Testing long-run monetary neutrality in Malaysia: Revisiting divisia money

Transcription

1 MPRA Munich Personal RePEc Archive Tesing long-run moneary neuraliy in Malaysia: Revisiing divisia money Chin-Hong Puah and Muzafar Shah Habibullah and Evan Lau and Shazali Abu Mansor Faculy of Economics and Business, Universii Malaysia Sarawak, Faculy of Economics and Managemen, Universii Pura Malaysia April 2006 Online a hps://mpra.ub.uni-muenchen.de/3750/ MPRA Paper No. 3750, posed 2. June 20 5:48 UTC

2 Tesing Long-Run Moneary Neuraliy in Malaysia: Revisiing Divisia Money Chin-Hong Puah a, Muzafar Shah Habibullah b, Evan Lau a and Shazali Abu Mansor a a Faculy of Economics and Business, Universii Malaysia Sarawak b Faculy of Economics and Managemen, Universii Pura Malaysia Absrac This sudy re-examines he long-run neuraliy (LRN) of money on real oupu in Malaysia using quarerly Divisia money daa from 98: o 2004:4 based on Fisher and Seaer s (993) nonsrucural reduced form bivariae ARIMA model. Special aenion has been given in idenifying he number of uni roo and coinegraing vecor, as a meaningful LRN es is criically depends on such properies. Empirical resuls indicae ha LRN is deviaed from Malaysian economy when Divisia money is used. In paricular, Divisia moneary expansion seems o have long-run posiive effec on real oupu in Malaysia. JEL Classificaion: C2, C43, E50 Keywords: ARIMA Model, Divisia Money, Long-run Neuraliy of Money. Inroducion There are voluminous sudies in finding ou he influence of money on real economy aciviy. Differen mehodologies and models have been applied o es he empirical validiy of money in order o give he answer wheher money is a viable moneary policy variable. Lieraure shows ha mos of he earlier sudies on money use differen level of moneary aggregaes such as M, M2 and M3 o evaluae he performance of money. Unil 26 years ago, he radiional or simple sum moneary aggregaes are he mos commonly used measure of money in he empirical lieraure. However, he use of simple sum moneary aggregaes has received much criicism because i is inconsisen wih he microeconomics and saisical index number heory. In he simple sum aggregaion mehod, differen moneary componen asses are given he same weigh. In oher words, he componen asses are regarded perfec subsiues. This implies ha coins and noes are assumed o provide he same ransacions or liquidiy services as ineres-bearing deposis such as savings deposis or ime deposis wihin he broad moneary aggregae. Clearly, his assumpion is no valid in he conex of opimizing agen. In realiy, neverheless, i is obviously ha he differen moneary componens included in he broad moneary aggregaes are acually imperfec subsiues. Generally, he role of coins and noes are limied o he funcion of media of exchange o faciliae marke ransacions. They do no serve as a Corresponding auhor: Chin-Hong Puah, Deparmen of Economics, Faculy of Economics and Business, Universii Malaysia Sarawak, Koa Samarahan, Sarawak, Malaysia. chpuah@feb.unimas.my

