7/29/2019 jeko_33-4

1/22

Jlfnla! Ekollomi Malaysia 33 (1999) 65 - 86Financial Liberalization in ASEAN and the FisherHypothesisSiong Hook LawHui Boon TanAhmad Zubaidi Baharumshah

ABSTRACT

This sll/dy examines the long-rull relationship between inflation andIlOlI/illal interest rates in the 1990s by utilizing the JolulIlsen-Juse/iuslIlultivariate cointegration technique. The evidence supports the taxadjusted form of Fisher hypothesis for three ASEAN couHlries, namelySingapore. Malaysia alld Thailalld. Thus, the assumption of a stablereal imerest rate appears to have empirical support for these low-inflation economies. We also demonstrate that inflation rate is both weaklyalld strongly exogenous in these three systems. However, the weakfonnof the hypothesis is decisively rejected for the inflation prone cOllntrieslike the Philippines and Indonesia. III general, Ottr results suggest thar;'1 a deregulated eliviromneHls real interest rate is insulated fmJ1l nominal shocks alld money is neutral.ABSTRAK

Kajiall ini meneliti hubuJlgan jangka panjang antara inflasi dengankadar bunga 1I0millai dalam talilm 1990an del/gem menggwwkan teknikkoinregrasi }ohansefl-Jllselius. Keplltusan kajiaH lIlenyokong bentukpelarasan c{{kai Hipotesis Fisher unruk liga buah negara ASEAN, iaituSingapura , Malaysia dan Thailand. Maka alldafelll kesrabilan kadarbUl/ga bellar Ilampaknya menerima sokongan empirik di lIegara yangkadar inJlasin)'a rendah. Artikel jlli juga melllmjukkan ballawa kadarinflasi adalall 'weakly ' dan 'strongly exogenous' dalam ketiga-tiganegara tersebut. Walaubagaimana PUll. benwk hipotesis 'weakly'iniditolak bagi negara yang cendenmg kepada inj1asi yal/g til/ggf sepertiFilipilla dan Indonesia. Umumnya, kepUlusan kajian menerangkallbahawa di dalam Sllasalla deregu/asi, kadar bunga benar tidakdipengaruhi oleh kejutGn l1om;,wl dan wang adalall 'Ilelllral'.

7/29/2019 jeko_33-4

2/22

66 illmal Eko1!omi Malays ia 33INTRODUCTION

Irving Fisher (1930) hypothesizes that nominal interest rates adjust one-for-one with respect to the changes in the expected inflation rate and itforms the foundation to the interest rates theory. It implies that realinterest rates will not be affected by changes in expected infl ation.Changes in inflation will be reflected in nominal rates, leaving real ratesconstant, ceteris paribus. However, this does not suggest that real rateswi ll be constant over tim e. In fact, changes in real intere st rates ma y beinfluenced by real economic factors (Kinal & Lahiri 1988). Thi shypoth esis is often referred to as the Fisher effect and has an importantimplication for monetary policy. Indeed, if the hypothesis is upheld theninflation has no real impact on the real sector. In other words, inflationary movements will be absorbed in nominal rates and cannot beinfluenced by monetary policy.

The behavioral relationship between real intere st rate and inflationhas received a great dea l of attention in the past several decades; seefor instance, Fama (1975), Silvapulle (1987), Atkins ( 1989), Rose (1988),Garcia (1993), Yuhn (1996), Thornton (1996), Payne and Ewing (1997)and Law (1997). With the exception Garcia (1993), Inder and S ilvapulle(1993), Thornton (1996), Payne and Ewing (1997) and Law ( 1997),studies on the validity of the Fisher hypothesis were based on sa mpleof the developed countries with low inflation rates. Nevertheless, theresults based on the data for the United States are far from conclusivewith Fisher effect estimates significantly less than the implied value of1.0 or greater, depending upon whether taxes are included or not in themodel (see, Summers 1983; Crowder and Hoffmen 1996). The conflicting resu lts obtained from previous studies in part are due to methodology formation of inflationary expectations, treatment of taxes, shift inmonetary regime, as well as the country examined. The empirical evidence seems to suggest the possibility that economi c agen ts suffer frommoney illusion, which is a clear violation of rational expectations.

Recent advancement in time-se rie s analysis particularly thecointegration technique provides some advanced techniques to te st forthe Fisher hypothesis as a long-run proposition. Unfortunately, studiesthat have utilized the techniques have also produced mixed results, andhence the validity of the hypothesis remains an empirical issue. Forinstance, Rose (1988) and Inder and Silvapulle (1993) reject cointegration(long-run) relationship between inflation and nominal interest rates, whileAtkins (1989) and Thornton (1996) show strong ev idence of cointegrating

l

l

ctl

tJtI

eane,pa,citablh:Sh,fc

7/29/2019 jeko_33-4

3/22

Financial Liberalization ;n ASEAN alld the Fi sher Hypotliesis 67relationship between them. Garcia (1993) and Thornton (1996) reportthat the Fisher hypothesis holds for Brazil and Mexico, respectively,whereas Dull and Ghosh (1995) al so using the coinlegration methodsobserve that the hypothesis is overwhelmingly rejected for the Canadaunder both fi xed and floating exchange regimes. Contrary to thjs study.MacDonald and Myrphy (1989) found mixed results with the Fi sherrelation be ing upheld (not upheld) under fixed (floating) exchange ratesystem.

