The Impact of Governance on IFRS Restatement Quality

The Impact of Governance on IFRS Restatement Qualty Authors: Arnt Verrest* Ann Gaeremynck Contact Informaton: *Contactng Author: Katholeke Unverstet Leuven Etenne Sabbelaan 53 B-8500 Kortrjk Arnt.verrest@kuleuven-kortrjk.be Frst Draft August 2008 Abstract As pror research manly focused on the nfluence of IFRS adopton on accountng qualty and ts economc effects (Leuz and Verreccha, 2000; Daske, 2006; Barth et al. 2007; Chrstensen et al. 2008), ths paper examnes the mpact of corporate governance on the qualty of the IFRS restatement process tself. The unque feature of ths study s our ablty to nvestgate accountng qualty through our focus on three features of the IFRS restatement process: (1) the dsclosure qualty of the IFRS restatement process; (2) the mmedate adopton of an IFRS standard, even though an extended adopton perod s allowed; and (3) the earnngs management actvty by restatng fnancal fgures from local GAAP to IFRS. Thereby we hypothesze that n a changng accountng envronment, better governed frms wll provde a better fnancal restatement qualty. Our results largely confrm ths predcton. It turns out that hgh qualty governance frms are more lkely to dsclose the equty, earnngs and cash flow mpact of IFRS n detal, are less lkely use the extended adopton perod of IAS 39 (but only when the effect on equty s negatve) and do not engage n earnngs management around the tme of adopton but nstead apply IFRS n a rgorous manner. Interestngly, the composton of the board and the qualty of the audt commttee n partcular determne the qualty of IFRS restatements. Key Words: Internatonal Accountng, IFRS Restatements, Corporate Governance, Dsclosure, IAS39, Earnngs Qualty JEL classfcatons : G14, G15, G30, M38, K22 We thank Laurens Cherchye, Wouter Dutlleux and Marleen Wllekens for provdng helpful comments.

1. Introducton Ths study nvestgates the mpact of corporate governance on the qualty of the IFRS restatement process. The qualty of the restatement process s assessed by three of ts components: (1) the dsclosure qualty of the restatement process, (2) the mmedate adopton of IAS39, even f an extended adopton perod s allowed, and (3) the earnngs management actvty around the tme of IFRS adopton. The adopton of IFRS was made mandatory by the European Commsson from 1 January 2005 onwards. Under IFRS 1 frms are requred to restate ther prevous year s fnancal fgures n accordance wth IFRS for reasons of comparablty and to allow nvestors to evaluate performance over tme. In practce, ths means that fnancal fgures are avalable usng local and IFRS numbers for the year 2004. Furthermore, next to dong the restatement, companes are requred to dsclose a detaled explanaton n the notes of the man adjustments made due to the IFRS transton. Fnally, wth the ntroducton of IFRS the EU allowed an extended adopton perod for IAS 39,.e. mmedate applcaton n 2004 or postponement of the applcaton untl 2005 1. Ths change n the accountng envronment from local to nternatonal standards, wth a choce for applcaton of IAS 39, offers us the possblty to evaluate the qualty of the restatement process on several dstnct aspects: the qualty of dsclosure, the mmedate applcaton or the use of the extended adopton perod and the earnngs management behavour n ths transton phase. Pror research manly focuses on the queston whether the ntroducton of IFRS led to real economc effects, by nvestgatng ts market (Leuz and Verreccha 2000; Daske 2006) as well as ts accountng consequences (Barth et al. 2007; Chrstensen et al. 2008). Barth et al. (2007) fnd sgnfcantly postve effects of IFRS wth respect to 1 The IAS39 standard deals wth the recognton and valuaton of fnancal nstruments and turned out to be one of the most debated and contested standards because of ts effect on equty value n some cases, especally for bankng and nsurance companes. Gebhardt et al. (2003) reckon that, contrary to assertons from the ndustry, hedge accountng adequately reflects the economcs of bankng actvtes. 1

the value relevance, accrual qualty and conservatsm component of earnngs. Buldng further on ths lterature, ths paper nvestgates not only whether IFRS results nto a better accountng qualty but also evaluates the accountng qualty of the restatement process tself. Ths s done by comparng hand-collected local GAAP data wth IFRS fnancal nformaton for a constant sample of 152 lsted European manufacturng and servce frms from the MSCI Pan Euro Index. The use of handcollected data n ths context enables us to provde an answer to the followng questons: (1) Is the restatement process tself used to manpulate earnngs? (2) If there s an applcaton choce for a partcular standard, do frms apply ths standard mmedately or not? (3) To what extent do frms dsclose nformaton about the restatement process? (4) What s the role of corporate governance n ths restatement process? We contrbute to IFRS lterature by nvestgatng the qualty of the restatement process tself and by searchng for potental determnants. Furthermore, an appealng feature of our study s the combnaton of a dsclosure element, an early adopton ssue of a new standard and an earnngs qualty component n one settng 2. Fnally, buldng on prevous studes nvestgatng the lnk between fnancal statement qualty and corporate governance, we focus on the role of governance n an accountng change process for a dverse spectrum of accountng qualty proxes. Our descrptve results show a sgnfcant relaton between corporate governance qualty, dsclosure qualty of IFRS restatements and the level of earnngs management n the restatement phase. In a second step we analyse each of these three components separately. We fnd sgnfcant evdence that governance qualty s postvely assocated wth dsclosure qualty. Next, we show that frms wth hgh qualty governance are more lkely to adopt IAS39 n a tmely manner, only when the mpact of the standard on equty s negatve. If not so, there s no relaton between early adopton choce and governance. Fnally, we report fndngs consstent wth the noton that earnngs management actvty s less prevalent n envronments wth hgh 2 Horton and Serafem (2007) notce sgnfcant dfferences between frms wth respect to the tmng of IFRS dsclosures. We do not consder the tmng of IFRS restatements. Tmng s not relevant n our settng snce we nvestgate the mpact of corporate governance on the qualty of the restatement process nstead of valuaton effects. 2

qualty governance. Board ndependence, board functonng and qualty of the audt commttee n partcular appear to be mportant sngle governance ndcators n explanng dfferences n restatement qualty across frms. Evdence for our hypothess s confrmed when we ntroduce the fnancal and nsurance companes n the sample. These results hold when accountng for possble endogenety ssues. Fnally, n order to assess the overall qualty of the restatement process we use the beneft-of-the-doubt method (Cherchye et al., 2008) to aggregate the three components n one score. We report an average overall qualty of the restatement process of 82.5%. The remander of the paper s structured as follows. Secton 2 revews relevant pror lterature and bulds the hypothess. Secton 3 dscusses the sample selecton. Secton 4 provdes varable defntons and Secton 5 the descrptve statstcs. Secton 6 presents regresson results. Secton 7 provdes results on extensons. Secton 8 concludes the study. 2. Lterature and motvaton Pror research on IFRS manly focuses on the queston whether the ntroducton of IFRS entals sgnfcant captal market (Leuz and Verreccha, 2000; Daske, 2006) and accountng qualty effects (Barth et al., 2007; Chrstensen et al., 2008). No unform proof s found that IFRS results nto a better accountng qualty. Bartov et al. (2005) fnd evdence that accountng numbers usng IFRS are more value relevant than those based on German GAAP. Conversely, Eccher and Healy (2003) fnd no evdence that IFRS earnngs are more value relevant than local GAAP ones, for a sample of Chnese frms owned by foregn nvestors. Also, Hung and Subramanyam (2007) detect no sgnfcant IFRS effect n Germany. The potental reason for the lack of consstent evdence s that not all frms adopt IFRS n a serous and rgorous manner, but rather consder IFRS a label and not as a commtment to provde nvestors wth hgher qualty fnancal nformaton (Daske et al., 2007). The fndng that earnngs qualty ncreases consderably more for voluntary than for mandatory adopters (Chrstensen et al., 2008) also supports ths argument. Fnally, serous 3

adopton and real economc effects are more lkely to occur n strong nsttutonal envronments (Ball et al., 2003; Burgstahler et al., 2006; Daske et al., 2008). In assessng earnngs qualty effects, Barth et al. (2007) compare earnngs management, tmely loss recognton and value relevance pre and post IFRS usng a constant sample of frms. They fnd that the varablty of net ncome and the varablty of net ncome relatve to cash flows ncrease around the tme of IFRS adopton, suggestng an mprovement n accountng qualty. Buldng on ths study and further contrbutng to IFRS lterature, ths paper nvestgates whether IFRS results nto a better accountng qualty around the tme of IFRS adopton, and not after the adopton. In other words, we evaluate the qualty of the accountng restatement process tself. We do so by comparng local GAAP fgures wth IFRS fgures for a constant sample of large European frms usng fnancal data from one and the same fscal year. As a consequence, we can produce a clean measure of the level of earnngs management when IFRS s appled for the frst tme. Furthermore, ths approach allows us to answer research questons that have not been studed yet n IFRS lterature, such as: (1) Do frms use the restatement process tself as a means to manage earnngs? (2) If an extended adopton perod for a certan standard s allowed, do frms apply ths standard early or not? (3) Is ths choce opportunstcally done? (4) How detaled do frms dsclose restatement nformaton? (6) What s the mpact of corporate governance on the qualty of ths restatement process? Next to the feature of nvestgatng the qualty of the restatement process tself, our study combnes a dsclosure ssue, an early adopton ssue of a new IFRS standard and an earnngs management ssue n one settng, whch s qute unque n our mnd 3. Specfcally, we focus on corporate governance as a potental drver of restatement qualty, next to the role of country governance. Whle the mpact of country governance has been documented n lterature (e.g. Daske et al., 2008), frm-specfc governance has been far less the subject of nvestgaton. However, corporate 3 Horton and Serafem (2007) notce sgnfcant dfferences between frms wth respect to the tmng of IFRS dsclosures. We do not consder the tmng of IFRS restatements. Tmng s not relevant n our settng snce we nvestgate the mpact of corporate governance on the qualty of the restatement process nstead of valuaton effects. 4

governance mechansms and structures are potentally mportant determnants of accountng qualty (e.g. Klen, 2002; Larcker et al., 2007). Therefore, n ths study we manly focus on the mpact of corporate governance on restatement qualty, thereby controllng for country-specfc governance measures. Corporate governance s typcally defned as a set of structures that montor a frm s operatons (Shlefer and Vshny, 1998; Larcker et al., 2007) and lmts agency problems, orgnatng from the separaton of ownershp and control wthn a frm. Agency problems between nsders and outsders can be lmted by provdng more transparent nformaton. Pror lterature has documented that effectve governance mechansms are postvely assocated wth dsclosure qualty (e.g. Karamanou and Vafeas, 2005; Marques, 2006) and earnngs qualty, although results are rather weak n some cases (Larcker et al., 2007). Klen (2002) and Chtourou et al. (2001) fnd a negatve assocaton between earnngs management and corporate governance measures such as board ndependence and functonng of the audt commttee. Fan and Wong (2002) report sgnfcant negatve relatons between earnngs qualty and a frm s degree of ownershp concentraton. DeFond and Jambalvo (1991) report that the overstatement of earnngs s less lkely among frms wth audt commttees. Evdence on the mpact of corporate governance on the early adopton of a new accountng standard s non-exstent although prevous lterature shows that early recognton of a new accountng standard s encouraged when t results nto an ncrease n earnngs (Langer and Lev, 1993 and Amr and Zv, 1997a and 1997b). In sum, we contrbute to lterature by nvestgatng the mpact of corporate governance on accountng qualty n a changng accountng envronment (.e. the mandatory adopton of IFRS), thereby consderng a broader spectrum of accountng qualty proxes. Ths results n the followng hypothess: Hypothess: Better frm-specfc governance mechansms and structures result nto an IFRS restatement process of a hgher qualty,.e. good corporate governance leads to (1) a more detaled dsclosure on the mpact of IFRS; (2) an mmedately applcaton of the standards wth an 5