3 sore of value funcion because hey are non ineres-bearing. On he oher hand, savings deposis and ime deposis as well as oher newly emerging financial asses are ineres-bearing componens, which have a mixure of ransacions and sore of value characerisics varying from asse o asse and over ime. Due o hese differences, i is inappropriae o assume all he componens in he simple sum aggregae are perfecly subsiuable and given hem equal weigh in he consrucion of he aggregae. In he las hree decades, his deficiency becomes more apparen as financial innovaion and deregulaion inroduce many newly issue financial insrumens in he marke, in which provide boh ransacion and liquidiy services. As a resul, simple sum aggregaion of financial asses is subopimal wih respec o ransacions services of money. The more heerogeneous of he moneary componens, he less valid he simple sum aggregae as a measure of money. During he 980s, almos all Asian counries, including Malaysia in his sudy, have liberalized heir domesic financial sysems. The key reforms were aimed a liberalizing ineres raes, reducing conrols on credi, enhancing compeiion and efficiency in financial sysem, srenghening supervisory framework and promoing he growh and deepening of financial markes. The liberalizaion of he financial sysem was accompanied by he relaxaion of resricions on inernaional capial flows and a shif oward more flexible exchange rae arrangemens (Tseng and Corker 993, pp. 9). Financial liberalizaion ha brings o a more efficien and sophisicaed financial sysem, however, may lead o a breakdown of he sable and predicable relaionship beween moneary aggregaes and economic aciviy. In line wih he increase in he sophisicaion of he financial sysem, financial innovaion and deregulaion have pervaded in Malaysian financial markes. The emerging of newly issued financial asses has blurred he definiion of money because mos of hese financial asses are immediae or easy access, and given he marke-relaed ineres raes. The radiional simple sum aggregae canno disinguish beween he ransacion service and sore of value funcion provided by hem. Thus, here is a need for reassessmen of he appropriae insrumens of moneary policy in he conex of Malaysian economy. Alhough he use of weighed moneary aggregaes has gained much aenion, he exploraion in his area is relaively limied in he conex of Malaysian economy. Therefore, i moivaes us o address his issue by consrucing he weighed moneary aggregaes for Malaysia, and hen use hem o conduc a sudy o examine heir longrun impac on real economic aciviy. Specifically, we inend o re-examine he LRN of money sudy in Malaysia by Habibullah, e al. (2003) using Divisia moneary aggregaes proposed by Barne (980). 2. Divisia Index Hulen (973) poins ou ha in coninuous ime, he Divisia quaniy index (see Divisia, 925) is exac for he unknown moneary services (quaniy) aggregae. In D paricular, he coninuous ime Divisia index, M, is given by he differenial equaion: 2

4 D d log( M n ) d log( mi ) = si d d i= () where m i is he opimum moneary asse and he expendiure share for he i h moneary asse is defined as: s i miπi = n m π i= i i (2) I is clearly shown in he coninuous ime Divisia quaniy index, he growh rae of D M is equal o he share-weighed average of he growh raes of he moneary componen quaniies. Unlike he simple sum index, which simply assumes ha all componen moneary asses are perfec subsiues, he Divisia quaniy index assigns weigh o each of is componens according o he degree ha hey provide moneary services. Diewer (976) demonsraes ha here exiss a class of superlaive saisical index numbers, which are exac for second-order approximaions o unknown economic aggregaor in he linearly homogeneous funcion in discree ime. One of he mos imporan superlaive index numbers is he Tornqvis-Theil discree ime approximaion o Divisia coninuous ime quaniy index. For moneary aggregaion, he Tornqvis-Theil moneary quaniy index (or more commonly known as Divisia index) is defined as follows: ( s i + s i, 2 ) n = TT TT mi M = M (3) i mi, Diewer (976) shows ha Divisia index is exac for he ranslog flexible funcional form, and i provides a second-order approximaion o he unknown subuiliy funcion obained from he microeconomic opimizaion. Barne (980) advocaes he use of Divisia index due o is sraighforward inerpreaion, which can be seen by aking he logarihms of Equaion (3): n i i i, i= TT TT log M log M = s (log m log m ) (4) where s i = ( s +, ) 2 i si is he average expendiure share for all i. Equaion (4) clearly indicaes ha he growh rae of he Divisia index is simply a weighed average of he growh raes of componen moneary asses. The Divisia index is in line of microeconomic heory. They are regarded high qualiy saisical approximaions of he rue, bu unknown aggregaes in he uiliy funcion. Therefore, i is clear ha he Divisia index is, a leas heoreically, superior o he simple sum index. The relaive performance of hese wo indexes in empirical applicaions, however, is acually an empirical issue. 3

