AN EMPIRICAL TESTING OF CAPITAL ASSET PRICING MODEL IN BANGLADESH

Transcription

1 Journal o Research (Scence), Bahauddn Zakarya Unversty, Multan, Pakstan. Vol.17, No.4, October 2006, pp ISSN AN EMPIRICAL TESTING OF CAPITAL ASSET PRICING MODEL IN BANGLADESH Md. Mostazur Rahman School o Busness, South East Unversty, Dhaka, Bangladesh Md. Azzul Baten Department o Statstcs, Shah Jalal Unversty o Scence and Technology, Sylhet-3114, Bangladesh Abstract Captal Asset Prcng Model (CAPM) provdes an equlbrum lnear relatonshp between expected return and rsk o an asset. The purpose o ths paper s to nvestgate a rsk-return relatonshp wthn the CAPM ramework. The study also ams at explorng whether CAPM s a good ndcator o asset prcng n Bangladesh. For ths study, a perod have been consdered. Fama-French [1992] methodology on ve varables-stock market return, beta, book to market value, sze (Market captalzaton) and sze 1 (sales) were used to test ths model. In the present ndngs on the CAPM t has been shown that the varables studed have sgncant relatonshp wth stock return, are stll too alve on ths ground. Keywords: Captal asset prcng model, emprcal test, equlbrum lnear relatonshp, return, sze. INTRODUCTION Under certanty the savng can be nvested n one knd asset or certan but n case o uncertanty, whch appears to be unavodable n the real world, anyone must have to undertake the lablty o rsk due to uncertan uture earnngs on assets or securtes. The potental nvestors, ndvdual or corporaton are aced wth a captal market o consderable sophstcaton oerng a wde range o nvestment opportuntes. The Asset Prcng models (CAPM) descrbe an equlbrum relatonshp between expected return and rsk n the securtes market under the assumpton that uncertan uture returns o securtes can be descrbed n terms o moments o ther probablty dstrbutons. The most common verson s the mean-varance CAPM n whch two moments are assumed to be relevant. The relatonshp between expected return and rsk s the central theme o the asset prcng theory. It s wth the problems surroundng choce under uncertanty that Markowtz [1952] and Tobn [1958] rst concerned themselves. The orgn o the asset prcng theory les wth Markowtz [1952] who was a poneer n demonstratng ormally that dverscaton o securty holdngs reduces the rsk, unless the 225 J. res. Sc., 2006, 17(4),

