Accounting Conservatism and Information Asymmetry: Evidence from Taiwan

Transcription

1 Inernaonal Busness Research; Vol. 6, No. 7; 03 ISSN E-ISSN Publshed by Canadan Cener of Scence and Educaon Accounng Conservasm and Informaon Asymmery: Evdence from Tawan Juo-Len Wang Deparmen of Accounng, Chaoyang Unversy of Technology, Tachung, Tawan Correspondence: Juo-Len Wang, Deparmen of Accounng, Chaoyang Unversy of Technology, 68, Jfeng E. Rd., Wufeng Dsrc, Tachung 4349, Tawan, R. O. C. Tel: E-mal: Receved: Aprl 8, 03 Acceped: May 9, 03 Onlne Publshed: June 4, 03 do:0.5539/br.v6n7p3 URL: hp://dx.do.org/0.5539/br.v6n7p3 Absrac Ths paper adops nformaon neuraly perspecve o examne he role of earnngs conservasm o explore he effec of accounng conservasm on nformaon asymmery n Tawan. Resuls reveal ha when corporae earnngs are ncreasngly conservave, nformaon asymmery s more severe generally. In addon, when earnngs conservasm s excessve or nsuffcen, varyng effecs are produced beween earnngs conservasm levels and nformaon asymmery. Specfcally, when corporae conservasm s nsuffcen, he relaonshp beween earnngs conservasm and nformaon asymmery s sgnfcanly negave. Conversely, when accounng earnngs s much more conservave, he nfluence on nformaon asymmery s posve. The emprcal resuls of hs sudy suppor he curren developmen of fnancal sandards. The emprcal resuls also show ha when he corporaon performance demonsraes good news or bad news separaely, nvesors percepon regardng he nformaveness of accounng conservasm s dfferen. Invesors are more lkely o denfy wh relevan resuls despe an overly conservave accounng recognon f he curren perod of a corporaon provdes good news. Keywords: accounng conservasm, nformaon asymmery, fahful represenaon. Inroducon Conservasm s also known as he prudence prncple. I s defned ha accounng conservasm as a more srngen verfcaon level requred for accounng personnel when recognzng good news compared o ha requred when recognzng bad news (Basu, 997). In oher words, when accounng earnngs reflec operang resuls, he melness of recognzng economc losses exceeds ha of economc profs, resulng n asymmerc melness for recognon. In accounng sandards, examples of praccal applcaons of he conservasm prncple as an accounng ool nclude adopng he mehod of he lower of cos or ne realzable value of nvenory, mparmen esng for he value of asses, and conngency accounng. Basu (997) conended ha conservasm resuls n nconssen recognon crera for profs and losses n accounng sandards, whch resuls n a slower earnngs response o good news compared o bad news. Basu used hs dfference o measure he conservasm of fnancal saemens, and defned hs dscrepancy as he asymmerc melness of earnngs. Numerous emprcal sudes have exended Basu model and adoped varous perspecves, ncludng conrac heory, agency heory, and lgaon heory, o explan corporae demands for conservasm (Ahmed, Bllngs, Moron, & Sandford-Harrs, 00; La Fond & Was, 008; Shackelford & Shevln, 00; Was, 003a, 003b). Among hese sudes, Was (003a, 003b) adoped conrac heory perspecve o examne he conservasm demands of corporae fnancal saemen nformaon, and argued ha conservave accounng s an effecve conrac mechansm for resolvng ehcal crses caused by nformaon asymmery and lmed lably among he pares nvolved n corporae conracs. LaFond and Was (008) examned wheher nformaon asymmery beween he nsders and ousders resuled n demand for conservasm, and concluded ha conservasm can reduce earnngs manpulaon ncenves for he nsders of a corporaon, decrease agency problems among sakeholders, and furher reduce nformaon asymmery and ncrease corporae value. Ch and Wang (00) and Ch, Lu, & Wang (009) examned he relaonshp beween accounng conservasm and nformaon asymmery based on daa from Tawan. Ther fndngs suppored he assumpon ha conservave accounng can reduce he occurrence of nformaon asymmery and, hus, can be 3

2 Inernaonal Busness Research Vol. 6, No. 7; 03 employed as a corporae governance mechansm o ncrease corporae value and cash flow. In he heorecal framework of generally acceped accounng prncples, he ulmae goal for accounng nformaon s decson usefulness. Fnancal saemens provde useful nformaon for exsng and poenal nvesors, credors, and oher users o evaluae he fuure nvesmen cash flow or cred and o conduc raonal decson-makng. The Concepual Framework for Fnancal Reporng whch developed by he Inernaonal Accounng Sandards Board (IASB) and U.S. Fnancal Accounng Sandards Board (FASB)(hereafer he Boards) esablshed frmly he basc qualy characerscs of accounng are relevance and fahful represenaon. In oher words, he prncple of conservasm s no longer ncluded. However, he conservasm prncple has sgnfcanly affeced he preparaon of accounng nformaon over he long erm. Thus, n he pas 30 years, accounng pracces have ended o be more conservave (Was, 003a), emphaszng he need for he conservasm prncple. Varous subsequen sudes (Cheng, Wu, & Shue, 0) have also suppored he conrac heory perspecve of Was (003a). Har and Moore (995) conended ha agency problems are prmarly caused by nformaon asymmery beween corporae funders and fund users. Insders such as corporae managers and conrollng shareholders possess complee nformaon, whereas nvesors such as credors and ousde shareholders have nsuffcen nformaon. Therefore, he nsders may, based on moves o pursue prvae neres, engage n behavor ha s dermenal o he enre corporaon prof. Ths suaon s known as an agency conflc, whch ofen becomes more severe because of nformaon asymmery among pares nvolved n he conrac. The fahful represenaon