extended adopton perod; and (3) lower ncentves to manage earnngs around the tme of adopton. 3. Sample desgn and data The sample we use to nvestgate our hypothess contans all companes whch are part of the MSCI Pan Euro Index over the perod January 1 st 2005 untl June 30 th 2006 4. From the ntal sample of 299 frms, 89 fnancal and nsurance frms are elmnated due to specfc ndustry characterstcs. Furthermore, 39 early IFRS adopters and 15 US GAAP are wthdrawn as no IFRS restatement nformaton s avalable for the year 2004. Fnally, 4 frms are elmnated due to lack of IFRS restatement nformaton. Ths results n a fnal sample of 152 frms shown n Table 1. The majorty of frms belong to the French (71) or the Englsh (60) nsttutonal orented envronment. To provde nformaton to dfferent stakeholders about the mpact of IFRS on accountng numbers, frms have dfferent communcaton channels. In ths study, the qualty of the IFRS restatement process s assessed usng the fnancal statements snce the latter remans an mportant publc channel for provdng nformaton to dfferent stakeholders (Chang et al., 2007). Furthermore, as the nformaton provded n the annual report s audted, ths nformaton source guarantees a hgh level of relablty (Francs and Wang, 2007). As stated n our hypothess, we assess the qualty of the restatement process by means of three dfferent elements. Whle electronc databases such as Thomson Datastream could be used to measure the earnngs management behavour around the IFRS restatement as well, nformaton on the dsclosure qualty and on the mmedate applcaton of IAS standards can only be gathered usng the hard copy verson of the 4 The MSCI Pan-Euro Index contans 287 securtes wth a free float-adjusted market captalzaton of 4,348 bllon, selected from 16 European countres. 6

fnancal statements 5. On order to be consstent n the nformaton source used for each of the three components, we collect all nformaton on the restatement process from the hard copy verson of the fnancal statements. Fnally, as IFRS became compulsory n 2005 and early applers are elmnated from the sample, the notes of fnancal statements from the year 2005 are the man source to evaluate the qualty of the restatement process 6. As only for the year 2004 local and IFRS fgures are avalable reflectng the same economc event, the 2004 data wll be used to evaluate the qualty of the IFRS restatement process. To test for corporate governance as a determnant of restatement qualty, we consder the corporate data base from Rsk Metrcs (formerly known as the Demnor database) whch contans detaled corporate governance nformaton for the MSCI Pan Euro Index 7. The ratng s based on more than 300 corporate governance crtera and the maxmum score s 40. Appendx A provdes a detaled overvew of the dfferent tems ncluded n the corporate governance score. 4. Model development To answer our research queston, we develop a regresson model n whch corporate governance (CORPGOV) s lnked to the dfferent proxes for the qualty of the IFRS restatement process (QRIFRS). The general form s the followng one: QRIFRS = α + β CORPGOV + β ControlVar ables + ε (1). 1 2 5 Although restated fgures are avalable n Datastream n most cases, they contan a serous amount of error. 6 For frms not voluntarly adoptng IFRS wth a fscal year-end not equal to 31st of December, the pre-adopton year s 2004, the transton year s 2005 and the adopton year s 2006. From now on we refer to the pre-adopton, transton and adopton year as beng 2003, 2004 and 2005 for all frms n the sample. 7 Demnor Ratngs have been used n pror studes such as Bauer et al., (2004), Van der Bauwhede and Wllekens (2008) and Renders and Gaeremynck (2008). 7

CORPGOV stands for the company corporate governance ratng from Rsk Metrcs of 2004 8. Based on prevous lterature (Klen, 2002; Karamanou and Vafeas, 2005; Larcker et al. 2007) we predct a postve coeffcent for the corporate governance varable, as governance has the purpose to decrease asymmetry n nformaton by provdng more transparent and accurate nformaton. Our dependent varable QRIFRS ether measures (1) the qualty of the dsclosure on IFRS restatements (DISCL), (2) the early adopton of IAS39 (EARLYADOP) or (3) the use of the restatement process to manage earnngs (DIFFACC). Consequently, we defne three separate regresson models. In our core analyses, we prefer testng separate regresson models above one aggregated model, as control varables dffer across our three restatement qualty measures. In senstvty analyses, we use the beneft-of-the-doubt method to aggregate these three qualty elements to aggregate them nto one overall restatement qualty score (also referred to as QRIFRS). To measure the dsclosure qualty of the restatement process, we construct a dsclosure score based on quanttatve nformaton about the IFRS restatement process n the frst regresson. The dsclosure ndex s based on 5 tems relevant for fnancal statements users: the effect on the (1) net ncome of IFRS n 2004 (2) the book value of equty at the begnnng of 2004 (= book value at the end of 2003), (3) the book value of equty at the end of 2004, (4) the total revenues or sales of 2004 and (5) the operatng cash flow of 2004 9. We choose to focus on net ncome and equty as these two measures are frequently used for valuaton purposes. Moreover, frms were mandated by IFRS 1 to provde nformaton on the restatement of these fnancal tems. Conversely, dsclosure on cash flow and sales restatements was to a large extent voluntary. However, n our opnon restatement nformaton on these 8 Results prove to be robust for usng 2004 or 2005 ratngs. Ths s not surprsng, as corporate governance tends to be rather stcky over tme. For 10 observatons the corporate governance ratng of 2005 has to be used as the 2004 ratng was not avalable. Regresson results reman qualtatvely the same wthout those 10 observatons. 9 In the dsclosure ndex equty s consdered twce as mportant as the other two tems, net ncome and operatng cash flows because prevous evdence shows that the restatement to IFRS has the largest mpact on equty. In the senstvty analyss the dsclosure ndex on three tems (sales, net ncome and operatng cash flows of 2004) s computed. Results reman the same. 8

tems s potentally mportant for nvestors and other stakeholders. Operatng cash flow s an mportant alternatve performance measure for net ncome. The sales fgure s also relevant, because t determnes the sze of a frm s market share. In ths respect, our dsclosure score conssts of mandatory (net ncome and equty fgures) and voluntary dsclosure tems (cash flow and sales fgures). To receve a 1 for one of those tems, a frm has to allocate the dfferences between the IFRS and the prevous local GAAP number to the each specfc nternatonal standard provokng (part of) ths dfference. In other words, the mere dsclosure of the dfference between the two numbers on tself (whch s requred under IFRS 1) s not suffcent to receve a 1 for that specfc tem. The dsclosure of the restatement nformaton for net ncome and book value of equty done by the Spansh company Grupo Ferroval s depcted n Appendx B as a best practce example. We defne ths score as DISCL. Based on prevous lterature, we ntroduce both country and frm control varables n the dsclosure model. Most notably, we construct a composte ndctor IPR usng 4 nsttutonal varables from LaPorta et al. (1998 and 2006) to control for countryspecfc dfferences n the nsttutonal envronment (see Appendx C for methodologcal detals on composte ndcators and a defnton of the sngle nsttutonal varables capturng nvestor protecton). In ths case, we prefer a composte ndcator above the use of sngle ndcators because t provdes us wth a more accurate nsttutonal measure (Cherchye et al., 2008). Among others, t helps us to avod multcollnearty problems between dfferent nsttutonal varables. The materalty of the IFRS restatement s the frst frm control varable. We predct a postve sgn for the absolute value of the mpact of IFRS on net ncome (ABS( NI)) and for book value of equty (ABS( BVE)), as frms experencng a large mpact are expected to dsclose more nformaton (Chang et al., 2007). We further control for frm sze (LNMV) (Raffourner, 1996; Brown et al., 1999; Holland, 2005), frm age (LNAGE) (Holland, 2005), sales growth (SLSGR), performance (ROA), amount of debt (LEVERAGE), ownershp dffuson 9

(OWNDIFF) and a US GAAP dummy varable (US GAAP). Fnally, we ntroduce ndustry dummes n the dsclosure model to control for ndustry effects (INDUSTRY). The subsequent regresson model testng dsclosure looks as follows: DISCL β LNMV 5 + β 11 = α + β CORPGOV USGAAP 1 2 c 3 ( ) 4 ( ) + β LNAGE 6 j + β IPR + β ROA + β SLSGR + β INDUSTRY 7 j 8 + ε + β ABS NI + β ABS BVE + β LEVERAGE 9 + + β OWNDIFF 10 (2). The second aspect of IFRS restatement qualty under consderaton s the choce of early applcaton of IAS39 n 2004. Ths qualty aspect s defned as EARLYADOP. If a frm mmedately apples the IAS 39 standard n 2004, the EARLYADOP gets a value of 1. A value of 0 s assgned f the frm prefers the extended adopton perod and only apples the standard for the frst tme n 2005 10. In ths case, our man varable of nterest remans CORPGOV. Further, we ntroduce a varable that measures the sgned mpact of IAS39 on equty (IMP(IAS39)). We expect a postve sgn because the hgher IMP(IAS39), the more lkely IAS39 wll be appled mmedately. Consstent wth the prevously developed early adopton models (Langer and Lev, 1993; Amr and Zv, 1997a and 1997b), and the dsclosure model, the total IFRS effects on net ncome and on equty are ntroduced n the model, together wth frm-specfc control varables. Subsequently, we test the followng regresson model: EARLYADOP= α + β CORPGOV β USGAAP+ ε 1 + β2 ( 39) β LNMV + β LNAGE + β ROA + β SLSGR + β LEVERAGE + β OWNDIFF+ 6 12 7 8 9 IMP IAS 10 + β 3IPR + β 4 NI + + β 5 BVE c 11 (3). Fnally, the thrd component of the restatement qualty reflects whether the restatement process s used to manage earnngs. As opposed to Barth et al. (2007) who compare the accruals n the year of adopton usng IFRS wth accruals n the pre 10 Addtonally, as a robustness check, we provde frms that experence a negatve mpact on equty value of IAS39 and early adoptng the standard wth a 2, as these frms are clearly havng an ncentve not to adopt the standard early. Results prove to be robust. 10