5 3. The Fisher and Seaer Mehodology and Daa The classical heory of macroeconomics assers ha here exiss a classical dichoomy in which nominal variables has no effec on real economic aciviy in he long run. This line of research has araced grea academic ineres for a long period. There are various economeric procedures in esing his classical quaniy heoreic proposiion. One of he leading approaches in esing he LRN of money is he use of a nonsrucural reduced form bivariae ARIMA model developed by Fisher and Seaer (993, hereafer FS). FS propose he use of a simple and relaively srucural free model because srucural deails are no relevan o LRN as i does no depends on he shor-run dynamics of he economy. Their es is criically depending on he order of inegraion of money and real series. In paricular, a conex of valid LRN es exiss only when he order of inegraion of he money and real series is a leas equal o one, and equal for boh series. Furhermore, here should be no common rend exiss wihin he wo-variable se daa. Le m be he log of nominal Divisia money and y is he log of real oupu: a ( L) d( L) m y m y = b ( L) = c ( L) y m y + u m + w (5) where represen he firs difference operaor, a(l), b(l), c(l) and d(l) are disribued lag polynomials in he lag operaor L, wih a 0 = d 0 =, and b 0 and c 0 are no resriced. m and y are he orders of inegraion of he money and real oupu. The LRN can be defined in erms of he long-run derivaive (LRD) of y wih respec o a permanen change in m as follows: LRD y / u + k y, m limk (6) m+ k / u where lim k m +k / u 0. If lim k m +k / u = 0, here will be no permanen innovaions in he level of money and hus he neuraliy proposiions canno be esed. LRD y,m expresses he ulimae impac of an exogenous money disurbance on y relaive o ha disurbance s ulimae impac on m. There are permanen changes in boh m and y when m and y. If he variables have he same order of inegraion, m = y, LRD y,m can be reaed as he long-run elasiciy of y wih respec o m and i can be evaluaed using he impulse response represenaion of Equaion (5). The special case occur when m = y =, hen he LRD y,m = c()/d(). LRN requires ha LRD y,m = 0 if y is a real variable. When he error vecor (u w ) is iid (0,σ 2 ), and he money supply is exogenous, he erm c()/d() is he Barle esimaor of frequency-zero coefficien in a regression of y y on m m, and i can be esimaed by lim k β k, where β k is he slope coefficien from he following regression: 4

6 k y k m y j = k + βk j= 0 j= 0 α m j + ε k (7) LRN is esable when m = y =, and Equaion (7) becomes: ( y y k ) = α k + βk ( m m k ) + ε k (8) The null hypohesis of LRN is β k = 0. The rejecion of he null hypohesis implies he deviaion of LRN. Quarerly ime series of real GDP and Divisia money a wo levels of aggregaion (M and M2) from 98Q o 2004Q4 were used in his sudy. Since he Divisia money daa are no available, we need o consruc hem using he asse componens and raes of reurn for each componen asse. For real oupu, we use GDP deflaed by CPI a a base year of All he daa can be obained from he Quarerly Saisical Bullein of Malaysia published by Bank Negara Malaysia he cenre bank of Malaysia. All variables were in he naural logarihm form. 4. The Empirical Resuls Since FS s es criically depends on he order of inegraion of he variables, he nonsaionariy propery of he daa needs o be examined properly. In doing so, we uilize he augmened Dickey-Fuller (ADF) (Said and Dickey, 984), PP (Phillips and Perron, 988), and he saionariy KPSS (Kwiakowski e al., 992) ess o check for he presence of a uni roo in he daa. Resuls of he uni roo ess are repored in Table. For he ADF and PP ess, i is clearly shown ha he daa are nonsaionary in levels bu conain a uni roo in heir firs differences. Meanwhile, in he KPSS es, he null hypohesis of rend saionariy is rejeced for all daa in levels form. However, we canno rejec he null of level saionariy in he firs differences of he daa in KPSS es. As such, we conclude ha he order of inegraion of he daa is one, and herefore, he LRN resricion c()/d() is hereby esable. Table : Resuls of Uni Roo Tess Series ADF PP KPSS Level Divisia M -.86(0) -.662(3) 0.265(4) Divisia M2 -.62(0) -.882(4) 0.89(4) Real GDP (5) (4) 0.87(4) Firs Difference Divisia M (3) (4) 0.095(4) Divisia M (0) -8.93(3) 0.00(4) Real GDP (4) (4) 0.054(4) Noes: Aserisks () and () denoe significan a he 5% and % levels respecively. The opimal lag lenghs for ADF, PP and KPSS ess were chosen based on Schwarz Informaion Crierion (SIC), Newey-Wes using Barle kernel and Schwer (987) formula, where k = [4(T/00) /4 ] respecively. As saed by FS, a meaningful LRN is esable in he absence of coinegraion beween money and real oupu. The reason behind is ha in order for money o be LRN wih respec o real variable, i mus exhibi insances of permanen change and 5

