Suggested running head: Determinants of COE of M sian Firms

Transcription

1 CS3-H2 The 12th Internatonal Conventon of the East Asan Economc Assocaton LG Conventon Hall, Internatonal Educaton Buldng, Ewha Womans Unversty, Seoul, 2-3 October 2010 Conventon Theme: Asa and the Global Economc Recovery DETERMINANTS OF THE COST OF EQUITY OF MALAYSIAN FIRMS Authors: FOONG SWEE SIM Department of Appled Statstcs Faculty of Economcs and Admnstraton Unversty of Malaya Kuala Lumpur Malaysa Emal: Telephone: Fax No.: GOH KIM LENG Department of Appled Statstcs Faculty of Economcs and Admnstraton Unversty of Malaya Kuala Lumpur Malaysa Emal: Telephone: Fax No.: Suggested runnng head: Determnants of COE of M san Frms 1

2 Abstract A number of Captal Asset Prcng Model (CAPM) and ts varants are consdered n ths paper for explorng the determnants of the cost of equty of Malaysan frms. The sem-devaton approach s shown to yeld the hghest explanatory power on the returns of frms. The estmates of cost of equty from the sem-devaton approach were regressed on a lst of potental determnants n a panel model analyss. The results show that frm sze, book-to-market rato, payout rato and return to equty are negatvely related to the cost of equty. Keywords: CAPM, cost of equty, determnant, frm. JEL classfcaton codes: G12, G31 1. Introducton Cost of equty, or more commonly known as requred rate of return, s the mnmum return needed to compensate the common stockholders of a frm. Accurate estmaton of the cost of equty s vtal for makng many fnancal decsons, for example captal structure choce, captal budgetng analyss, performance assessment, and frm valuaton. There are varous methods n whch a frm s cost of equty can be obtaned, for example, the dscountng cash flow method and the Captal Asset Prcng Model (CAPM), there s no consensus n the lterature as to whch s the best model. Survey evdence (see for example, Bruner et al., 1998 and McLaney et al., 2004) ndcates that although nvestors use a wde varety of asset prcng model for calculatng cost of equty, they favour the CAPM. In the quest to search for the best asset prcng model for the case of Malaysa, the current paper works wth the CAPM snce t s most wdely appled among practtoners. The choce bears some merts snce practtoners are the end users who rely on the model for fnancal decson makng. To begn wth, we have the Local CAPM (LCAPM) whch s sutable for markets that are segmented. Then, there s the Global CAPM (GCAPM) desgned to capture the varaton n frm s returns that s explaned by global market returns. Accordng to Stulz (1981), GCAPM should be used for markets that are fully ntegrated, and the cost of equty estmates usng LCAPM and GCAPM should be sgnfcantly dfferent. However, Mshra and O Bren (2001), Koedjk et al. (2002), Harrset al. (2003), and Koedjk and van Djk (2004) seemed to share the vew that the local and global asset prcng models, on average, do not produce substantal dfference n cost of equty estmates. Due to the beleve that emergng markets le wthn the contnuum of full ntegraton at one end and full segmentaton at the other end, specfc adjustments were made to the CAPM n order to better sut an emergng market settng. Some examples nclude Lessard s (1996) model, Godfrey and Espnosa s (1996) model, Perero s (2001) adjusted hybrd CAPM, and Damodaran s (2002) model. The present study dentfes a gap n Perero s (2006) comprehensve lst of asset prcng models for emergng markets n that they do not consder both local and global factors smultaneously. If Malaysa s partally ntegrated to the world captal market, then a model 2

3 whch consders both local and global factors mght offer greater explanatory power on frm s stock returns. Hence, better cost of equty estmates could be obtaned. To compare the performance of the varous models, we appled the commonly used R 2 and adjusted R 2. After the best model s selected, ts cost of equty estmates were then used for explorng the determnants of frm s cost of equty. The exstng lterature s lackng of studes that examne determnants of cost of equty. In most studes, explorng for cost of equty determnants s not the core objectve, but a perpheral product of the analyss on the mpact of dfferent factors such as fnancal lberalzaton (Ameer, 2007), lqudty (Ln et al., 2009), earnngs forecast (Rakow, 2010), and corporate governance (Chen et al., 2009, Guedham and Mshra, 2009) on cost of equty. In the present study, potental cost of equty determnants are dvded nto frm s fundamental-related factors (debt-to-equty, return-to-equty, sales-to-assets and payout rato) and prce-related factors (frm sze, market-to-book, stock lqudty, dvdend yeld, earnngs per share, beta, sem-devaton and standard devaton of returns). Our emprcal results reveal two nterestng features. Frst, we fnd that downsde rsk measures are better than ther standard rsk counterparts, a fndng that s n lne wth Estrada (2000, 2001, 2002). Further, model whch consders both local and global rsk factors has hgher explanatory power than model that consders only ether a sngle rsk factor. Second, contrary to the wsdom n the lterature on asset prcng, we do not fnd the book-to-market rato to be postvely related to frm s cost of equty. Instead, we fnd the book-to-market and cost of equty are negatvely related. Our result seems to support the fndngs of Dchev (1998) and Zaretzky and Zumwalt (2007) that most dstressed frms have low book-to-market ratos. Therefore, frms wth low book-to-market ratos should ncur hgher cost of equty. The rest of the paper s organzed as follows. Secton 2 presents the methodology and data. Results and dscussons are gven n Secton 3. Fnally, concludng remarks are presented n Secton Methodology and Data Ths secton s dvded nto three parts. The frst part dscusses the varous CAPM varants and two non-capm based models for estmatng frm s cost of equty. The second part provdes a lst of potental cost of equty determnants as well as ther expected relatonshp wth frm s cost of equty. The thrd part dscusses the data used n the study. 2.1 Measurng Cost of Equty Modern fnancal economcs assumes that nvestors rsk percepton of a frm s reflected n the cost of equty of the frm. Beng rsk-adverse, nvestors wll demand a hgher return when the perceved rsk s larger. Ths transforms nto a smple method for computng the frm s cost of equty by stackng up the rsk-free rate and the premum for systematc rsk, whch s the product of the beta for frm ( β ) and the benchmark market rsk premum, as follows: Cost of Equty = Rsk-Free Rates + (Rsk Measure x Market Rsk Premum) CE = R + β R R (1) or ( ) f m f 3

