88 วารสารเศรษฐศาสตร์ มหาวิทยาลัยเชียงใหม่
14.02 กรกฎาคม-ธันวาคม 2553
Stock Market Development and Economic Growth: ARDL Causality in Thailand R ossarinOsathanunkul1 NisaFusrinual2 ChaiwatNimanussornkul3
Abstract
This paper endeavors to investigate the relationship between the stock market developmentandeconomicgrowthinThailand.The61quarterlydataduringthetimeperiod formMarch,1995toJune,2010utilizedinthisstudyarethegrowthrateofGrossDomestic Product(G DP)ofThailand,thegrowthrateofMarketcapitalizationofGovernmentBonds (BO)andthegrowthrateofMarketcapitalizationoftheStockexchangeofThailand(MC). Twostationaritytestsnamely,DF-G LStestandNg-Perrontestareemployedtofindthe integratingorderofthevariableswhichtheresultsrevealthatallvariablesarestationaryat theirlevelformortheyhavetheintegratedoforderzero,I(0).Totestlong-runrobustness, ARDLboundstestingtechniqueisapplied.Thefindingrevealsthatthereexistapositive relationship stock market development and economic growth in Thailand implying that stockmarketdevelopmentisanimportantingredientforeconomicgrowthinThailand.The findingofthestudysuggeststhatthereisaneedofpoliciestowardrapiddevelopmentof thestockmarketinThailand. Keywords:ARDL,StockMarketDevelopment,EconomicGrowth.
1 2 3
Professor of Faculty of Economics, Chiang Mai University. Research Assistant of Economic Research Park of Faculty of Economics, Chiang Mai University. Lecturer of Faculty of Economics, Chiang Mai University.
Stock Market Development and Economic Growth: ARDL Causality in Thailand 89
CMU. Journal of Economics
1.Introduction Thestockmarkethasbeencrucialandbecomeanimportantwheelforeconomicgrowth sinceitdoesnotprovideonlysourcesofexternalfinancingforfirmsandallocatecapitalto corporatesectorswhichimprovesresourceallocationbutalsoprovidethechangeofstock priceasaconsequenceofchanginginwealthwhichcanaffectthedemandforconsumption andinvestmentgoods,therebystimulatingrealeconomicactivityandboostupeconomic growth.Itappearsthatstockmarketscanstimulateeconomicgrowthinseveralways.First, stockmarketsplayanimportantroleinallocationofcapitaltocorporatesectorwhichresult inanincreaseinrealeconomicactivities(Shahbaz,et.al,2008).Second,stockmarketsmay encourageeconomicgrowththroughincreasingtheliquidityoffinancialassetswhichseems tobecrucialindevelopingcountries.Third,stockmarketsprovideinvestmentopportunities bymobilisingdomesticsavingswhichinturnpromotewiserinvestmentdecisions(Caporale et.al,2004).Thiscanbeseenthroughfinancialliberalizationinanumberofdeveloping countriesespeciallyacountrylikeThailandduringthepastdecadefinancialliberalization hasbeenrecognizedasasignificantpartofaneconomicpolicyindevelopingcountries. Itisbelievedthattheresultoffinancialliberalizationcanattractbothinternationaland domesticcapitalwhichisexpectedtoincreaseresourcesavailablefordomesticinvestment. There is plenty of research concerning the relationship between the stock market development and economic growth, including Bencivenga and Smith (1992), Atje and Jovanic(1993),GreenwoodandSmith(1997)andBellandRousseau(2001).Bencivenga andSmith(1 992)foundthatanewstockmarketcanleadtoeconomicgrowthbyreducing holdingsofliquidassetsandincreasingthegrowthrateofphysicalcapital.Similarly,Atje