Stock Market and Economic Activity in Malaysia

Transcription

1 Sock Marke and Economic Aciviy in Malaysia AUTHORS ARTICLE INFO JOURNAL FOUNDER Hawai Janor Noreha Halid Aisyah Abdul Rahman Hawai Janor, Noreha Halid and Aisyah Abdul Rahman (2005). Sock Marke and Economic Aciviy in Malaysia. Invesmen Managemen and Financial Innovaions, 2(4) "Invesmen Managemen and Financial Innovaions" LLC Consuling Publishing Company Business Perspecives NUMBER OF REFERENCES 0 NUMBER OF FIGURES 0 NUMBER OF TABLES 0 The auhor(s) This publicaion is an open access aricle. businessperspecives.org

2 116 Invesmen Managemen and Financial Innovaions, 4/2005 Sock Marke and Economic Aciviy in Malaysia Hawai Janor, Noreha Halid, Aisyah Abdul Rahman Absrac This sudy examines he sock marke as a predicor of he economic aciviy in Malaysia and is sensiiviy o differen sock marke condiions. In doing so, we employ he Johansen coinegraion, variance decomposiion and he Auoregressive Disribued Lags bound es. For he whole period under sudy, he Johansen coinegraion and he VDC show ha he Malaysian sock marke lead changes in economic aciviy. However, he resuls from he ARDL show no relaion beween he wo variables. Differen findings are found for differen sub-periods. All he hree ess sugges ha he sock marke lead changes in economic aciviy only during sub-period of 1986m5-1998m7. This implies ha he sock marke as predicor o economic aciviy is sensiive o differen sock marke condiions. This sudy also highlighs he usefulness of ARDL mehod especially when saionariy becomes an issue. Key words: Sock marke, Economic aciviy, Coinegraion, ARDL JEL classificaion: C22; E44 1. Inroducion Sock price is widely believed o be he predicor of economic aciviy. Theoreically, he link beween sock price and economic aciviy can be explained hrough he sock valuaion model and he wealh effec. The sock valuaion model argues ha sock prices reflec expecaion abou fuure economy, hus i can somehow indicae wha will happen o he economy. The wealh effec conends ha sock prices lead economic aciviy by simulaing or no simulaing he consumpion paern of invesors ha will laer on influence demand and producion of he economy. Empirical researches wih respec o he sock marke performance as a predicor of he economy yield mixed resuls. Earlier sudies such as Peek and Rosengren (1988) and Barro (1989) found ha sock marke someimes does predic he economy in he US. Peek and Rosengren (1988) found ha ou of eleven cases of a declining sock marke, only six were followed by recessions. Barro (1989) found ha sock marke performances successfully prediced eigh ou of nine recessions. A more recen sudy by Muradoglu e al. (2000) found evidence ha sock reurns lead economic aciviy in India and Mexico. However sudies such as Kwon and Shin (1999), Binswanger (2000) and Hondroyiannis and Papaerou (2001) found ha sock reurns do no lead changes in economic aciviy in Korea, US and Greece respecively. Given hese inconclusive evidence, furher research on his subjec is herefore needed. Mos of he sudies cied above examined daa for developed counries or he indusrialized counries. Wih he increase of financial inegraion and he emergence of new sock markes, paricularly in he Asian region, hese new sock markes have become he focus of inernaional invesors. Therefore, sudies ha can predic he economy of hese markes are useful for he inernaional invesors. In addiion, if i is proven ha sock marke reurns can predic he economy, policy makers can have some insighs regarding policy implemenaion in order o achieve a desired resul. Mos of he sudies underaken have employed various echniques such as regressions (Fischer and Meron, 1984; Barro, 1989; Binswanger, 2000), coinegraion ess, (Kwon and Shin, 2000; Hondroyiannis and Papaperou, 2001) and causaliy ess (Mahdavi and Sohrabian, 1991, Muradoglu e al., 2000). A poenial weakness of hese echniques is ha hey require he underlying ime series o be non-saionary. While hese sudies presen evidence from ess, which suppor he presence of a uni roo in he underlying series, i is well known ha such uni roo Hawai Janor, Noreha Halid, Aisyah Abdul Rahman, 2005