4 where CE represents the cost of equty for frm, 4 R f s the return on the rsk-free asset, R m s the return on the benchmark market ndex and β measures the senstvty of frm s returns to the benchmark market returns. The above settng dscounts out frm level unsystematc rsk as nvestors beleve that frm specfc rsks can be dversfed away and hence should not be ncorporated nto calculatng the cost of equty for frm. What matters n evaluatng a frm performance s by lookng at β, where conventonally t can be estmated va a CAPM where: rt = α + β ( rmt rft ) + ε t (1a) where r t s the return seres for frm, r mt s the returns for the market portfolo and r ft s the cov( r, rm ) rsk-free return seres. The parameter α represents the ntercept, and β = 2 s the σ m regresson coeffcent capturng the senstvty of frm to the market rsk. The contrbuton of the CAPM s the dea of benchmarkng the frm to the overall market or systematc rsk the comovement of frm wth the market. Ths s powerful n practse as t has avoded the tedous calculaton of modern portfolo theory that requres the extremely large portfolo covarance/correlaton matrx for establshng an effcent portfolo. By benchmarkng to the market, the calculaton s reduced from ( n 2 n) /2 to n, where n the case of 100 frms, nstead of ( ) /2 =4,950, we only need to calculate the rsk for 100 frms. Ths smplstc feature may be part of the reason for the CAPM s wdespread popularty among practtoners despte the many debates assocated wth the use of market beta as the only factor that explans varaton n stock returns Local CAPM (LCAPM) The proponents of segmented world captal market may use the LCAPM. Equaton (1) n a local settng s gven by: Cost of Equty = Rsk-Free Rates + Premum for Local Systematc Rsk CE = RF + β ( RM RF ) (2) where CE represents the cost of equty for frm, R F s the return on the rsk-free asset, R M s the return on the local market ndex and β s obtaned by regressng frm s returns on the local market returns: rt = α + β ( rmt ) + ε t (2a) where r t s the compoundng returns for frm, r Mt s the compoundng returns for the market portfolo and r Ft s the compoundng local rsk-free rates. The parameter α and β are the ntercept and coeffcent, respectvely Global CAPM (GCAPM) In an ntegrated captal market settng, the expected return s determned by the beta wth respect to the world market portfolo multpled by the world rsk premum. Extendng equaton (1) to a global settng, the GCAPM s gven by:

5 Cost of Equty = Global Rsk-Free Rates + Premum for Global Systematc Rsk G G G G CE = RF + β ( RM RF ) (3) G G G where R F s the global rsk-free rates, R M the global portfolo returns, and β s obtaned by regressng frm s returns on the world market returns: G G G G rt = α + β ( rmt ) + ε t (3a) G where r Mt s the compoundng returns for global market portfolo and r G Ft s the compoundng G G global rsk-free rates. The parameter α and β are the ntercept and coeffcent, respectvely Two-factor CAPM (2F-CAPM) The precedng sectons have dscussed CAPM under two extreme assumptons,.e., ether the world captal market s fully segmented or s fully ntegrated. Tests of the classc CAPM under the hypothess of full market ntegraton have rejected a sngle source of rsk as beng adequate n descrbng cross-secton varatons of returns across dfferent countres (see Harvey, 1991). Ths rejecton sgnal could mean that the world captal market s not ntegrated. Drven by the belef that the world captal market s probably nether fully segmented nor fully ntegrated, as well as the fndngs of Bekaert and Harvey (1995) and Bekaert et al. (2005) that some emergng markets are partally ntegrated nto world captal market, ths study proposes a two-factor model whch ntroduces a global market factor nto the classc CAPM, hereafter denoted as 2F-CAPM. 1 In the case of 2F-CAPM, the model ncludes both knds of premums, one for the securty s exposure to the return on the local market portfolo and another for the exposure to the return on the world market portfolo. Therefore, the model captures the senstvty of a frm s returns not only to the local market movements, but also to the global factor. The cost of equty under 2F- CAPM s gven by: Cost of Equty = Rsk-Free Rates + Premum for Local Systematc Rsk + Premum for Global Systematc Rsk G G CE = R + β R R + β R R (4) F L ( ) ( ) M F G M Note that β L and β G n equaton (4) s denoted dfferently from the β coeffcents n equaton (2) and (3). Ths s to hghlght the fact that they are coeffcents measurng partal senstvty of frm returns to the local and world market movements, respectvely. The beta estmaton for the 2F-CAPM s gven as below: G G r = α + β r r + β r r + ε (4a) t L ( ) ( ) t Mt Ft G Mt Ft In conjuncton wth the fndngs of Estrada (2002) and Chen and Chen (2004) that downsde beta has a stronger explanatory power on stock returns than the standard beta, ths study proposes a downsde verson of the LCAPM, GCAPM and 2F-CAPM. The three models are dscussed n more detals n the followng secton Downsde CAPM (DCAPM) The calculaton of downsde beta nvolves solatng nstances when both the frm and the local market ndex returns are less than the rsk-free rate. From here, two new downsde seres were F 5