andJovanic(1993)concludedthatstockmarketshavebeenlong-runaffectedoneconomic growth and manipulate economic growth through a number of channels including liquidity,riskdiversifications,acquisitionofinformationaboutfirms,corporategovernance and savings mobilization. Greenwood and Smith (1997) also indicated that large stock marketscandecreasethecostofmobilizingsavings,thusfacilitatinginvestmentinmost productivetechnologies.BellandRousseau(2001)investigatedthelinkageamongindividual macroeconomicindicatorsandmeasureoffinancialdevelopmentinIndiawhichrevealsthat thefinancialsectorhasbeeninstrumentalinpromotingeconomicperformance. Therefore,thepaperproceedsasfollows,sectiontwobrieflyprovidesthedataand methodology,whichconsistoftheunitroottestbyusingtheDickey-FullerGeneralizesLeast Square(D F-GLS)testandNg-Perron(NP)test.Autoregressivemodelswithdistributedlags
90 วารสารเศรษฐศาสตร์ มหาวิทยาลัยเชียงใหม่
14.02 กรกฎาคม-ธันวาคม 2553
(ARDL)thenwasusedtoestimatetherelationshipbetweenthestockmarketdevelopment and economic growth in Thailand. Section three discusses the principle results of the econometricaltest.Thepaperendswithconclusion. Themainobjectiveofthestudyis: Toinvestigatetherelationshipbetweenthestockmarketdevelopmentandeconomic growthinThailand. 2.M ethodologicalFramework
2 .1TheModel Thisstudyaimstoinvestigatetherelationshipbetweenstockmarketdevelopmentand economicgrowthinThailandbyusingthefollowingmodel;
GDPt = f (GDPt–i , MCt , MCt–j , BOt , BOt–k)
Gt = log(
 where GDPt–i = ThegrowthrateofthepreviousiquarterofGrossDomesticProductof Thailand(i=1,…p). MCt = Theg rowthrateo fM arketc apitalizationo ftheS tocke xchange ofT hailand forcurrentquarter. MCt–j = ThegrowthrateofpreviousjquarterofMarketcapitalization oftheStock exchangeofThailand( j=1,…q). BOt = ThegrowthrateofMarketcapitalizationofThaiGovernmentBondsfor currentquarter. BOt–k = Thegrowthrateofprevious kquarterofMarketcapitalization ofThai GovernmentBonds(k=1,… r). Thegrowthrateofeachvariableattimetiscalculatedasfollows: )
 whereGt isthegrowthrateofthevariables, Yt andYt–1 arevariablesusinginthisstudy whichareG DP,MCandBO fortandt-1, respectively.
Stock Market Development and Economic Growth: ARDL Causality in Thailand 91
CMU. Journal of Economics
2.2Data
DatawasobtainedfromthebankofThailandandEcowindatabase.GDP(expressedas millionofThaiBHT)wasobtainedfromEcowinandthedataofstockdevelopmentincluding MarketcapitalizationoftheStockexchangeofThailandandMarketcapitalizationofThai GovernmentBondswereobtainedformthebankofThailand. 2.3Analysis
Inthisstudy,thekeystepsofananalysisaredefinedasfollows: • Theunitroottest TwotypeofunitroottestsnamelyDickey-FullerGeneralizesLeastSquare(DF-GLS) testandNg-Perron(NP)testwereusedtocheckthestationarityofvariables. • ARDLboundstestingtechnique Autoregressivemodelswithdistributedlags(ARDL)isemployedtoestimatethecausality relationshipbetweenstockmarketdevelopmentandeconomicgrowthofThailand. 2.3.1TheUnitRootTest Aunitroottestisrequiredtotestwhetherthevariablesinthisstudystationaryor non-stationaryandwhataretheorderofintegratedofthesevariables.Thus,weemploytwo typesofthestationaritytestsnamely,theDickey-FullerGeneralizesLeastSquare(DF-GLS) testandNg-Perron(NP)test. 2.3.1.1Dickey-FullerGeneralizesLeastSquare(DF-GLS)test DF-GLStestwasdevelopedbyElliotetal.(1996)whichiscalledde-trendingtest.Itis similartoAugmentedDickey-Fuller(ADF)test.However,ithasanadvantageovertheADF testwhenthereareasmallnumberofobservations.Thisde-trendingisdonebytakingthe explanatoryvariablesoutofthedata(see,Elliott,RothenbergandStock,1996).Thefollowing equationisthenestimatedtotestforaunitrootinthevariable:
