Accountng Standards and Internatonal Portfolo Holdngs: Analyss of Cross-border Holdngs Followng Mandatory Adopton of IFRS Gwen Yu Unversty of Mchgan gowoonyu@bus.umch.edu Jan 2009 Abstract: Pror lterature shows that nvestors under-nvest n foregn frms due to nformaton asymmetry problems. I post that dfferences n local accountng standards are a source of the nformaton asymmetry among nvestors. Usng securty-level holdngs of nternatonal mutual funds, I fnd that harmonzng accountng standards (adopton of IFRS) ncreases foregn mutual fund holdngs. Harmonzng accountng standards ncreases cross-border holdngs 1) drectly by reducng the nformaton processng cost of foregn nvestors and 2) ndrectly by reducng the effect of other barrers on cross-border nvestments such as geographc dstance. Further analyss suggests that harmonzaton across regmes s a more effectve means to attract foregn captal than a unlateral mprovement n a country s reportng regme. I thank my dssertaton char, Russell Lundholm, for hs contnuous gudance and members of my dssertaton commttee, Ila Dchev, Greg Mller, Venky Nagar, Tyler Shumway, and Jng Zhang for ther thoughtful comments and suggestons. I also thank Phl Berger (Edtor), Bjorn Jorgensen, Nemt Shroff, Don Yu, an anonymous referee, and semnar partcpants at the Unversty of Mchgan for ther helpful feedback on the paper. I gratefully acknowledge fnancal support from the Center of Internatonal Busness and Educaton, the Delotte Foundaton, Harry H. Jones Endowment Fund, and the Paton Accountng Fellowshp. All errors are my own.
1. Introducton It s well establshed that nvestors are reluctant to hold securtes outsde ther domestc markets. The underweghtng of foregn nvestments, referred to as home bas, has been consstently observed across dfferent classes of nvestments and types of nvestors (Ahearne et al. 2004, Kang and Stulz 1997, and Lews 1999). The most common explanaton for equty home bas s that addtonal nformaton costs exst when nvestng abroad (Coval and Moskowtz 2001, Van Neuwerburgh and Veldkamp 2009). Ths paper examnes how local accountng standards, and the changes nduced by adoptng nternatonal accountng standards, nfluence the cross-border holdngs decsons of nvestors. Accountng nformaton consttutes one of the key nputs n portfolo nvestment decsons. Therefore, nformaton presented under dfferent accountng standards or practces s lkely to be a source of addtonal processng costs. Such processng costs consst of drect costs to learn dfferent accountng standards as well as ndrect costs arsng from the need to nterpret fnancal statements n lght of local practces. When accountng nformaton s presented under the same standards, nvestors fnd t easer to process the nformaton and are more lkely to rely on fnancal statements relatve to other prvate nformaton. The dffculty n examnng the effect of accountng nformaton on cross-border nvestment s that determnants of frm s accountng choces are commonly assocated wth other factors that affect nvestors holdng decsons. Thus, t s hard to attrbute the full extent of the observed assocaton to accountng standards alone rather than to other factors such as qualty of nsttutons or reportng ncentves. However, a natural experment recently occurred when 89 countres mandated the adopton of Internatonal Fnancal Reportng Standards (IFRS) for most, but not all, of ther local companes. I examne the effect of IFRS adopton on cross-border nvestment decsons usng securty- 1
level holdngs data of nternatonal mutual funds. The change n the regulatory envronment provdes a powerful settng to dentfy how lowerng nformaton cost affects nvestment decsons across borders. Furthermore, snce not all frms n a country were requred to adopt IFRS, and some have already adopted t, there s a natural control for country-level changes n non-nformaton-based motves for nternatonal portfolo holdngs. Whle the deal research desgn would be to have frms randomly assgned to the adopter and non-adopter group, such a natural experment s not possble. Hence, I explot the varaton n the captal sources wthn an adoptng frm and examne the changes n frm-level holdngs from varous countres followng IFRS adopton. By focusng on changes n the holdngs from dfferent countres, and relatng t to the varyng degree of harmonzaton trggered by the IFRS adopton, the research desgn effectvely uses nvestors from other countres nvestng n the same frm as a control. Hence, I control for changes n frm-level factors that affect all nvestors equally. I frst start by examnng changes n holdngs of domestc and foregn nvestors to establsh the aggregate holdngs pattern followng IFRS adopton. Consstent wth accountng nformaton lowerng nformaton processng costs, I fnd that foregn mutual funds ncrease ther holdngs by 2.0% of total outstandng shares for frst-tme IFRS adopters relatve to frms that were not requred to adopt IFRS. 1 Whle there was also a general ncrease n domestc mutual fund holdngs durng ths perod, the ncrease n ownershp by foregn funds s hgher for the mandatory adopters than the ncrease n ownershp by domestc funds. Consstent wth ths, frms that were not requred to adopt IFRS show almost no change (0.3%) n ownershp by foregn funds, yet show a meanngful ncrease (2.2%) n the 1 Untabulated results show that ths fndng s robust to alternatve control groups (.e., non-adoptng frms matched on sze and frms from countres that dd not adopt IFRS). Also, frms that voluntarly adopt IFRS before the year of mandate show an ncrease n foregn mutual fund holdngs (2.8%) smlar to that of the mandatory adopters n the year of the IFRS mandate. Ths s consstent wth mandatory IFRS adopton ncreasng comparablty even for frms that were not drectly affected by the mandate (Barth et al. 2009, DeFond et al. 2009). 2
