his sudy also adops he crosssecional modified Jones model o esimae discreionary accruals and obain similar resuls. However, owing o possible error, his sudy sill feaures in imeseries version of he modified Jones model. 2.2. Daa and sample selecion. To some exen, earnings managemen is an overall accouning arrangemen, and ime is required for adjusmens o discreionary accruals o feed hrough. If managers manipulae earnings, he effecs of such manipulaion will ulimaely unwind and evenually be reversed a he same amoun, albei coming ino play during subsequen years. On he implemenaion of he SOX Ac, accouning officers would have needed ime o reac o he change in he accouning 112 environmen. This sudy hus adops a pre-sox sample period comprising 1999 o 2001 o ensure a sufficienly large sample, as well as a pos-sox sample period covering 2002-2004. The daa was obained from he COMPUSTAT daabase for he period of 1989-2004 o obain finance daa o esimae earnings managemen proxy. Sample firms mus have all of he necessary relaed financial daa. This resricion inroduces a survivorship bias o he sample resuling from larger and more successful enrepreneurs. Firms closed during he sample period are excluded from he sample. Many of hese firms may confron financial difficuly before erminaion and aemp o manage earnings aggressively, and herefore he earnings managemen measures of hese firms may be much larger han hose of oher firms and become he exremely values of he sample. We expec ha his survivorship bias reduces he variaion in earnings managemen measures, making i a conservaive es of he research quesion. Banks and financial insiuions (SIC codes 6000-6999) were excluded from he sample because of heir differen accrual procedures. To conrol for he possible influence of exreme observaions, his sudy winsorizes all observaions below he 1 s and above he 99 h percenile of observaions. Afer implemening hese filers, he sample comprises 1,149 firms (6,894 firm-years) wih he presence of 66 separae wo-digi SIC codes, indicaing a paricularly wide selecion of indusries. 2.3. The models. This sudy firs ess he impac of he SOX Ac on he pervasiveness of earnings managemen by esimaing he following pooled OLS regression: ADA SOX ROA LTA GROWTH MB OPP, (4) i, 0 1 i, 2 i, 3 i, 4 i, 5 i, 6 i, i, where ADA i is he proxy of earnings managemen explained above, and SOX i is a dummy variable ha equals o 1 for all pos-sox periods, oherwise 0. If earnings managemen becomes less pervasive afer he implemenaion of he SOX Ac, his sudy predics ha he SOX coefficien will be significanly negaive. Equaion (4) also includes proxies for oher facors ha migh affec earnings manipulaion. In many companies, he sock price and managers compensaion are ied o earnings performance, his may moivae managers o engage in earnings manipulaion. A posiive relaionship beween discreionary accruals and firm profiabiliy is found by Lee e al. (2006); however, a negaive relaionship is also found by Chung e al. (2009). Following ha, his sudy adops reurn on asses (ROA) as a proxy o capure firm performance bu no direcion is prediced. Managers of large firms may have greaer incenives o manipulae earnings in order o reduce coss; on he oher hand, since hey are acively followed by ouside capial markes, such firms may be less able o hide earnings managemen behavior. This sudy hus uses he logarihm of oal asses (LTA) as a proxy o capure firm size and informaion environmen; however, no direcion is prediced. Given ha i is much more difficul o scruinize he aciviies of rapidly-growing firms, i is much easier for rapidly-growing firms o manage heir earnings han slower-growing firms. Dechow e al. (1996) demonsrae ha firms which are alleged o have violaed GAAP by oversaing heir repored earnings have higher marke-o-book raios vis-à-vis a conrol group, and sugges ha invesors expec hese firms o have higher growh opporuniies.