of accounng nformaon, o some exen, eases and resolves conflcs of neres caused by nformaon asymmery. Regardng economc developmen, Tawan s currenly an emergng economy marke. Dfferences n he framework for regulaory sysems and culural envronmen backgrounds exs beween Tawan and developed counres. The exsence of prvae nformaon s more common n he markes of emergng economes; hence, ousde nvesors are more lkely o be exploed by nsders. In some degree, he developmen of marke sze s parally deermned by wheher nformaon asymmery beween nsders and ousders can be reduced. Compared o sock markes n Europe or he Uned Saes, he majory of he parcpans n Tawan sock marke are ndvdual nvesors. In hs conex, he nformaon asymmery n Tawan sock marke s exreme. Thus, for corporaons, mananng fahful represenaon n accounng nformaon and a neural perspecve nsead of overly emphaszng accounng conservasm can posvely nfluence he developmen of Tawan sock marke. Snce 999, Tawan has gradually ncreased and amended accounng sandards o converge on he Inernaonal Fnancal Reporng Sandards (IFRS). By referencng relevan IFRS regulaons, accounng bullens or repors have been conssenly ncreased and amended, and lsed, OTC, and emergng companes are requred o drecly adop he IFRS for preparaon of fnancal repors or saemens begnnng n 03. The excluson of he conservasm prncple from he basc accounng framework wll nevably and drecly nfluence nformaon qualy regardng accounng earnngs recognonn Tawanese corporaons. Basu, Huang, Msudome and Wenrop (005) assered ha conservasm exss n he accounng earnngs of Tawanese corporaons, and Ch and Wang (00) referenced Tawanese daa and obaned emprcal resuls verfyng ha conservasm reduces nformaon asymmery. However, Ch and Wang (00) adoped Basu (997) as he emprcal model framework, alhough Basu s earnngs conservasm esmaon shows ceran lmaons. Furhermore, durng he esmaon process, Basu assumed ha all frms n he ndusry are homogeneous and he frm s operang characerscs are no change over me (Khan & Was, 009). Therefore, Basu mehod could no deermne wheher varyng degrees of conservasm had an dencal nfluence on nformaon asymmery. For hs reason, hs sudy deems necessary o reexamne he nfluence ha earnngs conservasm has on nformaon asymmery n Tawan s capal marke o undersand he role of conservasm n accounng nformaon and provde references for corporaons when preparng fnancal saemens n he fuure. Invesors and he accounng sandard makers have argued ha earnngs conservasm can resul n an underesmaon of curren earnngs and an overesmaon of fuure earnngs. Ths affecs he neuraly and relevance of fnancal nformaon dsclosures and furher ncreases nformaon asymmery beween he nsders and ousders of he corporaon. The majory of prevous sudes adoped conrac heory perspecve and, hus, suppored he reenon of he conservasm prncple (e.g., Ahmed e al., 00; Ahmed & Duellman, 007; Khan & Was, 009). In conras o prevous sudes ha prmarly suppored conservave accounng o reduce managemen manpulaon of earnngs, hs sudy focuses on () he effecs ha conservasm has on nformaon asymmery under varous accounng conservasm degrees, and () conrbung leraure on conservave accounng n nformaon neuraly. Recenly, he Boards concep of he conservasm prncple or prudence has 33

3 Inernaonal Busness Research Vol. 6, No. 7; 03 been ransformed no he fahful represenaon of accounng nformaon. Consequenly, hs sudy adops nformaon neuraly perspecve o examne he role of earnngs conservasm. Prevous sudes have prmarly used lnear relaonshps o measure he correlaon beween earnngs conservasm and nformaon asymmery whn he marke, and subsequenly suggesed ha accounng conservasm can reduce nformaon asymmery, ncrease conrac effcency, and enhance he qualy of fnancal saemen nformaon (La Fond & Was, 008; Ch & Wang, 00; Ch e al., 009). Ths sudy exends he research of prevous sudes regardng he correlaon beween he conservasm and nformaon asymmery. As a resul, he nfluence ha earnngs conservasm has on corporae nformaon asymmery s analyzed, and he nfluence s assumed o be nonlnear. In oher words, hs sudy assumes ha boh excessve and nsuffcen conservasm nduce nvesor doubs, and, hus, he percepon ha he nformaveness of corporae accounng repors should be enhanced. Lsed and OTC companes n Tawan were adoped as he research samples. The emprcal resuls ndcae ha Tawan s corporae accounng earnngs are conservave. Ths sudy also employed he frm-year ndcaor of accounng conservasm esablshed by Khan and Was (009) o verfy relevan nferences. Overall, he resuls showed ha when corporae earnngs are ncreasngly conservave, nformaon asymmery s more severe. Earnngs conservasm causes nvesors o perceve ha publshed accounng nformaon s less neuraly and relevan, subsequenly wdenng he nformaon asymmery beween he nernal personnel and exernal pares of a corporaon. Ths resuls n nvesors rasng sock bd-ask spreads for self-proecon. In addon, when earnngs conservasm s excessve or nsuffcen, varyng effecs are produced beween earnngs conservasm levels and nformaon asymmery. Specfcally, when corporae conservasm s nsuffcen, he relaonshp beween earnngs conservasm and nformaon asymmery s sgnfcanly negave. Conversely, when accounng earnngs s much more conservave, he nfluence on nformaon asymmery s posve. Therefore, when corporaons recognze revenue and expense, exremy conservasm (.e., eher excessvely or nsuffcenly conservave) would enhance he nformaon asymmery beween nsders and ousders. The emprcal resuls also show ha when he corporaon performance demonsraes good news or bad news separaely, nvesors