ownershp of domestc funds. Taken together, these fndngs suggest that fund managers ncreased ther holdngs of foregn securtes followng IFRS adopton because of the reducton n nformaton processng costs, and they decreased holdngs of domestc securtes because IFRS reduced ther local nformaton advantage. I explore two channels through whch harmonzng accountng standards ncreases foregn holdngs. Frst, I examne whether the adopton of IFRS ncreases cross-border holdngs by drectly reducng the nformaton processng cost of foregn nvestors. Usng measures of nternatonal GAAP dfferences (Bae, Tan, and Welker 2008), I fnd that dfferences between the local accountng standards of the reportng frm and the local accountng standards of the nvestng fund (.e., accountng dstance) can explan the country varaton n the holdngs of each frm. Specfcally, I fnd that frms have dmnshng ownershp from funds n countres wth greater accountng dstance. Furthermore, I fnd that ncrease n holdngs followng adopton of IFRS s drven by captal flows from countres where IFRS trggered a large reducton n the dfferences between the nvestng country s and the adoptng country s accountng standards. Second, I examne whether harmonzng accountng standards ncreases cross-border holdngs by reducng the effect of other nformaton barrers. Pror lterature fnds many other barrers to crossborder nvestments that are unrelated to accountng standards (.e. geographc dstance, economc dstance, and cultural dstance). Many of these barrers arse from a general lack of famlarty, and thus have an nformaton asymmetry component (Ahearne et al. 2004, Portes and Rey 2005). 2 Consequently, a natural queston s to what extent harmonzng accountng standards can reduce the mpact of other nformaton barrers. For example, harmonzng accountng standards can reduce the effect of geographc dstance by nducng nvestors to rely less on prvate nformaton sources. 2 For example, geographc dstance has been nterpreted as a measure of nformaton acquston cost or the lkelhood of nvestors sufferng nformaton asymmetry (Portes and Rey 2005). 3
Identfyng the margnal effect of other nformaton barrers followng IFRS adopton, however, presents a challenge snce dfferent nformaton barrers share common varatons,.e., accountng standards are lkely to be more dfferent among dstant countres that have less nformaton flows. Hence, smply comparng the mpact of nformaton barrers before and after IFRS adopton wll yeld nconclusve results, especally f the global IFRS adopton was clustered among countres wth less nformaton flows. Therefore, I hold the level of accountng harmonzaton constant by comparng the changes n holdngs of nvestors who experenced equal reducton n accountng dstance but faced varyng level of nformaton barrers. That s, I compare the effect a unt reducton of accountng dstance had on holdngs of nvestors wth hgh nformaton barrers and low nformaton barrers wthn an adoptng frm. I fnd that harmonzng accountng standards has a stronger effect on captal flows among countres wth hgh nformaton barrers,.e., dstant countres and countres wth less phone traffc. Ths ndcates that reducng accountng dstance has an nteractve effect of reducng the effect of other nformaton barrers, perhaps because better knowledge of accountng standards makes prvate nformaton less necessary. That s, harmonzng accountng standards encourages nvestors to seek remote nvestment opportuntes that they would not have pursued f the nformaton were presented under dfferent accountng standards. Fnally, I ask whether reducng the common nose wthn the boundares of each accountng standard can reduce the effect of accountng barrers. 3 In partcular, I examne whether the effect of accountng dstance on cross-border holdngs dffers by level of common nose shared across all nvestors. Usng analyst forecast errors as the measure of common nose n the accountng nformaton, I fnd that dfferences n accountng standards mpede cross-border nvestments more for frms wth less common nose n the accountng nformaton. I show that ths s consstent wth an 3 Common nose s defned as nose n the accountng nformaton sgnal that has common varance for both domestc and foregn nvestors. Refer to secton 2.2 for detals. 4
equlbrum model of asset demand where all nvestors observe a publc sgnal wth common nose but have dfferng abltes to nterpret the nformaton. The model shows that when there s less common nose shared across all nvestors, the weght of nose from accountng dstance ncreases, makng foregn nvestors more senstve to ther nformaton dsadvantage. As a result, when there s less common nose n the accountng nformaton, dfferences n accountng standards become a bndng constrant for foregn nvestors who are less knowledgeable about the local standards. Hence, when foregn nvestors have lmted understandng of the local accountng standards, reducng common nose n the local accountng standards can exacerbate the nformaton dsadvantage of the foregn nvestors. Ths suggests that harmonzaton across regmes s a more effectve means to attract remote foregn captal than a unlateral mprovement n a country s reportng regme. Ths paper contrbutes to the lterature by drectly showng how frctons from accountng standards nfluence cross border holdngs. A strand of emprcal studes shows that proxes of nformaton cost can explan a large porton of the varaton n cross-border holdngs (Obstfeld and Rogoff 2000, Portes and Rey 2005). However, the exact nature of the nformaton costs that causes the asymmetry among nvestors has receved less attenton. I show that dfferences n accountng standards create barrers to cross-border nvestments even for mutual funds that are among the most sophstcated users of fnancal nformaton (Benesh and Yohn 2008). My fndngs also suggest a possble explanaton for the economc beneft of adoptng IFRS. Pror lterature shows that the beneft of IFRS stems from both the standards beng of hgher qualty (Armstrong et al. 2007 and Barth et al. 2008) as well as from accountng nformaton beng presented n a form that s more famlar to nvestors (Bradshaw et al. 2004, Aggarwal et al. 2005). 4 In ths paper, I focus on how adoptng IFRS changes the mx of nvestors n each frm. By focusng on changes n 4 Prevous studes refer to the former as the nformaton beneft and the later as the harmonzaton beneft. 5
