Quality of IFRS Adoption*

Transcription

1 Qualty of IFRS Adopton* Arnt Verrest Tlburg Unversty, the Netherlands Ann Gaeremynck Katholeke Unverstet Leuven, Belgum Danel B. Thornton Queen s Unversty, Canada Current Draft: October 2009 Abstract: We nvestgate factors assocated wth European frms IFRS adopton qualty, usng three proxes for adopton qualty: early applcaton of IAS 39, transparency of transton-year dsclosure and rgor of ntal applcaton of IFRS n the ncome statement. In contrast to pror studes, we focus on the IFRS adopton qualty n the restatement phase and explore ts assocaton wth frms accrual qualty before and after IFRS adopton. We hypothesze and fnd that better governed frms dsclose more detals concernng the mpact of IFRS and apply the new standards n a more rgorous manner. However, governance qualty s not unformly assocated wth early adopton of IAS 39. Rather, better-governed frms use the adopton flexblty n a conservatve fashon. If IAS 39 has a negatve mpact on equty book value, well-governed frms tend to adopt the standard early and vce versa. Our results suggest that poltcal nvolvement n the mplementaton process of new accountng standards s undesrable. Frms are lkely to use carve-out optons opportunstcally or n ways that are nconsstent wth the IFRS Conceptual Framework. Key Words: Internatonal Accountng, IFRS Restatements, Corporate Governance, Dsclosure, IAS39 JEL classfcatons : G14, G15, G30, M38, K22, M41 We apprecate the helpful comments of Ulf Bruggemann, Lesbeth Bruynseels, Laurens Cherchye, Hans Chrstensen, Peter Easton, Stephan Hollander, Laurence van Lent, Frank Lu, Don Stokes, Marleen Wllekens and Yachang Zeng. Ths paper also benefted from beng presented at the 2010 IAS Mdyear Conference and from comments from the workshop partcpants at Tlburg Unversty and the Katholeke Unverstet Leuven. The INTACCT research Network and the Fund for Scentfc Research (FWO) s gratefully acknowledged for fnancal support. 1 Electronc copy avalable at:

2 1. Introducton The ntroducton of Internatonal Fnancal Reportng Standards (IFRS) for lsted companes n more than 100 countres s one of the most sgnfcant regulatory events n accountng hstory. For EU lsted companes, European Regulaton made IFRS compulsory from 2005 onwards, wth the objectve of enhancng the qualty of fnancal reportng. Although the standards are ssued by an ndependent accountng standard setter, the Internatonal Accountng Standards Board (IASB), adopton of IFRS by European countres s not a straghtforward process. Each standard must pass a two-tered endorsement mechansm nvolvng a possble carve-out opton, under whch frms may elect to delay or avod applyng a standard. Moreover, the rgor or serousness wth whch frms apply the measurement and dsclosure provsons of ndvdual standards s unlkely to be unform, dependng on frm-specfc ncentves and country-specfc factors. Ths study nvestgates how IFRS adopton qualty dffers across frms and how adopton qualty relates to frm-specfc governance qualty, durng the restatement phase from local to nternatonal standards. We focus on the qualty of the IFRS adopton process tself. Companes are requred to restate ther fnancal statements from local GAAP to IFRS durng the transton year 1. Pror studes have prmarly focused on the change n accountng qualty exhbted by frms n perods followng IFRS adopton (Cujpers and Bujnk, 2005; Chrstensen et al., 2007; Barth et al., 2008; Garca-Osma and Pope, 2009). In contrast to these studes, we focus on the change n qualty around the tme of the actual adopton of IFRS. We employ three measures of IFRS adopton qualty: (1) early adopton choce of a controversal standard relatng to fnancal nstruments, IAS 39, n 2004 nstead of makng use of a carve-out opton proposed by the EU; (2) transparency of dsclosure of IFRS mpact on fve specfc fnancal statement tems; and (3) rgor of IFRS restatements, measured by 1 For companes wth a year-end of December 31, ths s 2004; for companes wth a year-end dfferng from December 31, the transton year s Electronc copy avalable at:

3 dfferences n the magntudes of accruals computed under IFRS vs. local GAAP for the transton year. For a sample of 223 European fnancal and nonfnancal lsted frms belongng to the MSCI Pan Euro Index, we hand-collect fnancal statement data relatng to varous dsclosures, the early adopton of IAS39, and accruals under local GAAP and IFRS for the transton year. All frms are mandatory adopters of IFRS. Our accrual qualty measure s smlar to earnngs management measures used by Barth et al. (2008) and Chrstensen et al. (2008). However, our same frm-year approach facltates the evaluaton of the qualty of IFRS adopton because frm-level dfferences between accruals under local GAAP and those under IFRS are not subject to changes n the economc envronment or n dsclosure ncentves over tme. Ths same frm-year approach has been used n market-based studes (e.g. Wang and Welker, 2008; Hung and Subramanyam, 2007; Clarkson et al., 2008) but not n accountng-based studes nvestgatng the mpact of IFRS on earnngs qualty. Our descrptve results reveal sgnfcant dfferences n the qualty of the IFRS restatement process across frms and countres. Only 44% of the sample frms early adopt IAS 39; on average, however, frms make extensve dsclosures about the fnancal statement mpact of IFRS adopton. We also detect a sgnfcant ncrease n accrual qualty durng the transton. Next, and perhaps more mportantly, our unvarate results ndcate a postve relatonshp between corporate governance qualty and IFRS adopton qualty. The lack of a clear relatonshp between frm governance and the early adopton of IAS 39 s further nvestgated n a multvarate settng. We fnd that well-governed frms tend to adopt IAS 39 n a conservatve fashon. They adopt IAS 39 early (or voluntarly) when the applcaton of the standard has a negatve mpact on equty. Our multvarate analyses confrm the sgnfcant and postve assocaton between corporate governance qualty on the one hand, and the transparency of IFRS dsclosures and IFRS restatement rgor on the other. 3