percepon regardng he nformaveness of accounng conservasm s dfferen. The remander of hs paper s organzed as follows. Secon dscusses he exan leraure on earnngs conservasm and nformaon asymmery, followed n Secon 3 by a descrpon of he daa used, an explanaon of he research desgn and presenaon of he mehods used o denfy earnngs conservasm and nformaon asymmery. The emprcal resuls are presened n Secon 4, wh he conclusons drawn from hs sudy beng presened n Secon 5.. Relaed Leraure and Hypoheses Varous hypoheses are esablshed n hs sudy o verfy he nfluence ha accounng conservasm has on nformaon asymmery beween he nernal and exernal pares of an enerprse. Frs, under perfec machng prncple, all revenues ha have occurred or can be recognzed, as well as all expenses (and losses) relevan o he revenues should be recognzed n he same perod. However, under he conservave accounng prncple, nose exss n he correlaon beween ncome and expenses. Ths nferor form of machng prncple may reduce he exen of nformaon dsclosures and ncrease he level of nformaon asymmery. Hence, he Concepual Framework for Fnancal Reporng publshed by he Boards defned and esablshed he fundamenal qualave characerscs for accounng as relevance and fahful represenaon. The Boards perceve ha when corporaons face unceranes, fnancal saemen should be evaluaed more judcously. Abdng by he convenonal conservasm prncple n such crcumsances can easly cause fnancal nformaon bases, whch conflcs wh he neuraly of he fahful represenaon characersc, leadng o volaons of nformaon qualy. Therefore, he new concepual framework publshed n 00 no longer ncluded he conservasm (or prudence) prncple. However, LaFond and Was (008) assered ha, under earnngs conservasm, revenue s no overesmaed and expenses and losses are no underesmaed; hus, a ceran level of proecon s provded for exernal pares. When he nformaon asymmery beween nsders and ousders s more severe, he exernal pares demand a greaer proecon. In oher words, corporae earnngs mus be more conservave. The emprcal resuls repored by LaFond and Was (008) ndcae ha a posve correlaon exss beween he varaons n nformaon asymmery and he varaons n earnngs conservasm of curren year. Changes n nformaon asymmery boh lead and are conemporaneous he conservasm. Furhermore, asymmery decreases n subsequen perod for he group wh he greaes ncrease n nformaon asymmery, ndcang ha nformaon asymmery changes lead conservasm, whch reduces nformaon asymmery n he followng perod. Overall, LaFond and Was (008) proposed ha conservasm can reduce he manager s ncenves and ably o manpulae earnngs, 34

4 Inernaonal Busness Research Vol. 6, No. 7; 03 mgae agency problems beween neresed pares, and furher lower nformaon asymmery and ncrease corporae value. However, because of he lmaons of Basu (997) model, prevous sudes have no drecly verfed he nfluence ha earnngs conservasm has on nformaon. The followng effecs were only ndrecly denfed: () A more severe level of nformaon asymmery beween nsder and ousder resuls n more sgnfcan earnngs conservasm for corporae fnancal saemens; and () changes n nformaon asymmery lead o an ncrease n he level of earnngs conservasm (La Fond & Was, 008; Ch & Wang, 00). Therefore, hs sudy used he earnngs conservasm ndcaor proposed by Khan and Was (009) o drecly examne he average nfluence ha conservave accounng sandards have on nformaon asymmery. An ncrease n he ransparency of corporae nformaon can elevae marke lqudy, mgae nformaon asymmery, and reduce he bd-ask spread. Easley and O Hara (004) conended ha mgang nformaon asymmery o reduce he cos of capal s one consderaon facor of corporae nformaon dsclosures. Greaer nformaveness released n corporae fnancal nformaon resuls n greaer reducons n he cos for exernal pares o oban nformaon. Ths enables capal markes o arac exernal pares and ncrease her wllngness nransacons, reduces he proporon of prvae nformed ransacons, and ndrecly decreases he occurrence of nformaon asymmery. Ths sudy adoped he Boards perspecve, whch proposes ha f he accounng sandards overly emphasze conservasm, he qualy of corporae earnngs canno reflec frms acual performance. Consequenly, he fnancal saemen releases nsuffcen level of accounng nformaon, whch mgh damage he fahful represenaon characersc and furher ncreases nformaon asymmery beween he nernal and exernal pares of a corporaon. Thus, hs sudy proposes Hypohess as follows: H: Overall, greaer earnngs conservasm n fnancal saemen resuls n more severe nformaon asymmery. Alhough earnngs conservasm may reduce he nformaveness of fnancal repor, he convenonal conservasm prncple s beleved o reduce agency coss and ncrease conrac effcency. Basu (997) assered ha because managers possess prvae nformaon regardng corporae operang performance and he asses value, and because he fuure profs of frm exss uncerany, managers may conceal nformaon harmful for he repored earnngs. When encounerng hs ype of ehcal crss, adopng conservasm prncple appropraely can reduce he doubs of exernal pares and ncrease conrac effcency. LaFond and Was (008) argued ha accounng conservasm even can be used as a corporae governance mechansm o nduce nformaon ransparency. Based on he above-menoned analyss, hs sudy proposes ha alhough ousde nvesors perceve ha excessve accounng conservasm reduces he nformaveness of fnancal saemen, hey do no desre ha frms demonsrae dencal recognon crera for revenue and expenses (and losses). By conras, nvesors ancpaean approprae level of conservasm as a form of proecon. Therefore, hs sudy expecs ha a nonlnear relaonshp exss beween accounng conservasm and nformaon asymmery. In oher words, an exremy n accounng conservasm of recognon (.e., eher excessve or nsuffcenly conservasm) ncreases he severy of nformaon asymmery. Thus, hs sudy proposes he second hypohess: H: A nonlnear relaonshp exss beween corporae earnngs conservasm and IA. The followng wo conceps are proposed n H: Ha: When corporae earnngs conservasm s exremely low, accounng conservasm negavely nfluence nformaon asymmery. Hb: When corporae earnngs conservasm s exremely hgh, accounng conservasm posvely nfluence nformaon asymmery. 