holdngs of nvestors from dfferent countres, I control for frm level factors that affect all nvestors equally, such as an overall mprovement n nformaton qualty. Ths s n contrast to most prevous studes that examne the effect of IFRS adopton from an average nvestor s perspectve (Armstrong et al. 2007, Daske et al. 2008). Thus, my analyss provdes drect evdence on how harmonzng accountng standards affects nvestment decsons by reducng nformaton asymmetry among nvestors. The rest of the paper s organzed as follows. In secton 2, I revew the lterature and develop the major hypotheses based on a smple model of equlbrum asset demand under asymmetrc nformaton. In secton 3, I descrbe the data and present a smple analyss of aggregate foregn holdngs. In secton 4, I emprcally test how harmonzng accountng standards changes cross-border holdngs of nternatonal mutual funds from dfferent countres. In secton 5, I test the robustness of the results, and conclude n secton 6. 2. Background and hypotheses development 2.1 Accountng nformaton and cross-border nvestment decsons Pror lterature documents that cross-border nvestment decsons are related to nternatonal dfferences n accountng standards. Ths relatonshp s observed because foregn nvestors have preferences towards nformaton presented n a famlar form when makng cross-border nvestment decsons (Leuz and Verreccha 2000). Barth, Clnch, and Shbano (1999) analytcally show that foregn nvestors prefer harmonzed accountng standards because t reduces the costs of acqurng expertse comparable to local nvestors. In addton, Leuz, Lns, and Warnock (2009) show that foregn nvestors prefer nvestng n countres wth hgher-qualty dsclosure rules because local nvestors are less lkely to have access to prvate nformaton n such countres. However, when nformaton s presented under local accountng standards, foregn nvestors fnd t costly to process the nformaton, leadng to equty home bas. Taken together, the lterature argues that foregn 6
nvestors prefer nternatonal standards because t reduces ther nformaton dsadvantage relatve to the local nvestors. Recently, the IASB has promoted adopton of IFRS around the world clamng that one of the motvatons for countres to adopt IFRS s to attract foregn captal flows by reducng nformaton barrers. However, there are skeptcal vews on whether adoptng an nternatonal accountng standard tself wll brng sgnfcant changes n nvestment decsons. Accountng standards are complementary to the nsttutonal envronment, thus, t s unclear whether adoptng nternatonal accountng standards wll provde better nformaton to nvestors n all countres (Holthausen 2003). In addton, gven concerns around the mplementaton and enforcement of standards, crtcs clam that the beneft of mandatng IFRS s not mmedately obvous (Ball et al. 2003, Holthausen and Watts 2000). Nonetheless, several studes show that adoptng nternatonal accountng standards affects nvestors nvestment decsons. Adoptng IFRS changes nvestors belefs because the new standards requre ncreased level of dsclosure (Barth et al. 2008). And even f the level of dsclosure remans unchanged, adopton of new accountng standards can affect nvestors by ncreasng comparablty across countres (Bradshaw et al. 2004, DeFond et al. 2009). A recent survey of 187 nternatonal mutual fund managers n the European Unon fnds that 72% of the surveyed fund managers consder adopton of IFRS to have nfluenced nvestment decsons, and nearly half of the 72% state that IFRS changed ther holdngs decsons. 5 Fund managers consder the key beneft of IFRS to be the ncreased level of consstency n fnancal reportng between jursdcton and sectors, suggestng that ncreased comparablty from the adopton of IFRS s sgnfcant enough to alter nvestment decsons. Another channel through whch adoptng nternatonal accountng standards can affect nvestment decsons s by ncreasng vsblty of securtes n less famlar countres. Investors often 5 PrcewaterhouseCoopers MORI Survey (2006) IFRS, the European nvestors vew. 7
overlook certan groups of securtes wth low vsblty (Merton 1987, Lehavy and Sloan 2008, and Bushee and Mller 2009). Usng harmonzed accountng standards can be one way to ncrease vsblty of remote nvestments that have been overlooked. Consstent wth ths argument, studes show how adoptng IFRS ncreases prvate equty nvestment (Cummng and Johan 2007), foregn drect nvestment (Marquez-Ramos 2008, L and Shroff 2009), nsttutonal nvestment (Florou and Pope 2009), and tradng actvtes of ndvdual nvestors (Brüggemann et al. 2009). However, a common crtcsm of these fndngs s that t s dffcult to dstngush whether the observed beneft s from ncreased famlarty or smply from IFRS provdng hgher-qualty nformaton. In the followng secton, I address ths queston usng a model of equlbrum asset demand. 2.2 Hypothess development To gude my emprcal work, I present a smple characterzaton of the equlbrum asset demand when two nvestors, domestc and foregn, observe publc sgnals but have dfferng abltes to nterpret the nformaton. I study how the demand of the two nvestors s affected by changes n the nformaton structure for a gven frm. Each nvestor j s endowed wth shares n two rsky assets, one domestc and one foregn, and a sngle global rskless asset, wth a normalzed prce of one. 6 V s the value of a rsky asset n each country ( =1, 2). The termnal payoffs of both rsky assets ( V, =1, 2) are ndependent and normally dstrbuted wth mean and varance f, 1 f1 0 V, 0 f 2 2. 6 Thus, the model assumes that any exchange rsk or purchasng power rsk can be completely hedged. 8
Indvdual j s wealth s the sum of ther exogenous endowment j W0 and the net returns (V P ) from holdng domestc each country. That s, (D ) and foregn j j j (D j) rsky assets, where P s the prce of the rsky asset n W W D (V P ) D (V P ). (1) j j j j 0 1 1 1 2 2 2 Each nvestor n country j receves nformaton j (Y ) about the value of each asset, whch s used to derve the demand for each asset. The crtcal assumpton s that the precson of the nformaton s asymmetrc for domestc and foregn nvestors. In partcular, the nose n nvestors nformaton has j two components: a common nose component ( ) for all nvestors, and an dosyncratc nose j component( ) for foregn nvestors only. 7 If = j (domestc nvestor) Y V, where N(0, ), j j j and f j (foregn nvestor) Y V, where N(0, ). j j j j j j represents the common nose shared by domestc and foregn nvestors, and the precson depends only on the country where the asset s located. Ths nose s due to an mperfect fnancal reportng envronment arsng from management opportunsm or weak enforcement. j represents the addtonal nose for nvestors when they nvest n foregn assets, where the precson depends on the locaton of the nvestors. Ths addtonal nose can arse from nvestors havng a murky understandng of the busness n foregn countres, but also from ther lmted ablty to nterpret what the accountng nformaton mples about the termnal asset value V. 8 It should be noted that my framework 7 An alternatve specfcaton s to nclude dosyncratc nose for the domestc nvestors that have lower varance than the dosyncratc nose of foregn nvestors. As long as the dosyncratc nose of the domestc and foregn nvestors are ndependent, the specfcaton above s a reduced form of ths alternatve specfcaton, hence yelds dentcal mplcatons. 8 Ths s a common assumpton n the analytcal models of home bas where the nformaton dsadvantage of foregn nvestors s exogenously gven. Gehrg (1993) presents a two-country model where domestc nvestors are endowed wth superor prvate nformaton and shows that ths leads to undernvestment n foregn assets. Gordon and Bovenberg (1996) endogenze the nformaton dsadvantage of foregn nvestors by allowng them the opton to acqure costly nformaton. As long as domestc nvestors are endowed wth an ntal nformaton advantage, there s undernvestment of foregn nvestment n equlbrum. 9