6 generated and the beta was calculated for these seres, usng smple lnear regresson. Ths beta s D called downsde beta, denoted β for frm : Cost of Equty = Rsk-Free Rates + Premum for Downsde Systematc Rsk D CE = RF + β ( RM RF ) (5) D E[ mn{ ( rt ),0} mn{ ( rmt ),0} ] where β = E mn r r,0 2 (5a) {[ ( Mt Ft ) ] } s estmated from the regresson of the two newly generated downsde seres of mn{ ( rt ),0} and mn{ ( r r ),0} (see Estrada, 2002). Mt Ft Downsde GCAPM (DGCAPM) Usng Estrada s approach, the downsde rsk model can be extended to GCAPM. The ratonale s that even f the market s globally ntegrated, nvestors mght stll have a preference for asymmetrc rsk. We thus nclude the downsde verson of the GCAPM where we term as DGCAPM, as shown below: Cost of Equty = Rsk-Free Rates + Premum for Global Downsde Systematc Rsk G DG G G CE = R + β R R (6) where DGCAPM: F ( M F ) G G G E[ mn{ ( rt ),0 } mn{ ( rmt ),0 }] DG = G G E [ mn( r r ),0 ] 2 { Mt Ft } β (6a) Two-factor Downsde CAPM (2F-DCAPM) Smlarly, the downsde betas for the two-factor CAPM were estmated from the followngs: where Cost of Equty = Rsk-Free Rates + Premum for Local Downsde Systematc Rsk + Premum for Global Downsde Systematc Rsk D D G G CE = R + β R R + β R R (7) D L F L ( M F ) G ( M F ) [ mn{ ( rt ),0 } mn{ ( rmt ),0 }] E{ [ mn( r ),0 ] } 2 Mt G G G [ mn{ ( rt ),0 } mn{ ( rmt ),0 }] G G E [ mn( r r ),0 ] 2 E β = (7a) E β = (7b) D G { Mt Ft } D D where β L s the downsde local beta and β G the downsde global beta (wth respect to the U.S. market) The Non-CAPM Cost of Equty: Estrada Model Exstng emprcal evdence has questoned the valdty of the classcal CAPM for emergng markets. For example, Harvey (1995) and Estrada (2000) showed that standard betas are not correlated wth returns computed for the world market. In addton, the beta values seem to be too small to reflect cost of equty that most nvestors deem as reasonable. These problems have 6

7 led some scholars to look for measures of rsk beyond the realm of CAPM. One of such alternatves s offered n Estrada (2000, 2001). In the classcal one-factor CAPM, beta coeffcent s used as the only rsk measure n the calculaton of cost of equty. However, Estrada (2000, 2001) argued that beta s not approprate to estmate the cost of equty for emergng market and suggests several rsk varables, namely, total rsk as measured by the standard devaton of returns and downsde rsks as measured by the sem-devaton of returns and downsde beta Standard Devaton of Returns (Total Rsk) From a local nvestor perspectve, the general framework of Estrada s model can be gven as: Cost of Equty = Rsk-Free Rates + Premum for Total Rsk CE = R f + σ ( Rm R f ) (8) The total rsk for the stock returns of any partcular frm s bascally gven by the smple standard T 1 devaton of the return seres, = ( ) σ r t r (8a) T t= Sem-Devaton of Returns (Downsde Rsk) Downsde rsk s not a new concept. It was frst suggested by Roy (1952) who beleves nvestors wll prefer safety of prncpal frst and wll set some mnmum acceptable return that wll preserve the prncpal. Roy s concept became nfluental n the development of downsde rsk measures. The cost of equty measure for ths model can be wrtten as: Cost of Equty = Rsk-Free Rates + Premum for Downsde Rsk CE = R f + δ R ( Rm R f ) ft, (9) The sem-devaton measures the average devaton of returns below the rsk-free rate: T 1 2 R = ( {( rt rft ) }) ft, mn, 0 T t = 1 δ R ft, 2 δ (9a) The measure obtaned was then appled to equaton (8) n replacement of σ to calculate the frm-level cost of equty. 2.2 Determnants of Cost of Equty In the search for potental factors that determne cost of equty, we consder some of the frm s fnancal ratos and a few other measures that nvestors mght consder when nvestng. The determnants and the hypotheszed relatonshp wth cost of equty are as below: a) Debt-to-Equty Rato, DE (postve): The fnance lterature stressed a drect relatonshp between debt-to-equty rato and cost of equty of a frm. The argument s that the tax beneft of debt dmnshes beyond a certan pont, and the addtonal fnancal rsk outweghs the lower nomnal cost of debt, thereby ncreasng the cost of equty, reflectng the ncrease n the fnancal rsk of a frm. DE s defned as total debt dvded by total market value of equty rato at the end of year. 7