3. Mehodology and Daa Source 3. Measurng Conservasm Based on he resuls of prevous sudes (Basu, 997), conservasm s defned as mmedaely recognzng possble expenses and losses and only recognzng possble or expeced profs when hey have been realzed, whch resuls n he recognon of accounng earnngs beng more mely or concurrenly sensve n reflecng bad news han good news. Basu (997) employed he dspary beween he speed of earnngs reflecng good news compared o ha for bad news and subsequenly measured accounng conservasm. Numerous sudes have adoped he asymmerc melness of earnngs model proposed by Basu (997) as a bass for measurng he level of accounng conservasm (Ball, Kohar, & Nkolaev, 00; Pae, 007; Cheng e al., 0; La Fond & Was, 008). Sock prces reflec dsclosed nformaon n he marke; herefore, Basu (997) used marke reurns as an ndcaor of good and bad news durng accounng. Equaon () shows he asymmerc melness of earnngs model proposed by Basu (997): 35

5 Inernaonal Busness Research Vol. 6, No. 7; 03 X, / P, 0 DR, R, 3R, DR,, () where X, /P - represens he connung operang ncome afer ax for frm a year dvded by he equy marke value a he begnnng of year, wh he purpose of deflang he ne ncome or prof usng marke value; R ndcaes buy and hold reurn whch s calculaed from May of year o Aprl 30 of year + (Noe ); and DR, s a dummy varable. When R, s a negave value, DR = ; oherwse, DR = 0. Accordng o Eq. (), when he marke response reflecs posve news (.e., R>0), he response coeffcen of earnngs for good news s β, and when he marke response reflecs bad news (.e., R<0), he response coeffcen of earnngs for good news s β +β 3.β 3 denoes he dspary beween he accounng recognon of good news compared o bad news. Accordng o Basu (997), f earnngs conservasm exss n accounng recognon, β 3 wh be greaer han 0. Alhough he mos popular mehod for esmang earnngs conservasm s Basu s asymmerc melness of earnngs mehod (Ryan, 006), hs esmaon mehod has ceran lmaons. For example, durng he esmaon process, he conservasm measure s esmaed eher for an ndusry-year usng a cross-secon of frms n he ndusry or for a frm usng a me-seres of frm-years. I mples ha conservasm whn he same year for all frms n he ndusry are homogeneous or assumes ha he frm s operang characerscs are saonary (Khan & Was, 009); however, hs s mpossble. To resolve hese problems, Khan and Was (009) broadened he assumpon of homogeney and saonary n Basu s mehod and developed a frm-year conservasm measure. Thus, hs sudy adoped he mehod proposed by Khan and Was (009) as an ndcaor for measurng accounng conservasm. The Khan and Was (009) mehod s prmarly based on Basu s model (997), wh exended frm-year conservasm ndcaors comprsng corporae sze (oal asses), developmen poenal (marke-o-book rao), and fnancal leverage (deb rao). To esmae he melness wh whch frms reflec boh good news and conservasm, Khan and Was (009) defned he melness of good news (G_score) and he ncremenal melness of bad news (C_score) each year as lnear funcons of he hree frm-specfc characerscs each year: G _ score, SIZE, 3MB, 4LEV, C _ score, 3 SIZE, 3MB, 4LEV, (a) (b) β represens he response o good news of he earnngs n Eq. (), β 3 ndcaes he dspary n responses oward good and bad news for earnngs n Eq. (), SIZE, denoes he logarhmc value for he oal asses of he corporaon, MB, s he frm marke-o-book rao, and LEV, symbolzes he deb rao. Snce he equaons are esmaed from annual cross-seconal regressons, all corporaons possess dencal μ and λ values n he same year; however, μ and λ vary n dfferen years. By subsung Eqs. (a) and (b) no he Basu (997) esmaon model represened n Eq. (), he annual cross-secon regresson model s: X / P DR R ( SIZE MB LEV ) 0 R DR ( SIZE MB LEV ) ( SIZE MB LEV DR SIZE DR MB DR LEV ) Afer organzng Eq. (3), he followng regresson esmaon s obaned: X / P DR 0 SIZE MB LEV 3 ( DR SIZE ) ( DR MB ) ( DR LEV ) R ( R SIZE ) ( R MB ) ( R LEV ) ( R DR ) ( R DR SIZE ) ( R DR MB ) ( R DR LEV ) Usng a cross-seconal model o esmae Eq. (4), he annual μ and λ values can be obaned. By combnng hese values wh he frm sze, marke-o-book rao, and deb rao for each year, and subsung he oucome no Eqs. (a) and (b), he C_score conservasm measure s acqured. A greaer C_score ndcaes more conservave n corporae fnancal reporng (3) (4) 36