Mishkin (1992) conducts the test on u.s. data to show that the Fi shereffect appears to be strong only for particular sample period but not forolhers. Hawtrey ( 1997) using Austra lian data shows that while the Fishereffec t fail s prior to the financial deregulation of the 1980s, there isev idence fo llowing deregulation th at the relationship is res tored. Thus,implying that the test results are sensitive to time period over whi ch therelatjon is estimated. Similarly, Evans and Lewis (1995) show the longrun relationship between nominal interes t rates and inflation but rejectthe strong form of the hypothesis. They ascribed th eir observed lessthan-unity expec ted inflation coefficient to the changing dynamics ofinflation over the chosen sample period. Payne and Ew ing ( 1997)examined the Fi sher hypothesis for a sample of nine less developedcountries and rejected the weak form of the hypothes is for five coun-tries of which include Argentine, Fiji , India, and Thailand. Indeed, theyargued th at the source for the absence of co integration in these countries was that rea l interests were non-stat ionary in level s. For Singapore.the evidence suggests that long-run Fisher relationship is upheld but thestrong form is rejected at the 10% significance level. The data providesev idence of unit proportional relationship between pre-tax nominal ratesand inflation for Malaysia, Pakistan and Sri Lanka.

The primary purpose of this study is to in ves tigate the empiricalrelationship between nominal interes t rates and inflation, in five A$EANeconomies: Singapore, Malaysia, Thailand, Indonesia, and the Philippines. IMF classifies Malaysia, Indonesia, Thailand and the Philippinesas emerging markets while Singapore is included as deve loped financial market. In thi s paper the ana lysis is based primarily on the aftertax rate of interest since earlier studies that run the Fi sher equation usingbefore-tax nom ina l rates are in general not supporti ve of the Fisherhypo thesis fo r the ASEAN economies for 1975-96 period, except forSingapore (see al so Law 1997). In this study, the weak form of thehypothesis is upheld for the period 1990-1996. Nevertheless, the strongform of the hypothesis is decisively rejected for all of the ASEAN coun-

7/29/2019 jeko_33-4

4/22

68 Jllrnal Ekollomi Malaysia 33tries, excep t for Si ngapore. In addition, we utilized the Johanse ncointegration approach to overcome the problem of non-stationarity oftime series data.Several reasons have motivated this study: First, as pointed byHawtrey (1997), the real rate of interest is important in determiningeco nomic growth, saving and investment, and any di stortion in post-taxreal financial asset will have adverse impact on economic performance.The real rate of interest affects th e real exchange rate, an in turn tradeand capital flows. It is of interest to economists, therefore, whether themacroeconomic determination of the real rate is ' purc' and undistorted.Second, if the Fi sher neutrality hypothes is holds then nom inal interestrates is an un biased predictor of future expected inflation. Bus inessesand government may use the relationship for decision-making purposes .Third, the ASEAN countries have taken major steps in 1990s to liberali zetheir fi nancial markets in order to promote economic e ffi ciency. Financial liberalization is expected to have some impact on th e interest rateinflation relationship. Specifically, we expect to find the Fisher effec tto hold in the pos t deregulation era. Thus, th e resu lts in this paper throwsome lights into this debate.

The ASEAN-S shares th e mix characteristics of both the developedand developing economies. The currencies of these countries are managed floats with varying weights given to major trading partner currencies. The five economies adopted different degree of financial liberalization as they took major steps to reform their fi nancia l system in thepast decades. While most of these countries have relatively open cap italaccount durin g the period of study, the Philippines still maintained controlover capital move ment s. Hence, it is interesting to investigate whetherthe ir past and current reforms have any impact on th e intere st ratesinflation re lation. Uncertainty in the reg ional financial cri sis in the firs tthree quaflers of 1998 forced Malaysia to implement selecti ve capitalcontrols. On September 1 I ] 998, the Malaysia ringgit was made inter-nationally non-convertible, and was fixed at RM3.80/us dollar. At thesame time, capital had to stay on shol1 for at leas t 12 1110nths and notmore than RM 10 , 000 could be taken out of the country.

In thi s study, the standard classical unit root procedures are employed, in th e first step, to examine th e time series properties of theti me series. Next, the Johansen multivariate cointegration procedure isutilized to determine the presence (or absence) of a long-run equilibrium relationship between after-tax nominal interest rate and inflation.The dynamic behav ior of interest rates and inflation is also examined

b(h\ IITtl

TET

a'TdlahIg. .