devates from ratonal expectaton models because foregn nvestors suffer from ths nformatonal dsadvantage, yet do not use prce to nfer the sgnals of domestc nvestors. 9 That s, there exsts some cost to arbtrage (Shlefer 2000) or lmted attenton (Hrshlefer 2001) for foregn nvestors that causes them to not attend to the nformaton mpounded n prce. Wth ths framework, I can derve the nvestor s demand functon by maxmzng the expected condtonal utlty j j E UW Y. Appendx A derves the demand functon of domestc and foregn nvestors assumng a negatve exponental utlty wth a rsk averson factor of one. For a gven sgnal realzaton the demand functon smply states that each nvestor compares hs expected value to prce and weghts the dfference by hs posteror precson. To characterze the equlbrum demand as a functon of the nformaton structure, I determne the market-clearng prce by equatng aggregate demand to aggregate supply, Z ( 1,2), and then substtute ths equlbrum prce back nto the demand functon. 10 Averagng over sgnal realzatons, Appendx A derves the equlbrum expected demand for domestc and foregn nvestors gven as: domestc nvestors demand for asset 1 s and foregn nvestors demand for asset 1 s 1 1 1 f1 1 E D1 Z 1 (2) 1 1 1 2 1 (f 1 1 ) f1 1 1 2 1 2 f 1 ( 1 ) E D1 Z 1. (3) 1 1 1 2 1 (f 1 1 ) f1 1 Fgure 1 llustrates how the equlbrum demand of domestc and foregn nvestor s affected by varyng levels of dosyncratc nose and common nose. Two major mplcatons can be drawn from the demand functons. Frst, holdng all else constant, equlbrum demand of foregn nvestors s a 9 Smlar mplcatons can be drawn from standard nosy ratonal expectaton models, where nvestors use prce to nfer nformaton about the domestc nvestors. As long as there s a random nose component that prevents prce from perfectly revealng, dentcal mplcatons can be drawn for the equlbrum demand. See Gehrg (1998). 10 In other words, the equlbrum demand s derved from the ndrect demand functon wth market-clearng prce. 10
decreasng functon of the dosyncratc nose of foregn nvestors j, llustrated n Panel A. Ths also mples that domestc nvestors demand s ncreasng n the dosyncratc nose of foregn nvestors. Thus, f dfferences n accountng standards add addtonal nose to foregn nvestors nformaton sgnals, ths lowers the demand for foregn securtes consstent wth equty home bas. Ths leads to the followng frst hypothess, H1: Frms wll have dmnshng ownershp from funds n countres wth greater accountng dstance (.e., dfferences n the local accountng standards of the reportng frm and the nvestng fund). The second observaton from equatons (2) and (3) s that a lower level of common nose wll cause nvestor s demand to be more senstve to the dosyncratc nose j. Panel B of Fgure 1 shows that dfference n the equlbrum demand of domestc and foregn nvestors ncreases as the common nose shared across all nvestors s reduced. Ths occurs because the holdngs of nvestors are determned by the relatve weght of the dosyncratc nose and the common nose. As the common nose decreases, the relatve porton of the dosyncratc nose becomes larger, makng the demand of nvestors more senstve to the dosyncratc nose. As long as domestc nvestors contnue to have an advantage over foregn nvestors ( j >0), smply mprovng the nformaton qualty of all nvestors (.e. reducng ) wll exacerbate the nformaton asymmetry among nvestors. Ths also mples that n the cross-secton, frms wth low common nose n the accountng nformaton (.e. lower ) wll have more undernvestment from dfferences n accountng standards than frms wth hgh common nose. H2: Undernvestment from funds n countres wth greater accountng dstance wll be more severe for frms wth less common nose n ther accountng nformaton than for frms wth more common nose. Fnally, I examne how accountng harmonzaton (.e., adopton of IFRS) wll affect cross-border holdngs decsons of dfferent nvestors. Adopton of IFRS wll reduce the dosyncratc nose for 11
nvestors that are more famlar wth IFRS. 11 More broadly, adoptng IFRS wll have a greater effect on foregn nvestors who experence a greater reducton n the dosyncratc nose ( j ) followng the adopton. Hence, I expect frms adoptng IFRS to experence greater ncrease n holdngs from countres where dfferences n accountng standards reduced the most (a greater change n j ) followng IFRS adopton. Relatng changes n accountng dstance to the changes n holdngs s a drect examnaton of the downward-slopng demand n Fgure 1, Panel A. H3: Holdng common nose constant, frms adoptng IFRS wll have greater ncrease n holdngs from countres where adopton of IFRS resulted n greater reducton n accountng dstance (greater change n j ). nose Note that IFRS adopton s lkely to reduce the common nose ( ) as well as the dosyncratc j ( ) n the accountng sgnals. 12 Furthermore, Panel B of Fgure 1 shows that reducng accountng dstance (.e., dosyncratc nose) has a greater effect on foregn nvestors demand when there s less common nose shared across all nvestors. Whle ths mples that the relatonshp n H3 s lkely to be stronger for frms that have less common nose n ther accountng nformaton, the predcton of H3 between hgh and low common nose frms s unclear because the level of common nose s lkely to change followng IFRS adopton. Snce I do not know the resultng level of common nose followng IFRS adopton, I have no dfferental predcton of H3 for each group of hgh and low common nose frms. In the emprcal tests that follow, I examne these hypotheses usng cross-border holdngs of nternatonal mutual funds as a proxy for nvestors demand from dfferent countres. 11 Pror lterature refers to ths as the harmonzaton beneft of adoptng IFRS. (Barth et al. 1999) 12 IFRS wll reduce the common nose n accountng nformaton f nternatonal accountng standards lead to ncreased dsclosure wth better qualty. Pror lterature refers to ths as the nformaton beneft (Barth et al. 2008). 12