6 Inernaonal Busness Research Vol. 6, No. 7; Measurng Informaon Asymmery Despe nernal personnel (e.g., managers) and exernal pares (e.g., mnory shareholders) possessng an equal amoun of nformaon regardng he macro marke, f nformaon asymmery exss for ndvdual corporaons, nernal personnel hold nformaon ha exernal pares lack. Hence, n he ransacon marke, marke makers recognze an opporuny for exploave behavor of nsders and subsequenly ncrease he nformed ransacon probably. However, marke makers are no sure he amoun of prvae nformaon ha he nernal personnel have and he level of exploaon. I would resuls n hgher radng coss; hus marke makers end o creae greaer quoed spreads and proec hemselves. Glosen and Mlgrom (985) assered ha he marke makers bd-ask spreads ncrease wh he number of nformed ransacons, reflecng a rse n he level of nformaon asymmery. Therefore, hs sudy employed he daly las bd and las offer prce o calculae he annually average bd-ask spread of he ndvdual frms as he measure of he level of nformaon asymmery (IA): IA, annually average of a, b, (5) ( a, b, ) / Where a and b denoe he daly las bd prce and las ask prce a year for frm. 3.3 The Models Generally, corporae fnancal nformaon s announced lags behnd marke response. To examne he nfluence ha he level of earnngs conservasm has on nformaon asymmery, hs sudy mached he conservasm of prevous perods o he nformaon asymmery ndcaor of he curren perod for he sample frms. For hs, he followng model was employed: IA, 0 C _ score, PRICE, 3VOLUME, 4DEV, (6) Where IA, s he annually average of las bd-ask spread; C_score, s Khan and Was (009) measure of conservasm; PRICE, represens he naural logarhm of he annual average daly closng prce, VOLUME, represens he naural logarhm of he annual average daly radng volume, and DEV, denoes he annual devaon of he daly rae or reurn. Accordng o prevous sudes (Soll, 978, 000; Van Ness, Van Ness, & Warr, 00), he bd-ask spread s affeced by prce, ransacon volume, and volaly n rae of reurn; herefore, we conrolled he relevan varables. If accounng conservasm can reduce he level of nformaon asymmery beween nsders and ousders, he regresson coeffcen α of he C_score n Eq. (5) should be sgnfcanly negave, and vce versa. To comprehensvely descrbe he effec of accounng conservasm on nformaon asymmery, hs sudy furher observed he nfluence ha varyng levels of accounng conservasm have on IA. The C_score was arranged n sequenal order from hgh o low and dvded no four groups. Subsequenly, hs sudy examned wheher he nfluences of frm conservasm on IA were dencal across he C_Score quarers. Furhermore, nvesor percepons regardng he level of conservasm for a corporaon when recevng good news or bad news may be dfferen. For example, when bad news occurs for a corporaon, f accounng recognon s overly conservave, nvesors may perceve ha corporaons are akng a bg bah and opporunscally explong ousders. Conversely, when good news occurs, f corporae accounng recognon s oo conservave, nvesors are more lkely o perceve ha corporaons are cauous raher han concealng nformaon. So long as corporaons provde good news, nvesors can even perceve exreme conservasm as accepable and do no vew conservasm as reducng he nformaveness of corporae fnancal saemens. In oher words, nvesor responses and accepance regardng he nformaveness of hese wo scenaros may dffer compleely. Therefore, n addon o conducng a general analyss of all samples, he samples were dvded no subsamples of good news (.e., R>0) and bad news (.e., R<0), and he effec of earnngs conservasm on IA was examned. 3.4 Daa and Sample Selecon Lsed or OTC companes n Tawan were adoped as he research subjecs of hs sudy o analyze he correlaon beween corporae earnngs conservasm and nformaon asymmery n Tawan beween 00 and 0. Daa were sourced from he fnance and share prce daabase of he Tawan Economc Journal (TEJ). For he reason ha marke response s ahead of fnancal nformaon announcemen, hs sudy mached he prevous year C_score wh IA of curren year for ndvdual sample frm o examne he nfluence ha he level of earnngs conservasm has on nformaon asymmery. Specfcally, C_score from year 00 o 00 was mached wh IA and conrol varables from year 00 o 0. The fnancal, nsurance, secures ndusres and he sample wh mssng daa were excluded. Observaons fallng n he op or boom % of share prce or fnancal daa values 37

7 Inernaonal Busness Research Vol. 6, No. 7; 03 were excluded o reduce he effecs of oulers on he emprcal resuls. The samples employed oaled 6,387 observaon values. 4. Emprcal Resuls 4. Descrpve Sascs Table shows he descrpve sascs of he prmary varables. The means for X/P, R, and he C_score were 8.733%, 5.447%, and 0.68, respecvely. The marke prces of he sample frms were approxmaely.549 mes ha of her book value, he mean for he deb rao was approxmaely 43%, and he bd-ask spread was approxmaely 0.608% of he share prce. Table. Descrpve sascs of he varables Varables Descrpve Sascs N Mean Max. Mn. Sd. Dev. Measuremen ndexes of conservasm X/P (%) 6, R (%) 6, C_score 6, MB 6, SIZE 6, LEV (%) 6, Informaon asymmery and conrol varables IA (%) 6, PRICE 6, VOLUME 6, DEV 6, Noes: X, /P - represens he connung operang ncome afer ax for frm a year dvded by he equy marke value a he begnnng of year ; R, ndcaes buy and hold reurn whch s calculaed from May of year o Aprl 30 of year +; C_score s Khan and Was (009) measure of conservasm; SIZE denoes he logarhmc value for he oal asses of he corporaon, MB, s he frm marke-o-book rao, and LEV, symbolzes he deb rao. IA s he annually average of las bd-ask spread; PRICE represens he naural logarhm of he annual average daly closng prce, VOLUME represens he naural logarhm of he annual average daly radng volume, and DEV denoes he annual devaon of he daly rae or reurn. Table shows he Pearson and Spearman correlaon coeffcen marx of he varables, hereby provdng some basc analyss of he correlaon beween varables. C_score and nformaon asymmery (IA) exhb a posve correlaon, whch ndcaes ha corporaons wh greaer accounng earnngs conservasm show greaer endences oward IA n he followng perod. In addon, IA exhbs a sgnfcanly negave correlaon wh share prce (PRICE) and ransacon volume (VOLUME), and exhbs a posve correlaon wh he level of varaon n share