4 Specfcally, we fnd that a well-functonng board s assocated wth better accountng qualty n the restatement phase. All results are robust to senstvty checks and the employment of two-stage least squares methods to allow for the endogenety of corporate governance strength and fnancal reportng choces. Furthermore, we document an emprcal regularty suggestng that frms IFRS adopton rgor s not assocated wth frms pre-ifrs earnngs qualty. We do fnd a postve assocaton between IFRS adopton qualty and subsequent earnngs qualty under IFRS. These fndngs are consstent wth the noton that IFRS adopton represents a watershed n fnancal reportng n whch frms wth low earnngs qualty under local GAAP do not apply IFRS n a less serous manner. Fnally, we report an ncrease n earnngs qualty after mandatory IFRS adopton, confrmng earler fndngs n the lterature (Barth et al., 2008; Chrstensen et al., 2008). Both sgned and unsgned abnormal accruals are sgnfcantly lower after IFRS adopton, ndcatng an mprovement n accrual qualty. We contrbute to the lterature n three ways. Frst, unlke many pror studes, ths study consders accountng qualty changes around the tme of IFRS adopton rather than lookng at accountng qualty pre- and post- IFRS adopton, employng a same-year approach. We also document dsclosure precson and adopton choces that frms make durng ths transton perod. Second, unlke prevous studes that measured governance qualty at the country-level, such as Daske et al. (2008), ths study consders the nfluence of frm-specfc governance qualty on the applcaton qualty of IFRS. These results are mportant as there s consderable varaton n governance qualty wthn countres. Thrd, to provde evdence on the mpact of mandatory IFRS adopton on accountng qualty, we employ a potentally superor accrual qualty measure based on Francs and Wang (2008). To the best of our knowledge ths metrc has not been used n the lterature to document earnngs qualty changes around IFRS transton. 4

5 The paper contnues as follows. Secton 2 revews pror lterature leadng to the formulaton of hypotheses. Secton 3 dscusses the sample selecton. Secton 4 explans the model desgn. Secton 5 provdes descrptve statstcs and unvarate analyses. Secton 6 presents regresson results and robustness checks. Secton 7 concludes. 2. Lterature Revew and Hypothess Development Pror lterature focusng on the benefts of voluntary IFRS adopton provdes mxed results. Bartov et al. (2005) report that IFRS numbers are more value relevant than German GAAP accountng numbers. For a sample of Chnese frms owned by foregn nvestors, however, Eccher and Healy (2003) fnd that IFRS accountng measures are less value relevant than those based on Chnese GAAP. Pae et al. (2008) fnd that mnorty shareholders of EU frms wth the hghest agency costs are most lkely to beneft from IFRS adopton. Bae et al. (2008) report that the extent to whch GAAP dffers between country pars s negatvely related to both the number of foregn analysts followng frms and analysts forecast accuracy; hence IFRS adopton s lkely to enhance nternatonal analyst scrutny, promotng hgher qualty earnngs and a lower cost of equty captal. Usng a same frm-year desgn smlar to ours, Hung and Subramanyam (2007) fnd that, for 80 voluntary IFRS adopters, book value restatements are value relevant but ncome restatements are not. Clarkson et al. (2008) fnd that followng IFRS adopton, the value relevance of accountng numbers decreases for frms n common law countres. They fnd nsgnfcant changes for frms n code law countres. The average effects of voluntary IFRS adopton on market lqudty and cost of equty captal are reported to be modestly postve (e.g. Daske et al., 2008; Barth et al., 2005; Cujpers and Bujnk, 2005). In fact, enhanced accountng qualty and sgnfcant economc effects are only observed only for serous IFRS adopters as opposed to label IFRS 5

6 adopters, where serous IFRS adopters are defned as those that report comparatvely large decreases n the level of accruals due to IFRS adopton (Daske et al., 2007). Moreover, the extent of countres nvestor protecton and legal enforcement are postvely assocated wth the benefcal mpact of changes n accountng standards (Wang and Yu, 2008; Ball et al., 2003; Burgstahler et al., 2006; Daske et al., 2008). For a sample of German frms, Van Tendeloo and Vanstraelen (2005) and Goncharov and Zmmermann (2006) fnd lttle evdence that IFRS adopton curbs earnngs management relatve to German GAAP. However, usng a tme-constant sample of frms n 21 countres, Barth et al. (2008) fnd a sgnfcant ncrease n accountng qualty (.e. less earnngs smoothng, less management of earnngs towards a target and more tmely loss recognton) followng IFRS adopton. On balance, pror lterature suggests that there are economc benefts from adoptng IFRS, and that the benefts of adopton are postvely assocated wth the rgor wth whch frms mplement IFRS. We extend the lterature by assessng the determnants of IFRS adopton qualty around the tme of transton tself, usng three qualty measures: (1) Early applcaton of IAS39: Frms rgorously applyng IFRS are defned as those that apply IAS39 n 2004 rather than postponng the frst applcaton to 2005 by usng the proposed carve-out opton, whch was avalable followng poltcal nvolvement by the European Unon. 2 Ultmately the carve-out opton dd not materalze and all EU frms had to adopt IAS 39 n 2005 (Brackney and Wtmer, 2005). (2) Transparency of dsclosure n fnancal statement notes: Frms consderng IFRS as a means to enhance transparency are expected to dsclose more nformaton on adopton. We 2 Ths proxy relates to prevous studes nvestgatng ncentves to opt for the mmedate applcaton of a new accountng standard (Langer and Lev, 1993; Amr and Zv, 1997a and 1997b; Gaeremynck and Van de Gucht, 2004). 6