2 226 Md. Mostazur Rahman and Md. Azzul Baten returns to the securtes are perectly correlated. He theorzed that nvestors could dversy away all sorts o rsks except the rsk that comes wth holdng stocks n general. The CAPM model usually attrbuted to 1990 Nobel Laureate Wllam Sharpe [1964], was also developed by Lnter [1965], Mossn [1966] and Fsher [1972]. In general CAPM provdes an equlbrum lnear relatonshp between expected return and rsk assocated wth an asset. The average return anomales o the CAPM suggest that, asset prcng s ratonal, a multactor verson o Merton s [1973] nternatonal CAPM or Ross [1976] arbtrage prcng theory (APT) can provde a better descrpton o average returns. Fama and Macbeth [1973] was the rst reported study n Amerca o the lnear relaton as predcted by the CAPM. They reported statstcs or the slope coecents as beng between 0.7 and 1.73 or the years and respectvely. Ball et al. [1976] reported evdence o a robust postve lnear relatonshp between rsk and returns n Australa. Ar and Johnson [1990] ound a strong lnear relatonshp n the Sngapore share market, thus suggestons that portolo rsk and returns are postve and lnear. Ther reported coecent o varaton was hgh as 70% especally over long perod tests o 18 years, but both the slope coecent and the coecent o varaton n tests done n short perods were nsgncant. Fama and French [1989] dented two useul varables or orecastng expected asset returns: the deault and term spread, whch depend upon the monetary envronment ounded by Jensen [1996]. He [1993] shows that the CAPM model s a good descrpton o returns on portolo ormed on sze and book to market equty. Fama and French [1992] reported that the relatonshp between beta and average returns dsappeared durng the perod. They ound that there s a smple postve relatonshp between average return and beta durng the pre-1969 perod; no sgncant relatonshp was ound or perod. Fama and French [1994] use the model also to explan ndustry returns. The captal asset prcng model n ts varous orms has been extensvely tested or the developed captal markets such as those o U.S.A., Europe and Australa and to a lesser extent or the developng captal markets. It s needless to say that there have been very lttle tests o ths model n the Bangladesh envronment despte the exstence o an organzed captal market or a long perod. Moreover, the applcablty o the western theores to Bangladesh captal markets s suspect owng to several derences between the developed captal markets and the developng ones. There are economc and nsttutonal derences, sze related varatons, lqudty condtons, dsclosure requrements, ntegraton o the nancal system etc. Thus, the motvaton or the study s to generate comparatve test results wth n the CAPM ramework or a developng captal market such as Bangladesh. Ths study ams at testng the applcablty o the model to descrbe rsk-return relatonshp on the Bangladesh captal market. Ths paper explores whether the CAPM s a sutable descrpton o asset prcng n Bangladesh context. The paper has been organzed as ollows: Next secton descrbes the data and the varables used n the study or analyzng the applcablty o Fama-French s CAPM n DSE. Then a justcaton o choosng the sutable analyss technque to denty the determnants o CAPM n an emergng market s dscussed. Proceedng secton dscusses the speccaton o CAPM wth assumptons. Then emprcal ndngs wth dscusson have been presented. Fnally, n the last secton are gven concludng remarks.

3 AN EMPIRICAL TESTING OF CAPITAL ASSET PRICING MODEL IN BANGLADESH 227 DATA DESCRIPTION Ths survey s conducted to nvestgate the Fama-French s CAPM n developng countres especally n Bangladesh based on the sources o normaton rom Dhaka Stock Exchange (DSE) emphaszng only on non nancal sector. The data range s rom 1999 to Non-nancal sectors have been ncorporated or analyzng the applcablty o Fama-French s CAPM n DSE and to judge the multactor varable eect on DSE. SAMPLE SIZE The nal sample conssts o 123 Dhaka Stock Exchange lsted non-nancal companes. SAMPLE PERIOD Fve years perod ( ) s consdered or ths study. There were 93 companes lsted n the DSE n 1988 but that ncreased to 105 n 1989, to 116 n 1990, and to 209 n 1997 and to 248 n So, t s observable that the lsted companes o the DSE are ncreasng every year because o new lsted companes. Ths study consdered all the DSE lsted non-nancal companes or the 5 years perod as the sample,.e., t conducts panel study, however, as the sample sze s not same or every year but rather the sample sze ncreases every year. Ths study conducts ve yearly average cross-secton models and polled models (tme-seres and cross-secton together). So, there s no problem to conduct pooled regresson analyss would be ((Sample sze * Number o years) Mssng cases) because o the unbalanced panel data. EMPIRICAL PART: TESTING FAMA-FRENCH S CAPITAL ASSET PRICING MODEL ON THE DSE DATA The Dhaka Stock Exchange lsted all non-nancal sector companes over perod o are prmarly consdered the sample o ths emprcal phase. However, as we have already been mentoned earler, a ew number o companes are excluded rom the sample because ether all o the company or market data o those companes are unavalable. So, the sample sze became smaller than the actual companes lsted n the DSE. Thereore, the nal sample conssts o 123 DSE lsted non-nancal sector companes or ths research. All the company data are collected rom the annual reports o the lsted non-nancal sector companes o the perod o The market data ( ) are collected rom the DSE prce database. However, the macro-economc data are collected the publshed reports o Natonal Board o Revenue o Bangladesh. Bre descrpton o varous varables Name o the Varables Proxes Calculatons Stock Market Return Natural Log o Stock Return LN ( I I I ) Beta beta actor by Scholes and Wllam [1977] and Dmsons [1979] / t 1 Usng Regresson Model Book to Market Book value/market value Book value/market value Sze Market captalzaton and Sales LN (Market captalzaton and Sales)