; - . " " ,YearFIGURE 5. Three-month treasury bill rate and CPI in flat ion

of the PhiJippines

,

METHODOLOGY AND DATATh is study employs the multivariate co inLegration analys is to examinethe long-run equilibrium relationship between the nominal interest rates

Fin

an(an(bebet iObeforfraJo

UNIt ibeanipoteqlPhiinl(proptcrit

T ilha\lf tCotiabeno ta Igelwilshupro(19(cooft

7/29/2019 jeko_33-4

9/22

Fillancial Liberalization ill ASEAN and the Fisher HypO/flesis 73and expected in fl ation rates. As pointed out by DUll and Ghosh (1995)and o th ers the procedure has the added advantage of distingui shi ngbetween th e weak foml of the Fi sher hypo th esis (long-run comovement)between interest rates and inflation from th e strong fO fm (unit proportional co movement). The weak: form of the Fisher hypothesis is said tobe uphe ld i f th e two vari ables are fou nd to be cointegrated. The strongfo rm of the hypoth esis can easi ly be tested within th e l ohansen-luseliusframework by computing th e likelihood ra tio (LR) test as decribed inJohansen and Juselius (1990).UN tT ROOT TESTS AND ORDERS OF INTEGRAT IONIt is important to determine the characteri stic of th e individual seriesbefore conducting the cointegration analysis. Th is is due to the fac t thato nl y var iables of the sa me o rder of integ ra ti on may constitu te apotential cointegration relationship. In this study, two asymptotica llyequiva lent procedures, Ihe Augmented Dickey Full er (A DF) and thePh jllips and Pe rron (pp) tests are employed 10 determi ne the order ofintegration for all the se ri es. The cri tica l va lues for these tests areprovided in MacKin non (1991). [n apply in g both of these lests, theoptimal lag structures are determined using the Akaike info rmationcriteria (AIC) and Schwarz criteria (SC).JOHANSEN MULTIVARIATE CorNTEGRATION TESTThe cointegrat ion analysis is performed iF the variabl es in th e systemhave the same order of integration as indica ted by the unit rool tests.If the are integrated of di fferent orders, they cannot be co integrated.Co integralion among variabl es means th at one or mo re li near comb in ation of these variables is stationary, even though the variables might notbe stationary individuall y. I f th ese variables are co integrated, they cannot move 'too far' apart from each other. From a statistical perspective,a long-run equilibrium relat ionsh ip means that the variables move together over time so that any short-run deviation from the long-term trendwill be correc ted. These series are said to be cointegrated and the reforeshared a common root stochastic trend.

In this study, the mult ivariate max imum like li hood co integrationprocedure developed by Johansen ( 1988) and Johansen and Juselius( 1990) is used to determ ine th e number of co integrat ing vectors(cointegration rank) in th e system. This procedure has improved someof the limitations of the bivariate model introduced by Engle and Granger

7/29/2019 jeko_33-4

10/22

74 JUnlal Ekollomi Malaysia 33(1987). Gonzalo (1994) has shown that the Johansen method perfonnsbetter than the single equation methods as weU as alternative multivariate methods (see also Johansen 1988; Enders 1995). This proceduresreconsider the n-variable first-order VAR given by

(I)By subtracting \ .1 from each side of the equation, Equation (I) can berewri Hen as

Llx, = Ar ,o} - x/oj + c/= (A 1 - l) X,.I + E,;;;; 7r x /oj + , (2)

where XI and , are (II x 1) vectors; A is an (II x 11) matrix of parameters;I is an (n x n) identity matrix; and" is defined to be (A 1 - f). The rankof 1t equals the number of cointegratin g vectors.

The model in equation (2) can be generalized to allow for a higher-order autoregress ive process, that is

. - 1Ax :;;: L 7r1Ax'_i + 1! x'_m + 1I i= ]

(3)

and the key feature is still the rank of " which is equal to the numberof independent cointegrating vectors. As we know that the rank of amatrix is equal to the number of its characteristic roots that differ fromzero, thus the number of distinct cointegrating vectors in thi s model canbe determined by chec king the significance of the characteristic rootsof n. The test for the number of characteristic roots that are insignificantly different from zero can be conducted usi ng the following twotest statistics:

A, ( r ) = - T I n ( l - ~ ) (4)i" , r + L

A, ( r,r + I) = -T In (1- A, ) (5)mu r+ 1where).. are the estimated va lues of the characteristic roots or eigenvalues obtained from the estimated n matrix, and T is the number of usableobservations. The first statistic tests the null hypothesis that the number

Fi

oalthmeioVt

esIe;e(vaobe

x

wa 'enInwniJ

7/29/2019 jeko_33-4

11/22

Financial Liberalization ill ASEAN alld the Fisher Hypothesis 75of cointegrating vectors is less than or equal to r against a generalalternative. It is referred to as the trace test. The second statistic teststhe null that the number of cointegrating vectors is r against the alternative of (r + I ) cointegrating vectors. It is known as the maximumeigenva lue tes t. Critical values for both the tests are tabulated inOsterwald-Lenum (1992).VECTOR ERRORCORRECTION MODELING (VECM)When a system of variables are cointegrated, their time paths are influenced by the extent of any deviation from long-run equilibrium. I f thesystem is to return to the long-run equilibrium. the movements of atleast some of the variables must respond to the magnitude of the disequilibrium. The main purpose of vector error-correction modeling(VECM) is to focus on the short-ru n dynamics of the variables, whi le theadjustment on long-run disequilibrium is also captured through the setupof error-correction terms in the model.

The testing of causal relationship in the environment of VECM canbe represented by the following example:

tlx cc /3 ( L)/3 (L) /3 (L) tlx yz" " 11 '0 " I I . , - Itlx cc, /3 (L)/3 (L) /3 (L) 6., Y zX " 2\ 22 '0 " 2. 1 - I; ; + +

tlx cc /3 ( L)/3 (L ) /3 (L) tl x Y z"' 0' 0' "' "' " M. , - I

(L)OO I. ,O(L)OO 2. I+OO(L) o . ,

where XI is an (n x I) vector of variables in the system, a's representa vector of constant terms,j3's are estimable parameters, is a difference operator, L is a lag operator,ji(L) and (L) are finite polynomialsin the lag operator, ZI_l'S are error-correction term s, and c('s are jointwhite-noise processes.