8 b) Return-to-Equty, ROE (postve/negatve): In general, the hgher the ROE, the better off are the frm s common stockholders. On the other hand, t can be argued that snce only stockholder s equty appears n the denomnator, the ROE s nfluenced drectly by the amount of debt. Gven two frms wth the same earnngs, the frm that uses more debt fnancng wll appear to have hgher ROE. Ths s because wth relatvely hgher debt fnancng, the frm s able to spread ts earnngs over a smaller base of stockholders equty. Therefore, t s debatable whether the sgn should be postve or negatve. ROE s defned as earnngs avalable for common stockholders dvded by common stock equty. c) Sales-to-Assets, SA (negatve): Ang et al. (2000) argued that asset turnover rato measures management s effcency n utlzng assets. Frms wth hgher asset turnover rato have lower cost of equty n the framework of Ang et al. (2000) because t s a reflecton of lower manageral agency problem. Ther fndngs are supported by Sngh and Nejadmalayer (2004) who suggested that manageral effcency n utlzaton of frm resources has a postve effect on frm s cost of equty. SA s defned as total sales dvded by total assets. d) Payout rato, POUT (postve): The payout rato measures the proporton of earnngs that s pad out as dvdends. Consstent wth Myers s (1984) peckng order theory, frms prefer nternal fnancng (retaned earnngs) to external fnancng (debt and external equty fnancng) because of floataton costs of new securty ssues. Relatvely, nternal fnancng s less costly than external fnancng. Hgh payout rato means the porton of retaned earnngs avalable for renvestment purposes wll be less. As a result, frms may have to use external fnancng when nvestment opportuntes occur. Antcpatng larger rsks n terms of hgher debt ratos n the future, nvestors wll demand hgher returns. Therefore, hgher cost of equty. POUT s defned as dvdends per share dvded by earnngs per share for the last fnancal perod. e) Frm Sze, SIZE (negatve): Bloomfeld and Mchaely (2004) reported that analysts expect large frms to have slghtly less rsk and therefore hgher returns compared to small frms. In such a case, there should be a negatve relatonshp between sze and the cost of captal. Recent studes (Hal and Leuz, 2006; Chen et al., 2004) have found a sgnfcant negatve relatonshp between frm sze and the cost of equty. SIZE s defned as natural logarthm of market value of a frm s outstandng common stock at the end of each year. f) Book-to-Market Rato, BM (posttve/negatve): Fama and French (1993) showed that the BM rato s an mportant valuaton measure for explanng average stock returns. The rato may act as a proxy for dstress rsk factor snce fnancally dstressed frms are lkely to have hgh BM. Gode and Mohanram (2003) also pont out that hgher BM reflects hgher perceved rsk. On the contrary, Erb et al. (1996) found the BM rato to be negatvely related to a country s economc rsk ratng, whch ndcates that as the rsk ratng mproves, the rato decreases. As stated by Ameer (2007), one of the mportant mplcatons of Erb et al. s fndng s that f cash flows to emergng markets are related to better economc 8

