The Value Relevance of Foreign Income: An Australian, Canadian, and British Comparison

Transcription

1 The Value Relevance of Foreign Income: An Ausralian, Canadian, and Briish Comparison Gordon M. Bodnar Morris W. Offi rofessor of Inernaional Finance aul H. Nize School of Advanced Inernaional Sudies Johns Hopkins Universiy Washingon, DC (202) Lee Seok Hwang Deparmen of Accouning Chinese Universiy of Hong Kong and School of Business Adminisraion Baruch College, New York, N.Y and Joseph Weinrop* San Ross rofessor Deparmen of Accouning School of Business Adminisraion Baruch College, New York, N.Y (646) Ocober 10, 2001 * conac auhor The auhors wish o hank paricipans a he 1999 American Accouning Associaion Annual Meeings and an anonymous referee for commens. We also acknowledge he Disclosure Company for providing us wih World Scope and Columbia Universiy for funding hrough a CIBER gran.

2 The Value Relevance of Geographical Segmen Disclosure: An Ausralian, Canadian, and Briish Comparison Absrac In his paper we examine he value-relevance of geographical earnings disclosures for firms lised and domiciled in Ausralia, Canada and he Unied Kingdom. We find ha foreign earnings in all hree counries are valued differenly han domesic earnings. The esimae of he associaion coefficien for foreign earnings changes wih reurns is posiive in all hree counries and saisically larger han he associaion coefficien for domesic earnings changes in Canada and he Unied Kingdom. Furher ess show ha his difference is relaed o relaive growh opporuniies of overseas operaions o domesic operaions. These findings are similar o resuls for foreign earnings associaion coefficiens for American-based mulinaionals found in Bodnar and Weinrop (1997). These resuls indicae ha across counries he marke perceives he resuls of foreign operaions as value relevan and suggess ha, greaer emphasis should be placed on he required disclosure of segmenal daa raher han on he concern ha all counries prepare he segmenal informaion using a common GAA. Keywords: Foreign Income, Valuaion, Growh Opporuniies,

3 1. Inroducion As he world coninues o globalize, he foreign operaions of corporaions are becoming increasingly relevan as a source of value. However, in only a few counries are firms required o make deailed disclosures of heir foreign operaions, despie he fac ha foreign operaions can experience paerns of profiabiliy, growh, and risk ha differ from hose of he consolidaed operaions. Given he increasing imporance of foreign operaions, surprisingly lile research has been done on he valuerelevance of such disclosures for invesors and wha any differenial marke reacions ell us abou he economic characerisics of foreign operaions relaive o hose of domesic operaions. In one of he earlies sudies on geographic disclosures, Boasman, Behn and az (1993) demonsraed ha regional foreign income disclosures were no value-relevan for US mulinaional firms reurns beyond heir par in oal income. However, in a more recen evaluaion of he value relevance of regional geographic daa, Thomas (2000) finds some suppor for he value relevance of geographical segmenal daa wih sock reurns over hree-year windows. Bodnar and Weinrop (1997) consider he value relevance bu consider he less ambiguous decomposiion of earnings ino jus domesic and foreign componens. For a large se of U.S. mulinaional firms, hey find value relevance for foreign income disclosures by demonsraing ha changes in foreign income are more posiively associaed wih annual equiy reurns changes in domesic income. Afer furher invesigaions hey conclude ha he difference resuls from he marke pricing greaer opporuniies for fuure income growh from successful foreign operaions han successful domesic operaions. A second line of invesigaion is offered in Garrod and Rees (1998). They use foreign segmen daa o exend he Ohlson (1995) valuaion model and sudy he earnings and book values for mulinaional firms in he Unied Kingdom. They find ha Unied Kingdom mulinaional firms are more highly valued han Unied Kingdom domesic firms. They also invesigae he value impac of foreign operaions. While hey fail o deec a significan premium for he oal foreign operaions, hey do repor ha operaions in he U.S. are more highly valued han operaions han oher areas. 1

4 In his paper, we follow he approach of Bodnar and Weinrop (1997), and look a he issue of value relevance of aggregae foreign income disclosures for a se of non-u.s. firms. The reason for preferring his approach for non-u.s. counries is wo-fold. Firs, firms in counries oher he U.S., if hey make geographic disclosures, are more likely o disclosure only non domesic income, and as such rying o look a regional geographic disclosures resuls in an exremely small sample. 1 Second, if firms do disclose he regions ha hey operae in, a review of financial saemens demonsraes ha here is no commonaliy in he selecion of he paricular segmens. Given his mehodological decision, he goals of his paper are o deermine if foreign income disclosures are value-relevan under GAA environmens oher han he U.S. and o deermine if he marke s percepion of he economic characerisics of foreign income in hese counries are consisen wih hose found in he U.S. by Bodnar and Weinrop (1997). Ideally, such a sudy would like o examine a large se of counries in order o carry ou cross secional ess; however, unil recenly, very few counries have mandaed he disclosure of income on a geographic basis for heir mulinaional firms. As a resul, we examine he value-relevance of foreign income in hose hree counries ha have mandaed foreign income segmen disclosures since 1990: Ausralia, Canada, and he Unied Kingdom. While he differences in GAA among hese counries are no ha significan, his cross secion of counries allow us o deermine if he value relevance of foreign income disclosures is a fundamenal economic phenomenon or a peculiariy of he US marke or accouning sysem. Furher, specific ess use changes in income componens ha miigae concerns over specific differences in accouning policies. Using daa from 444 Ausralian firm-years, 891 Canadian firm-years and 738 Briish firm-years, from , we find ha changes in foreign income have a posiive associaion wih annual reurns for all hree counries. For Canada and he Unied Kingdom, he foreign income associaion is significanly larger han he domesic income associaion wih annual reurns while for Ausralia he difference is only marginally larger across he full sample of firms. This evidence suggess ha, as in he 1 Foreign revenues are also commonly disclosed. For purposes of our ess, here is limied evidence of he value relevance for geographic revenue disclosures for domesic or foreign firms. 2