3. Sample selecton and analyss of aggregatge foregn holdngs 3.1 Mandatory adopton of IFRS I determne the year of country-level mandatory adopton of IFRS from Internatonal Accountng Standards Plus, an annual newsletter publshed by Delotte Touche Tohmatsu, and the country reports of the Unted Natons Economc and Socal Councl (ECOSOC). Frm-level adopton dates are collected from Thompson Datastream. Snce these establshed databases are known to have codng errors (Daske et al. 2008), I cross-check the subset of all non-adoptng frms wth two other databases, Amadeus and Orbs, both provded by Bureau van Djk Electronc Publshng. For the 86 frms that seem to show dscrepances across databases, I manually search ther fnancal statements and ndvdually code the adopton years. At the country level, the IAS board has made sgnfcant progress n promotng IFRS worldwde, although there s stll consderable dvergence across countres n the rate and the extent of the adopton. For ths study I classfy all countres that have endorsed the full verson of IFRS (.e., the EU countres) and countres that chose to gradually adjust natonal standards n lne wth IFRS (.e., Australa, Hong Kong, and New Zealand) as mandatory adopters. 13 Ths results n an ntal sample of 89 countres. Elmnatng countres wth mssng fnancal data n Thompson Datastream reduces the fnal sample to 28 countres. 14 Table 1 shows the percentage of frms adoptng IFRS n the 28 countres from 2000 to 2007. Wthn each country, the adopton process also exhbts consderable heterogenety across frms. Two fndngs stand out from Table 1. Frst, a sgnfcant number of frms voluntarly adopted IFRS before 13 Countres that have selectvely adopted only a subset of the standards (.e., Inda, Malaysa, and Thaland) are not ncluded n the sample. 14 Frm-level fnancal data s often mssng for the emergng countres, especally n the Afrcan contnent and the former Republcs of the Sovet Unon. 13
t was mandated. 15 Second, even after the year of mandate there are frms who stll report under the local standards. Ths s because mandatory adopton of IFRS was lmted to consoldated accounts. Companes wthout consoldated accounts or those that qualfy for the small-medum entty exempton contnue to report under local accountng standards. The defnton of a small-medum entty vares by each country s legslaton and s based on the nature of the entty rather than on ts sze (PrcewaterhouseCoopers, IFRS for SMEs Pocket Gude 2007). For example, Aero Inventory Plc., a UK company wth total assets over USD 800 mllon, qualfed for the small-medum entty exempton because t was lsted on the Alternatve Investment Market (AIM). 16 Also, Sad S.P.A, one of the largest constructon companes n Italy, dd not report under IFRS n year 2005 because t was not requred to report consoldated fnancal statements. 17 Both voluntary adopters and non-adopters serve as useful control samples. Due to ths varaton n the adopton process of IFRS, I dvde the sample nto three groups: frms that adopted early, frms that adopted n the year of mandate, and frms that dd not adopt durng or after the year of mandatory adopton. Emprcally, I categorze frms that adopted IFRS n fscal years before IFRS was requred as voluntary adopters. 18 Mandatory adopters are frms that frst adopt IFRS wthn two years after IFRS was mandated at the country level. 19 All frms that dd not adopt IFRS by the end of the sample perod (2007) are categorzed as non-adopters. To ensure contnuty, I delete 15 Early adopton was permtted n many countres before the year of mandatory adopton. For example, Germany and Austra have allowed frms to adopt for ther consoldated accounts as early as 1998. For countres lke the UK where early adopton was not permtted, frms had the opton to cross-lst ther shares and report consoldated accounts under other nternatonal standards. 16 Frms lsted on the Alternatve Investment Market (AIM) were requred to adopt IFRS startng after January 2007. 17 In 2007, Sad S.P.A merged as Sad Servz Industral Group and started reportng ts consoldated fnancal statements under IFRS. 18 Followng pror lterature (Daske et al. 2008), I also dstngush early voluntary adopters that adopt before the announcement of the mandatory adopton and late voluntary adopters that adopt after the announcement date. Untabulated results show that there are no observable dfferences n the holdngs of the two types of voluntary adopters. 19 For countres that have decded to adjust natonal standards to be n lne wth IFRS (.e., Australa, New Zealand, and Phlppnes), all frms are consdered to have mandatorly adopted IFRS n the year of mandate. 14
frms that delst before IFRS became mandatory at the country level and frms that exst only after the mandatory adopton. Also, frms cross-lsted n the US are excluded to ensure that IFRS s the only nternatonal standard n consderaton. After mergng wth the nternatonal mutual fund database, the fnal sample conssts of 4,744 frms (662 voluntary adopters, 3,800 mandatory adopters, and 282 non-adopters) across 28 countres. Table 2 compares the characterstcs of the dfferent types of adopters over years pror to IFRS adopton. Panel A shows that voluntary adopters are on average larger and more proftable. The medan sze of the mandatory adopters and non-adopters are comparable, but the dfference n the mean suggests that there are large outlers n the upper tal for mandatory adopters. Although sze was one reason non-adopters were exempt from the mandatory requrement, the non-adopters n my sample stll nclude many large frms because I requre frms to have holdngs from nternatonal mutual funds to be ncluded n the sample. In the emprcal tests of aggregate foregn holdngs, I examne changes n holdngs of mandatory adopters usng both non-adopters and voluntary adopters as benchmark groups. The regresson analyss ncludes varous frm-level and country-level controls to account for determnants of how regulaton and a frm s reportng choce sorted the adoptng and non-adoptng frms. In the man emprcal tests of Secton 4, I focus on the mandatory adopters and explot the varaton n the captal sources by breakng down the aggregate foregn holdngs by each country. Ths specfcaton effectvely uses nvestors from other countres nvestng n the same frm as the control group. 3.2 Cross-border holdngs of nternatonal mutual funds I examne how accountng standards affect cross-border holdngs usng securty-level holdngs data of nternatonal mutual funds. Securty-level holdngs data are compled by Thompson Fnancal Securtes (TFS) and contans cross-border mutual fund holdngs n 39 countres. Earler years of the 15