The Granger-causality can be detected through the statistical significance of the t-statistic for the lagged error-correction term and lor

7/29/2019 jeko_33-4

12/22

76 Jurnaf Ekollolll i Malaysia 33the F-statistic appl ied 10 the joint signifi cance of the sum of lags ofeach explanatory variable. The non-significa nce of both the {.:stat isticand F-statistic in the system indicates econom etric exogeneity of thedependent variable. In addition to indicating the direction of causali tyamongst variables, VECM al so allows us to discriminate between theshort-and long-run Granger-causa li ty. The F-test of the exp lanaLOryvariables (in their first differences) indicates the "short-run" causaleffects, whereas the "Iong- run" adjustment of the dependent variablewith respect to dev iati on from equil ibrium is implied through th e significance of the I-stati stic of the error-correction lCIlTI .THE FISHER EQUATIONFisher (1930) argued that within a country, the nominal interest rateconsists of two components, a real rate and a premium for expectedinflation. The we ll known Fi sher effect states that changes in the expectedinflation rate will be reflected in the nominal interest rate one-for-one,ceteris paribus, and hence keeping the rea l interest rate constant over time.The Fi sher equation can be estimated by using the following model:

" = a + J3 J( , + , (6)where it is the changes in nominal interest rates in period t, 1(, is thechanges in expected inflation in a given time period T, stated in logarithm terms; and 1 is the residual , assumed to have a zero mean andrinite variance. If i, and '](, are both integrated of order one, and th eyare cointegrated then Eq uation 6 can be said to depict the whole formof Fisher effect. The absence of co integration relationship when bot hvariables are I( I) would imply the absence of long-run Fisher effec l.

The Johansen proced ure allows us to do hypotheses tes ts on theparameters from Equation 6. The one-ror-one movement of nominalinterest rate and expected innation, is tes ted by imposing the res trict ionJ3 = 1 on the cointegrating regression. The tes t on the null hypothes isHo: jJ = 1 is co nsidered as the strong form of Fisher hypothes is. Thehypothesis can be co nducted using standard asymptotic ch i- square testsunder the Johansen maximum likelihood approach. 1l is in this specialcase that the nominal interes t moves one-ror-one with the rate of inflation and that the full Fi sher effect is said to hold.Th e adequacy of the regression model is checked by employing ase ries of diagnostic tests to ensure the validity of the results for stati s-

lithinPisInindeaaminininsuth,inexMSisaD.fOth,doN,SiIn.arcbythtof19

7/29/2019 jeko_33-4

13/22

Fillallcial Liberalization in ASEAN and the Fisher Hypothesis 77tical influence . Th ey are the No rm tes ts for normality of the res iduals.the ARCH test for Autoregressive Conditional Heteroscedastic ity effectsin residuals, HET tes t for res iduals heterosceda sticity te sts and the BoxPierce S t a t i s t i c test for autocorrelation.SOURCE OF DATAIn this study, the three-month Treasury bill is used as th e nominalinterest rates for Malaysia, Singapore and the Philippines . Due to thedeficiency of data, the nominal interest rates for Indonesia and Thailandare proxies by the three months S8t (Bank Indonesia Certificate) rateand th e money market rate, respec ti ve ly. A ll the interest rates aremarket determinant rates, which are free From monetary authorities 'intervention. The consumer price index (CPI) is used as a meas ure ofinflation for each of these countries.

Several studi es have focus on the behavior of interest rate s andinflation during different pel'iods (e.g. Mi shkin 1992). These analysissuggest that the interest rates and inflation have not been invariant overthese sub-periods due to institutional factors and the forecastability ofinnation. To overcome this problem our study is based mainly on theexperiences of 1990s. Monthly data spanning from Janua ry 1990 toMarch 1997 (87 observations) are used in the analysis for Malaysia,Singapore and Indones ia. Due to some problems in data co llection, thesample pe riod for Philippines onl y covers from January 1990 toDecember 1996 (84 observations). Since monthl y data are not avai lablefor Thailand, the analysis is conducted using quarterly frequency datathat cover from 1982: I to 1996:4 (60 observations).

The data of trea sury bill and CPJ for Malaysia, S ingapore and In-donesia are co llected, respectively, from Monthly Bulletin of BankNegara Malaysia, Monthly Statistical Bulletin of Monetary Au thority ofSi ngapore, and Indonesia Financial Statistics, published by th e Bank ofIndonesia. The Treasury bill and CPt for Thailand and the Philippinesare both co llected from the International Financial Statistics, publishedby LMF .The expected inflation rate for each countries is comp uted, usingthe rational expec tation approach, as a function of the distribution lagsof the past rate of inflation (see Fisher 1930; Fama 1975; and Amsler1986). The results are presented in Table I.