5 U.S., foreign income disclosures are incremenally value-relevan o invesors and ha foreign income is more highly associaed wih annual reurns han domesic income. We also conduc ess o examine wheher he larger foreign associaion coefficien is relaed o greaer growh opporuniies for foreign operaions. Using relaive sales growh as a measure of relaive fuure growh opporuniies, we find ha he foreign income associaion coefficiens in all hree counries are larger han domesic income associaion coefficiens when foreign growh opporuniies are higher han domesic growh opporuniies. Finally, we demonsrae ha he resuls are robus o he eliminaion of negaive income realizaions and he inclusion of income levels. We provide furher evidence ha, despie differences in income deerminaion across hese counries, he economic consequences associaed wih he disclosure of foreign income are he same. Marke paricipans evaluae income from foreign sources differenly from income from domesic operaions, and in paricular is changes are more highly associaed wih annual reurns. This underlines he imporance of geographic segmenal disclosures across markes. These findings have implicaion for he inernaional accouning regulaory bodies as hey consider moves oward global sandards. This suggess ha a movemen owards weaker segmenal disclosure requiremens, such as hose of he new Unied Saes segmenal disclosure rule, SFAS No. 131, will resul in a reducion of value-relevan informaion for invesors abou mulinaional firms. 2 The remainder of he paper is organized as follows. The nex secion develops he mehodology and discusses he relevan heoreical predicion for he empirical secion. Secion 3 describes he sample selecion and Secion 4 presens he basic empirical findings. Secion 5 provides some invesigaion of he economic explanaions of he resuls and Secion 6 provides some invesigaion of he economic explanaion for he resuls. Finally, Secion 7 concludes. 2 Herman and Thomas (2000) demonsrae ha he new SFAS No. 131 disclosure requiremens have resuled in smaller percenage of U.S. firms reporing geographic income relaive o he disclosure requiremens of he previous SFAS No

6 2. Mehodology: The concep of a muli-counry sudy of he value-relevance of income goes back o Alford, Jones, Lefwich and Zmijewski (1993) who address he issue of he value-relevance of annual earnings prepared under differen GAA applicaions. For a sample of 18 counries from 1982 o 1990, hey documen a posiive associaion beween earnings and reurns. Among he 18 counries hey examined were Ausralia, Canada, and he Unied Kingdom. Wih he esablishmen of he commonaliy of an earnings/reurns relaion, researchers have moved o deermine if firms supplemenal disclosures are also associaed wih annual reurns across counries ha use differen GAAs. We choose geographically segmened income daa as he supplemenal disclosure because of he imporance ha foreign operaions play for he larges firms in hese counries. Disclosures of foreign operaions should be of imporance o invesors of firms wih significan foreign operaions if foreign and domesic operaions are subjec o differen risks and reurns as is commonly believed in pracice. The issue of supplemenal foreign daa has been an area ha accouning regulaors have long deemed as imporan o invesors. In he Unied Saes, he Securiies and Exchange Commission firs mandaed disclosure of geographic segmen informaion in 1969 as par of heir segmened disclosure requiremens in Accouning Series Releases and exended i o all publicly raded firms in According o hese rules, a firm was required o idenify segmens providing 10 percen or more of consolidaed revenue and/or earnings. In 1976, he Financial Accouning Sandards Board (FASB) adoped measures (SFAS No. 14) very similar in naure o he SEC requiremen and added he addiional crierion ha disclosure was required if more han 10 percen of he oal asses of he firm was in one segmen. Segmens were defined as boh service/produc lines and geographic locaion, as well as expor sales or major cusomers. Cenral o he reasons for including geographic segmens was he belief of he 3 See Swaminahan (1991) for an exended discussion on he hisory of he expansion of foreign daa disclosures in he Unied Saes. 4

7 regulaory bodies ha informaion conained in geographic operaions is relevan o financial saemen users. 4 Ouside he Unied Saes, geographic segmen disclosures are no as common, hough here is a growing recogniion of heir imporance. This view is refleced in he geographic segmenal disclosure requiremens currenly oulined wihin he conex of he Inernaional Accouning Sandards No. 14 (IAS #14), Financial Reporing Informaion by Segmen. 5 This March 1981 regulaion was inended o aid he users of financial saemen,...o access he effec ha operaions in differen indusries and in differen geographical area may have on he enerprise as a whole. 6 The major disclosure requiremens are o provide daa on segmenal revenues, income and asses when a segmen exceeded a 10% hreshold. In addiion o his accouning sandard, he European Economic Communiy recommends geographic disclosures for boh financial and non-financial resuls of a firm s mulinaional operaions. Their Fourh Direcive requires ha sales be segregaed by caegories of aciviy and geographic marke o he exen ha hese caegories and markes differ subsanially from one anoher. The Organizaion of Economic Cooperaion and Developmen s Guides for Mulinaional Enerprises also suggess a series of geographic disclosures of operaing revenues and sales by geographic area. They also ask for disclosure of significan new capial projecs and he average number of employees in each geographical area. For he hree counries of our sudy, Ausralia, Canada, and he Unied Kingdom, foreign segmen disclosures have been mandaed since a leas Unlike many oher foreign counries where disclosures are volunary, and largely limied o revenues, hese hree counries all mandae similar disclosures as in he U.S., requiring reporing of foreign sales, asses and income. In all cases, he geographic disclosures we sudy are provided in erms of he GAA of he primary reporing counry. Thus, a company incorporaed in he Ausralia, will use Ausralian GAA for is supplemenal disclosures 4 More recenly, he U.S. has adoped SFAS No One implicaion of his new saemen for fiscal years 1998 and beyond is he dropping of he requiremen of mandaed foreign income disclosures. 5 We direc he reader o Choi and Mueller (1992), pp for a more deailed hisory of hese disclosure issues. 6 Ibid., par We direc he reader o Choi and Mueller Table 7.1. In his able, he disclose requiremens for hireen sock markes are provided. Of ineres o his sudy is iem 4g, segmenal financial informaion in Ausralia, Canada and he Unied Kingdom. 5