Park and Shin (2004) also find earnings managemen o be posiively correlaed wih firm growh opporuniies. Alhough marke-o-book raio and sales growh boh measure firm growh opporuniies, here is lile difference beween hem. For firms considered o have high profi growh in he near fuure, while heir realized revenue does no increase, heir marke-o-book raio indicaes high revenue growh. This sudy hus measures he curren and fuure growh opporuniies using ne revenue growh (GROWTH) and marke-o-book raio (MB), respecively. The esimaed coefficiens of he conrol variable for GROWTH and MB are posiive. Burgsahler and Dichev (1997) also argue ha firms wih high levels of curren asses and curren liabiliies were likely o find i relaively cheaper o manage earnings; his variable is ermed manipulaion DA ROA 6 BELOW 0 1 LTA 7 HIGH GROWTH 8 2 SOX 3 BM 9 Invesmen Managemen and Financial Innovaions, Volume 8, Issue 4, 2011 opporuniy (OPP) and calculaed as follows: curren asses plus curren liabiliies less cash a he end of year 1, scaled by lagged asses. This sudy predics ha he esimaed coefficien for OPP is posiive. Nex, his sudy aemps o isolae incenives for earnings managemen by comparing pre-managed earnings and arge earnings. To avoid he backing-ou problem (Peasnell e al., 2005), his work uses cash flow from operaions as he insrumen for pre-managed earnings (PME ). This work invesigaes wheher he inroducion of he SOX Ac influenced he likelihood of upward (downward) earnings managemen when PME undershoos (considerably exceeds) he arges, by separaely esimaing he following pooled OLS regression for boh earnings hresholds: 4SOX BELOW OPP. 10 SOX 5 HIGH (5) The absolue value of discreionary accruals (ADA) does no conain he informaion on upward or downward manipulaion of repored earnings invesigaed in equaion (5). This sudy hus uses he original discreionary accruals proxy, DA, for his es. HIGH and BELOW are dummy variables. Equaion (5) has wo benchmarks: zero and repored earnings for he previous year (EARN -1 ). Therefore, boh HIGH and BELOW have wo definiions: 1. For he regressions where pre-managed earnings (PME ) is benchmarked agains zero, HIGH is 1 if PME PME scaled by oal asses for Toal asses firm i in period exceeds he 3 rd quarile of he disribuion of posiive PME scaled by oal asses PME in he indusry, and 0 oherwise. Toal asses Meanwhile, BELOW akes he value 1 if PME is negaive, and 0 oherwise. 2. For he regressions where PME is benchmarked agains EARN -1, his sudy defines HIGH as 1 if PME minus EARN -1, scaled by oal asses PME EARN for firm i in period, exceeds Toal asses he 3 rd quarile of he disribuion of posiive premanaged earnings changes in Toal asses PME EARN he indusry, and oherwise as 0. Meanwhile, BE- LOW akes a value of 1 if PME < EARN -1, and 0 oherwise. If firms really manage earnings upward (downward) when pre-managemen earnings are exremely low (high), he coefficien of BELOW (HIGH ) would be significanly posiive (negaive). Moreover, if he Ac successfully improves financial disclosure accuracy, hen even for firms wih srong incenives o manage earnings, his sudy predics ha upwards and downwards manipulaion would be declined afer he inroducion of he SOX Ac and he coefficien of SOX BELOW (SOX HIGH ) would be significanly negaive (posiive). 3. Empirical resuls 3.1. Descripive saisics. Table 1 liss he summary saisics for he absolue value of discreionary accruals (ADA) and oher financial variables, wih Panel A including he descripive saisics for he enire sample, and Panels B and C respecively lising he descripive saisics for he pre- and pos- SOX periods. Consisen wih Cohen e al. (2008), ADA represens approximaely 7.44 percen of oal asses, and ranges beween 97.9 percen and 0.0004 percen, whils he mean absolue discreionary accruals are 7.95 percen of oal asses for he pre- SOX period, and 6.92 percen of oal asses for he pos-sox period. In general, he firm characerisics of he wo periods appear o be differen. To es his, we perform -ess and Wilcoxon rank sum ess (woailed) of he equaliy of he variables. The es resuls find ha he earnings managemen degree of he pre-sox observaions significanly exceed heir pos-sox counerpars a he 1 percen level and iniially verify ha earnings managemen declined 113