9 fundamentals and the stock valuaton reflects these, then there should be a negatve relatonshp between the BM rato and cost of equty. BM s defned as the market value of the ordnary (common) equty dvded by the balance sheet value of the ordnary (common) equty. g) Stock lqudty, SL (negatve): Stock lqudty s an mportant attrbute snce hghly lqud stocks can be bought and sold wth mnmal mpact on stock prces. On the contrary, an llqud stock wll ncrease cost of tradng because of the dffculty to trade the stocks. The nfluence of tradng costs on nvestors requred returns was examned by Amhud and Mendelson (1986), Brennan and Subrahmanyam (1996) and Jacoby et al. (2000), that showed a drect lnk between lqudty and frm s cost of captal. Followng Brennan et al. (1998) and Chorda et al. (2001), the natural logarthm of annual tradng volume s used as the proxy for SL. h) Dvdend Yeld, DY (postve): The noton of usng dvdend yeld to forecast returns s not new. Evdence to support the hypothess can be found n the study of Rozeff (1984), Campbell and Shller (1988), Fama and French (1988) and Campbell (1991), among many others. Ther fndngs are n accord wth the ntuton that stock prces are low relatve to dvdends when dscount rates (cost of equty) and expected returns are hgh. DY n ths study s computed as dvdend per share dvded by stock prce. ) Earnngs per Share, EPS (postve): EPS has smlar ntuton as DY accordng to Fama and French (1988). Therefore, a postve relatonshp s expected. EPS s defned as earnngs avalable for common stockholders dvded by number of shares outstandng. In sum, a total of nne potental ndependent varables have been dentfed. As some varables n the lst measure smlar attrbutes, we expect hghly correlated ndependent varables. Those wth correlaton coeffcents exceedng 50 percent wll be taken out from the model. Remanng factors are tested n the followng emprcal settng: CEt = α t + β1 VAR1 + β2var2 + β3var βnvarn + εt (9) where CE t s frm s cost of equty, α t represents ntercept, β 1, β 2, β 3, β n are the regresson coeffcents, VAR 1, VAR 2, VAR 3, VAR n are the possble cost of equty determnants and ε t the error term. The above settng s estmated wth a panel regresson whch allows us to control for the frm effect. 2.3 Data Descrpton Weekly data were used n the estmaton of all rsk measures. The sample perod for ths study covers 5 January 2000 untl 31 December The rsk measures were estmated for every year of the sample perod based on the weekly observatons of the relevant year. All data were collected from DataStream, whch nclude the weekly prces of stocks lsted on the Man Board of Bursa Malaysa as well as the market ndces of the U.S. Weekly frequency s preferable because daly seres has more nose that may affect the qualty of the cost of equty estmates. 2 The annual averages of the monthly 3-month Treasury bll rates of Malaysa and U.S. were used for the local and global rsk-free rate, respectvely. The varables used for explorng the determnants of frm s cost of equty were also obtaned from the DataStream database on yearly 9

10 bass. The calculaton of costs of equty nvolves the local and global market rsk premums. Followng Damodaran (2002), the soveregn bond premum approach was used to solve the problem assocated wth the estmaton of market rsk premum for emergng markets. Accordngly, the Malaysan equty rsk premum was computed as the sum of the premum of a developed market (.e., the U.S. for ths study). Gven that only annual rsk premums are avalable, the costs of equty were calculated on annual bass. We nclude frms from seven sectors of the Man Board n Bursa Malaysa. After flterng out new frms whch were lsted after 2000 because they do not have a complete seres of data for the full sample perod, we have a total of 413 frms avalable for analyss. They are from Constructon (32 frms), Consumer Products (57 frms), Industral (138 frms), Plantatons (23 frms), Propertes (58 frms), Trade & Servces (92 frms) and Technology (13 frms). Fnance sector s excluded from the study due to dfferent nterpretaton of certan fnancal ratos. Mnng s also excluded because only two frms passed the flterng process. 3. Results and Dscusson 3.1 Selecton of the Best Model Table 1 shows the annual returns of Malaysan frms by sector, both local and global rsk-free rates and market rsk premums (extracted from Damodaran s webste) for local as well as global market. Overall, there are large fluctuatons n the frm annual returns. Negatve returns were recorded n 2000 but n 2001, huge mprovement can be seen for all frms, wth the Consumer Products, Technology and Plantatons sectors recorded postve returns. The annual returns deterorated n the followng year but mproved n Nevertheless, all the sectors show postve annual returns n 2007, a major mprovement from year In 2008, all the sectors experenced a sharp declne n ther average annual returns. Local and global rsk-free rates have been declnng snce 2000 and showed sgn of ncrease only n However, both rates declned agan n 2007 and experenced a further drop n On the contrary, both local and global market rsk premum has remaned relatvely stable around 6.5% and 5%, respectvely. Insert Table 1 about here Estmated rsk measures from equatons (2a), (3a), (4a), (5a), (6a), (7a, b), (8a) and (9a) are presented n Table 2. In lne wth Estrada s (2000, 2001) fndngs, our sem-devaton estmates are lower than those of standard devaton, whle estmated downsde betas are greater than standard betas for both the one-factor and two-factor models. Estmated betas for CAPM s roughly three tmes hgher than GCAPM, suggestng frm returns are more responsve to the varatons n the local market than to the world market movements. The estmated betas for fve out of seven sectors have average fgures of greater than one. Ths means the fve sectors have hgher rsk exposure than the market. The other two sectors, Consumer Products and Plantatons have lower average betas of 0.78 and 0.97, respectvely. On the contrary, the estmated betas for GCAPM are less than 0.5, suggestng the stock returns are less responsve to global market returns. Estmated downsde betas have been consstently above one for both the CAPM and GCAPM models. When both the local and global factors appear together n the two-factor model, 10