8 (similarly for Canadian and U.K. firms). For he purposes of our sudy, each of hese counries reflecs he regulaory policies of he local sock exchange (he Sydney Sock Exchanges for Ausralia, he Torono Sock Exchange for Canada and he London Sock Exchange for he Unied Kingdom). 2.2 Model Developmen Our ineres is o examine he associaion of foreign income changes wih annual equiy reurns for he sample firms. This allows us o es for he incremenal value-relevance of foreign income beyond oal income and provides us insighs abou marke percepions of fuure invesmen and producion opporuniies ouside of he home counry. Foreign operaions, wherever locaed, face risks (currency, legal and poliical) and opporuniies (new markes) ha differ fundamenally from domesic operaions. These differences sugges ha foreign income disclosures could have incremenal explanaory power for reurns han domesic income. The naure of any incremenal explanaory power could provide insighs ino he differenial economic characerisics of foreign income relaive o domesic income. The sandard associaion regression in accouning was originally esed by Ball and Brown (1968) for U.S. firms and also used by Alford e al. (1993) for non U.S. firms. In i, changes in he firm s marke value are regressed agains changes in componens of consolidaed earnings 8. V = α 0 + α 1 TERN + ε. (1) where V = he change in he marke value of firm i over he reporing period TERN = he change in oal earnings of firm i over period. The associaion coefficiens, α 1, provide a measure of he sensiiviy of firm value o repored earnings, respecively. Theoreically, if markes are efficien and earnings follow a simple random walk, he associaion coefficien would be equal o one over he firm s cos of capial. 9 8 The sudies cied rely on he ne income. As we will demonsrae laer, we are consrained by daa availabiliy and we are only able secure foreign operaing income and hus modify our heory appropriaely. 9 In a simple model, where earnings sreams are assumed o evolve randomly and he value of he firm is equal o he presen discouned value of he earnings sream, i is possible o show ha he associaion coefficien beween an earnings change and 6

9 For he purposes of sudying foreign operaing income disclosures, one can modify equaion (1) by adding he change in foreign income. V = γ + γ TERN + γ FOERN + ε (2) 0 1 Where: FERN is he change in foreign income for firm i in period. In his specificaion, γ 1 now represens he associaion coefficien for he domesic porion of he income change and γ 2 is he incremenal associaion coefficien for foreign income. This is because of he perfec decomposiion of oal income ino domesic and foreign income. The full associaion of he foreign income wih reurns is given by γ 1 + γ 2. Thus from equaion (2), γ 2 is he coefficien of ineres in deermining wheher he associaion of foreign income is significanly differen from ha of domesic operaing income. A significan incremenal foreign associaion coefficien, γ 2, would imply ha foreign income disclosures are incremenally value relevan. There are several economic argumens as o wheher he sign of he incremenal foreign associaion coefficien, γ 2, should be posiive or negaive. One argumen is ha foreign earnings could be valued less highly han domesic earnings by he marke because hey are perceived o be less reliable and/or more uncerain. Foreign income sreams can be more uncerain in economic value as hey face more risk facors han domesic income, including greaer business cycle volailiy and currency gyraions. In addiion, acions by foreign governmens can make he availabiliy o shareholders of fuure and curren foreign earnings less cerain han domesic income. Because of his greaer risk and lower reliabiliy, foreign operaions may require a higher of rae of reurn on such aciviies. This higher rae of reurn implies a higher discoun rae for foreign income han domesic income resuling in a lower associaion of foreign income changes wih reurns. Repored foreign operaing income may also be less represenaive of rue economic condiions han domesic operaing earnings because of addiional 2 he marke value of he firm will be equal o one over he marke discoun rae for ha sream of earnings (see, e.g., Collins and Kohari (1989)). 7

10 accouning echniques required o ranslae hem from foreign currency ino domesic currency making here repored value less value relevan. 10 On he oher hand, here are reasons ha foreign associaion coefficiens could be larger han domesic associaion coefficiens. One explanaion is ha despie heir added volailiy exchange rae changes work o make foreign operaing income changes more persisen because exchange rae changes are hemselves highly persisen, (see, e.g., Messe and Rogoff (1983) and Frankel and Rose (1995)). If he influence of exchange rae resuls in foreign earnings being more persisen ha domesic operaing earnings, a change in foreign earnings oday would have a larger impac on value han an equivalen change in domesic earnings. 11 Anoher reason could be he greaer use of conservaive accouning mehods in newer (i.e., foreign) operaions. For example, in a new segmen, he audior may ask for more aggressive expense recogniion and more conservaive revenue recogniion unil he segmen esablishes iself and some uncerainy has been resolved. If audiors require a greaer use of conservaive accouning procedures in foreign aciviies because of heir newness, his would resul in larger muliples for foreign income. An addiional reason for a larger foreign associaion coefficien relaes o differences in growh opporuniies beween foreign and domesic operaions. Collins and Kohari (1989) demonsrae ha associaion coefficiens are a posiive funcion of he firm s fuure growh opporuniies. Because in he normal evoluion of mulinaional firms, foreign operaions ypically follow he developmen of successful domesic operaions, i is possible ha successful foreign operaions indicae greaer opporuniies for fuure growh han he more esablished domesic operaions. This would be consisen wih he view ha foreign operaions represen expansions ino new, less exploied markes (see, e.g., 10 This is rue of he emporal mehod of foreign currency consolidaion used by foreign operaions ha have designaed he U.S. dollar as funcional currency. For more on his problem, see Barov and Bodnar (1996). 11 ermanen changes in earnings by definiion should lead o a larger impac on price oday as he presen value of he impac is larger han for a similar-sized emporary change in earnings. Eason and Harris (1991) argue ha under he specificaion of earnings as a random walk, changes in earnings represen permanen componens and levels of earnings represen emporary componens. They include boh earnings levels and changes in a single regression wih reurns and find he earnings response coefficien for changes o be higher han i is for levels. Of course he influence of he exchange rae will only resul in larger foreign associaion coefficiens o he exen ha changes in foreign currency income are no less predicable or perceived less permanen han changes in domesic income. 8