7/29/2019 jeko_33-4

14/22

78 lurnai Ekonomi Malays ia 33

CountryMalaysia

Singapore

Indonesia

Thailand

TABLE 1. Estimation of the expected inflation(The Model: 7[ = ao + a j 1tt _1 + .. ,ak1tt _k + e t )

Regression equation R'rr = 0.9999 + 0.7550rr,, 0.6702(13.140) (4.270)rr = 0.5942 + 0.7728rr ., 0.6496(12.550) (3.598)rr = 2.4020 + 0.926Irr,. , - 0.2068" 0.6913

(8.8 16 ) (-2 .177) (5134)rr = 2.0454 + 0.5940rr" 0.3750

(5.899) (3.870)

DW Lag1.8679

1.6333

1.440 1 2

1.6034

Philippines rr = 1.689 1 + 0.9874rr,., + 0.0539rr,. , 0.8204 1.2701 30.2008]("

(9.070) (0.3484) (-1.898) (3.255)

Note : The values in parentheses represent the I-statistics.

ESTIMATION RESULTSAs a preliminary step towards testing the Fi sher relation, we investigatethe time-series properties of the data used in the analysis. To determinethe order of integration of each series under study, the AugmentedDickey-Fuller (ADF) and the Phillips-Perron (pp) unit root tests areconducted first on the leve ls and then on the first differences of thelogarithms of nominal interest rate and expected inflation. Te st resu ltsfor levels are rcported in Table I panel A, and those for the first differences are summarized in panel B of the same table. As shown inpanel A of Table I, the unit root hypothesis cannot be rejected for anystati stic test, indicating that the series are non-stationary in their levels.However, the series attain their stationary after the first differencing inall cases. Results presented indicate that the null hypothesis of a unitroot is rejected for all the cases at the 5% significance level or better.In addition, the hypothesis of a deterministic (linear) trend in the datais tested and rejected. Thus, confirming the presence of one unit root

I fDf(alrasiC(lab(c(C,b,n(p'ficP'thI f loecFimccthge2)itJooconacaaunoth.afl

7/29/2019 jeko_33-4

15/22

Fillallcial Liberalization in ASEAN alld the Fisher Hypothesis 79in our data series. The results are not surprising given the evidence inOut! and Ghosh (1995) and Payne and Ewing (1997), among others.

We conclude that both the nominal interes t rates and inflation ratesfor the ASEAN economies are integrated of order one, that is, they areall """ I( I). The results from univariate unit root tests suggest that interestrates and inflation are driven by nonstationarities and justify their inclusion in the VAR system in order to investigate the existence of long-runcomovements in interest rates and inflation. If the textbook Fisher relation holds, it implies that they should share a common stochastic trend.

The presence (or absence) of a long-run equilibrium relationshipbetween the nominal interest rates and expected inflation can beconfirmed by conducting the lohansen (1988) max imum likelihood procedure. In the preliminary analysis, we first run the Fisher equation usingbefore-tax nominal interest rates. The evidence obtained are in generalnot in favor of the Fisher effect for all countries examined in the sampleperiod, except for Singapore. The model also suffers from misspecification so, only the post-tax results are reported and discussed in thispaper. In most case, normality test reject the hypothesis of normality inthe error terms. The equations also exhibit significa nt ARC H effect.

The Fisher hypothesis was re-examined using after-tax nominalinterest rates as dependent variables. Table 2 presents both the resultsof the trace and maximum eigenvalue tests for the tax adjusted Fisherequation. The model was estimated using lag length based on AkaikeFinal Prediction Error (FPE) criteria. Both the computed trace andmaximum eigenvalue (A-max) statistics consistently arrive at the sameconclusion. There is one cointegrating vector in the bivariate system forthe case of Ma laysia, Singapore, and Thailand. Columns 2 and 4 suggest that while the null hypothesis of r :;: I (agai nst the alternative r :;:2) cannot be rejected for Singapore, Ma laysia, and Thailand. However,it can be easily rejected for the case of the Philippines and Indonesia.Johansen (\ 988) demonstrates that in a special case when the memberof cointegration vectors is equal to number of variables these is nocommon trend between the variables, and that each variable is stationary. For this reason, further analysis relating to the Fisher relationcannot be performed for Indonesia and the Philippines.

These test results are insensitive to the chosen lag length in theautoregressive specification. Experimentation with high lag length didnot produce any significant different results especially with respect tothe number of cointegrating vectors in the system. This implies thatafter-tax effects on changes in inflation in the autoregressive specifica-

7/29/2019 jeko_33-4

16/22

80 JUnlaf Ekollollli Malaysia 33TABLE 2 . Un it root results

Variable ADF Tes t PP TestConstant Con stant Constant Constantwithout trend with trend without trend wi th Lrend . IA. Levels

[merest RateMalaysia - 1.6856( I) - 1.9 146( 1) - 1.9158(1) -2. 1202( 1)Singapore -2.5892(2) -3.1533(2) -2.7616( I) -3.25 18(1 ) ilIndones ia - 1.1077(0) - 1.5 192(0) - 1.1 733( 1) - 1.5780( I) iTh ai land -2.8098(0) -2.8666(0) -2.8013(1) -2.8872( 1) hPhi lippines - 1.5047(4) -2.7224(4) - 1.5506( I) -2.6956(1 ) f(