4 228 Md. Mostazur Rahman and Md. Azzul Baten JUSTIFICATION OF CHOOSING THE ANALYSIS TECHNIQUE IN EMPIRICAL PHASE Ths phase consdered multple regresson analyss approach to denty the determnants o Captal Asset Prcng Model n an emergng market. Ths method best suts ths study because we took the CAPM theores, whch comprses o beta o the stock and we take ths on the day. The day beore and the next day are consderng or computng marketng beta. Not only that we also consder book to market value and sze both n the market captalzaton and sales aspect. As ths study consders the CAPM theores to denty the determnants o stock return, ths s completely new n ths area, whch adds new value to the research and also attempts to mnmze the gap between theoretcal studes and emprcal studes. However, ths study brngs the dvdend theores nto the emprcal nvestgaton, whch wll obvously help to mnmze the gap between theoretcal and emprcal study. As prevous researchers suggest that averagng works very well wth the unbalanced panel data, that motvates to conduct ve yearly average crosssectons and pooled multple regresson analyss or ths study. However, multple regresson analyss s more sutable to deal wth the research problem and data set or the current research. To have a better understandng about the CAPM and ater an ntensve revew o the prevous emprcal studes on the CAPM, we ound nterestng to conduct a study on the Fama-French s CAPM. However, as we know that a number o studes conducted on the CAPM but a very ew are n the developng markets and vrtually no study on the DSE, thereore, ths s ndeed rght attempt to conduct such a study on the DSE data. MODEL SPECIFICATIONS AND TESTABLE HYPOTHESES Fama and French ntroduced three actors CAPM model. Ther model assumes that the expected return o a portolo n excess o the rsk-ree rate E R R explaned by the senstvty o ts return to three actors: ( ) m R, () The excess return on a broad market portolo ( ) m R () The derence between the return on a portolo o small stocks and the return on a portolo o large stocks (SMB, small mnus bg), and () The derence between the return on a portolo o hgh-book-to market stocks and the return on a portolo o low-book to market stocks (HML, hgh mnus low). Speccally, the expected excess return on portolo s R R Where E ( R ) m [( Rm ) R ] + s ( SMB) + h ( HML) ε = α + b + R, E(SMB) and E(HML) are expected premums, and the actor senstvtes or loadngs, b, s, h are slopes n the tme seres regresson. One thng that s nterestng s that Fama-French stll sees hgh returns as a reward or takng on hgh rsk; n partcular that means returns ncrease wth book/prce, then stocks wth a hgh book/prce rato must be more rsky than average-exactly the opposte o what a tradtonal busness analyss would tell that the derence comes rom whether one beleve n the ecent market theory.