adopton year usng local standards, we employ hand-collected data from one and the same fscal year (the transton year 2004). Measurng accruals under IFRS and under local GAAP for 2004 provdes us wth a cleaner estmate of earnngs management behavour around the tme of adopton 11. Followng Leuz et al. (2003), Francs and Wang (2007) and Daske et al. (2007), we consder both the sgned and unsgned dfference n magntude of accruals calculated under local GAAP ( ACC LocalGaap(0) / TA LocalGaap(0) ) and calculated under IFRS ( ACC IFRS(0) / TA IFRS(0) ) to assess earnngs management behavour 12. To calculate the sgned dfference n accrual magntude, we subtract local GAAP accruals from IFRS accruals. A hgher level for ths dfference s consdered as an ndcaton of more earnngs management when frms restate. When consderng the unsgned dfference, we nterpret larger dfference values as evdence of a more rgorous applcaton of IFRS for the frst tme. Frms showng practcally no dfference between ther accruals under IFRS and local standards are suspected to use the dscreton avalable under IFRS. We defne ths thrd qualty measure of the restatement process as DIFFACC (sgnfyng the sgned dfference) and DIFFACC (sgnfyng the unsgned dfference). Two mportant remarks have to be made wth respect to the DIFFACC measures. Frst, accruals under IFRS are based on net ncome fgures excludng any nfluence from IAS39 because the early applcaton nfluences the amount of accruals. Second, accruals are calculated as net ncome mnus net operatng cash flows and scaled by the frm s total assets as done n Hrbar and Collns (2002). Smlar to our prevous models, the composte ndcator IPR s ntroduced as an nsttutonal measure. Control varables are also smlar to the dsclosure and the early adopton model except for the varable ACC(PREADOP) ndcatng the level of 11 When comparng accruals of dfferent years wth each other, the amount of accruals can be dfferent not only because of a change n the accountng standards but also because of economc transactons. By nvestgatng dfferences n accruals of the same year, the economc base s the same and the dfference n accruals can be more cleanly attrbuted to the change n accountng standards from local GAAP to IFRS. 12 We choose to scale by total assets from the current year, because these are avalable under IFRS. Scalng both accrual measures wth lagged total assets (under local standards) does not alter results. As a robustness check, we redo all analyses scalng by absolute value of operatng cash flows. Agan, results reman qualtatvely the same. 11

earnngs management n the pre-adopton year. Our thrd regresson model looks as follows: DIFFACC = α + β CORPGOV + β IPR + β ACC PREADOP + β LNMV + 1 2 c 3 ( ) β ROA + β SLSGR + β OWNDIFF + β USGAAP+ ε 5 6 7 8 4 (4). In ths respect, we acknowledge that the dfference between local GAAP and IFRS accruals s also nfluenced by the dfference between the local GAAP and the IFRS accountng regme, and not only by how serous frms apply IFRS. Therefore, n addtonal tests we use a two-stage regresson to deal wth ths ssue. In a frst stage DIFFACC and DIFFACC are regressed on the country-level varable DISTANCE, measured as the average rank of the number of nconsstences and absences that exst between the local standards and IFRS, based on a comparson study of bg 4 audt frms (GAAP 2000 and 2001; Wulandar and Rahman, 2004). The larger a country s DISTANCE, the larger we expect the dfference to be between local GAAP accruals and IFRS accruals. In the second step, the resdual of the frst stage s consdered as a proxy for the level of earnngs management around the tme of IFRS adopton. Results prove to be consstent wth our man analyss. To summarze, Table 2 provdes a lst of all relevant varables and ther defnton used n the dfferent models. 5. Descrptve statstcs Panel A of Table 3 detals frm-specfc descrptve statstcs of the restatement qualty ndcators, the corporate governance ratng and the control varables. Descrptve statstcs show the exstence of large dfferences across the sample for all three restatement qualty tems. Whle the average frm scores 3.82 on 5 for the DISCL varable, only 44% mmedately apples IAS 39 n 2004. Fnally, as expected, the average and medan DIFFACC s negatve (resp. -0.0075 and -0.0050), ndcatng lower accruals under IFRS. In absolute terms, ths dfference ( DIFFACC ) amounts to 1.73% on average. The mean CORPGOV score equals 25.47, on a maxmum of 12

40. Also, the standard devaton (5.6) and quartle range (11.25) of CORPGOV are qute large ndcatng a consderable varaton n the sample n terms of overall governance qualty, despte these beng all large and mature frms. Further, Panel A of Table 3 reveals that NI (the sgned mpact of IFRS adopton on net ncome) s postve wth an average 1.28% of market value. ABS( NI) s on average 1.62% of a frm s market value 13. BVE s sgnfcantly negatve wth a value of 1.93%. Its magntude (ABS( BVE)) s qute large wth an average of more than 5% of market value. For a quarter of our sample, ABS( BVE) s 7.37% or above, n terms of market value. The medan mpact of IAS39 on book value of equty (IMP(IAS39)) s slghtly postve, but not sgnfcantly dfferent from zero (0.13% of market value). ABSIMP(IAS39) reaches an average of 1.4% of market value. As an addtonal measure, we calculate the mpact of IAS39 n 2004 on a frm s equty/debt rato, as ths standard also nfluences labltes. Contrary to the average mpact on equty, the mean equty/debt mpact s negatve as the rato decreases by an average of 0.83% (not reported). Further, the average frm n our sample s 62 years old (AGE), has a market value above 2 bllon USD (MV), a return on assets of 5.67% (ROA), a sales growth of almost 9.3% (SLSGR) and a leverage rato of 66.68% (LEVERAGE). The average free float s 65.65% (OWNDIFF). Only 4% of the companes prevously reported under US GAAP. Panel B of Table 3 provdes medan values on country level of the restatement qualty, governance ndcators and restatement magntudes. We select country type followng La Porta et al. (1998) and dstngush between law systems accordng to ther Englsh, French, German or Scandnavan orgn. DISCL s lowest n German countres (2.33 on 5) and hghest n the Englsh envronment (4.00). Indvdually, France obtans the hghest score (4.39), followed by Ireland (4.33) and the Netherlands (4.18), whereas South-European companes dsclose the lowest amount of restatement nformaton. Only 25% of the Englsh law frms apply IAS39 early, 13 We prefer to scale all IFRS effects wth market value of the frm as ths s the only true scale measure of a frm (Easton and Sommers, 2003), nstead of measurng an ndex (e.g. Gordon and Weetman, 2006) whch s subject to unwanted scale effects. 13

whle all German frms do so. More nterestngly, we dscover large dfferences between countres n terms of DIFFACC. In German countres, accruals are slghtly larger under IFRS than local standards resultng n a postve DIFFACC (0.009), whle for the UK and Ireland the medan dfference s negatve (-0.013). Regardng French and Scandnavan law countres, the sgn of DIFFACC vares across countres but s negatve on average (-0.004 n both cases). The unsgned accrual dfference s largest for the UK and Ireland once agan, ndcatng that these countres have made the largest restatements. The followng four columns of Panel B depct dfferences between countres for NI, BVE, ABS( NI) and ABS( BVE). Interestngly, n all countres except Belgum, Greece and Portugal, IFRS adopton leads on average to hgher earnngs. Conversely, IFRS brngs along a lower book value of equty n most countres, as BVE s on average -0.54%. Portuguese frms experence the largest negatve effect (-19.35%). In absolute terms, Portuguese and Norwegan frms exhbt the largest ABS( NI) wth values of 3.03% and 1.34%, respectvely. For most countres, BVE s larger than NI. Overall, we conclude that the mpact of IFRS on earnngs and equty s consderable n magntude and vares sgnfcantly across countres. We also report corporate and country governance qualty fgures n Panel B of Table 3. Governance s of a hgher qualty n a common law envronment than n a code law one. Dutch (24.52), French (21.48) and Scandnavan (23.44) companes also gve proof of hgh CORPGOV values. Swss (14.81), Portuguese (16.53), Greek (18.34) and German (18.51) frms generally engage n low qualty corporate governance mechansms. Investor protecton (IPR) s measured by a composte ndcator (see Appendx 3 for detals). The UK, France and the Netherlands offer the hghest nvestor protecton (100%), whle Germany, Belgum and Greece score lowest on IPR (below 70%). Panel C of Table 3 shows that detaled equty and earnngs effects of IFRS are dsclosed by most frms, whle only 64.5% and 35.5% of the frms dsclose the effect on total revenues and operatng cash flows, respectvely. A thrd of the sample frms provde detals on all fve dsclosure tems (47 out of 152), whle 12.5% of the frms 14

only provde detals on two tems or less (19 out of 152). Three frms do not provde restatement detals at all n ther annual report. Panel D of Table 3 reports how frms dffer n ther choce to early adopt IAS39, dependng on the mpact the standard has on ther book value of equty (left part) and on ther equty/debt rato (rght part). Overall, results ndcate that the choce of applyng IAS39 early s to a large extent drven by the standard s effect on equty and labltes. In other words, ths result s consstent wth the noton of an opportunstc choce. Of all frms for whch IMP(IAS39) s smaller than zero, a majorty of 71% delays the adopton. Ths rate s only 43% when the mpact s postve. Ths dfference proves to be hghly sgnfcant (p=0.001). Our results are qualtatvely the same when consderng the equty/debt rato (p=0.000). Panel E of Table 3 compares accrual magntude and ts assocaton wth cash flows between local and nternatonal standards. These metrcs have been used n pror studes (e.g. Barth et al., 2007 and Leuz et al., 2003). Recall that we compare data from one and the same fscal year. Frstly, mean and medan accruals are smaller n magntude under IFRS than under local GAAP (one-sded p-values are 0.038 and 0.027 respectvely), provdng evdence for the noton that earnngs dscreton s lower under IFRS. Secondly, the Pearson and rank correlatons between sgned accruals and cash flows are more negatve under local GAAP than under IFRS ndcatng less smoothng actvty; although dfferences are not sgnfcant (p-values are 0.13 and 0.19). Table 4 shows Pearson and rank correlatons between the four restatement qualty components, our governance measures and all relevant control varables. Already, these unvarate tests provde evdence consstent wth our hypothess, as CORPGOV correlates postvely wth DISCL (0.376) and DIFFACC (0.165). The correlaton between CORPGOV and DIFFACC s sgnfcantly negatve, as predcted (-0.314). Fnally, EARLYADOP does not correlate postvely wth CORPGOV as hypotheszed. 15