Invesmen Managemen and Financial Innovaions, Volume 8, Issue 4, 2011 afer he enforcemen of he SOX Ac. I also shows ha companies of he pos-sox period has significanly larger firm size (LTA) and has a significanly smaller profiabiliy (ROA), revenue growh (GROWTH), marke-o-book raio (MB) and earnings manipulaion opporuniy(opp). I migh be because firm grows up over he years, and he firm size becomes larger. In addiion, he SOX Ac and oher economic siuaions a ha ime make hese variables srucurally changed. Table 1. Descripive saisics of he variables Pearson and Spearman correlaion coefficiens are conduced and lised in Table 2, hereby providing some basic analysis of he correlaion beween variables. The correlaion coefficiens reveal ha ADA is negaively correlaed wih profiabiliy (ROA), firm size (LTA) and revenue growh (GROWTH), and posiively correlaed wih marke-o-book raio (MB) and manipulaion opporuniy (OPP). In some degree, i iniially fis in wih he esimaed relaionship beween earnings managemen and conrolled variables. Variables a Mean Median Max. Min. Panel A. Toal sample (6,894 firm-years) b ADA(%) 7.442 4.647 97.999 4x10 4 ROA(%) 3.188 4.293 61.926-125.609 LTA 5.987 6.140 10.733 0.248 GROWTH 0.089 0.064 2.921-0.778 MB 2.591 1.831 34.146-20.305 OPP 0.594 0.578 4.212 0.042 Panel B. Pre-SOX sample (3,447 firm-years) c ADA(%) 7.955 4.997 97.999 4x10 4 ROA(%) 3.536 4.555 61.926-125.609 LTA 5.918 6.043 10.477 0.028 GROWTH 0.101 0.063 2.921-0.737 MB 2.728 1.762 34.129-12.921 OPP 0.611 0.603 1.722 0.042 Panel C. Pos-SOX sample (3,447 firm-years) c ADA(%) 6.929 4.359 94.432 0.002 ROA(%) 2.840 4.063 61.345-99.708 LTA 6.055 6.237 10.733-0.248 GROWTH 0.077 0.064 1.796-0.778 MB 2.452 1.881 34.146-20.305 OPP 0.577 0.555 4.212 0.052 Noes: a ADA is he absolue value of he ime-series version of he modified Jones model of discreionary accruals; ROA is he reurn on asses; LTA represens he naural log of oal asses a he end of he year; GROWTH refers o ne revenue growh; MB is he marke-o-book raio; OPP is he curren asses plus curren liabiliies less cash a he end of year 1 scaled by lagged asses. b The oal sample comprises of 1,149 lised firms (6,894 firm-year observaions) covering he period from 1999 o 2004. c The pre-sox sub-sample conains observaions for he period of 1999-2001, whils he pos-sox sub-sample conains observaions for he period of 2002-2004. Table 2. Correlaion coefficiens ac Variables b ADA ROA LTA GROWTH MB OPP ADA -0.228 *** -0.259 *** 0.015 0.039 *** 0.188 *** ROA -0.127 *** 0.231 *** 0.248 *** 0.077 *** -0.007 LTA -0.270 *** 0.152 *** 0.055 *** 0.148 *** -0.217 *** GROWTH -0.048 *** 0.358 *** 0.096 *** 0.130 *** -0.074 *** MB -0.050 * 0.415 *** 0.300 *** 0.260 *** -0.017 OPP 0.212 *** 0.050 *** -0.201 *** -0.060 *** -0.046 *** Noes: a The sample comprises of 1,149 lised firms (6,894 firm-year observaions) covering he period of 1999-2004. *** indicaes significance a he 1% level; ** indicaes significance a he 5% level; and * indicaes significance a he 10% level (wo-ail es). b ADA is he absolue value of he ime-series version of he modified Jones model of discreionary accruals; ROA is he reurn on asses; LTA represens he naural log of oal asses a he end of he year; GROWTH refers o ne revenue growh; MB is he markeo-book raio; OPP is he curren asses plus curren liabiliies less cash a he end of year 1 scaled by lagged asses. c Pearson correlaions are presened below he diagonal and Spearman correlaions are presened above he diagonal. 3.2. Effecs of he SOX Ac. 3.2.1. Tes of earnings managemen surrounding he SOX Ac. The effecs of he SOX Ac vary wih firm size (Chhaochharia and Grinsein, 2007). To es wheher he SOX Ac works well on earnings managemen for boh large firms and small firms, his sudy hus sors he 1,149 114