11 Kogu (1983) and Sopford and Wells (1972)). If foreign operaions currenly offer greaer growh opporuniies han domesic operaions, hen one would expec a larger associaion beween firm value and foreign income han beween firm value and domesic income. 3. Daa and Sample Selecion To es he value relevance of foreign income disclosures in Ausralia, Canada and he Unied Kingdom, i is necessary o obain samples of firms ha disclose foreign income and (for subsequen ess) foreign revenues. We obain his informaion from he WorldScope daabase from Disclosure. The version of WorldScope we use is Ocober 1999 and i provides daa on firms disclosure of heir foreign operaions for he 10-year period from 1989 o 1998 in he form of percenages of oal. 12 In hese hree counries, WorldScope repors percenages for foreign sales, foreign operaing income, and foreign asses, as percenages of he consolidaed oals. Daa on consolidaed values for sales, operaing income, and asses are aken from WorldScope and cross-checked wih daa from Compusa s Global Vanage for each firm. These consolidaed figures are combined wih he WorldScope raios o consruc he nominal figures for foreign sales, operaing income and asses. All values are repored in he firms home currency. I is also imporan o noe ha he geographic segmen disclosures in hese counries are available only a he operaing income level and no he ne income level. In he nex secions, all of our comparisons of domesic versus foreign associaion coefficiens are done a he operaing income level. Since non-operaing income may also be value relevan, we also require daa on he oal non-operaing income (he difference beween operaing income and ne income) for each firm. This daa is available only on a consolidaed basis. 12 Worldscope repors he foreign percenages down o basis poins. I also provides daa on he absolue amouns of foreign variables in unis of billions of local currency. These figures are oo crude o accuraely measure changes for mos firms. 9

12 In order for a firm o remain in he sample, we require wo consecuive years of daa for operaing income and sales as well as he foreign percenages for each. The wo consecuive years of accouning daa are needed o calculae change variables. The dependen variable in our analysis is a welve-monh domesic-marke adjused reurn (MAR) for he firm, measured in he domesic currency. To calculae his variable, we need sock prices, dividends, and share changes which we obain from he Global Vanage. The reurn is calculaed from year-end prices adjused for sock splis and sock/cash dividends plus he value of cash dividends per share disribued during he holding period. To conrol for common reurn influences wihin a counry, we adjus each firm s welve-monh reurn by he corresponding annual reurn o he local marke index. 13 The local marke index is deermined forming an equally-weighed porfolio of all of he firms on he exchange. In our sudy, he placemen of he window o accumulae he annual reurn is based on he counry-specific regulaions regarding he public release of he accouning daa. 14 For example in Grea Briain, firms are required o file financial saemens 120 days afer he year-end, so we calculae our marke-adjused reurns over a welve monh period ending four monhs afer he close of he fiscal year end. 15 We also use he price a he beginning of he reurn measuremen period o normalize he income variables. For he purposes of our sudy, we are only ineresed in he securiies of he firms in he counry of heir incorporaion. Thus a firm incorporaed in Canada (wih is head office in Canada) is considered only in he Canadian ess even if he firm was lised on oher exchanges around he world. In addiion we delee firms in he financial services indusries because heir foreign revenues disclosures are no readily comparable o he non-financial firms ha dominae he sample. 13 We obain fundamenally similar resuls wih all our ess when we use raw reurns insead of marke adjused reurns. 14 Alford e al. (1993, Table 1) provide daa on he differen reporing requiremens of firms when i comes o making heir daa publicly available. 15 We direc he reader o Alford e al. (Table 1). We found no deviaions from his lising and so hese represen he iming lags we used o calculae our window of marke adjused reurns. 10

13 These daa requiremens resul in a sample of 444 firm-year Ausralian observaions, 891 Canadian firm-year observaions, and 738 U.K. firm-year observaions. Table 1 provides some summary saisics on hese firms. Across he hree counries, as expeced when dealing wih mulinaionals, hese are large firms, wih median asses of approximaely in he range of US$400MM in he U.K. o US$550MM in Ausralia. Moreover, he median observaions indicae a significan degree of foreign revenues and income, wih he hird quarile observaions in boh Canada and he UK having a majoriy of heir income coming from abroad. Table 2 repors he summary saisics for he regression variables used in he nex secion. 4. Empirical Tess We begin our empirical ess by verifying ha each counry in our sample demonsraes a significan posiive relaion beween changes in annual earnings and reurns similar in naure o hose repored by Alford e al. (1993). Equaion (3) provides empirical specificaion where he change in oal ne income is broken ino he change in oal operaing income and he change in non-operaing income. MAR TOES = α 0 + α1 1 TNOES + α1 1 + ε (3) where: MAR, is marke adjused change in firm s i sock price over he year, TOES is he change in oal operaing earnings for firm i in year, TNOES is he change in oal non-operaing earnings for firm i in year, -1, is firm i s share price a he beginning of he reurn accumulaion period, ime -1, used o normalize he income per share info ε is he error erm for firm i a ime. 16 We are forced o use his specificaion since our daa on foreign income is a he operaing level. The resuls of regression (3) for each counry are repored in anel A of Table As expeced, all hree 11