Table 5 provdes further unvarate support for the relatonshp between corporate governance and the qualty of the IFRS restatement process. In ths case, we splt the sample nto four equally-szed groups dependng on the frms governance ratngs. Class 0 contans the quarter of the sample wth the worst governance ratngs (mean CORPGOV s 17.9), whle we classfy companes wth the hghest governance qualty n class 3 (mean CORPGOV s 32.5). Results shown n Table 5 confrm our hypothess, as frms n class 3 provde the most detaled fnancal restatements (DISCL s 4 or 5 on 5 n 92% of the cases) and exhbt the largest declnes n total accruals (medan DIFFACC equals -1.61% of total assets). Contrary to our predcton, frms n class 3 adopt IAS39 early n only 26% of the cases, whereas ths percentage s 58% for class 0 frms. Most Brtsh frms do not adopt the fnancal standard n a tmely manner. For class 0 frms, the medan DIFFACC s not sgnfcantly dfferent from zero (-0.06% of total assets) and the medan DIFFACC s lower as compared to better governance frms (0.8%). In classes 1 and 2, the dfference s sgnfcantly negatve (resp. -0.45% and -0.84%), but not of the same magntude as class 3. Dfferences between all classes are sgnfcant n most cases. These results support the noton that n frms wth low qualty governance mechansms managers use the dscreton provded under IFRS to manpulate fnancal statements. Mean fgures confrm ths result as can be derved from the second part of Table 5. 6. Regresson results Regresson results for the dsclosure model (model 2) are depcted n Table 6. In the control model (specfcaton 1), the R² reaches 47.4%. In the second specfcaton, CORPGOV, our man varable of nterest, s added and found to be postvely sgnfcant (p=0.014). We note an ncrease n adjusted R² value of 3.5%. Thereby, we also control for country governance qualty (IPR), whch also has a sgnfcantly postve coeffcent (p=0.000). As for the other control varables, LNMV (p=0.040) and several ndustry dummes appear to be sgnfcant. Dfferent from what we expected, the materalty of the IFRS restatement captured by ABS( NI) and 16

ABS( BVE) does not nfluence the qualty of the dsclosure on the restatement process. It does not come as a surprse that several control varables are nsgnfcant n explanng ths dsclosure varable. We do not consder all nformaton a frm s dsclosng, but only that fracton of the annual report dealng wth the IFRS transton process. In the thrd specfcaton, a two stage least square model (2-SLS) s used to control for possble endogenety problems found n prevous corporate governance studes (Bhagat and Black, 2002; Durnev and Km, 2005; Black et al., 2006). In a reduced form equaton (5), we choose country-level characterstcs (captured by country dummy varables) as nstruments, smlar to Klapper and love (2004) 14. Afterwards, n the second stage, CORPGOV s replaced by ts predcted value from the frst stage represented by equaton (5). As a consequence, IPR s left out of the model, because country effects are already ntroduced as nstruments n the frst stage. CORPGOV carres, consstent wth our expectatons, a sgnfcantly postve coeffcent (p=0.002). The results are robust for addng other varables, such as market to book value (as an alternatve performance or growth measure), captal ntensty, ntangble assets, nternatonalty (measured as the percentage of foregn sales on total sales) and a dummy varable for early adopters. Audtor type s not consdered snce all sample frms have BIG 4 audtors. 14 The reduced form equaton from the frst stage looks as follows: CORPGOV = α + β COUNTRYDUMY + ε (5). c A good nstrument s typcally strongly correlated wth the endogenous varable, but not wth the error term of the structural equaton. In practce, t s mpossble to fnd a perfect nstrument: ether the nstruments are not exogenous (sem-endogenous nstruments) or they have a low correlaton wth the endogenous varable (weak nstruments). We prefer country dummes above alternatves, as they are exogenous and prove to be a strong nstrument (as the R² s about 0.65 n the frst stage). In senstvty analyses, we employ legal orgn (as n Durnev and Km, 2005) and our composte nvestor protecton ndcator IPR as alternatve nstruments. IPR s a strong nstrument but not an exogenous one, whle legal orgn s an exogenous but weak nstrument. As another senstvty check, we nclude the exogenous control varables from the second stage (n ths case ndustry dummes and frm age). c 17

Results for the early adopton model are depcted n Table 7. In the control model (frst specfcaton), we notce that the sgned mpact of IAS39 on equty, IMP(IAS39), s a sgnfcant drver of the early adopton choce (p=0.012). Ths fndng s consstent wth results found n Panel D of Table 3 and robust for the use of the sgned mpact of IAS39 on the equty-debt rato. In the second specfcaton, CORPGOV has a margnally sgnfcant negatve coeffcent (p=0.086). At frst sght, ths fndng looks counterntutve as better governance frms are a pror expected to early adopt the standard. Lookng closer, ths result s due to the large number of Brtsh frms present n our sample and reluctant to apply IAS39 early. In the thrd specfcaton, IMP(IAS39) s dropped and replaced by a dummy varable (D_IMP(IAS39)) that equals 1 f the frm experences a negatve mpact on ts book value of equty from applyng IAS39, and zero f the equty effect s postve. An nteracton varable between CORPGOV and D_IMP(IAS39) measures how governance ncrementally nfluences the choce of early adopton for frms havng a negatve effect. We expect the coeffcent on the nteracton varable to be postve as we hypothesze that hgh governance qualty frms wll be more wllng to apply the standard early, even f these frms have an ncentve not to, as opposed to frms wth worse governance practces. In addton, the unsgned IAS39 mpact s ncluded as a control varable n specfcatons 3 and 4 (ABSIMP(IAS39)). Interestngly, n the thrd specfcaton the dummy ntercept D_IMP(IAS39) has the expected negatve sgn (p=0.001). We detect a sgnfcantly postve coeffcent on the nteracton varable, ndcatng a postve ncremental effect of corporate governance on the chance of early adopton, for frms havng a negatve mpact (p=0.008). The overall mpact of corporate governance on the chance of early applcaton for frms wth a negatve effect s postve, as the sum of β 1 and β 3 s +0.43 (-0.162+0.205). For frms that have a postve IAS39 effect, corporate governance appears to play a negatve role, as β 1 s negatve (p=0.006). The negatve coeffcent on ABS( BVE) shows that the larger the restatement due to other standards s, the more reluctant frms are to apply IAS39 n 2004 (p=0.021). In the fourth specfcaton, we estmate a 2-SLS model. Consstent wth prevous results 18

and predctons, the mpact dummy remans negatve (p=0.002) and the nteracton varable contnues to show a sgnfcantly postve coeffcent (p=0.013). Overall, regresson results shown n Table 7 llustrate the postve mpact of the qualty of corporate governance n the choce to early apply IAS39 when ts mpact on equty value s negatve. Fnally, n Table 8 the results for the earnngs management model (4) are shown. In Panel A (Panel B), we report regresson results wth DIFFACC ( DIFFACC ) as dependent varable. In each test we drop extreme values of DIFFACC (1% on each sde), because they may have unwanted effects on estmated coeffcents. Results do not change when leavng these extremes n the sample. Recall that we nterpret hgher (lower) values of DIFFACC ( DIFFACC ) as more earnngs management and a less accurate or rgorous frst tme applcaton of IFRS. In the frst specfcaton of Panel A, the coeffcent on IPR s sgnfcantly negatve (p=0.010). Control varables are not sgnfcant n most cases. In the second specfcaton the coeffcent on CORPGOV s negatve as predcted (p=0.002). Ths result confrms the noton that better corporate governance leads to a more rgorous applcaton of IFRS. The adjusted R² value rses from 6.4% to 12.2% when ncludng CORPGOV n the model. Enterng the predcted governance score from the reduced equaton (2-SLS), confrms the result as the coeffcent on CORPGOV carres a negatve and hghly sgnfcant coeffcent (p=0.000). The equvalent analyses shown n Panel B convey smlar results. CORPGOV carres as predcted a sgnfcantly postve sgn, even when controllng for country-specfc governance qualty (p=0.035). Ths result confrms fndngs from Table 5 and holds when executng a 2-SLS regresson (p=0.001). Overall, we conclude that results n Table 8 support the noton that better corporate governance mechansms lead to a more rgorous applcaton of IFRS, provdng managers wth less ncentves to manpulate earnngs durng the restatement. However, we acknowledge that these regressons are not wthout ts shortcomngs. Foremost, we are unable to refer to any pror lterature, and therefore we nterpret regresson results of Table 8 wth a degree 19

of cauton 15. Further, the use of accrual measures for earnngs qualty has always been up for debate n accountng lterature (e.g. Guay et al., 1996). 7. Addtonal analyses A drawback of prevous analyses s each of the three elements of fnancal restatement qualty s estmated separately. Therefore, t s dffcult to obtan an overall measure of a frm s restatement qualty. In ths frst of four addtonal analyses, we test a model that explans varaton n overall restatement qualty, assessed by means of composte ndcators. An appealng feature of the use of composte ndcators s that restatement qualty s measured by means of one sngle fgure capturng nformaton from multple relevant characterstcs. We apply the beneft-of-the-doubt methodology to construct the ndcators, usng flexble weghtng (see also Cherchye et al, 2008). In Appendx C we provde detals on ths procedure. The outcome of the analyss s QRIFRS, a varable between 0 and 1, where hgher values sgnfy a better overall restatement qualty. The average frm attans an overall restatement qualty of 82.5%, ndcatng that the average frm s 17.5% removed from the best restatement qualty frms. The medan overall restatement qualty s 83%, the mnmum 43% and the maxmum 100%. We test a tobt specfcaton as the composte ndcator vares between 0 and 1. We execute the regresson wth CORPGOV as our man varable of nterest, n addton to the common control varables used n the three prevous analyses. In the control model shown n Table 9, IPR (p=0.002), LNMV (p=0.003), USGAAP (p=0.099) and OWNDIFF (p=0.095) are sgnfcant varables, all carryng the predcted sgns. The second regresson and the 2-SLS regresson confrm our hypothess: CORPGOV carres a sgnfcantly 15 In explanng dfferences n abnormal accruals, Geger and North (2006) control for changes n earnngs, cash flows, and other frm control varables. All studes nvestgatng change n accruals measure these changes over two years. However, n our settng, we measure dfferences n accruals from the same fscal year. 20

postve coeffcent (p=0.020 and p=0.009). Also, the ntroducton of CORPGOV leads to an ncrease n R² of almost 3%. All of the prevous analyses were done wth nonfnancal frms, as t s dffcult to measure earnngs qualty n a comparable way for fnancal, nsurance and nonfnancal companes. However, as fnancals represent 28% of the ntal sample, the dsclosure and the early adopton model are rerun for the total sample of 223 frms (18 from the 89 fnancal and nsurance frms dsappear due to the lack of restatement data). We even expect more sgnfcant results as fnancal frms usually experence a larger mpact from IAS39. Control varables dffer somewhat because certan varables are not comparable between fnancal and nonfnancal frms. We replace SLSGR and ROA by respectvely market-to-book value (MTBV) and a loss dummy (LOSS). LEVERAGE s dropped 16. Panel A of Table 10 reports the results of the dsclosure model for the total sample. It should be noted that n ths case DISCL s a frm-specfc dsclosure score on 4 nstead of 5. Dsclosure of the IFRS mpact on total revenues s not consdered n ths case because these are less relevant for fnancal frms and less comparable wth nonfnancal frms sales numbers. As expected, the sgnfcance of the CORPGOV further ncreases wth or wthout control for endogenety. Panel B of Table 10 shows the EARLYADOP results for the full sample. Agan, we fnd evdence strokng wth our research hypothess. For frms possessng hgher qualty governance mechansms, dsclosure on IFRS restatements appears sgnfcantly hgher and the early adopton of IAS39 s more lkely, when the mpact on equty s negatve. Moreover, results prove to be stronger than those prevously reported n Tables 6 and 7. Fnally, results are confrmed by the two stage approach accountng for a possble endogenety problem. As a thrd extenson, we nvestgate how the four categores of corporate governance qualty each explan varaton n restatement qualty. The governance ratngs contan 16 Leavng leverage and return on assets n the equaton however, does not change the results. 21