3 Invesmen Managemen and Financial Innovaions, 4/ ess have low power (Akins and Coe, 2002). Paricularly, if he series is persisen, ye saionary, hen any inference ha can be made abou he validiy of he relaion beween he series is condiional on he assumpion ha he series are I(1) which means condiioning he resuls of economic hypohesis on shaky saisical ground (Canova, 1994). The objecive of his paper is o examine sock prices as a predicor of he economic aciviy in Malaysia. We also invesigae wheher he resuls hold up in differen sample period. In line wih mos of he sudies, we apply he coinegraion and he VAR echniques. In addiion, we also employ he ARDL bounds es o overcome saionariy issues ha may be encounered in he previous echniques ha required he non-saionariy of he ime series involved. The main advanage of he ARDL bounds es lies in he fac ha i can be applied irrespecive of wheher he regressors are I(0) or I(1), and his avoids he preesing problems associaed wih sandard coinegraion analysis which requires he classificaion of he variables ino I(1) and I(0). 2. Theoreical and empirical framework Theoreically, he sock valuaion model and he wealh effec sugges ha he sock marke predics economic aciviy. The sock valuaion model argues ha he sock marke is forward-looking; hus, curren prices reflec he fuure earnings poenial or profiabiliy of corporaions. Since sock prices picure he expeced profiabiliy, and profiabiliy iself is direcly linked o economic aciviy, flucuaion in sock prices are implied o lead he economic direcion. For insance, if he economy is expeced o ener ino a booming (recession) sage, he sock marke will anicipae his by bidding up (down) he prices of socks. Sock prices will be influenced by expecaions abou fuure economy because a firm s profi has a direc relaionship wih he behavior of he real economy. For insance, if invesors predic an economic growh in fuure, hen expeced profis will improve (assuming dividend and profi has a posiive relaionship) and value of he sock will increase; and vice versa for he opposie scenario. Thus, if predicions by invesors are fruiful, sock price movemens will lead he direcion of he economy. According o Fama (1990), he level of real economic aciviy is expeced o have a posiive effec on fuure cash flows and hus is relaed o sock prices. His sudy showed ha sock reurns were acually significan in explaining fuure real aciviy in US for he whole period from 1953 o The sock marke o predic economic aciviy can also be explained by he wealh effec hrough he resul of wealhy invesors consumpions. Pearce (1983) argues ha flucuaions in sock prices have a direc effec on aggregae spending. When he sock marke is rising, invesors are wealhier and end o spend more. This will increase he demand for goods and hus expand he economy. Mahdavi and Sohrabian (1991) explore he effec of he rae of growh of sock prices on he rae of growh of GNP in he Unied Saes using quarerly daa over a period covering he pas hree decades; ha is, from he firs quarer 1960 o he second quarer Employing he Granger causaliy mehod (Granger, 1969), hey documen evidence ha suppored Fama (1990) and sugges he leading indicaor role of sock prices. However, Binswanger (2000) also using US daa, finds ha he sock marke does no possess any explanaory power on he economy over he sub-samples from 1984 o 1995, bu he resuls hold for he whole period from 1953 o The resuls presen evidence ha sock reurns ceased o lead real economic aciviy during he recen sock marke boom (since he early 1980s) compared o he firs sock marke boom afer World War II from he lae 1940s o he mid-1960s. For he pos-war Unied Saes, Lee (1992) invesigaes causal relaions and dynamic ineracions among asse reurns, real aciviy, and inflaion. The main resul indicaes ha sock reurns help explain movemens in real aciviy and suppor he findings of Fama (1990). All he above sudies focus on he US daa. Kwon and Shin (1999) invesigae wheher economic aciviies in Korea can explain sock marke reurns for he period from January 1980 o December Applying he coinegraion es and a Granger causaliy es from a vecor error correcion model, he sudy found ha he sock price indices are no a leading indicaor of economic aciviy. This sudy is inconsisen wih he previous findings, which conclude ha he sock marke raionally signals changes in real aciviies such as in Fama (1990). Similar resul was