14 counries produce a posiive and highly significan coefficien on he change in operaing and nonoperaing earnings per share. Since he change in ne income is by definiion he sum of hese wo erms, he associaion coefficien for ne income is he sum of he wo associaion coefficiens (α 1 +α 2 ). I is no surprising ha ne income is slighly, bu significanly, more associaed wih annual reurn han is operaing income. These resuls are logically consisen wih Alford e al. (1993) and indicae he value relevance of hese wo ne income componens. Having esablished he associaion of operaing and ne income wih reurns in hese hree counries, we now move on o es wheher he foreign operaing income disclosure is value-relevan. We do his by modifying equaion (3) in he same fashion ha we modified equaion (1) by adding he change in foreign operaing income over price ( FOES / -1 ) as an addiional variable o he original regression. MAR TOES = δ 0 + δ 1 1 TNOES + δ 2 1 FOES + δ 3 1 ε + (4) anel B of Table 3 provides he resuls of his regression for Ausralia, Canada and he Unied Kingdom. For all hree counries he associaion coefficien on he oal change in operaing income, which now measures he associaion of he domesic componen of operaing income, has decreased relaive o he esimaes in anel A, bu remain saisically significan. The associaion coefficien on he change in non-operaing income remains posiive and saisically significan. The associaion coefficien on he change in foreign operaing income is posiive and saisically significan a convenional levels for Canada and he Unied Kingdom, and is posiive and significan a he 11% level for Ausralia. The consisen posiive value indicaes ha he associaion of foreign operaing income is larger han ha of domesic operaing income for all hree counries. For Ausralia, he incremenal foreign associaion coefficien is 0.43, implying a oal foreign operaing income associaion coefficien of 1.44, for Canada, 16 The use of a naive predicion for operaing and non-operaing earnings expecaions on a per share is consisen wih Boasman e al. (1993) and Alford e al. (1993). See Chrisie (1987) for a discussion of he division by price a he beginning of he period. 12

15 he incremenal foreign associaion coefficien is 0.62 implying a oal foreign operaing income associaion coefficien of 1.53, and for he U.K, he incremenal foreign associaion coefficien is 0.92, implying a oal foreign operaing income associaion coefficien of This finding of a meaningfully posiive incremenal foreign income associaion coefficien is consisen wih similar resuls for he U.S. shown in Bodnar and Weinrop (1997). Togeher hese resuls suppor he claim ha marke paricipans find foreign income disclosures incremenally useful when deermining he value of mulinaional firms and ha i appears o be an economic phenomenon ha changes in foreign income are more posiively relaed wih reurns han are changes in domesic income. The commonaliy of his resul across four counries despie differences in GAA suggess ha his is because of because of fundamenal characerisics of foreign income. In he nex secion we examine if he same economic explanaion for his phenomenon in he US is suppored in hese counries. 5. Economic Explanaion The resul of a posiive incremenal associaion coefficien for foreign income for Canada, he U.K., and Ausralia (hough marginally significan) is consisen wih he U.S. resuls of Bodnar and Weinrop (1997). As menioned above, here are several possible explanaion for he more posiive associaion of foreign income han domesic income. Bodnar and Weinrop (1997) invesigae several of hese possibiliies and find suppor for he growh opporuniies sory by demonsraing ha he size of he incremenal foreign associaion coefficien is relaed o he relaive growh opporuniies in he foreign segmen proxied for by relaive sales growh each segmen. 18 Drawing upon he fac ha previous research had demonsraed a posiive link beween growh opporuniies and associaion coefficiens, hey inerpreed heir resul as indicaing he marke s percepion of greaer growh opporuniies from successful foreign operaions. In his subsecion we carry ou a similar analysis o see if he same 17 The regressions are screened for overly influenial observaions using a sandard Cook s Disance es and ess winsorizing he op and boom 1% do no change he qualiaive naure of he conclusions. 13

16 economic phenomenon will explain he posiive incremenal foreign associaion coefficiens in hese hree counries. Following he approach of Bodnar and Weinrop (1997), we use he firm s segmenal disclosures on foreign sales from WorldScope o calculae percenage changes in foreign and domesic sales over he previous year. The sample is hen porioned ino wo groups based upon he relaive sales growhs in hese wo regions. In he firs group are he observaions where domesic sales growh is larger han foreign sales growh over he previous year and in he second group are hose observaions where foreign sales growh is as large as, or larger han domesic sales growh. If relaive foreign domesic sales growh proxies for he marke s percepion of relaive growh opporuniies beween foreign and domesic operaions, and greaer growh opporuniies are relaed o more posiive associaion coefficiens, hen we should expec o see differen resuls across sub samples wih differen relaive regional sales growh. For he case where domesic sales growh is greaer han foreign sales growh, here is no signal of greaer foreign growh opporuniies and we would expec ha here should be no significan incremenal foreign associaion coefficien. The reason we migh no expec he foreign associaion coefficiens o significanly smaller is ha we know ha uncondiionally he effec is posiive and i is ofen he case ha firms expand overseas only afer exhausing mos of he growh opporuniies a home so his siuaion migh no necessarily indicae subsanially more growh opporuniies a home han abroad. In he second case, where foreign sales growh surpasses domesic sales growh signaling greaer foreign growh opporuniies, we should expec o see large and significan incremenal foreign associaion coefficiens. This es is run on equaion (4), and he resuls on he wo subsamples are repored in Table 4. In anel A, where domesic sales growh oupaces foreign sales growh, we see no evidence ha he foreign associaion coefficien is saisically larger han he domesic associaion coefficien. For wo of he hree counries (Canada and he U.K.) he incremenal foreign associaion coefficien is no significanly 18 They also invesigaed he issue of he persisence of he exchange rae change by decomposing he foreign income ino exchange rae and foreign currency componens. The removal of he foreign exchange rae componen had no deecible effec 14