Δydt = αydt–1 +
βpΔydt–p + vt
(2.3.1)
whereΔ isthedifferenceoperator,ydt istheGeneralisedLeastSquaresde-trendedvalue ofthevariable,α andβp arecoefficientstobeestimatedandvt istheindependentlyand identicallydistributederrorterm.AsinthecaseoftheADFtest,atestforaunitrootofthe variable y involvesexaminationofwhetherthecoefficientoftheAR(1)term,inthiscase
92 วารสารเศรษฐศาสตร์ มหาวิทยาลัยเชียงใหม่
14.02 กรกฎาคม-ธันวาคม 2553
α,inequation(2 .1)is α =0ortheseriesisnon-stationaryagainstthealternativeof α ≠0 ortheseriesisstationary.Inmakinginferences,thecriticalvaluestabulatedinElliott,
RothenbergandStock(1996)areused. FollowingDF-GLStestestablishedbyElliotetal.(1996),thenequation(2.1)canbe specifiedas:
ΔGDPdt = αGDP GDPdt–1 +
ΔMCdt = αMC MCdt–1 +
ΔBOdt = αBO BOdt–1 +
k=
MZda = [T–1(Ddt)2 – f0] /(2k)
MZdt = MZa × MSB
MSBd = (k/f0)1/2
MPdt = {[
βGDP, pΔGDPdt–p + vGDPt βMC, pΔMCdt–p + vMCt
(2.3.2) (2.3.3)
βBO, pΔBOdt–p + vBOt
(2.3.4)
2.3.1.2Ng-Perron(NP)test Ng-Perron (2001) developed four statistical tests by utilizing GLS de-trended data sets Ddt .ThecalculatedvaluesofthesetestsbasedontheformsofPhilip-P erron(1988) Zα and Zt statistics,Bhargava(1986)R1 statistics,Elliot,RotherbergandStock(1996) that createdbestoptimalstatistics.Thetermsaredefinedasfollows: (Ddt–1)2 / T2
 Whilede-trendedGLStailoredstatisticsareasgivenbelow:
–
(Ddt)2]/f0 , and, (
+ (1–
)T–1(Ddt)2/f }0
 Ifx t ={1}infirstcase,and xt ={1,t} insecond. NP test is a non-parametric approach to correct the residual autocorrelation. The regressionofthistestwasestimated.
Stock Market Development and Economic Growth: ARDL Causality in Thailand 93
CMU. Journal of Economics
Δydt = (δ–1) ydt–1 +
Δydt–p + ut
(2.3.5)
p
Thenullhypothesisofequation(2.3.5)is H0 : δ =1 ortheseriesisnon-stationary. FollowingNPtest(2001),thenequation(2.3.5)canbespecifiedas:
ΔGDPdt = (δGDP –1) GDPdt–1 +
ΔMCdt = (δMC –1) MCdt–1 +
ΔBOdt = (δBO –1) BOdt–1 +
GDPydt–p + uGDPt
GDP, p
ΔMCdt–p + uMCt
MC, p
ΔBOdt–p + uBOt
BO, p
(2.3.6) (2.3.7) (2.3.8)
2 .3.2AutoregressiveModelswithDistributedLags(ARDL) Therelationshipb etweenthestockm arketd evelopmenta nde conomicg rowthinThailand wasconductedbyAutoregressivemodelswithdistributedlags(ARDL).Oneofthereasons forpreferringtheARDListhatARDLcanalsobeusedtoexaminetherelationshipthestock marketdevelopmentandeconomicgrowthinThailandinpreviously.Inaddition,ARDLis morerobustandperformbetterforsmallsamplesizethanothercointegrationtechnique. ARDLcanbewrittenasfollows;
GDPt = ω0 +
αi GDPt–i + β0 MCt +
βj MCt–j + γ0BOt +
γkBOt–k + εt
(2.3.9)
where GDP = ThegrowthrateofGrossDomesticProductofThailand. MC = ThegrowthrateofMarketcapitalizationoftheStockexchangeofThailand. BO = ThegrowthrateofMarketcapitalizationofGovernmentBonds. t = Time(t=1,…,n). εt = Independentlydistributedrandomerrorterm,withzeromeanand constant varianceattimet. ω0 , αi , β0 , βj , γ0 , γk =Theparameterstobeestimated.