11 the local betas end up wth average values greater than the global betas. Ths s also true for ts downsde verson. Ths fndng s consstent wth the observaton for the one-factor model. Insert Table 2 about here A panel regresson analyss s performed where actual returns for all frms were regressed aganst each of the dfferent rsk measures and the explanatory power of the estmated models s compared. Rsk measures that have good explanatory power are also better measures for the calculaton of cost of equty. The annual rsk measures as well as the annual actual returns of all the 413 frms were stacked by year and by frm. The panel regresson controls for frm specfc effects as well as perod effects. Table 3 and Table 4 report the R 2 and adjusted R 2 fgures, respectvely, for the dfferent rsk measures accordng to sectors. The rsk measure wth the hghest R 2 and adjusted R 2 s consdered to yeld the best model. The result generally shows that downsde rsk measures are better than ts standard rsk counterparts (except for the sngle factor model). Ths fndng s n lne wth Estrada (2000, 2001, 2002). Not only that, the model whch consders both local and global rsk factors has hgher explanatory power than model that consders only a sngle rsk factor. Based on the average rankngs from the selecton crtera, the sem-devaton approach s ranked one and therefore, yelds the best model. Ths model explans about 40% of varatons n stock returns and for some sectors, the fgure goes up to more than 50%. Therefore, cost of equty estmates were obtaned usng sem-devaton and used n the subsequent step of analyss, that s to explore for determnants of cost of equty. Insert Table 3 about here Insert Table 4 about here 3.2 Determnants of the Asset Prcng Based Cost of Equty The strength of lnear relatonshp between explanatory varables was examned to check for potental multcollnearty. The results from Table 5 ndcate that DE has hgh correlaton wth BM and SL s hghly correlated to SIZE. Therefore, DE and SL were dropped to avod the problem of multcollnearty. The statonarty propertes of all the varables were establshed. The results are not shown here to conserve space but are avalable upon request. Bascally, the result mples that all the panel seres are statonary at level, or they are all I(0) seres. Results on the panel regressons of the potental determnants on frm s cost of equty are gven n Table 6. Results from both R 2 and adjusted R 2 suggest that the varables have reasonably strong explanatory power on the cost of equty of Malaysan frms. Four varables turned out to be sgnfcant. These are SIZE, BM, POUT and ROE. Consstent wth the studes of Hal and Leuz (2006) and Chen et al. (2004), we found a sgnfcant negatve relatonshp between SIZE and the cost of equty. Ths supports the vew that larger frms are able to gan economes of scale n rasng funds and thus should have a lower cost of equty compared to smaller frms. BM s negatvely related to cost of equty. Ths result seems to support the fndng of Erb et al. (1996). The varable PAYOUT s sgnfcantly negatve at 1% level. Ths ndcates that a hgh payout rato, or where more of a frm s earnngs are payout as dvdend, wll lead to lower cost of equty and vce versa. Ths observaton nonetheless, does 11

12 not provde support for Myers s (1984) peckng order theory. The results show that that ROE s negatvely related to the cost of equty. Generally, nvestors favour frms wth hgher ROE, at least for the case of Malaysa. 4. Concluson The am of ths study s to explore for determnants of cost of equty for the Malaysan frms. Unlke prevous studes where the model used for estmatng cost of equty s pre-determned, we consdered a few alternatve models and selected that wth the hghest explanatory power for the analyss. A lst of potental factors was dentfed and the results show that frm s sze, book-tomarket rato, payout rato and return to equty explan the cost of equty sgnfcantly. Our study reveals some nterestng fndngs on the relatonshp between cost of equty and ts determnants, specfcally the book-to-market rato and payout rato. Fama and French (1996) suggested that frms wth hgh book-to-market are a sgn of fnancal dstress and therefore wll command a hgher premum for takng these addtonal rsks. However, our results show the book-to-market rato s negatvely related to cost of equty, suggestng fnancal dstressed frms could have low book-to-market rato, thereby demandng a hgher requred return. Apart from the fndngs of Erb et al. (1996) whch provded support for our results, Dchev (1998) and Zaretzky and Zumwalt (2007) also reported that most dstressed frms have low book-to-market ratos. Our observaton on the payout rato does not conform to the conventonal wsdom, too. We fnd that payout rato s negatvely related to frm s cost of equty, suggestng hgher dvdend payouts leads to lower cost of equty. Perhaps the concept of tme value of money can be used to explan ths phenomenon and ths suggests the need for further studes. Acknowledgement The frst author s a PhD canddate at the Unversty of Malaya. She s grateful for the sponsorshp under the Unversty of Scence Malaysa Fellowshp Scheme. Ths research s supported by the Unversty of Malaya PPP Research Grant (PS005/2009A). The usual dsclamer regardng errors and omssons apples. Footnotes 1. A two-factor settng s common n the lterature of asset prcng for partally ntegrated markets. However, there are a few dfferent approaches to deal wth partally ntegrated prcng, see for example, Errunza and Losq (1985), Errunza et al. (1992), Kearney (2000) and Gérard et al. (2003). 2. For the weekly seres, Wednesday closng prces are collected to avod the Monday and Frday effects. References Ameer, R., 2007, Tme-varyng cost of equty captal n Southeast Asan Countres. Asan Economc Journal, 21(2), pp Amhud, Y. and H. Mendelson, 1986, Asset prcng and the bd-ask spread. Journal of Fnancal Economcs, 17, pp