crtera from 4 categores of governance mechansms: (1) shareholders rghts and dutes, (2) range of takeover defences, (3) board structure and functonng and (4) dsclosure on corporate governance (for detals see Appendx A). A pror, we predct a sgnfcant mpact from the shareholders rghts ratng and from the board structure and functonng ratng (Chtourou et al., 2000 and Bhagat and Black, 2002). Furthermore, we predct board ndependence and the qualty and functonng of the audt commttee to be the most mportant sngle drvers of these two corporate governance ndcators (Klen et al., 2002; Dechow et al., 1996; DeFond and Jambalvo, 1991) 17. Our results (not shown n tables) convey that n the second stage, n whch the precedng frst stage s stll equal to equaton (5), board functonng s hghly sgnfcant n explanng the level of DISCL. In determnng EARLYADOP, board characterstcs turn out to be the most explcatve varable of all four ratngs, as ts nteracton coeffcent wth D_IMP(IAS39) s postve and sgnfcant n the second stage. Further, nvestgatng specfc governance ndcators reveals that the proporton of ndependent drectors has the hghest mpact on DISCL and DIFFACC. In comparson wth board structure ndcators, board functonng measures are less sgnfcant n explanng DISCL, but hghly sgnfcant n determnng EARLYADOP. The separaton of CEO and charman of the board does not appear to be sgnfcant. Consstent wth our predctons, we fnd a sgnfcantly postve (respectvely negatve) assocaton between the qualty of the audt commttee and DISCL (respectvely DIFFACC). The sze, workngs and composton (.e. ndependence) of the audt commttee all turn out to be the most relevant sngle corporate governance determnants of IFRS restatement qualty. In a fnal extenson, we nvestgate the dsclosure of IAS39 effects for the 85 late adopters. These frms are requred to report the mpact of the fnancal standard on ther equty value at the begnnng of 2005. As t turns out, these frms dsclose ths nformaton n a varyng degree of accuracy, once agan. We nvestgate the 17 We are restrcted to use 2005 data for these specfc governance ndcators, as detaled governance data of 2004 are not avalable. 22

dsclosure qualty of ths specfc tem and to whch extent governance plays a role n ths partcular dsclosure process. We make a dstncton between four dsclosure levels of IAS39 effects. Frms reportng the mpact on each balance sheet tem and makng the mpact mmedately vsble on ther key fnancal statements (.e. on the balance sheet, not n the notes of the 2005 annual report) represent the hghest dsclosure class. Frms reportng the mpact of IAS39 only on book value of equty, but dsclosng the mpact on the balance sheet construct the second class. Frms reportng on IAS39 effects only n ther notes are placed n the thrd class. The last class contans frms that report nothng about the mpact of the fnancal standard. The number of frms n each class s 14, 42, 23 and 6, gong from the hghest to the lowest dsclosure group. Consstent wth predctons, our results (not shown n tables) convey that governance qualty plays a sgnfcant role n determnng dsclosure qualty on the mpact of IAS39. 8. Concluson Ths study documents on the propertes and determnants of the qualty of the IFRS restatement process. Whle prevous lterature manly focused on the accountng qualty and economc consequences of the IFRS, ths paper studes the qualty of the IFRS restatement process tself. Ths process provdes us wth a unque settng to further nvestgate IFRS adopton snce two sets of fnancal statements under local GAAP and IFRS are avalable for one and the same fscal year. Around the adopton tmng, frms have to make choces about the level of dsclosure on the restatement process, the early adopton of IAS39 or not, and the extent to whch they use the restatement process as an opportunty to manage earnngs. In partcular, we nvestgate to what extent corporate governance nfluences the outcome of the restatement process. Usng a hand-collected dataset of large lsted European frms, our descrptve analyses show that dsclosure qualty of the restatement process s qute hgh, the choce to apply IAS39 early or extendedly s done opportunstcally and there s a sgnfcant amount of earnngs management around the tme of the adopton. 23

Regresson results further confrm our hypothess that well governed frms () dsclose more nformaton on restatement process, () choose to adopt IAS39 early, but only when experencng a negatve mpact from ths standard, and () engage less n earnngs management around the tme of adopton and apply IFRS more serously. Furthermore, results hold when fnancal companes are added to the sample. Fnally, board functonng, board ndependence and qualty of the audt commttee turn out to be sgnfcant drvers of the qualty of the restatement process. Our results are potentally nterestng for polcy makers. If new IFRS standards are ntroduced leavng frms a choce to apply them or not, standard setters may expect frm to choose opportunstcally, especally those lackng governance qualty. Our study also shows that corporate governance can fulfl the role of an enforcement mechansm n a reportng change process. We acknowledge that ths study s subject to a number of caveats. Frst, as we do not nvestgate economc consequences of IFRS restatements, we are constraned n our conclusons about the accountng qualty effects of IFRS. Another lmtaton s the external valdty of our fndngs, as only large lsted companes are ncluded n the sample. However, f the results hold for large companes, whch are more lkely to nstall an optmal level of governance, t can be argued that governance wll have an mpact on the qualty of the restatement process n smaller companes as well. 24

References Amr E. and A. Zv (1997a) Economc consequence of alternatve adopton rules for new accountng standards Contempory Accountng Research, Vol 14, pp. 543-68. Amr E. and A. Zv (1997b) Recognton, dsclosure or Delay: Tmng the adopton of SFAS NR 106, Journal of Accountng Research, Vol. 35, pp. 61-81. Archambault, J.J., and Archambault, M.E., (2003) A multnatonal test of determnants of corporate dsclosure, The Internatonal Journal of Accountng, 38, 173-194. Ashbaugh, H., and M. Pncus (2001), Domestc Accountng Standards, Internatonal Accountng Standards, and the Predctablty of Earnngs, Journal of Accountng Research 39, 417 434. Ball R., S.P. Kothar and A. Robn (2000), The Effect of Internatonal Factors on Propertes of Accountng Earnngs, Journal of Accountng and Economcs 29, 1-52. Ball, R., Sadka, G. and Robn, A. (2006) "Is Fnancal Reportng Shaped by Equty Markets or by Debt Markets? An Internatonal Study of Tmelness and Conservatsm", Workng Paper. Barth, M.E., W.R. Landsman and M.H. Lang (2007), Internatonal Accountng Standards and Accountng Qualty, Workng Paper, Stanford Unversty and Unversty of North Carolna. Bartov, E., S.R. Goldberg, and M. Km (2005) Comparatve value relevance among German, U.S., and Internatonal Accountng Standards: A German stock market perspectve, Journal of Accountng, Audtng and Fnance 20(2): 95-119. Bauer, R., Günster, N., Otten, R. (2004) Emprcal evdence on corporate governance n Europe: The effect on stock returns, frm value, and performance, Journal of Asset Management 5, 91-104. Bhagat, S. and B. Black (2002) The non-correlaton between board ndependence and long-term frm performance, Journal of Corporate Law 27, 231-272. Burgstahler, D.C., L. Hal, and C. Leuz (2006), The Importance of Reportng Incentves: Earnngs Management n European Prvate and Publc Frms, The Accountng Revew 81, 983 1017. Cherchye, L., A. Gaeremynck and A. Verrest (2008), "Insttutonal Characterstcs and Frm Proftablty", Workng paper, Katholeke Unverstet Leuven. 25

Chrstensen, Hans Bonde Lee, Edward and Walker M. (2007), "Do IFRS/UK-GAAP Reconclatons Convey New Informaton?" Internatonal Journal of Accountng Chtourou, M., Bedard, J. and Courteau, L. (2001), "Corporate Governance and Earnngs Management", workng paper. Daske, H. (2006) Economc Benefts of Adoptng IFRS or US-GAAP Have the Expected Costs of Equty Captal really decreased? Journal of Busness Fnance & Accountng 33, 329 373. Daske, H., Hal, L., Leuz, C. and Verd, R. "Adoptng a Label: Heterogenety n the Economc Consequences of IFRS Adoptons" Workng paper. Daske, H., Hal, L., Leuz, C. and Verd, R. "Adoptng a Label: Heterogenety n the Economc Consequences of IFRS Adoptons" Workng paper. Dechow, P., and I. Dchev. 2002. The qualty of accruals and earnngs: The role of accrual estmaton error. The Accountng Revew 77: 35-59. DeFond, and C. W. Park. 2001. The reversal of abnormal accruals and the market valuaton of earnngs surprses, The Accountng Revew 76 (July): 375-404. Damond, D. and Verreccha, R., 1991, Dsclosure, Lqudty and the cost of captal, Journal of Fnance 46, 1325-1360. Fan, Joseph P. H., and T. J. Wong. 2002. Corporate ownershp structure and the nformatveness of accountng earnngs n East Asa. Journal of Accountng and Economcs 33 (August): 401 26. Francs J.R., and D. Wang. 2007. The jont effect of nvestor protecton and bg 4 audts on earnngs qualty around the world. Contemporary Accountng Research (forthcomng). GAAP 2000: A survey of natonal Accountng rules n 53 countres. Arthur Andersen, BDO, Delotte Touche Tohmatsu, Ernst and Young Internatonal, Grant Thornton, KPMG, and PrcewaterhouseCoopers; edtor: Nobes, C.W. GAAP 2001: A survey of natonal Accountng rules n 53 countres. Arthur Andersen, BDO, Delotte Touche Tohmatsu, Ernst and Young Internatonal, Grant Thornton, KPMG, and PrcewaterhouseCoopers; edtor: Nobes, C.W. Geger, M.A., and D.S. North (2006). Does hrng a new CFO change thngs? An nvestgaton of changes n dscretonary accruals. The Accountng Revew 81. Gompers, P., Ish, J., Metrck, A., 2001. Corporate governance and equty prces. Quarterly Journal of Economcs 116, 107-155. 26