database were used n several studes aggregated at the country level (Chan et al. 2005, Hau and Rey 2008). In contrast, I use holdngs at the securty level to account for the varaton n the IFRS adopton process wthn each country. The holdngs cover an eght-year perod from 2000 to 2007. Snce funds have dfferent reportng frequences, I undertake my analyss on an annual bass usng the latest avalable reported holdngs for each calendar year. 20 In the TFS mutual fund database, holdngs of funds n dfferent countres are avalable at the securty level. The securtes held by the funds cover a wde range of countres around the world. For each securty, I compute aggregate holdngs held by all mutual funds n the latest reportng perod for each year. Thus, mutual fund demand n country j s measured as the aggregate percentage shares held by all funds n country j for a gven frm. An mportant assumpton underlyng my analyss s that the country of a fund s ncorporaton represents where the nvestor s captal orgnates. Albet ndrect, I argue that a fund s country of ncorporaton characterzes the nvestment decsons of local nvestors for two reasons. Frst, legal restrctons requre mutual funds to rase captal prmarly from domestc nvestors. For example, Fdelty holds afflates across the world but restrcts captal nflows to the local regons where each afflate operates. Purchase orders of non-resdents outsde the regon are not accepted for tax and regulatory reasons. 21 Wth the excepton of offshore funds, whch are excluded from the sample, captal nflows to each fund are restrcted to regonal nvestors. 22 Second, the nvestment behavor of a 20 About one-half of the funds report ther holdngs on a sem-annual bass, and a thrd report on a quarterly bass. 21 Smlarly, US nvestors cannot purchase mutual funds ssued n another country unless the funds are regstered wth the SEC. The only excepton would be for mutual funds that ssue prvately to fewer than 100 people or ssue only to sophstcated nvestors. Hedge funds, for example, crcumvent the mandatory reportng requrements by qualfyng for ths exempton. 22 Offshore funds are defned as funds ncorporated n offshore fnancal centers as defned n the IMF s year 2000 Report on Offshore Fnancal Centers. Whether fnancal centers also qualfy as offshore funds s dsputed n the lterature. In ths study, I do not consder major fnancal centers (e.g., London, Hong Kong, and Sngapore) as offshore, snce they play a major role n processng nformaton n the captal market (Gehrg 1998). 16
partcular mutual fund reflects the preference of local nvestors who have delegated the nvestment decson to the fund managers. Investors select from a varety of funds whch dffer n nvestment strategy and geographc focus. Thus, holdngs of partcular funds reflect the revealed preference of local nvestors that nvest n each fund. Table 3 shows the descrptve statstcs of fund holdngs over the sample perod 2000 to 2007 for the sample frms n ths study. Panel A shows that the total percentage of shares held by mutual funds s 9.9 % for an average frm, whch s slghtly lower than the fndngs from mutual fund holdngs n US securtes (Falkensten 1996). Domestc funds hold 6.5% of these shares, and the remanng 3.5% s held by foregn funds. Ths drectly reflects the reluctance of mutual funds to nvest n foregn securtes. Panel B of Table 2 examnes the dstrbuton of foregn and domestc holdngs by dfferent types of IFRS adopters. Panel B clearly shows that the level of foregn ownershp s related to a frm s tendency to adopt IFRS. For non-adopters, only 1.3% of the 12.6% shares held by mutual funds are held by foregn funds. On the other hand, for the voluntary adopters more than half (5.0%) of the 9.5% mutual fund shares are held by foregn funds, far above the 3.4% for mandatory adopters. In the followng secton, I test how the adopton of IFRS affected the holdngs of dfferent type of adopters. 3.3 Multvarate analyss of aggregate holdngs In ths secton, I examne the changes n aggregate holdngs for the mandatory adopters relatve to the voluntary adopters and the non-adopters usng the followng dfference-n-dfference model, Holdngs D D D D D D,t 0 0 Mandatory adopter Post mandate 0 Non-adopter 1 Non-adopter Post mandate 0 Voluntary adopter N 1 Voluntary adopter Post mandate 0,n n=1 D D country control frm control+. M m=1 1,m,t (4) 17
Holdngs.t s the percentage of total outstandng shares held by mutual funds for frm n year t. DMandatory adopter s an ndcator varable set to one f a frm s a mandatory adopter. DPost mandate ndcator varable that takes a value of one for fscal years endng on or after the mandatory adopton. The 0 coeffcent s the man term of nterest, whch measures the average ncrease n fund holdngs s an of mandatory adopters followng IFRS adopton. D Voluntary adopter s an ndcator varable set to one f a frm s a voluntary adopter, and zero otherwse. Also, D Non-adopter s an ndcator varable set to one f a frm s a non-adopter. The ncrease n the holdngs of mandatory adopters relatve to the ncrease of non-(voluntary) adopters can be examned by comparng the 0 coeffcent to 1 ( 1 ). Ths dfference-n-dfference desgn mplctly controls for any dfferental holdngs n the years followng IFRS adopton that affected holdngs of all frms and for any dfferental n holdngs among dfferent types of adopton frms not attrbutable to IFRS. Nonetheless, I also nclude a rch set of country and frm level controls from pror lterature to account for determnants of cross-border holdngs that can dfferentally affect dfferent type of adopters. Country-level controls nclude both macroeconomc factors and polcy factors that affect decsons to nvest n a certan country. Specfcally, I nclude market captalzaton/gdp (Market cap), Market return (Mkt return), GDP (GDP) and GDP growth (Growth) to capture macroeconomc performance. I also nclude wthholdng tax rate of dvdends (Wthholdng tax), market turnover (Turnover) and exchange rate regme (Exchange) to account for polcy factors that affect transacton costs of nvestng abroad. Wthholdng tax measures the drect transacton cost arsng from captal taxes (Graham et al. 2008), Turnover s a proxy for tradng actvtes that measures ndrect cost from barrers to arbtrage (Bloun et al. 2009). Exchange represents the stablty of the local currency and s an ndcator (0 5) that takes a hgher value f the currency s free floatng (Renhart and Rogoff 2003). Fnally, I nclude measures of the qualty of the nsttutons (e.g., legal orgn (Code law), ant-drector law ndex (SHrght) and rule of law ndex (Enforce)), whch 18