7 develop a dsclosure score representng the most relevant restatement tems for valuaton purposes (the mpact of IFRS on equty, earnngs, cash flows and sales). 3 (3) Rgor of ntal applcaton of IFRS n the ncome statement: Ths s measured as the change n accruals due to the change from local GAAP to IFRS. In contrast to Barth et al. (2008), we use IFRS and local GAAP data from the same fscal year (.e. 2004). Ths samefrm-year research desgn allows us to gauge the mpact of the accountng standards per se on the level of accruals wthout needng to control for bases ntroduced by frm-level changes n economc ncentves over tme. Frms that more rgorously apply IFRS on the transton date are expected to report larger decreases n accrual levels (Daske et al., 2008; Barth et al., 2008). Although the three IFRS adopton qualty ndcators are dstnct from each other, we hypothesze all of them to be postvely assocated wth the strength of frms corporate governance practces, after controllng for other factors assocated wth fnancal reportng ncentves reported n Daske et al. (2008),.e., frm sze, proftablty, fnancng needs, growth opportuntes and ownershp concentraton. Compared wth US frms, European frms are more lkely to choose non-value maxmzng governance levels because they are characterzed by concentrated ownershp structures (Facco and Lang, 2002), sgnfcant prvate benefts for nsders (Leuz et al., 2003; Dyck and Zngales, 2004; Pae et al., 2008) and the absence of an actve takeover market (Mkkelson and Partch, 1997). Thus, stronger governance s expected to have a sgnfcant and negatve mpact on nformaton asymmetry for frms that are more transparent at the tme of the change from local GAAP to IFRS (Shlefer and Vshny, 1997; Larcker et al., 2007; Goodwn et al., 2009). Ths leads to our man hypothess that corporate 3 Horton and Serafem (2007) report sgnfcant dfferences between frms wth respect to the tmng of IFRS dsclosures. We do not consder the tmng of the IFRS restatements. Tmng s not relevant n our settng snce we nvestgate determnants and not the nformaton content of the restatement process. 7

8 governance strength s postvely assocated wth IFRS adopton qualty durng the transton phase. 4 After testng ths hypothess, we explore the assocaton between the thrd measure of IFRS adopton qualty (.e. the change n accruals n gong from local GAAP to IFRS) and frms accrual qualty (1) before and (2) after adoptng IFRS (.e. the estmated abnormal accruals). We do not advance any drectonal predctons for these assocatons. As to the frst assocaton, arguably low agency-cost frms wth low accrual qualty under local GAAP are lkely to seze the opportunty to effect a clean-up,.e., to make larger decreases n accruals on IFRS adopton so that there s less overhang (such as unrecognzed actuaral penson losses) to wrte off after IFRS adopton. However, hgh agency-cost frms wth low accountng qualty under local GAAP may be less lkely to consder IFRS as a vehcle for hgher transparency, and therefore report a lower ncrease (or no ncrease at all) n earnngs qualty. It can even be argued that there s no relatonshp between accrual qualty under local standards and IFRS adopton qualty, snce each frm starts wth a clean sheet of paper applyng IFRS 1 for the frst tme. As to the second assocaton, arguably frms adoptng IFRS rgorously for the frst tme are lkely to ascrbe more value to supplyng transparent nformaton and therefore provde hgher accrual qualty afterwards as well. However, t s also arguable that IFRS adopton qualty reflects a one-tme opportunty to clean up the balance sheet, so t s unrelated to future earnngs qualty. Thus, for these assocatons we test the null hypotheses of no assocaton and comment on the results, ex post. Further, we predct governance qualty to be postvely assocated wth future accrual qualty under IFRS as prevous studes have documented dfferental qualty effects of IFRS, dependng on (country-level) governance qualty (e.g. Daske et al., 2008) and ncentve structures (Chrstensen et al., 2008). 4 Our study complements prevous lterature reportng postve assocatons between governance strength and both dsclosure qualty and earnngs qualty (e.g. DeFond and Jambalvo, 1994; Klen 2002; Karamanou and Vafeas, 2005; Peasnell et al., 2005; Marques, 2006). 8

9 3. Sample selecton and data The ntal sample conssts of companes belongng to the MSCI Pan Euro Index for the perod January 1, 2005 through June 30, From the ntal sample of 299 frms, 18 US GAAP frms are deleted from the sample 6 ; 54 frms are voluntary IFRS adopters; 3 frms were exempted from applyng IFRS mandatorly and 1 frm merged durng the sample perod. Ths procedure results n a fnal sample of 223 frms across 15 countres. 7 Early IFRS adopters are elmnated for three reasons: (1) Some frms already apply IFRS (and IAS) for long perods, makng t costly to track back the annual reports. (2) Voluntary adopters were not requred to dsclose detaled restatement nformaton and dd not face the opton to apply IAS39. (3) Prevous research suggests that there are sgnfcant dfferences between voluntary and mandatory IFRS applers. Table 1 shows the dstrbuton by country as well as the splt between fnancal and nonfnancal frms. As expected, the number of nonfnancal frms (152) s sgnfcantly larger than the number of fnancal frms (71). The UK has the largest number of observatons n the sample (76) followed by France (44), Italy (21) and Span (19). Germany, Austra and Swtzerland lose a lot of observatons because of the elmnaton of voluntary adopters. Insert Table 1 about here 5 The MSCI Pan-Euro Index contans securtes wth a free float-adjusted market captalzaton of 4,348 bllon, selected from 16 European countres. 6 These are frms that ONLY report under US GAAP. Frms that report under both IFRS and US GAAP are ncluded n the sample. 7 Our sample selecton crtera lmt our analyss to large frms only. Although we are aware of ths external valdty problem, we argue that the large frms n our sample represent a substantal fracton of the market value of all European publc companes. In that sense, our results matter a great deal. Furthermore, the varance n corporate governance practces between frms s smaller n large than n small companes. If we fnd a relatonshp between governance and IFRS restatement qualty for those large frms, t can be expected that the relatonshp wll also exst and wll probably be stronger for smaller companes. 9

10 The transton year 2004 s used n a same frm-year desgn to nvestgate the determnants of IFRS adopton qualty and ts varaton across frms. Ths s the only year for whch both local and IFRS accountng numbers are avalable reflectng the same economc events. Ths nformaton s hand-gleaned from data n the 2005 fnancal statement notes. Ths s the actual IFRS adopton year. 8 Corporate governance data for 2004 come from the Rsk Metrcs corporate database (formerly the Demnor database), whch contans detaled corporate governance nformaton for frms n the MSCI Pan Euro Index. 9 The ratng s based on more than 300 corporate governance crtera, measured n Appendx A provdes a lstng of the tems ncluded n the corporate governance score. Aggregated data as well as more detaled data on ndvdual corporate governance tems (e.g. audt commttee, board ndependence and functonng of the board) are avalable. 4. Model development To test the hypothess that IFRS adopton qualty s postvely assocated wth governance strength at the frm-level, we use three regresson models n whch corporate governance strength (CORPGOV) s an explanatory varable for the three IFRS adopton qualty proxes (QRIFRS) as follows: QRIFRS α 1 CORPGOV 2Control varables ε (1) 8 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 Henceforth we refer to the pre-adopton, transton and adopton year as beng 2003, 2004 and 2005 for all frms n the sample. 9 Demnor Ratngs have been used n pror studes such as Bauer et al. (2004), Bozec et al. (2007), Florou et al. (2007) and Van der Bauwhede and Wllekens (2008). 10