I'tpected Inflation slMa lays ia -2.5505(6) -2 .79 13(6) -2.4354( I) -3.3859(1) bSingapore -2. 1702(0) -2.6303(0) -2.09 13(1 ) -2.5595( I) sIndones ia -2.6430(0) -2.4620(0) -2.8247(1 ) -2.6482( I) FThailand -2.5941 (0) -3.2523(0) -2.4505(1) -3.35 16(1)Philippines - 1.4 175(0) -2 .1 292(0) -1.5862(1 ) -2.3 143(1) it

B. Fi rst Differences e,Im erest Ra te CMal aysia -3.4764(6) ' -3.4844(6) ' -1 1.535(1)* - 11 .479( 1) ' foSingapore -3.9338(7)' -4.0733(7)' - 11.076(1)' - 11.0 17( 1) ' InIndones ia -3.0919(4) ' -3.5576(4)' -8. 2683(1)' -8.3482( I) ' 10Th ailand -3.6864(3)* -3.6642(3) ' -8.2571 ( 1)* -8.2599( I) ' diPhilippines -4.9963(2) ' -4.9828(2)' -7.9870( 1) ' -7.7570( 1)' InExpected Inj1alioll reMalaysia -4.0640(3)' -4.07 18(3) ' -10.272( I) ' - 10.3 17( 1) ' peSingapore -3 .5535(5)' -3 .5219(5) ' -10.484(1) ' -10.432( I) ' 111Indonesia -3 .6403(7) ' -3.9702(7) ' -8.2683(1)' -8.3482( I) ' TtThailand -3 .9033(4) ' -3.9054(4) ' -8.4025( 1)* -8.2599( 1)* foPh ilippines -3.3130(3)' -3.3186(3)' -7.9870(1 ) ' -7.7570(1) ' 10

noNOles : The null hypothesis is that the series is J(l). Th e cri tical val ues for rejection arc pe-2.86 at a significam level of 5% for mode ls without trend and -3.41 for modelswith trend . Th ese values are provided by the SHAZAM OUiput based on MacKinnon C3I

(199 1). cmca!

7/29/2019 jeko_33-4

17/22

Fillancial Liberalization in ASEAN alld the Fisher Hypothesis 81tion are reflected in nominal interest rates in the three economies. Theadequacy of the specification of these models is checked by the diagnostic tests (residual normality. autoregressive, heteroscedas ticity andARCH effect) and is summarized in th e last three columns of Table 2.In general, the result of the diagnostic tests revealed that the VAR modelsare we ll specified.

Cointegration is only a necessary but not a sufficient to establishnominal interest rates adjust one-for-one lO inflation innovations. Themarket is said to be efficient, in th e sense of interest rates, if nominalinterest rates fully sunmlarize the infonnatjon on future inflation ratesif the strong form of the Fisher hypothesis holds. The test resu lts of thehypothesis thatjl = [I - I] are also summarized in Table 2. The strongfonn of the hypothesis is tested under th e c-ointegration regression forMalaysia, Singapore, and Thailand. The results presented in Table 2show that the null hypothesis of Fisher effect in the strong form can notbe rejected for these three economics. Based on the likelihood ratio teststatistic we conclude that the evidence support the existence of the fullFisher effect in three out of five ASEAN countries. This accords withintuition: we would expect agents to act on after-tax basis and investorsto protect their rea l returns, net of tax and inflation.

To gain some further insight about the dynamics of the system, weestimate the vector error correction model (VECM) that is the dual to thecoi ntegration model above. Table 3 presents a summary of the resultsfor the three ASEAN countries . The error-correction coefficient (Ecr),measures the one-period response of eac h of the endogenous variablesto a deviation from the equilibrium is corrected. The coefficient of ECTdisplays the expected negative sign implying that interest rates adjustin the opposi te direction of last period's deviation foml the equi libriumrelationship. These terms are statistica lly significant when the independent variable is interest rate, implyi ng that inflation adjusts to shortrun deviations from long-run equilibrium (ex-post real interest rates).This estimates seems to suggest a high speed, with around 57 percentfor Malaysia and 55 percent for Singapore, of any deviation from thelong-run ex post rea l interest rates in the current month. ft is worthnoting here that si ngle period response is smallest in Thailand, about 40percent in a quarter. However, these results must be interpreted withcaution since the analysis is based on different data frequency.

The F-tesls reveal that the lagged interest changes have no significant effect on inflation growth at a significance level of 5 percent in allcases. For example, in the equation for Malaysia tbe estimated coeffi-

7/29/2019 jeko_33-4

18/22

82 Jumal Eko11omi Malaysia 33TABLE 3. Johansen coinlegration resultsCri tical Trace Critical

Country A-max Value Test Value Lag Norm Arch Helro Q(8)Ma laysia (1990:0 1 - 1997:03)p=o 17.43* 14.1 20.87 15.4 10 5.92 10.08 52.55 3.44

3.44 3.8 3.44 3.8[H o: 3= IJ = 1.26(0.26)Si ngapore ( 1990: 0 1 - 1997 :03)p=o 16.39* 14 .1 19.25* 15 .4 8 1.84 3 .73 24.65 0.8 12.86 3.8 2.86 3.8[Ho: 3= I J = 2. 19(0.14)Thailand ( 1982: 1 - 1996:4)p=o 15.5 1* 14.1 17 .34* 15.4 5 0.82 1.56 3.39

1.83 3.8 1.83 3.8[Ho: 3=IJ = 3.36(0.06)Indonesia ( 1990:0 1 - 1997:03)p=o 14.08* 14 .1 16.6* 15.4 6 2.89 4.3 1 26.05 3.36