5 AN EMPIRICAL TESTING OF CAPITAL ASSET PRICING MODEL IN BANGLADESH 229 The busness analyst doesn t beleve t, so they would say hgh book/prce ndcates a buyng opportunty; the stock looks cheap. But t does beleve n EMT then t s to beleve cheap stocks can only be cheap or a good reason, namely that nvestors thnk they re rsky! So, one s nterested to see whether Fama-French s varables are vald or the Bangladesh market or not. The proposed Fama-French s model consders Scholes and Wllam s [1977] beta actor, book or market value and sze (market captalzaton and sales) or the Bangladesh market. The proposed model s: R R = α + b [( Rm ) R ] + s ( Sze) + h ( BM ) + ε Where t s same as beore the actor senstvtes or loadngs, b, s, h are the slopes n the tme seres regresson. In ths regard the natural log o daly return s taken wth the consderaton o lag and lead. In case o beta, the daly beta o the stock s chosen. In ths regard the orm s lke ths: R = α + β1rmt ( 1) + β2rmt + β3r mt ( + 1) Where α s ntercept and β 1 s the coecents o the day beore (t-1) the stock return and β 2 ( t) s the coecents o the current day the stock return and β 3 s the coecents o the day ater (t+1) the stock return. Through regresson we get the β o the stock. Agan daly market value s taken and n terms o sze the natural log o both the market captalzaton and sales s taken. In regard both the cross sectonal and regresson analyss are taken to the data set In equlbrum, the CAPM speces expected returns as a lnear uncton o rsk n the orm E( R ) = R + [ E( R m ) R ] β (1) Where : ndcates any asset that s expected to produce a cash low, m: ndcates the market or a set o smlar assets traded n an asset market, : the yeld on a deault-ree asset wth dentcal nterval o tme as the asset, and R: represents the returns over ntervals o tme. Equaton (1) s n terms o expected returns. But mplcatons must be tested wth data on perod-by-perod securty or portolo returns. Gven the stochastc generalzaton o (1) the present study attempts to test the ollowng hypothess: (a) A postve relaton between expected return E(R) and the systematc rsk β s hypotheszed,.e., the slope o the CAPM equaton (1) s postve. (b) The slope s equal to the derence between the expected return on the portolo and the rsk-ree rate. (c) The expected return on any β = 0 asset s the rsk-ree return.e. the ntercept o CAPM equaton (1) s. (d) The relatonshp between expected return and rsk ( β ) s lnear. R R

6 230 Md. Mostazur Rahman and Md. Azzul Baten (e) Only β s mportant n derentatng among securty returns,.e., there are no terms other than the rsk-ree-rate and the premum or β that determne expected returns. In other words, the market wll not prce the resdual rsk o any stock. EMPIRICAL FINDINGS Through 5 yearly average cross sectonal analyss and polled multple regresson analyss has been done or ths study. There are two types o Ordnary Least Square (OLS) regresson run to applcablty o three actor Fama- French [1993] model and the justcaton o CAPM n DSE as a representatve o emergng markets n developng countres: one, ve yearly ( ) average cross secton regresson model, and two, polled regresson model. 4 years (5-1) dummes are consdered or 5 years ( ). In average cross sectonal analyss the proxy varables sales and market captalzaton were nterchanged. In polled regresson t s taken the mpulse dummy 1999 year as 1 (one) and other 2003-year as (0) zero. In all aspect the years are sgncant whch ndcates the mpact o tme on the model as well as the sze (both the market captalzaton and sales). As tme have the mpact on the model, so ncorporaton o year dummy has rather mprove the overall sgncance o the regresson model. From the cross secton and regresson analyss one gets the derent oc. In rst case (Table 1) the stock return s taken as dependent varable and the ndependent varables are sze (market captalzaton) market beta and book to market value. The CAPM relates the senstvty o an ndvdual company s stock returns to the returns o the market as a whole. Estmatng a model or a partcular rm requres data on the market rate o return, the rsk-ree rate o return (usually a short-term treasury bll), and stock returns rom the non-nancal nsttutons. The data or ths example consst o daly observatons rom January 1999 through December 2003 on the market return, the rsk-ree rate. Rsk premum s the excess return o a securty over the rsk-ree rate or, rather, the 2 extra return that nvestors requre or bearng rsk. The R value o means that about 2% o the varaton n the stock returns can be explaned by the ndependent varables o the market. The correlaton among the dependent and ndependent varables s , whch shows the nterdependency among the varables. Table 1: Fve Yearly Average Cross Sectonal Analyss: (Sze, Beta, BM) Varables n the equaton (General value). Varable B SE (Standard Beta T Sg T Error Beta) B Beta BM SIZE (Constant) Multple R Adjusted R R Square F-statstc Sgn F