11 In equaton (1), CORPGOV represents the company s corporate governance ratng and appears n all three models as the man explanatory varable of nterest. 10 Consstent wth prevous lterature, we predct a postve coeffcent for CORPGOV snce governance strength s negatvely assocated nformaton asymmetry between the frm and outsde nvestors (Klen, 2002; Karamanou and Vafeas, 2005; Larcker et al., 2007). The dependent varable QRIFRS s ether (1) early adopton of IAS39 (EARLYADOP), (2) dsclosure qualty of restatements (DISCL) or (3) rgor of the restatement (DIFFACC REST ). Each of these adopton qualty measures s descrbed n detal below, along wth the accompanyng control varables. IFRS adopton qualty measure #1: EARLYADOP The frst measure of IFRS adopton qualty s EARLYADOP, an ndcator varable takng the value 1 f a frm voluntary adopts IAS 39 n 2004 and 0 the frm delays IAS 39 applcaton untl The frst control varable, IMP(IAS39), proxes for the mpact or the materalty of changes n accountng numbers due to applyng IAS 39 (Langer and Lev, 1993; Amr and Zv, 1997a and 1997b). It s defned as the mpact of IAS 39 adopton on transtonyear book value of equty, scaled by market value of equty (at the end of 2004). As the mpact of IAS39 becomes more materal, early adopton of the new standard s more lkely. To control for the materalty of IFRS restatement n total, we nclude the unsgned mpact of IFRS adopton on net ncome, ABS(ΔNI), where ΔNI s defned as transton year net ncome under IFRS mnus net ncome under local GAAP, scaled by market value of equty. We also nclude the unsgned mpact of IFRS adopton on book value of equty, ABS(ΔBVE), where ΔBVE s defned as transton year book value of equty under IFRS mnus book value of equty under local GAAP, scaled by market value of equty. 10 Results are robust to usng 2004 or 2005 ratngs. Ths s expected snce corporate governance tends to be stcky over tme. We prefer 2004 fgures to mtgate endogenety. For 10 observatons we had to use the corporate governance ratng of 2005 because the 2004 ratng was not avalable. Regresson results reman qualtatvely the same wthout those 10 observatons. 11

12 The fourth control varable s a country-level measure of nvestor protecton rghts, IPR, whch s the ant-drector rghts ndex from LaPorta et al. (1998). Ths ndex s based on the presence of sx elements of nvestor protecton n a country s corporate legslaton. IPR thus controls for country-level ncentves for frms to rgorously apply IFRS (Ball et al., 2003; Burgstahler et al., 2006; Daske et al., 2008; Wang and Yu, 2008). In strong legal envronments the IFRS restatement process s expected to be hgher n qualty. Frm-level ncentves are captured by frm sze (LNMV), ownershp dffuson (OWNDIFF), proftablty (LOSS), growth opportuntes (MTBV), the age of the frm (LNAGE) and whether the frm also reports under USGAAP or not (Daske et al., 2008; Raffourner, 1996) 11. Frms that are larger and have a more dspersed ownershp are expected to be more lkely to commt to transparency n the IFRS change process. In contrast to pror studes, we do not control for audtor type because all sample frms employ a Bg-4 audtor. The early adopton model for IAS 39, estmated usng a logt model, s as follows: EARLYADOP α 6 9 CORPGOV 1 ABS( MKTBV BVE) IMP(IAS39) LNMV OWNDIFF LNAGE IPR USGAAP c 8 LOSS ε 4ABS( NI) (2) IFRS adopton qualty measure #2: DISCL The second measure of IFRS adopton restatement qualty s DISCL, an ndex of the transparency of companes dsclosure of the mpact of IFRS adopton. DISCL s based on the presence or absence of the followng fve quanttatve dsclosures of the restatement mpact of IFRS adopton: (1) net ncome n 2004; (2) book value of equty at the begnnng of 2004; (3) 11 To make fgures comparable between fnancal and other frms, a loss dummy and MTBV are preferred above ROA and Growth n Sales to capture performance and growth, respectvely. However, results reman the same when replacng both varables for alternatves. 12

13 book value of equty at the end of 2004; (4) operatng cash flow for 2004; (5) total revenues or sales for Net ncome and equty are ntroduced n the dsclosure ndex because, under IFRS 1, frms must provde detaled nformaton on the restatement of these tems. Dsclosure of quanttatve nformaton concernng the restatement of cash flows s largely voluntary but these tems could be mportant to nvestors as an alternatve performance measure. To earn a score of 1 for a partcular tem, a frm must dsclose the dfferences between the IFRS and the prevous local GAAP number for each specfc nternatonal standard gvng rse to the dfference. As the sales varable s only relevant for the nonfnancal frms, two dsclosure varables are created. DISCL5 ncludes the sales restatement and only apples to the nonfnancal sample. DISCL4 wthout the sales restatement apples to the total sample. Recall that DISCL4 and DISCL5 do not focus on the overall ncrease n dsclosure due to the ntroducton of IFRS. Rather, they focus on the dsclosure of the IFRS mpact of the restatement phase tself. Durng a restatement phase, the avalablty of nformaton about the IFRS mpact on the equty, earnngs, cash flows and sales s hghly relevant for nvestors. In the dsclosure model, control varables are smlar to those n the early adopton model except for the omsson of the IMP(IAS39) and the addton of ndustry dummes whch are found to be sgnfcantly assocated wth dsclosures n pror lterature (Raffourner, 1996). The other control varables are based on prevous lterature (Raffourner, 1995; Brown et al., 1999; Holland, 2005); they control for frms ncentves to be more transparent at the tme of the swtch from local to nternatonal standards. The OLS regresson specfcaton for testng the dsclosure model s as follows: 12 In the dsclosure ndex equty s consdered twce because both begnnng and and of the year book equty should be restated accordng to IFRS. Prevous evdence has also shown that the restatement to IFRS has the largest mpact on equty. 13