have been shown to promote foregn nvestments (La Porta et al. 1998). Detaled defnton and sources of each control varable are descrbed n Table 5, Panel A. Snce the estmaton of equaton (4) uses securty-level holdngs, I also nclude frm-level attrbutes shown to affect cross-border holdngs n pror lterature (Aggarwal et al. 2005, Barth et al. 2008). Frm-level controls nclude measures of frm sze (sze, # Analysts), performance (ROA, ROE, Dv. yeld), and growth potental (Leverage, MB, and PE). I also nclude governance measures to control for the preference of foregn nvestors towards well-governed frms (Leuz et al. 2009),.e., an ndcator varable for frms that are audted by a bg fve audt frm (Bg 5 audt) and the percentage of float shares avalable to ordnary nvestors (Free float). 23 Table 4 presents the results of estmatng equaton (4). I predct that f adoptng IFRS lowers the nformaton-processng cost of foregn nvestments, the changes n foregn ownershp wll be hgher for the mandatory adopters than for the non-adopters or the voluntary adopters. Coeffcents show that the average foregn holdngs of mandatory adopters ncrease from 2.7% ( 0 ) to 5.0% ( 0 0) followng adopton of IFRS. Ths s n contrast to the average foregn holdngs of non-adoptng frms, whch changes from 1.2% ( 0 0 ) to 1.23% ( 0 0 1) durng the same tme perod. Statstcal sgnfcance of all coeffcents are assessed after correctng for cross-sectonal and tme-seres dependence by clusterng standard errors on both year and country-ndustry. 24 Also, each model s estmated wth ndustry dummes to control for mutual funds preference for certan ndustres. In Table 4, I estmate equaton (4) both wth and wthout the frm-level and country-level controls 23 A large number of observatons are mssng for these two governance varables. Frms wth mssng audtor nformaton are assumed to have an ndcator value of zero. Frms wth mssng float data are assumed to have all shares avalable to ordnary nvestors. 24 Estmatng equaton (4) wth unbalanced number of frm observatons across dfferent countres can result n excessve weghts beng placed on large countres that cannot be completely corrected by clusterng the standard errors. To address ths concern, secton 5 tests the senstvty of the results usng the country-mean as the unt of observaton. 19
descrbed earler. Includng the controls yeld smlar coeffcent estmates, wth slghtly reduced sgnfcance. Thus, hereafter I only dscuss the estmaton results wthout ncludng the controls. Interestngly, the changes n the holdngs of voluntary adopters are very smlar to the changes of mandatory adopters. Coeffcents show that the average foregn holdngs of voluntary adopters ncrease from 4.1% ( 0 0) to 6.9% ( 0 0 1) followng adopton of IFRS. Although the voluntary adopters were not drectly affected by the mandate, these frms experenced an ncrease n foregn holdngs of mandatory adopters. A possble explanaton s that these frms also beneft from ncreased comparablty (DeFond et al. 2009). 25 Ths suggests that n addton to IFRS reducng the nformaton processng cost arsng from dfferent local standards, mprovng comparablty s another mportant channel through whch adoptng IFRS reduces the nformaton processng cost of nvestors. Fgure 2 graphcally llustrates the average changes n the foregn and domestc holdngs estmated from equaton (4). Panel A shows that percentage of foregn holdngs ncreases by 2.3% ( 0 from Table 4, t-stat=3.96) for mandatory adopters whle non-adopters have only a 0.3% ncrease ( 0 from Table 4, t-stat=1.36) durng the same perod. Ths s n stark contrast to the ncrease n domestc holdngs as shown n Panel B, whch shows a smaller ncrease for the mandatory adopters ( 0 =1.0%, t- stat=3.96) than for the non-adopters ( 0 =2.2%, t-stat=5.74). As n Table 4, the changes n the foregn and domestc holdngs of voluntary adopters are very smlar to the changes of mandatory adopters. Taken together, the dfferences n the changes of foregn and domestc ownershp suggest that fund managers have ncreased ther holdngs of foregn securtes after the adopton of IFRS, and have reduced holdngs of domestc securtes where IFRS reduced ther nformaton advantage. 25 Consstent wth ths, untabulated results show that the ncreases n holdngs of these voluntary adopters are drven by funds that nvest prmarly (.e., more than 50% of ther asset-under-management) n countres that mandated IFRS. 20
4. Analyss of the effect of accountng dstance on holdngs from dfferent countres 4.1 Accountng dstance and other determnants of blateral holdngs In the prevous secton, I examned changes n aggregate holdngs followng IFRS adopton. In ths secton, I explore the country-varaton n the fund holdngs by breakng down the foregn holdngs nto holdngs from dfferent countres. Usng a par-wse measure of dfferences n local accountng standards between two countres, I drectly test whether dfferences between the local accountng standards of the reportng frm and the local accountng standards of the nvestng fund can explan the country varaton n the holdngs for each frm. Dsaggregatng the foregn holdngs by each country also allows me to examne the sgnfcance of accountng dstance after controllng for other determnants of blateral cross-border holdngs (.e., geographc, economc and cultural dstance). Also, I can drectly test whether the captal flows followng IFRS adopton were drven by countres where dfferences n accountng standards reduced the most followng IFRS adopton. The measure of dfferences n local accountng standards s based on Bae, Tan, and Welker (2008). 26 Ths measure s constructed based on a survey of seven global accountng frms (Nobes 2002) desgned to examne how much natonal accountng standards devate from IFRS. 27 The two-year survey presents a detaled comparson of dfferent accountng rules, and classfes the accountng rules to be ether the same or dfferent from IFRS. Bae et al. (2008) dentfy 21 accountng standards based on whether the standards show suffcent varaton across countres and also have been recognzed as a 26 There exst other measures of dfferences n accountng standards across countres; however, these often relate to frms accountng choce (Bradshaw et al. 2004) or ther reportng behavors (Hung 2000). Snce my man nterest s the extent to whch accountng standards dffer between two countres, I use a modfed measure of Bae et al. (2008), whch compares the natonal accountng standards of each country-par. 27 GAAP 2001 (Survey of Natonal Accountng Rules Benchmarked Aganst Internatonal Accountng Standards) s a comprehensve two-year study that presents a detaled comparson of each local GAAP n 80 dfferent accountng dmensons. Comparsons of each local standard are based on both the actual dfference n the standards and the dfference n accountng standards as perceved by the nvestors. 21