4 118 Invesmen Managemen and Financial Innovaions, 4/2005 also found by Hondroyiannis and Papaperou (2001) for Greece for he period from 1984 o 1999 employing he Johansen coinegraion and variance decomposiion analysis. Murodoglu e al. (2000) examined causal relaionships beween macroeconomic variables (inflaion, ineres raes, real economic aciviy, and exchange raes) and sock reurns in nineeen emerging markes using a mulivariae approach. The resuls suggesed ha sock reurns lead real economic aciviy as proxied by indusrial producion in India and Mexico. He gave wo possible explanaions for such findings. Firs, sock reurns migh simply be leading real economic aciviy, in which case he relaionship mus be undersood as one of a lead-lag relaionship. In his case, sock reurns migh be used as a baromeer. Being able o adjus o informaion regarding governmen policy raher insananeously, changes in sock reurns migh be he indicaors of changes in oher variables. An alernaive explanaion migh be relaed o he size of he sock markes in hese counries. If sock markes were no hin, hey migh serve as a proxy for he financial wealh in he counry. In ha case, unidirecional causaliy from sock reurns o he variables mus be inerpreed as he effec of changes on financial wealh on he variables. 3. Daa, Mehodology and Findings The daa consis of monhly ime series spanning January 1980 hrough December The economic aciviy is measured by he indusrial producion index 1 and he sock marke by Kuala Lumpur Composie Index (KLCI). The indusrial producion index was obained from he IMF Inernaional Financial Saisics and KLCI from he Daasream. All he variables are expressed in logarihmic. We examine he sock marke, measured by KLCI (LCI) as a predicor o he economic aciviy, measured by indusrial producion index (LIP) by applying he Johansen coinegraion, he variance decomposiion and he auoregressive disribued lag (ARDL) bounds es. In order o examine he linkage for differen periods, his sudy divides he monhly daa ino hree periods based on he break in rends of he sock index: 1980m1-1986m3; 1986m5-1998m7; 1998m9-2004m12. We run he daa for he whole period from 1980 o 2004 and compare he resuls over he hree sub-samples. Johansen Coinegraion The Johansen (1991) procedure involves he idenificaion of rank of m by m marix in he specificaion as follows: k 1 X X X, (1) 1 k 1 where X is a column vecor of he m variables, and represen coefficien marices, is a difference operaor, k denoes he lag lengh, is a consan. If has zero rank, no saionary linear combinaion can be idenified. If he rank r of is greaer han zero, however, here will exis r possible saionary linear combinaions and may be decomposed ino wo marices and, (each m x r) such ha =. In his represenaion conains he coefficiens of he r disinc coinegraing vecors ha render X saionary even hough X is iself non-saionary, and conains he speed of adjusmen coefficiens for he equaion. To examine he long run relaionship beween sock reurns and economic aciviy, firs we employ he Johansen coinegraion echnique (Johansen, 1990, 1991). Firs, he saionariy of each series is examined by using he augmened Dickey-Fuller (ADF). The ADF uni roo ess fail o rejec he null hypohesis of he exisence of a uni roo in log levels bu can be rejeced in he log firs difference of he indusrial producion series for he whole period and for he hree subperiods using he Schwarz Bayesian Crierion (SBC) bu no he AIC. Therefore based on he 1 Indusrial producion index is used as proxy o real economic aciviy by sudies such as Muradoglu e al. (2000) and Hondroyiannis and Papaperou (2001).