Hal, L. and C. Leuz (2006). Internatonal Dfferences n Cost of Equty Captal: Do Legal Insttutons and Securtes Regulaton Matter? Journal of Accountng Research 44, 485-531. Healy, P, and Palepu, K (2001). "Informaton asymmetry, corporate dsclosure, and the captal markets: A revew of the emprcal dsclosure lterature", Journal of Accountng and Economcs 31, 405-440. Horton, Joanne and Serafem, George, "Market Reacton & Valuaton of IFRS Reconclaton Adjustments: Frst Evdence from the UK", Workng Paper. Hrbar, P., and D.C. Nchols. The use of unsgned earnngs qualty measures n tests of earnngs management. Workng Paper. Cornell Unversty, 2006. Hung, M. (2001). Accountng Standards and Value Relevance of Fnancal Statements: an Internatonal Analyss. Journal of Accountng and Economcs, 30(3), pp. 401-420. Hung, M., and K.R. Subramanyam, 2007, Fnancal Statement Effects of Adoptng Internatonal Accountng Standards: The Case of Germany, Revew of Accountng Studes, forthcomng. Jagg B. and P. Low (2000). Impact of Culture, Market Forces, and Legal System on Fnancal Dsclosures. The Internatonal Journal of Accountng, 35(4), pp. 495-519. Jensen, M.C. and Mecklng, W.H. (1976), Theory of the frm; manageral behavor, agency costs and ownershp structure, Journal of Fnancal Economcs, 4(3), pp. 305-360 Karamanou, I. and N. Vafeas (2005) The assocaton between corporate boards, audt commttees and management earnngs forecasts: an emprcal analyss, Journal of Accountng Research, Vol. 43, N. 3, pp. 453-486. Klapper, L., Love, I., 2004. Corporate governance, nvestor protecton, and performance n emergng markets. Journal of Corporate Fnance 10, 703-728. Klen, A. (2002) Audt commttee, board of drector characterstcs and earnngs managements,journal of Accountng and Economcs, Vol. 33, pp. 375-400. Langer, R. and B. Lev (1993), The FASB s Polcy of Extended Adopton for New Standards: An Examnaton of FAS Nr 87, The Accountng Revew, Vol. 68, pp. 515 33. Larcker, D., Rchardson, S., Tuna, I., 2006. How mportant s corporate governance? Workng Paper, Unversty of Pennsylvana. 27

La Porta, R., F. Lopez-De-Slanes, A. Shlefer and R. Vshny (1997). Legal Determnants of External Fnance. Journal of Fnance, 52, 1131-1150. La Porta, R., F. Lopez-De-Slanes, A. Shlefer and R. Vshny (1998). Law and Fnance. Journal of Poltcal Economy, 106(6), 1113-1154. La Porta, R., F. Lopez-de-Slanes, A Shlefer, AND R. Vshny. (2000) Investor protecton and corporate governance. Journal of Fnancal Economcs 58 (October/November 2000): 3-27. La Porta, R., F. Lopez-de-Slanes, AND A Shlefer. What works n securtes laws? Journal of Fnance 61 (February 2006): 1-32. Lambert, R.A., C. Leuz, and R.E. Verreccha, 2007b, Informaton Asymmetry, Informaton Precson, and the Cost of Captal, Workng paper, Unversty of Pennsylvana and Unversty of Chcago. Leuz, C. (2003). IAS versus US GAAP, Informaton Asymmetry-Based Evdence from Germany s New Market. Journal of Accountng Research, Vol. 41, No. 3 (June), pp. 445-472. Leuz, C., D. Nanda and P.D. Wysock (2003). Earnngs Management and Investor Protecton: An Internatonal Comparson. Journal of Fnancal Economcs, Vol. 69, pp. 505 527. Leuz, C., and R.E. Verreccha, 2000, The Economc Consequences of Increased Dsclosure, Journal of Accountng Research 38, 91 124. Marques, A. (2006) SEC nterventons and the frequency and usefulness of non- GAAP fnancal measures, Revew of Accountng Studes, forthcomng. Renders, A. and A. Gaeremynck (2008), A Cross-country Study of the Impact of Corporate Governance on Company Performance, Workng Paper. Shlefer, A., Vshny, R., 1997. A survey of corporate governance. Journal of Fnance 52, 737-783. Vander Bauwhede H and M. Wllekens (2008). Dsclosure on corporate governance n the European Unon, Corporate Governance. Van Tendeloo, B., and A. Vanstraelen, 2005, Earnngs management under German GAAP versus IFRS, European Accountng Revew 14, 155 180. Weetman, P. and P. Gordon. 2006. Communcatng sgnfcant change n accountng measurement: the transton to IFRS n the UK. Workng paper. 28

Table 1: Sample Outlne Country Type Country No Observatons Englsh 60 Ireland 3 UK 57 French 71 Belgum 1 France 36 Greece 1 Italy 8 Netherlands 11 Portugal 1 Span 13 German 6 Germany 5 Swtzerland 1 Scandnavan 15 Denmark 3 Fnland 2 Norway 2 Sweden 8 Total 152 29

Table 2: Varable Defntons Dependent var. DISCL EARLYADOP DIFFACC DIFFACC Independent var. CORPGOV IPR NI BVE ABS( NI) ABS( BVE) IMP(IAS39) D_IMP(IAS39) ABSIMP(IAS39) LNMV LNAGE ROA SLSGR LEVERAGE OWNDIFF US GAAP INDUSTRYD Defnton = Dsclosure measure equallng the number of restatement tems from a lst of four (net ncome of transton year, book value at the end of pre-adopton year, book value at end of transton year and operatng cash flow of transton year) of whch detaled IFRS restatement nformaton s provded; maxmum score s 4. An example s provded n Appendx B. = Dummy varable takng the value of 1 for frms early adoptng IAS 39. = ACC IFRS(0) / TA IFRS(0) - ACC LG(0) / TA LG(0), wth ACC IFRS(0) calculated as restated transton year net ncome (NI IFRS(0) ), before IAS 39, mnus restated net cash flow from operatng actvtes (CFO IFRS(0) ). = Absolute value of DIFFACC = The corporate governance ratng from Demnor on frm level, based on a grd of 300 crtera; maxmum score attanable s 40; see Appendx A. = Composte ndcator measurng nvestor protecton on country level. For measurement of the composte ndcators, see Appendx C for detals. = Net ncome of the transton year under IFRS mnus net ncome under local GAAP, scaled by market value. = Book value of equty of the transton year under IFRS mnus book value of equty under local GAAP, scaled by market value. = Absolute value of NI. = Absolute value of BVE. = The effect of IAS 39 on book value of equty n the transton year, scaled by market value. = Dummy varable equal to 1 f frm has negatve effect of IAS39 on book value of equty. = Absolute value of SIZE (IAS 39). = Ln(market value of equty at end of the transton year n thousands of USD) = Ln(transton year year of foundaton or ncorporaton) 18 = Net ncome on total assets 19 = Average growth n sales over last 5 years pror to adopton year = 1 (book value of equty (local GAAP) / total assets) = Percentage of shares not closely held by drectors or bg shareholders = 1 f company reported prevously under US GAAP = Set of ndustry dummy varables based on two-dgt SIC codes 18 For more than half of the frms, the foundaton year s avalable. If the date s mssng, we use the date of ncorporaton. If nether of these two are avalable, we take the year the frm was avalable n Datastream. 19 Net ncome, book value of equty and total assets, requred to calculate ROA and LEVERAGE, are local GAAP fgures for the transton year. 30

Table 3: Descrptve Statstcs Panel A shows descrptve statstcs on all test and control varables, defned n Table 2. Panel B lsts descrptve data of the four restatement qualty varables (DISCL, EARLYADOP, DIFFACC and DIFFACC ) and two governance ndcators (CORPGOV and IPR) per country. For DISCL and EARLYADOP, the mean fgures are depcted. For the accrual dfference measures and the restatement varables of net ncome and equty, the medan fgure s shown nstead of the mean to avod the nfluence of outlers. Panel C provdes data on the dsclosure qualty per tem nvestgated, together wth the number of frms per dsclosure score. Panel D shows the percentage frms applyng IAS 39 early, dependng on the mpact frms experence on ther equty value (BVE) and ther equty/debt rato and sgnfcance levels of the dfference. Panel E depcts accrual propertes of local GAAP earnngs and IFRS earnngs of the transton year. ACC (0) /TA (0) s the absolute level of accruals scaled by total assets. Accruals are measured by subtractng operatng cash flows from net earnngs. Lower values of ACC (0) /TA (0) sgnfy lower earnngs dscreton. Both mean and medan values are shown, together wth sgnfcance levels of the dfference. ρ(acc (0),CFO (0) ) s the correlaton between sgned accruals and operatng cash flows, both scaled by current year total assets. Larger negatve values of ρ(acc (0),CFO (0) ) sgnfy more earnngs smoothng. Both ordnary and rank correlatons are measured, together wth the sgnfcance of the dfference. Panel A: N Mean Medan MIN MAX Q1 Q3 STDEV DISCL 152 3,82 4,00 0,00 5,00 3,00 5,00 1,14 EARLYADOP 152 0,44 0,00 0,00 1,00 0,00 1,00 0,50 DIFFACC 152-0,0075-0,0050-0,1085 0,1877-0,0164 0,0021 0,0280 DIFFACC 152 0,0173 0,0105 0,0001 0,1877 0,0039 0,0216 0,0231 CORPGOV 152 25,47 24,99 12,99 35,96 20,50 31,75 5,60 NI 152 0,0128 0,0069-0,0565 0,1534 0,0004 0,0195 0,026 BVE 152-0,0193-0,0043-0,3891 0,1722-0,0485 0,0225 0,080 ABS( NI) 152 0,0162 0,0088 0,0000 0,1534 0,0029 0,0206 0,024 ABS( BVE) 152 0,0532 0,0331 0,0001 0,3891 0,0127 0,0737 0,063 IMP(IAS39) 152 0,0013 0,0003-0,1508 0,1299-0,0067 0,0065 0,026 ABSIMP(IAS39) 152 0,0140 0,0066 0,0000 0,1508 0,0015 0,0159 0,022 MV n '000$ 152 24064060 12013065 1432864 209475900 7949500 23003870 33853667 AGE n years 152 62 55 1 261 17 102 47 ROA 152 0,0567 0,0459-0,1172 0,3538 0,0278 0,0814 0,0596 SLSGR 152 9,2886 7,3500-23,9000 98,0700 2,3050 13,8400 13,6726 LEVERAGE 152 0,6668 0,6854 0,2173 1,2332 0,5605 0,7864 0,1648 OWNDIFF 152 65,65 64,50 6,00 100,00 48,00 86,00 22,53 US GAAP 152 0,04 0,00 0,00 1,00 0,00 0,00 0,20 31