14 DISCL α 5 9 CORPGOV 1 LNMV 6 OWNDIFF LNAGE 10 2 IPR 7 c USGAAP 3 LOSS ABS( j 8 NI) DINDUSTRY MTBV 4 ABS( j ε BVE) (3) IFRS adopton qualty measure #3: DIFFACC REST Our thrd measure of adopton qualty assesses the rgor of IFRS restatement n the ncome statement. Ths s represented by DIFFACC REST, the sgned dfference between the magntude of accruals computed under IFRS and the magntude of accruals computed under local GAAP, both scaled by total assets. 13 Pror lterature mples that smaller (more negatve) values of DIFFACC REST ndcate more rgorous applcaton of IFRS (Leuz et al., 2003; Barth et al., 2008; Daske et al., 2008). Unlke those studes, ths study uses a same frm-year research desgn n the year of the restatement (Hung and Subramanyam, 2007; Clarkson et al., 2008) to proxy the IFRS restatement qualty of the accruals. The chef advantage of ths approach s that the qualty measure s mmune from confoundng nfluences over tme, so that dfferences n accruals can be totally attrbuted to dfference n accountng standards 14. In contrast to Leuz et al. (2003), Barth et al. (2008) and Garca-Osma and Pope (2009), we do not use the rato of the standard devaton of net operatng ncome (deflated by assets) dvded by the standard devaton of operatng cash flows (deflated by total assets) or the Spearman Rank correlaton changes n accruals and changes n cashflows as proxes for earnngs smoothng. The reason s that the mpact of IFRS n the restatement process s estmated on frm level and s then lnked to frm governance. Ths also explans why tmely loss recognton as well as value relevance are not consdered. 13 We scale by total assets n the current nstead of prevous 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 the absolute value of operatng cash flows. Agan, results reman qualtatvely the same. 14 Consstent wth Hrbar and Collns (2002), accruals are calculated as net ncome mnus net operatng cash flows scaled by total assets. Accruals under IFRS are based on net ncome fgures excludng any nfluence from IAS39. 14

15 Control varables for reportng ncentves to supply hgh qualty nformaton are smlar to those used n models (2) and (3) except for sales growth and control varables specfc to accrual qualty models. The ncrease n sales s preferred as the growth measure because the lnk to the level of accruals s strong. DIFFACC REST s not consdered for fnancal frms. To control for pror performance we ntroduce lagged (.e., preadopton year) operatng cash flows scaled by total assets (CFO(PREADOP)). Worse performng frms are more lkely to use ncome ncreasng accruals (Bowen et al., 2008; Hbron and Nchols, 2007). Further, we nclude lagged accruals proxes for earnngs qualty under local GAAP. Hence, model (4) s as follows: DIFFACC REST α 4 7 CORPGOV 1 ACC(PREADOP) 3LNMV OWNDIFF 8 2 IPR c USGAAP 3 ε CFO(PREADOP) SLGR 6 (4) Insert Table 2 about here 5. Descrptve Statstcs and Unvarate Analyses Panel A of Table 2 contans descrptve statstcs for the varables descrbed above. We note large dfferences across frms for the three IFRS adopton qualty proxes. Only 44% of sample frms early-adopt IAS39. The mean of DISCL4 s 2.9 out of a possble frms obtan a score of 4.0, manly French and Brtsh frms. When the sample s lmted to the nonfnancal frms, the mean of DISCL5 s DIFFACC REST has a negatve mean (medan) of 0.75% ( 0.50%) of total assets, ndcatng that frms generally report lower accruals under IFRS than under local GAAP. The sample conssts of large frms wth a mean (medan) market captalzaton of 24 (12) bllon US dollars. The mean (medan) value of CORPGOV s only (24.80) out of a possble 40, wth a large standard devaton (5.89). Ths ndcates consderable varaton n the sample n terms of overall governance strength. On average, IFRS adopton yelds a 1.26 % 15

16 ncrease n net ncome (ΔNI) and a 0.95% decrease n book value of equty (ΔBVE). 15 The mean of IMP(IAS39), the mpact of applyng IAS 39 on transton-year book value of equty for both early and late adopters, s slghtly postve (0.67%) whle the mean, unsgned mpact s 2.4%. As for other control varables, the average frm exhbts a market to book value of 3.18, 7% of the frms report losses and only 3% of the companes prevously reported under US GAAP. Panel B n Table 2 provdes country-level statstcs on restatement qualty, governance ndcators and restatement magntudes. We observe large dfferences across countres for all IFRS restatement qualty varables. The lkelhood of early adopton of IAS39 s low n the UK (0.32) and Sweden (0.23); however, t s hgh n Fnland (1.00) and Southern European countres. Even though some dsclosure tems are compulsory, Southern European companes exhbt low DISCL scores. Fnally, n Germany accruals are slghtly larger under IFRS than under local standards. Ths results n a postve DIFFACC (0.0049). In contrast, for the UK and Ireland the medan dfference s negatve ( and ). The followng two columns depct dfferences between countres for ΔNI and ΔBVE. Interestngly, IFRS adopton leads to hgher earnngs n most countres. Book value of equty decreases n some countres (e.g. n the Netherlands and the UK) and ncreases n others (e.g. all Scandnavan countres). Overall, we conclude that IFRS restatement qualty and the mpact on equty and earnngs sgnfcantly vares across countres. Panel B also ndcates that CORPGOV s hgher n common law than code law countres. Dutch, French and Scandnavan companes also exhbt hgher CORPGOV. Swss, Portuguese and Greek frms score lower on average. Panel C n Table 2 shows that 73% of sample frms late-adopt IAS 39 when the mpact on book value of equty s negatve but only 42% late-adopt when the mpact s postve. These 15 We scale all IFRS adopton measures by market value of equty (Easton and Sommers, 2003) nstead of an ndex (Weetman and Gordon, 2006). 16