7 AN EMPIRICAL TESTING OF CAPITAL ASSET PRICING MODEL IN BANGLADESH 231 The F score = or cross sectonal analyss and sgncant at 2% level and the beta s sgncant at 1% level other varables BM and Sze (market captalzaton) are not sgncant to explan the dependent varable. So t can be sad that beta have the relatonshp wth the stock return. In Table 2, agan when stock return s taken as dependent varable and the ndependent varables are Sze (sales) market beta and stock to market value. The value o means that about 3% o the varaton n the stock returns can be explaned by the ndependent varables o the market. Table 2: Fve Yearly Average Cross Sectonal Analyss: (Sze 1, Beta, BM) Varables n the equaton (General value). Varable B SE (Standard Beta T Sg T Error Beta) B Beta BM SIZE (Constant) Multple R Adjusted R R Square F-statstc Sgn F The F score = or cross sectonal analyss and the beta and Sze 1 (sales) s sgncant at 1% level and varables BM s not sgncant to explan the dependent varable. So t can be sad that beta and Sze 1 (sales) have the relatonshp wth the stock return. In Table 3, year 1999 was taken as mpulse dummy. The multple R s and R square s Here one observes hghly mpact o year over the 2 ndependent varables. The R value o means that about 47% o the varaton n the stock returns can be explaned by the ndependent varables o the market. Here the mpact o year s noted. Table 3: Pooled Regresson Analyss: (Sze, 1999 mpulse dummy) Varables n the equaton (General value). Varable B SE (Standard Beta T Sg T Error Beta) B Beta BM x SIZE DUM DUM DUM DUM (Constant) Multple R Adjusted R R Square F-statstc Sgn F

8 232 Md. Mostazur Rahman and Md. Azzul Baten F score The overall =3.038 and or cross-sectonal and pooled regresson models respectvely wth the consderaton o Sze (market captalzaton) and year 2000, 2002, 2003 are sgncant at 1% level but the other varables BM and the year 2001 are not sgncant. For Table 4, year 2003 was taken as mpulse dummy. The multple R s and R square s Here one agan observes hghly mpact o year over the 2 ndependent varables. The R value o means that about 47% o the varaton n the stock returns can be explaned by the ndependent varables o the market. Table 4: Pooled Regresson Analyss: (Sze, 1997 mpulse dummy) Varables n the equaton (General value). Varable B SE (Standard Beta T Sg T Error Beta) B Beta BM x SIZE DUM DUM DUM DUM (Constant) Multple R Adjusted R R Square F-statstc Sgn F F score The overall = or pooled regresson analyss and the Sze (market captalzaton) s sgncant at 5% level and the varable beta and BM s not sgncant to explan the dependent varable. So t can be sad that Sze (market captalzaton) have the relatonshp wth the stock return. In pooled regresson models respectvely wth the consderaton o Sze as Market captalzaton and taken year 2003 as mpulse dummy. Here one nds that Sze (market captalzaton) and year 1999, 2000, 2001, 2002 are sgncant at 1% level but the other varables BM and the year 1995 are not sgncant. Table 5: Pooled Regresson Analyss: (Sze 1, 2003 mpulse dummy) Varables n the equaton (General value). Varable B SE(Standard Error Beta T Sg T Beta) B Beta BM 5.61E SIZE DUM DUM DUM DUM (Constant) Multple R Adjusted R R Square F-statstc Sgn F