sample firms by oal asses in he year 1999 ino four quariles, and labels he quarile of firms wih he larges (smalles) oal asses as he large ( small ) firms. Boh he large and small sub-samples conain 287 firms (1,722 firm-years). The regression resuls of equaion (4), for all, large and small firms, are presened in Table 3. The variance inflaion facors (VIFs) measure he exen o which mulicollineariy exiss in he seleced explanaory variables. The VIFs of all he independen variables are below 2, indicaing ha he mulicollineariy problem does no exis. This sudy also follows he regression diagnosic suggesed by Belsley e al. (1980) o explore he collineariy of he independen variables and compue he condiion indexes (CI). The larges CI in he empirical resuls presened in his sudy was 10.30, Invesmen Managemen and Financial Innovaions, Volume 8, Issue 4, 2011 well below he rule of humb of CI = 30. Consequenly, he above resuls indicae ha he high collineariy problem does no exis 1. The significanly negaive coefficien esimae on SOX of oal sample in Table 3 is consisen wih he hypohesis ha he passage of he SOX Ac is associaed wih reduced earnings managemen and is insensiive o he inclusion of conrol variables designed o measure oher aspecs of firm governance srucures. Consisen wih prior sudies (Park and Shin, 2004; and Burgsahler e al., 2006), his sudy finds ha greaer earnings managemen may be associaed wih poor profiabiliy, small size or rapid growh. The coefficien on OPP is significanly posiive, suggesing ha firms wih greaer curren asses and curren liabiliies have higher levels of absolue discreionary accruals. Table 3. OLS regression resuls of he absolue discreionary accruals on he passage of SOX a Variables c Prediced sign Toal sample Large firms Small firms Coefficien -value b Coefficien -value b Coefficien -value b Inercep +/- 9.456 *** 19.96 8.132 *** 4.36 9.218 *** 7.13 SOX - -0.755 *** -3.80-0.757 *** -2.55-1.391 *** -2.55 ROA(%) +/- -0.164 *** -8.43-0.219 *** -3.63-0.123 *** -4.54 LTA +/- -0.775 *** -13.90-0.479 ** -2.33-0.462-1.46 GROWTH + 2.688 *** 3.62 2.423 ** 2.01 1.895 1.27 MB + 0.226 *** 4.18 0.355 *** 3.84 0.205 ** 2.10 OPP + 4.554 *** 11.13 2.474 *** 3.49 4.895 *** 5.80 F value 175.30 19.46 20.69 Adj. R 2 (%) 13.25 6.37 6.75 Toal No. of firm-years b 6,894 1,722 1,722 Noes: a We repor one-ailed ess for he variables wih prediced signs; wo-ailed for hose wih no predicion. *** indicaes significance a he 1% level; ** indicaes significance a he 5% level; and * indicaes significance a he 10% level. b All -values of coefficien are calculaed using Whie (1980) robus sandard errors o correc for heeroscedasiciy. c ADA is he absolue value of he ime-series version of he modified Jones model of discreionary accruals; ROA is he reurn on asses; LTA represens he naural log of oal asses a he end of he year; GROWTH refers o ne revenue growh; MB is he marke-o-book raio; OPP is he curren asses plus curren liabiliies less cash a he end of year 1 scaled by lagged asses. In Table 3, he coefficiens of SOX of boh large and small firms are significanly negaive, which means ha he degree of earnings managemen is reduced afer he implemenaion of he SOX Ac. The earnings managemen proxies for large firms (5.80 and 4.70 percen for pre- and pos-sox, respecively) are, on he whole, less han hose for small firms (11.97 and 10.33 percen for pre- and pos-sox, respecively). A similar case is also shown in Teoh e al. (1998), wih he smalles absolue discreionary curren accruals quarile ending o conain larger firms. Unabulaed regression is performed o es if he impac of he SOX Ac on earnings managemen of he large firms is differen from which of he small firms. However, no evidence shows ha size effec exiss. The effec of he Ac on improving financial ransparency is boh for small firms and large ones, essenially because he SOX Ac comes ino force for all lised firms. 3.2.2. Beaing he benchmarks. Our basic model for capuring he impac of he SOX Ac on abnormal accrual aciviy is provided by equaion (5), from which we conduc hree ess of accruals managemen around he benchmark zero and prior earnings. The resuls are repored in Table 4 and Table 5, respecively. The firs es (T11 in Table 4 and T21 in Table 5) considers all observaions o invesigae wheher firms manage heir earnings upwards (downwards) when pre-managemen earnings are prey low (high). This es reveals he posiive (negaive) and significan coefficien esimaes on BELOW (HIGH) in T11 and T21. 1 1 Belsley e al. (1980) propose ha a CI of 30 o 100 indicaes moderae o srong collineariy. 115