5 Invesmen Managemen and Financial Innovaions, 4/ SBC, he indusrial producion index (LIP) is inegraed of order one, i.e., I(1). For he KLCI series (LCI), according o he AIC and SBC, he uni roo es fails o rejec he null hypohesis of he exisence of a uni roo in log levels bu can be rejeced in he log firs difference for he whole period and for he firs wo sub-periods only. Therefore based on he AIC and SBC, he KLCI (LCI) and IP (LIP) are I(1) excep for he sub-period of 1998m9-2004m12 which is I(0). Nex, he number of significan coinegraing vecors are esed for he I(1) series. The likelihood raio (LR) ess are performed o deermine he lag lengh of he VAR. The selecion of lag order of he unresriced VAR is deermined by Akaike informaion crierion. Then we proceed wih coinegraion es according o equaion (1). Table 1 reveals ha here exiss a coinegraion relaionship beween he sock price index and he indusrial producion index for he whole period and sub-period of 1986m5-1998m7. Boh he race es and he maximum eigenvalue give similar resul ha is r=1 a 95% significance level. This implies ha sock prices and indusrial producion are in equilibrium in he long run. However for he sub-period of 1980m1-1986m3, here is no coinegraion beween he wo series. Tess for he number of coinegraing vecors of LCI and LIP The whole period of Order of VAR =2 Null Alernaive Max eigenvalue Trace es r = 0 r = * r<= 1 r = Sub-period of 1980m1-1986m4. Order of VAR =2 r = 0 r = r<= 1 r = Sub-period of 1986m5-1998m7. Order of VAR =12 r = 0 r = * r<= 1 r = *95% significance level; No inegraion for period of 1980m1-1986m4 Table 1 For he whole period of Dependen variable is dlip ECM for variable LIP and LCI Coefficien Sandard Error ecm1(-1).30071* Dependen variable is dlci ecm1(-1) For sub-period of 1986m5-1998m7. Dependen variable is dlip ecm1(-1).36915* Dependen variable is dlci ecm1(-1) *95% significance level Table 2 Nex he coinegraed series is represened as a Vecor Error Correcion Model (VECM) which specifies as:

6 120 Invesmen Managemen and Financial Innovaions, 4/2005 x az lagged( x, y ) y a z lagged( x, y ) The represenaion implies ha changes in dependen variable are a funcion of he level of disequilibrium in he coinegraing relaionship as well as changes in oher explanaory variables. More formally, if x and y are boh inegraed of order 1, and hey are coinegraed so ha Z xay is I(0), hen Engel and Granger (1987) demonsraed ha i mus be he above ECM is correc. If he error-correcion coefficien ( 1 or 2 ) in any equaion is insignifican, ha implies ha he corresponding dependen variable of he equaion is exogenous (i.e. i does no depend on he deviaions of oher variables). Bu if he coefficien is significan, ha implies ha he corresponding dependen variable is endogenous. The resul of ECM from Table 2 reveals ha for he whole period error correcion erm for he composie index equaion (LCI) is insignifican (ECT of ), which implies i is exogenous. This means i does no depend on he deviaions in he indusrial producion. I also implies ha i is a leading variable and iniially receives he exogenous shocks resuling in deviaions from equilibrium and ransmis he shocks o oher variables. When indusrial producion (LIP) is he dependen variable, he error correcion coefficien is significan (ECT of ), which implies ha i is endogenous (i does depend on he deviaions of he composie index). Similar resul is found for he sub-period of 1986m5-1998m7 where he ECT for LIP is significan. Variance Decomposiion (VDC) and Impulse Response Funcion (IRF) The VECM, however, does no provide he relaive degree of endogeneiy or exogeneiy among he variables. We hen proceed o es he VAR equaions for variance decomposiion (VDCs) and impulse response funcion (IRF). From Table 3 he repored numbers indicae he percenage of he forecas error in each variable ha can be aribued o innovaions in oher variables a nine differen ime horizons. For he composie index, 99% of is variabiliy