9 AN EMPIRICAL TESTING OF CAPITAL ASSET PRICING MODEL IN BANGLADESH 233 Year 2003 was taken as mpulse dummy n Table 5. The multple R s and R square s Here one agan notces hghly mpact o year over the 2 ndependent varables. The R value o means that about 50% o the varaton n the stock returns can be explaned by the ndependent varables o the market. Here one agan nds the mpact o year. The overall F score = or pooled regresson analyss and the Sze 1 (Sales) s sgncant at 5% level and the varable beta and BM s not sgncant to explan the dependent varable. So t can be sad that Sze 1 (Sales) have the relatonshp wth the stock return. In pooled regresson models respectvely wth the consderaton o Sze 1 as Sales and taken year 2003 as mpulse dummy. Here t s ound that Sze 1 (Sales) and year 1999, 2000, 2001, 2002 are sgncant at 1% level but the other varables BM and beta are not sgncant. CONCLUDING REMARKS Fama and French s three-actor model o CAPM denes the sgncance o other varables n the market. And to walk n that way, the CAPM model s desgned wth the consderaton o beta, book to market value and Sze (Market captalzaton and Sales). The results o the emprcal nvestgaton strongly support the relatonshp among the varables to determne the stock return also evdenced that beta s not only the actor to determne the stock return but the other varables as taken also sgncantly mportant. In ths research paper the mpact o tme s observed and as t s seen that tme varablty causes the stock return to vary and all varables become sgncant wth the tme actor. So the varables beta, book to market value and Sze but the tme mpact also has sgncant mportance. In present ndngs on the CAPM t has shown that the varables studed have sgncant relatonshp wth stock return, are stll too alve on ths ground. Here t s also ound that the mpact o tme and the year mpact create mportance n Bangladesh market, whch s the newly ssue or CAPM model. Reerences Ar, M. and Johnson, L.W. (1990) Ex ante rsk prema on macroeconomc actors n the prcng o stocks: an analyss usng arbtrage prcng theory, Chapter 18, In: M. Ar and L.W. Johnson (Eds.), Securtes Markets and Stock Prcng, Longman Press, Sngapore. pp Ball, R., Brown, P. and Ocer, R.R. (1976) Asset prcng n the australan ndustral equty market, Aust. J. Manag., 1(1), Dmsons, E. (1979) Rsk measurement when shares are subjected to nrequent tradng, J. Fnan. Eco., 7(2), Fama, E.F. and French, K.R. (1989) Busness condtons and expected returns on stocks and bonds, J. Fnan. Eco., 25, Fama, E.F. and French, K.R. (1992) The cross-secton o expected stock returns, J. Fnan., 47(4),

10 234 Md. Mostazur Rahman and Md. Azzul Baten Fama, E.F. and French, K.R. (1993) Common rsk actors n the returns on stocks and bonds, J. Fnan. Eco., 33, Fama, E.F. and French, K.R. (1994) Industry Costs o Equty, Workng paper, Graduate School o Busness, Unversty o Chcago, Chcago, IL. revsed July Fama, E.F. and French, K.R. (1995) Sze and book to market actors n earnngs and return, J. Fnan., 50, Fama, E.F. and French, K.R. (1996) The CAPM s wanted, dead or alve, J. Fnan., 51, Fama, E.F. and Mac Beth, J. (1973) Rsk, return, and equlbrum: emprcal tests. J. Polt. Eco., 71(4), Fscher, B. (1972) Captal market equlbrum wth restrcted borrowng, J. Busness, 45, Gupta, O.P. and Sanjay, S. (1993) An emprcal testng o captal asset prcng model n Inda, Fnan. Inda, 4, Lnter, J. (1965) The valuaton o rsk assets and the selecton o rsky nvestments n stock portolos and captal budgets, Rev. Eco. and Stat., 47, Markowtz, H. (1952) Portolo selecton, J. Fnan., 7, Merton, R.C. (1973) An Internatonal Captal Asset Prcng Model, Econometrca, 41, Mchael, C.J. (1993) The modern ndustral revoluton, ext and the alure o nternal control systems, J. Fnan., Reprnted n: J. Appl. Corpor. Fnan., (1994) 6(4), and n: Fnanzmarkt und Portolo Management, (1993), 7(3), Reprnted n: Theory o the Frm: Governance, Resdual Clams, and Organzatonal Forms, (2000) Harvard Unversty Press. Mossn, J. (1966) Equlbrum n a captal asset market, Econometca, 35, Ross, S.A. (1976) Arbtrage theory o captal asset prcng, J. Eco. Theory, 13, Scholes, M.S. and Wllam, J. (1977) Estmatng beta rom non-synchronous data, J. Fnan. Eco., 5(5), Sharpe, W.F. (1964) Captal asset prces: a theory o market equlbrum under condtons o rsk, J. Fnan., 19, Tobn, J. (1958) Lqudty preerence as behavor towards rsk, Rev. Eco. Stud., 67,