7 ha he respecive sochasic rends driving moneary and real variables are uncorrelaed in he long run. As repored in Table 2, he Johansen and Juselius (990) coinegraion es shows ha we canno rejec he null hypohesis of no coinegraion beween real oupu and boh Divisia M and M2. This furher indicaes ha he condiions necessary for meaningful LRN es hold for our daa. Table 2: Resuls of Johansen and Juselius Coinegraion Tes Series k λ-max λ-race H 0 : r=0 H 0 : r H 0 : r=0 H 0 : r Divisia M Divisia M Noes: Aserisks () indicae significan a he 5% level. Lag selecion is based on Schwarz Bayesian Crierion (SBC). For he nex sep, we proceed o apply Equaion (8) o es for LRN. The esimaed resuls are hen presened in graphical form in Figures and 2. We found ha he poin esimaes of β k in Figures and 2 end o move away from he zero line as he lag lengh (k) increases. The 95 percen confidence bands do no include zero excep for k < 2 for Divisia M and k < 5 for Divisia M2. This resul implies ha LRN does no hold in he economy of Malaysia when Divisia money is used. In paricular, we noice ha he Divisia moneary simulus do have long-run posiive impac on real oupu in Malaysia. FIGURE : REAL GDP on DIVISIA M: 98:-2004: Esimaed coefficiens Lag lengh (k) [k=,...,28] FIGURE 2: REAL GDP on DIVISIA M2: 98:-2004:4 Esimaed coefficiens Lag lengh (k) [k=,...,28] 6

8 5. Conclusion In his paper, we re-examine he LRN of money sudy in Malaysia by Habibullah, e al. (2003) using Divisia moneary aggregaes proposed by Barne (980). The purpose of our sudy is o provide alernaives o he moneary aggregaes currenly used by he Bank Negara Malaysia. Special aenion has been given o he nonsaionariy and coinegraion properies of he daa, since meaningful FS ess criically depend on such properies. We discover ha all of he series are I() and he Divisia money series do no coinegraed wih real oupu. We found evidence agains LRN, indicaing he permanen shocks o he level of Divisia money do have imporan effec on real economic performance. Our finding is in line wih Habibullah, e al. (2003) in which hey found ha he LRN hypohesis is no suppored when quarerly official moneary aggregaes are used. To conclude, his sudy provides empirical evidence ha suppor he usefulness of he Divisia moneary aggregaes as alernaive inermediae variable for he case of Malaysia. Acknowledgemen: The auhors acknowledge he financial suppor rendered by Universii Malaysia Sarawak (UNIMAS) hrough he Fundamenal Research Gran No. 03(74)554/2005(53). References: Bank Negara Malaysia. Monhly Saisical Bullein, various issues. Kuala Lumpur: Bank Negara Malaysia. Barne, W.A. (980). Economic moneary aggregaes: An applicaion of index number and aggregaion heory. Journal of Economerics, 4(), -48. Diewer, W.E. (976). Exac and superlaive index numbers. Journal of Economerics, 4(2), Divisia, F. (925). L Indice Monéaire e la Théorie de Ia Monnaie. Revue d Economie Poliique, 39, Fisher, M.E., & Seaer, J.J. (993). Long-run neuraliy and superneuraliy in an ARIMA framework. American Economic Review, 83, Habibullah, M.S., Mohamed, A., Puah, C.H., & Baharumshah, A.Z. (2002). The relevance of moneary aggregaes for moneary policy purposes in Malaysia. Borneo Review, 3(), -2. Hulen, C.R. (973). Divisia index numbers. Economnerica, 63, Johansen, S., & Juselius, K. (990). Maximum likelihood esimaed and inference on coinegraion wih applicaion o he demand for money. Oxford Bullein of Economics and Saisics, 52,

9 Kwiakowski, D., Phillips, P.C.B., Schmid, P., & Shin, Y. (992). Tesing he null hypohesis saionariy agains he alernaive of a uni roo: How sure are we ha economic ime series have a uni roo? Journal of Economerics, 54, Phillips, P.C.B., & Perron, P. (988). Tesing for a uni roo in ime series regression, Biomerika, 75(2), Said, S.E., & Dickey, D.A. (984). Tesing for uni roo in auoregressive-moving average of unknown order. Biomerika, 7, Schwer, G.W. (987). Effecs of model specificaion ess for uni roo in macroeconomic daa. Journal of Moneary Economics, 20, Tseng, W., & Corker, R. (933). SEACEN sudy on moneary policy and financial reform. In A. Talid (Ed.). Moneary policy in he SEACEN counries: An updae. Kuala Lumpur: The Souh Eas Asian Cenral Banks (SEACEN) Research and Training Cenre. 8