17 statstcally sgnfcant dfferences n early adopton rates are consstent wth frms preferrng to delay reportng unfavorable changes n equty and leverage 16. In other words, the choce of applcaton seems to exhbt opportunsm. There are no sgnfcant dfferences n the percentage of late adopton n the total sample vs. the subsample of nonfnancal frms. Panel D of Table 2 llustrates that dfferences n dsclosure behavor are mostly due to not reportng the IFRS mpact on net ncome, operatng cash flows and sales. The voluntary nature of these dsclosures probably explans why less than one thrd of the total sample dscloses the IFRS effect on operatng cash flows, but not why 15% of the frms do not report the compulsory effect on net ncome. Panel E of Table 2 depcts how earnngs qualty changed movng from local to IFRS based on total and abnormal accrual measures. DIFFACC REST conssts of the accruals under local GAAP ACC Local(04) / TA Local (04) and accruals under IFRS ACC IFRS(04) / TA IFRS(04). As expected, mean and medan accruals are smaller n magntude under IFRS than under local GAAP (one-sded p-values are and respectvely). As n Barth et al. (2008), the accruals under local GAAP ( ACC Local(04) / TA Local(04) ) are sgnfcantly lower n comparson wth IFRS accruals n the adopton year ACC IFRS(05) / TA IFRS(05). When the decrease n accruals s compared to the decrease usng the same frm-year desgn, the larger drop n accruals (from to vs. from to 0.061) shows that confoundng effects can lead to erroneous nferences about changes over tme. Fnally, a more refned measure of earnngs management s developed by computng abnormal accruals measured as n Francs and Wang (2008) and DeFond and Park (2001) 17. The hghly sgnfcant dfferences between 16 In addton to the effect on equty, we also capture the effect on equty/debt. Results are qualtatvely the same. 17 The abnormal accruals usng the DeFond and Park (2001) measure are computed as: ABNACC t = Total Accruals t - Predcted Accruals t, where Predcted Accruals t = ([Sales t x (Current Accruals t-1 / Sales t-1 )] [Gross PPE t x (Deprecaton t-1 / Gross PPE t-1 ]) / Total Assets t-1 Local GAAP accruals are based on data; IFRS accruals are based on data. Negatve (postve) abnormal accruals evnce ncome-decreasng (ncome-ncreasng) earnngs management. Hgher unsgned abnormal accruals sgnfy lower predctable accruals and thus more earnngs management. Hrbar and Nchols (2007) argue that sgned accruals are a better measure of earnngs qualty than unsgned accruals. 17

18 sgned accruals ndcate less ncome ncreasng earnngs management under IFRS. Dfferences between unsgned accruals mply a hgher predctablty of accruals and thus less earnngs management n general. Insert Tables 3 and 4 about here Table 3 provdes correlaton coeffcents for all relevant varables descrbed above. As expected, CORPGOV s postvely correlated wth DISCL and negatvely wth the sgned DIFFACC REST. Contrary to expectatons, the dummy varable for early adopton (EARLYADOP) correlates negatvely to CORPGOV. Table 4 provdes further descrptve nformaton on unvarate assocatons between corporate governance (CORPGOV) and the three adopton qualty measures (EARLYADOP, DISCL, and DIFFACC REST ). Results are gven for the total sample as well as the sample of nonfnancal frms. In ths table, we splt the total sample of 223 frms arbtrarly nto four groups of 56 frms, dependng on CORPGOV: Group 1 contans the 25% worst governed frms (mean CORPGOV = 16.58) and Group 4 contans the 25% best-governed ones (mean CORPGOV = 32.23). Contrary to expectatons, we do not fnd a sgnfcant relatonshp between corporate governance and the early adopton of IAS39 for the total sample. For the nonfnancal sample, frms wth worse governance apply IAS39 n a tmeler manner 18. Wth regard to DISCL4, the sgnfcant Ch-square value for both samples llustrates that dsclosure scores ncrease sgnfcantly when movng from Group 1 to Group 4. Fnally, we notce that DIFFACC REST becomes sgnfcantly larger n magntude gong from Group 1 to 4. Moreover, Accrual qualty can also be compared by computng the correlaton between cash flows and accruals. The Pearson and rank correlatons between sgned accruals and cash flows are more negatve under local GAAP than under IFRS, ndcatng less smoothng actvty. However, as n the current study the lnk between corporate governance and IFRS qualty s nvestgated, we do not use ths proxy further. 18 A more detaled analyss shows that most Brtsh frms (whch usually have hgh qualty governance) do not adopt IAS39 n an early fashon. 18

19 DIFFACC REST s sgnfcantly dfferent from zero only for the two hghest governance classes. These results suggest that better governed frms engage n a more rgorous or serous applcaton of IFRS for the frst tme than worse corporate governance frms (p=0.000). 6. Multvarate analyses Early Adopton Model Table 5 contans regresson results of the early adopton logt model. EARLYADOP s the dependent varable. We report results on the total sample (Models 1-4) as well as for both subsamples of nonfnancal (Model 5) and fnancal frms (Model 6). Model (1) s a control model wthout CORPGOV as an explanatory varable. The probablty of early-adoptng IAS 39 s postvely assocated wth IAS 39 mpact (p = 0.000), confrmng the result n Panel C of Table 2. Restatement value of equty (p = 0.011) s negatvely assocated wth EARLYADOP. Apparently, frms that already face reportng sgnfcant changes n ther book value of equty because of IFRS transton are less lkely to pre-adopt IAS39. In Model (2), CORPGOV has a negatve but nsgnfcant coeffcent, suggestng that stronger governance s not assocated wth early adopton n general. Insert Table 5 about here In Model (3) the unsgned mpact of IAS39 s replaced by a dummy varable, D_IMP(IAS39), whch equals 1 f the frm experences a negatve mpact on ts book value of equty from applyng IAS39 and zero otherwse. An nteracton varable between CORPGOV and D_IMP(IAS39) measures how governance strength ncrementally nfluences the early adopton choce for frms facng a negatve effect from IAS 39. The unsgned IAS39 mpact s retaned n the model to nvestgate whether the sze stll matters, gven a postve or negatve mpact of IAS39. The negatve coeffcent for D_IMP(IAS39) (p=0.000) mples that frms 19