prce raes of reurn (DEV). These resuls are dencal o hose repored by Demsez (968) and Soll (978). Table. Correlaon coeffcens ab X/P R C_score MB SIZE LEV IA PRICE VOLUME DEV X/P 0.89 *** *** 0.36 *** 0.04 *** *** -0.3 *** 0.97 *** *** -0.9 *** R *** *** 0.3 *** *** *** 0.68 *** 0.55 *** *** C_score *** 0.37 *** -0.5 *** * *** *** -0.7 *** *** *** MB *** 0.65 *** -0.6 *** 0.05 *** -0.9 *** *** 0.65 *** *** * SIZE *** *** *** *** 0.75 *** *** 0.08 *** 0.77 *** LEV *** *** *** 0.78 *** *** -0.3 *** 0.0 *** 0.34 *** IA -0.7 *** *** *** *** *** 0.03 ** *** *** 0.89 *** PRICE 0.5 *** 0.70 *** *** 0.70 *** 0.3 *** -0.7 *** *** 0.84 *** VOLUME *** 0.43 *** *** *** *** 0.7 *** *** 0.95 *** 0.4 *** DEV -0.3 *** ** *** *** 0.4 *** *** Noes: a The sample comprses of 6,387 frm-year observaons. *** ndcaes sgnfcance a he % level; ** ndcaes sgnfcance a he 5% level; and * ndcaes sgnfcance a he 0% level (wo-al es). Pearson correlaons are presened below he dagonal and Spearman correlaons are presened above he dagonal. b The varables are defned n Table. 38

8 Inernaonal Busness Research Vol. 6, No. 7; The Level of Earnngs Conservasm n Tawan beween 00 and 00 Frsly, hs sudy employed he asymmerc melness of earnngs model (Basu, 997) o examne he conservasm degree for lsed and OTC corporaons n Tawan. The resuls showed ha when reurns are posve (.e., good news), he response of earnngs o good news s, and when reurns are negave (bad news), he response of earnngs o bad news s. The dfference n he response of earnngs oward good news and bad news s represened by. If conservasm exss n corporae earnngs, s sgnfcanly greaer han zero. Table 3 lss he level of annual earnngs response oward posve and negave raes of reurn beween 00 and 00(Noe ). Excludng 003 and 006, he earnngs of Tawan s lsed and OTC companes were conservave regardng accounng earnngs, whch ndcaes ha earnngs conservasm commonly exss n Tawan s corporae accounng sysem. Table 3. Accounng conservasm rends of Tawan lsed and OTC companes X, / P, 0 DR, R, 3R, DR,, Year Obs. β β Adj.R (4.430) *** (0.650) (0.90) (.670) * (9.750) *** (-0.300) (3.80) *** (.900) ** (0.60) *** (-.50) (.600) (0.730) (3.0) *** (0.330) (3.590) *** (4.860) *** (0.0) *** (-.450) (6.00) *** (3.80) *** (9.70) *** (-.970) ** (3.00) *** (.060) (.870) *** (0.700) (4.40) *** (3.50) *** (4.600) *** (0.730) (-0.400) (3.880) *** (4.50) *** (-0.00) (6.780) *** (.580) ** (.390) *** (0.90) (6.70) *** (4.480) *** Noes: *** ndcaes sgnfcance a he % level; ** ndcaes sgnfcance a he 5% level; and * ndcaes sgnfcance a he 0% level (wo-al es). X/P represens he connung operang ncome afer ax dvded by he equy marke value a he begnnng of year; R ndcaes buy and hold reurn whch s calculaed from May of year o Aprl 30 of year +. DR s a dummy varable, when R s a negave value, DR = ; oherwse, DR = Effec of Conservasm on Informaon Asymmery Ths sudy nvesgaed wheher earnngs conservasm causes nvesors o perceve ha publshed accounng nformaon s less melyness and relevan, and wheher hs ncreases nformaon asymmery beween nsders and ousders, subsequenly nducng sock bd-ask spreads for self-proecon. The emprcal resuls are lsed n Table 4. Table 4ndcaes daa regardng H and shows he regresson resuls for all samples, as well as curren year good news and curren year bad news subsamples, obaned usng Eq. (6). In Table 4, good news subsamples nvolve corporaons wh posve raes of reurn for he annual buy-and-hold of he year-(r>0), whereas R < 0 represen bad news subsamples. If H s suppored, nformaon asymmery (IA) becomes more severe as he earnngs conservasm (C_score) ncreases. Specfcally, he coeffcen for he C_score should sgnfcanly exceed zero. As shown n Table 4, he regresson resuls for all samples and he bad news subsamples exhb sgnfcanly posve coeffcens for he C_score, ndcang ha accounngs conservasm and nformaon asymmery possess a posve correlaon. Noably, for he good news subsamples, he C_score coeffcens were non-sgnfcan, ndcang ha when corporaons show good news n he curren year, nvesors are less sensve o he degree of accounng conservasm. The above resuls parally suppor H, greaer earnngs conservasm resuls n more severe nformaon asymmery, especally for he frms wh bed news. 39

9 Inernaonal Busness Research Vol. 6, No. 7; 03 Table 4. Informaon asymmery and accounng conservasm IA, 0 C _ score, PRICE, 3VOLUME, 4DEV, All samples Good news (R>0) Bad news (R<0) Varables coeff. OLS -sascs coeff. OLS -sascs coeff. OLS -sascs Inercep *** *** *** C_score *** *** PRICE *** *** *** VOLUME *** *** *** DEV *** *** *** ADJ R F N 6,387 3,65,77 Noe: Good news subsamples nvolve corporaons wh posve buy-and-hold reurn (R > 0) n year-, whereas R < 0 represen bad news subsamples. Dependen varable IA s he annually average of las bd-ask spread. C_score s Khan and Was (009) measure of conservasm; PRICE represens he naural logarhm of he annual average daly closng prce; VOLUME represens he naural logarhm of he annual average daly radng volume; and DEV denoes he annual devaon of he daly rae or reurn. *** ndcaes sgnfcance a he % level; ** ndcaes sgnfcance a he 5% level; and * ndcaes sgnfcance a he 0% level. To verfy Ha and Hb, ha s, wheher dfferences exs n he correlaons beween hgh or low earnngs conservasm, hs sudy dvded he conservasm ndcaor (C_score)no four groups (Q-Q4) organzed n sequenal order from low o hgh. Q represens corporaons wh he lowes C_score; means he accounng conservasm level of frms n Q s low. By conras, Q4 comprsed corporaons wh he hghes C_score; hus, he sample frms n Q4 adoped hgher accounng