3.52* 3.8 3.52* 3.8Phi lippines ( 1990:0 1 - 1997:03)p=o 20.85* 14.1 26.29** 15.4 8 15. 12 6.80 42.43 3.2 1

5.45* 3.8 5.45* 3.8NOles: P indicate s the number of cointegraling vectors. The optimal lag-structure for eac h

mode l is determined through the likelihood-ratio test. Monthly frequency data areutilized fo r all the sample countries, except for Thailand.

cients of the lagged first-difference INF are jointly insigni ficant (F=O.989;p-value:0 .404). Thu s, inflation does n01Granger-cause interest ra les inthe short run. This result ho lds for the other two equ ations and so wecan conclude that infl ation is also stron gly exogenous to th e parametersof th e cointegarting system.

Nc

radafmalslatfaifivpraftco i

7/29/2019 jeko_33-4

19/22

Financial Liberalization in ASEAN and the Fi sher Hypothesis 83TABLE 4. Vector error correction models

Independent VariableDependent F - statistics Coefficient(std. error)Variable L1!NT MNF ECT[I_11MalaysiaL1!NT 0.2 14 0.989 -0.573(0.886) (0.404) (0.233)*L1!NF 0.293 1.220 -0.237(0.880) (0.3 10) (0.264)

SingaporeL1INT 1.460 1.445 -0.553(0.223) (0.228) (0.247)*MNF 0.390 0.426 -0.338(0.815) (0.789) (0.304)ThailandL1 INT 0.655 1.044 -0.47 1(0.590) (0.397) (0. 183)*L1!NF 0.575 1.1 73 -0.1 19

(0 .639) (0.348) (0.254)NOles: Standard errors in parentheses and p-value for the F-tests are given in brackets.

'*' Indicates significance at 5 percent.

CONCLUSIONS AND IMPLICATIONSThi s study examines the long-run relationship between nominal interestralesand inflation for five AS EAN economies in the 19905 using monthlydata (except for Thailand). Both nominal interest rates and inflation werefound to be non statioary in levels but stationary in first difference. Wealso fou nd that inclusion of taxes affect the infl ation-interest rates relation. Results based on the Johansen mu ltivariate coi ntegration approachfai led to reject the null hypothesis of non-cointegrat ion in th ree out offive of the ASEAN countries - Singapore, Malaysia and Thailand. Thus,provid ing empirical su pport for stable long-run co-movements betweenafter-tax nominal interest rates and inflation. What is supri sing is thatcointegration can be found in less than ten years of data for these econo-

7/29/2019 jeko_33-4

20/22

84 JUrl/al Ekonomi Malaysia 33mies. The weak form of the Fisher hypothesis appears not hold for th ecase of Indon esia and the Phi lippines.

The strong form of the hypothe sis was tested and was found tohold for countries wit h more efficient capital markets like Singaporeand Malaysia. In addition, our results show that the Fi sher effect holdsfor Thailand, which delayed their financial reforms till 1990. It is worthpo inting out here that for Thailand, ou r analys is is based on quarterlydata and the data begins from 1982: 1 and the results may not be com-pared to the rem aining ASEAN economies where the starting date beginin January 1990. While it may be imp ossible to draw strong conclu-sions, these results are at least suggestive that deregulation of thefi nancia l sectors in the recent year has impact on the behavior of thefina nci al variabl es. As expected in a deregulation environment realinterest rate is insula ted from nominal shocks and money is superneutral.

The Fisher effect is eas ily rejected for the case of the less libera-lized financial systems of Indonesia and the Philippines. These twocountries experience high and vo latile rates of inf1ation and interest ratesduring the episodes of currency and financia l instability. In fact. theevidence showed that cointegration was weak for countries with mana-ged exchange rate and capital controls like the Philjppines. While mostof the ASEAN countires have relatively open accoun ts in the 1990s, capitalcontrol stills remain in the Philippines. Partial adj ustment to anticipatedinflation meant that real rates was systematically affected by shocks tonominal money supp ly. Thus, in these two countries the monetaryauthority can affect real activities since money is not neutral.

Lastly, we limit our analysis to the period before the 1997 Asianfinancial crisis. The cri sis was proceeded by high interest rates, badbank loans and slow economic growth. During the crisis most of thesa mple countries (perhaps except for Singapore) have suffered majorpolicy reversals and some political instability which may ca use too muchnoise for cointegration method to detect the stable long-run relationshipas suggested by the Fisher equation. The recent cap ital controls inMalaysia may adversely affect the inflati on -interest rate rela tion. Ofcourse all the issues merit further research.REFERENCESAtkins, Frank 1. 1989. Cointegration, Error Correction and the Fi sher Effecl.Applied Ecollomics 21:161 1-1620.Amsler, Christine E. L986. The Fisher Effect: Sometimes Invcl1cd , Sometimes

Not? SOl/them Economic JOllmal 52:832-835 .

I

7/29/2019 jeko_33-4

21/22

Financial Liberalization in ASEAN alld the Fisher Hypothesis 85Crowder W.J. & D.L. Hoffman. 1996. The Long-Run Relationsh ip between

Nominal interest rates and infl ation: The Fisher Equation Revisi ted ,Journal of Money, Credit and Banking 281: 102-11 8.