20 facng a negatve mpact of IAS 39 on equty tend to postpone the applcaton of the standard. Consstent wth our hypothess, the sgnfcantly postve coeffcent for the nteracton varable mples that governance strength ncrementally enhances early adopton probablty for frms negatvely mpacted by the standard (p=0.003). In Model (4), we test a two-stage least square model (2-SLS) to allow for the possble endogenety of IFRS adopton qualty and corporate governance (Larcker et al., 2007; Zhao and Chen, 2008). We choose country-level characterstcs (represented by country dummy varables) as nstruments, smlar to Klapper and Love (2004). 19 In the second stage, CORPGOV s replaced by ts predcted value from the frst stage. IPR s not ncluded n the second stage because collnear country effects were already ntroduced as nstruments n the frst stage. Results of Model (4) wth control for possble endogenety problems are smlar to the results found n Model (3). Models (5) and (6) ndcate a sgnfcant effect of IMP(IAS39) on the choce of applcaton of IAS39. The coeffcent on the nteracton term wth corporate governance s sgnfcantly postve for both samples and larger n magntude for n the fnancal sample. Well-governed fnancal as well as nonfnancal frms prefer to report conservatvely by choosng early applcaton of IAS39 when the mpact of IAS39 conveys bad news. Dsclosure model Table 6 contans the results of estmatng the dsclosure-qualty model for the total sample as well as for both subsamples separately. As the restatement of sales from local 19 The reduced form equaton from the frst stage looks as follows: CORPGOV α COUNTRYDUMMY ε c 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 to alternatves because they are exogenous and prove to be a strong nstrument (R² s about 0.65 n the frst stage). 20

21 GAAP to IFRS s not avalable for the subsample of fnancal frms, results are gven usng DISCL4 as the dependent varable for the total sample. In the control specfcaton (1), the level of dsclosure n the IFRS restatement phase s postvely assocated wth nsttutonal nvestor rghts (p=0.055) and frm sze (p=0.027). In Models (2) and (3), as hypotheszed, CORPGOV s postvely assocated wth dsclosure qualty and s robust to two-stage least squares estmaton. Fnally, Model 4 and 5 confrm the postve mpact of governance on the extent of dsclosure for fnancal as well as nonfnancal frms. If the dsclosure score s extended from 4 to 5 tems, the assocaton between governance and dsclosure n the restatement phase becomes even stronger. Ths s expected, gven the voluntary nature of the IFRS mpact on sales dsclosure. Insert Table 6 about here Rgorous applcaton model The last proxy used to assess adopton qualty n the restatement phase s the change n accruals employng a same frm-year desgn for the subsample of nonfnancal frms. In ths model we wnsorze at the 1% level. 20 Recall that hgher values of DIFFACC REST proxy for more earnngs management and less rgorous frst-tme applcaton of IFRS. Regresson results are presented n Table 7. In Model (1), nvestor protecton rghts are postvely assocated wth restatement rgor (p = 0.004). Some control varables are sgnfcantly assocated wth restatement rgor such as frm sze (p=0.044) and sales growth (p=0.072). Model (2) ncludes corporate governance strength as an explanatory varable. As hypotheszed, the coeffcent on CORPGOV s negatve, but only margnally sgnfcant (p=0.081). Model (3) s a two-stage least squares formulaton whch uses the predcted value 20 We choose to wnsorze the data nstead of droppng extremes because the sample contans a lmted number of observatons. However, our results are not senstve to the ncluson or excluson of extreme observatons of DIFFACC REST. 21

22 of CORPGOV as an explanatory varable. Ths formulaton yelds a sgnfcantly negatve coeffcent for CORPGOV (p = 0.015). Insert Table 7 about here Assocaton between IFRS adopton qualty and pre-and post-adopton accrual qualty As explaned n Secton 2, we are nterested n explorng the assocaton between IFRS adopton qualty and pre- and post-adopton accrual qualty. Panel A of Table 8 regresses DIFFACC REST on ABNACC Local. In Panel B we regress ABNACC IFRS on DIFFACC REST. Abnormal accruals under local standards are measured usng the approach of Francs and Wang (2008). Explanatory varables are the same as n Table 7. Snce we do not make any drectonal predctons for ABNACC Local, we test the null hypotheses of no assocaton Panel A. We expect that frms applyng IFRS n a rgorous fashon ntally wll contnue to do so afterwards (.e. we expect a postve correlaton between ABNACC IFRS and DIFFACC REST ). In Table 8 we fnd that DIFFACC REST s not assocated wth ABNACC Local. 21 However, ABNACC IFRS s postvely assocated wth DIFFACC REST. These results are consstent wth the noton that IFRS adopton represents a watershed n fnancal reportng n whch frms wth low accrual qualty under local GAAP do not exhbt lower IFRS adopton qualty, but subsequent accrual qualty under IFRS s postvely assocated wth IFRS adopton qualty. We cauton, however, that ths second assocaton s characterzed by substantal randomness snce R 2 s low by conventonal standards. Corporate governance qualty was prevously found to be sgnfcantly assocated wth frst-tme IFRS adopton qualty (p=0.003, see Panel A). Ths effect dsappears, however, when relatng governance qualty to subsequent accrual qualty, although we stll have a negatve coeffcent (p=0.210). Insert Table 8 about here 21 In ths case the ncluson or excluson of outlers does have a sgnfcant effect on the estmated coeffcents. We decde to leave the top and bottom observaton out of the analyss. On top, one observaton dsappear because the frm ceased to exst n Ths results n a sample of 149 observatons. 22