Invesmen Managemen and Financial Innovaions, Volume 8, Issue 4, 2011 116 Table 4. OLS regression resuls of he discreionary accruals on he passage of SOX, esing earnings agains he benchmark zero a (T12) All observaions wihou hose wih PME < las period earnings (T13) Observaions where prior period earnings failed o mee zero Variables b Prediced sign (T11) All observaions Coefficien -value c Coefficien -value c Coefficien -value c Inercep -1.717 *** 2.89-1.719 *** -2.74 0.218 0.13 BELOW + 8.949 *** 11.74 8.580 *** 5.93 6.606 *** 4.59 HIGH - -6.579 *** -15.54-6.054 *** -13.81-7.162 *** -4.34 SOX - 0.790 2.86 0.853 2.87 0.345 0.39 SOX BELOW - 0.142 0.13 2.093 1.11 1.707 0.96 SOX HIGH + 1.562 *** 2.89 1.469 *** 2.66 1.363 0.66 ROA (%) +/- 0.641 *** 24.52 0.619 *** 19.91 0.720 *** 14.56 LTA +/- 0.010 0.16 0.050 0.69 0.044 0.26 GROWTH + -3.786-2.43-4.140-2.23-3.002-1.86 MB + -0.093-1.42-0.099-1.39-0.156-1.09 OPP + -1.234-2.49-2.185-3.95-0.016-0.01 F value 183.90 134.15 47.97 Adj. R 2 (%) 22.13 20.14 25.86 Toal Obs. of firm-years 6,480 5,329 1,386 Obs. where BELOW = 1 620 224 305 Obs. where HIGH = 1 1,470 1,430 172 Noes: a We repor one-ailed ess for he variables wih prediced signs; wo-ailed for hose wih no predicion. *** indicaes significance a he 1% level; ** indicaes significance a he 5% level; and * indicaes significance a he 10% level. b DA is he value of he ime-series modified-jones model of discreionary accruals; SOX is a dummy variable which is equal o 1 for all pos-sox periods, oherwise 0; BELOW is an indicaor variable aking he value of 1 if pre-managed earnings (proxied by operaing cash flow) are below zero (oherwise 0); HIGH is an indicaor variable aking he value of 1 if pre-managed earnings (proxied by operaing cash flow) exceed zero by a large margin (oherwise 0); ROA is he reurn on asses; LTA represens he naural log of oal asses a he end of he year; GROWTH refers o ne revenue growh; MB is he marke-o-book raio; OPP is he curren asses plus curren liabiliies less cash a he end of year 1 scaled by lagged asses. c All -value of coefficien are calculaed using Whie (1980) robus sandard errors o correc for heeroscedasiciy. Table 5. OLS regression resuls of he discreionary accruals on he passage of SOX, esing benchmark prior-period earnings a (T22) All observaions wihou hose wih PME < zero (T23) Observaions where prior period earnings failed o mee he benchmark (T21) All observaions Variables b Prediced sign Coefficien -value c Coefficien -value c Coefficien -value c Inercep -2.555 *** -4.34-3.007 *** -5.1-1.189-1.33 BELOW + 5.068 *** 10.24 3.194 *** 6.09 5.267 *** 5.82 HIGH - -1.022 ** -1.84-1.308 *** -2.37 0.104 0.13 SOX - 1.143 4.33 1.003 3.86 0.451 1.07 SOX BELOW - 0.127 0.18 0.698 0.93 0.442 0.35 SOX HIGH + 1.358 ** 1.82 1.201 ** 1.66 0.589 0.58 ROA(%) +/- 0.461 *** 19.52 0.503 *** 17.98 0.497 *** 13.31 LTA +/- -0.040-0.58 0.098 1.39-0.041-0.40 GROWTH + -1.968-2.27-2.532-2.75-2.850-2.20 MB + -0.148-2.16-0.235-3.46-0.113-1.11 OPP + -0.695-1.36-1.653 3.10-2.571-3.11 F value 112.39 98.82 51.07 Adj. R 2 (%) 14.80 14.31 15.55 Toal Obs. of firm-years 6,480 5,860 2,784 Obs. where BELOW = 1 1,151 755 343 Obs. where HIGH = 1 1,359 1,273 814 Noes: a We repor one-ailed ess for he variables wih prediced signs; wo-ailed for hose wih no predicion. *** indicaes significance a he 1% level; ** indicaes significance a he 5% level; and * indicaes significance a he 10% level. b DA is he value of he ime-series modified Jones model of discreionary accruals; SOX is a dummy variable which is equal o 1 for all pos-sox periods, oherwise 0; BELOW is he indicaor variable aking he value of 1 if pre-managed earnings (proxied by operaing cash flow) are below zero (oherwise 0); HIGH is an indicaor variable aking he value of 1 if pre-managed earnings (proxied by operaing cash flow) exceed prior-period earnings by a large margin (oherwise 0). c All -value of coefficien are calculaed using Whie (1980) robus sandard errors o correc for heeroscedasiciy.