17 differen from zero, while for Ausralia, i is acually significanly smaller han he domesic associaion coefficien. However, in anel B, when foreign sales growh oupaces domesic sales growh, we see consisen and significan evidence of a posiive incremenal foreign associaion coefficien for all hree counries. In his sub sample, he incremenal foreign associaion coefficien is larger and more significan han in he uncondiional regression repored in anel B of Table 3. Togeher he wo panels of Table 4 indicae ha he phenomenon of significanly larger foreign associaion coefficiens is relaed o relaive regional sales growh. This resul is qualiaively he same as ha found for he U.S. firms in Bodnar and Weinrop (1997). The foreign associaion coefficiens are significan and posiive and he incremenal foreign associaion coefficien is relaed o he relaive regional sales growh (which acs as a proxy for incremenal growh opporuniies). Thus, hese resuls poin o he conclusion ha his effec is no unique o he U.S. or is accouning sysem, bu more likely a fundamenal feaure of he marke s percepion of firms foreign income. As such, i poins o he imporance and value-relevance of providing invesor s specific informaion on firm s foreign operaions. 6. Robusness Checks In his secion we consider wo specific adjusmens of he specified model o examine he robusness of our resuls. Firs, a possible concern abou inerpreing hese resuls is he omied variables problem. Recen work on he relaion beween earnings and reurns (see, e.g., Eason and Harris (1995) and Ali and Zarowin (1992)) has demonsraed ha in addiion o he change in earnings, he level of earnings is an imporan explanaory variable for annual reurns. The raionale behind he imporance of he level of earnings relaes o is abiliy o capure ransiory componens of earnings above he permanen componen of earnings ha are capured in he earnings change. Based upon hese sudies, we add he level of oal operaing income, foreign operaing income and oal non-operaing income on he significance of he incremenal foreign associaion coefficien. As a resul, we do no repea his invesigaion here. 15

18 (normalized by beginning of period price) o our empirical specificaion. This allows o us o check wheher our incremenal foreign associaion coefficien resul is driven by he omission of he level of earnings. While he relaion beween income levels are equiy reurns are no well defined beyond saisical issues, we do no provide predicions for he coefficiens on he income level variables, oher han o sugges ha foreign income should on average be more permanen, suggesing a less imporan relaion beween reurns and foreign income levels han wih reurns and domesic income levels. A second poenial problem in inerpreing he size of associaion coefficiens is he impac of negaive income realizaions. As negaive income is no a sensible long run forecas (as he firm would evenually go under) negaive income realizaion mus be expeced o be emporary. Therefore, he change in income, when income urns negaive, will have a lower associaion wih reurns han income changes when income is posiive (see Hayn (1995)). This bias could be driving our resuls if he firms in our sample repored negaive income realizaions more frequenly for domesic income han for foreign income. To deermine wheher his problem is driving he resuls in anel B of Table 4, we creae dummy variables o separae ou hose cases where repored income, domesic, foreign, in boh changes and levels, is negaive. These new variables are NEG_ TOES/, NEG_ FOES/, for negaive change in oal and foreign operaions and NEG_TOES/ and NEG_FOES/ o represen negaive levels of oal and foreign operaions, respecively. These variable are assigned he value of he income variable when he realized value of he income variable is less han zero for ha period, and are zero oherwise. As a resul, hese variables represen he differences in he associaion coefficiens when he paricular income realizaion We es he combined effecs of he levels and changes and he issues of negaive earnings in equaions (5). MAR TOES = ω NEG _ FOES 6 1 NEG _ TOES ω 2 1 FOES 7 1 TOES 3 1 NEG _ FOES 8 1 NEG _ TOES 4 1 TNOES 9 1 FOES 5 ω 10 1 TNOES 1 + ε (5) 16

19 The resuls of esimaing equaion (5) are repored in Table 5. In anel A of Table 5, we repor our findings where he domesic sales growh is greaer han he foreign sales growh. Consisen wih previous research he level of oal operaing income is saisically significan for all hree counries. As suggesed, he associaion of he foreign income level is eiher no differen from domesic income or is significanly smaller (in he case of Ausralia). While he change in oal operaing income remains significan in Canada and he U.K., he incremenal associaion coefficien for he change in foreign operaing income remains insignifican excep for he U.K. were i is now posiive (0.51) and significan. anel B of Table 5 reveals resuls for subsample when he foreign sales growh is greaer han domesic sales growh. I is ineresing o noe ha he level of oal income is no as significan as in anel A, perhaps because he income change is perceived as more permanen. Once again, he coefficien on he level of foreign income is no significanly differen from zero. In suppor of our conclusions from Table 4, even wih he addiion of income levels and he separaion of negaive income realizaions he incremenal associaion coefficien on he foreign income change is posiive. For Canada and he U.K. i is significan while for Ausralia i is marginally significan a he 11% level. Thus despie our limied sample sizes hese robusness ess ha include income levels and separae ou negaive income realizaions do no provide any compelling evidence agains original findings or our inerpreaion of he phenomena. 7. Summary and Conclusions In he Unied Saes here have been mandaed disclosure requiremens of geographic revenues, income and asses since Bodnar and Weinrop (1997) demonsrae he value relevance of foreign income disclosure in he U.S. In his paper we provide evidence for he value-relevance of foreign income disclosures for Ausralia, Canada and he U.K. Similar o he findings in he U.S. we demonsrae ha foreign income changes are more posiively associaed wih annual reurns han are domesic income changes. Taken ogeher hese resuls sugges ha foreign income is perceived by he marke o have 17