is aribued o shocks in iself hroughou he whole period. This shows ha he LCI is he mos exogenous. For he indusrial producion variable, own shocks accoun for 99% of he forecas variance for he firs monh. In he long period, is variaions are explained by 40% of is own changes, while approximaely 60% are aribued o composie index, hus provide evidence ha i is endogenous and can be prediced by he composie index. For he sub-period of 1986m5-1998m7, 97% of he variabiliy in indusrial producion is aribued o is own shock in he firs monh, and in he laer period, 60% of is variaion is explained by he composie index. The IRF gives similar resuls as he variance decomposiion (he resul is no repored here). (2) Orhogonalized variance decomposiion For he whole period of Table 3 VDC of LCI VDC of LIP Horizon LCI LIP LCI LIP E

7 Invesmen Managemen and Financial Innovaions, 4/ For sub-period of 1986m5-1998m7 Table 3 (coninuous) VDC of LCI VDC of LIP Horizon LCI LIP LCI LIP E The ARDL bounds es The ARDL mehod as suggesed by (Pesaran e al., 2001) does no require he assumpion ha he variables are boh I(1) or boh I(0). This es is employed o consider he saionariy issues ha may be encounered in he previous echniques ha required he non-saionariy of he ime series involved. In esing for he ARDL we concenrae on he F-saisic value of he ime series o check wheher here exiss he long-run relaionship beween LIP and LCI variables. Firs we check he F-saisic for indusrial producion as dependen variable F(LIP LCI). Reference is made o he able by Pesaran and Shin (1996), which has he lower and upper bound for k=1 a 10% significan level as (4.042, 4.788). The F-saisic value is lower han lower bound for he whole period, subperiod of 1980m1-1986m3 and 1998m9-2004m12, herefore he null hypohesis of no coinegraion and long run relaionship canno be rejeced. On he oher hand, we find ha F(LIP LCI)= is well above he upper bound for 1986m5-1998m7 indicaing ha sock price is he long run forcing variable in explaining economic aciviy. Similar procedure is done by aking he sock price as dependen variable. F-saisic F(LCI LIP) value is lower han lower bound for he whole period, sub-period of 1980m1-1986m3 and 1986m5-1998m7. However, for sub-period of 1998m9-2004m12, F(LCI LIP)= is well above he upper bound suggesing ha economic aciviy is he long run forcing variable in explaining sock price. Nex is o specify he Error Correcion Model (ECM). The error correcion version of he ARDL (p,q) model of he equaion for he indusrial producion index is given by: k i i i u 1 ij, j, j 1 j 1 j 1 j 1 k. (3) Herei is he indusrial producion a ime and +1 is he sock price a ime +1. The empirical analysis is based on monhly daa for Malaysia. Since he observaions are monhly, for maximum order of he lags in he ARDL model we choose 12. The hypohesis ha we will be esing is he null of nonexisence of he long-run relaionship defined by = =0. Under he alernaive of indusrial producion 0 and 0, here is a sable long-run level relaionship beween he variables. The ECM represenaion for he ARDL model is seleced by using he Schwarz Bayesian Crierion, where in his sudy we specify he maximum order is 12. The ECM represenaion using his crierion is given in Table 4. The ECM implies ha changes in he dependen variable are he funcion of he level disequilibria in he coinegraing relaionship i.e. he deparure from he long run equilibrium as well as changes in oher explanaory variables. The error correcion coefficiens, esimaed a for sub-period of 1986m5-1998m7 and for sub-period of