The resuls are consisen wih he predicion ha managers manipulae earnings upwards (downwards) when pre-managed earnings are less han benchmarks (significanly exceed benchmarks). The esimaed coefficien for he ineracion erm SOX BELOW is no significan a convenional levels, which suggess no evidence of firms wih poor pre-managed earnings reducing heir income-increasing earnings manipulaion wih he enforcemen of he SOX Ac. In conras, he esimaed coefficiens on he SOX HIGH ineracion erm are posiive and significan, suggesing ha income-decreasing earnings managemen behavior is reduced afer he inroducion of he SOX Ac. The above es of all observaions migh be confounded by manager aemps o mee anoher benchmark. For example, T11 is designed o es wheher firms manage heir earnings upwards (downwards) when pre-managemen earnings agains he benchmark zero are very low (high). The es resuls migh be confounded by firms wih pre-managemen earnings failing o achieve anoher benchmark prior-period earnings. To accoun for such poenial confounding, his sudy excludes observaions where pre-managed earnings (PME) are below prior-period earnings in he second es T12 in Table 4. Similarly, his sudy excludes observaions where pre-managed earnings (PME) are below zero in T22 in Table 5. The resuls lised in T12 and T22 are consisen wih he firs esimaion, he coefficiens on BELOW (HIGH) are significanly posiive (negaive), wih he coefficien on SOX BELOW being insignifican, whils he SOX HIGH ineracion erm is posiive and significan. Burgsahler and Dichev (1997) demonsrae ha incenives o avoid losses or reduced earnings increase when firms fail o mee he benchmarks in he prior period. Therefore, in he hird es (T13) lised in Table 4, his sudy explores he impac of he SOX Ac on firm earnings managemen following a period of negaive earnings, and focuses on hose observaions wih poor prior-period earnings. Idenically, his sudy examines he impac of he SOX on he exen of earnings mana- References Invesmen Managemen and Financial Innovaions, Volume 8, Issue 4, 2011 gemen by firms following a period of earnings decreases in T23 in Table 5. The resuls show ha he coefficiens on boh SOX i BELOW i and SOX i HIGH i are insignifican and sugges ha for firms wih poor prior earnings, here was no evidence o show ha he manipulaion of earnings had been consrained by he SOX Ac. According o Tables 4 and 5, he coefficiens of SOX i BELOW i are all insignifican, while several coefficiens of SOX i HIGH i are significanly posiive. These figures prove he predicion ha he effec of diminishing earnings manipulaion for firms wih abnormally low pre-managed earnings may be less han ha for firms wih abnormally high premanaged earnings afer he SOX Ac. Conclusion The purpose of he SOX Ac was o reinforce corporae governance and reduce he likelihood of missaemens. To he exen ha earnings manipulaion imposes coss on marke paricipans, his sudy predics ha he SOX Ac should aim o consrain such managemen. The empirical resuls sugges ha he pervasiveness of earnings manipulaion has significan reducions afer he SOX Ac. This sudy also finds ha firms wih high pre-managed earnings had reduced heir income-decreasing earnings managemen behavior afer he SOX Ac. In conras, we find no evidence o sugges ha he SOX Ac has consrained income-increasing manipulaion for firms wih poor pre-managed earnings. Finally, we find no evidence o show ha he SOX Ac has reduced he upward or downward earnings managemen of hose firms wih greaer pressure o presen posiive and/or increased profis. I migh imply ha firms end o manage earnings upward (downward) when he pre-managed earning is exraordinarily poor (good). Thus, afer he enforcemen of he Ac, he firms wih poor premanaged earning sill end o engage in incomeincreasing behavior while he firms wih good premanaged earnings desis from income-decreasing. 