20 differen economic characerisics han domesic income. Moreover, also consisen wih he resuls from he U.S., we demonsrae ha he incremenal foreign associaion coefficien is posiively relaed o a proxy for he incremenal degree of growh opporuniies in foreign operaions. We claim his indicaes marke paricipans value foreign income more highly because of he greaer degree of growh opporuniies hey perceive. Finally, hese resuls are largely robus o he addiion of income level variables and adjusmens for income negaive income realizaions. 18

21 References 17 CFR Regulaion paragraph (h) General Noes o Financial Saemens - Income Tax Expense. Al A. and. Zarowin, The Role of Earnings Levels in Annual Earnings-Reurns Sudies, Journal of Accouning Research, (Auumn 1992): Alford, A., J. Jones, R. Lefwich and M. Zmijewsk "The Relaive Informaiveness Accouning Disclosures in Differen Counries," Journal of Accouning Research, (Supplemen 1993): Ball, R., and. Brown, "An Empirical Evaluaion of Accouning Numbers," Journal of Accouning Research, (Auumn 1968): Barh, M., and G. Clinch, Inernaional Accouning Differences and Their Relaion o Share rices: Evidence from he U.K., Ausralian, and Canadian Firms, Conemporary Accouning Research, Vol. 13 No. 1 (Spring 1996) pp Bhaga, S. and I. Welch, "Corporae Research & Developmen Invesmens Inernaional Comparisons", Journal of Accouning and Economics (March-May 1995): Belsley, D.A., E. Kuh, and R. E. Welsch, Regression Diagnosics. New York:: John Wiley and Sons, Boasman, J. R., B. K. Behn and D. H. az, "A Tes of he Use of Geographical Segmen Disclosures," Journal of Accouning Research, (Supplemen 1993): Bodnar, G. M., and J. Weinrop, "The Valuaion of Foreign Earnings: A Growh Opporuniies erspecive," Journal of Accouning and Economics, Vol 24, Chan, K., and G. Seow, The Associaion Beween Sock Reurns and Foreign GAA Earnings Versus Earnings Adjused o U.S. GAA, Journal of Accouning and Economics, Volume 21, No. 1, February 1996, pp Cho F.D.S. and G.G. Mueller, Inernaional Accouning, Second Ediion. Englewood Cliffs: renice Hall, Collins, D. W. and S.. Kohar "An Analysis of he Ineremporal and Cross-secional Deerminans of Earnings Response Coefficiens," Journal of Accouning and Economics (July 1989): Chrisie, A., "On Cross-Secional Analysis in Accouning Research", Journal of Accouning And Economics, (1987,9): Financial Accouning Sandards Board, Saemen of Financial Accouning Sandards No. 14: Financial Reporing for Segmens of a Business Enerprise. Samford, Conn.: FASB, Garrond, N. and W. Rees, "Inernaional Diversificaion and Firm Value," Journal of Business Finance and Accouning, (November/December 1998): Hayn, C., "Informaion Conen of Losses", Journal of Accouning And Economics, (1995,20):

22 Kinney, W., "Covariabiliy of Segmen Earnings and Mulisegmen Company Reurns," Accouning Review, (April 1972): Herrmann, D., and W. B. Thomas, An Analysis of Segmen Disclosures under SFAS No. 131 and SFAS No. 14, Accouning Horizons, (Sepember, 2000) vol. 14, no. 3, pp Ohlson, J., "Earnings, Book Values and Dividends in Equiy Valuaion", Conemporary Accouning Research (Spring 1995): Rajan, R.G. and L. Zingales, "Wha Do We Know abou Capial Srucure? Some Evidence from Inernaional Daa," Journal of Finance (December 1995): Saudagaran, S., "Discussion of A Tes of he Use of Geographical Segmen Disclosure," Journal of Accouning Research, (Supplemen 1993): Securiies and Exchange Commission, Securiies Exchange Ac of 1933, Release No , Chicago: Commerce Clearing House, July Securiies and Exchange Commission, Securiies Exchange Ac of 1933, Release No , Chicago: Commerce Clearing House, February Securiies and Exchange Commission, Securiies Exchange Ac of 1934, Release No , Chicago: Commerce Clearing House, Ocober Securiies and Exchange Commission, General Revision of Regulaion SX, Federal Regiser Vol. 48, No. 188, Sepember 23, Swaminahan, S., "The Impac of SEC Mandaed Segmen Daa on rice Variabiliy and Divergence of Beliefs," Accouning Review, (January 1991): Thomas, W., "The Value-relevance of Geographic Segmen Earnings Disclosures Under SFAS 14", forhcoming, Journal of Inernaional Finance and Accouning, Whie, H. "A Heeroscedasiciy-Consisen Marix Esimaor and a Direc Tes for Heeroscedasiciy." Economerica (May 1098):

23 Table 1 Sample Characerisics for Ausralia, Canada, and he Unied Kingdom from ANEL A : Ausralia - Quariles of ooled Sample Variable Q1 Median Q3 N Asses (A$MM) , Marke Value (A$MM) , Foreign Revenues (% of oal) Foreign Operaing Inc. (% of oal) Foreign Asses (% of oal) ANEL B: Canada - Quariles of ooled Samples Variable Q1 Median Q3 N Asses (C$MM) , Marke Value (C$MM) , Foreign Revenues (% of oal) Foreign Operaing Inc. (% of oal) Foreign Asses (% of oal) ANEL C: Unied Kingdom - Quariles of ooled Samples Variable Q1 Median Q3 N Asses ( MM) Marke Value ( MM) Foreign Revenues (% of oal) Foreign Operaing Inc. (% of oal) Foreign Asses (% of oal) Table Noes: Asses and marke values are drawn from Compusa (Global Vanage). Foreign asses, revenues and earnings are from he Disclosure. Q1 is he 25 h percenile, Median is he 50 h percenile, and Q3 is he 75 h percenile. N is he number of firm-year observaions. 21