8 122 Invesmen Managemen and Financial Innovaions, 4/ m9-2004m12, are saisically highly significan, have he correc sign, and sugges moderae speed of convergence o equilibrium. In our case, for sub-period of 1986m5-1998m7, i implies ha LIP is endogenous ha is depends on he deviaion of LCI and oher variables (consan). I also implies ha he dependen variable DIP bears he brun of shor run adjusmen o bring abou long run equilibrium among coinegraing variables. Conversely, for sub-period of 1998m9-2004m12, LCI is endogenous ha is depends on he deviaion of LIP and oher variables (consan). 4. Conclusion In his sudy we examine wheher sock marke can predic economic aciviy in Malaysia. For he whole economy, he Johansen coinegraion and VDC ess seem o confirm previous findings of Fama (1990) and ohers who found evidence ha sock marke can lead changes o he economic aciviy. However he ADRL es is no able o capure any long run relaionship beween he variables. Bu when sub-samples are examined, all mehods demonsrae ha sock marke can lead economic aciviy for sub-period of 1986m5-1998m7. Ye for sub-period of 1998m9-2004m12, ARDL shows economic aciviy leads sock marke. Coinegraion ess canno be applied for his sub-period because all he variables are I(0). The resuls sugges ha sub-samples sudy is imporan o consider when esing he sock marke and he economic aciviy. This sudy also highlighs he usefulness of ARDL mehod especially when saionariy becomes an issue. ARDL bounds es Table 4 Error Correcion Represenaion for he Seleced ARDL Model ARDL(12,12) seleced based on Schwarz Bayesian Crierion For he sub-period of 1986m5-1998m7 Dependen variable is dlip Regressor Coefficien Sandard Error T-Raio[Prob] ecm(-1) [.000] ecm(-1) = *LCI *C Error Correcion Represenaion for he Seleced ARDL Model ARDL(10,0) seleced based on Schwarz Bayesian Crierion For he sub-period of 1998m9-2004m12 Dependen variable is dlci Regressor Coefficien Sandard Error T-Raio[Prob] ecm(-1) [.000] ecm(-1) = *LCI *C References 1. Akins, F.J., P.J. Coe. An ARDL Bounds Tes of he Long-run Fisher Effec in he Unied Saes and Canada // Journal of Macroeconomics No 24. pp Barro, R.J. The Sock Marke and he Macroeconomy: Implicaions of he Ocober 1987 Crash, in Rober W. Kamphuis e al. (eds.). Black Monday and Fuure of Financial Marke. New York: Irwin, Binswanger, M. Sock Reurns and Real Aciviy: Is There Sill A Connecion? // Applied Financial Economics, No 10. pp Canova, F. Tesing Long-run Moneary Neuraliy: Empirical Evidence for G7 Counries wih Special Emphasis on Germany // Carnegie-Rocheser Conference Series on Public Policy, No 41. pp

9 Invesmen Managemen and Financial Innovaions, 4/ Engle, R.F., C.W.J. Granger. Coinegraion and Error Correcion: Represenaion, Esimaion, and Tesing // Economerica, No 55. pp Fama, E.F. Sock Reurns, Expeced Reurns and Real Aciviy // Journal of Finance, No 45. pp Fama, E. F. Efficien Capial Marke: II // Journal of Finance, No 46. pp Fischer, S., R.C. Meron. Macroeconomics and Finance: The Role of he Sock Marke // Carnegie-Rocheser Conference Series on Public Policy, No 21. pp Hondroyiannis, G., E. Papaperou. Macroeconomic Influences on he Sock Marke // Journal of Economics and Finance, No 25. pp Johansen, S., K. Juselius. Maximum Likelihood Esimaion and Inference on Coinegraion-Wih Applicaion o Demand for Money // Oxford Bullein of Economics and saisics, No 52. pp Kwon, C.S., T.S. Shin. Coinegraion and Causaliy beween Macroeconomic Variables and Sock Marke Reurns // Global Finance Journal, No 10(1). pp Lee, B.S. Causal relaions among sock reurns, ineres raes, real aciviy and inflaion // Journal of Finance, No 47. pp Mahdavi, S., A. Sohrabian. The Link beween he Rae of Growh of Sock Prices and he Rae of Growh of GNP in he Unied Saes: A Granger Causaliy Tes // The American Economis, No 35(2). pp Muradoglu, G., F. Taskin, I. Bigan. Causaliy beween Sock Reurns and Macroeconomic Variables in Emerging Markes // Russian and Eas European Finance and Trade, No 36(6). pp Pearce, D. K. Sock Prices and he Economy // Federal Reserve Bank of Kansas Ciy Economic Review, No November. pp Peek, J., E.S. Rosengren. The Sock Marke and Economic Aciviy // New England Economic Review, No May/June. pp Pesaran, H.M., Y. Shin, R.J. Smih. Bounds esing Approaches To The Analysis Of Long-Run Relaionships // Journal of Applied Economerics, No 16. pp