13 Ang, J.S., R. Cole, and J. Ln, 2000, Agency costs and ownershp structure. Journal of Fnance, 55, pp Banz, R.W., 1981, The relatonshp between return and market value of common stocks. Journal of Fnancal Economcs, 9(1), pp Basu, S., 1977, Investment performance of common stocks n relaton to ther prce-earnngs ratos: a test of the effcent market hypothess. Journal of Fnance, 32(3), pp Bekaert, G. and C.R. Harvey, 1995, Tme-varyng world market ntegraton. Journal of Fnance, 50(2), pp Bekaert, G., C.R. Harvey, and A. Ng, 2005, Market ntegraton and contagon. Journal of Busness, 78(1), pp Bloomfeld, R. and R. Mchaely, 2004, Rsk or msprcng? From the mouths of professonals. Fnancal Management, 33, pp Brennan, M. J. and A. Subrahmanyam, 1996, Market mcrostructure and asset prcng: on the compensaton for llqudty n stock returns. Journal of Fnancal Economcs, 41, pp Brennan, M.J., T. Chorda, and A. Subrahmanyam, 1998, Alternatve factor specfcatons, securty characterstcs, and the cross-secton of expected stock returns. Journal of Fnancal Economcs, 49, pp Bruner, R.F., K.M. Eades, R.S. Harrs and R.C. Hggns, 1998, Best practces n estmatng the cost of captal: survey and synthess. Fnancal Practce and Educaton, pp Campbell, J.Y., 1991, A varance decomposton for stock returns. Economc Journal, 101(405), pp Campbell, J.Y. and R.J. Shller, 1988, Stock prces, earnngs and expected dvdends. Journal of Fnance, 43(3), pp Chen, J.G. and Chen, D.H., 2004, The downsde rsk and equty valuaton: emergng market evdence. Journal of Emergng Market Fnance, 3(1), pp Chen, K.C.W., Z. Chen and K.C.J. We, 2004, Dsclosure, corporate governance, and the cost of equty captal n emergng markets. Workng paper. Hong Kong Unversty of Scence and Technology, Hong Kong. Chen, K.C.W., Z. Chen, and K.C.J. We, 2009, Legal protecton of nvestors, corporate governance, and the cost of equty captal. Journal of Corporate Fnance, 15, pp

14 Chorda, T., A. Subrahmanyam, and R.V. Anshuman, 2001, Tradng actvty and expected stock returns. Journal of Fnancal Economcs, 59, pp Damodaran, A, 2002, Investment valuaton, Wley, New York. Dchev, I.D., 1998, Is the rsk of bankruptcy a systematc rsk? Journal of Fnance, 53(3), pp Erb, C. B., R. H. Campbell and T. E. Vskanta, 1996, Poltcal rsk, economc rsk and fnancal rsk. Fnancal Analyst Journal, 52(6), pp Errunza, V., and E. Losq, 1985, Internatonal asset prcng under mld segmentaton: theory and tests. Journal of Fnance, 40, pp Errunza, V., E. Losq, and P. Padmanabhan, 1992, Tests of ntegraton, mld segmentaton and segmentaton hypotheses. Journal of Bankng and Fnance, 16, pp Estrada, J., 2000, The cost of equty n emergng markets: a downsde rsk approach. Emergng Markets Quarterly, pp Estrada, J., 2001, The cost of equty n emergng markets: a downsde rsk approach (II). Emergng Markets Quarterly, pp Estrada, J., 2002, Systematc rsk n emergng markets: the D-CAPM. Emergng Markets Revew, 3, pp Fama, E.F. and K.R. French, 1988, Dvdend yelds and expected stock returns. Journal of Fnancal Economcs, 22, pp Fama, E.F. and K.R. French, 1993, Common rsk factors n the returns on stocks and bonds. Journal of Fnancal Economcs, 33, pp Fama, E.F. and K.R. French, 1996, Multfactor explanatons of asset prcng anomales. Journal of Fnance, 51(1), pp Gérard, B., K. Thanyalakpark and J.A. Batten, 2003, Are the East Asan Market ntegrated? Evdence from the ICAPM. Journal of Economcs and Busness, 55, pp Gode, D., and P. Mohanram, 2003, Inferrng the cost of captal usng the Ohlson-Juettner model. Revew of Accountng Studes, 8, pp Godfrey, S. and R. Espnosa, 1996, A practcal approach to calculatng costs of equty for nvestments n emergng markets. Journal of Appled Corporate Fnance, pp Guedham, O. and D. Mshra, 2009, Excess control, corporate governance and mpled cost of equty: nternatonal evdence. The Fnancal Revew, 44, pp