Panel B: COUNTRY N DISCL EARLYADOP DIFFACC DIFFACC NI BVE ABS( NI) ABS( BVE) CORPGOV IPR mean mean medan medan medan medan medan medan medan country stat. ENGLISH 60 4,00 0,25-0,0134 0,0152 0,0095-0,0073 0,0114 0,0348 32,07 100,00% IRELAND 3 4,33 0,33-0,0042 0,0151 0,0029-0,0263 0,0029 0,0263 24,85 80,72% UK 57 3,98 0,25-0,0149 0,0159 0,0096-0,0030 0,0131 0,0353 32,14 100,00% FRENCH 71 3,94 0,58-0,0041 0,0125 0,0077-0,0164 0,0088 0,0344 21,31 100,00% BELGIUM 1 2,00 1,00 0,0085 0,0085-0,0063-0,0595 0,0063 0,0595 18,72 59,16% FRANCE 36 4,39 0,61-0,0044 0,0095 0,0108-0,0195 0,0125 0,0449 21,48 100,00% GREECE 1 3,00 1,00 0,0068 0,0068-0,0085 0,0240 0,0085 0,0240 18,52 62,62% ITALY 8 3,88 0,50 0,0024 0,0075 0,0030 0,0130 0,0056 0,0359 20,55 80,72% NETHERLANDS 11 4,18 0,27-0,0085 0,0085 0,0091-0,0154 0,0091 0,0248 24,52 100,00% PORTUGAL 1 4,00 1,00-0,0140 0,0140 0,0116-0,0959 0,0116 0,0959 16,51 92,20% SPAIN 13 2,77 0,69-0,0031 0,0031 0,0011-0,0094 0,0043 0,0336 19,98 85,30% GERMAN 16 2,33 1,00 0,0092 0,0124 0,0040 0,0098 0,0082 0,0373 18,29 65,06% GERMANY 5 2,80 1,00 0,0049 0,0049 0,0022 0,0050 0,0042 0,0294 19,37 50,60% SWITZERLAND 1 0,00 1,00 0,0134 0,0134 0,0110 0,0156 0,0110 0,0156 16,11 79,52% SCANDINAVIAN 15 3,07 0,33-0,0037 0,0103 0,0054 0,0109 0,0054 0,0117 23,44 71,42% DENMARK 3 2,33 0,33-0,0002 0,0129 0,0210-0,0075 0,0210 0,0639 17,50 77,05% FINLAND 2 3,50 1,00-0,0046 0,0046 0,0040 0,0158 0,0040 0,0158 23,06 85,30% NORWAY 2 3,00 0,00 0,0003 0,0093 0,0134 0,0784 0,0134 0,0784 21,96 80,00% SWEDEN 8 3,25 0,25-0,0049 0,0078 0,0048 0,0095 0,0048 0,0095 24,37 71,42% MEAN 10 3,20 0,60-0,0018 0,0106 0,0061-0,0041 0,0087 0,0374 21,83 81,11% MEDIAN 3 3,25 0,58-0,0037 0,0095 0,0054-0,0030 0,0085 0,0344 21,31 80,72% MIN 1 0,00 0,00-0,0149 0,0031-0,0085-0,0959 0,0029 0,0095 16,11 50,60% MAX 57 4,39 1,00 0,0134 0,0252 0,0210 0,0784 0,0210 0,0959 32,14 100,00% 1

Panel C: Proporton of Frms provdng Dsclosure Detals Restatement Item DISCL No Obs 0 3 1. Equty Value 2003 92,11% 1 3 2. Equty Value 2004 92,11% 2 13 3. Net Income 2004 84,21% 3 28 4. Total Revenues 2004 64,47% 4 58 5. Operatng Cash Flow 2004 35,53% 5 47 Total 152 Panel D: Neg Effect BVE p-value Pos Effect BVE Neg Eff Equty/Debt p-value Pos Eff Equty/Debt EARLY ADOP 22 (29%) 45 (57%) 22 (26%) 26 (63%) 0,001 0,000 LATE ADOP 51 (71%) 34 (43%) 64 (74%) 15 (27%) TOTAL 73 (100%) 79 (100%) 86 (100%) 41 (100%) Panel E: Accrual Propertes under IFRS and Local GAAP No Obs Local GAAP p-value (1-sded) IFRS ACC (0) / TA (0) ACC (0) / TA (0) Medan 152 0,056 0,027 0,048 Mean 148 0,067 0,038 0,059 ρ(acc (0),CFO (0) ) ρ(acc (0),CFO (0) ) Rank Corr 152-0,589 0,130-0,505 Pearson Corr 149-0,596 0,190-0,535 2

Table 4: Correlatons Ordnary (Pearson) correlatons are shown n the upper rght trangle wth correspondng sgnfcance levels. Rank (Spearman) correlatons are shown n the lower left trangle of the table. All varables are defned n Table 2. DISCL EARLY ADOP DIFF ACC DIFF ACC CORP GOV IPR ABS ( NI) DISCL 1,000-0,078-0,018 0,043 0,376*** 0,40*** 0,016 0,003 0,147* 0,021 0,131-0,078 0,187** -0,091-0,056 EARLYADOP -0,025 1,000 0,046 0,003-0,223*** -0,23*** 0,026-0,148** 0,029 0,006 0,107 0,114 0,070-0,093 0,160** DIFFACC 0,008 0,113 1,000 0,034-0,316*** -0,169** -0,152** -0,029-0,119 0,028-0,027-0,046 0,016-0,24*** 0,068 DIFFACC -0,049-0,010-0,463*** 1,000 0,165** 0,081 0,294*** 0,087 0,194** -0,145* 0,119-0,053 0,050-0,147* -0,091 CORPGOV 0,359*** -0,22*** -0,339*** 0,241*** 1,000 0,53*** 0,125-0,040 0,043-0,158** 0,135* 0,031-0,034 0,090-0,147* IPR 0,418*** -0,182** -0,278*** 0,148* 0,495*** 1,000 0,163** -0,030 0,088 0,006 0,087 0,019 0,021 0,007-0,329*** ABS( NI) -0,003-0,090-0,166** 0,297*** 0,005 0,193** 1,000 0,329*** 0,100-0,003 0,023-0,032 0,038-0,31*** -0,101 ABS( BVE) 0,020-0,101 0,051 0,038-0,039 0,071 0,394*** 1,000-0,084-0,065-0,066 0,003 0,209*** -0,22*** -0,015 LNMV 0,137* 0,076-0,109 0,071 0,039 0,086-0,017-0,073 1,000-0,034 0,046 0,073-0,151** 0,176** -0,007 LNAGE 0,069 0,029 0,132-0,143* -0,137* -0,047-0,049-0,110-0,024 1,000 0,071-0,241 0,041 0,083-0,048 OWNDIFF 0,150* 0,118-0,041 0,171** 0,163** 0,108 0,030-0,081 0,031 0,060 1,000-0,163** 0,084-0,133* 0,023 SLSGR -0,119 0,150* 0,011-0,018-0,062-0,090-0,066-0,112 0,110-0,20*** -0,111 1,000-0,123 0,116-0,029 LEVERAGE 0,154* 0,051-0,024 0,112-0,092 0,042 0,158* 0,253*** -0,137* 0,053 0,070-0,052 1,000-0,31*** 0,013 ROA -0,063-0,118-0,103 0,010 0,148* -0,018-0,332*** -0,340*** 0,269*** 0,061-0,083 0,081-0,360*** 1,000 0,014 USGAAP -0,002 0,160** 0,167** -0,122-0,162** -0,24*** -0,143* 0,024-0,053-0,029 0,022-0,024 0,011-0,007 1,000 ABS ( BVE) LNMV LNAGE OWN DIFF SLSGR LEVER AGE ROA USGAAP 3

Table 5: Earnngs Management Actvty per Corporate Governance Category Panel A shows the number of frms scorng 0, 1, 2, 3, 4 or 5 on DISCL subdvded per corporate governance class. The sgnfcance of the Pearson Ch² statstc shows how these frequences dffer across governance classes. Panel B shows the number of frms scorng 0 or 1 on EARLYADOP subdvded per corporate governance class. The sgnfcance of the Pearson Ch² statstc shows how these frequences dffer dependent on the governance class. Panel C shows unvarate results for four groups of frms, categorzed by ther corporate governance ratng: category 0 (3) contans the worst (best) governance frms. Medan fgures of DIFFACC and DIFFACC (for defnton see Table 2) are shown, together wth the sgnfcance levels (1-sded p-values) of all mutual dfferences the four categores, measured respectvely wth a Rank Sum test. Cl. 0 Cl. 1 Cl. 2 Cl. 3 (n=38) (n=38) (n=38) (n=38) Mean CORPGOV 17,9 21,43 26,81 32,5 1. DISCL (#) DISCL= 0 3 0 0 0 DISCL= 1 1 1 1 0 DISCL= 2 6 4 3 0 DISCL= 3 11 9 5 3 DISCL= 4 11 14 13 20 DISCL= 5 6 10 16 15 SUM 38 38 38 38 Ch-square = 30,10 p-value = 0,012 2. EARLYADOP (#) EARLYADOP= 0 16 19 22 28 EARLYADOP= 1 22 19 16 10 SUM 38 38 38 38 Ch-square = 8,41 p-value = 0,038 3a. DIFFACC (MEDIAN) -0,0006-0,0045-0,0084-0,0161 Rank Sum Test p-values (0-1-2-3) 0,05 0,08 0,13 p-value (0-2) 0,00 p-value (1-3) 0,01 p-value (0-3) 0,00 3b. DIFFACC (MEDIAN) 0,0080 0,0092 0,0110 0,0161 Rank Sum Test p-values (0-1-2-3) 0,12 0,27 0,11 p-value (0-2) 0,05 p-value (1-3) 0,03 p-value (0-3) 0,00 4

Table 6: Regresson Analyss of Dsclosure Qualty Ths table presents coeffcents and p-values (n talc) from regresson model (2) wth DISCL as dependent varable. The test varable CORPGOV, measured by the governance ratng s entered n the second model. It s entered n the thrd model as the predcted value of the frst stage regresson (5). All varables are defned n Table 2. Varable pred. sgn DISCL DISCL DISCL 1 2 3 2-SLS CORPGOV + 0,052 0,079 0,014 0,002 IPR + 4,126 3,263 0,000 0,000 ABS( NI) + -4,530-6,002-3,265 0,291 0,157 0,473 ABS( BVE) + -1,395-1,559-1,770 0,378 0,314 0,298 LNMV + 0,269 0,248 0,289 0,029 0,040 0,028 LNAGE + -0,024 0,028-0,001 0,783 0,755 0,988 ROA?/+ -1,160-2,147-2,073 0,524 0,240 0,292 SLSGR? -0,010-0,008-0,008 0,198 0,328 0,338 LEVERAGE?/+ 0,853 1,026 1,365 0,199 0,116 0,056 OWNDIFF + 0,004 0,000 0,002 0,282 0,934 0,653 USGAAP?/+ 0,169 0,032-0,565 0,735 0,949 0,259 Constant? -4,158-4,680-3,113 0,050 0,025 0,168 INDUSTRYD Incl. Incl. Incl. R² 0,474 0,504 0,404 Adjusted R² 0,251 0,286 0,151 No Obs 152 152 152 5