94 วารสารเศรษฐศาสตร์ มหาวิทยาลัยเชียงใหม่
14.02 กรกฎาคม-ธันวาคม 2553
3.E mpiricalResults
Inthissection,DF-GLStestandNPtestwereemployedtotestthestationarityofthe variablesthenARDLwasusedtoestimatetherelationshipofthestockmarketdevelopment and economic growth of Thailand. The 61 quarterly data during the time period from March,1995toJune,2010wasusedtodeterminetherelationshipbetweenstockmarket developmentandeconomicgrowthinThailand.Table2.1showsthedescriptivestatistictest ofthevariables. Table2.1DescriptiveStatisticsofthevariables
Variables Mean
GDP MC BO
Median Maximum Minimum Std. Dev. Skewness Kurtosis Jarque-Bera Probability
0.007 0.012 0.060
0.006 0.023 0.022
0.094 0.408 1.557
-0.082 -0.429 -0.272
0.044 0.164 0.270
0.366 -0.155 4.045
2.503 3.670 21.178
1.988 1.385 1006.232
0.370 0.500 0.000
3.1Unitroottest
Forthestationaritytesting,weemployedDF-G LStestandNg-P errontestratherthan ADFtest.ThereasonforusingthesetwotechniquesisthatDF-G LSandNg-Perronaremore powerfulandmoresuggestiveteststhanADFtestwhentherearesmallsamplesizesinceADF testisnotreliableforsmallsample(Dejongetal,1992andHarris,2003).Thestationarity testresultsareshownintable3.1 Table3.1Resulto fTheA ugmentedDickeyF uller(ADF)test,D ickey-F uller GeneralizesLeastSquare (DF-GLS)testandNg-Perron(NP)test
Variables
DF-GLS test
GDP MC BO
-3.378** [4] -6.250***[0] -3.812***[0]
MZa
-0.773 -28.626*** -18.983**
MZt
-0.421 -3.782*** -3.080**
Note: The number in bracket is the optimal lag length and bandwidth. Optimal lag length for ADF test is determined by AIC. Modified AIC is used determined the lag length in DF-GLS and NP tests. **,*** denotes the 5% and 1% level of significance, respectively.
NP test
MSB
0.545*** 0.132*** 0.162**
MPT
61.844 3.191*** 4.804**
Lag
[3] [0] [0]
Stock Market Development and Economic Growth: ARDL Causality in Thailand 95
CMU. Journal of Economics
ThestationaritytestresultsbaseonDF-GLStestandNPtestintable3.1showthatboth DF-GLStestandNPtestdisplaythenullhypothesiscanberejectedforthegrowthrateof MarketcapitalizationoftheStockexchangeofThailand(MC)andthegrowthrateofMarket capitalizationofGovernmentBonds(BO)implyingthatMCandBOarestationaryattheir levelformat1%and5%significantlevel. AlthoughonlyMSBstatisticaltestofNPtestrevealsthatthenullhypothesiscanbe rejectedforGDP,DF-GLStestrevealsthatthenullhypothesiscanberejectedat1%and5% levelofsignificant.Therefore,wecanconcludethatGDPisstationaryatitslevelform. Overallthestationarityevidencesshowthatallvariablesarestationaryattheirlevel formortheyhavetheintegratedoforderzero,I(0). 3.2 Theautoregressivemodelswithdistributedlags(A RDL)estimates The relationship and direction of causal relationship between the stock market developmentandeconomicgrowthinThailandisestimatedbyAutoregressivemodelswith distributedlags(ARDL),shownintable3.2 Table3.2Theautoregressivem odelsw ithd istributedlags(ARDL)estimates.
Variable C GDPt–1 GDPt–2 GDPt–3 GDPt–4 MC MCt–1 MCt–2 MCt–3 MCt–4 BO BOt–1 BOt–2
Coefficient
0.011 -0.168 -0.529*** -0.194 0.432*** 0.026** 0.081*** 0.022 0.057*** -0.005 -0.065*** 0.043*** 0.000
t-Statistic
2.614 -1.210 -4.633 -1.457 3.650 1.693 5.485 1.266 3.505 -0.281 -5.604 2.777 0.031
Note: **,*** Significant at critical value at 5% and 1% significance level, respectively.