conservasm creron. Q o Q4 were used o esmae Eq. (6) respecvely and o deermne he effecs of dfferen conservasm levels havng on IA. The emprcal resuls for all samples (Table 5) ndcae ha he IA average ncreased from 0.55% n Q o 0.76% n Q4. Ths fndng also suppors H: earnngs conservasm s posvely correlaed wh IA. In addon, for corporaons caegorzed n he group wh he lowes conservasm level (Q), he correlaon beween earnngs conservasm and IA s sgnfcanly negave. Conversely, he correlaon beween he wo s sgnfcanly posve for corporaons n Q4. Therefore, exremy conservasm (.e., eher excessvely or nsuffcenly conservave) ncreases he severy of nformaon asymmery beween nsders and ousders. Table 5. The effecs ha dfferen conservasm levels have on IA all samples IA, 0 C _ score, PRICE, 3VOLUME, 4DEV, Q Q Q3 Q4 Varables coeff. OLS -sascs coeff. OLS -sascs coeff. OLS -sascs coeff. OLS -sascs Inercep *** *** *** *** C_score *** *** PRICE *** *** *** *** VOLUME *** *** *** *** DEV *** *** *** *** ADJ R F N,596,597,597,597 Mean of IA (%) Noe: C_score s Khan and Was (009) measure of conservasm; PRICE represens he naural logarhm of he annual average daly closng prce; VOLUME represens he naural logarhm of he annual average daly radng volume; and DEV denoes he annual devaon of he daly rae or reurn. *** ndcaes sgnfcance a he % level; ** ndcaes sgnfcance a he 5% level; and * ndcaes sgnfcance a he 0% level. The emprcal resuls of usng good news and bad news subsamples o analyze Eq. (6) are shown n Tables 6 and 7 respecvely. The esmaon resuls ndcae ha for corporaons n Q, he correlaon beween earnngs conservasm and IA s sgnfcanly negave. Conversely, he correlaon beween he wo s sgnfcanly posve when a corporaon s n Q4.Hence, exremy conservasm (.e., eher excessvely or nsuffcenly conservave) ncreases he severy of nformaon asymmery beween he exernal pares and nernal 40

10 Inernaonal Busness Research Vol. 6, No. 7; 03 personnel of a corporaon. The resul s smlar o Table 5. In oher words, f he accounng nformaon s more neural, he correlaon beween earnngs conservasm and nformaon asymmery s no sgnfcan; hus, Ha and Hb are suppored. Ths resul ndrecly confrms he Boards perspecve on excludng he conservasm prncple. In Table 6, he mean of he IA ncreased from 0.549% n Q o 0.594% n Q4. Table 7 shows IA mean ncrease from 0.495% n Q o 0.95% n Q4. In comparson, he speed of IA ncrease for he bad news subsample s greaer han ha of he good news subsample. Hence, f he curren perod of a corporaon provdes bad news and he corporae accounng recognon s overly conservave (Q4 group), ousde nvesors may suspec ha corporaons are akng a bg bah, whch negavely mpacs he nformaveness of he publshed fnancal saemens. Conversely, f he curren perod of a corporaon provdes good news, nvesors are more lkely o denfy wh relevan resuls despe an overly conservave accounng recognon. Ths demonsraes he asymmery of nvesor percepons regardng he accounng nformaveness of good and bad news. Table 6. The effecs ha dfferen conservasm levels have on IA good news (R>0) IA, 0 C _ score, PRICE, 3VOLUME, 4DEV, Q Q Q3 Q4 Varables coeff. OLS -sascs coeff. OLS -sascs coeff. OLS -sascs coeff. OLS -sascs Inercep.990.*** *** *** *** C_score *** ** *** PRICE *** *** *** *** VOLUME *** *** *** *** DEV *** *** *** *** ADJ R F N mean of IA(%) Noe: C_score s Khan and Was (009) measure of conservasm; PRICE represens he naural logarhm of he annual average daly closng prce; VOLUME represens he naural logarhm of he annual average daly radng volume; and DEV denoes he annual devaon of he daly rae or reurn. *** ndcaes sgnfcance a he % level; ** ndcaes sgnfcance a he 5% level; and * ndcaes sgnfcance a he 0% level. Table 7. The effecs ha dfferen conservasm levels have on IA bed news (R<0) IA, 0 C _ score, PRICE, 3VOLUME, 4DEV, Q Q Q3 Q4 Varables coeff. OLS -sascs coeff. OLS -sascs coeff. OLS -sascs coeff. OLS -sascs Inercep *** *** *** *** C_score *** * *** PRICE *** *** *** *** VOLUME *** *** *** *** DEV *** *** 0.66 *** *** ADJ R F N mean of IA(%) Noe: C_score s Khan and Was (009) measure of conservasm; PRICE represens he naural logarhm of he annual average daly closng prce; VOLUME represens he naural logarhm of he annual average daly radng volume; and DEV denoes he annual devaon of he daly rae or reurn. *** ndcaes sgnfcance a he % level; ** ndcaes sgnfcance a he 5% level; and * ndcaes sgnfcance a he 0% level. 5. Concluson Fnancal saemens n Tawan are sgnfcanly nfluenced by accounng conservasm. Prevous sudes have 4

11 Inernaonal Busness Research Vol. 6, No. 7; 03 explaned corporaons demands for conservasm from numerous perspecves, ncludng conrac heory, agency heory, and acon heory. These heores suppor ha conservasm s posve and benefcal for decson-usefulness. However, fnancal saemens mus possess he qualy characersc of fahful represenaon whch ncludes neuraly. When he operang performance of corporaons exhbs unceranes, he conservasm prncple may resul n he underesmaon of ne asses, whch conflcs wh he qualy characersc of fahful represenaon. Therefore, he Boards excluded he conservasm prncple from he basc framework of accounng. Ths sudy adops he perspecve of nformaon beween nsders and ousders o examne he value of he exsence of earnngs conservasm. We ancpae ha he resuls of hs sudy provde emprcal evdence ha suppors he curren developmen of fnancal sandards. Ths sudy adoped lsed and OTC corporaons n Tawan as research samples. In summary, he resuls ndcaed ha for corporaons n Tawan, when earnngs conservasm s more sgnfcan, nformaon asymmery becomes