24 Table 2 Sample Characerisics for Ausralia, Canada, and he Unied Kingdom from ANEL A : Ausralia Sample Characerisics Variable Mean Median Min Max Q3-Q1 Correlaion N TOES / FOES / MAR (%) ANEL B: Canada Sample Characerisics Variable Mean Median Min Max Q3-Q1 Correlaion N TOES / FOES / MAR (%) ANEL C: Unied Kingdom Sample Characerisics Variable Mean Median Min Max Q3-Q1 Correlaion N TOES / FOES / MAR (%) Table Noes: The able displays disribuion saisics for he variables used in he regression analysis. TOES is he change in oal operaing income per share from fiscal year -1 o fiscal year ; FOES is he change in he foreign componen of operaing income per share from fiscal year -1 o fiscal year. -1 is he price of he firm s share a he beginning of he reurn accumulaion period. This price is used o normalize all he accouning variables defined above. MAR is he cumulaive abnormal reurn over he welve-monh period ending in he monh of he required public release of accouning daa. Q3 - Q1 is he iner-quarile range and N is he number of firmyear observaions. All correlaion values are saisically significan a he 1% level. 22

25 Table 3 Cross-Secional Associaion Regressions of Marke Adjused Reurns and Changes in Earnings by Counry anel A: Toal Operaing Earnings and Non-Operaing Earnings MAR TOES = α0 + α1 + α -1 2 TNOES -1 + ε 1 α 0 α 1 α 2 Adj. R 2 N Ausralia (0.04) (6.07) a (5.56) a Canada (-1.60) b (9.07) a (1.91) b Unied Kingdom (-4.45) a (9.32) a (3.01) a ANEL B: Toal Operaing Income, Non-Operaing Earnings and Foreign Earnings TOES TNOES MAR = δ0 + δ1 + δ FOES δ3 + ε - 1 δ 0 δ 1 δ 2 δ 3 Adj R 2 N Ausralia (-0.21) (5.62) a (5.18) a (1.26) Canada (-1.78) b (7.25) a (1.99) b (3.22) a Unied Kingdom (-4.51) a (8.17) a (3.00) a (1.89) b 3 Table Noes: anel A: MAR is he cumulaive marke adjused reurn for firm i over he 12-monh period. TOES is he change in oal operaing earnings for firm i from fiscal year -1 o fiscal year ; TNOES is he change in non operaing income (he difference beween a firm s operaing income and ne income) for firm i from fiscal year -1 o fiscal year and FOES is he change in he foreign componen of earnings for firm i from fiscal year -1 o fiscal year. -1 is he price of firm i s equiy a ime -1, used as a normalizaion for he income daa. All regression are OLS. -saisics are shown in parenheses based upon he Whie (1980) correced sandard errors. a, b and c represen saisical significance a he 1, 5, and 10 percen levels respecively for oneailed ess. 23

26 Table 4 Relaion Beween Marke Adjused Annual Reurn and Toal, Foreign, and Oher Earnings By Relaive Growh Raes of Foreign and Domesic Sales MAR TOES = δ 0 + δ 1 TNOES + δ 2 1 FOES + δ anel A: Domesic Sales Growh > Foreign Sales Growh + ε δ 0 δ 1 δ 2 δ 3 Adj R 2 N Ausralia (0.48) (5.13) a (4.44) a (-1.64) c Canada (-0.61) (5.52) a (1.27) (0.60) Unied Kingdom (-3.00) a (6.86) a (1.98) b (0.93) anel B: Foreign Sales Growh Domesic Sales Growh δ 0 δ 1 δ 2 δ 3 Adj R 2 N Ausralia (-0.99) (3.22) a (3.71) a (3.19) a Canada (-2.27) b (4.83) a (2.34) a (4.03) a Unied Kingdom (-3.58) a (4.44) a (2.27) a (1.91) b Table Noes: MAR is he cumulaive marke adjused reurn for firm i over he 12-monh period. TOES is he change in oal operaing earnings for firm i from fiscal year -1 o fiscal year ; TNOES is he change in non operaing income (he difference beween a firm s operaing income and ne income) for firm i from fiscal year -1 o fiscal year and FOES is he change in he foreign componen of earnings for firm i from fiscal year -1 o fiscal year. -1 is he price of firm i s equiy a ime -1, used as a normalizaion for he income daa. Observaions are broken down across panels based on relaive sales growh for he foreign and domesic segmen from he prior year. -saisics based upon Whie (1980) correced sandard error are repored in parenheses. a, b and c represen saisical significance a he 1, 5 and 10 percen levels respecively for one-ailed ess. 24

27 MAR Table 5 Relaion Beween Marke Adjused Annual Reurn and Toal, Foreign and Oher Earnings Changes and Levels Conrolling for Negaive Earnings for Firms By Relaive Growh Raes of Foreign and Domesic Sales TOES = ω 1 0 NEG _ FOES NEG _ TOES ω 2 1 FOES 7 1 TOES 3 1 NEG _ FOES 8 1 NEG _ TOES 4 1 TNOES 9 1 FOES TNOES + ε anel A: Domesic Sales Growh > Foreign Sales Growh Ausralia Canada Unied Kingdom ω (-1.96) b (-0.83) (-6.93) a ω (4.32) a (3.23) a (5.33) a ω (-1.27) (-0.82) (-1.19) ω (0.66) (2.97) a (7.06) a ω (0.68) 0.21 (0.56) (-3.96) a ω (-2.24) 0.23 (1.19) (-0.61) ω (0.69) 0.48 (0.80) (-2.55) a ω (0.86) (-0.16) 0.51 (1.78) b ω (-1.68) (-0.39) 0.66 (2.35) b ω (3.02) a (4.03) a (2.45) b ω (0.64) 0.00 (0.90) 0.10 (1.40) Observaions R