Dutt, Swarna D. & D. Ghosh. 1995. The Fi sher Hypothesis: Ex amining theCanadian Experience. Applied Economics 27:1025-1030.

Enders. W. 1995. Applied Econometric Tim e Series. New York: John Wiley &Sons, Inc.

Engle, R.E & G.W.J. Granger. 1987. Cointegration and Error Correction ;Representation, Estimation and Tes ting. Econometrics 35: 143- 159 .

Evans, M & K. Lewis. 1995. Do Expected Shifts in Inflation Affect Esti matesof the Long-Run Fisher Relation? l Ol/mal of Finance 50:225-253.

Fama, E.F. 1975. Short-term Interest Rates as Pred ictors of Inflation. AmericanEconomic Review 65:269-282.

Feldstein . M. 1976. Inflation, Income Ta xes, and the Rate of Interest: A Theoret ical Analys is. American Ecollomic Review 66:809-820 .Fisher. L 1930. Th e Th eory of Illterest. New York: Macmillan .Garcia. P. 1993. The Fi sher Effect in a Signal EXlraction Framework: The Recent

Brazilian Expericnce. Journal of Developmelll Economics 41:71-93.Gonzalo, J. 1994. Fi ve Alternative Methods of Estimating Long-Run Equilib

rium Relationships. Jouma! of Ecollometrics 60:203-33.Hawtrey, K.M. 1997. The Fisher Effec t and Australian Interest Rates. Applied

Fillancial Ecol/omics 7:337-346.Inde r. B. & P. Silvapulle. 1993. Does th e Fisher Effect App ly in Australia?

Applied Economics 25:839-843.Johansen, S. 1988. Statistical Analysis of Co integ ration Vectors. Journal of

Economic DYllamics alld COI/frol 12:231-254.Johansen, S. & K. Juseliu s. 1990. The Full In formation Maximum Likelihood

Procedure for rnference on Co integral ion - with Application to the Demand for Money. Oxfo rd Bu llerin of Ecol/omics alld SWri slics 52: 169-2 1O.

Kinal, T. & K. Lahiri . 1988 . A Model for Ex-ante Reallnterest Rates and DerivedInfl a ti on Forecasts. Jou rnal of American Sratistical Associatioll83:665-673.

Law, S.H. 1997. The Fi sher Effect of Asean Economies: Evidence Based onCointegration and Spectral Analysis. Unpublished Master 's lhes is. UniversitiPutra Malaysia.

MacDonald. R., & P.O. Murphy. 1989. Testing for the Long Run Relationshipbetween Interest Rates and Infl ation Using Cointegralion Techniques.Applied Econometrics 21 :439-447.

MacKinnon. J. 199 1. Crit ical Values for Cointegration Tests in RF . In edited byEngle and Granger. Long-Rull Ecollomic Relar iol/ships: Reading illCoillregrarioll. New York: Oxford University Press .

Mishkin, ES. 1992. Is the Fi sher Effect for Real? A Reezamination of theRelationship bctween Inflation and In terest Rates. Journal of MonetaryEcol1omics 30: 195-2 15.

7/29/2019 jeko_33-4

22/22

86 Juntai Ekollomi Malays ia 33O s t e r w a l d ~ L e n u m , M. 1992. A Note with Quantilies of the Asymptotic Distri

bution of the Maximum Likelihood Cointegration Rank Test Statistics.Oxford Bulletill oj Economics and Statistics 54:461.472.Owen, P.O. 1993. Cointegration Analysis of the Fisher Hypothesis: The Roleof Real Rate and th e Fi sher Identity. Applied Financial Economics3:2 1-26.

Payne, lE . & B.T. Ewing. 1997. Evidence from Less Developi ng Countries onthe Fi sher Hypothesis: A Cointcgration Analys is. Applied EconomicsLeuers 4:683-687.

Rose, A.K. 1988. Is the Real Interest Rate Stable? JOllrnal of FinQll ce 43:1095-1112 .

Sargent, R. 1973. Interest Rates and Prices in the Long Run: A study of theGibson Paradox. JOlll71al of MOlley. Credit and Banking 86:212-225.

Silvapulle, P. 1987. A model for Ex-ante Real Intere st Rate in Australia: IsIt Constant? Unpulished Master 's Master thes is, Australian NationalUnive rsity.

Summers, L.H. 1983. The Non-Adjustment of No minal Lnterest Rates: A Studyof the Fisher Effect. In edited by 1. TobinoA Symposium ill Honor ofArthurOkun, Washington, D.C. Brookings In stitution.

Tanzi, V. 1980. Inflationary Expectations, Economic Activity, Taxes, and Interest Rates. American Economic Review 70: 12-2 1.

Thornton, 1. 1996. The Adjustment of Nominal Interes t Rates in Mexico: Studyof Fi sher Effect. Applied Economics Letters 5:385-349.Yuhn, K. 1996. Is the Fishe r Effect Robust? Further Evidence. Applied

Economics Leffer 3:4 1-44 .

Fakulti Ekonomi dan PengurusanUniversi ti PUlTa Malaysia43400 UPM SerdangSelangor Danl! EhsanMalaysia

J

FZN

AiA.aJk,dik.k.seyokeseSbtbdaibedi

A1thopabeCOinmHla