ncreasngly severe. In addon, when earnngs conservasm s exreme or nsuffcen, s correlaon wh nformaon asymmery produces varyng effecs and nfluences. Specfcally, when he conservasm level of accounng nformaon s low, he correlaon beween earnngs conservasm and nformaon asymmery s sgnfcanly negave, whereas he nfluence on nformaon asymmery s sgnfcanly posve when he conservasm n corporae earnngs s hgh. Therefore, when corporaons recognze revenue and expenses, an exremy n conservasm causes nvesors o perceve ha he nformaveness of he fnancal saemens publshed by a corporaon s nsuffcen. As a resul, nformaon asymmery s ncreased and nvesors hen ncrease sock bd-ask spreads for self-proecon. The resuls also ndcae ha when corporaons provde bad news n he curren perod and corporae accounng recognon s overly conservave, ousde nvesors may perceve ha corporaons are akng a bg bah, whch negavely affecs he nformaveness of he publshed fnancal saemens. Conversely, f he curren perod of a corporaon provdes good news, nvesors are more lkely o denfy wh relevan resuls despe an overly conservave accounng recognon. Ths demonsraes he asymmery of nvesor percepons regardng he accounng nformaveness of good and bad news. Fnally, he emprcal resuls of hs sudy suppor he curren developmen of fnancal sandards. However, he formulaon of fnancal accounng sandards possesses a wde scope of nfluence. And hs sudy conends ha an exremy n conservasm ncreases he nformaon asymmery beween he nernal personnel and exernal pares of a corporaon. Thus, he boards should endeavor o acheve balance whle emphaszng fahful represenaon. References Ahmed, A. S., & Duellman, S. (007). Accounng conservasm and board of drecor characerscs: An emprcal analyss. Journal of Accounng and Economcs, 43, hp://dx.do.org/0.06/j.jacceco Ahmed, A. S., Bllngs, B. K., Moron, R. M., & Sandford-Harrs, M. (00). The role of accounng conservasm n mgang bondholder-shareholder conflcs over dvdend polcy and n reducng deb coss. The Accounng Revew, 77, hp://dx.do.org/0.308/accr Ball, R., Kohar, S. P., & Nkolaev, V. V. (03). Economercs of he basu asymmerc melness coeffcen and accounng conservasm. Chcago Booh Research Paper, No hp://dx.do.org/0.39/ssrn Basu, S. (997). The conservasm prncple and he asymmerc melness of earnngs. Journal of Accounng and Economcs, 4, hp://dx.do.org/0.06/s065-40(97)0004- Basu, S., Huang, A., Msudome, T., & Wenrop, J. (005). Tmelness and conservasm of awanese earnngs. Asa-Pacfc Journal of Accounng and Economcs,, hp://dx.do.org/0.080/ Cheng, C. H., Wu, P. C., & Shue, F. J. (0). Accounng conservasm and earnngs perssence: A consderaon of conservasm ndex componens. The Inernaonal Journal of Accounng Sudes, 5, (n Chnese) Ch, W., & Wang, C. (00). Accounng conservasm n a seng of nformaon asymmery beween majory and mnory shareholders. The Inernaonal Journal of Accounng, 45, hp://dx.do.org/0.06/j.nacc Ch, W., Lu, C., & Wang, T. (009). Wha affecs accounng conservasm: A corporae governance perspecve. Journal of Conemporary Accounng and Economc, 5, hp://dx.do.org/0.06/j.jcae

12 Inernaonal Busness Research Vol. 6, No. 7; 03 Demsez, H. (968). The coss of ransacng. Quarerly Journal of Economcs, 8, hp://dx.do.org/0.307/8844 Easley, D., & O Hara, M. (004). Informaon and he cos of capal. The Journal of Fnance, 59, hp://dx.do.org/0./j x Glosen, L., & Mlgrom, P. (985). Bd, ask and ransacon prces n a specals marke wh heerogeneously nformed raders. Journal of fnancal Economcs, 4, hp://dx.do.org/0.06/ x(85) Har, O., & Moore, J. (995). Deb and senory: an analyss of he role of hard clams n consranng managemen. Amercan Economc Revew, 85, Khan, M., & Was, R. L. (009). Esmaon and emprcal properes of a frm-year measure of accounng conservasm. Journal of Accounng and Economcs, 48, hp://dx.do.org/0.06/j.jacceco LaFond, R., & Was, R. L. (009). The nformaon role of conservasm. The Accounng Revew, 83, hp://dx.do.org/0.308/accr Pae, J. (007). Unexpeced accruals and condonal accounng conservasm. Journal of Busness Fnance & Accounng, 34, hp://dx.do.org/0./j x Ryan, S. (006). Idenfyng condonal conservasm. European Fnancal Revew, 5, hp://dx.do.org/0.080/ Shackelford, D. A., & Shevln, T. (00). Emprcal ax research n accounng. Journal of Accounng and Economcs, 3, hp://dx.do.org/0.06/s065-40(0)000- Soll, H. R. (978). The supply of dealer servces n secures markes. Journal of Fnance, 33, hp://dx.do.org/0./j b0053.x Soll, H. R. (000). Frcon. Journal of Fnance, 55, hp://dx.do.org/0./ Van Ness, B. F., Van Ness, R. A., & Warr, R. S. (00). Is he adverse selecon componen really hgher on he NYSE/AMEX han on he Nasdaq? Journal of Busness Fnance & Accounng, 9, hp://dx.do.org/0./ Was, R. L. (003a). Conservasm n accounng par I: Explanaons and mplcaons. Accounng Horzons, 7, 07-. hp://dx.do.org/0.308/acch Was, R. L. (003b). Conservasm n accounng par II: Evdence and research opporunes. Accounng Horzons, 7, hp://dx.do.org/0.308/acch Noes Noe. Accordng o regulaons, fnancal saemens for lsed or OTC companes on Tawan s sock marke mus be made publc whn four monhs afer he end of he fscal year. Snce 0, hs perod was reduced o hree monhs. Noe. Table 3 shows he samples of all lsed and OTC companes n Tawan afer he mssng values and oulers of Eq. () varables were deduced; hus, he number of samples n Table 3 exceeds ha of Table. Copyrghs Copyrgh for hs arcle s reaned by he auhor(s), wh frs publcaon rghs graned o he journal. Ths s an open-access arcle dsrbued under he erms and condons of he Creave Commons Arbuon lcense (hp://creavecommons.org/lcenses/by/3.0/). 43