key accountng tem n pror lterature (Comprx et al. 2003). A composte measure of accountng dstance can be constructed for each country-par by summng the 21 bnary values assgned after comparng ndvdual accountng standards. Bae et al. (2008) consder two accountng standards as smlar when both rules comply wth IFRS or when both rules follow local standards that are noncomplant wth IFRS. However, t s unclear that a par of non-complant local accountng standards should be consdered more smlar to each other than a par of local accountng standards where only one comples wth IFRS. Thus, I modfy Bae et al. (2008) and consder two non-complant local accountng standards to be smlar only f the two countres are from the same legal orgn (AD1). Ths gves a measure of accountng dstance across all country-pars wth a scale from 0 to 21. For robustness, the emprcal tests also nclude the orgnal measure of Bae et al. 2008 (AD2), treatng two local standards that are non-complant to IFRS as smlar, and another measure of accountng dstance that assumes two non-complant local standards to be dfferent (AD3). Note that global adopton of IFRS ntroduced tme-varaton n the accountng dstance measure by changng the smlarty of accountng standards across all countres. Countres that adopted IFRS experenced a reducton n accountng dstance wth other countres that adopted IFRS at the same tme. Countres not partcpatng n the global adopton, on the other hand, became more dstant from the accountng standards of the rest of the world. The emprcal test uses ths change n the accountng dstance trggered by the IFRS adopton ( AD) to dentfy the effect of accountng harmonzaton on cross-border nvestments. Barrers that cause frctons to blateral holdngs go beyond accountng standards. Prevous studes have shown that both explct and mplct barrers play a role n cross-border nvestment decsons. Explct barrers such as captal controls exst but have been found to be non-bndng (Tesar and Werner 1995). On the other hand, mplct barrers such as a general lack of famlarty due to 22
geographc or cultural dstance have had greater emprcal success. The leadng explanaton, known as the gravty model, fnds that the majorty of varance n cross-border captal flows can be explaned by geographc dstance. 28 In contrast to real goods, fnancal assets have no weght and geographc dstance cannot be a measure of transportaton costs. Thus, for fnancal assets, geographc dstance has been nterpreted as a measure of the cost of nformaton acquston, and therefore the lkelhood of nvestors sufferng nformaton asymmetry (Portes and Rey 2005). Table 5 summarzes the other determnants of blateral cross-border holdngs ncluded n ths study. I nclude varables whch proxy for other nformaton barrers: geographc dstance and phone traffc volume. Geographc dstance (ldst ) s a proxy for nformaton cost and s motvated by the strong emprcal support from prevous studes. I also nclude measures of economc dstance (e.g., phone traffc volume, and trade ntensty) and cultural dstance (e.g., common language, common border, and an egaltaransm ndex) between the two countres. Phone traffc volume(teleph ) s the volume of phone traffc between two countres scaled by the geometrc average of each country s populaton. Trade ntensty ( ltrade n bllon US dollars. Common language ( comlang ) s measured as the log of total net exports between two countres ) s an ndcator varable set to one when two countres use a common language. Border ( border ) s an ndcator varable set to one when two countres share a common border. Egaltaransm ndex ( egal countres cultural egaltaransm score (Segel, Lcht, and Schwartz 2008). ) s the squared dfference of the two Correlatons n Panel B of Table 5 show that accountng dstance s related to other barrers of cross-border nvestments. Accountng dstance s postvely related to geographc dstance and negatvely related to nformaton flows as measured by telephone traffc and common language. The 28 The gravty model has been the workhorse model for cross-border trade n real goods snce the 1960s. Geographc dstance s shown to explan 70% of the varaton n blateral trade of real goods (Anderson and van Wncoop 2005). However, even n trade of fnancal assets where tradng costs are unlkely to be affected by physcal dstance, ths varable has been shown to be the strongest determnant of cross-border holdngs (Portes and Rey 2005). 23
correlatons between accountng dstance and other determnants make t hard to dentfy accountng nformaton as a sngle bndng constrant on cross-border holdngs. Thus, I use the changes n the accountng dstance trggered by the IFRS adopton to examne how reducng accountng dstance affects cross-border holdng decsons. 4.2 Multvarate analyss of accountng dstance and holdngs from dfferent countres I examne the effect of accountng dstance on holdngs of each country-par usng the regresson model n equaton (5). Ths s a test of my frst hypothess statng that frms wll have dmnshng ownershp from funds n countres wth greater dfferences n ther local accountng standards. The specfcaton of baselne holdngs s from Martn and Rey (2000) where holdngs of nvestors from a source country n a destnaton country are determned by the market sze of the two, and the estmates of tradng costs between the two countres. Holdngs MVsource MVdest AD controls (5),t 0 c,t 1,t 0,t 1,m,t,t m1 Holdngs,t s the percentage of frm s shares held by all funds from country c n year t. 29 For every frm-year, I calculate the percentage of shares held by funds from country c usng the shares of the M latest reportng perod. MVsourcec,t s the weght of mutual funds n the fund s country c relatve to the world s mutual fund holdngs. It should be noted that MVsourcec,t corresponds to the predcted holdngs of the nternatonal CAPM, where shares held by nvestors from a certan country equal the country s weght n the world s portfolo for any assets. Thus, absent any market frctons, the theoretcal value of 0 equals 1. MVdest,t s the weght of mutual funds n frm s country relatve to the world s mutual fund holdngs and measures the economc mass of the country where the frm s located. 29 Aggregatng Holdngs,t over all countres that nvest n frm wll yeld the Holdngs,t varable n Secton 3. 24