23 Robustness checks DIFFACC REST can depend on the structure of local standards as well as on frm ncentves. If local standards dffer sgnfcantly from IFRS, dfferences n accruals are lkely to be larger. Therefore, we employ a two-stage approach. In the frst stage DIFFACC REST s regressed on the number of standards for whch local GAAP dffers from IFRS plus the number of standards omtted n local GAAP. Data are obtaned from a survey executed by audtng frms (GAAP, 2000 and 2001). In the second stage the resdual from the frst stage (.e. when the change n accruals not due to the dfference n standards) s regressed on governance qualty. Results reman qualtatvely the same. Further, we use legal orgn (Durnev and Km, 2005) and nvestor protecton rghts (IPR) as alternatve nstruments to deal wth the possble endogenety problem between accountng qualty and corporate governance. In both cases, results are smlar. Next, we nvestgate how the four categores of corporate governance qualty each explan varaton n restatement qualty. The governance ratngs contan 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 of corporate governance. Overall, we conclude that there s not one partcular aspect of governance that outperforms the others n explanng varaton across frms n terms of restatement qualty of IFRS. Wthn the thrd category, the board functonng ndcator appears to be the one provdng the most consstent results across the three restatement qualty tems, scorng better than audt commttee, board sze and board ndependence. A last robustness check s done by decomposng the dsclosure transparency nto voluntary and compulsory components. As expected, frm governance has an especally large mpact on voluntary dsclosure. However, for the fnancal frms, stronger governance s also assocated wth dsclosure of requred tems. 23

24 7. Concluson Ths study nvestgates factors assocated wth European frms IFRS adopton qualty, usng three proxes for adopton qualty: early applcaton of IAS39, transparency of transton-year dsclosure n fnancal statement notes, and rgor of ntal applcaton of IFRS n the ncome statement. Although the same frm-year desgn has been used n market-based studes, t has not been used to date n accountng-based studes measurng the mpact of mandatory IFRS transton on accountng qualty. We predct and document an mportant role of governance qualty (at the frm and country level), as frms have consderable dscreton n how they adopt IFRS for the frst tme. Further, we explore the assocatons between IFRS adopton qualty and frms accrual qualty before and after IFRS adopton. The nvestgaton s facltated by the characterstcs of the European nsttutonal envronment, vz., a two-ter system of acceptng IFRS standards, whch allows the possblty of carve-out optons n adoptng IFRS; concentrated ownershp structures; and consderable varaton n corporate governance strength across frms. The use of hand-collected data for the transton year 2004 enables us to nvestgate dfferences between local GAAP and IFRS accountng numbers, ndependent of changes n the economc and nformaton ncentves over tme. We hypothesze and fnd that better governed frms exhbt greater transparency of dsclosure and ntal applcaton rgor of IFRS, after controllng for other factors that are lkely to be assocated wth adopton qualty. However, governance strength s not unformly assocated wth early IAS 39 adopton. Rather, we fnd that well governed frms use adopton flexblty n a conservatve manner. If IAS 39 has a negatve mpact on equty book value or 24

25 leverage, these frms tend to adopt the standard early; when t has a postve mpact, they tend to postpone adopton. We document that frms IFRS adopton rgor s unassocated wth pre-ifrs accrual qualty. We fnd that post-ifrs accrual qualty s postvely assocated wth IFRS adopton rgor. These fndngs are consstent wth the noton that IFRS adopton represents a watershed n fnancal reportng n whch frms wth low accrual qualty under local GAAP do not exhbt lower IFRS adopton qualty. We conjecture that the reason for ths s that IFRS affords companes a fresh start wth regard to ther accountng qualty. Furthermore, we fnd that IFRS adopton has sgnfcantly mproved frms accountng qualty, both durng the transton perod as well as n the subsequent perod. Our fndngs add to the lterature that s consstent wth IFRS adopton makng fnancal statements more useful as hypotheszed n the IFRS conceptual framework. However, our results suggest that carve-out optons hnder comparablty under IFRS, snce frms used the IAS 39 carve-out n a conservatve way. Such conservatsm s nconsstent wth the conceptual framework s prncple that nformaton should be unbased to ensure relablty and neutralty (Barth, 2008). Cauton s warranted n nterpretng the results for a number of reasons. Frst, the external valdty of these fndngs could be questoned snce the sample comprses only large, lsted companes. Second, IFRS adopton qualty s measured for a sample of mandatory complers; pror lterature ndcates that mprovements n accountng qualty may dffer between voluntary and compulsory applers. However, usng a homogenous sample of large and compulsory adopters of IFRS s lkely to work aganst us fndng sgnfcant results. Thrd, whle measures for the early adopton of IAS39 and the rgor of IFRS applcaton can be reasonably well captured wth proxy varables, our dsclosure qualty measure only consders a lmted number of dsclosure tems. However, we attempt to measure dsclosure of the most 25

26 relevant accountng numbers. Fourth, the endogenous character of our governance varables s a potental lmtaton. The absence of well-specfed theory and the dffculty of dentfyng strctly exogenous nstruments make t dffcult to address endogenous assocatons. We mtgate these concerns by employng two-stage least squares procedures. Many of our results reman smlar to those wthout any adjustment for endogenety. Fnally, the study s done n the EU, where a unque, two-tered system of IFRS endorsement exsts. Future research could compare IFRS adopton qualty n dfferent nsttutonal envronments such as Canada, where the ntent s to adopt all IFRS standards wthout modfcaton. Despte these caveats, our results should be of nterest to regulators and polcy makers n many countres. The study suggests that IFRS adopton represents an mprovement n fnancal reportng snce we fnd no assocaton between pre-adopton accrual qualty and IFRS adopton qualty, but IFRS adopton qualty s postvely assocated wth post-adopton accrual qualty. Furthermore, we fnd that IFRS adopton qualty s postvely assocated wth governance strength at the frm-level. Hence, for countres that have not yet adopted IFRS, stronger governance gudelnes are lkely to promote hgher qualty adopton, and dsclosures of governance measures are lkely to be useful to analysts and other fnancal statement readers n nferrng the relablty of the ntal IFRS numbers. Our results also suggest that poltcal nvolvement n the IFRS adopton process decreases comparablty of fnancal nformaton snce frms are lkely to use carve-out optons opportunstcally or n ways that are nconsstent wth the IFRS Conceptual Framework. 26