Standard error
0.004 0.139 0.114 0.133 0.118 0.015 0.015 0.018 0.016 0.018 0.012 0.015 0.012
96 วารสารเศรษฐศาสตร์ มหาวิทยาลัยเชียงใหม่
14.02 กรกฎาคม-ธันวาคม 2553
Theresultsoftherelationshipbetweenstockmarketdevelopmentandeconomicgrowth canbewrittenas; GDPt = –0.529*** GDPt–2 + 0.432*** GDPt–4 + 0.026* MCt + 0.081*** MCt–1 + 0.057*** MCt–3 –0.07*** BOt + 0.04*** BOt–1 (3.1)
 Equation (3.1) presents the relationship between stock market development and economicgrowth,accordingtotheresultsthatGDPt–4 ,MCt ,MCt–1 , MCt–3 and BOt–1 havesignificantpositiveimpacton GDPt ,whereas GDPt–2 and BOt havesignificant negativeimpacton GDPt .Theanalysisoftheresultsshowthatinlong-runeconomic growthinThailandisstronglyinfluencefrompreviousGDPandfinancialliberalizationon stockmarketdevelopmentinThailand. 4.C onclusion
This paper attempts to explore the causal relationship between the stock market development and economic growth by utilizing ARDL. The data set depended on the availabilityofthedataserieswhichistakenfromEcowindatabase,duringMarch1995to June2010period. Forthestationaritytest,wehaveDF-G LStestandNg-Perrontesttofindtheintegrating orderofthevariablesutilizedinthestudy.Thestationarityevidencesrevealthatallvariables arestationaryattheirlevelformortheyhavetheorderofintegrationzero,I(0). The causality and relationship between stock market development and economic growthisestimatedbyARDL.Thefindingsrevealthatthemarketcapitalizationhasthe positiveimpactoneconomicgrowthimplyingthatgreaterstockmarketliquidityortheability totradetheequityeasilyreducethedownsideriskandcostofinvestinginprojects.Thus, moreliquidityinstockmarketmayacceleratethegrowthofmarketcapitalization,thereby stimulatingtheeconomicactivitiesandimprovingresourceallocation.Consequently,this canboosteconomicgrowthinThailand. Thefindingconfirmsthecausalitybetweenstockmarketdevelopmentandeconomic growthincaseofThailandandindicatesthatstockmarketdevelopmentleadstoeconomic growthatleastfortheperiodunderstudyoftheconsideration,whichsuggeststhatthestock
Stock Market Development and Economic Growth: ARDL Causality in Thailand 97
CMU. Journal of Economics
marketdevelopmentthroughfinancialliberalizationpolicyhasbecomeanimportantwheel foreconomicgrowthofThailand.Therefore,itissuggestedthatThailandneedstocontinue thedevelopmentofitsstockmarketthroughgovernmentpolicy.Thepolicyshouldfacilitate investmentaswellasincreasestockmarketliquiditywhichinturnincreasesincentiveof investors.
98 วารสารเศรษฐศาสตร์ มหาวิทยาลัยเชียงใหม่
14.02 กรกฎาคม-ธันวาคม 2553
References
A tje,RaymondandJovanovic,Boyan.1993,“Stockmarketsanddevelopment.”European EconomicReview.37:632-4 0. Bell,C.andRousseau,P.L.2001. “Post-independenceinIndia:acaseoffinancelend industrialization.”J ournalofDevelopmentEconomics.65:153-175. Bencivenga,etal.1996.“Liquidityofsecondarycapitalmarkets:allocativeefficiencyand thematuritycompositionofthecapitalstock”J ournalofEconomicTheory,Springer. 7,1:19-50. Bencivenga,ValerieR.andSmith,BruceD.1992.“Deficits,inflation,andthebankingsystem indevelopingcountries:theoptimaldegreeoffinancialrepression”,O xfordEconomic Papers.OxfordUniversityPress.44,4:767-90. Caporale,Howellsandsoliman. 2004. “Stockmarketdevelopmentandeconomicgrowth: acasuallinkage”JournalofEconomicDevelopment29:33-50. Dejong,D.N.;Nankervis,J.C.,andSavin,N.E.1992.“IntegrationVersusTrendStationarity inTimeSeries”,E conometrica.60:423-33. Elliot,G.;Rothenberg,T.J.;andStock,J.H.1996.“EfficienttestsforanautoregressiveUnit Root” E conometrica.64:813-36. Greenwood,JandSmith,BD.1997.“Financialmarketsindevelopment,andthedevelopment offinancialmarkets.”JournalofEconomicDynamicsandControl.21:145-181. Muhammad, S.; Nadeem, A., and Liaquat, A. 2008. “Stock market development and economicgrowth:ArdlcausalityinPakistan” InternationalResearchJournalofFinance andEconomics.14.