15 Hal, L. and C. Leuz, 2006, Internatonal dfferences n cost of captal: do legal nsttutons and securtes regulaton matter? Journal of Accountng Research, 44, pp Harrs, R.S., F.C. Marston, D.R. Mshra and T.J. O Bren, 2003, Ex ante cost of equty estmates of S&P frms: the choce between global and domestc CAPM. Fnancal Management, pp Harvey, C.R., 1991, The world prce of covarance rsk. Journal of Fnance, 46(1), pp Jacoby, G., D.J. Fowler and A.A. Gottesman, 2000, The captal asset prcng model and the lqudty effect: a theoretcal approach. Journal of Fnancal Markets, 3, pp Kearney, C., 2000, The determnaton and nternatonal transmsson of stock market volatlty. Global Fnance Journal, 11, pp Koedjk, K.G. and M.A. van Djk, 2004, Global rsk factors and the cost of captal. Fnancal Analysts Journal, 60(2), pp Koedjk, K.G., C.J.M. Kool, P.C. Schotman, and M.A. van Djk, 2002, The cost of captal n nternatonal fnancal markets: local or global? Journal of Internatonal Money and Fnance, 21, pp Lessard, D., 1996, Incorporatng country rsk n the valuaton of offshore projects. Journal of Appled Corporate Fnance, 9(3), pp Ln, J.C., A.K. Sngh and W. Yu, 2009, Stock splts, tradng contnuty, and the cost of equty captal. Journal of Fnancal Economcs, 93, pp Ltzenberger, R. and K. Ramaswamy, 1982, The effects of dvdends on common stock prces: tax effects or nformaton effects? Journal of Fnance, 37(2), pp McLaney, E., J. Ponton, M. Thomas and J. Tucker, 2004, Practtoners perspectves on the UK cost of captal. The European Journal of Fnance, 10, pp Mshra, D.R. and T.J. O Bren, 2001, A comparson of cost of equty estmates of local and global CAPMs. The Fnancal Revew, 36, pp Myers, S.C., 1984, The captal structure puzzle. Journal of Fnance, 39, pp Perero, L.E., 2001, The valuaton of closely-held companes n Latn Amerca. Emergng Markets Revew, 2, pp Perero, L.E., 2006, The practce of nvestment valuaton n emergng markets: evdence from Argentna. Journal of Multnatonal Fnancal Management, 16, pp

16 Rakow, K.C., 2010, The effect of management earnngs forecast characterstcs on cost of equty captal. Advances n Accountng, ncorporatng Advances n Internatonal Accountng, 26, pp Renganum, M.R., 1981, Msspecfcaton of captal asset prcng: emprcal anomales based on earnngs yelds and market values. Journal of Fnancal Economcs, 9(1), pp Rozeff, M., 1984, Dvdend yelds are equty rsk premums. Journal of Portfolo Management, pp Sngh, M. and A. Nejadmalayer, 2004, Internatonalzaton, captal structure and cost of captal: evdence from French corporatons. Journal of Multnatonal Fnancal Management, 14, pp Stulz, R., 1981, On the effects of barrers to nternatonal nvestment. Journal of Fnance, 36(4), pp Zaretzky, K. and J.K. Zumwalt, 2007, Relaton between dstress rsk, book-to-market rato and return premum. Manageral Fnance, 33(10), pp

17 Table-1 Annual Average of Frm Returns, Rsk-Free Rates and the Market Rsk Premums (n percent) Year Frm Returns Constructon Consumer Products Industry Plantatons Property Technology Tradng/Servces Rsk-free Rate Local Global Market Rsk Premum Local Global Grand Mean 17

18 Table-2 Average Rsk Measures by Sector Constructon Consumer Products Industry Plantatons Propertes Technology Tradng Grand Mean β D β G β DG β β L β G D β L D β G σ δ R ft, Table-3 R 2 from Regresson of Frm Returns on Varous Rsk Measures Model Constructon Consumer Products Industry Plantatons Propertes Technology Tradng Grand Mean CAPM GCAPM DCAPM DGCAPM FCAPM FDCAPM SMSTD STD

19 Table-4 Adjusted R 2 from Regresson of Frm Returns on Varous Rsk Measures Constructon Consumer Products Industry Plantatons Propertes Technology Tradng Grand Mean CAPM GCAPM DCAPM DGCAPM F-CAPM F-DCAPM SMSTD STD Notes: SMSTD refers to sem-devaton and STD refers to standard devaton of returns. 19

20 Table-5 Correlaton Matrx of the Potental Determnants of Cost of Equty COE DY EPS SIZE ROE BM CR SA DE SL POUT COE DY EPS SIZE ROE BM CR SA DE# * SL# * POUT (*) hgh correlaton, (#) to be dropped to avod multcollnearty. 20

21 Table-6 Estmates of the Panel Model for Cost of Equty Varables Coeffcent P-value C (0.0000)** DY EPS SIZE (0.0002)** ROE (0.0989)* BM (0.0715)* CR SA POUT (0.0005)** R-squared Adjusted R-squared Durbn-Watson statstc F-statstc (0.0000)** Note: * and ** denote sgnfcance at the 0.10 and 0.01, respectvely. 21