1. Barov, E., D. Givoly, C. Hayn. The Rewards o Meeing or Beaing Earnings Expecaions // Journal of Accouning and Economics, 2002. 33. pp. 173-204. 2. Barov, E., F. Gul, J. Tsui. Discreionary Accrual Models and Audi Qualificaions // Journal of Accouning and Economics, 2000. 30. pp. 421-452. 3. Belsley D.A., E. Kuh, R.E. Welsch. Regression diagnosics. New York: Wiley, 1980. 4. Burgsahler, D., I. Dichev. Earnings Managemen o Avoid Earnings Decreases and Losses // Journal of Accouning and Economics, 1997. 24. pp. 99-126. 5. Burgsahler, D., L. Hail, C. Leuz. The Imporance of Reporing Incenives: Earnings Managemen in European Privae and Public Firms // Accouning Review, 2006. 81. pp. 983-1016. 6. Chhaochharia, V., Y. Grinsein. Corporae Governance and Firm Value: The Impac of he 2002 Governance Rules // Journal of Finance, 2007. 62. pp. 1789-1825. 7. Chung, H., H.J. Sheu, J.L. Wang. Do Firms Earnings Managemen Pracices Affec Their Equiy Liquidiy? // Finance 117
Invesmen Managemen and Financial Innovaions, Volume 8, Issue 4, 2011 Research Leers, 2009. 6. pp. 152-158. 8. Cohen, D.A., A. Dey, T.Z. Lys. Real and Accrual-Based Earnings Managemen in he Pre- and Pos- Sarbanes Oxley Periods // Accouning Review, 2008. 83. pp. 757-787. 9. Dechow, P., I. Dichev. The Qualiy of Accruals and Earnings Managemen: The Role of Accruals Esimaion Errors // Accouning Review, 2002. 77. pp. 35-59. 10. Dechow, P., R. Sloan, A. Sweeney. Deecing Earnings Managemen // Accouning Review, 1995. 70. pp. 193-225. 11. Dechow, P., R. Sloan, A. Sweeney. Causes and Consequences of Earnings Manipulaion: An Analysis of Firms Subjec o Enforcemen Acions by he SEC // Conemporary Accouning Research, 1996. 13. pp. 1-36. 12. Degeorge, F., J. Pael, R. Zeckhauser. Earnings Managemen o Exceed Thresholds // Journal of Business, 1999. 72. pp. 1-33. 13. Healy, P., J. Wahlen. A Review of he Earnings Managemen Lieraure and Is Implicaions for Sandard Seing // Accouning Horizons, 1999. 13. pp. 365-383. 14. Jain, P.K., J.-C. Kim, Z. Rezaee. The Sarbanes-Oxley Ac of 2002 and Marke Liquidiy // Financial Review, 2008. 43. pp. 361-382. 15. Jones, J. Earnings Managemen during Impor Relief Periods // Journal of Accouning Research, 1991. 29. pp. 193-228. 16. Kalelkar, R., E.T. Nwaeze. Sarbanes-Oxley Ac and he Qualiy of Earnings and Accruals: Marke-Based Evidence // Journal of Accouning and Public Policy, 2011. 30. pp. 275-294 17. Lee, C.W. J., L.Y. L H. Yue. Performance, Growh and Earnings Managemen // Review of Accouning Sudies, 2006. 11. pp. 305-334. 18. Leuz, C., D. Nanda, P.D. Wysocki. Earnings Managemen and Invesor Proecion: an Inernaional Comparison // Journal of Financial Economics, 2003. 69. pp. 505-527. 19. Leuz, C., A.J. Trianis, T.Y. Wang. Why Do Firms Go Dark? Causes and Economic Consequences of Volunary SEC Deregisraions // Journal of Accouning and Economics, 2008. 45. pp. 181-208. 20. L H., M. Pincus, S.O. Rego. Marke Reacion o Evens Surrounding he Sarbanes-Oxley Ac of 2002 and Earnings Managemen // Journal of Law and Economics, 2008. 51. pp. 111-134. 21. Park, Y.W., H.H. Shin. Board Composiion and Earnings Managemen in Canada // Journal of Corporae Finance, 2004. 10. pp. 431-457. 22. Peasnell, K.V., P.F. Pope, S. Young. Board Monioring and Earnings Managemen: Do Ouside Direcors Influence Abnormal Accruals? // Journal of Business Finance & Accouning, 2005. 32. pp. 1311-1346. 23. Shaw, K.W. Corporae Disclosure Qualiy, Earnings Smoohing, and Earnings Timeliness // Journal of Business Research, 2003. 56. pp. 1043-1050. 24. Teoh, S.H., I. Welch, T.J. Wong. Earnings Managemen and he Long-Run Marke Performance of Iniial Public Offerings // Journal of Finance, 1998. 53. pp. 1935-1975. 25. Whie, H. A Heeroskedasiciy-Consisen Covariance Marix Esimaor and a Direc Tes for Heeroskedasiciy // Economerica, 1980. 48. pp. 817-838. 118