Table 7: Logstc Regresson Analyss of Early Adopton of IAS39 Ths table lsts coeffcents and p-values (n talc) from the logstc regresson model (3) wth EARLYADOP as dependent varable n each case. The frst specfcaton s a control model. In the second specfcaton, the sgned mpact of IAS 39 on book value of equty of the transton year (IMP(IAS39)) s ncluded. In the thrd specfcaton, a dummy varable measurng the IAS 39 mpact (D_IMP(IAS39)) s ncluded together wth an nteracton dummy varable between CORPGOV and D_IMP(IAS39). IMP(IAS39) s dropped from the model and replaced by ts absolute value (ABS SIZE 39). CORPGOV s entered n the fourth model as the predcted value of the frst stage regresson (5). All varables are defned n Table 2. Varable pred. sgn EARLYADOP EARLYADOP EARLYADOP EARLYADOP 1 2 3 4 2-SLS CORPGOV +/? -0,069-0,162-0,204 0,086 0,006 0,002 IMP(IAS39) + 24,461 25,856 0,012 0,010 D_IMP(IAS39) - -6,602-6,811 0,001 0,002 D_IMP(IAS39)*CORPGOV + 0,205 0,214 0,008 0,013 ABSIMP(IAS39) + 13,943 11,131 0,184 0,271 IPR? -3,053-1,632-1,920 0,069 0,359 0,324 ABS( NI) - 8,961 11,225 14,734 13,154 0,321 0,222 0,118 0,165 ABS( BVE) - -10,716-10,013-10,637-9,354 0,017 0,019 0,021 0,049 LNMV + 0,141 0,156 0,171 0,192 0,497 0,452 0,434 0,380 LNAGE? -0,070-0,105-0,120-0,127 0,677 0,546 0,518 0,486 ROA? -3,347-3,760-2,701-2,224 0,360 0,286 0,451 0,515 SLSGR? 0,018 0,021 0,003 0,028 0,226 0,167 0,110 0,068 LEVERAGE? 1,052 0,824 0,815 0,581 0,367 0,476 0,527 0,644 OWNDIFF + 0,013 0,017 0,022 0,019 0,121 0,062 0,024 0,044 USGAAP + 1,353 1,288 1,709 2,005 0,258 0,280 0,189 0,123 Constant? -0,837-0,498 3,056 1,171 0,826 0,897 0,474 0,779 McFadden R² 0,145 0,154 0,230 0,214 No Obs 152 152 152 152 6

Table 8: Regresson Analyss of Earnngs Management Actvty Panel A and Panel B of ths table present coeffcents and p-values (n talc) from regresson model (4) wth DIFFACC and DIFFACC as dependent varables, respectvely. The test varable CORPGOV, measured by the governance ratng s entered n the second model. It s entered n the thrd model as the predcted value of the frst stage regresson (1). All varables are defned n Table 2. Panel A: Panel B: Varable pred. sgn DIFF ACC DIFF ACC DIFF ACC Varable pred. sgn DIFF ACC DIFF ACC DIFF ACC 1a 2a 3a 1b 2b 3b 2-SLS 2-SLS CORPGOV - -0,001 0,002 CORPGOV + 0,001 0,001 0,002 0,000 0,035 0,001 IPR - -0,039-0,016 IPR + 0,017 0,004 0,010 0,337 0,174 0,793 ACC (PREADOP) + 0,010 0,007 0,005 ACC (PREADOP) - -0,001 0,000 0,002 0,691 0,777 0,828 0,956 0,985 0,914 LNMV - 0,000-0,001-0,001 LNMV + 0,006 0,006 0,006 0,891 0,639 0,631 0,001 0,001 0,000 ROA? -0,078-0,057-0,052 ROA? -0,014-0,024-0,028 0,015 0,072 0,094 0,592 0,365 0,279 SLSGR? 0,002 0,000 0,000 SLSGR? 0,000 0,000 0,000 0,101 0,073 0,069 0,513 0,591 0,625 OWNDIFF - 0,000 0,000 0,000 OWNDIFF + 0,000 0,000 0,000 0,875 0,700 0,886 0,213 0,376 0,233 US GAAP? 0,004 0,005 0,008 US GAAP? -0,007-0,008-0,007 0,709 0,570 0,382 0,373 0,311 0,324 Constant? 0,036 0,048 0,047 Constant? -0,099-0,103-0,112 0,311 0,157 0,166 0,001 0,001 0,000 R² 0,108 0,169 0,153 R² 0,120 0,147 0,171 Adjusted R² 0,064 0,122 0,111 Adjusted R² 0,076 0,099 0,131 No Obs 150 150 150 No Obs 150 150 150 7

Table 9: Regresson Analyss of Overall Restatement Qualty (QRIFRS) Ths table presents coeffcents and p-values (n talc) from the followng tobt regresson model: QRIFRS = α + β CORPGOV+ β IPR + β NIREST + β BVEREST+ β LNMV + 1 2 c 3 β6lnage + β7roa + β8slsgr + β9leverage + β10owndiff + β11usgaap + ε QRIFRS s the dependent varable, defned n Appendx C. The test varable CORPGOV, measured by the governance ratng s entered n the second model. It s entered n the thrd specfcaton as the predcted value of the frst stage regresson (5). All varables are defned n Table 2. 4 5 Varable pred. sgn QRIFRS QRIFRS QRIFRS 1 2 3 2-SLS CORPGOV + 0,003 0,003 0,020 0,009 IPR + 0,165 0,106 0,002 0,075 NIREST + 0,242 0,160 0,267 0,417 0,588 0,375 BVEREST + -0,229-0,230-0,238 0,034 0,031 0,030 LNMV + 0,021 0,023 0,024 0,003 0,001 0,001 LNAGE?/+ -0,002 0,001 0,000 0,729 0,890 0,975 ROA?/+ -0,209-0,267-0,251 0,078 0,025 0,038 SLSGR?/+ 0,000 0,000 0,000 0,883 0,867 0,901 LEVERAGE?/+ 0,061 0,066 0,078 0,130 0,095 0,053 OWNDIFF + 0,005 0,003 0,000 0,095 0,214 0,110 USGAAP?/+ 0,056 0,051 0,029 0,099 0,134 0,375 Constant? 0,275 0,227 0,278 0,037 0,084 0,039 R² 0,206 0,233 0,190 Adjusted R² 0,143 0,167 0,127 No Obs 152 152 152 8

Table 10: Regresson Analyss wth Full Sample Panel A presents coeffcents and p-values (n talc) from regresson model (2) wth DISCL as dependent varable. DISCL s n ths case a score on 4 nstead of 5. Dsclosure on sales s not consdered because of the presence of fnancal companes. The test varable CORPGOV, measured by the governance ratng s entered n the frst model. It s entered n the second model as the predcted value of the frst stage regresson (5). All varables are defned n Table 2. Panel B lsts coeffcents and p-values (n talc) from the logstc regresson model (3) wth EARLYADOP as dependent varable n each specfcaton. In the frst specfcaton, a dummy varable measurng the IAS 39 mpact (D_IMP(IAS39)) s ncluded together wth an nteracton dummy varable between CORPGOV and D_IMP(IAS39). IMP(IAS39) s dropped from the model and replaced by ts absolute value (ABS SIZE 39). CORPGOV s entered n the second model as the predcted value of the frst stage regresson (5). The sample that s consdered n both Panel A and B contans 223 observatons, ncludng 71 fnancal and nsurance companes. All varables are defned n Table 2. Panel A: Varable pred. sgn DISCL DISCL 1 2 2-SLS CORPGOV + 0,061 0,062 0,000 0,001 IPR + 1,174 0,098 ABS( NI) + 2,163 3,756 0,531 0,293 ABS( BVE) + 0,532 0,230 0,638 0,845 LNMV + 0,127 0,169 0,165 0,074 LNAGE + 0,063 0,035 0,371 0,641 LOSS? -0,263-0,331 0,421 0,331 MTBV?/+ -0,003-0,003 0,806 0,813 OWNDIFF + 0,002 0,004 0,494 0,246 USGAAP?/+ -0,147-0,283 0,735 0,517 Constant? -1,964-1,467 0,215 0,364 INDUSTRYD Incl Incl R² 0,380 0,324 Adjusted R² 0,191 0,123 No Obs 223 223 9

Panel B: Varable pred. sgn EARLYADOP EARLYADOP 1 2 2-SLS CORPGOV? -0,107-0,123 0,010 0,007 D_IMP(IAS39) - -5,887-4,550 0,000 0,005 D_IMP(IAS39)*CORPGOV + 0,173 0,121 0,003 0,057 ABSIMP(IAS39) + 9,635 8,388 0,037 0,058 IPR? -0,122 0,934 ABS( BVE) - -6,936-5,561 0,031 0,078 LNMV + -0,103-0,018 0,566 0,920 LNAGE? -0,243-0,294 0,096 0,047 LOSS? 0,737 0,606 0,249 0,337 MTBV? -0,172-0,163 0,039 0,057 OWNDIFF + 0,025 0,023 0,001 0,002 USGAAP + 1,017 0,726 0,334 0,464 Constant? 4,515 3,693 0,177 0,252 McFadden R² 0,219 0,205 No Obs 223 223 10

APPENDIX A: Detals on Corporate Governance Ratngs 1. RIGHTS AND DUTIES OF SHAREHOLDERS a. Treatment of the one share one vote one dvdend prncple. Respect/volaton of the One Share One Vote prncple. Respect/volaton of the One Share One Dvdend prncple b. Votng ssues. Range of ssues that are oblgatorly submtted to the General Meetng. Quorum and majorty requrements. Percentage of Board members elected by the General Meetng c. Shareholder proposals and dervatve acton. Shareholder proposals. Dervatve actons. Dervatve acton for damages v. Acton for annulment d. Votng procedures and methods. Legal tmng for the convenng of an Annual General Meetng. AGM admsson formaltes. Postal and Internet votng v. Proxy votng v. Confdental votng v. Votng methods at the General Meetng v. Follow-up of the General Meetng e. Attendance rate f. Mantenance of pre-emptve rghts of exstng shareholders 2. RANGE OF TAKE-OVER DEFENCES a. Presence of structural take-over defences. Captal structure. Board Insulaton. Votng rght dstortons v. Ownershp rght dstortons b. Presence of captal take-over defences. Share repurchase. Authorsed captal c. Presence of other take-over defences 11

3. DISCLOSURE ON CORPORATE GOVERNANCE a. Dsclosure on general nformaton. Avalablty and language of documents. Accountng standards. Complance wth a Code of Best Practce v. Audtors mandates v. Poltcal and chartable nformaton v. Envronmental nformaton b. Informaton on the company s captal and shareholder structure c. Informaton on the company board. Composton and functonng of the Board. Remuneraton of the Board d. Informaton on the company s commttees e. Informaton on stock optons 4. BOARD STRUCTURE AND FUNCTIONING a. Electon of the members of the company s bodes. General descrpton. Electon of Company drectors. Age lmtaton v. Sze of the Board b. Composton of the company board. Number of Executve Drectors. Independent Drectors. Dversty v. Separaton of CEO and Charman c. Functonng of the board. Workngs of the Board. Internal conduct d. Remuneraton of the company s bodes. Remuneraton of Board Members. Remuneraton of Executve Drectors for ther executve functons. Stock Opton Plans e. Commttees of the board 12

APPENDIX B: Restatement nformaton dsclosed by Grupo Ferroval 1. Restatement of Book Value of Equty n Y0 2. Restatement of Net Income 3. Restatement of Book Value of Equty n Y1 13